blob: 766c596585631e1001bdd64c80639554c6ab7831 [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
Joe Perchesafd465032012-03-12 07:03:32 +000068#define pr_fmt(fmt) "IPv4: " fmt
69
Herbert Xuf4c50d92006-06-22 03:02:40 -070070#include <linux/err.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/errno.h>
72#include <linux/types.h>
73#include <linux/socket.h>
74#include <linux/in.h>
75#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/module.h>
77#include <linux/sched.h>
78#include <linux/timer.h>
79#include <linux/string.h>
80#include <linux/sockios.h>
81#include <linux/net.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080082#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#include <linux/fcntl.h>
84#include <linux/mm.h>
85#include <linux/interrupt.h>
86#include <linux/stat.h>
87#include <linux/init.h>
88#include <linux/poll.h>
89#include <linux/netfilter_ipv4.h>
David S. Millerb3da2cf2007-03-23 11:40:27 -070090#include <linux/random.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090091#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/inet.h>
96#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020097#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Herbert Xu73cc19f2008-12-15 23:41:09 -080099#include <net/checksum.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#include <net/ip.h>
101#include <net/protocol.h>
102#include <net/arp.h>
103#include <net/route.h>
104#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700105#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/tcp.h>
107#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800108#include <net/udplite.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000109#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110#include <linux/skbuff.h>
111#include <net/sock.h>
112#include <net/raw.h>
113#include <net/icmp.h>
114#include <net/ipip.h>
115#include <net/inet_common.h>
116#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700117#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118#ifdef CONFIG_IP_MROUTE
119#include <linux/mroute.h>
120#endif
121
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123/* The inetsw table contains everything that inet_create needs to
124 * build a new socket.
125 */
126static struct list_head inetsw[SOCK_MAX];
127static DEFINE_SPINLOCK(inetsw_lock);
128
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800129struct ipv4_config ipv4_config;
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800130EXPORT_SYMBOL(ipv4_config);
131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132/* New destruction routine */
133
134void inet_sock_destruct(struct sock *sk)
135{
136 struct inet_sock *inet = inet_sk(sk);
137
138 __skb_queue_purge(&sk->sk_receive_queue);
139 __skb_queue_purge(&sk->sk_error_queue);
140
Hideo Aoki95766ff2007-12-31 00:29:24 -0800141 sk_mem_reclaim(sk);
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700144 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 sk->sk_state, sk);
146 return;
147 }
148 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700149 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 return;
151 }
152
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700153 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
154 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
155 WARN_ON(sk->sk_wmem_queued);
156 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000158 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000159 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700160 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700161 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700163EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
165/*
166 * The routines beyond this point handle the behaviour of an AF_INET
167 * socket object. Mostly it punts to the subprotocols of IP to do
168 * the work.
169 */
170
171/*
172 * Automatically bind an unbound socket.
173 */
174
175static int inet_autobind(struct sock *sk)
176{
177 struct inet_sock *inet;
178 /* We may need to bind the socket. */
179 lock_sock(sk);
180 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000181 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 if (sk->sk_prot->get_port(sk, 0)) {
183 release_sock(sk);
184 return -EAGAIN;
185 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000186 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 }
188 release_sock(sk);
189 return 0;
190}
191
192/*
193 * Move a socket into listening state.
194 */
195int inet_listen(struct socket *sock, int backlog)
196{
197 struct sock *sk = sock->sk;
198 unsigned char old_state;
199 int err;
200
201 lock_sock(sk);
202
203 err = -EINVAL;
204 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
205 goto out;
206
207 old_state = sk->sk_state;
208 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
209 goto out;
210
211 /* Really, if the socket is already in listen state
212 * we can only allow the backlog to be adjusted.
213 */
214 if (old_state != TCP_LISTEN) {
Jerry Chu83368862012-08-31 12:29:12 +0000215 /* Check special setups for testing purpose to enable TFO w/o
216 * requiring TCP_FASTOPEN sockopt.
217 * Note that only TCP sockets (SOCK_STREAM) will reach here.
218 * Also fastopenq may already been allocated because this
219 * socket was in TCP_LISTEN state previously but was
220 * shutdown() (rather than close()).
221 */
222 if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
223 inet_csk(sk)->icsk_accept_queue.fastopenq == NULL) {
224 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
225 err = fastopen_init_queue(sk, backlog);
226 else if ((sysctl_tcp_fastopen &
227 TFO_SERVER_WO_SOCKOPT2) != 0)
228 err = fastopen_init_queue(sk,
229 ((uint)sysctl_tcp_fastopen) >> 16);
230 else
231 err = 0;
232 if (err)
233 goto out;
234 }
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800235 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 if (err)
237 goto out;
238 }
239 sk->sk_max_ack_backlog = backlog;
240 err = 0;
241
242out:
243 release_sock(sk);
244 return err;
245}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700246EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Eric Dumazetbe776282007-03-27 13:53:04 -0700248u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700249EXPORT_SYMBOL(inet_ehash_secret);
250
Eric Dumazetbe776282007-03-27 13:53:04 -0700251/*
252 * inet_ehash_secret must be set exactly once
Eric Dumazetbe776282007-03-27 13:53:04 -0700253 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700254void build_ehash_secret(void)
255{
Eric Dumazetbe776282007-03-27 13:53:04 -0700256 u32 rnd;
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000257
Eric Dumazetbe776282007-03-27 13:53:04 -0700258 do {
259 get_random_bytes(&rnd, sizeof(rnd));
260 } while (rnd == 0);
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000261
262 cmpxchg(&inet_ehash_secret, 0, rnd);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700263}
264EXPORT_SYMBOL(build_ehash_secret);
265
David S. Millerf9242b62012-06-19 18:56:21 -0700266static inline int inet_netns_ok(struct net *net, __u8 protocol)
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700267{
Alexey Dobriyan32613092009-09-14 12:21:47 +0000268 const struct net_protocol *ipprot;
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700269
Eric Dumazet04f258c2008-11-23 15:42:23 -0800270 if (net_eq(net, &init_net))
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700271 return 1;
272
David S. Millerf9242b62012-06-19 18:56:21 -0700273 ipprot = rcu_dereference(inet_protos[protocol]);
274 if (ipprot == NULL) {
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700275 /* raw IP is OK */
276 return 1;
David S. Millerf9242b62012-06-19 18:56:21 -0700277 }
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700278 return ipprot->netns_ok;
279}
280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281/*
282 * Create an inet socket.
