blob: 09d78d4a3cff0ede57130b6c0e6cae11aaa61c5c [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#include <net/inet_common.h>
115#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700116#include <net/net_namespace.h>
Eric Dumazetaebda152013-04-29 05:58:52 +0000117#include <net/secure_seq.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
248/*
249 * Create an inet socket.
250 */
251
Eric Paris3f378b62009-11-05 22:18:14 -0800252static int inet_create(struct net *net, struct socket *sock, int protocol,
253 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254{
255 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 struct inet_protosw *answer;
257 struct inet_sock *inet;
258 struct proto *answer_prot;
259 unsigned char answer_flags;
260 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700261 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800262 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
264 sock->state = SS_UNCONNECTED;
265
266 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700267lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800268 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700270 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700272 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 /* Check the non-wild match. */
274 if (protocol == answer->protocol) {
275 if (protocol != IPPROTO_IP)
276 break;
277 } else {
278 /* Check for the two wild cases. */
279 if (IPPROTO_IP == protocol) {
280 protocol = answer->protocol;
281 break;
282 }
283 if (IPPROTO_IP == answer->protocol)
284 break;
285 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800286 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 }
288
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700289 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700290 if (try_loading_module < 2) {
291 rcu_read_unlock();
292 /*
293 * Be more specific, e.g. net-pf-2-proto-132-type-1
294 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
295 */
296 if (++try_loading_module == 1)
297 request_module("net-pf-%d-proto-%d-type-%d",
298 PF_INET, protocol, sock->type);
299 /*
300 * Fall back to generic, e.g. net-pf-2-proto-132
301 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
302 */
303 else
304 request_module("net-pf-%d-proto-%d",
305 PF_INET, protocol);
306 goto lookup_protocol;
307 } else
308 goto out_rcu_unlock;
309 }
310
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000312 if (sock->type == SOCK_RAW && !kern &&
313 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316 sock->ops = answer->ops;
317 answer_prot = answer->prot;
318 answer_no_check = answer->no_check;
319 answer_flags = answer->flags;
320 rcu_read_unlock();
321
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700322 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
324 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700325 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 if (sk == NULL)
327 goto out;
328
329 err = 0;
330 sk->sk_no_check = answer_no_check;
331 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000332 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800335 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000337 inet->nodefrag = 0;
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000340 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 if (IPPROTO_RAW == protocol)
342 inet->hdrincl = 1;
343 }
344
345 if (ipv4_config.no_pmtu_disc)
346 inet->pmtudisc = IP_PMTUDISC_DONT;
347 else
348 inet->pmtudisc = IP_PMTUDISC_WANT;
349
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000350 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
352 sock_init_data(sock, sk);
353
354 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 sk->sk_protocol = protocol;
356 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
357
358 inet->uc_ttl = -1;
359 inet->mc_loop = 1;
360 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000361 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 inet->mc_index = 0;
363 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000364 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700366 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000368 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 /* It assumes that any protocol which allows
370 * the user to assign a number at socket
371 * creation time automatically
372 * shares.
373 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000374 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 /* Add to protocol hash chains. */
376 sk->sk_prot->hash(sk);
377 }
378
379 if (sk->sk_prot->init) {
380 err = sk->sk_prot->init(sk);
381 if (err)
382 sk_common_release(sk);
383 }
384out:
385 return err;
386out_rcu_unlock:
387 rcu_read_unlock();
388 goto out;
389}
390
391
392/*
393 * The peer socket should always be NULL (or else). When we call this
394 * function we are destroying the object and from then on nobody
395 * should refer to it.
396 */
397int inet_release(struct socket *sock)
398{
399 struct sock *sk = sock->sk;
400
401 if (sk) {
402 long timeout;
403
David S. Millerc58dc012010-04-27 15:05:31 -0700404 sock_rps_reset_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 /* Applications forget to leave groups before exiting */
407 ip_mc_drop_socket(sk);
408
409 /* If linger is set, we don't return until the close
410 * is complete. Otherwise we return immediately. The
411 * actually closing is done the same either way.
412 *
413 * If the close is due to the process exiting, we never
414 * linger..
415 */
416 timeout = 0;
417 if (sock_flag(sk, SOCK_LINGER) &&
418 !(current->flags & PF_EXITING))
419 timeout = sk->sk_lingertime;
420 sock->sk = NULL;
421 sk->sk_prot->close(sk, timeout);
422 }
423 return 0;
424}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700425EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
427/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700428int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700429EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
431int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
432{
433 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
434 struct sock *sk = sock->sk;
435 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000436 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 unsigned short snum;
438 int chk_addr_ret;
439 int err;
440
441 /* If the socket has its own bind function then use it. (RAW) */
442 if (sk->sk_prot->bind) {
443 err = sk->sk_prot->bind(sk, uaddr, addr_len);
444 goto out;
445 }
446 err = -EINVAL;
447 if (addr_len < sizeof(struct sockaddr_in))
448 goto out;
449
Marcus Meissnerc349a522011-07-04 01:30:29 +0000450 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400451 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
452 * only if s_addr is INADDR_ANY.
