blob: 24b384b7903ea7a59a11e7a4cbf06db996498924 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * PF_INET protocol family socket handler.
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Florian La Roche, <flla@stud.uni-sb.de>
11 * Alan Cox, <A.Cox@swansea.ac.uk>
12 *
13 * Changes (see also sock.c)
14 *
15 * piggy,
16 * Karl Knutson : Socket protocol table
17 * A.N.Kuznetsov : Socket death error in accept().
18 * John Richardson : Fix non blocking error in connect()
19 * so sockets that fail to connect
20 * don't return -EINPROGRESS.
21 * Alan Cox : Asynchronous I/O support
22 * Alan Cox : Keep correct socket pointer on sock
23 * structures
24 * when accept() ed
25 * Alan Cox : Semantics of SO_LINGER aren't state
26 * moved to close when you look carefully.
27 * With this fixed and the accept bug fixed
28 * some RPC stuff seems happier.
29 * Niibe Yutaka : 4.4BSD style write async I/O
30 * Alan Cox,
31 * Tony Gale : Fixed reuse semantics.
32 * Alan Cox : bind() shouldn't abort existing but dead
33 * sockets. Stops FTP netin:.. I hope.
34 * Alan Cox : bind() works correctly for RAW sockets.
35 * Note that FreeBSD at least was broken
36 * in this respect so be careful with
37 * compatibility tests...
38 * Alan Cox : routing cache support
39 * Alan Cox : memzero the socket structure for
40 * compactness.
41 * Matt Day : nonblock connect error handler
42 * Alan Cox : Allow large numbers of pending sockets
43 * (eg for big web sites), but only if
44 * specifically application requested.
45 * Alan Cox : New buffering throughout IP. Used
46 * dumbly.
47 * Alan Cox : New buffering now used smartly.
48 * Alan Cox : BSD rather than common sense
49 * interpretation of listen.
50 * Germano Caronni : Assorted small races.
51 * Alan Cox : sendmsg/recvmsg basic support.
52 * Alan Cox : Only sendmsg/recvmsg now supported.
53 * Alan Cox : Locked down bind (see security list).
54 * Alan Cox : Loosened bind a little.
55 * Mike McLagan : ADD/DEL DLCI Ioctls
56 * Willy Konynenberg : Transparent proxying support.
57 * David S. Miller : New socket lookup architecture.
58 * Some other random speedups.
59 * Cyrus Durgin : Cleaned up file for kmod hacks.
60 * Andi Kleen : Fix inet_stream_connect TCP race.
61 *
62 * This program is free software; you can redistribute it and/or
63 * modify it under the terms of the GNU General Public License
64 * as published by the Free Software Foundation; either version
65 * 2 of the License, or (at your option) any later version.
66 */
67
Joe Perchesafd465032012-03-12 07:03:32 +000068#define pr_fmt(fmt) "IPv4: " fmt
69
Herbert Xuf4c50d92006-06-22 03:02:40 -070070#include <linux/err.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/errno.h>
72#include <linux/types.h>
73#include <linux/socket.h>
74#include <linux/in.h>
75#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/module.h>
77#include <linux/sched.h>
78#include <linux/timer.h>
79#include <linux/string.h>
80#include <linux/sockios.h>
81#include <linux/net.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080082#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#include <linux/fcntl.h>
84#include <linux/mm.h>
85#include <linux/interrupt.h>
86#include <linux/stat.h>
87#include <linux/init.h>
88#include <linux/poll.h>
89#include <linux/netfilter_ipv4.h>
David S. Millerb3da2cf2007-03-23 11:40:27 -070090#include <linux/random.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090091#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/inet.h>
96#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020097#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Herbert Xu73cc19f2008-12-15 23:41:09 -080099#include <net/checksum.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#include <net/ip.h>
101#include <net/protocol.h>
102#include <net/arp.h>
103#include <net/route.h>
104#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700105#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/tcp.h>
107#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800108#include <net/udplite.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000109#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110#include <linux/skbuff.h>
111#include <net/sock.h>
112#include <net/raw.h>
113#include <net/icmp.h>
114#include <net/ipip.h>
115#include <net/inet_common.h>
116#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700117#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118#ifdef CONFIG_IP_MROUTE
119#include <linux/mroute.h>
120#endif
121
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123/* The inetsw table contains everything that inet_create needs to
124 * build a new socket.
125 */
126static struct list_head inetsw[SOCK_MAX];
127static DEFINE_SPINLOCK(inetsw_lock);
128
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800129struct ipv4_config ipv4_config;
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800130EXPORT_SYMBOL(ipv4_config);
131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132/* New destruction routine */
133
134void inet_sock_destruct(struct sock *sk)
135{
136 struct inet_sock *inet = inet_sk(sk);
137
138 __skb_queue_purge(&sk->sk_receive_queue);
139 __skb_queue_purge(&sk->sk_error_queue);
140
Hideo Aoki95766ff2007-12-31 00:29:24 -0800141 sk_mem_reclaim(sk);
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700144 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 sk->sk_state, sk);
146 return;
147 }
148 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700149 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 return;
151 }
152
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700153 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
154 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
155 WARN_ON(sk->sk_wmem_queued);
156 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000158 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000159 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700160 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700161 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700163EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
165/*
166 * The routines beyond this point handle the behaviour of an AF_INET
167 * socket object. Mostly it punts to the subprotocols of IP to do
168 * the work.
169 */
170
171/*
172 * Automatically bind an unbound socket.
173 */
174
175static int inet_autobind(struct sock *sk)
176{
177 struct inet_sock *inet;
178 /* We may need to bind the socket. */
179 lock_sock(sk);
180 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000181 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 if (sk->sk_prot->get_port(sk, 0)) {
183 release_sock(sk);
184 return -EAGAIN;
185 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000186 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 }
188 release_sock(sk);
189 return 0;
190}
191
192/*
193 * Move a socket into listening state.
194 */
195int inet_listen(struct socket *sock, int backlog)
196{
197 struct sock *sk = sock->sk;
198 unsigned char old_state;
199 int err;
200
201 lock_sock(sk);
202
203 err = -EINVAL;
204 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
205 goto out;
206
207 old_state = sk->sk_state;
208 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
209 goto out;
210
211 /* Really, if the socket is already in listen state
212 * we can only allow the backlog to be adjusted.
