blob: 68f6a94f7661999095d9a02539aee956cd52e8cc [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 Dumazet08dcdbf2013-02-21 12:18:52 +0000251u32 ipv6_hash_secret __read_mostly;
252EXPORT_SYMBOL(ipv6_hash_secret);
253
Eric Dumazetbe776282007-03-27 13:53:04 -0700254/*
Eric Dumazet08dcdbf2013-02-21 12:18:52 +0000255 * inet_ehash_secret must be set exactly once, and to a non nul value
256 * ipv6_hash_secret must be set exactly once.
Eric Dumazetbe776282007-03-27 13:53:04 -0700257 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700258void build_ehash_secret(void)
259{
Eric Dumazetbe776282007-03-27 13:53:04 -0700260 u32 rnd;
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000261
Eric Dumazetbe776282007-03-27 13:53:04 -0700262 do {
263 get_random_bytes(&rnd, sizeof(rnd));
264 } while (rnd == 0);
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000265
Eric Dumazet08dcdbf2013-02-21 12:18:52 +0000266 if (cmpxchg(&inet_ehash_secret, 0, rnd) == 0)
267 get_random_bytes(&ipv6_hash_secret, sizeof(ipv6_hash_secret));
David S. Millerb3da2cf2007-03-23 11:40:27 -0700268}
269EXPORT_SYMBOL(build_ehash_secret);
270
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271/*
272 * Create an inet socket.
273 */
274
Eric Paris3f378b62009-11-05 22:18:14 -0800275static int inet_create(struct net *net, struct socket *sock, int protocol,
276 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277{
278 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 struct inet_protosw *answer;
280 struct inet_sock *inet;
281 struct proto *answer_prot;
282 unsigned char answer_flags;
283 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700284 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800285 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Eric Dumazet04f258c2008-11-23 15:42:23 -0800287 if (unlikely(!inet_ehash_secret))
288 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
289 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 sock->state = SS_UNCONNECTED;
292
293 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700294lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800295 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700297 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700299 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 /* Check the non-wild match. */
301 if (protocol == answer->protocol) {
302 if (protocol != IPPROTO_IP)
303 break;
304 } else {
305 /* Check for the two wild cases. */
306 if (IPPROTO_IP == protocol) {
307 protocol = answer->protocol;
308 break;
309 }
310 if (IPPROTO_IP == answer->protocol)
311 break;
312 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800313 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 }
315
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700316 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700317 if (try_loading_module < 2) {
318 rcu_read_unlock();
319 /*
320 * Be more specific, e.g. net-pf-2-proto-132-type-1
321 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
322 */
323 if (++try_loading_module == 1)
324 request_module("net-pf-%d-proto-%d-type-%d",
325 PF_INET, protocol, sock->type);
326 /*
327 * Fall back to generic, e.g. net-pf-2-proto-132
328 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
329 */
330 else
331 request_module("net-pf-%d-proto-%d",
332 PF_INET, protocol);
333 goto lookup_protocol;
334 } else
335 goto out_rcu_unlock;
336 }
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000339 if (sock->type == SOCK_RAW && !kern &&
340 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 sock->ops = answer->ops;
344 answer_prot = answer->prot;
345 answer_no_check = answer->no_check;
346 answer_flags = answer->flags;
347 rcu_read_unlock();
348
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700349 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700352 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 if (sk == NULL)
354 goto out;
355
356 err = 0;
357 sk->sk_no_check = answer_no_check;
358 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000359 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
361 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800362 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000364 inet->nodefrag = 0;
365
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000367 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 if (IPPROTO_RAW == protocol)
369 inet->hdrincl = 1;
370 }
371
372 if (ipv4_config.no_pmtu_disc)
373 inet->pmtudisc = IP_PMTUDISC_DONT;
374 else
375 inet->pmtudisc = IP_PMTUDISC_WANT;
376
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000377 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
379 sock_init_data(sock, sk);
380
381 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 sk->sk_protocol = protocol;
383 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
384
385 inet->uc_ttl = -1;
386 inet->mc_loop = 1;
387 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000388 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 inet->mc_index = 0;
390 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000391 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700393 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000395 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 /* It assumes that any protocol which allows
397 * the user to assign a number at socket
398 * creation time automatically
399 * shares.
400 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000401 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 /* Add to protocol hash chains. */
403 sk->sk_prot->hash(sk);
404 }
405
406 if (sk->sk_prot->init) {
407 err = sk->sk_prot->init(sk);
408 if (err)
409 sk_common_release(sk);
410 }
411out:
412 return err;
413out_rcu_unlock:
414 rcu_read_unlock();
415 goto out;
416}
417
418
419/*
420 * The peer socket should always be NULL (or else). When we call this
421 * function we are destroying the object and from then on nobody
422 * should refer to it.
423 */
424int inet_release(struct socket *sock)
425{
426 struct sock *sk = sock->sk;
427
428 if (sk) {
429 long timeout;
430
David S. Millerc58dc012010-04-27 15:05:31 -0700431 sock_rps_reset_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 /* Applications forget to leave groups before exiting */
434 ip_mc_drop_socket(sk);
435
436 /* If linger is set, we don't return until the close
437 * is complete. Otherwise we return immediately. The
438 * actually closing is done the same either way.
439 *
440 * If the close is due to the process exiting, we never
441 * linger..