453 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000454 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400455 if (addr->sin_family != AF_UNSPEC ||
456 addr->sin_addr.s_addr != htonl(INADDR_ANY))
457 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000458 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700459
Eric W. Biederman35946982012-11-16 03:03:12 +0000460 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461
462 /* Not specified by any standard per-se, however it breaks too
463 * many applications when removed. It is unfortunate since
464 * allowing applications to make a non-local bind solves
465 * several problems with systems using dynamic addressing.
466 * (ie. your servers still start up even if your ISDN link
467 * is temporarily down)
468 */
469 err = -EADDRNOTAVAIL;
470 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700471 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700472 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 chk_addr_ret != RTN_LOCAL &&
474 chk_addr_ret != RTN_MULTICAST &&
475 chk_addr_ret != RTN_BROADCAST)
476 goto out;
477
478 snum = ntohs(addr->sin_port);
479 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000480 if (snum && snum < PROT_SOCK &&
481 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 goto out;
483
484 /* We keep a pair of addresses. rcv_saddr is the one
485 * used by hash lookups, and saddr is used for transmit.
486 *
487 * In the BSD API these are the same except where it
488 * would be illegal to use them (multicast/broadcast) in
489 * which case the sending device address is used.
490 */
491 lock_sock(sk);
492
493 /* Check these errors (active socket, double bind). */
494 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000495 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 goto out_release_sock;
497
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000498 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000500 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
502 /* Make sure we are allowed to bind here. */
503 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000504 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 err = -EADDRINUSE;
506 goto out_release_sock;
507 }
508
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000509 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
511 if (snum)
512 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000513 inet->inet_sport = htons(inet->inet_num);
514 inet->inet_daddr = 0;
515 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 sk_dst_reset(sk);
517 err = 0;
518out_release_sock:
519 release_sock(sk);
520out:
521 return err;
522}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700523EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000525int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 int addr_len, int flags)
527{
528 struct sock *sk = sock->sk;
529
Changli Gao6503d962010-03-31 22:58:26 +0000530 if (addr_len < sizeof(uaddr->sa_family))
531 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 if (uaddr->sa_family == AF_UNSPEC)
533 return sk->sk_prot->disconnect(sk, flags);
534
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000535 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000537 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700539EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Yuchung Cheng783237e2012-07-19 06:43:07 +0000541static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542{
543 DEFINE_WAIT(wait);
544
Eric Dumazetaa395142010-04-20 13:03:51 +0000545 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000546 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548 /* Basic assumption: if someone sets sk->sk_err, he _must_
549 * change state of the socket from TCP_SYN_*.
550 * Connect() does not allow to get error notifications
551 * without closing the socket.
552 */
553 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
554 release_sock(sk);
555 timeo = schedule_timeout(timeo);
556 lock_sock(sk);
557 if (signal_pending(current) || !timeo)
558 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000559 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000561 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000562 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 return timeo;
564}
565
566/*
567 * Connect to a remote host. There is regrettably still a little
568 * TCP 'magic' in here.
569 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000570int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
571 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572{
573 struct sock *sk = sock->sk;
574 int err;
575 long timeo;
576
Changli Gao6503d962010-03-31 22:58:26 +0000577 if (addr_len < sizeof(uaddr->sa_family))
578 return -EINVAL;
579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 if (uaddr->sa_family == AF_UNSPEC) {
581 err = sk->sk_prot->disconnect(sk, flags);
582 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
583 goto out;
584 }
585
586 switch (sock->state) {
587 default:
588 err = -EINVAL;
589 goto out;
590 case SS_CONNECTED:
591 err = -EISCONN;
592 goto out;
593 case SS_CONNECTING:
594 err = -EALREADY;
595 /* Fall out of switch with err, set for this state */
596 break;
597 case SS_UNCONNECTED:
598 err = -EISCONN;
599 if (sk->sk_state != TCP_CLOSE)
600 goto out;
601
602 err = sk->sk_prot->connect(sk, uaddr, addr_len);
603 if (err < 0)
604 goto out;
605
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900606 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
608 /* Just entered SS_CONNECTING state; the only
609 * difference is that return value in non-blocking
610 * case is EINPROGRESS, rather than EALREADY.
611 */
612 err = -EINPROGRESS;
613 break;
614 }
615
616 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
617
618 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000619 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
620 tcp_sk(sk)->fastopen_req &&
621 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
622
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000624 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 goto out;
626
627 err = sock_intr_errno(timeo);
628 if (signal_pending(current))
629 goto out;
630 }
631
632 /* Connection was closed by RST, timeout, ICMP error
633 * or another process disconnected us.
634 */
635 if (sk->sk_state == TCP_CLOSE)
636 goto sock_error;
637
638 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
639 * and error was received after socket entered established state.
640 * Hence, it is handled normally after connect() return successfully.
641 */
642
643 sock->state = SS_CONNECTED;
644 err = 0;
645out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 return err;
647
648sock_error:
649 err = sock_error(sk) ? : -ECONNABORTED;
650 sock->state = SS_UNCONNECTED;
651 if (sk->sk_prot->disconnect(sk, flags))
652 sock->state = SS_DISCONNECTING;
653 goto out;
654}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000655EXPORT_SYMBOL(__inet_stream_connect);
656
657int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
658 int addr_len, int flags)
659{
660 int err;
661
662 lock_sock(sock->sk);
663 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
664 release_sock(sock->sk);
665 return err;
666}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700667EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
669/*
670 * Accept a pending connection. The TCP layer now gives BSD semantics.