213 */
214 if (old_state != TCP_LISTEN) {
Jerry Chu83368862012-08-31 12:29:12 +0000215 /* Check special setups for testing purpose to enable TFO w/o
216 * requiring TCP_FASTOPEN sockopt.
217 * Note that only TCP sockets (SOCK_STREAM) will reach here.
218 * Also fastopenq may already been allocated because this
219 * socket was in TCP_LISTEN state previously but was
220 * shutdown() (rather than close()).
221 */
222 if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
223 inet_csk(sk)->icsk_accept_queue.fastopenq == NULL) {
224 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
225 err = fastopen_init_queue(sk, backlog);
226 else if ((sysctl_tcp_fastopen &
227 TFO_SERVER_WO_SOCKOPT2) != 0)
228 err = fastopen_init_queue(sk,
229 ((uint)sysctl_tcp_fastopen) >> 16);
230 else
231 err = 0;
232 if (err)
233 goto out;
234 }
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800235 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 if (err)
237 goto out;
238 }
239 sk->sk_max_ack_backlog = backlog;
240 err = 0;
241
242out:
243 release_sock(sk);
244 return err;
245}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700246EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Eric Dumazetbe776282007-03-27 13:53:04 -0700248u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700249EXPORT_SYMBOL(inet_ehash_secret);
250
Eric Dumazetbe776282007-03-27 13:53:04 -0700251/*
252 * inet_ehash_secret must be set exactly once
Eric Dumazetbe776282007-03-27 13:53:04 -0700253 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700254void build_ehash_secret(void)
255{
Eric Dumazetbe776282007-03-27 13:53:04 -0700256 u32 rnd;
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000257
Eric Dumazetbe776282007-03-27 13:53:04 -0700258 do {
259 get_random_bytes(&rnd, sizeof(rnd));
260 } while (rnd == 0);
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000261
262 cmpxchg(&inet_ehash_secret, 0, rnd);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700263}
264EXPORT_SYMBOL(build_ehash_secret);
265
David S. Millerf9242b62012-06-19 18:56:21 -0700266static inline int inet_netns_ok(struct net *net, __u8 protocol)
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700267{
Alexey Dobriyan32613092009-09-14 12:21:47 +0000268 const struct net_protocol *ipprot;
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700269
Eric Dumazet04f258c2008-11-23 15:42:23 -0800270 if (net_eq(net, &init_net))
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700271 return 1;
272
David S. Millerf9242b62012-06-19 18:56:21 -0700273 ipprot = rcu_dereference(inet_protos[protocol]);
274 if (ipprot == NULL) {
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700275 /* raw IP is OK */
276 return 1;
David S. Millerf9242b62012-06-19 18:56:21 -0700277 }
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700278 return ipprot->netns_ok;
279}
280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281/*
282 * Create an inet socket.
283 */
284
Eric Paris3f378b62009-11-05 22:18:14 -0800285static int inet_create(struct net *net, struct socket *sock, int protocol,
286 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287{
288 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 struct inet_protosw *answer;
290 struct inet_sock *inet;
291 struct proto *answer_prot;
292 unsigned char answer_flags;
293 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700294 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800295 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296
Eric Dumazet04f258c2008-11-23 15:42:23 -0800297 if (unlikely(!inet_ehash_secret))
298 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
299 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700300
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 sock->state = SS_UNCONNECTED;
302
303 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700304lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800305 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700307 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700309 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 /* Check the non-wild match. */
311 if (protocol == answer->protocol) {
312 if (protocol != IPPROTO_IP)
313 break;
314 } else {
315 /* Check for the two wild cases. */
316 if (IPPROTO_IP == protocol) {
317 protocol = answer->protocol;
318 break;
319 }
320 if (IPPROTO_IP == answer->protocol)
321 break;
322 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800323 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 }
325
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700326 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700327 if (try_loading_module < 2) {
328 rcu_read_unlock();
329 /*
330 * Be more specific, e.g. net-pf-2-proto-132-type-1
331 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
332 */
333 if (++try_loading_module == 1)
334 request_module("net-pf-%d-proto-%d-type-%d",
335 PF_INET, protocol, sock->type);
336 /*
337 * Fall back to generic, e.g. net-pf-2-proto-132
338 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
339 */
340 else
341 request_module("net-pf-%d-proto-%d",
342 PF_INET, protocol);
343 goto lookup_protocol;
344 } else
345 goto out_rcu_unlock;
346 }
347
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000349 if (sock->type == SOCK_RAW && !kern &&
350 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700353 err = -EAFNOSUPPORT;
354 if (!inet_netns_ok(net, protocol))
355 goto out_rcu_unlock;
356
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 sock->ops = answer->ops;
358 answer_prot = answer->prot;
359 answer_no_check = answer->no_check;
360 answer_flags = answer->flags;
361 rcu_read_unlock();
362
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700363 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
365 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700366 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 if (sk == NULL)
368 goto out;
369
370 err = 0;
371 sk->sk_no_check = answer_no_check;
372 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000373 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
375 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800376 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000378 inet->nodefrag = 0;
379
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000381 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 if (IPPROTO_RAW == protocol)
383 inet->hdrincl = 1;
384 }
385
386 if (ipv4_config.no_pmtu_disc)
387 inet->pmtudisc = IP_PMTUDISC_DONT;
388 else
389 inet->pmtudisc = IP_PMTUDISC_WANT;
390
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000391 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 sock_init_data(sock, sk);
394
395 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 sk->sk_protocol = protocol;
397 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
398
399 inet->uc_ttl = -1;
400 inet->mc_loop = 1;
401 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000402 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 inet->mc_index = 0;
404 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000405 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700407 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000409 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 /* It assumes that any protocol which allows
411 * the user to assign a number at socket
412 * creation time automatically
413 * shares.
414 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000415 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 /* Add to protocol hash chains. */
417 sk->sk_prot->hash(sk);
418 }
419
420 if (sk->sk_prot->init) {
421 err = sk->sk_prot->init(sk);
422 if (err)
423 sk_common_release(sk);
424 }
425out:
426 return err;
427out_rcu_unlock:
428 rcu_read_unlock();
429 goto out;
430}
431
432
433/*
434 * The peer socket should always be NULL (or else). When we call this
435 * function we are destroying the object and from then on nobody
436 * should refer to it.