283 */
284
Eric Paris3f378b62009-11-05 22:18:14 -0800285static int inet_create(struct net *net, struct socket *sock, int protocol,
286 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287{
288 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 struct inet_protosw *answer;
290 struct inet_sock *inet;
291 struct proto *answer_prot;
292 unsigned char answer_flags;
293 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700294 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800295 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296
Eric Dumazet04f258c2008-11-23 15:42:23 -0800297 if (unlikely(!inet_ehash_secret))
298 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
299 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700300
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 sock->state = SS_UNCONNECTED;
302
303 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700304lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800305 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700307 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700309 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 /* Check the non-wild match. */
311 if (protocol == answer->protocol) {
312 if (protocol != IPPROTO_IP)
313 break;
314 } else {
315 /* Check for the two wild cases. */
316 if (IPPROTO_IP == protocol) {
317 protocol = answer->protocol;
318 break;
319 }
320 if (IPPROTO_IP == answer->protocol)
321 break;
322 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800323 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 }
325
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700326 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700327 if (try_loading_module < 2) {
328 rcu_read_unlock();
329 /*
330 * Be more specific, e.g. net-pf-2-proto-132-type-1
331 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
332 */
333 if (++try_loading_module == 1)
334 request_module("net-pf-%d-proto-%d-type-%d",
335 PF_INET, protocol, sock->type);
336 /*
337 * Fall back to generic, e.g. net-pf-2-proto-132
338 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
339 */
340 else
341 request_module("net-pf-%d-proto-%d",
342 PF_INET, protocol);
343 goto lookup_protocol;
344 } else
345 goto out_rcu_unlock;
346 }
347
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 err = -EPERM;
Eric Parisc84b3262009-11-05 20:45:52 -0800349 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700352 err = -EAFNOSUPPORT;
353 if (!inet_netns_ok(net, protocol))
354 goto out_rcu_unlock;
355
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 sock->ops = answer->ops;
357 answer_prot = answer->prot;
358 answer_no_check = answer->no_check;
359 answer_flags = answer->flags;
360 rcu_read_unlock();
361
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700362 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
364 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700365 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 if (sk == NULL)
367 goto out;
368
369 err = 0;
370 sk->sk_no_check = answer_no_check;
371 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000372 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
374 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800375 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000377 inet->nodefrag = 0;
378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000380 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 if (IPPROTO_RAW == protocol)
382 inet->hdrincl = 1;
383 }
384
385 if (ipv4_config.no_pmtu_disc)
386 inet->pmtudisc = IP_PMTUDISC_DONT;
387 else
388 inet->pmtudisc = IP_PMTUDISC_WANT;
389
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000390 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 sock_init_data(sock, sk);
393
394 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 sk->sk_protocol = protocol;
396 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
397
398 inet->uc_ttl = -1;
399 inet->mc_loop = 1;
400 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000401 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 inet->mc_index = 0;
403 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000404 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700406 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000408 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 /* It assumes that any protocol which allows
410 * the user to assign a number at socket
411 * creation time automatically
412 * shares.
413 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000414 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 /* Add to protocol hash chains. */
416 sk->sk_prot->hash(sk);
417 }
418
419 if (sk->sk_prot->init) {
420 err = sk->sk_prot->init(sk);
421 if (err)
422 sk_common_release(sk);
423 }
424out:
425 return err;
426out_rcu_unlock:
427 rcu_read_unlock();
428 goto out;
429}
430
431
432/*
433 * The peer socket should always be NULL (or else). When we call this
434 * function we are destroying the object and from then on nobody
435 * should refer to it.
436 */
437int inet_release(struct socket *sock)
438{
439 struct sock *sk = sock->sk;
440
441 if (sk) {
442 long timeout;
443
David S. Millerc58dc012010-04-27 15:05:31 -0700444 sock_rps_reset_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700445
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 /* Applications forget to leave groups before exiting */
447 ip_mc_drop_socket(sk);
448
449 /* If linger is set, we don't return until the close
450 * is complete. Otherwise we return immediately. The
451 * actually closing is done the same either way.
452 *
453 * If the close is due to the process exiting, we never
454 * linger..
455 */
456 timeout = 0;
457 if (sock_flag(sk, SOCK_LINGER) &&
458 !(current->flags & PF_EXITING))
459 timeout = sk->sk_lingertime;
460 sock->sk = NULL;
461 sk->sk_prot->close(sk, timeout);
462 }
463 return 0;
464}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700465EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700468int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700469EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
472{
473 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
474 struct sock *sk = sock->sk;
475 struct inet_sock *inet = inet_sk(sk);
476 unsigned short snum;
477 int chk_addr_ret;
478 int err;
479
480 /* If the socket has its own bind function then use it. (RAW) */
481 if (sk->sk_prot->bind) {
482 err = sk->sk_prot->bind(sk, uaddr, addr_len);
483 goto out;
484 }
485 err = -EINVAL;
486 if (addr_len < sizeof(struct sockaddr_in))
487 goto out;
488
Marcus Meissnerc349a522011-07-04 01:30:29 +0000489 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400490 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
491 * only if s_addr is INADDR_ANY.
492 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000493 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400494 if (addr->sin_family != AF_UNSPEC ||
495 addr->sin_addr.s_addr != htonl(INADDR_ANY))
496 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000497 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700498
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900499 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
501 /* Not specified by any standard per-se, however it breaks too
502 * many applications when removed. It is unfortunate since
503 * allowing applications to make a non-local bind solves
504 * several problems with systems using dynamic addressing.