671 */
672
673int inet_accept(struct socket *sock, struct socket *newsock, int flags)
674{
675 struct sock *sk1 = sock->sk;
676 int err = -EINVAL;
677 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
678
679 if (!sk2)
680 goto do_err;
681
682 lock_sock(sk2);
683
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000684 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700685 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000686 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
687 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
689 sock_graft(sk2, newsock);
690
691 newsock->state = SS_CONNECTED;
692 err = 0;
693 release_sock(sk2);
694do_err:
695 return err;
696}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700697EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
699
700/*
701 * This does both peername and sockname.
702 */
703int inet_getname(struct socket *sock, struct sockaddr *uaddr,
704 int *uaddr_len, int peer)
705{
706 struct sock *sk = sock->sk;
707 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700708 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
710 sin->sin_family = AF_INET;
711 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000712 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
714 peer == 1))
715 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000716 sin->sin_port = inet->inet_dport;
717 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000719 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000721 addr = inet->inet_saddr;
722 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 sin->sin_addr.s_addr = addr;
724 }
725 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
726 *uaddr_len = sizeof(*sin);
727 return 0;
728}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700729EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
731int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
732 size_t size)
733{
734 struct sock *sk = sock->sk;
735
David S. Millerc58dc012010-04-27 15:05:31 -0700736 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700737
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000739 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
740 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 return -EAGAIN;
742
743 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
744}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700745EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
Changli Gao7ba42912010-07-10 20:41:55 +0000747ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
748 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749{
750 struct sock *sk = sock->sk;
751
David S. Millerc58dc012010-04-27 15:05:31 -0700752 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700753
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000755 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
756 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 return -EAGAIN;
758
759 if (sk->sk_prot->sendpage)
760 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
761 return sock_no_sendpage(sock, page, offset, size, flags);
762}
Changli Gao7ba42912010-07-10 20:41:55 +0000763EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
Tom Herbertfec5e652010-04-16 16:01:27 -0700765int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
766 size_t size, int flags)
767{
768 struct sock *sk = sock->sk;
769 int addr_len = 0;
770 int err;
771
David S. Millerc58dc012010-04-27 15:05:31 -0700772 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700773
774 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
775 flags & ~MSG_DONTWAIT, &addr_len);
776 if (err >= 0)
777 msg->msg_namelen = addr_len;
778 return err;
779}
780EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782int inet_shutdown(struct socket *sock, int how)
783{
784 struct sock *sk = sock->sk;
785 int err = 0;
786
787 /* This should really check to make sure
788 * the socket is a TCP socket. (WHY AC...)
789 */
790 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
791 1->2 bit 2 snds.
792 2->3 */
793 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
794 return -EINVAL;
795
796 lock_sock(sk);
797 if (sock->state == SS_CONNECTING) {
798 if ((1 << sk->sk_state) &
799 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
800 sock->state = SS_DISCONNECTING;
801 else
802 sock->state = SS_CONNECTED;
803 }
804
805 switch (sk->sk_state) {
806 case TCP_CLOSE:
807 err = -ENOTCONN;
808 /* Hack to wake up other listeners, who can poll for
809 POLLHUP, even on eg. unconnected UDP sockets -- RR */
810 default:
811 sk->sk_shutdown |= how;
812 if (sk->sk_prot->shutdown)
813 sk->sk_prot->shutdown(sk, how);
814 break;
815
816 /* Remaining two branches are temporary solution for missing
817 * close() in multithreaded environment. It is _not_ a good idea,
818 * but we have no choice until close() is repaired at VFS level.
819 */
820 case TCP_LISTEN:
821 if (!(how & RCV_SHUTDOWN))
822 break;
823 /* Fall through */
824 case TCP_SYN_SENT:
825 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
826 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
827 break;
828 }
829
830 /* Wake up anyone sleeping in poll. */
831 sk->sk_state_change(sk);
832 release_sock(sk);
833 return err;
834}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700835EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
837/*
838 * ioctl() calls you can issue on an INET socket. Most of these are
839 * device configuration and stuff and very rarely used. Some ioctls
840 * pass on to the socket itself.
841 *
842 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
843 * loads the devconfigure module does its configuring and unloads it.
844 * There's a good 20K of config code hanging around the kernel.