437 */
438int inet_release(struct socket *sock)
439{
440 struct sock *sk = sock->sk;
441
442 if (sk) {
443 long timeout;
444
David S. Millerc58dc012010-04-27 15:05:31 -0700445 sock_rps_reset_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 /* Applications forget to leave groups before exiting */
448 ip_mc_drop_socket(sk);
449
450 /* If linger is set, we don't return until the close
451 * is complete. Otherwise we return immediately. The
452 * actually closing is done the same either way.
453 *
454 * If the close is due to the process exiting, we never
455 * linger..
456 */
457 timeout = 0;
458 if (sock_flag(sk, SOCK_LINGER) &&
459 !(current->flags & PF_EXITING))
460 timeout = sk->sk_lingertime;
461 sock->sk = NULL;
462 sk->sk_prot->close(sk, timeout);
463 }
464 return 0;
465}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700466EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
468/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700469int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700470EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
473{
474 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
475 struct sock *sk = sock->sk;
476 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000477 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 unsigned short snum;
479 int chk_addr_ret;
480 int err;
481
482 /* If the socket has its own bind function then use it. (RAW) */
483 if (sk->sk_prot->bind) {
484 err = sk->sk_prot->bind(sk, uaddr, addr_len);
485 goto out;
486 }
487 err = -EINVAL;
488 if (addr_len < sizeof(struct sockaddr_in))
489 goto out;
490
Marcus Meissnerc349a522011-07-04 01:30:29 +0000491 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400492 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
493 * only if s_addr is INADDR_ANY.
494 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000495 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400496 if (addr->sin_family != AF_UNSPEC ||
497 addr->sin_addr.s_addr != htonl(INADDR_ANY))
498 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000499 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700500
Eric W. Biederman35946982012-11-16 03:03:12 +0000501 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
503 /* Not specified by any standard per-se, however it breaks too
504 * many applications when removed. It is unfortunate since
505 * allowing applications to make a non-local bind solves
506 * several problems with systems using dynamic addressing.
507 * (ie. your servers still start up even if your ISDN link
508 * is temporarily down)
509 */
510 err = -EADDRNOTAVAIL;
511 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700512 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700513 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 chk_addr_ret != RTN_LOCAL &&
515 chk_addr_ret != RTN_MULTICAST &&
516 chk_addr_ret != RTN_BROADCAST)
517 goto out;
518
519 snum = ntohs(addr->sin_port);
520 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000521 if (snum && snum < PROT_SOCK &&
522 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 goto out;
524
525 /* We keep a pair of addresses. rcv_saddr is the one
526 * used by hash lookups, and saddr is used for transmit.
527 *
528 * In the BSD API these are the same except where it
529 * would be illegal to use them (multicast/broadcast) in
530 * which case the sending device address is used.
531 */
532 lock_sock(sk);
533
534 /* Check these errors (active socket, double bind). */
535 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000536 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 goto out_release_sock;
538
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000539 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000541 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543 /* Make sure we are allowed to bind here. */
544 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000545 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 err = -EADDRINUSE;
547 goto out_release_sock;
548 }
549
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000550 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
552 if (snum)
553 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000554 inet->inet_sport = htons(inet->inet_num);
555 inet->inet_daddr = 0;
556 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 sk_dst_reset(sk);
558 err = 0;
559out_release_sock:
560 release_sock(sk);
561out:
562 return err;
563}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700564EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000566int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 int addr_len, int flags)
568{
569 struct sock *sk = sock->sk;
570
Changli Gao6503d962010-03-31 22:58:26 +0000571 if (addr_len < sizeof(uaddr->sa_family))
572 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 if (uaddr->sa_family == AF_UNSPEC)
574 return sk->sk_prot->disconnect(sk, flags);
575
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000576 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000578 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700580EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
Yuchung Cheng783237e2012-07-19 06:43:07 +0000582static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583{
584 DEFINE_WAIT(wait);
585
Eric Dumazetaa395142010-04-20 13:03:51 +0000586 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000587 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
589 /* Basic assumption: if someone sets sk->sk_err, he _must_
590 * change state of the socket from TCP_SYN_*.
591 * Connect() does not allow to get error notifications
592 * without closing the socket.
593 */
594 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
595 release_sock(sk);
596 timeo = schedule_timeout(timeo);
597 lock_sock(sk);
598 if (signal_pending(current) || !timeo)
599 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000600 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000602 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000603 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 return timeo;
605}
606
607/*
608 * Connect to a remote host. There is regrettably still a little
609 * TCP 'magic' in here.
610 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000611int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
612 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613{
614 struct sock *sk = sock->sk;
615 int err;
616 long timeo;
617
Changli Gao6503d962010-03-31 22:58:26 +0000618 if (addr_len < sizeof(uaddr->sa_family))
619 return -EINVAL;
620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 if (uaddr->sa_family == AF_UNSPEC) {
622 err = sk->sk_prot->disconnect(sk, flags);
623 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
624 goto out;
625 }
626
627 switch (sock->state) {
628 default:
629 err = -EINVAL;
630 goto out;
631 case SS_CONNECTED:
632 err = -EISCONN;
633 goto out;
634 case SS_CONNECTING:
635 err = -EALREADY;
636 /* Fall out of switch with err, set for this state */
637 break;
638 case SS_UNCONNECTED:
639 err = -EISCONN;
640 if (sk->sk_state != TCP_CLOSE)
641 goto out;
642
643 err = sk->sk_prot->connect(sk, uaddr, addr_len);
644 if (err < 0)
645 goto out;
646
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900647 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 /* Just entered SS_CONNECTING state; the only
650 * difference is that return value in non-blocking
651 * case is EINPROGRESS, rather than EALREADY.
652 */
653 err = -EINPROGRESS;
654 break;
655 }
656
657 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
658
659 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000660 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
661 tcp_sk(sk)->fastopen_req &&
662 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
663
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000665 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 goto out;
667
668 err = sock_intr_errno(timeo);
669 if (signal_pending(current))
670 goto out;
671 }
672
673 /* Connection was closed by RST, timeout, ICMP error
674 * or another process disconnected us.
675 */
676 if (sk->sk_state == TCP_CLOSE)
677 goto sock_error;
678
679 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
680 * and error was received after socket entered established state.
681 * Hence, it is handled normally after connect() return successfully.
