blob: d01be2a3ae53170c1075a57a32f2b11e1e75885b [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
Eric Dumazetbe776282007-03-27 13:53:04 -0700248u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700249EXPORT_SYMBOL(inet_ehash_secret);
250
Eric Dumazet08dcdbf2013-02-21 12:18:52 +0000251u32 ipv6_hash_secret __read_mostly;
252EXPORT_SYMBOL(ipv6_hash_secret);
253
Eric Dumazetbe776282007-03-27 13:53:04 -0700254/*
Eric Dumazet08dcdbf2013-02-21 12:18:52 +0000255 * inet_ehash_secret must be set exactly once, and to a non nul value
256 * ipv6_hash_secret must be set exactly once.
Eric Dumazetbe776282007-03-27 13:53:04 -0700257 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700258void build_ehash_secret(void)
259{
Eric Dumazetbe776282007-03-27 13:53:04 -0700260 u32 rnd;
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000261
Eric Dumazetbe776282007-03-27 13:53:04 -0700262 do {
263 get_random_bytes(&rnd, sizeof(rnd));
264 } while (rnd == 0);
Eric Dumazet49e8ab02010-08-19 06:10:45 +0000265
Eric Dumazetaebda152013-04-29 05:58:52 +0000266 if (cmpxchg(&inet_ehash_secret, 0, rnd) == 0) {
Eric Dumazet08dcdbf2013-02-21 12:18:52 +0000267 get_random_bytes(&ipv6_hash_secret, sizeof(ipv6_hash_secret));
Eric Dumazetaebda152013-04-29 05:58:52 +0000268 net_secret_init();
269 }
David S. Millerb3da2cf2007-03-23 11:40:27 -0700270}
271EXPORT_SYMBOL(build_ehash_secret);
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273/*
274 * Create an inet socket.
275 */
276
Eric Paris3f378b62009-11-05 22:18:14 -0800277static int inet_create(struct net *net, struct socket *sock, int protocol,
278 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279{
280 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 struct inet_protosw *answer;
282 struct inet_sock *inet;
283 struct proto *answer_prot;
284 unsigned char answer_flags;
285 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700286 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800287 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
Eric Dumazet04f258c2008-11-23 15:42:23 -0800289 if (unlikely(!inet_ehash_secret))
290 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
291 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 sock->state = SS_UNCONNECTED;
294
295 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700296lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800297 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700299 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700301 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 /* Check the non-wild match. */
303 if (protocol == answer->protocol) {
304 if (protocol != IPPROTO_IP)
305 break;
306 } else {
307 /* Check for the two wild cases. */
308 if (IPPROTO_IP == protocol) {
309 protocol = answer->protocol;
310 break;
311 }
312 if (IPPROTO_IP == answer->protocol)
313 break;
314 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800315 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 }
317
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700318 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700319 if (try_loading_module < 2) {
320 rcu_read_unlock();
321 /*
322 * Be more specific, e.g. net-pf-2-proto-132-type-1
323 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
324 */
325 if (++try_loading_module == 1)
326 request_module("net-pf-%d-proto-%d-type-%d",
327 PF_INET, protocol, sock->type);
328 /*
329 * Fall back to generic, e.g. net-pf-2-proto-132
330 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
331 */
332 else
333 request_module("net-pf-%d-proto-%d",
334 PF_INET, protocol);
335 goto lookup_protocol;
336 } else
337 goto out_rcu_unlock;
338 }
339
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000341 if (sock->type == SOCK_RAW && !kern &&
342 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344
345 sock->ops = answer->ops;
346 answer_prot = answer->prot;
347 answer_no_check = answer->no_check;
348 answer_flags = answer->flags;
349 rcu_read_unlock();
350
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700351 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352
353 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700354 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 if (sk == NULL)
356 goto out;
357
358 err = 0;
359 sk->sk_no_check = answer_no_check;
360 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000361 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
363 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800364 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000366 inet->nodefrag = 0;
367
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000369 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 if (IPPROTO_RAW == protocol)
371 inet->hdrincl = 1;
372 }
373
374 if (ipv4_config.no_pmtu_disc)
375 inet->pmtudisc = IP_PMTUDISC_DONT;
376 else
377 inet->pmtudisc = IP_PMTUDISC_WANT;
378
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000379 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
381 sock_init_data(sock, sk);
382
383 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 sk->sk_protocol = protocol;
385 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
386
387 inet->uc_ttl = -1;
388 inet->mc_loop = 1;
389 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000390 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 inet->mc_index = 0;
392 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000393 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700395 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000397 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 /* It assumes that any protocol which allows
399 * the user to assign a number at socket
400 * creation time automatically
401 * shares.
402 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000403 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 /* Add to protocol hash chains. */
405 sk->sk_prot->hash(sk);
406 }
407
408 if (sk->sk_prot->init) {
409 err = sk->sk_prot->init(sk);
410 if (err)
411 sk_common_release(sk);
412 }
413out:
414 return err;
415out_rcu_unlock:
416 rcu_read_unlock();
417 goto out;
418}
419
420
421/*
422 * The peer socket should always be NULL (or else). When we call this
423 * function we are destroying the object and from then on nobody
424 * should refer to it.
425 */
426int inet_release(struct socket *sock)
427{
428 struct sock *sk = sock->sk;
429
430 if (sk) {
431 long timeout;
432
David S. Millerc58dc012010-04-27 15:05:31 -0700433 sock_rps_reset_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700434
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 /* Applications forget to leave groups before exiting */
436 ip_mc_drop_socket(sk);
437
438 /* If linger is set, we don't return until the close
439 * is complete. Otherwise we return immediately. The
440 * actually closing is done the same either way.