845 */
846
847int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
848{
849 struct sock *sk = sock->sk;
850 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900851 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
853 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700854 case SIOCGSTAMP:
855 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
856 break;
857 case SIOCGSTAMPNS:
858 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
859 break;
860 case SIOCADDRT:
861 case SIOCDELRT:
862 case SIOCRTMSG:
863 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
864 break;
865 case SIOCDARP:
866 case SIOCGARP:
867 case SIOCSARP:
868 err = arp_ioctl(net, cmd, (void __user *)arg);
869 break;
870 case SIOCGIFADDR:
871 case SIOCSIFADDR:
872 case SIOCGIFBRDADDR:
873 case SIOCSIFBRDADDR:
874 case SIOCGIFNETMASK:
875 case SIOCSIFNETMASK:
876 case SIOCGIFDSTADDR:
877 case SIOCSIFDSTADDR:
878 case SIOCSIFPFLAGS:
879 case SIOCGIFPFLAGS:
880 case SIOCSIFFLAGS:
881 err = devinet_ioctl(net, cmd, (void __user *)arg);
882 break;
883 default:
884 if (sk->sk_prot->ioctl)
885 err = sk->sk_prot->ioctl(sk, cmd, arg);
886 else
887 err = -ENOIOCTLCMD;
888 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 }
890 return err;
891}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700892EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000894#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000895static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000896{
897 struct sock *sk = sock->sk;
898 int err = -ENOIOCTLCMD;
899
900 if (sk->sk_prot->compat_ioctl)
901 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
902
903 return err;
904}
905#endif
906
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800907const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800908 .family = PF_INET,
909 .owner = THIS_MODULE,
910 .release = inet_release,
911 .bind = inet_bind,
912 .connect = inet_stream_connect,
913 .socketpair = sock_no_socketpair,
914 .accept = inet_accept,
915 .getname = inet_getname,
916 .poll = tcp_poll,
917 .ioctl = inet_ioctl,
918 .listen = inet_listen,
919 .shutdown = inet_shutdown,
920 .setsockopt = sock_common_setsockopt,
921 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000922 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700923 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800924 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000925 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800926 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800927#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800928 .compat_setsockopt = compat_sock_common_setsockopt,
929 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000930 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800931#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700933EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800935const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800936 .family = PF_INET,
937 .owner = THIS_MODULE,
938 .release = inet_release,
939 .bind = inet_bind,
940 .connect = inet_dgram_connect,
941 .socketpair = sock_no_socketpair,
942 .accept = sock_no_accept,
943 .getname = inet_getname,
944 .poll = udp_poll,
945 .ioctl = inet_ioctl,
946 .listen = sock_no_listen,
947 .shutdown = inet_shutdown,
948 .setsockopt = sock_common_setsockopt,
949 .getsockopt = sock_common_getsockopt,
950 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700951 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800952 .mmap = sock_no_mmap,
953 .sendpage = inet_sendpage,
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_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
962/*
963 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
964 * udp_poll
965 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800966static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800967 .family = PF_INET,
968 .owner = THIS_MODULE,
969 .release = inet_release,
970 .bind = inet_bind,
971 .connect = inet_dgram_connect,
972 .socketpair = sock_no_socketpair,
973 .accept = sock_no_accept,
974 .getname = inet_getname,
975 .poll = datagram_poll,
976 .ioctl = inet_ioctl,
977 .listen = sock_no_listen,
978 .shutdown = inet_shutdown,
979 .setsockopt = sock_common_setsockopt,
980 .getsockopt = sock_common_getsockopt,
981 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700982 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800983 .mmap = sock_no_mmap,
984 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800985#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800986 .compat_setsockopt = compat_sock_common_setsockopt,
987 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000988 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800989#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990};
991
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +0000992static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 .family = PF_INET,
994 .create = inet_create,
995 .owner = THIS_MODULE,
996};
997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998/* Upon startup we insert all the elements in inetsw_array[] into
999 * the linked list inetsw.
1000 */
1001static struct inet_protosw inetsw_array[] =
1002{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001003 {
1004 .type = SOCK_STREAM,
1005 .protocol = IPPROTO_TCP,
1006 .prot = &tcp_prot,
1007 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001008 .no_check = 0,
1009 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001010 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001011 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001013 {
1014 .type = SOCK_DGRAM,
1015 .protocol = IPPROTO_UDP,
1016 .prot = &udp_prot,
1017 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001018 .no_check = UDP_CSUM_DEFAULT,
1019 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001021
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001022 {
1023 .type = SOCK_DGRAM,
1024 .protocol = IPPROTO_ICMP,
1025 .prot = &ping_prot,
1026 .ops = &inet_dgram_ops,
1027 .no_check = UDP_CSUM_DEFAULT,
1028 .flags = INET_PROTOSW_REUSE,
1029 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001032 .type = SOCK_RAW,
1033 .protocol = IPPROTO_IP, /* wild card */
1034 .prot = &raw_prot,
1035 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001036 .no_check = UDP_CSUM_DEFAULT,
1037 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 }
1039};
1040
Denis Chengc40f6ff2007-09-16 16:39:25 -07001041#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042
1043void inet_register_protosw(struct inet_protosw *p)
1044{
1045 struct list_head *lh;
1046 struct inet_protosw *answer;
1047 int protocol = p->protocol;
1048 struct list_head *last_perm;
1049
1050 spin_lock_bh(&inetsw_lock);
1051
1052 if (p->type >= SOCK_MAX)
1053 goto out_illegal;
1054
1055 /* If we are trying to override a permanent protocol, bail. */
1056 answer = NULL;
1057 last_perm = &inetsw[p->type];
1058 list_for_each(lh, &inetsw[p->type]) {
1059 answer = list_entry(lh, struct inet_protosw, list);
1060
1061 /* Check only the non-wild match. */
1062 if (INET_PROTOSW_PERMANENT & answer->flags) {
1063 if (protocol == answer->protocol)
1064 break;
1065 last_perm = lh;
1066 }
1067
1068 answer = NULL;
1069 }
1070 if (answer)
1071 goto out_permanent;
1072
1073 /* Add the new entry after the last permanent entry if any, so that
1074 * the new entry does not override a permanent entry when matched with
1075 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001076 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 * system automatically returns to the old behavior.