505 * (ie. your servers still start up even if your ISDN link
506 * is temporarily down)
507 */
508 err = -EADDRNOTAVAIL;
509 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700510 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700511 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 chk_addr_ret != RTN_LOCAL &&
513 chk_addr_ret != RTN_MULTICAST &&
514 chk_addr_ret != RTN_BROADCAST)
515 goto out;
516
517 snum = ntohs(addr->sin_port);
518 err = -EACCES;
519 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
520 goto out;
521
522 /* We keep a pair of addresses. rcv_saddr is the one
523 * used by hash lookups, and saddr is used for transmit.
524 *
525 * In the BSD API these are the same except where it
526 * would be illegal to use them (multicast/broadcast) in
527 * which case the sending device address is used.
528 */
529 lock_sock(sk);
530
531 /* Check these errors (active socket, double bind). */
532 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000533 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 goto out_release_sock;
535
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000536 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000538 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540 /* Make sure we are allowed to bind here. */
541 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000542 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 err = -EADDRINUSE;
544 goto out_release_sock;
545 }
546
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000547 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
549 if (snum)
550 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000551 inet->inet_sport = htons(inet->inet_num);
552 inet->inet_daddr = 0;
553 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 sk_dst_reset(sk);
555 err = 0;
556out_release_sock:
557 release_sock(sk);
558out:
559 return err;
560}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700561EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000563int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 int addr_len, int flags)
565{
566 struct sock *sk = sock->sk;
567
Changli Gao6503d962010-03-31 22:58:26 +0000568 if (addr_len < sizeof(uaddr->sa_family))
569 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 if (uaddr->sa_family == AF_UNSPEC)
571 return sk->sk_prot->disconnect(sk, flags);
572
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000573 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000575 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700577EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
Yuchung Cheng783237e2012-07-19 06:43:07 +0000579static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
581 DEFINE_WAIT(wait);
582
Eric Dumazetaa395142010-04-20 13:03:51 +0000583 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000584 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 /* Basic assumption: if someone sets sk->sk_err, he _must_
587 * change state of the socket from TCP_SYN_*.
588 * Connect() does not allow to get error notifications
589 * without closing the socket.
590 */
591 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
592 release_sock(sk);
593 timeo = schedule_timeout(timeo);
594 lock_sock(sk);
595 if (signal_pending(current) || !timeo)
596 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000597 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000599 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000600 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 return timeo;
602}
603
604/*
605 * Connect to a remote host. There is regrettably still a little
606 * TCP 'magic' in here.
607 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000608int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
609 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610{
611 struct sock *sk = sock->sk;
612 int err;
613 long timeo;
614
Changli Gao6503d962010-03-31 22:58:26 +0000615 if (addr_len < sizeof(uaddr->sa_family))
616 return -EINVAL;
617
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 if (uaddr->sa_family == AF_UNSPEC) {
619 err = sk->sk_prot->disconnect(sk, flags);
620 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
621 goto out;
622 }
623
624 switch (sock->state) {
625 default:
626 err = -EINVAL;
627 goto out;
628 case SS_CONNECTED:
629 err = -EISCONN;
630 goto out;
631 case SS_CONNECTING:
632 err = -EALREADY;
633 /* Fall out of switch with err, set for this state */
634 break;
635 case SS_UNCONNECTED:
636 err = -EISCONN;
637 if (sk->sk_state != TCP_CLOSE)
638 goto out;
639
640 err = sk->sk_prot->connect(sk, uaddr, addr_len);
641 if (err < 0)
642 goto out;
643
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900644 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646 /* Just entered SS_CONNECTING state; the only
647 * difference is that return value in non-blocking
648 * case is EINPROGRESS, rather than EALREADY.
649 */
650 err = -EINPROGRESS;
651 break;
652 }
653
654 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
655
656 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000657 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
658 tcp_sk(sk)->fastopen_req &&
659 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000662 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 goto out;
664
665 err = sock_intr_errno(timeo);
666 if (signal_pending(current))
667 goto out;
668 }
669
670 /* Connection was closed by RST, timeout, ICMP error
671 * or another process disconnected us.
672 */
673 if (sk->sk_state == TCP_CLOSE)
674 goto sock_error;
675
676 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
677 * and error was received after socket entered established state.
678 * Hence, it is handled normally after connect() return successfully.
679 */
680
681 sock->state = SS_CONNECTED;
682 err = 0;
683out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 return err;
685
686sock_error:
687 err = sock_error(sk) ? : -ECONNABORTED;
688 sock->state = SS_UNCONNECTED;
689 if (sk->sk_prot->disconnect(sk, flags))
690 sock->state = SS_DISCONNECTING;
691 goto out;
692}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000693EXPORT_SYMBOL(__inet_stream_connect);
694
695int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
696 int addr_len, int flags)
697{
698 int err;
699
700 lock_sock(sock->sk);
701 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
702 release_sock(sock->sk);
703 return err;
704}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700705EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
707/*
708 * Accept a pending connection. The TCP layer now gives BSD semantics.
709 */
710
711int inet_accept(struct socket *sock, struct socket *newsock, int flags)
712{
713 struct sock *sk1 = sock->sk;
714 int err = -EINVAL;
715 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
716
717 if (!sk2)
718 goto do_err;
719
720 lock_sock(sk2);
721
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000722 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700723 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000724 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
725 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726
727 sock_graft(sk2, newsock);
728
729 newsock->state = SS_CONNECTED;
730 err = 0;
731 release_sock(sk2);
732do_err:
733 return err;
734}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700735EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
737
738/*
739 * This does both peername and sockname.