442 */
443 timeout = 0;
444 if (sock_flag(sk, SOCK_LINGER) &&
445 !(current->flags & PF_EXITING))
446 timeout = sk->sk_lingertime;
447 sock->sk = NULL;
448 sk->sk_prot->close(sk, timeout);
449 }
450 return 0;
451}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700452EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700455int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700456EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457
458int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
459{
460 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
461 struct sock *sk = sock->sk;
462 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000463 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 unsigned short snum;
465 int chk_addr_ret;
466 int err;
467
468 /* If the socket has its own bind function then use it. (RAW) */
469 if (sk->sk_prot->bind) {
470 err = sk->sk_prot->bind(sk, uaddr, addr_len);
471 goto out;
472 }
473 err = -EINVAL;
474 if (addr_len < sizeof(struct sockaddr_in))
475 goto out;
476
Marcus Meissnerc349a522011-07-04 01:30:29 +0000477 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400478 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
479 * only if s_addr is INADDR_ANY.
480 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000481 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400482 if (addr->sin_family != AF_UNSPEC ||
483 addr->sin_addr.s_addr != htonl(INADDR_ANY))
484 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000485 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700486
Eric W. Biederman35946982012-11-16 03:03:12 +0000487 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
489 /* Not specified by any standard per-se, however it breaks too
490 * many applications when removed. It is unfortunate since
491 * allowing applications to make a non-local bind solves
492 * several problems with systems using dynamic addressing.
493 * (ie. your servers still start up even if your ISDN link
494 * is temporarily down)
495 */
496 err = -EADDRNOTAVAIL;
497 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700498 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700499 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 chk_addr_ret != RTN_LOCAL &&
501 chk_addr_ret != RTN_MULTICAST &&
502 chk_addr_ret != RTN_BROADCAST)
503 goto out;
504
505 snum = ntohs(addr->sin_port);
506 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000507 if (snum && snum < PROT_SOCK &&
508 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 goto out;
510
511 /* We keep a pair of addresses. rcv_saddr is the one
512 * used by hash lookups, and saddr is used for transmit.
513 *
514 * In the BSD API these are the same except where it
515 * would be illegal to use them (multicast/broadcast) in
516 * which case the sending device address is used.
517 */
518 lock_sock(sk);
519
520 /* Check these errors (active socket, double bind). */
521 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000522 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 goto out_release_sock;
524
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000525 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000527 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
529 /* Make sure we are allowed to bind here. */
530 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000531 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 err = -EADDRINUSE;
533 goto out_release_sock;
534 }
535
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000536 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
538 if (snum)
539 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000540 inet->inet_sport = htons(inet->inet_num);
541 inet->inet_daddr = 0;
542 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 sk_dst_reset(sk);
544 err = 0;
545out_release_sock:
546 release_sock(sk);
547out:
548 return err;
549}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700550EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000552int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 int addr_len, int flags)
554{
555 struct sock *sk = sock->sk;
556
Changli Gao6503d962010-03-31 22:58:26 +0000557 if (addr_len < sizeof(uaddr->sa_family))
558 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 if (uaddr->sa_family == AF_UNSPEC)
560 return sk->sk_prot->disconnect(sk, flags);
561
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000562 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000564 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700566EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
Yuchung Cheng783237e2012-07-19 06:43:07 +0000568static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
570 DEFINE_WAIT(wait);
571
Eric Dumazetaa395142010-04-20 13:03:51 +0000572 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000573 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574
575 /* Basic assumption: if someone sets sk->sk_err, he _must_
576 * change state of the socket from TCP_SYN_*.
577 * Connect() does not allow to get error notifications
578 * without closing the socket.
579 */
580 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
581 release_sock(sk);
582 timeo = schedule_timeout(timeo);
583 lock_sock(sk);
584 if (signal_pending(current) || !timeo)
585 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000586 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000588 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000589 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 return timeo;
591}
592
593/*
594 * Connect to a remote host. There is regrettably still a little
595 * TCP 'magic' in here.
596 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000597int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
598 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
600 struct sock *sk = sock->sk;
601 int err;
602 long timeo;
603
Changli Gao6503d962010-03-31 22:58:26 +0000604 if (addr_len < sizeof(uaddr->sa_family))
605 return -EINVAL;
606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 if (uaddr->sa_family == AF_UNSPEC) {
608 err = sk->sk_prot->disconnect(sk, flags);
609 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
610 goto out;
611 }
612
613 switch (sock->state) {
614 default:
615 err = -EINVAL;
616 goto out;
617 case SS_CONNECTED:
618 err = -EISCONN;
619 goto out;
620 case SS_CONNECTING:
621 err = -EALREADY;
622 /* Fall out of switch with err, set for this state */
623 break;
624 case SS_UNCONNECTED:
625 err = -EISCONN;
626 if (sk->sk_state != TCP_CLOSE)
627 goto out;
628
629 err = sk->sk_prot->connect(sk, uaddr, addr_len);
630 if (err < 0)
631 goto out;
632
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900633 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635 /* Just entered SS_CONNECTING state; the only
636 * difference is that return value in non-blocking
637 * case is EINPROGRESS, rather than EALREADY.
638 */
639 err = -EINPROGRESS;
640 break;
641 }
642
643 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
644
645 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000646 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
647 tcp_sk(sk)->fastopen_req &&
648 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
649
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000651 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 goto out;
653
654 err = sock_intr_errno(timeo);
655 if (signal_pending(current))
656 goto out;
657 }
658
659 /* Connection was closed by RST, timeout, ICMP error
660 * or another process disconnected us.