1078 */
1079 list_add_rcu(&p->list, last_perm);
1080out:
1081 spin_unlock_bh(&inetsw_lock);
1082
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 return;
1084
1085out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001086 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 goto out;
1088
1089out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001090 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 p->type);
1092 goto out;
1093}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001094EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
1096void inet_unregister_protosw(struct inet_protosw *p)
1097{
1098 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001099 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 p->protocol);
1101 } else {
1102 spin_lock_bh(&inetsw_lock);
1103 list_del_rcu(&p->list);
1104 spin_unlock_bh(&inetsw_lock);
1105
1106 synchronize_net();
1107 }
1108}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001109EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001111/*
1112 * Shall we try to damage output packets if routing dev changes?
1113 */
1114
Brian Haleyab32ea52006-09-22 14:15:41 -07001115int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001116
1117static int inet_sk_reselect_saddr(struct sock *sk)
1118{
1119 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001120 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001121 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001122 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001123 struct rtable *rt;
1124 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001125 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001126
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001127 inet_opt = rcu_dereference_protected(inet->inet_opt,
1128 sock_owned_by_user(sk));
1129 if (inet_opt && inet_opt->opt.srr)
1130 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001131
1132 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001133 fl4 = &inet->cork.fl.u.ip4;
1134 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001135 sk->sk_bound_dev_if, sk->sk_protocol,
1136 inet->inet_sport, inet->inet_dport, sk, false);
1137 if (IS_ERR(rt))
1138 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001139
Changli Gaod8d1f302010-06-10 23:31:35 -07001140 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001141
David S. Miller6e869132011-05-06 16:18:04 -07001142 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001143
1144 if (new_saddr == old_saddr)
1145 return 0;
1146
1147 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001148 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1149 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001150 }
1151
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001152 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001153
1154 /*
1155 * XXX The only one ugly spot where we need to
1156 * XXX really change the sockets identity after
1157 * XXX it has entered the hashes. -DaveM
1158 *
1159 * Besides that, it does not check for connection
1160 * uniqueness. Wait for troubles.
1161 */
1162 __sk_prot_rehash(sk);
1163 return 0;
1164}
1165
1166int inet_sk_rebuild_header(struct sock *sk)
1167{
1168 struct inet_sock *inet = inet_sk(sk);
1169 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001170 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001171 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001172 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001173 int err;
1174
1175 /* Route is OK, nothing to do. */
1176 if (rt)
1177 return 0;
1178
1179 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001180 rcu_read_lock();
1181 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001182 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001183 if (inet_opt && inet_opt->opt.srr)
1184 daddr = inet_opt->opt.faddr;
1185 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001186 fl4 = &inet->cork.fl.u.ip4;
1187 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001188 inet->inet_dport, inet->inet_sport,
1189 sk->sk_protocol, RT_CONN_FLAGS(sk),
1190 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001191 if (!IS_ERR(rt)) {
1192 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001193 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001194 } else {
1195 err = PTR_ERR(rt);
1196
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001197 /* Routing failed... */
1198 sk->sk_route_caps = 0;
1199 /*
1200 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1201 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1202 */
1203 if (!sysctl_ip_dynaddr ||
1204 sk->sk_state != TCP_SYN_SENT ||
1205 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1206 (err = inet_sk_reselect_saddr(sk)) != 0)
1207 sk->sk_err_soft = -err;
1208 }
1209
1210 return err;
1211}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001212EXPORT_SYMBOL(inet_sk_rebuild_header);
1213
Herbert Xua430a432006-07-08 13:34:56 -07001214static int inet_gso_send_check(struct sk_buff *skb)
1215{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001216 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001217 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -07001218 int proto;
1219 int ihl;
1220 int err = -EINVAL;
1221
1222 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1223 goto out;
1224
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001225 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001226 ihl = iph->ihl * 4;
1227 if (ihl < sizeof(*iph))
1228 goto out;
1229
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001230 proto = iph->protocol;
1231
1232 /* Warning: after this point, iph might be no longer valid */
Herbert Xua430a432006-07-08 13:34:56 -07001233 if (unlikely(!pskb_may_pull(skb, ihl)))
1234 goto out;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001235 __skb_pull(skb, ihl);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001236
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001237 skb_reset_transport_header(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001238 err = -EPROTONOSUPPORT;
1239
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001240 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001241 if (likely(ops && ops->callbacks.gso_send_check))
1242 err = ops->callbacks.gso_send_check(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001243
1244out:
1245 return err;
1246}
1247
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001248static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
Eric Dumazet3347c962013-10-19 11:42:56 -07001249 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001250{
1251 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001252 const struct net_offload *ops;
Eric Dumazet3347c962013-10-19 11:42:56 -07001253 unsigned int offset = 0;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001254 bool udpfrag, encap;
David S. Millerf9242b62012-06-19 18:56:21 -07001255 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001256 int proto;
Eric Dumazet3347c962013-10-19 11:42:56 -07001257 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001258 int ihl;
1259 int id;
1260
Herbert Xubbcf4672006-07-03 19:38:35 -07001261 if (unlikely(skb_shinfo(skb)->gso_type &
1262 ~(SKB_GSO_TCPV4 |
1263 SKB_GSO_UDP |
1264 SKB_GSO_DODGY |
1265 SKB_GSO_TCP_ECN |
Pravin B Shelar68c33162013-02-14 14:02:41 +00001266 SKB_GSO_GRE |
Eric Dumazetcb32f512013-10-19 11:42:57 -07001267 SKB_GSO_IPIP |
Eric Dumazet61c1db72013-10-20 20:47:30 -07001268 SKB_GSO_SIT |
Pravin B Shelar9b3eb5e2013-05-02 16:14:19 +00001269 SKB_GSO_TCPV6 |
Pravin B Shelar73136262013-03-07 13:21:51 +00001270 SKB_GSO_UDP_TUNNEL |
Simon Horman0d89d202013-05-23 21:02:52 +00001271 SKB_GSO_MPLS |
Herbert Xubbcf4672006-07-03 19:38:35 -07001272 0)))