740 */
741int inet_getname(struct socket *sock, struct sockaddr *uaddr,
742 int *uaddr_len, int peer)
743{
744 struct sock *sk = sock->sk;
745 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700746 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
748 sin->sin_family = AF_INET;
749 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000750 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
752 peer == 1))
753 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000754 sin->sin_port = inet->inet_dport;
755 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000757 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000759 addr = inet->inet_saddr;
760 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 sin->sin_addr.s_addr = addr;
762 }
763 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
764 *uaddr_len = sizeof(*sin);
765 return 0;
766}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700767EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
770 size_t size)
771{
772 struct sock *sk = sock->sk;
773
David S. Millerc58dc012010-04-27 15:05:31 -0700774 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000777 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
778 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 return -EAGAIN;
780
781 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
782}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700783EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Changli Gao7ba42912010-07-10 20:41:55 +0000785ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
786 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787{
788 struct sock *sk = sock->sk;
789
David S. Millerc58dc012010-04-27 15:05:31 -0700790 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700791
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000793 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
794 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 return -EAGAIN;
796
797 if (sk->sk_prot->sendpage)
798 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
799 return sock_no_sendpage(sock, page, offset, size, flags);
800}
Changli Gao7ba42912010-07-10 20:41:55 +0000801EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Tom Herbertfec5e652010-04-16 16:01:27 -0700803int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
804 size_t size, int flags)
805{
806 struct sock *sk = sock->sk;
807 int addr_len = 0;
808 int err;
809
David S. Millerc58dc012010-04-27 15:05:31 -0700810 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700811
812 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
813 flags & ~MSG_DONTWAIT, &addr_len);
814 if (err >= 0)
815 msg->msg_namelen = addr_len;
816 return err;
817}
818EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
820int inet_shutdown(struct socket *sock, int how)
821{
822 struct sock *sk = sock->sk;
823 int err = 0;
824
825 /* This should really check to make sure
826 * the socket is a TCP socket. (WHY AC...)
827 */
828 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
829 1->2 bit 2 snds.
830 2->3 */
831 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
832 return -EINVAL;
833
834 lock_sock(sk);
835 if (sock->state == SS_CONNECTING) {
836 if ((1 << sk->sk_state) &
837 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
838 sock->state = SS_DISCONNECTING;
839 else
840 sock->state = SS_CONNECTED;
841 }
842
843 switch (sk->sk_state) {
844 case TCP_CLOSE:
845 err = -ENOTCONN;
846 /* Hack to wake up other listeners, who can poll for
847 POLLHUP, even on eg. unconnected UDP sockets -- RR */
848 default:
849 sk->sk_shutdown |= how;
850 if (sk->sk_prot->shutdown)
851 sk->sk_prot->shutdown(sk, how);
852 break;
853
854 /* Remaining two branches are temporary solution for missing
855 * close() in multithreaded environment. It is _not_ a good idea,
856 * but we have no choice until close() is repaired at VFS level.
857 */
858 case TCP_LISTEN:
859 if (!(how & RCV_SHUTDOWN))
860 break;
861 /* Fall through */
862 case TCP_SYN_SENT:
863 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
864 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
865 break;
866 }
867
868 /* Wake up anyone sleeping in poll. */
869 sk->sk_state_change(sk);
870 release_sock(sk);
871 return err;
872}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700873EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875/*
876 * ioctl() calls you can issue on an INET socket. Most of these are
877 * device configuration and stuff and very rarely used. Some ioctls
878 * pass on to the socket itself.
879 *
880 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
881 * loads the devconfigure module does its configuring and unloads it.
882 * There's a good 20K of config code hanging around the kernel.
883 */
884
885int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
886{
887 struct sock *sk = sock->sk;
888 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900889 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
891 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700892 case SIOCGSTAMP:
893 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
894 break;
895 case SIOCGSTAMPNS:
896 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
897 break;
898 case SIOCADDRT:
899 case SIOCDELRT:
900 case SIOCRTMSG:
901 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
902 break;
903 case SIOCDARP:
904 case SIOCGARP:
905 case SIOCSARP:
906 err = arp_ioctl(net, cmd, (void __user *)arg);
907 break;
908 case SIOCGIFADDR:
909 case SIOCSIFADDR:
910 case SIOCGIFBRDADDR:
911 case SIOCSIFBRDADDR:
912 case SIOCGIFNETMASK:
913 case SIOCSIFNETMASK:
914 case SIOCGIFDSTADDR:
915 case SIOCSIFDSTADDR:
916 case SIOCSIFPFLAGS:
917 case SIOCGIFPFLAGS:
918 case SIOCSIFFLAGS:
919 err = devinet_ioctl(net, cmd, (void __user *)arg);
920 break;
921 default:
922 if (sk->sk_prot->ioctl)
923 err = sk->sk_prot->ioctl(sk, cmd, arg);
924 else
925 err = -ENOIOCTLCMD;
926 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 }
928 return err;
929}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700930EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000932#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000933static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000934{
935 struct sock *sk = sock->sk;
936 int err = -ENOIOCTLCMD;
937
938 if (sk->sk_prot->compat_ioctl)
939 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
940
941 return err;
942}
943#endif
944
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800945const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800946 .family = PF_INET,
947 .owner = THIS_MODULE,
948 .release = inet_release,
949 .bind = inet_bind,
950 .connect = inet_stream_connect,
951 .socketpair = sock_no_socketpair,
952 .accept = inet_accept,
953 .getname = inet_getname,
954 .poll = tcp_poll,
955 .ioctl = inet_ioctl,
956 .listen = inet_listen,
957 .shutdown = inet_shutdown,
958 .setsockopt = sock_common_setsockopt,
959 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000960 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700961 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800962 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000963 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800964 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800965#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800966 .compat_setsockopt = compat_sock_common_setsockopt,
967 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000968 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800969#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700971EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800973const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800974 .family = PF_INET,
975 .owner = THIS_MODULE,
976 .release = inet_release,
977 .bind = inet_bind,
978 .connect = inet_dgram_connect,
979 .socketpair = sock_no_socketpair,
980 .accept = sock_no_accept,
981 .getname = inet_getname,
982 .poll = udp_poll,
983 .ioctl = inet_ioctl,
984 .listen = sock_no_listen,
985 .shutdown = inet_shutdown,
986 .setsockopt = sock_common_setsockopt,
987 .getsockopt = sock_common_getsockopt,
988 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700989 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800990 .mmap = sock_no_mmap,
991 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800992#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800993 .compat_setsockopt = compat_sock_common_setsockopt,
994 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000995 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800996#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700998EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000/*
1001 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
1002 * udp_poll
1003 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001004static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001005 .family = PF_INET,
1006 .owner = THIS_MODULE,
1007 .release = inet_release,
1008 .bind = inet_bind,
1009 .connect = inet_dgram_connect,
1010 .socketpair = sock_no_socketpair,
1011 .accept = sock_no_accept,
1012 .getname = inet_getname,
1013 .poll = datagram_poll,
1014 .ioctl = inet_ioctl,
1015 .listen = sock_no_listen,
1016 .shutdown = inet_shutdown,
1017 .setsockopt = sock_common_setsockopt,
1018 .getsockopt = sock_common_getsockopt,
1019 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001020 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001021 .mmap = sock_no_mmap,
1022 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001023#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001024 .compat_setsockopt = compat_sock_common_setsockopt,
1025 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001026 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001027#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028};
1029
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001030static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 .family = PF_INET,
1032 .create = inet_create,
1033 .owner = THIS_MODULE,
1034};
1035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036/* Upon startup we insert all the elements in inetsw_array[] into
1037 * the linked list inetsw.