441 *
442 * If the close is due to the process exiting, we never
443 * linger..
444 */
445 timeout = 0;
446 if (sock_flag(sk, SOCK_LINGER) &&
447 !(current->flags & PF_EXITING))
448 timeout = sk->sk_lingertime;
449 sock->sk = NULL;
450 sk->sk_prot->close(sk, timeout);
451 }
452 return 0;
453}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700454EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
456/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700457int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700458EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
460int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
461{
462 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
463 struct sock *sk = sock->sk;
464 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000465 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 unsigned short snum;
467 int chk_addr_ret;
468 int err;
469
470 /* If the socket has its own bind function then use it. (RAW) */
471 if (sk->sk_prot->bind) {
472 err = sk->sk_prot->bind(sk, uaddr, addr_len);
473 goto out;
474 }
475 err = -EINVAL;
476 if (addr_len < sizeof(struct sockaddr_in))
477 goto out;
478
Marcus Meissnerc349a522011-07-04 01:30:29 +0000479 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400480 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
481 * only if s_addr is INADDR_ANY.
482 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000483 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400484 if (addr->sin_family != AF_UNSPEC ||
485 addr->sin_addr.s_addr != htonl(INADDR_ANY))
486 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000487 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700488
Eric W. Biederman35946982012-11-16 03:03:12 +0000489 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 /* Not specified by any standard per-se, however it breaks too
492 * many applications when removed. It is unfortunate since
493 * allowing applications to make a non-local bind solves
494 * several problems with systems using dynamic addressing.
495 * (ie. your servers still start up even if your ISDN link
496 * is temporarily down)
497 */
498 err = -EADDRNOTAVAIL;
499 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700500 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700501 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 chk_addr_ret != RTN_LOCAL &&
503 chk_addr_ret != RTN_MULTICAST &&
504 chk_addr_ret != RTN_BROADCAST)
505 goto out;
506
507 snum = ntohs(addr->sin_port);
508 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000509 if (snum && snum < PROT_SOCK &&
510 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 goto out;
512
513 /* We keep a pair of addresses. rcv_saddr is the one
514 * used by hash lookups, and saddr is used for transmit.
515 *
516 * In the BSD API these are the same except where it
517 * would be illegal to use them (multicast/broadcast) in
518 * which case the sending device address is used.
519 */
520 lock_sock(sk);
521
522 /* Check these errors (active socket, double bind). */
523 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000524 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 goto out_release_sock;
526
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000527 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000529 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
531 /* Make sure we are allowed to bind here. */
532 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000533 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 err = -EADDRINUSE;
535 goto out_release_sock;
536 }
537
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000538 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
540 if (snum)
541 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000542 inet->inet_sport = htons(inet->inet_num);
543 inet->inet_daddr = 0;
544 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 sk_dst_reset(sk);
546 err = 0;
547out_release_sock:
548 release_sock(sk);
549out:
550 return err;
551}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700552EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000554int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 int addr_len, int flags)
556{
557 struct sock *sk = sock->sk;
558
Changli Gao6503d962010-03-31 22:58:26 +0000559 if (addr_len < sizeof(uaddr->sa_family))
560 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 if (uaddr->sa_family == AF_UNSPEC)
562 return sk->sk_prot->disconnect(sk, flags);
563
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000564 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000566 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700568EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
Yuchung Cheng783237e2012-07-19 06:43:07 +0000570static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
572 DEFINE_WAIT(wait);
573
Eric Dumazetaa395142010-04-20 13:03:51 +0000574 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000575 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
577 /* Basic assumption: if someone sets sk->sk_err, he _must_
578 * change state of the socket from TCP_SYN_*.
579 * Connect() does not allow to get error notifications
580 * without closing the socket.
581 */
582 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
583 release_sock(sk);
584 timeo = schedule_timeout(timeo);
585 lock_sock(sk);
586 if (signal_pending(current) || !timeo)
587 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000588 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000590 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000591 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 return timeo;
593}
594
595/*
596 * Connect to a remote host. There is regrettably still a little
597 * TCP 'magic' in here.
598 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000599int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
600 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
602 struct sock *sk = sock->sk;
603 int err;
604 long timeo;
605
Changli Gao6503d962010-03-31 22:58:26 +0000606 if (addr_len < sizeof(uaddr->sa_family))
607 return -EINVAL;
608
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 if (uaddr->sa_family == AF_UNSPEC) {
610 err = sk->sk_prot->disconnect(sk, flags);
611 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
612 goto out;
613 }
614
615 switch (sock->state) {
616 default:
617 err = -EINVAL;
618 goto out;
619 case SS_CONNECTED:
620 err = -EISCONN;
621 goto out;
622 case SS_CONNECTING:
623 err = -EALREADY;
624 /* Fall out of switch with err, set for this state */
625 break;
626 case SS_UNCONNECTED:
627 err = -EISCONN;
628 if (sk->sk_state != TCP_CLOSE)
629 goto out;
630
631 err = sk->sk_prot->connect(sk, uaddr, addr_len);
632 if (err < 0)
633 goto out;
634
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900635 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637 /* Just entered SS_CONNECTING state; the only
638 * difference is that return value in non-blocking
639 * case is EINPROGRESS, rather than EALREADY.
640 */
641 err = -EINPROGRESS;
642 break;
643 }
644
645 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
646
647 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000648 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
649 tcp_sk(sk)->fastopen_req &&
650 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
651
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000653 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 goto out;
655
656 err = sock_intr_errno(timeo);
657 if (signal_pending(current))
658 goto out;
659 }
660
661 /* Connection was closed by RST, timeout, ICMP error
662 * or another process disconnected us.