1273 goto out;
1274
Eric Dumazet3347c962013-10-19 11:42:56 -07001275 skb_reset_network_header(skb);
1276 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001277 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001278 goto out;
1279
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001280 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001281 ihl = iph->ihl * 4;
1282 if (ihl < sizeof(*iph))
1283 goto out;
1284
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001285 id = ntohs(iph->id);
1286 proto = iph->protocol;
1287
1288 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001289 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001290 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001291 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001292
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001293 encap = SKB_GSO_CB(skb)->encap_level > 0;
1294 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001295 features = skb->dev->hw_enc_features & netif_skb_features(skb);
1296 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001297
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001298 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001299
Herbert Xuf4c50d92006-06-22 03:02:40 -07001300 segs = ERR_PTR(-EPROTONOSUPPORT);
1301
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001302 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001303 if (likely(ops && ops->callbacks.gso_segment))
1304 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001305
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001306 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001307 goto out;
1308
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001309 udpfrag = !!skb->encapsulation && proto == IPPROTO_UDP;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001310 skb = segs;
1311 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001312 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001313 if (udpfrag) {
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001314 iph->id = htons(id);
1315 iph->frag_off = htons(offset >> 3);
1316 if (skb->next != NULL)
1317 iph->frag_off |= htons(IP_MF);
Eric Dumazet3347c962013-10-19 11:42:56 -07001318 offset += skb->len - nhoff - ihl;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001319 } else {
Pravin B Shelar25c77042013-03-24 17:36:29 +00001320 iph->id = htons(id++);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001321 }
Eric Dumazet3347c962013-10-19 11:42:56 -07001322 iph->tot_len = htons(skb->len - nhoff);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001323 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001324 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001325 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001326 skb->network_header = (u8 *)iph - skb->head;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001327 } while ((skb = skb->next));
1328
1329out:
1330 return segs;
1331}
1332
Herbert Xu73cc19f2008-12-15 23:41:09 -08001333static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1334 struct sk_buff *skb)
1335{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001336 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001337 struct sk_buff **pp = NULL;
1338 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001339 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001340 unsigned int hlen;
1341 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001342 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001343 int flush = 1;
1344 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001345
Herbert Xua5b1cf22009-05-26 18:50:28 +00001346 off = skb_gro_offset(skb);
1347 hlen = off + sizeof(*iph);
1348 iph = skb_gro_header_fast(skb, off);
1349 if (skb_gro_header_hard(skb, hlen)) {
1350 iph = skb_gro_header_slow(skb, hlen, off);
1351 if (unlikely(!iph))
1352 goto out;
1353 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001354
David S. Millerf9242b62012-06-19 18:56:21 -07001355 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001356
1357 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001358 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001359 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001360 goto out_unlock;
1361
Herbert Xua5ad24b2009-02-08 18:00:39 +00001362 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001363 goto out_unlock;
1364
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001365 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001366 goto out_unlock;
1367
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001368 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001369 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001370 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001371
1372 for (p = *head; p; p = p->next) {
1373 struct iphdr *iph2;
1374
1375 if (!NAPI_GRO_CB(p)->same_flow)
1376 continue;
1377
1378 iph2 = ip_hdr(p);
1379
Herbert Xua5ad24b2009-02-08 18:00:39 +00001380 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001381 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1382 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001383 NAPI_GRO_CB(p)->same_flow = 0;
1384 continue;
1385 }
1386
1387 /* All fields must match except length and checksum. */
1388 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001389 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001390 (iph->tos ^ iph2->tos) |
Eric Dumazet5b9b6262013-06-12 14:23:15 -07001391 (__force int)((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)) |
Herbert Xua5ad24b2009-02-08 18:00:39 +00001392 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001393
1394 NAPI_GRO_CB(p)->flush |= flush;
1395 }
1396
1397 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001398 skb_gro_pull(skb, sizeof(*iph));
1399 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001400
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001401 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001402
1403out_unlock:
1404 rcu_read_unlock();
1405
1406out:
1407 NAPI_GRO_CB(skb)->flush |= flush;
1408
1409 return pp;
1410}
1411
1412static int inet_gro_complete(struct sk_buff *skb)
1413{
Herbert Xu73cc19f2008-12-15 23:41:09 -08001414 __be16 newlen = htons(skb->len - skb_network_offset(skb));
David S. Millerf9242b62012-06-19 18:56:21 -07001415 struct iphdr *iph = ip_hdr(skb);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001416 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001417 int proto = iph->protocol;
1418 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001419
1420 csum_replace2(&iph->check, iph->tot_len, newlen);
1421 iph->tot_len = newlen;
1422
1423 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001424 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001425 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001426 goto out_unlock;
1427
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001428 err = ops->callbacks.gro_complete(skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001429
1430out_unlock:
1431 rcu_read_unlock();
1432
1433 return err;
1434}
1435
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001436int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001437 unsigned short type, unsigned char protocol,
1438 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001439{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001440 struct socket *sock;
1441 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001442
1443 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001444 *sk = sock->sk;
1445 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001446 /*
1447 * Unhash it so that IP input processing does not even see it,
1448 * we do not wish this socket to see incoming packets.