1038 */
1039static struct inet_protosw inetsw_array[] =
1040{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001041 {
1042 .type = SOCK_STREAM,
1043 .protocol = IPPROTO_TCP,
1044 .prot = &tcp_prot,
1045 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001046 .no_check = 0,
1047 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001048 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001049 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001051 {
1052 .type = SOCK_DGRAM,
1053 .protocol = IPPROTO_UDP,
1054 .prot = &udp_prot,
1055 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001056 .no_check = UDP_CSUM_DEFAULT,
1057 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001059
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001060 {
1061 .type = SOCK_DGRAM,
1062 .protocol = IPPROTO_ICMP,
1063 .prot = &ping_prot,
1064 .ops = &inet_dgram_ops,
1065 .no_check = UDP_CSUM_DEFAULT,
1066 .flags = INET_PROTOSW_REUSE,
1067 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
1069 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001070 .type = SOCK_RAW,
1071 .protocol = IPPROTO_IP, /* wild card */
1072 .prot = &raw_prot,
1073 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001074 .no_check = UDP_CSUM_DEFAULT,
1075 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 }
1077};
1078
Denis Chengc40f6ff2007-09-16 16:39:25 -07001079#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080
1081void inet_register_protosw(struct inet_protosw *p)
1082{
1083 struct list_head *lh;
1084 struct inet_protosw *answer;
1085 int protocol = p->protocol;
1086 struct list_head *last_perm;
1087
1088 spin_lock_bh(&inetsw_lock);
1089
1090 if (p->type >= SOCK_MAX)
1091 goto out_illegal;
1092
1093 /* If we are trying to override a permanent protocol, bail. */
1094 answer = NULL;
1095 last_perm = &inetsw[p->type];
1096 list_for_each(lh, &inetsw[p->type]) {
1097 answer = list_entry(lh, struct inet_protosw, list);
1098
1099 /* Check only the non-wild match. */
1100 if (INET_PROTOSW_PERMANENT & answer->flags) {
1101 if (protocol == answer->protocol)
1102 break;
1103 last_perm = lh;
1104 }
1105
1106 answer = NULL;
1107 }
1108 if (answer)
1109 goto out_permanent;
1110
1111 /* Add the new entry after the last permanent entry if any, so that
1112 * the new entry does not override a permanent entry when matched with
1113 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001114 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 * system automatically returns to the old behavior.
1116 */
1117 list_add_rcu(&p->list, last_perm);
1118out:
1119 spin_unlock_bh(&inetsw_lock);
1120
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 return;
1122
1123out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001124 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 goto out;
1126
1127out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001128 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 p->type);
1130 goto out;
1131}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001132EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
1134void inet_unregister_protosw(struct inet_protosw *p)
1135{
1136 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001137 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 p->protocol);
1139 } else {
1140 spin_lock_bh(&inetsw_lock);
1141 list_del_rcu(&p->list);
1142 spin_unlock_bh(&inetsw_lock);
1143
1144 synchronize_net();
1145 }
1146}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001147EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001149/*
1150 * Shall we try to damage output packets if routing dev changes?
1151 */
1152
Brian Haleyab32ea52006-09-22 14:15:41 -07001153int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001154
1155static int inet_sk_reselect_saddr(struct sock *sk)
1156{
1157 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001158 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001159 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001160 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001161 struct rtable *rt;
1162 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001163 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001164
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001165 inet_opt = rcu_dereference_protected(inet->inet_opt,
1166 sock_owned_by_user(sk));
1167 if (inet_opt && inet_opt->opt.srr)
1168 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001169
1170 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001171 fl4 = &inet->cork.fl.u.ip4;
1172 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001173 sk->sk_bound_dev_if, sk->sk_protocol,
1174 inet->inet_sport, inet->inet_dport, sk, false);
1175 if (IS_ERR(rt))
1176 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001177
Changli Gaod8d1f302010-06-10 23:31:35 -07001178 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001179
David S. Miller6e869132011-05-06 16:18:04 -07001180 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001181
1182 if (new_saddr == old_saddr)
1183 return 0;
1184
1185 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001186 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1187 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001188 }
1189
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001190 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001191
1192 /*
1193 * XXX The only one ugly spot where we need to
1194 * XXX really change the sockets identity after
1195 * XXX it has entered the hashes. -DaveM
1196 *
1197 * Besides that, it does not check for connection
1198 * uniqueness. Wait for troubles.