682 */
683
684 sock->state = SS_CONNECTED;
685 err = 0;
686out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 return err;
688
689sock_error:
690 err = sock_error(sk) ? : -ECONNABORTED;
691 sock->state = SS_UNCONNECTED;
692 if (sk->sk_prot->disconnect(sk, flags))
693 sock->state = SS_DISCONNECTING;
694 goto out;
695}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000696EXPORT_SYMBOL(__inet_stream_connect);
697
698int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
699 int addr_len, int flags)
700{
701 int err;
702
703 lock_sock(sock->sk);
704 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
705 release_sock(sock->sk);
706 return err;
707}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700708EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
710/*
711 * Accept a pending connection. The TCP layer now gives BSD semantics.
712 */
713
714int inet_accept(struct socket *sock, struct socket *newsock, int flags)
715{
716 struct sock *sk1 = sock->sk;
717 int err = -EINVAL;
718 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
719
720 if (!sk2)
721 goto do_err;
722
723 lock_sock(sk2);
724
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000725 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700726 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000727 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
728 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729
730 sock_graft(sk2, newsock);
731
732 newsock->state = SS_CONNECTED;
733 err = 0;
734 release_sock(sk2);
735do_err:
736 return err;
737}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700738EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
740
741/*
742 * This does both peername and sockname.
743 */
744int inet_getname(struct socket *sock, struct sockaddr *uaddr,
745 int *uaddr_len, int peer)
746{
747 struct sock *sk = sock->sk;
748 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700749 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 sin->sin_family = AF_INET;
752 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000753 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
755 peer == 1))
756 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000757 sin->sin_port = inet->inet_dport;
758 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000760 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000762 addr = inet->inet_saddr;
763 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 sin->sin_addr.s_addr = addr;
765 }
766 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
767 *uaddr_len = sizeof(*sin);
768 return 0;
769}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700770EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
772int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
773 size_t size)
774{
775 struct sock *sk = sock->sk;
776
David S. Millerc58dc012010-04-27 15:05:31 -0700777 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000780 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
781 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return -EAGAIN;
783
784 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
785}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700786EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Changli Gao7ba42912010-07-10 20:41:55 +0000788ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
789 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790{
791 struct sock *sk = sock->sk;
792
David S. Millerc58dc012010-04-27 15:05:31 -0700793 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700794
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000796 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
797 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 return -EAGAIN;
799
800 if (sk->sk_prot->sendpage)
801 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
802 return sock_no_sendpage(sock, page, offset, size, flags);
803}
Changli Gao7ba42912010-07-10 20:41:55 +0000804EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
Tom Herbertfec5e652010-04-16 16:01:27 -0700806int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
807 size_t size, int flags)
808{
809 struct sock *sk = sock->sk;
810 int addr_len = 0;
811 int err;
812
David S. Millerc58dc012010-04-27 15:05:31 -0700813 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700814
815 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
816 flags & ~MSG_DONTWAIT, &addr_len);
817 if (err >= 0)
818 msg->msg_namelen = addr_len;
819 return err;
820}
821EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
823int inet_shutdown(struct socket *sock, int how)
824{
825 struct sock *sk = sock->sk;
826 int err = 0;
827
828 /* This should really check to make sure
829 * the socket is a TCP socket. (WHY AC...)
830 */
831 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
832 1->2 bit 2 snds.
833 2->3 */
834 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
835 return -EINVAL;
836
837 lock_sock(sk);
838 if (sock->state == SS_CONNECTING) {
839 if ((1 << sk->sk_state) &
840 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
841 sock->state = SS_DISCONNECTING;
842 else
843 sock->state = SS_CONNECTED;
844 }
845
846 switch (sk->sk_state) {
847 case TCP_CLOSE:
848 err = -ENOTCONN;
849 /* Hack to wake up other listeners, who can poll for
850 POLLHUP, even on eg. unconnected UDP sockets -- RR */
851 default:
852 sk->sk_shutdown |= how;
853 if (sk->sk_prot->shutdown)
854 sk->sk_prot->shutdown(sk, how);
855 break;
856
857 /* Remaining two branches are temporary solution for missing
858 * close() in multithreaded environment. It is _not_ a good idea,
859 * but we have no choice until close() is repaired at VFS level.
860 */
861 case TCP_LISTEN:
862 if (!(how & RCV_SHUTDOWN))
863 break;
864 /* Fall through */
865 case TCP_SYN_SENT:
866 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
867 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
868 break;
869 }
870
871 /* Wake up anyone sleeping in poll. */
872 sk->sk_state_change(sk);
873 release_sock(sk);
874 return err;
875}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700876EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
878/*
879 * ioctl() calls you can issue on an INET socket. Most of these are
880 * device configuration and stuff and very rarely used. Some ioctls
881 * pass on to the socket itself.
882 *
883 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
884 * loads the devconfigure module does its configuring and unloads it.
885 * There's a good 20K of config code hanging around the kernel.