1449 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001450 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001451
1452 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001453 }
1454 return rc;
1455}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001456EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1457
Tejun Heo7d720c32010-02-16 15:20:26 +00001458unsigned long snmp_fold_field(void __percpu *mib[], int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001459{
1460 unsigned long res = 0;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001461 int i, j;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001462
1463 for_each_possible_cpu(i) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001464 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1465 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001466 }
1467 return res;
1468}
1469EXPORT_SYMBOL_GPL(snmp_fold_field);
1470
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001471#if BITS_PER_LONG==32
1472
1473u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1474{
1475 u64 res = 0;
1476 int cpu;
1477
1478 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001479 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001480 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001481 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001482 unsigned int start;
1483
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001484 bhptr = per_cpu_ptr(mib[0], cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001485 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1486 do {
1487 start = u64_stats_fetch_begin_bh(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001488 v = *(((u64 *) bhptr) + offt);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001489 } while (u64_stats_fetch_retry_bh(syncp, start));
1490
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001491 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001492 }
1493 return res;
1494}
1495EXPORT_SYMBOL_GPL(snmp_fold_field64);
1496#endif
1497
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001498int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001499{
1500 BUG_ON(ptr == NULL);
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001501 ptr[0] = __alloc_percpu(mibsize, align);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001502 if (!ptr[0])
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001503 return -ENOMEM;
1504#if SNMP_ARRAY_SZ == 2
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001505 ptr[1] = __alloc_percpu(mibsize, align);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001506 if (!ptr[1]) {
1507 free_percpu(ptr[0]);
1508 ptr[0] = NULL;
1509 return -ENOMEM;
1510 }
1511#endif
Herbert Xu5e0f0432007-04-24 21:53:35 -07001512 return 0;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001513}
1514EXPORT_SYMBOL_GPL(snmp_mib_init);
1515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001517static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001519 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520};
1521#endif
1522
Alexey Dobriyan32613092009-09-14 12:21:47 +00001523static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001524 .early_demux = tcp_v4_early_demux,
1525 .handler = tcp_v4_rcv,
1526 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001527 .no_policy = 1,
1528 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529};
1530
Alexey Dobriyan32613092009-09-14 12:21:47 +00001531static const struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001532 .early_demux = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 .handler = udp_rcv,
1534 .err_handler = udp_err,
1535 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001536 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537};
1538
Alexey Dobriyan32613092009-09-14 12:21:47 +00001539static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001541 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001542 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001543 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544};
1545
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001546static __net_init int ipv4_mib_init_net(struct net *net)
1547{
Tejun Heo7d720c32010-02-16 15:20:26 +00001548 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001549 sizeof(struct tcp_mib),
1550 __alignof__(struct tcp_mib)) < 0)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001551 goto err_tcp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001552 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001553 sizeof(struct ipstats_mib),
1554 __alignof__(struct ipstats_mib)) < 0)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001555 goto err_ip_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001556 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001557 sizeof(struct linux_mib),
1558 __alignof__(struct linux_mib)) < 0)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001559 goto err_net_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001560 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001561 sizeof(struct udp_mib),
1562 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001563 goto err_udp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001564 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001565 sizeof(struct udp_mib),
1566 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001567 goto err_udplite_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001568 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001569 sizeof(struct icmp_mib),
1570 __alignof__(struct icmp_mib)) < 0)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001571 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001572 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1573 GFP_KERNEL);
1574 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001575 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001576
1577 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001578 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001579
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001580err_icmpmsg_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001581 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001582err_icmp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001583 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001584err_udplite_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001585 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001586err_udp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001587 snmp_mib_free((void __percpu **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001588err_net_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001589 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001590err_ip_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001591 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001592err_tcp_mib:
1593 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001594}
1595
1596static __net_exit void ipv4_mib_exit_net(struct net *net)
1597{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001598 kfree(net->mib.icmpmsg_statistics);
Tejun Heo7d720c32010-02-16 15:20:26 +00001599 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1600 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1601 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1602 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1603 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1604 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001605}