1199 */
1200 __sk_prot_rehash(sk);
1201 return 0;
1202}
1203
1204int inet_sk_rebuild_header(struct sock *sk)
1205{
1206 struct inet_sock *inet = inet_sk(sk);
1207 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001208 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001209 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001210 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001211 int err;
1212
1213 /* Route is OK, nothing to do. */
1214 if (rt)
1215 return 0;
1216
1217 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001218 rcu_read_lock();
1219 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001220 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001221 if (inet_opt && inet_opt->opt.srr)
1222 daddr = inet_opt->opt.faddr;
1223 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001224 fl4 = &inet->cork.fl.u.ip4;
1225 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001226 inet->inet_dport, inet->inet_sport,
1227 sk->sk_protocol, RT_CONN_FLAGS(sk),
1228 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001229 if (!IS_ERR(rt)) {
1230 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001231 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001232 } else {
1233 err = PTR_ERR(rt);
1234
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001235 /* Routing failed... */
1236 sk->sk_route_caps = 0;
1237 /*
1238 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1239 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1240 */
1241 if (!sysctl_ip_dynaddr ||
1242 sk->sk_state != TCP_SYN_SENT ||
1243 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1244 (err = inet_sk_reselect_saddr(sk)) != 0)
1245 sk->sk_err_soft = -err;
1246 }
1247
1248 return err;
1249}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001250EXPORT_SYMBOL(inet_sk_rebuild_header);
1251
Herbert Xua430a432006-07-08 13:34:56 -07001252static int inet_gso_send_check(struct sk_buff *skb)
1253{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001254 const struct net_protocol *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001255 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -07001256 int proto;
1257 int ihl;
1258 int err = -EINVAL;
1259
1260 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1261 goto out;
1262
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001263 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001264 ihl = iph->ihl * 4;
1265 if (ihl < sizeof(*iph))
1266 goto out;
1267
1268 if (unlikely(!pskb_may_pull(skb, ihl)))
1269 goto out;
1270
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001271 __skb_pull(skb, ihl);
1272 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001273 iph = ip_hdr(skb);
David S. Millerf9242b62012-06-19 18:56:21 -07001274 proto = iph->protocol;
Herbert Xua430a432006-07-08 13:34:56 -07001275 err = -EPROTONOSUPPORT;
1276
1277 rcu_read_lock();
1278 ops = rcu_dereference(inet_protos[proto]);
1279 if (likely(ops && ops->gso_send_check))
1280 err = ops->gso_send_check(skb);
1281 rcu_read_unlock();
1282
1283out:
1284 return err;
1285}
1286
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001287static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1288 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001289{
1290 struct sk_buff *segs = ERR_PTR(-EINVAL);
Alexey Dobriyan32613092009-09-14 12:21:47 +00001291 const struct net_protocol *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001292 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001293 int proto;
1294 int ihl;
1295 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001296 unsigned int offset = 0;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001297
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001298 if (!(features & NETIF_F_V4_CSUM))
1299 features &= ~NETIF_F_SG;
1300
Herbert Xubbcf4672006-07-03 19:38:35 -07001301 if (unlikely(skb_shinfo(skb)->gso_type &
1302 ~(SKB_GSO_TCPV4 |
1303 SKB_GSO_UDP |
1304 SKB_GSO_DODGY |
1305 SKB_GSO_TCP_ECN |
1306 0)))
1307 goto out;
1308
1309 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001310 goto out;
1311
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001312 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001313 ihl = iph->ihl * 4;
1314 if (ihl < sizeof(*iph))
1315 goto out;
1316
Herbert Xubbcf4672006-07-03 19:38:35 -07001317 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001318 goto out;
1319
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001320 __skb_pull(skb, ihl);
1321 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001322 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001323 id = ntohs(iph->id);
David S. Millerf9242b62012-06-19 18:56:21 -07001324 proto = iph->protocol;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001325 segs = ERR_PTR(-EPROTONOSUPPORT);
1326
1327 rcu_read_lock();
1328 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001329 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001330 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001331 rcu_read_unlock();
1332
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001333 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001334 goto out;
1335
1336 skb = segs;
1337 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001338 iph = ip_hdr(skb);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001339 if (proto == IPPROTO_UDP) {
1340 iph->id = htons(id);
1341 iph->frag_off = htons(offset >> 3);
1342 if (skb->next != NULL)
1343 iph->frag_off |= htons(IP_MF);
1344 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1345 } else
1346 iph->id = htons(id++);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001347 iph->tot_len = htons(skb->len - skb->mac_len);
1348 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001349 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001350 } while ((skb = skb->next));
1351
1352out:
1353 return segs;
1354}
1355
Herbert Xu73cc19f2008-12-15 23:41:09 -08001356static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1357 struct sk_buff *skb)
1358{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001359 const struct net_protocol *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001360 struct sk_buff **pp = NULL;
1361 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001362 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001363 unsigned int hlen;
1364 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001365 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001366 int flush = 1;
1367 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001368
Herbert Xua5b1cf22009-05-26 18:50:28 +00001369 off = skb_gro_offset(skb);
1370 hlen = off + sizeof(*iph);
1371 iph = skb_gro_header_fast(skb, off);
1372 if (skb_gro_header_hard(skb, hlen)) {
1373 iph = skb_gro_header_slow(skb, hlen, off);
1374 if (unlikely(!iph))
1375 goto out;
1376 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001377
David S. Millerf9242b62012-06-19 18:56:21 -07001378 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001379
1380 rcu_read_lock();
1381 ops = rcu_dereference(inet_protos[proto]);
1382 if (!ops || !ops->gro_receive)
1383 goto out_unlock;
1384
Herbert Xua5ad24b2009-02-08 18:00:39 +00001385 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001386 goto out_unlock;
1387
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001388 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001389 goto out_unlock;
1390
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001391 id = ntohl(*(__be32 *)&iph->id);
1392 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001393 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001394
1395 for (p = *head; p; p = p->next) {
1396 struct iphdr *iph2;
1397
1398 if (!NAPI_GRO_CB(p)->same_flow)
1399 continue;
1400
1401 iph2 = ip_hdr(p);
1402
Herbert Xua5ad24b2009-02-08 18:00:39 +00001403 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001404 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1405 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001406 NAPI_GRO_CB(p)->same_flow = 0;
1407 continue;
1408 }
1409
1410 /* All fields must match except length and checksum. */
1411 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001412 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001413 (iph->tos ^ iph2->tos) |
Herbert Xua5ad24b2009-02-08 18:00:39 +00001414 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001415
1416 NAPI_GRO_CB(p)->flush |= flush;
1417 }
1418
1419 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001420 skb_gro_pull(skb, sizeof(*iph));
1421 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001422
1423 pp = ops->gro_receive(head, skb);
1424
1425out_unlock:
1426 rcu_read_unlock();
1427
1428out:
1429 NAPI_GRO_CB(skb)->flush |= flush;
1430
1431 return pp;
1432}
1433
1434static int inet_gro_complete(struct sk_buff *skb)
1435{
Herbert Xu73cc19f2008-12-15 23:41:09 -08001436 __be16 newlen = htons(skb->len - skb_network_offset(skb));
David S. Millerf9242b62012-06-19 18:56:21 -07001437 struct iphdr *iph = ip_hdr(skb);
1438 const struct net_protocol *ops;
1439 int proto = iph->protocol;
1440 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001441
1442 csum_replace2(&iph->check, iph->tot_len, newlen);
1443 iph->tot_len = newlen;
1444
1445 rcu_read_lock();
1446 ops = rcu_dereference(inet_protos[proto]);
1447 if (WARN_ON(!ops || !ops->gro_complete))
1448 goto out_unlock;
1449
1450 err = ops->gro_complete(skb);
1451
1452out_unlock:
1453 rcu_read_unlock();
1454
1455 return err;
1456}
1457
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001458int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001459 unsigned short type, unsigned char protocol,
1460 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001461{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001462 struct socket *sock;
1463 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001464
1465 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001466 *sk = sock->sk;
1467 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001468 /*
1469 * Unhash it so that IP input processing does not even see it,
1470 * we do not wish this socket to see incoming packets.