663 */
664 if (sk->sk_state == TCP_CLOSE)
665 goto sock_error;
666
667 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
668 * and error was received after socket entered established state.
669 * Hence, it is handled normally after connect() return successfully.
670 */
671
672 sock->state = SS_CONNECTED;
673 err = 0;
674out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 return err;
676
677sock_error:
678 err = sock_error(sk) ? : -ECONNABORTED;
679 sock->state = SS_UNCONNECTED;
680 if (sk->sk_prot->disconnect(sk, flags))
681 sock->state = SS_DISCONNECTING;
682 goto out;
683}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000684EXPORT_SYMBOL(__inet_stream_connect);
685
686int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
687 int addr_len, int flags)
688{
689 int err;
690
691 lock_sock(sock->sk);
692 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
693 release_sock(sock->sk);
694 return err;
695}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700696EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
698/*
699 * Accept a pending connection. The TCP layer now gives BSD semantics.
700 */
701
702int inet_accept(struct socket *sock, struct socket *newsock, int flags)
703{
704 struct sock *sk1 = sock->sk;
705 int err = -EINVAL;
706 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
707
708 if (!sk2)
709 goto do_err;
710
711 lock_sock(sk2);
712
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000713 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700714 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000715 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
716 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
718 sock_graft(sk2, newsock);
719
720 newsock->state = SS_CONNECTED;
721 err = 0;
722 release_sock(sk2);
723do_err:
724 return err;
725}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700726EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728
729/*
730 * This does both peername and sockname.
731 */
732int inet_getname(struct socket *sock, struct sockaddr *uaddr,
733 int *uaddr_len, int peer)
734{
735 struct sock *sk = sock->sk;
736 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700737 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 sin->sin_family = AF_INET;
740 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000741 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
743 peer == 1))
744 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000745 sin->sin_port = inet->inet_dport;
746 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000748 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000750 addr = inet->inet_saddr;
751 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 sin->sin_addr.s_addr = addr;
753 }
754 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
755 *uaddr_len = sizeof(*sin);
756 return 0;
757}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700758EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
761 size_t size)
762{
763 struct sock *sk = sock->sk;
764
David S. Millerc58dc012010-04-27 15:05:31 -0700765 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000768 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
769 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 return -EAGAIN;
771
772 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
773}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700774EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Changli Gao7ba42912010-07-10 20:41:55 +0000776ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
777 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778{
779 struct sock *sk = sock->sk;
780
David S. Millerc58dc012010-04-27 15:05:31 -0700781 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000784 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
785 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 return -EAGAIN;
787
788 if (sk->sk_prot->sendpage)
789 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
790 return sock_no_sendpage(sock, page, offset, size, flags);
791}
Changli Gao7ba42912010-07-10 20:41:55 +0000792EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
Tom Herbertfec5e652010-04-16 16:01:27 -0700794int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
795 size_t size, int flags)
796{
797 struct sock *sk = sock->sk;
798 int addr_len = 0;
799 int err;
800
David S. Millerc58dc012010-04-27 15:05:31 -0700801 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700802
803 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
804 flags & ~MSG_DONTWAIT, &addr_len);
805 if (err >= 0)
806 msg->msg_namelen = addr_len;
807 return err;
808}
809EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811int inet_shutdown(struct socket *sock, int how)
812{
813 struct sock *sk = sock->sk;
814 int err = 0;
815
816 /* This should really check to make sure
817 * the socket is a TCP socket. (WHY AC...)
818 */
819 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
820 1->2 bit 2 snds.
821 2->3 */
822 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
823 return -EINVAL;
824
825 lock_sock(sk);
826 if (sock->state == SS_CONNECTING) {
827 if ((1 << sk->sk_state) &
828 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
829 sock->state = SS_DISCONNECTING;
830 else
831 sock->state = SS_CONNECTED;
832 }
833
834 switch (sk->sk_state) {
835 case TCP_CLOSE:
836 err = -ENOTCONN;
837 /* Hack to wake up other listeners, who can poll for
838 POLLHUP, even on eg. unconnected UDP sockets -- RR */
839 default:
840 sk->sk_shutdown |= how;
841 if (sk->sk_prot->shutdown)
842 sk->sk_prot->shutdown(sk, how);
843 break;
844
845 /* Remaining two branches are temporary solution for missing
846 * close() in multithreaded environment. It is _not_ a good idea,
847 * but we have no choice until close() is repaired at VFS level.
848 */
849 case TCP_LISTEN:
850 if (!(how & RCV_SHUTDOWN))
851 break;
852 /* Fall through */
853 case TCP_SYN_SENT:
854 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
855 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
856 break;
857 }
858
859 /* Wake up anyone sleeping in poll. */
860 sk->sk_state_change(sk);
861 release_sock(sk);
862 return err;
863}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700864EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866/*
867 * ioctl() calls you can issue on an INET socket. Most of these are
868 * device configuration and stuff and very rarely used. Some ioctls
869 * pass on to the socket itself.
870 *
871 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
872 * loads the devconfigure module does its configuring and unloads it.
873 * There's a good 20K of config code hanging around the kernel.