661 */
662 if (sk->sk_state == TCP_CLOSE)
663 goto sock_error;
664
665 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
666 * and error was received after socket entered established state.
667 * Hence, it is handled normally after connect() return successfully.
668 */
669
670 sock->state = SS_CONNECTED;
671 err = 0;
672out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 return err;
674
675sock_error:
676 err = sock_error(sk) ? : -ECONNABORTED;
677 sock->state = SS_UNCONNECTED;
678 if (sk->sk_prot->disconnect(sk, flags))
679 sock->state = SS_DISCONNECTING;
680 goto out;
681}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000682EXPORT_SYMBOL(__inet_stream_connect);
683
684int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
685 int addr_len, int flags)
686{
687 int err;
688
689 lock_sock(sock->sk);
690 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
691 release_sock(sock->sk);
692 return err;
693}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700694EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696/*
697 * Accept a pending connection. The TCP layer now gives BSD semantics.
698 */
699
700int inet_accept(struct socket *sock, struct socket *newsock, int flags)
701{
702 struct sock *sk1 = sock->sk;
703 int err = -EINVAL;
704 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
705
706 if (!sk2)
707 goto do_err;
708
709 lock_sock(sk2);
710
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000711 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700712 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000713 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
714 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
716 sock_graft(sk2, newsock);
717
718 newsock->state = SS_CONNECTED;
719 err = 0;
720 release_sock(sk2);
721do_err:
722 return err;
723}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700724EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
726
727/*
728 * This does both peername and sockname.
729 */
730int inet_getname(struct socket *sock, struct sockaddr *uaddr,
731 int *uaddr_len, int peer)
732{
733 struct sock *sk = sock->sk;
734 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700735 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
737 sin->sin_family = AF_INET;
738 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000739 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
741 peer == 1))
742 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000743 sin->sin_port = inet->inet_dport;
744 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000746 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000748 addr = inet->inet_saddr;
749 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 sin->sin_addr.s_addr = addr;
751 }
752 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
753 *uaddr_len = sizeof(*sin);
754 return 0;
755}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700756EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
759 size_t size)
760{
761 struct sock *sk = sock->sk;
762
David S. Millerc58dc012010-04-27 15:05:31 -0700763 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700764
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000766 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
767 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 return -EAGAIN;
769
770 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
771}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700772EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Changli Gao7ba42912010-07-10 20:41:55 +0000774ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
775 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776{
777 struct sock *sk = sock->sk;
778
David S. Millerc58dc012010-04-27 15:05:31 -0700779 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000782 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
783 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 return -EAGAIN;
785
786 if (sk->sk_prot->sendpage)
787 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
788 return sock_no_sendpage(sock, page, offset, size, flags);
789}
Changli Gao7ba42912010-07-10 20:41:55 +0000790EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
Tom Herbertfec5e652010-04-16 16:01:27 -0700792int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
793 size_t size, int flags)
794{
795 struct sock *sk = sock->sk;
796 int addr_len = 0;
797 int err;
798
David S. Millerc58dc012010-04-27 15:05:31 -0700799 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700800
801 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
802 flags & ~MSG_DONTWAIT, &addr_len);
803 if (err >= 0)
804 msg->msg_namelen = addr_len;
805 return err;
806}
807EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
809int inet_shutdown(struct socket *sock, int how)
810{
811 struct sock *sk = sock->sk;
812 int err = 0;
813
814 /* This should really check to make sure
815 * the socket is a TCP socket. (WHY AC...)
816 */
817 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
818 1->2 bit 2 snds.
819 2->3 */
820 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
821 return -EINVAL;
822
823 lock_sock(sk);
824 if (sock->state == SS_CONNECTING) {
825 if ((1 << sk->sk_state) &
826 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
827 sock->state = SS_DISCONNECTING;
828 else
829 sock->state = SS_CONNECTED;
830 }
831
832 switch (sk->sk_state) {
833 case TCP_CLOSE:
834 err = -ENOTCONN;
835 /* Hack to wake up other listeners, who can poll for
836 POLLHUP, even on eg. unconnected UDP sockets -- RR */
837 default:
838 sk->sk_shutdown |= how;
839 if (sk->sk_prot->shutdown)
840 sk->sk_prot->shutdown(sk, how);
841 break;
842
843 /* Remaining two branches are temporary solution for missing
844 * close() in multithreaded environment. It is _not_ a good idea,
845 * but we have no choice until close() is repaired at VFS level.
846 */
847 case TCP_LISTEN:
848 if (!(how & RCV_SHUTDOWN))
849 break;
850 /* Fall through */
851 case TCP_SYN_SENT:
852 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
853 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
854 break;
855 }
856
857 /* Wake up anyone sleeping in poll. */
858 sk->sk_state_change(sk);
859 release_sock(sk);
860 return err;
861}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700862EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864/*
865 * ioctl() calls you can issue on an INET socket. Most of these are
866 * device configuration and stuff and very rarely used. Some ioctls
867 * pass on to the socket itself.
868 *
869 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
870 * loads the devconfigure module does its configuring and unloads it.
871 * There's a good 20K of config code hanging around the kernel.