1471 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001472 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001473
1474 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001475 }
1476 return rc;
1477}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001478EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1479
Tejun Heo7d720c32010-02-16 15:20:26 +00001480unsigned long snmp_fold_field(void __percpu *mib[], int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001481{
1482 unsigned long res = 0;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001483 int i, j;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001484
1485 for_each_possible_cpu(i) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001486 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1487 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001488 }
1489 return res;
1490}
1491EXPORT_SYMBOL_GPL(snmp_fold_field);
1492
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001493#if BITS_PER_LONG==32
1494
1495u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1496{
1497 u64 res = 0;
1498 int cpu;
1499
1500 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001501 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001502 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001503 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001504 unsigned int start;
1505
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001506 bhptr = per_cpu_ptr(mib[0], cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001507 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1508 do {
1509 start = u64_stats_fetch_begin_bh(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001510 v = *(((u64 *) bhptr) + offt);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001511 } while (u64_stats_fetch_retry_bh(syncp, start));
1512
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001513 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001514 }
1515 return res;
1516}
1517EXPORT_SYMBOL_GPL(snmp_fold_field64);
1518#endif
1519
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001520int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001521{
1522 BUG_ON(ptr == NULL);
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001523 ptr[0] = __alloc_percpu(mibsize, align);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001524 if (!ptr[0])
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001525 return -ENOMEM;
1526#if SNMP_ARRAY_SZ == 2
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001527 ptr[1] = __alloc_percpu(mibsize, align);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001528 if (!ptr[1]) {
1529 free_percpu(ptr[0]);
1530 ptr[0] = NULL;
1531 return -ENOMEM;
1532 }
1533#endif
Herbert Xu5e0f0432007-04-24 21:53:35 -07001534 return 0;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001535}
1536EXPORT_SYMBOL_GPL(snmp_mib_init);
1537
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001538void snmp_mib_free(void __percpu *ptr[SNMP_ARRAY_SZ])
Herbert Xu5e0f0432007-04-24 21:53:35 -07001539{
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001540 int i;
1541
Herbert Xu5e0f0432007-04-24 21:53:35 -07001542 BUG_ON(ptr == NULL);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001543 for (i = 0; i < SNMP_ARRAY_SZ; i++) {
1544 free_percpu(ptr[i]);
1545 ptr[i] = NULL;
1546 }
Herbert Xu5e0f0432007-04-24 21:53:35 -07001547}
1548EXPORT_SYMBOL_GPL(snmp_mib_free);
1549
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001551static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001553 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554};
1555#endif
1556
Alexey Dobriyan32613092009-09-14 12:21:47 +00001557static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001558 .early_demux = tcp_v4_early_demux,
1559 .handler = tcp_v4_rcv,
1560 .err_handler = tcp_v4_err,
1561 .gso_send_check = tcp_v4_gso_send_check,
1562 .gso_segment = tcp_tso_segment,
1563 .gro_receive = tcp4_gro_receive,
1564 .gro_complete = tcp4_gro_complete,
1565 .no_policy = 1,
1566 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567};
1568
Alexey Dobriyan32613092009-09-14 12:21:47 +00001569static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 .handler = udp_rcv,
1571 .err_handler = udp_err,
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001572 .gso_send_check = udp4_ufo_send_check,
1573 .gso_segment = udp4_ufo_fragment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001575 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576};
1577
Alexey Dobriyan32613092009-09-14 12:21:47 +00001578static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 .handler = icmp_rcv,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001580 .err_handler = ping_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001581 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001582 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583};
1584
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001585static __net_init int ipv4_mib_init_net(struct net *net)
1586{
Tejun Heo7d720c32010-02-16 15:20:26 +00001587 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001588 sizeof(struct tcp_mib),
1589 __alignof__(struct tcp_mib)) < 0)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001590 goto err_tcp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001591 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001592 sizeof(struct ipstats_mib),
1593 __alignof__(struct ipstats_mib)) < 0)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001594 goto err_ip_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001595 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001596 sizeof(struct linux_mib),
1597 __alignof__(struct linux_mib)) < 0)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001598 goto err_net_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001599 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001600 sizeof(struct udp_mib),
1601 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001602 goto err_udp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001603 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001604 sizeof(struct udp_mib),
1605 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001606 goto err_udplite_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001607 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001608 sizeof(struct icmp_mib),
1609 __alignof__(struct icmp_mib)) < 0)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001610 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001611 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1612 GFP_KERNEL);
1613 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001614 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001615
1616 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001617 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001618
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001619err_icmpmsg_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001620 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001621err_icmp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001622 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001623err_udplite_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001624 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001625err_udp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001626 snmp_mib_free((void __percpu **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001627err_net_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001628 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001629err_ip_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001630 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001631err_tcp_mib:
1632 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001633}
1634
1635static __net_exit void ipv4_mib_exit_net(struct net *net)
1636{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001637 kfree(net->mib.icmpmsg_statistics);
Tejun Heo7d720c32010-02-16 15:20:26 +00001638 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1639 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1640 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1641 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1642 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1643 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001644}
1645
1646static __net_initdata struct pernet_operations ipv4_mib_ops = {
1647 .init = ipv4_mib_init_net,
1648 .exit = ipv4_mib_exit_net,
1649};
1650
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651static int __init init_ipv4_mibs(void)
1652{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001653 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654}
1655
1656static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657
Herbert Xu30e224d2005-07-05 14:40:10 -07001658/*
1659 * IP protocol layer initialiser
1660 */
1661
Stephen Hemminger7546dd92009-03-09 08:18:29 +00001662static struct packet_type ip_packet_type __read_mostly = {
Harvey Harrison09640e62009-02-01 00:45:17 -08001663 .type = cpu_to_be16(ETH_P_IP),
Herbert Xu30e224d2005-07-05 14:40:10 -07001664 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001665 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001666 .gso_segment = inet_gso_segment,
Herbert Xu73cc19f2008-12-15 23:41:09 -08001667 .gro_receive = inet_gro_receive,
1668 .gro_complete = inet_gro_complete,
Herbert Xu30e224d2005-07-05 14:40:10 -07001669};
1670
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671static int __init inet_init(void)
1672{
1673 struct sk_buff *dummy_skb;
1674 struct inet_protosw *q;
1675 struct list_head *r;
1676 int rc = -EINVAL;
1677
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001678 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
Amerigo Wange3826f12010-05-05 00:27:06 +00001680 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1681 if (!sysctl_local_reserved_ports)
1682 goto out;
1683
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 rc = proto_register(&tcp_prot, 1);
1685 if (rc)
Amerigo Wange3826f12010-05-05 00:27:06 +00001686 goto out_free_reserved_ports;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
1688 rc = proto_register(&udp_prot, 1);
1689 if (rc)
1690 goto out_unregister_tcp_proto;
1691
1692 rc = proto_register(&raw_prot, 1);
1693 if (rc)
1694 goto out_unregister_udp_proto;
1695
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001696 rc = proto_register(&ping_prot, 1);
1697 if (rc)
1698 goto out_unregister_raw_proto;
1699
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001701 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 */
1703
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001704 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Al Virobd7b1532008-07-15 16:00:59 -04001706#ifdef CONFIG_SYSCTL
1707 ip_static_sysctl_init();
1708#endif
1709
Glauber Costa3dc43e32011-12-11 21:47:05 +00001710 tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1711
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 /*
1713 * Add all the base protocols.
1714 */
1715
1716 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001717 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001719 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001721 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722#ifdef CONFIG_IP_MULTICAST
1723 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001724 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725#endif
1726
1727 /* Register the socket-side information for inet_create. */
1728 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1729 INIT_LIST_HEAD(r);
1730
1731 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1732 inet_register_protosw(q);
1733
1734 /*
1735 * Set the ARP module up
1736 */
1737
1738 arp_init();
1739
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001740 /*
1741 * Set the IP module up
1742 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
1744 ip_init();
1745
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001746 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747
1748 /* Setup TCP slab cache for open requests. */
1749 tcp_init();
1750
Hideo Aoki95766ff2007-12-31 00:29:24 -08001751 /* Setup UDP memory threshold */
1752 udp_init();
1753
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001754 /* Add UDP-Lite (RFC 3828) */
1755 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001757 ping_init();
1758
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 /*
1760 * Set the ICMP layer up
1761 */
1762
Denis V. Luneva5710d62008-02-29 11:14:50 -08001763 if (icmp_init() < 0)
1764 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765
1766 /*
1767 * Initialise the multicast router
1768 */
1769#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001770 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001771 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772#endif
1773 /*
1774 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001775 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776
Stephen Hemminger132adf52007-03-08 20:44:43 -08001777 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001778 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001779
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 ipv4_proc_init();
1781
1782 ipfrag_init();
1783
Herbert Xu30e224d2005-07-05 14:40:10 -07001784 dev_add_pack(&ip_packet_type);
1785
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 rc = 0;
1787out:
1788 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001789out_unregister_raw_proto:
1790 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791out_unregister_udp_proto:
1792 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001793out_unregister_tcp_proto:
1794 proto_unregister(&tcp_prot);
Amerigo Wange3826f12010-05-05 00:27:06 +00001795out_free_reserved_ports:
1796 kfree(sysctl_local_reserved_ports);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 goto out;
1798}
1799
Heiko Carstensa536e072006-04-28 15:19:17 -07001800fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801
1802/* ------------------------------------------------------------------------ */
1803
1804#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805static int __init ipv4_proc_init(void)
1806{
1807 int rc = 0;
1808
1809 if (raw_proc_init())
1810 goto out_raw;
1811 if (tcp4_proc_init())
1812 goto out_tcp;
1813 if (udp4_proc_init())
1814 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001815 if (ping_proc_init())
1816 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 if (ip_misc_proc_init())
1818 goto out_misc;
1819out:
1820 return rc;
1821out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001822 ping_proc_exit();
1823out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 udp4_proc_exit();
1825out_udp:
1826 tcp4_proc_exit();
1827out_tcp:
1828 raw_proc_exit();
1829out_raw:
1830 rc = -ENOMEM;
1831 goto out;
1832}
1833
1834#else /* CONFIG_PROC_FS */
1835static int __init ipv4_proc_init(void)
1836{
1837 return 0;
1838}
1839#endif /* CONFIG_PROC_FS */
1840
1841MODULE_ALIAS_NETPROTO(PF_INET);
1842