874 */
875
876int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
877{
878 struct sock *sk = sock->sk;
879 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900880 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700883 case SIOCGSTAMP:
884 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
885 break;
886 case SIOCGSTAMPNS:
887 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
888 break;
889 case SIOCADDRT:
890 case SIOCDELRT:
891 case SIOCRTMSG:
892 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
893 break;
894 case SIOCDARP:
895 case SIOCGARP:
896 case SIOCSARP:
897 err = arp_ioctl(net, cmd, (void __user *)arg);
898 break;
899 case SIOCGIFADDR:
900 case SIOCSIFADDR:
901 case SIOCGIFBRDADDR:
902 case SIOCSIFBRDADDR:
903 case SIOCGIFNETMASK:
904 case SIOCSIFNETMASK:
905 case SIOCGIFDSTADDR:
906 case SIOCSIFDSTADDR:
907 case SIOCSIFPFLAGS:
908 case SIOCGIFPFLAGS:
909 case SIOCSIFFLAGS:
910 err = devinet_ioctl(net, cmd, (void __user *)arg);
911 break;
912 default:
913 if (sk->sk_prot->ioctl)
914 err = sk->sk_prot->ioctl(sk, cmd, arg);
915 else
916 err = -ENOIOCTLCMD;
917 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 }
919 return err;
920}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700921EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000923#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000924static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000925{
926 struct sock *sk = sock->sk;
927 int err = -ENOIOCTLCMD;
928
929 if (sk->sk_prot->compat_ioctl)
930 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
931
932 return err;
933}
934#endif
935
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800936const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800937 .family = PF_INET,
938 .owner = THIS_MODULE,
939 .release = inet_release,
940 .bind = inet_bind,
941 .connect = inet_stream_connect,
942 .socketpair = sock_no_socketpair,
943 .accept = inet_accept,
944 .getname = inet_getname,
945 .poll = tcp_poll,
946 .ioctl = inet_ioctl,
947 .listen = inet_listen,
948 .shutdown = inet_shutdown,
949 .setsockopt = sock_common_setsockopt,
950 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000951 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700952 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800953 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000954 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800955 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800956#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800957 .compat_setsockopt = compat_sock_common_setsockopt,
958 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000959 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800960#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700962EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800964const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800965 .family = PF_INET,
966 .owner = THIS_MODULE,
967 .release = inet_release,
968 .bind = inet_bind,
969 .connect = inet_dgram_connect,
970 .socketpair = sock_no_socketpair,
971 .accept = sock_no_accept,
972 .getname = inet_getname,
973 .poll = udp_poll,
974 .ioctl = inet_ioctl,
975 .listen = sock_no_listen,
976 .shutdown = inet_shutdown,
977 .setsockopt = sock_common_setsockopt,
978 .getsockopt = sock_common_getsockopt,
979 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700980 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800981 .mmap = sock_no_mmap,
982 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800983#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800984 .compat_setsockopt = compat_sock_common_setsockopt,
985 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000986 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800987#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700989EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
991/*
992 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
993 * udp_poll
994 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800995static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800996 .family = PF_INET,
997 .owner = THIS_MODULE,
998 .release = inet_release,
999 .bind = inet_bind,
1000 .connect = inet_dgram_connect,
1001 .socketpair = sock_no_socketpair,
1002 .accept = sock_no_accept,
1003 .getname = inet_getname,
1004 .poll = datagram_poll,
1005 .ioctl = inet_ioctl,
1006 .listen = sock_no_listen,
1007 .shutdown = inet_shutdown,
1008 .setsockopt = sock_common_setsockopt,
1009 .getsockopt = sock_common_getsockopt,
1010 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001011 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001012 .mmap = sock_no_mmap,
1013 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001014#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001015 .compat_setsockopt = compat_sock_common_setsockopt,
1016 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001017 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001018#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019};
1020
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001021static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 .family = PF_INET,
1023 .create = inet_create,
1024 .owner = THIS_MODULE,
1025};
1026
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027/* Upon startup we insert all the elements in inetsw_array[] into
1028 * the linked list inetsw.
1029 */
1030static struct inet_protosw inetsw_array[] =
1031{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001032 {
1033 .type = SOCK_STREAM,
1034 .protocol = IPPROTO_TCP,
1035 .prot = &tcp_prot,
1036 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001037 .no_check = 0,
1038 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001039 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001040 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001042 {
1043 .type = SOCK_DGRAM,
1044 .protocol = IPPROTO_UDP,
1045 .prot = &udp_prot,
1046 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001047 .no_check = UDP_CSUM_DEFAULT,
1048 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001050
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001051 {
1052 .type = SOCK_DGRAM,
1053 .protocol = IPPROTO_ICMP,
1054 .prot = &ping_prot,
1055 .ops = &inet_dgram_ops,
1056 .no_check = UDP_CSUM_DEFAULT,
1057 .flags = INET_PROTOSW_REUSE,
1058 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
1060 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001061 .type = SOCK_RAW,
1062 .protocol = IPPROTO_IP, /* wild card */
1063 .prot = &raw_prot,
1064 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001065 .no_check = UDP_CSUM_DEFAULT,
1066 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 }
1068};
1069
Denis Chengc40f6ff2007-09-16 16:39:25 -07001070#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
1072void inet_register_protosw(struct inet_protosw *p)
1073{
1074 struct list_head *lh;
1075 struct inet_protosw *answer;
1076 int protocol = p->protocol;
1077 struct list_head *last_perm;
1078
1079 spin_lock_bh(&inetsw_lock);
1080
1081 if (p->type >= SOCK_MAX)
1082 goto out_illegal;
1083
1084 /* If we are trying to override a permanent protocol, bail. */
1085 answer = NULL;
1086 last_perm = &inetsw[p->type];
1087 list_for_each(lh, &inetsw[p->type]) {
1088 answer = list_entry(lh, struct inet_protosw, list);
1089
1090 /* Check only the non-wild match. */
1091 if (INET_PROTOSW_PERMANENT & answer->flags) {
1092 if (protocol == answer->protocol)
1093 break;
1094 last_perm = lh;
1095 }
1096
1097 answer = NULL;
1098 }
1099 if (answer)
1100 goto out_permanent;
1101
1102 /* Add the new entry after the last permanent entry if any, so that
1103 * the new entry does not override a permanent entry when matched with
1104 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001105 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 * system automatically returns to the old behavior.