872 */
873
874int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
875{
876 struct sock *sk = sock->sk;
877 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900878 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
880 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700881 case SIOCGSTAMP:
882 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
883 break;
884 case SIOCGSTAMPNS:
885 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
886 break;
887 case SIOCADDRT:
888 case SIOCDELRT:
889 case SIOCRTMSG:
890 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
891 break;
892 case SIOCDARP:
893 case SIOCGARP:
894 case SIOCSARP:
895 err = arp_ioctl(net, cmd, (void __user *)arg);
896 break;
897 case SIOCGIFADDR:
898 case SIOCSIFADDR:
899 case SIOCGIFBRDADDR:
900 case SIOCSIFBRDADDR:
901 case SIOCGIFNETMASK:
902 case SIOCSIFNETMASK:
903 case SIOCGIFDSTADDR:
904 case SIOCSIFDSTADDR:
905 case SIOCSIFPFLAGS:
906 case SIOCGIFPFLAGS:
907 case SIOCSIFFLAGS:
908 err = devinet_ioctl(net, cmd, (void __user *)arg);
909 break;
910 default:
911 if (sk->sk_prot->ioctl)
912 err = sk->sk_prot->ioctl(sk, cmd, arg);
913 else
914 err = -ENOIOCTLCMD;
915 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 }
917 return err;
918}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700919EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000921#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000922static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000923{
924 struct sock *sk = sock->sk;
925 int err = -ENOIOCTLCMD;
926
927 if (sk->sk_prot->compat_ioctl)
928 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
929
930 return err;
931}
932#endif
933
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800934const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800935 .family = PF_INET,
936 .owner = THIS_MODULE,
937 .release = inet_release,
938 .bind = inet_bind,
939 .connect = inet_stream_connect,
940 .socketpair = sock_no_socketpair,
941 .accept = inet_accept,
942 .getname = inet_getname,
943 .poll = tcp_poll,
944 .ioctl = inet_ioctl,
945 .listen = inet_listen,
946 .shutdown = inet_shutdown,
947 .setsockopt = sock_common_setsockopt,
948 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000949 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700950 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800951 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000952 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800953 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800954#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800955 .compat_setsockopt = compat_sock_common_setsockopt,
956 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000957 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800958#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700960EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800962const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800963 .family = PF_INET,
964 .owner = THIS_MODULE,
965 .release = inet_release,
966 .bind = inet_bind,
967 .connect = inet_dgram_connect,
968 .socketpair = sock_no_socketpair,
969 .accept = sock_no_accept,
970 .getname = inet_getname,
971 .poll = udp_poll,
972 .ioctl = inet_ioctl,
973 .listen = sock_no_listen,
974 .shutdown = inet_shutdown,
975 .setsockopt = sock_common_setsockopt,
976 .getsockopt = sock_common_getsockopt,
977 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700978 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800979 .mmap = sock_no_mmap,
980 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800981#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800982 .compat_setsockopt = compat_sock_common_setsockopt,
983 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000984 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800985#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700987EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
989/*
990 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
991 * udp_poll
992 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800993static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800994 .family = PF_INET,
995 .owner = THIS_MODULE,
996 .release = inet_release,
997 .bind = inet_bind,
998 .connect = inet_dgram_connect,
999 .socketpair = sock_no_socketpair,
1000 .accept = sock_no_accept,
1001 .getname = inet_getname,
1002 .poll = datagram_poll,
1003 .ioctl = inet_ioctl,
1004 .listen = sock_no_listen,
1005 .shutdown = inet_shutdown,
1006 .setsockopt = sock_common_setsockopt,
1007 .getsockopt = sock_common_getsockopt,
1008 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001009 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001010 .mmap = sock_no_mmap,
1011 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001012#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001013 .compat_setsockopt = compat_sock_common_setsockopt,
1014 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001015 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001016#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017};
1018
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001019static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 .family = PF_INET,
1021 .create = inet_create,
1022 .owner = THIS_MODULE,
1023};
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025/* Upon startup we insert all the elements in inetsw_array[] into
1026 * the linked list inetsw.
1027 */
1028static struct inet_protosw inetsw_array[] =
1029{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001030 {
1031 .type = SOCK_STREAM,
1032 .protocol = IPPROTO_TCP,
1033 .prot = &tcp_prot,
1034 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001035 .no_check = 0,
1036 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001037 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001038 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001040 {
1041 .type = SOCK_DGRAM,
1042 .protocol = IPPROTO_UDP,
1043 .prot = &udp_prot,
1044 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001045 .no_check = UDP_CSUM_DEFAULT,
1046 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001048
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001049 {
1050 .type = SOCK_DGRAM,
1051 .protocol = IPPROTO_ICMP,
1052 .prot = &ping_prot,
1053 .ops = &inet_dgram_ops,
1054 .no_check = UDP_CSUM_DEFAULT,
1055 .flags = INET_PROTOSW_REUSE,
1056 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057
1058 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001059 .type = SOCK_RAW,
1060 .protocol = IPPROTO_IP, /* wild card */
1061 .prot = &raw_prot,
1062 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001063 .no_check = UDP_CSUM_DEFAULT,
1064 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 }
1066};
1067
Denis Chengc40f6ff2007-09-16 16:39:25 -07001068#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
1070void inet_register_protosw(struct inet_protosw *p)
1071{
1072 struct list_head *lh;
1073 struct inet_protosw *answer;
1074 int protocol = p->protocol;
1075 struct list_head *last_perm;
1076
1077 spin_lock_bh(&inetsw_lock);
1078
1079 if (p->type >= SOCK_MAX)
1080 goto out_illegal;
1081
1082 /* If we are trying to override a permanent protocol, bail. */
1083 answer = NULL;
1084 last_perm = &inetsw[p->type];
1085 list_for_each(lh, &inetsw[p->type]) {
1086 answer = list_entry(lh, struct inet_protosw, list);
1087
1088 /* Check only the non-wild match. */
1089 if (INET_PROTOSW_PERMANENT & answer->flags) {
1090 if (protocol == answer->protocol)
1091 break;
1092 last_perm = lh;
1093 }
1094
1095 answer = NULL;
1096 }
1097 if (answer)
1098 goto out_permanent;
1099
1100 /* Add the new entry after the last permanent entry if any, so that
1101 * the new entry does not override a permanent entry when matched with
1102 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001103 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 * system automatically returns to the old behavior.