1606
1607static __net_initdata struct pernet_operations ipv4_mib_ops = {
1608 .init = ipv4_mib_init_net,
1609 .exit = ipv4_mib_exit_net,
1610};
1611
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612static int __init init_ipv4_mibs(void)
1613{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001614 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615}
1616
1617static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Herbert Xu30e224d2005-07-05 14:40:10 -07001619/*
1620 * IP protocol layer initialiser
1621 */
1622
Vlad Yasevich22061d82012-11-15 08:49:11 +00001623static struct packet_offload ip_packet_offload __read_mostly = {
1624 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001625 .callbacks = {
1626 .gso_send_check = inet_gso_send_check,
1627 .gso_segment = inet_gso_segment,
1628 .gro_receive = inet_gro_receive,
1629 .gro_complete = inet_gro_complete,
1630 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001631};
1632
Eric Dumazetcb32f512013-10-19 11:42:57 -07001633static const struct net_offload ipip_offload = {
1634 .callbacks = {
1635 .gso_send_check = inet_gso_send_check,
1636 .gso_segment = inet_gso_segment,
1637 },
1638};
1639
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001640static int __init ipv4_offload_init(void)
1641{
1642 /*
1643 * Add offloads
1644 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001645 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001646 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001647 if (tcpv4_offload_init() < 0)
1648 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001649
1650 dev_add_offload(&ip_packet_offload);
Eric Dumazetcb32f512013-10-19 11:42:57 -07001651 inet_add_offload(&ipip_offload, IPPROTO_IPIP);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001652 return 0;
1653}
1654
1655fs_initcall(ipv4_offload_init);
1656
1657static struct packet_type ip_packet_type __read_mostly = {
1658 .type = cpu_to_be16(ETH_P_IP),
1659 .func = ip_rcv,
1660};
1661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662static int __init inet_init(void)
1663{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 struct inet_protosw *q;
1665 struct list_head *r;
1666 int rc = -EINVAL;
1667
YOSHIFUJI Hideaki / 吉藤英明95c7e0e2013-01-09 07:19:37 +00001668 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Amerigo Wange3826f12010-05-05 00:27:06 +00001670 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1671 if (!sysctl_local_reserved_ports)
1672 goto out;
1673
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 rc = proto_register(&tcp_prot, 1);
1675 if (rc)
Amerigo Wange3826f12010-05-05 00:27:06 +00001676 goto out_free_reserved_ports;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
1678 rc = proto_register(&udp_prot, 1);
1679 if (rc)
1680 goto out_unregister_tcp_proto;
1681
1682 rc = proto_register(&raw_prot, 1);
1683 if (rc)
1684 goto out_unregister_udp_proto;
1685
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001686 rc = proto_register(&ping_prot, 1);
1687 if (rc)
1688 goto out_unregister_raw_proto;
1689
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001691 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 */
1693
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001694 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695
Al Virobd7b1532008-07-15 16:00:59 -04001696#ifdef CONFIG_SYSCTL
1697 ip_static_sysctl_init();
1698#endif
1699
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 /*
1701 * Add all the base protocols.
1702 */
1703
1704 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001705 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001707 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001709 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710#ifdef CONFIG_IP_MULTICAST
1711 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001712 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713#endif
1714
1715 /* Register the socket-side information for inet_create. */
1716 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1717 INIT_LIST_HEAD(r);
1718
1719 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1720 inet_register_protosw(q);
1721
1722 /*
1723 * Set the ARP module up
1724 */
1725
1726 arp_init();
1727
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001728 /*
1729 * Set the IP module up
1730 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732 ip_init();
1733
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001734 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
1736 /* Setup TCP slab cache for open requests. */
1737 tcp_init();
1738
Hideo Aoki95766ff2007-12-31 00:29:24 -08001739 /* Setup UDP memory threshold */
1740 udp_init();
1741
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001742 /* Add UDP-Lite (RFC 3828) */
1743 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001745 ping_init();
1746
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 /*
1748 * Set the ICMP layer up
1749 */
1750
Denis V. Luneva5710d62008-02-29 11:14:50 -08001751 if (icmp_init() < 0)
1752 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753
1754 /*
1755 * Initialise the multicast router
1756 */
1757#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001758 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001759 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760#endif
1761 /*
1762 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001763 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764
Stephen Hemminger132adf52007-03-08 20:44:43 -08001765 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001766 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001767
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 ipv4_proc_init();
1769
1770 ipfrag_init();
1771
Herbert Xu30e224d2005-07-05 14:40:10 -07001772 dev_add_pack(&ip_packet_type);
1773
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 rc = 0;
1775out:
1776 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001777out_unregister_raw_proto:
1778 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779out_unregister_udp_proto:
1780 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001781out_unregister_tcp_proto:
1782 proto_unregister(&tcp_prot);
Amerigo Wange3826f12010-05-05 00:27:06 +00001783out_free_reserved_ports:
1784 kfree(sysctl_local_reserved_ports);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 goto out;
1786}
1787
Heiko Carstensa536e072006-04-28 15:19:17 -07001788fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
1790/* ------------------------------------------------------------------------ */
1791
1792#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793static int __init ipv4_proc_init(void)
1794{
1795 int rc = 0;
1796
1797 if (raw_proc_init())
1798 goto out_raw;
1799 if (tcp4_proc_init())
1800 goto out_tcp;
1801 if (udp4_proc_init())
1802 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001803 if (ping_proc_init())
1804 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 if (ip_misc_proc_init())
1806 goto out_misc;
1807out:
1808 return rc;
1809out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001810 ping_proc_exit();
1811out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 udp4_proc_exit();
1813out_udp:
1814 tcp4_proc_exit();
1815out_tcp:
1816 raw_proc_exit();
1817out_raw:
1818 rc = -ENOMEM;
1819 goto out;
1820}
1821
1822#else /* CONFIG_PROC_FS */
1823static int __init ipv4_proc_init(void)
1824{
1825 return 0;
1826}
1827#endif /* CONFIG_PROC_FS */
1828
1829MODULE_ALIAS_NETPROTO(PF_INET);
1830