886 */
887
888int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
889{
890 struct sock *sk = sock->sk;
891 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900892 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
894 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700895 case SIOCGSTAMP:
896 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
897 break;
898 case SIOCGSTAMPNS:
899 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
900 break;
901 case SIOCADDRT:
902 case SIOCDELRT:
903 case SIOCRTMSG:
904 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
905 break;
906 case SIOCDARP:
907 case SIOCGARP:
908 case SIOCSARP:
909 err = arp_ioctl(net, cmd, (void __user *)arg);
910 break;
911 case SIOCGIFADDR:
912 case SIOCSIFADDR:
913 case SIOCGIFBRDADDR:
914 case SIOCSIFBRDADDR:
915 case SIOCGIFNETMASK:
916 case SIOCSIFNETMASK:
917 case SIOCGIFDSTADDR:
918 case SIOCSIFDSTADDR:
919 case SIOCSIFPFLAGS:
920 case SIOCGIFPFLAGS:
921 case SIOCSIFFLAGS:
922 err = devinet_ioctl(net, cmd, (void __user *)arg);
923 break;
924 default:
925 if (sk->sk_prot->ioctl)
926 err = sk->sk_prot->ioctl(sk, cmd, arg);
927 else
928 err = -ENOIOCTLCMD;
929 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 }
931 return err;
932}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700933EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000935#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000936static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000937{
938 struct sock *sk = sock->sk;
939 int err = -ENOIOCTLCMD;
940
941 if (sk->sk_prot->compat_ioctl)
942 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
943
944 return err;
945}
946#endif
947
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800948const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800949 .family = PF_INET,
950 .owner = THIS_MODULE,
951 .release = inet_release,
952 .bind = inet_bind,
953 .connect = inet_stream_connect,
954 .socketpair = sock_no_socketpair,
955 .accept = inet_accept,
956 .getname = inet_getname,
957 .poll = tcp_poll,
958 .ioctl = inet_ioctl,
959 .listen = inet_listen,
960 .shutdown = inet_shutdown,
961 .setsockopt = sock_common_setsockopt,
962 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000963 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700964 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800965 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000966 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800967 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800968#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800969 .compat_setsockopt = compat_sock_common_setsockopt,
970 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000971 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800972#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700974EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800976const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800977 .family = PF_INET,
978 .owner = THIS_MODULE,
979 .release = inet_release,
980 .bind = inet_bind,
981 .connect = inet_dgram_connect,
982 .socketpair = sock_no_socketpair,
983 .accept = sock_no_accept,
984 .getname = inet_getname,
985 .poll = udp_poll,
986 .ioctl = inet_ioctl,
987 .listen = sock_no_listen,
988 .shutdown = inet_shutdown,
989 .setsockopt = sock_common_setsockopt,
990 .getsockopt = sock_common_getsockopt,
991 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700992 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800993 .mmap = sock_no_mmap,
994 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800995#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800996 .compat_setsockopt = compat_sock_common_setsockopt,
997 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000998 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800999#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000};
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001001EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
1003/*
1004 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
1005 * udp_poll
1006 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001007static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001008 .family = PF_INET,
1009 .owner = THIS_MODULE,
1010 .release = inet_release,
1011 .bind = inet_bind,
1012 .connect = inet_dgram_connect,
1013 .socketpair = sock_no_socketpair,
1014 .accept = sock_no_accept,
1015 .getname = inet_getname,
1016 .poll = datagram_poll,
1017 .ioctl = inet_ioctl,
1018 .listen = sock_no_listen,
1019 .shutdown = inet_shutdown,
1020 .setsockopt = sock_common_setsockopt,
1021 .getsockopt = sock_common_getsockopt,
1022 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001023 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001024 .mmap = sock_no_mmap,
1025 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001026#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001027 .compat_setsockopt = compat_sock_common_setsockopt,
1028 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001029 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001030#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031};
1032
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001033static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 .family = PF_INET,
1035 .create = inet_create,
1036 .owner = THIS_MODULE,
1037};
1038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039/* Upon startup we insert all the elements in inetsw_array[] into
1040 * the linked list inetsw.
1041 */
1042static struct inet_protosw inetsw_array[] =
1043{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001044 {
1045 .type = SOCK_STREAM,
1046 .protocol = IPPROTO_TCP,
1047 .prot = &tcp_prot,
1048 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001049 .no_check = 0,
1050 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001051 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001052 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001054 {
1055 .type = SOCK_DGRAM,
1056 .protocol = IPPROTO_UDP,
1057 .prot = &udp_prot,
1058 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001059 .no_check = UDP_CSUM_DEFAULT,
1060 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001062
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001063 {
1064 .type = SOCK_DGRAM,
1065 .protocol = IPPROTO_ICMP,
1066 .prot = &ping_prot,
1067 .ops = &inet_dgram_ops,
1068 .no_check = UDP_CSUM_DEFAULT,
1069 .flags = INET_PROTOSW_REUSE,
1070 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
1072 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001073 .type = SOCK_RAW,
1074 .protocol = IPPROTO_IP, /* wild card */
1075 .prot = &raw_prot,
1076 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001077 .no_check = UDP_CSUM_DEFAULT,
1078 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 }
1080};
1081
Denis Chengc40f6ff2007-09-16 16:39:25 -07001082#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
1084void inet_register_protosw(struct inet_protosw *p)
1085{
1086 struct list_head *lh;
1087 struct inet_protosw *answer;
1088 int protocol = p->protocol;
1089 struct list_head *last_perm;
1090
1091 spin_lock_bh(&inetsw_lock);
1092
1093 if (p->type >= SOCK_MAX)
1094 goto out_illegal;
1095
1096 /* If we are trying to override a permanent protocol, bail. */
1097 answer = NULL;
1098 last_perm = &inetsw[p->type];
1099 list_for_each(lh, &inetsw[p->type]) {
1100 answer = list_entry(lh, struct inet_protosw, list);
1101
1102 /* Check only the non-wild match. */
1103 if (INET_PROTOSW_PERMANENT & answer->flags) {
1104 if (protocol == answer->protocol)
1105 break;
1106 last_perm = lh;
1107 }
1108
1109 answer = NULL;
1110 }
1111 if (answer)
1112 goto out_permanent;
1113
1114 /* Add the new entry after the last permanent entry if any, so that
1115 * the new entry does not override a permanent entry when matched with
1116 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001117 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 * system automatically returns to the old behavior.
1119 */
1120 list_add_rcu(&p->list, last_perm);
1121out:
1122 spin_unlock_bh(&inetsw_lock);
1123
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 return;
1125
1126out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001127 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 goto out;
1129
1130out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001131 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 p->type);
1133 goto out;
1134}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001135EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137void inet_unregister_protosw(struct inet_protosw *p)
1138{
1139 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001140 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 p->protocol);
1142 } else {
1143 spin_lock_bh(&inetsw_lock);
1144 list_del_rcu(&p->list);
1145 spin_unlock_bh(&inetsw_lock);
1146
1147 synchronize_net();
1148 }
1149}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001150EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001152/*
1153 * Shall we try to damage output packets if routing dev changes?
1154 */
1155
Brian Haleyab32ea52006-09-22 14:15:41 -07001156int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001157
1158static int inet_sk_reselect_saddr(struct sock *sk)
1159{
1160 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001161 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001162 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001163 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001164 struct rtable *rt;
1165 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001166 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001167
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001168 inet_opt = rcu_dereference_protected(inet->inet_opt,
1169 sock_owned_by_user(sk));
1170 if (inet_opt && inet_opt->opt.srr)
1171 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001172
1173 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001174 fl4 = &inet->cork.fl.u.ip4;
1175 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001176 sk->sk_bound_dev_if, sk->sk_protocol,
1177 inet->inet_sport, inet->inet_dport, sk, false);
1178 if (IS_ERR(rt))
1179 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001180
Changli Gaod8d1f302010-06-10 23:31:35 -07001181 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001182
David S. Miller6e869132011-05-06 16:18:04 -07001183 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001184
1185 if (new_saddr == old_saddr)
1186 return 0;
1187
1188 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001189 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1190 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001191 }
1192
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001193 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001194
1195 /*
1196 * XXX The only one ugly spot where we need to
1197 * XXX really change the sockets identity after
1198 * XXX it has entered the hashes. -DaveM
1199 *
1200 * Besides that, it does not check for connection
1201 * uniqueness. Wait for troubles.