1105 */
1106 list_add_rcu(&p->list, last_perm);
1107out:
1108 spin_unlock_bh(&inetsw_lock);
1109
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 return;
1111
1112out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001113 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 goto out;
1115
1116out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001117 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 p->type);
1119 goto out;
1120}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001121EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122
1123void inet_unregister_protosw(struct inet_protosw *p)
1124{
1125 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001126 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 p->protocol);
1128 } else {
1129 spin_lock_bh(&inetsw_lock);
1130 list_del_rcu(&p->list);
1131 spin_unlock_bh(&inetsw_lock);
1132
1133 synchronize_net();
1134 }
1135}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001136EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001138/*
1139 * Shall we try to damage output packets if routing dev changes?
1140 */
1141
Brian Haleyab32ea52006-09-22 14:15:41 -07001142int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001143
1144static int inet_sk_reselect_saddr(struct sock *sk)
1145{
1146 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001147 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001148 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001149 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001150 struct rtable *rt;
1151 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001152 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001153
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001154 inet_opt = rcu_dereference_protected(inet->inet_opt,
1155 sock_owned_by_user(sk));
1156 if (inet_opt && inet_opt->opt.srr)
1157 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001158
1159 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001160 fl4 = &inet->cork.fl.u.ip4;
1161 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001162 sk->sk_bound_dev_if, sk->sk_protocol,
1163 inet->inet_sport, inet->inet_dport, sk, false);
1164 if (IS_ERR(rt))
1165 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001166
Changli Gaod8d1f302010-06-10 23:31:35 -07001167 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001168
David S. Miller6e869132011-05-06 16:18:04 -07001169 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001170
1171 if (new_saddr == old_saddr)
1172 return 0;
1173
1174 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001175 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1176 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001177 }
1178
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001179 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001180
1181 /*
1182 * XXX The only one ugly spot where we need to
1183 * XXX really change the sockets identity after
1184 * XXX it has entered the hashes. -DaveM
1185 *
1186 * Besides that, it does not check for connection
1187 * uniqueness. Wait for troubles.
1188 */
1189 __sk_prot_rehash(sk);
1190 return 0;
1191}
1192
1193int inet_sk_rebuild_header(struct sock *sk)
1194{
1195 struct inet_sock *inet = inet_sk(sk);
1196 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001197 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001198 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001199 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001200 int err;
1201
1202 /* Route is OK, nothing to do. */
1203 if (rt)
1204 return 0;
1205
1206 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001207 rcu_read_lock();
1208 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001209 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001210 if (inet_opt && inet_opt->opt.srr)
1211 daddr = inet_opt->opt.faddr;
1212 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001213 fl4 = &inet->cork.fl.u.ip4;
1214 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001215 inet->inet_dport, inet->inet_sport,
1216 sk->sk_protocol, RT_CONN_FLAGS(sk),
1217 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001218 if (!IS_ERR(rt)) {
1219 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001220 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001221 } else {
1222 err = PTR_ERR(rt);
1223
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001224 /* Routing failed... */
1225 sk->sk_route_caps = 0;
1226 /*
1227 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1228 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1229 */
1230 if (!sysctl_ip_dynaddr ||
1231 sk->sk_state != TCP_SYN_SENT ||
1232 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1233 (err = inet_sk_reselect_saddr(sk)) != 0)
1234 sk->sk_err_soft = -err;
1235 }
1236
1237 return err;
1238}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001239EXPORT_SYMBOL(inet_sk_rebuild_header);
1240
Herbert Xua430a432006-07-08 13:34:56 -07001241static int inet_gso_send_check(struct sk_buff *skb)
1242{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001243 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001244 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -07001245 int proto;
1246 int ihl;
1247 int err = -EINVAL;
1248
1249 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1250 goto out;
1251
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001252 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001253 ihl = iph->ihl * 4;
1254 if (ihl < sizeof(*iph))
1255 goto out;
1256
1257 if (unlikely(!pskb_may_pull(skb, ihl)))
1258 goto out;
1259
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001260 __skb_pull(skb, ihl);
1261 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001262 iph = ip_hdr(skb);
David S. Millerf9242b62012-06-19 18:56:21 -07001263 proto = iph->protocol;
Herbert Xua430a432006-07-08 13:34:56 -07001264 err = -EPROTONOSUPPORT;
1265
1266 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001267 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001268 if (likely(ops && ops->callbacks.gso_send_check))
1269 err = ops->callbacks.gso_send_check(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001270 rcu_read_unlock();
1271
1272out:
1273 return err;
1274}
1275
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001276static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1277 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001278{
1279 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001280 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001281 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001282 int proto;
1283 int ihl;
1284 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001285 unsigned int offset = 0;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001286
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001287 if (!(features & NETIF_F_V4_CSUM))
1288 features &= ~NETIF_F_SG;
1289