1107 */
1108 list_add_rcu(&p->list, last_perm);
1109out:
1110 spin_unlock_bh(&inetsw_lock);
1111
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 return;
1113
1114out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001115 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 goto out;
1117
1118out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001119 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 p->type);
1121 goto out;
1122}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001123EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
1125void inet_unregister_protosw(struct inet_protosw *p)
1126{
1127 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001128 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 p->protocol);
1130 } else {
1131 spin_lock_bh(&inetsw_lock);
1132 list_del_rcu(&p->list);
1133 spin_unlock_bh(&inetsw_lock);
1134
1135 synchronize_net();
1136 }
1137}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001138EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001140/*
1141 * Shall we try to damage output packets if routing dev changes?
1142 */
1143
Brian Haleyab32ea52006-09-22 14:15:41 -07001144int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001145
1146static int inet_sk_reselect_saddr(struct sock *sk)
1147{
1148 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001149 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001150 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001151 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001152 struct rtable *rt;
1153 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001154 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001155
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001156 inet_opt = rcu_dereference_protected(inet->inet_opt,
1157 sock_owned_by_user(sk));
1158 if (inet_opt && inet_opt->opt.srr)
1159 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001160
1161 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001162 fl4 = &inet->cork.fl.u.ip4;
1163 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001164 sk->sk_bound_dev_if, sk->sk_protocol,
1165 inet->inet_sport, inet->inet_dport, sk, false);
1166 if (IS_ERR(rt))
1167 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001168
Changli Gaod8d1f302010-06-10 23:31:35 -07001169 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001170
David S. Miller6e869132011-05-06 16:18:04 -07001171 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001172
1173 if (new_saddr == old_saddr)
1174 return 0;
1175
1176 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001177 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1178 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001179 }
1180
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001181 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001182
1183 /*
1184 * XXX The only one ugly spot where we need to
1185 * XXX really change the sockets identity after
1186 * XXX it has entered the hashes. -DaveM
1187 *
1188 * Besides that, it does not check for connection
1189 * uniqueness. Wait for troubles.
1190 */
1191 __sk_prot_rehash(sk);
1192 return 0;
1193}
1194
1195int inet_sk_rebuild_header(struct sock *sk)
1196{
1197 struct inet_sock *inet = inet_sk(sk);
1198 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001199 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001200 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001201 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001202 int err;
1203
1204 /* Route is OK, nothing to do. */
1205 if (rt)
1206 return 0;
1207
1208 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001209 rcu_read_lock();
1210 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001211 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001212 if (inet_opt && inet_opt->opt.srr)
1213 daddr = inet_opt->opt.faddr;
1214 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001215 fl4 = &inet->cork.fl.u.ip4;
1216 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001217 inet->inet_dport, inet->inet_sport,
1218 sk->sk_protocol, RT_CONN_FLAGS(sk),
1219 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001220 if (!IS_ERR(rt)) {
1221 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001222 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001223 } else {
1224 err = PTR_ERR(rt);
1225
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001226 /* Routing failed... */
1227 sk->sk_route_caps = 0;
1228 /*
1229 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1230 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1231 */
1232 if (!sysctl_ip_dynaddr ||
1233 sk->sk_state != TCP_SYN_SENT ||
1234 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1235 (err = inet_sk_reselect_saddr(sk)) != 0)
1236 sk->sk_err_soft = -err;
1237 }
1238
1239 return err;
1240}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001241EXPORT_SYMBOL(inet_sk_rebuild_header);
1242
Herbert Xua430a432006-07-08 13:34:56 -07001243static int inet_gso_send_check(struct sk_buff *skb)
1244{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001245 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001246 const struct iphdr *iph;
Herbert Xua430a432006-07-08 13:34:56 -07001247 int proto;
1248 int ihl;
1249 int err = -EINVAL;
1250
1251 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1252 goto out;
1253
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001254 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001255 ihl = iph->ihl * 4;
1256 if (ihl < sizeof(*iph))
1257 goto out;
1258
1259 if (unlikely(!pskb_may_pull(skb, ihl)))
1260 goto out;
1261
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001262 __skb_pull(skb, ihl);
1263 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001264 iph = ip_hdr(skb);
David S. Millerf9242b62012-06-19 18:56:21 -07001265 proto = iph->protocol;
Herbert Xua430a432006-07-08 13:34:56 -07001266 err = -EPROTONOSUPPORT;
1267
1268 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001269 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001270 if (likely(ops && ops->callbacks.gso_send_check))
1271 err = ops->callbacks.gso_send_check(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001272 rcu_read_unlock();
1273
1274out:
1275 return err;
1276}
1277
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001278static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1279 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001280{
1281 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001282 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001283 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001284 int proto;
1285 int ihl;
1286 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001287 unsigned int offset = 0;
Pravin B Shelar73136262013-03-07 13:21:51 +00001288 bool tunnel;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001289
Herbert Xubbcf4672006-07-03 19:38:35 -07001290 if (unlikely(skb_shinfo(skb)->gso_type &
1291 ~(SKB_GSO_TCPV4 |
1292 SKB_GSO_UDP |
1293 SKB_GSO_DODGY |
1294 SKB_GSO_TCP_ECN |
Pravin B Shelar68c33162013-02-14 14:02:41 +00001295 SKB_GSO_GRE |