1202 */
1203 __sk_prot_rehash(sk);
1204 return 0;
1205}
1206
1207int inet_sk_rebuild_header(struct sock *sk)
1208{
1209 struct inet_sock *inet = inet_sk(sk);
1210 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001211 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001212 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001213 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001214 int err;
1215
1216 /* Route is OK, nothing to do. */
1217 if (rt)
1218 return 0;
1219
1220 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001221 rcu_read_lock();
1222 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001223 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001224 if (inet_opt && inet_opt->opt.srr)
1225 daddr = inet_opt->opt.faddr;
1226 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001227 fl4 = &inet->cork.fl.u.ip4;
1228 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001229 inet->inet_dport, inet->inet_sport,
1230 sk->sk_protocol, RT_CONN_FLAGS(sk),
1231 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001232 if (!IS_ERR(rt)) {
1233 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001234 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001235 } else {
1236 err = PTR_ERR(rt);
1237
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001238 /* Routing failed... */
1239 sk->sk_route_caps = 0;
1240 /*
1241 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1242 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1243 */
1244 if (!sysctl_ip_dynaddr ||
1245 sk->sk_state != TCP_SYN_SENT ||
1246 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1247 (err = inet_sk_reselect_saddr(sk)) != 0)
1248 sk->sk_err_soft = -err;
1249 }
1250
1251 return err;
1252}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001253EXPORT_SYMBOL(inet_sk_rebuild_header);
1254
Herbert Xua430a432006-07-08 13:34:56 -07001255static int inet_gso_send_check(struct sk_buff *skb)
1256{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001257 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001258 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -07001259 int proto;
1260 int ihl;
1261 int err = -EINVAL;
1262
1263 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1264 goto out;
1265
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001266 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001267 ihl = iph->ihl * 4;
1268 if (ihl < sizeof(*iph))
1269 goto out;
1270
1271 if (unlikely(!pskb_may_pull(skb, ihl)))
1272 goto out;
1273
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001274 __skb_pull(skb, ihl);
1275 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001276 iph = ip_hdr(skb);
David S. Millerf9242b62012-06-19 18:56:21 -07001277 proto = iph->protocol;
Herbert Xua430a432006-07-08 13:34:56 -07001278 err = -EPROTONOSUPPORT;
1279
1280 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001281 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001282 if (likely(ops && ops->callbacks.gso_send_check))
1283 err = ops->callbacks.gso_send_check(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001284 rcu_read_unlock();
1285
1286out:
1287 return err;
1288}
1289
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001290static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1291 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001292{
1293 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001294 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001295 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001296 int proto;
1297 int ihl;
1298 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001299 unsigned int offset = 0;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001300
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001301 if (!(features & NETIF_F_V4_CSUM))
1302 features &= ~NETIF_F_SG;
1303
Herbert Xubbcf4672006-07-03 19:38:35 -07001304 if (unlikely(skb_shinfo(skb)->gso_type &
1305 ~(SKB_GSO_TCPV4 |
1306 SKB_GSO_UDP |
1307 SKB_GSO_DODGY |
1308 SKB_GSO_TCP_ECN |
1309 0)))
1310 goto out;
1311
1312 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001313 goto out;
1314
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001315 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001316 ihl = iph->ihl * 4;
1317 if (ihl < sizeof(*iph))
1318 goto out;
1319
Herbert Xubbcf4672006-07-03 19:38:35 -07001320 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001321 goto out;
1322
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001323 __skb_pull(skb, ihl);
1324 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001325 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001326 id = ntohs(iph->id);
David S. Millerf9242b62012-06-19 18:56:21 -07001327 proto = iph->protocol;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001328 segs = ERR_PTR(-EPROTONOSUPPORT);
1329
1330 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001331 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001332 if (likely(ops && ops->callbacks.gso_segment))
1333 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001334 rcu_read_unlock();
1335
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001336 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001337 goto out;
1338
1339 skb = segs;
1340 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001341 iph = ip_hdr(skb);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001342 if (proto == IPPROTO_UDP) {
1343 iph->id = htons(id);
1344 iph->frag_off = htons(offset >> 3);
1345 if (skb->next != NULL)
1346 iph->frag_off |= htons(IP_MF);
1347 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1348 } else
1349 iph->id = htons(id++);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001350 iph->tot_len = htons(skb->len - skb->mac_len);
1351 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001352 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001353 } while ((skb = skb->next));
1354
1355out:
1356 return segs;
1357}
1358
Herbert Xu73cc19f2008-12-15 23:41:09 -08001359static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1360 struct sk_buff *skb)
1361{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001362 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001363 struct sk_buff **pp = NULL;
1364 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001365 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001366 unsigned int hlen;
1367 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001368 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001369 int flush = 1;
1370 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001371
Herbert Xua5b1cf22009-05-26 18:50:28 +00001372 off = skb_gro_offset(skb);
1373 hlen = off + sizeof(*iph);
1374 iph = skb_gro_header_fast(skb, off);
1375 if (skb_gro_header_hard(skb, hlen)) {
1376 iph = skb_gro_header_slow(skb, hlen, off);
1377 if (unlikely(!iph))
1378 goto out;
1379 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001380
David S. Millerf9242b62012-06-19 18:56:21 -07001381 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001382
1383 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001384 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001385 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001386 goto out_unlock;
1387
Herbert Xua5ad24b2009-02-08 18:00:39 +00001388 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001389 goto out_unlock;
1390
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001391 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001392 goto out_unlock;
1393
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001394 id = ntohl(*(__be32 *)&iph->id);
1395 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001396 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001397
1398 for (p = *head; p; p = p->next) {
1399 struct iphdr *iph2;
1400
1401 if (!NAPI_GRO_CB(p)->same_flow)
1402 continue;
1403
1404 iph2 = ip_hdr(p);
1405
Herbert Xua5ad24b2009-02-08 18:00:39 +00001406 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001407 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1408 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001409 NAPI_GRO_CB(p)->same_flow = 0;
1410 continue;
1411 }
1412
1413 /* All fields must match except length and checksum. */
1414 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001415 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001416 (iph->tos ^ iph2->tos) |
Herbert Xua5ad24b2009-02-08 18:00:39 +00001417 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001418
1419 NAPI_GRO_CB(p)->flush |= flush;
1420 }
1421
1422 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001423 skb_gro_pull(skb, sizeof(*iph));
1424 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001425
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001426 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001427
1428out_unlock:
1429 rcu_read_unlock();
1430
1431out:
1432 NAPI_GRO_CB(skb)->flush |= flush;
1433
1434 return pp;
1435}
1436
1437static int inet_gro_complete(struct sk_buff *skb)
1438{
Herbert Xu73cc19f2008-12-15 23:41:09 -08001439 __be16 newlen = htons(skb->len - skb_network_offset(skb));
David S. Millerf9242b62012-06-19 18:56:21 -07001440 struct iphdr *iph = ip_hdr(skb);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001441 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001442 int proto = iph->protocol;
1443 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001444
1445 csum_replace2(&iph->check, iph->tot_len, newlen);
1446 iph->tot_len = newlen;
1447
1448 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001449 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001450 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001451 goto out_unlock;
1452
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001453 err = ops->callbacks.gro_complete(skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001454
1455out_unlock:
1456 rcu_read_unlock();
1457
1458 return err;
1459}
1460
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001461int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001462 unsigned short type, unsigned char protocol,
1463 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001464{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001465 struct socket *sock;
1466 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001467
1468 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001469 *sk = sock->sk;
1470 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001471 /*
1472 * Unhash it so that IP input processing does not even see it,
1473 * we do not wish this socket to see incoming packets.