Herbert Xubbcf4672006-07-03 19:38:35 -07001290 if (unlikely(skb_shinfo(skb)->gso_type &
1291 ~(SKB_GSO_TCPV4 |
1292 SKB_GSO_UDP |
1293 SKB_GSO_DODGY |
1294 SKB_GSO_TCP_ECN |
Pravin B Shelar68c33162013-02-14 14:02:41 +00001295 SKB_GSO_GRE |
Herbert Xubbcf4672006-07-03 19:38:35 -07001296 0)))
1297 goto out;
1298
1299 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001300 goto out;
1301
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001302 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001303 ihl = iph->ihl * 4;
1304 if (ihl < sizeof(*iph))
1305 goto out;
1306
Herbert Xubbcf4672006-07-03 19:38:35 -07001307 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001308 goto out;
1309
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001310 __skb_pull(skb, ihl);
1311 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001312 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001313 id = ntohs(iph->id);
David S. Millerf9242b62012-06-19 18:56:21 -07001314 proto = iph->protocol;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001315 segs = ERR_PTR(-EPROTONOSUPPORT);
1316
1317 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001318 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001319 if (likely(ops && ops->callbacks.gso_segment))
1320 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001321 rcu_read_unlock();
1322
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001323 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001324 goto out;
1325
1326 skb = segs;
1327 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001328 iph = ip_hdr(skb);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001329 if (proto == IPPROTO_UDP) {
1330 iph->id = htons(id);
1331 iph->frag_off = htons(offset >> 3);
1332 if (skb->next != NULL)
1333 iph->frag_off |= htons(IP_MF);
1334 offset += (skb->len - skb->mac_len - iph->ihl * 4);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001335 } else {
1336 if (!(iph->frag_off & htons(IP_DF)))
1337 iph->id = htons(id++);
1338 }
Herbert Xuf4c50d92006-06-22 03:02:40 -07001339 iph->tot_len = htons(skb->len - skb->mac_len);
1340 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001341 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001342 } while ((skb = skb->next));
1343
1344out:
1345 return segs;
1346}
1347
Herbert Xu73cc19f2008-12-15 23:41:09 -08001348static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1349 struct sk_buff *skb)
1350{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001351 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001352 struct sk_buff **pp = NULL;
1353 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001354 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001355 unsigned int hlen;
1356 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001357 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001358 int flush = 1;
1359 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001360
Herbert Xua5b1cf22009-05-26 18:50:28 +00001361 off = skb_gro_offset(skb);
1362 hlen = off + sizeof(*iph);
1363 iph = skb_gro_header_fast(skb, off);
1364 if (skb_gro_header_hard(skb, hlen)) {
1365 iph = skb_gro_header_slow(skb, hlen, off);
1366 if (unlikely(!iph))
1367 goto out;
1368 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001369
David S. Millerf9242b62012-06-19 18:56:21 -07001370 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001371
1372 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001373 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001374 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001375 goto out_unlock;
1376
Herbert Xua5ad24b2009-02-08 18:00:39 +00001377 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001378 goto out_unlock;
1379
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001380 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001381 goto out_unlock;
1382
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001383 id = ntohl(*(__be32 *)&iph->id);
1384 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001385 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001386
1387 for (p = *head; p; p = p->next) {
1388 struct iphdr *iph2;
1389
1390 if (!NAPI_GRO_CB(p)->same_flow)
1391 continue;
1392
1393 iph2 = ip_hdr(p);
1394
Herbert Xua5ad24b2009-02-08 18:00:39 +00001395 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001396 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1397 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001398 NAPI_GRO_CB(p)->same_flow = 0;
1399 continue;
1400 }
1401
1402 /* All fields must match except length and checksum. */
1403 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001404 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001405 (iph->tos ^ iph2->tos) |
Herbert Xua5ad24b2009-02-08 18:00:39 +00001406 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001407
1408 NAPI_GRO_CB(p)->flush |= flush;
1409 }
1410
1411 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001412 skb_gro_pull(skb, sizeof(*iph));
1413 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001414
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001415 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001416
1417out_unlock:
1418 rcu_read_unlock();
1419
1420out:
1421 NAPI_GRO_CB(skb)->flush |= flush;
1422
1423 return pp;
1424}
1425
1426static int inet_gro_complete(struct sk_buff *skb)
1427{
Herbert Xu73cc19f2008-12-15 23:41:09 -08001428 __be16 newlen = htons(skb->len - skb_network_offset(skb));
David S. Millerf9242b62012-06-19 18:56:21 -07001429 struct iphdr *iph = ip_hdr(skb);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001430 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001431 int proto = iph->protocol;
1432 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001433
1434 csum_replace2(&iph->check, iph->tot_len, newlen);
1435 iph->tot_len = newlen;
1436
1437 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001438 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001439 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001440 goto out_unlock;
1441
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001442 err = ops->callbacks.gro_complete(skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001443
1444out_unlock:
1445 rcu_read_unlock();
1446
1447 return err;
1448}
1449
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001450int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001451 unsigned short type, unsigned char protocol,
1452 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001453{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001454 struct socket *sock;
1455 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001456
1457 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001458 *sk = sock->sk;
1459 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001460 /*
1461 * Unhash it so that IP input processing does not even see it,
1462 * we do not wish this socket to see incoming packets.