Pravin B Shelar9b3eb5e2013-05-02 16:14:19 +00001296 SKB_GSO_TCPV6 |
Pravin B Shelar73136262013-03-07 13:21:51 +00001297 SKB_GSO_UDP_TUNNEL |
Herbert Xubbcf4672006-07-03 19:38:35 -07001298 0)))
1299 goto out;
1300
1301 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001302 goto out;
1303
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001304 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001305 ihl = iph->ihl * 4;
1306 if (ihl < sizeof(*iph))
1307 goto out;
1308
Herbert Xubbcf4672006-07-03 19:38:35 -07001309 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001310 goto out;
1311
Pravin B Shelar73136262013-03-07 13:21:51 +00001312 tunnel = !!skb->encapsulation;
1313
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001314 __skb_pull(skb, ihl);
1315 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001316 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001317 id = ntohs(iph->id);
David S. Millerf9242b62012-06-19 18:56:21 -07001318 proto = iph->protocol;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001319 segs = ERR_PTR(-EPROTONOSUPPORT);
1320
1321 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001322 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001323 if (likely(ops && ops->callbacks.gso_segment))
1324 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001325 rcu_read_unlock();
1326
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001327 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001328 goto out;
1329
1330 skb = segs;
1331 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001332 iph = ip_hdr(skb);
Pravin B Shelar73136262013-03-07 13:21:51 +00001333 if (!tunnel && proto == IPPROTO_UDP) {
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001334 iph->id = htons(id);
1335 iph->frag_off = htons(offset >> 3);
1336 if (skb->next != NULL)
1337 iph->frag_off |= htons(IP_MF);
1338 offset += (skb->len - skb->mac_len - iph->ihl * 4);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001339 } else {
Pravin B Shelar25c77042013-03-24 17:36:29 +00001340 iph->id = htons(id++);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001341 }
Herbert Xuf4c50d92006-06-22 03:02:40 -07001342 iph->tot_len = htons(skb->len - skb->mac_len);
1343 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001344 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001345 } while ((skb = skb->next));
1346
1347out:
1348 return segs;
1349}
1350
Herbert Xu73cc19f2008-12-15 23:41:09 -08001351static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1352 struct sk_buff *skb)
1353{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001354 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001355 struct sk_buff **pp = NULL;
1356 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001357 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001358 unsigned int hlen;
1359 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001360 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001361 int flush = 1;
1362 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001363
Herbert Xua5b1cf22009-05-26 18:50:28 +00001364 off = skb_gro_offset(skb);
1365 hlen = off + sizeof(*iph);
1366 iph = skb_gro_header_fast(skb, off);
1367 if (skb_gro_header_hard(skb, hlen)) {
1368 iph = skb_gro_header_slow(skb, hlen, off);
1369 if (unlikely(!iph))
1370 goto out;
1371 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001372
David S. Millerf9242b62012-06-19 18:56:21 -07001373 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001374
1375 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001376 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001377 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001378 goto out_unlock;
1379
Herbert Xua5ad24b2009-02-08 18:00:39 +00001380 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001381 goto out_unlock;
1382
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001383 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001384 goto out_unlock;
1385
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001386 id = ntohl(*(__be32 *)&iph->id);
1387 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001388 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001389
1390 for (p = *head; p; p = p->next) {
1391 struct iphdr *iph2;
1392
1393 if (!NAPI_GRO_CB(p)->same_flow)
1394 continue;
1395
1396 iph2 = ip_hdr(p);
1397
Herbert Xua5ad24b2009-02-08 18:00:39 +00001398 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001399 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1400 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001401 NAPI_GRO_CB(p)->same_flow = 0;
1402 continue;
1403 }
1404
1405 /* All fields must match except length and checksum. */
1406 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001407 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001408 (iph->tos ^ iph2->tos) |
Herbert Xua5ad24b2009-02-08 18:00:39 +00001409 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001410
1411 NAPI_GRO_CB(p)->flush |= flush;
1412 }
1413
1414 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001415 skb_gro_pull(skb, sizeof(*iph));
1416 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001417
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001418 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001419
1420out_unlock:
1421 rcu_read_unlock();
1422
1423out:
1424 NAPI_GRO_CB(skb)->flush |= flush;
1425
1426 return pp;
1427}
1428
1429static int inet_gro_complete(struct sk_buff *skb)
1430{
Herbert Xu73cc19f2008-12-15 23:41:09 -08001431 __be16 newlen = htons(skb->len - skb_network_offset(skb));
David S. Millerf9242b62012-06-19 18:56:21 -07001432 struct iphdr *iph = ip_hdr(skb);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001433 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001434 int proto = iph->protocol;
1435 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001436
1437 csum_replace2(&iph->check, iph->tot_len, newlen);
1438 iph->tot_len = newlen;
1439
1440 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001441 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001442 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001443 goto out_unlock;
1444
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001445 err = ops->callbacks.gro_complete(skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001446
1447out_unlock:
1448 rcu_read_unlock();
1449
1450 return err;
1451}
1452
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001453int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001454 unsigned short type, unsigned char protocol,
1455 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001456{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001457 struct socket *sock;
1458 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001459
1460 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001461 *sk = sock->sk;
1462 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001463 /*
1464 * Unhash it so that IP input processing does not even see it,
1465 * we do not wish this socket to see incoming packets.