1474 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001475 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001476
1477 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001478 }
1479 return rc;
1480}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001481EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1482
Tejun Heo7d720c32010-02-16 15:20:26 +00001483unsigned long snmp_fold_field(void __percpu *mib[], int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001484{
1485 unsigned long res = 0;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001486 int i, j;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001487
1488 for_each_possible_cpu(i) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001489 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1490 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001491 }
1492 return res;
1493}
1494EXPORT_SYMBOL_GPL(snmp_fold_field);
1495
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001496#if BITS_PER_LONG==32
1497
1498u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1499{
1500 u64 res = 0;
1501 int cpu;
1502
1503 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001504 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001505 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001506 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001507 unsigned int start;
1508
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001509 bhptr = per_cpu_ptr(mib[0], cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001510 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1511 do {
1512 start = u64_stats_fetch_begin_bh(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001513 v = *(((u64 *) bhptr) + offt);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001514 } while (u64_stats_fetch_retry_bh(syncp, start));
1515
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001516 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001517 }
1518 return res;
1519}
1520EXPORT_SYMBOL_GPL(snmp_fold_field64);
1521#endif
1522
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001523int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001524{
1525 BUG_ON(ptr == NULL);
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001526 ptr[0] = __alloc_percpu(mibsize, align);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001527 if (!ptr[0])
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001528 return -ENOMEM;
1529#if SNMP_ARRAY_SZ == 2
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001530 ptr[1] = __alloc_percpu(mibsize, align);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001531 if (!ptr[1]) {
1532 free_percpu(ptr[0]);
1533 ptr[0] = NULL;
1534 return -ENOMEM;
1535 }
1536#endif
Herbert Xu5e0f0432007-04-24 21:53:35 -07001537 return 0;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001538}
1539EXPORT_SYMBOL_GPL(snmp_mib_init);
1540
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001541void snmp_mib_free(void __percpu *ptr[SNMP_ARRAY_SZ])
Herbert Xu5e0f0432007-04-24 21:53:35 -07001542{
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001543 int i;
1544
Herbert Xu5e0f0432007-04-24 21:53:35 -07001545 BUG_ON(ptr == NULL);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001546 for (i = 0; i < SNMP_ARRAY_SZ; i++) {
1547 free_percpu(ptr[i]);
1548 ptr[i] = NULL;
1549 }
Herbert Xu5e0f0432007-04-24 21:53:35 -07001550}
1551EXPORT_SYMBOL_GPL(snmp_mib_free);
1552
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001554static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001556 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557};
1558#endif
1559
Alexey Dobriyan32613092009-09-14 12:21:47 +00001560static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001561 .early_demux = tcp_v4_early_demux,
1562 .handler = tcp_v4_rcv,
1563 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001564 .no_policy = 1,
1565 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566};
1567
Vlad Yasevichde27d002012-11-15 08:49:12 +00001568static const struct net_offload tcp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001569 .callbacks = {
1570 .gso_send_check = tcp_v4_gso_send_check,
1571 .gso_segment = tcp_tso_segment,
1572 .gro_receive = tcp4_gro_receive,
1573 .gro_complete = tcp4_gro_complete,
1574 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001575};
1576
Alexey Dobriyan32613092009-09-14 12:21:47 +00001577static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 .handler = udp_rcv,
1579 .err_handler = udp_err,
1580 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001581 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582};
1583
Vlad Yasevichde27d002012-11-15 08:49:12 +00001584static const struct net_offload udp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001585 .callbacks = {
1586 .gso_send_check = udp4_ufo_send_check,
1587 .gso_segment = udp4_ufo_fragment,
1588 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001589};
1590
Alexey Dobriyan32613092009-09-14 12:21:47 +00001591static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 .handler = icmp_rcv,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001593 .err_handler = ping_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001594 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001595 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596};
1597
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001598static __net_init int ipv4_mib_init_net(struct net *net)
1599{
Tejun Heo7d720c32010-02-16 15:20:26 +00001600 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001601 sizeof(struct tcp_mib),
1602 __alignof__(struct tcp_mib)) < 0)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001603 goto err_tcp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001604 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001605 sizeof(struct ipstats_mib),
1606 __alignof__(struct ipstats_mib)) < 0)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001607 goto err_ip_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001608 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001609 sizeof(struct linux_mib),
1610 __alignof__(struct linux_mib)) < 0)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001611 goto err_net_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001612 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001613 sizeof(struct udp_mib),
1614 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001615 goto err_udp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001616 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001617 sizeof(struct udp_mib),
1618 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001619 goto err_udplite_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001620 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001621 sizeof(struct icmp_mib),
1622 __alignof__(struct icmp_mib)) < 0)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001623 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001624 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1625 GFP_KERNEL);
1626 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001627 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001628
1629 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001630 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001631
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001632err_icmpmsg_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001633 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001634err_icmp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001635 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001636err_udplite_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001637 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001638err_udp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001639 snmp_mib_free((void __percpu **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001640err_net_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001641 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001642err_ip_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001643 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001644err_tcp_mib:
1645 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001646}
1647
1648static __net_exit void ipv4_mib_exit_net(struct net *net)
1649{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001650 kfree(net->mib.icmpmsg_statistics);
Tejun Heo7d720c32010-02-16 15:20:26 +00001651 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1652 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1653 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1654 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1655 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1656 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001657}
1658
1659static __net_initdata struct pernet_operations ipv4_mib_ops = {
1660 .init = ipv4_mib_init_net,
1661 .exit = ipv4_mib_exit_net,
1662};
1663
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664static int __init init_ipv4_mibs(void)
1665{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001666 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667}
1668
1669static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670
Herbert Xu30e224d2005-07-05 14:40:10 -07001671/*
1672 * IP protocol layer initialiser
1673 */
1674
Vlad Yasevich22061d82012-11-15 08:49:11 +00001675static struct packet_offload ip_packet_offload __read_mostly = {
1676 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001677 .callbacks = {
1678 .gso_send_check = inet_gso_send_check,
1679 .gso_segment = inet_gso_segment,
1680 .gro_receive = inet_gro_receive,
1681 .gro_complete = inet_gro_complete,
1682 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001683};
1684
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001685static int __init ipv4_offload_init(void)
1686{
1687 /*
1688 * Add offloads
1689 */
1690 if (inet_add_offload(&udp_offload, IPPROTO_UDP) < 0)
1691 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
1692 if (inet_add_offload(&tcp_offload, IPPROTO_TCP) < 0)
1693 pr_crit("%s: Cannot add TCP protocol offlaod\n", __func__);
1694
1695 dev_add_offload(&ip_packet_offload);
1696 return 0;
1697}
1698
1699fs_initcall(ipv4_offload_init);
1700
1701static struct packet_type ip_packet_type __read_mostly = {
1702 .type = cpu_to_be16(ETH_P_IP),
1703 .func = ip_rcv,
1704};
1705
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706static int __init inet_init(void)
1707{
1708 struct sk_buff *dummy_skb;
1709 struct inet_protosw *q;
1710 struct list_head *r;
1711 int rc = -EINVAL;
1712
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001713 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
Amerigo Wange3826f12010-05-05 00:27:06 +00001715 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1716 if (!sysctl_local_reserved_ports)
1717 goto out;
1718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 rc = proto_register(&tcp_prot, 1);
1720 if (rc)
Amerigo Wange3826f12010-05-05 00:27:06 +00001721 goto out_free_reserved_ports;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722
1723 rc = proto_register(&udp_prot, 1);
1724 if (rc)
1725 goto out_unregister_tcp_proto;
1726
1727 rc = proto_register(&raw_prot, 1);
1728 if (rc)
1729 goto out_unregister_udp_proto;
1730
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001731 rc = proto_register(&ping_prot, 1);
1732 if (rc)
1733 goto out_unregister_raw_proto;
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001736 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 */
1738
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001739 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Al Virobd7b1532008-07-15 16:00:59 -04001741#ifdef CONFIG_SYSCTL
1742 ip_static_sysctl_init();
1743#endif
1744
Glauber Costa3dc43e32011-12-11 21:47:05 +00001745 tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1746
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 /*
1748 * Add all the base protocols.
1749 */
1750
1751 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001752 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001754 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001756 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757#ifdef CONFIG_IP_MULTICAST
1758 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001759 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760#endif
1761
1762 /* Register the socket-side information for inet_create. */
1763 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1764 INIT_LIST_HEAD(r);
1765
1766 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1767 inet_register_protosw(q);
1768
1769 /*
1770 * Set the ARP module up
1771 */
1772
1773 arp_init();
1774
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001775 /*
1776 * Set the IP module up
1777 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778
1779 ip_init();
1780
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001781 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
1783 /* Setup TCP slab cache for open requests. */
1784 tcp_init();
1785
Hideo Aoki95766ff2007-12-31 00:29:24 -08001786 /* Setup UDP memory threshold */
1787 udp_init();
1788
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001789 /* Add UDP-Lite (RFC 3828) */
1790 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001792 ping_init();
1793
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 /*
1795 * Set the ICMP layer up
1796 */
1797
Denis V. Luneva5710d62008-02-29 11:14:50 -08001798 if (icmp_init() < 0)
1799 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800
1801 /*
1802 * Initialise the multicast router
1803 */
1804#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001805 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001806 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807#endif
1808 /*
1809 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001810 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811
Stephen Hemminger132adf52007-03-08 20:44:43 -08001812 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001813 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001814
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 ipv4_proc_init();
1816
1817 ipfrag_init();
1818
Herbert Xu30e224d2005-07-05 14:40:10 -07001819 dev_add_pack(&ip_packet_type);
1820
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 rc = 0;
1822out:
1823 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001824out_unregister_raw_proto:
1825 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826out_unregister_udp_proto:
1827 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001828out_unregister_tcp_proto:
1829 proto_unregister(&tcp_prot);
Amerigo Wange3826f12010-05-05 00:27:06 +00001830out_free_reserved_ports:
1831 kfree(sysctl_local_reserved_ports);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 goto out;
1833}
1834
Heiko Carstensa536e072006-04-28 15:19:17 -07001835fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
1837/* ------------------------------------------------------------------------ */
1838
1839#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840static int __init ipv4_proc_init(void)
1841{
1842 int rc = 0;
1843
1844 if (raw_proc_init())
1845 goto out_raw;
1846 if (tcp4_proc_init())
1847 goto out_tcp;
1848 if (udp4_proc_init())
1849 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001850 if (ping_proc_init())
1851 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 if (ip_misc_proc_init())
1853 goto out_misc;
1854out:
1855 return rc;
1856out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001857 ping_proc_exit();
1858out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 udp4_proc_exit();
1860out_udp:
1861 tcp4_proc_exit();
1862out_tcp:
1863 raw_proc_exit();
1864out_raw:
1865 rc = -ENOMEM;
1866 goto out;
1867}
1868
1869#else /* CONFIG_PROC_FS */
1870static int __init ipv4_proc_init(void)
1871{
1872 return 0;
1873}
1874#endif /* CONFIG_PROC_FS */
1875
1876MODULE_ALIAS_NETPROTO(PF_INET);
1877