1463 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001464 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001465
1466 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001467 }
1468 return rc;
1469}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001470EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1471
Tejun Heo7d720c32010-02-16 15:20:26 +00001472unsigned long snmp_fold_field(void __percpu *mib[], int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001473{
1474 unsigned long res = 0;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001475 int i, j;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001476
1477 for_each_possible_cpu(i) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001478 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1479 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001480 }
1481 return res;
1482}
1483EXPORT_SYMBOL_GPL(snmp_fold_field);
1484
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001485#if BITS_PER_LONG==32
1486
1487u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1488{
1489 u64 res = 0;
1490 int cpu;
1491
1492 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001493 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001494 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001495 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001496 unsigned int start;
1497
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001498 bhptr = per_cpu_ptr(mib[0], cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001499 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1500 do {
1501 start = u64_stats_fetch_begin_bh(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001502 v = *(((u64 *) bhptr) + offt);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001503 } while (u64_stats_fetch_retry_bh(syncp, start));
1504
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001505 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001506 }
1507 return res;
1508}
1509EXPORT_SYMBOL_GPL(snmp_fold_field64);
1510#endif
1511
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001512int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001513{
1514 BUG_ON(ptr == NULL);
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001515 ptr[0] = __alloc_percpu(mibsize, align);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001516 if (!ptr[0])
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001517 return -ENOMEM;
1518#if SNMP_ARRAY_SZ == 2
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001519 ptr[1] = __alloc_percpu(mibsize, align);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001520 if (!ptr[1]) {
1521 free_percpu(ptr[0]);
1522 ptr[0] = NULL;
1523 return -ENOMEM;
1524 }
1525#endif
Herbert Xu5e0f0432007-04-24 21:53:35 -07001526 return 0;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001527}
1528EXPORT_SYMBOL_GPL(snmp_mib_init);
1529
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001530void snmp_mib_free(void __percpu *ptr[SNMP_ARRAY_SZ])
Herbert Xu5e0f0432007-04-24 21:53:35 -07001531{
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001532 int i;
1533
Herbert Xu5e0f0432007-04-24 21:53:35 -07001534 BUG_ON(ptr == NULL);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001535 for (i = 0; i < SNMP_ARRAY_SZ; i++) {
1536 free_percpu(ptr[i]);
1537 ptr[i] = NULL;
1538 }
Herbert Xu5e0f0432007-04-24 21:53:35 -07001539}
1540EXPORT_SYMBOL_GPL(snmp_mib_free);
1541
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001543static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001545 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546};
1547#endif
1548
Alexey Dobriyan32613092009-09-14 12:21:47 +00001549static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001550 .early_demux = tcp_v4_early_demux,
1551 .handler = tcp_v4_rcv,
1552 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001553 .no_policy = 1,
1554 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555};
1556
Vlad Yasevichde27d002012-11-15 08:49:12 +00001557static const struct net_offload tcp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001558 .callbacks = {
1559 .gso_send_check = tcp_v4_gso_send_check,
1560 .gso_segment = tcp_tso_segment,
1561 .gro_receive = tcp4_gro_receive,
1562 .gro_complete = tcp4_gro_complete,
1563 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001564};
1565
Alexey Dobriyan32613092009-09-14 12:21:47 +00001566static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 .handler = udp_rcv,
1568 .err_handler = udp_err,
1569 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001570 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571};
1572
Vlad Yasevichde27d002012-11-15 08:49:12 +00001573static const struct net_offload udp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001574 .callbacks = {
1575 .gso_send_check = udp4_ufo_send_check,
1576 .gso_segment = udp4_ufo_fragment,
1577 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001578};
1579
Alexey Dobriyan32613092009-09-14 12:21:47 +00001580static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001582 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001583 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001584 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585};
1586
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001587static __net_init int ipv4_mib_init_net(struct net *net)
1588{
Tejun Heo7d720c32010-02-16 15:20:26 +00001589 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001590 sizeof(struct tcp_mib),
1591 __alignof__(struct tcp_mib)) < 0)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001592 goto err_tcp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001593 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001594 sizeof(struct ipstats_mib),
1595 __alignof__(struct ipstats_mib)) < 0)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001596 goto err_ip_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001597 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001598 sizeof(struct linux_mib),
1599 __alignof__(struct linux_mib)) < 0)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001600 goto err_net_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001601 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001602 sizeof(struct udp_mib),
1603 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001604 goto err_udp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001605 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001606 sizeof(struct udp_mib),
1607 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001608 goto err_udplite_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001609 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001610 sizeof(struct icmp_mib),
1611 __alignof__(struct icmp_mib)) < 0)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001612 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001613 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1614 GFP_KERNEL);
1615 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001616 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001617
1618 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001619 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001620
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001621err_icmpmsg_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001622 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001623err_icmp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001624 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001625err_udplite_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001626 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001627err_udp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001628 snmp_mib_free((void __percpu **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001629err_net_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001630 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001631err_ip_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001632 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001633err_tcp_mib:
1634 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001635}
1636
1637static __net_exit void ipv4_mib_exit_net(struct net *net)
1638{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001639 kfree(net->mib.icmpmsg_statistics);
Tejun Heo7d720c32010-02-16 15:20:26 +00001640 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1641 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1642 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1643 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1644 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1645 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001646}
1647
1648static __net_initdata struct pernet_operations ipv4_mib_ops = {
1649 .init = ipv4_mib_init_net,
1650 .exit = ipv4_mib_exit_net,
1651};
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653static int __init init_ipv4_mibs(void)
1654{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001655 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656}
1657
1658static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659
Herbert Xu30e224d2005-07-05 14:40:10 -07001660/*
1661 * IP protocol layer initialiser
1662 */
1663
Vlad Yasevich22061d82012-11-15 08:49:11 +00001664static struct packet_offload ip_packet_offload __read_mostly = {
1665 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001666 .callbacks = {
1667 .gso_send_check = inet_gso_send_check,
1668 .gso_segment = inet_gso_segment,
1669 .gro_receive = inet_gro_receive,
1670 .gro_complete = inet_gro_complete,
1671 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001672};
1673
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001674static int __init ipv4_offload_init(void)
1675{
1676 /*
1677 * Add offloads
1678 */
1679 if (inet_add_offload(&udp_offload, IPPROTO_UDP) < 0)
1680 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
1681 if (inet_add_offload(&tcp_offload, IPPROTO_TCP) < 0)
1682 pr_crit("%s: Cannot add TCP protocol offlaod\n", __func__);
1683
1684 dev_add_offload(&ip_packet_offload);
1685 return 0;
1686}
1687
1688fs_initcall(ipv4_offload_init);
1689
1690static struct packet_type ip_packet_type __read_mostly = {
1691 .type = cpu_to_be16(ETH_P_IP),
1692 .func = ip_rcv,
1693};
1694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695static int __init inet_init(void)
1696{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 struct inet_protosw *q;
1698 struct list_head *r;
1699 int rc = -EINVAL;
1700
YOSHIFUJI Hideaki / 吉藤英明95c7e0e2013-01-09 07:19:37 +00001701 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702
Amerigo Wange3826f12010-05-05 00:27:06 +00001703 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1704 if (!sysctl_local_reserved_ports)
1705 goto out;
1706
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 rc = proto_register(&tcp_prot, 1);
1708 if (rc)
Amerigo Wange3826f12010-05-05 00:27:06 +00001709 goto out_free_reserved_ports;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710
1711 rc = proto_register(&udp_prot, 1);
1712 if (rc)
1713 goto out_unregister_tcp_proto;
1714
1715 rc = proto_register(&raw_prot, 1);
1716 if (rc)
1717 goto out_unregister_udp_proto;
1718
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001719 rc = proto_register(&ping_prot, 1);
1720 if (rc)
1721 goto out_unregister_raw_proto;
1722
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001724 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 */
1726
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001727 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728
Al Virobd7b1532008-07-15 16:00:59 -04001729#ifdef CONFIG_SYSCTL
1730 ip_static_sysctl_init();
1731#endif
1732
Glauber Costa3dc43e32011-12-11 21:47:05 +00001733 tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 /*
1736 * Add all the base protocols.
1737 */
1738
1739 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001740 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001742 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001744 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745#ifdef CONFIG_IP_MULTICAST
1746 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001747 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748#endif
1749
1750 /* Register the socket-side information for inet_create. */
1751 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1752 INIT_LIST_HEAD(r);
1753
1754 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1755 inet_register_protosw(q);
1756
1757 /*
1758 * Set the ARP module up
1759 */
1760
1761 arp_init();
1762
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001763 /*
1764 * Set the IP module up
1765 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
1767 ip_init();
1768
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001769 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770
1771 /* Setup TCP slab cache for open requests. */
1772 tcp_init();
1773
Hideo Aoki95766ff2007-12-31 00:29:24 -08001774 /* Setup UDP memory threshold */
1775 udp_init();
1776
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001777 /* Add UDP-Lite (RFC 3828) */
1778 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001780 ping_init();
1781
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 /*
1783 * Set the ICMP layer up
1784 */
1785
Denis V. Luneva5710d62008-02-29 11:14:50 -08001786 if (icmp_init() < 0)
1787 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788
1789 /*
1790 * Initialise the multicast router
1791 */
1792#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001793 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001794 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795#endif
1796 /*
1797 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001798 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799
Stephen Hemminger132adf52007-03-08 20:44:43 -08001800 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001801 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001802
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 ipv4_proc_init();
1804
1805 ipfrag_init();
1806
Herbert Xu30e224d2005-07-05 14:40:10 -07001807 dev_add_pack(&ip_packet_type);
1808
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 rc = 0;
1810out:
1811 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001812out_unregister_raw_proto:
1813 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814out_unregister_udp_proto:
1815 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001816out_unregister_tcp_proto:
1817 proto_unregister(&tcp_prot);
Amerigo Wange3826f12010-05-05 00:27:06 +00001818out_free_reserved_ports:
1819 kfree(sysctl_local_reserved_ports);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 goto out;
1821}
1822
Heiko Carstensa536e072006-04-28 15:19:17 -07001823fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824
1825/* ------------------------------------------------------------------------ */
1826
1827#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828static int __init ipv4_proc_init(void)
1829{
1830 int rc = 0;
1831
1832 if (raw_proc_init())
1833 goto out_raw;
1834 if (tcp4_proc_init())
1835 goto out_tcp;
1836 if (udp4_proc_init())
1837 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001838 if (ping_proc_init())
1839 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 if (ip_misc_proc_init())
1841 goto out_misc;
1842out:
1843 return rc;
1844out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001845 ping_proc_exit();
1846out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 udp4_proc_exit();
1848out_udp:
1849 tcp4_proc_exit();
1850out_tcp:
1851 raw_proc_exit();
1852out_raw:
1853 rc = -ENOMEM;
1854 goto out;
1855}
1856
1857#else /* CONFIG_PROC_FS */
1858static int __init ipv4_proc_init(void)
1859{
1860 return 0;
1861}
1862#endif /* CONFIG_PROC_FS */
1863
1864MODULE_ALIAS_NETPROTO(PF_INET);
1865