1466 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001467 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001468
1469 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001470 }
1471 return rc;
1472}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001473EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1474
Tejun Heo7d720c32010-02-16 15:20:26 +00001475unsigned long snmp_fold_field(void __percpu *mib[], int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001476{
1477 unsigned long res = 0;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001478 int i, j;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001479
1480 for_each_possible_cpu(i) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001481 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1482 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001483 }
1484 return res;
1485}
1486EXPORT_SYMBOL_GPL(snmp_fold_field);
1487
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001488#if BITS_PER_LONG==32
1489
1490u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1491{
1492 u64 res = 0;
1493 int cpu;
1494
1495 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001496 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001497 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001498 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001499 unsigned int start;
1500
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001501 bhptr = per_cpu_ptr(mib[0], cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001502 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1503 do {
1504 start = u64_stats_fetch_begin_bh(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001505 v = *(((u64 *) bhptr) + offt);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001506 } while (u64_stats_fetch_retry_bh(syncp, start));
1507
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001508 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001509 }
1510 return res;
1511}
1512EXPORT_SYMBOL_GPL(snmp_fold_field64);
1513#endif
1514
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001515int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001516{
1517 BUG_ON(ptr == NULL);
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001518 ptr[0] = __alloc_percpu(mibsize, align);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001519 if (!ptr[0])
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001520 return -ENOMEM;
1521#if SNMP_ARRAY_SZ == 2
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001522 ptr[1] = __alloc_percpu(mibsize, align);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001523 if (!ptr[1]) {
1524 free_percpu(ptr[0]);
1525 ptr[0] = NULL;
1526 return -ENOMEM;
1527 }
1528#endif
Herbert Xu5e0f0432007-04-24 21:53:35 -07001529 return 0;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001530}
1531EXPORT_SYMBOL_GPL(snmp_mib_init);
1532
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001533void snmp_mib_free(void __percpu *ptr[SNMP_ARRAY_SZ])
Herbert Xu5e0f0432007-04-24 21:53:35 -07001534{
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001535 int i;
1536
Herbert Xu5e0f0432007-04-24 21:53:35 -07001537 BUG_ON(ptr == NULL);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001538 for (i = 0; i < SNMP_ARRAY_SZ; i++) {
1539 free_percpu(ptr[i]);
1540 ptr[i] = NULL;
1541 }
Herbert Xu5e0f0432007-04-24 21:53:35 -07001542}
1543EXPORT_SYMBOL_GPL(snmp_mib_free);
1544
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001546static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001548 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549};
1550#endif
1551
Alexey Dobriyan32613092009-09-14 12:21:47 +00001552static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001553 .early_demux = tcp_v4_early_demux,
1554 .handler = tcp_v4_rcv,
1555 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001556 .no_policy = 1,
1557 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558};
1559
Vlad Yasevichde27d002012-11-15 08:49:12 +00001560static const struct net_offload tcp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001561 .callbacks = {
1562 .gso_send_check = tcp_v4_gso_send_check,
1563 .gso_segment = tcp_tso_segment,
1564 .gro_receive = tcp4_gro_receive,
1565 .gro_complete = tcp4_gro_complete,
1566 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001567};
1568
Alexey Dobriyan32613092009-09-14 12:21:47 +00001569static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 .handler = udp_rcv,
1571 .err_handler = udp_err,
1572 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001573 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574};
1575
Vlad Yasevichde27d002012-11-15 08:49:12 +00001576static const struct net_offload udp_offload = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001577 .callbacks = {
1578 .gso_send_check = udp4_ufo_send_check,
1579 .gso_segment = udp4_ufo_fragment,
1580 },
Vlad Yasevichde27d002012-11-15 08:49:12 +00001581};
1582
Alexey Dobriyan32613092009-09-14 12:21:47 +00001583static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001585 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001586 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001587 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588};
1589
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001590static __net_init int ipv4_mib_init_net(struct net *net)
1591{
Tejun Heo7d720c32010-02-16 15:20:26 +00001592 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001593 sizeof(struct tcp_mib),
1594 __alignof__(struct tcp_mib)) < 0)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001595 goto err_tcp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001596 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001597 sizeof(struct ipstats_mib),
1598 __alignof__(struct ipstats_mib)) < 0)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001599 goto err_ip_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001600 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001601 sizeof(struct linux_mib),
1602 __alignof__(struct linux_mib)) < 0)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001603 goto err_net_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001604 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001605 sizeof(struct udp_mib),
1606 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001607 goto err_udp_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001608 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001609 sizeof(struct udp_mib),
1610 __alignof__(struct udp_mib)) < 0)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001611 goto err_udplite_mib;
Tejun Heo7d720c32010-02-16 15:20:26 +00001612 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
Eric Dumazet1823e4c82010-06-22 20:58:41 +00001613 sizeof(struct icmp_mib),
1614 __alignof__(struct icmp_mib)) < 0)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001615 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001616 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1617 GFP_KERNEL);
1618 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001619 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001620
1621 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001622 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001623
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001624err_icmpmsg_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001625 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001626err_icmp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001627 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001628err_udplite_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001629 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001630err_udp_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001631 snmp_mib_free((void __percpu **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001632err_net_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001633 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001634err_ip_mib:
Tejun Heo7d720c32010-02-16 15:20:26 +00001635 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001636err_tcp_mib:
1637 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001638}
1639
1640static __net_exit void ipv4_mib_exit_net(struct net *net)
1641{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001642 kfree(net->mib.icmpmsg_statistics);
Tejun Heo7d720c32010-02-16 15:20:26 +00001643 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1644 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1645 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1646 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1647 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1648 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001649}
1650
1651static __net_initdata struct pernet_operations ipv4_mib_ops = {
1652 .init = ipv4_mib_init_net,
1653 .exit = ipv4_mib_exit_net,
1654};
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656static int __init init_ipv4_mibs(void)
1657{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001658 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659}
1660
1661static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662
Herbert Xu30e224d2005-07-05 14:40:10 -07001663/*
1664 * IP protocol layer initialiser
1665 */
1666
Vlad Yasevich22061d82012-11-15 08:49:11 +00001667static struct packet_offload ip_packet_offload __read_mostly = {
1668 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001669 .callbacks = {
1670 .gso_send_check = inet_gso_send_check,
1671 .gso_segment = inet_gso_segment,
1672 .gro_receive = inet_gro_receive,
1673 .gro_complete = inet_gro_complete,
1674 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001675};
1676
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001677static int __init ipv4_offload_init(void)
1678{
1679 /*
1680 * Add offloads
1681 */
1682 if (inet_add_offload(&udp_offload, IPPROTO_UDP) < 0)
1683 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
1684 if (inet_add_offload(&tcp_offload, IPPROTO_TCP) < 0)
1685 pr_crit("%s: Cannot add TCP protocol offlaod\n", __func__);
1686
1687 dev_add_offload(&ip_packet_offload);
1688 return 0;
1689}
1690
1691fs_initcall(ipv4_offload_init);
1692
1693static struct packet_type ip_packet_type __read_mostly = {
1694 .type = cpu_to_be16(ETH_P_IP),
1695 .func = ip_rcv,
1696};
1697
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698static int __init inet_init(void)
1699{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 struct inet_protosw *q;
1701 struct list_head *r;
1702 int rc = -EINVAL;
1703
YOSHIFUJI Hideaki / 吉藤英明95c7e0e2013-01-09 07:19:37 +00001704 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Amerigo Wange3826f12010-05-05 00:27:06 +00001706 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1707 if (!sysctl_local_reserved_ports)
1708 goto out;
1709
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 rc = proto_register(&tcp_prot, 1);
1711 if (rc)
Amerigo Wange3826f12010-05-05 00:27:06 +00001712 goto out_free_reserved_ports;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
1714 rc = proto_register(&udp_prot, 1);
1715 if (rc)
1716 goto out_unregister_tcp_proto;
1717
1718 rc = proto_register(&raw_prot, 1);
1719 if (rc)
1720 goto out_unregister_udp_proto;
1721
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001722 rc = proto_register(&ping_prot, 1);
1723 if (rc)
1724 goto out_unregister_raw_proto;
1725
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001727 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 */
1729
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001730 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Al Virobd7b1532008-07-15 16:00:59 -04001732#ifdef CONFIG_SYSCTL
1733 ip_static_sysctl_init();
1734#endif
1735
Glauber Costa3dc43e32011-12-11 21:47:05 +00001736 tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1737
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 /*
1739 * Add all the base protocols.
1740 */
1741
1742 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001743 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001745 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001747 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748#ifdef CONFIG_IP_MULTICAST
1749 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001750 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751#endif
1752
1753 /* Register the socket-side information for inet_create. */
1754 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1755 INIT_LIST_HEAD(r);
1756
1757 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1758 inet_register_protosw(q);
1759
1760 /*
1761 * Set the ARP module up
1762 */
1763
1764 arp_init();
1765
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001766 /*
1767 * Set the IP module up
1768 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
1770 ip_init();
1771
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001772 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
1774 /* Setup TCP slab cache for open requests. */
1775 tcp_init();
1776
Hideo Aoki95766ff2007-12-31 00:29:24 -08001777 /* Setup UDP memory threshold */
1778 udp_init();
1779
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001780 /* Add UDP-Lite (RFC 3828) */
1781 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001783 ping_init();
1784
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 /*
1786 * Set the ICMP layer up
1787 */
1788
Denis V. Luneva5710d62008-02-29 11:14:50 -08001789 if (icmp_init() < 0)
1790 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 /*
1793 * Initialise the multicast router
1794 */
1795#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001796 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001797 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798#endif
1799 /*
1800 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001801 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
Stephen Hemminger132adf52007-03-08 20:44:43 -08001803 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001804 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001805
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 ipv4_proc_init();
1807
1808 ipfrag_init();
1809
Herbert Xu30e224d2005-07-05 14:40:10 -07001810 dev_add_pack(&ip_packet_type);
1811
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 rc = 0;
1813out:
1814 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001815out_unregister_raw_proto:
1816 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817out_unregister_udp_proto:
1818 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001819out_unregister_tcp_proto:
1820 proto_unregister(&tcp_prot);
Amerigo Wange3826f12010-05-05 00:27:06 +00001821out_free_reserved_ports:
1822 kfree(sysctl_local_reserved_ports);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 goto out;
1824}
1825
Heiko Carstensa536e072006-04-28 15:19:17 -07001826fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827
1828/* ------------------------------------------------------------------------ */
1829
1830#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831static int __init ipv4_proc_init(void)
1832{
1833 int rc = 0;
1834
1835 if (raw_proc_init())
1836 goto out_raw;
1837 if (tcp4_proc_init())
1838 goto out_tcp;
1839 if (udp4_proc_init())
1840 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001841 if (ping_proc_init())
1842 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 if (ip_misc_proc_init())
1844 goto out_misc;
1845out:
1846 return rc;
1847out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001848 ping_proc_exit();
1849out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 udp4_proc_exit();
1851out_udp:
1852 tcp4_proc_exit();
1853out_tcp:
1854 raw_proc_exit();
1855out_raw:
1856 rc = -ENOMEM;
1857 goto out;
1858}
1859
1860#else /* CONFIG_PROC_FS */
1861static int __init ipv4_proc_init(void)
1862{
1863 return 0;
1864}
1865#endif /* CONFIG_PROC_FS */
1866
1867MODULE_ALIAS_NETPROTO(PF_INET);
1868