blob: 9532ee87151f5d184205eafc70729fc543a9eb82 [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
129/* New destruction routine */
130
131void inet_sock_destruct(struct sock *sk)
132{
133 struct inet_sock *inet = inet_sk(sk);
134
135 __skb_queue_purge(&sk->sk_receive_queue);
136 __skb_queue_purge(&sk->sk_error_queue);
137
Hideo Aoki95766ff2007-12-31 00:29:24 -0800138 sk_mem_reclaim(sk);
139
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700141 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 sk->sk_state, sk);
143 return;
144 }
145 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700146 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 return;
148 }
149
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700150 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
151 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
152 WARN_ON(sk->sk_wmem_queued);
153 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000155 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000156 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700157 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700158 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700160EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
162/*
163 * The routines beyond this point handle the behaviour of an AF_INET
164 * socket object. Mostly it punts to the subprotocols of IP to do
165 * the work.
166 */
167
168/*
169 * Automatically bind an unbound socket.
170 */
171
172static int inet_autobind(struct sock *sk)
173{
174 struct inet_sock *inet;
175 /* We may need to bind the socket. */
176 lock_sock(sk);
177 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000178 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 if (sk->sk_prot->get_port(sk, 0)) {
180 release_sock(sk);
181 return -EAGAIN;
182 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000183 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 }
185 release_sock(sk);
186 return 0;
187}
188
189/*
190 * Move a socket into listening state.
191 */
192int inet_listen(struct socket *sock, int backlog)
193{
194 struct sock *sk = sock->sk;
195 unsigned char old_state;
196 int err;
197
198 lock_sock(sk);
199
200 err = -EINVAL;
201 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
202 goto out;
203
204 old_state = sk->sk_state;
205 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
206 goto out;
207
208 /* Really, if the socket is already in listen state
209 * we can only allow the backlog to be adjusted.
210 */
211 if (old_state != TCP_LISTEN) {
Jerry Chu83368862012-08-31 12:29:12 +0000212 /* Check special setups for testing purpose to enable TFO w/o
213 * requiring TCP_FASTOPEN sockopt.
214 * Note that only TCP sockets (SOCK_STREAM) will reach here.
215 * Also fastopenq may already been allocated because this
216 * socket was in TCP_LISTEN state previously but was
217 * shutdown() (rather than close()).
218 */
219 if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
Ian Morris51456b22015-04-03 09:17:26 +0100220 !inet_csk(sk)->icsk_accept_queue.fastopenq) {
Jerry Chu83368862012-08-31 12:29:12 +0000221 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
222 err = fastopen_init_queue(sk, backlog);
223 else if ((sysctl_tcp_fastopen &
224 TFO_SERVER_WO_SOCKOPT2) != 0)
225 err = fastopen_init_queue(sk,
226 ((uint)sysctl_tcp_fastopen) >> 16);
227 else
228 err = 0;
229 if (err)
230 goto out;
Christoph Paaschdfea2aa2015-06-18 09:15:34 -0700231
232 tcp_fastopen_init_key_once(true);
Jerry Chu83368862012-08-31 12:29:12 +0000233 }
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800234 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 if (err)
236 goto out;
237 }
238 sk->sk_max_ack_backlog = backlog;
239 err = 0;
240
241out:
242 release_sock(sk);
243 return err;
244}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700245EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247/*
248 * Create an inet socket.
249 */
250
Eric Paris3f378b62009-11-05 22:18:14 -0800251static int inet_create(struct net *net, struct socket *sock, int protocol,
252 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
254 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 struct inet_protosw *answer;
256 struct inet_sock *inet;
257 struct proto *answer_prot;
258 unsigned char answer_flags;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700259 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800260 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
262 sock->state = SS_UNCONNECTED;
263
264 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700265lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800266 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700268 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700270 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 /* Check the non-wild match. */
272 if (protocol == answer->protocol) {
273 if (protocol != IPPROTO_IP)
274 break;
275 } else {
276 /* Check for the two wild cases. */
277 if (IPPROTO_IP == protocol) {
278 protocol = answer->protocol;
279 break;
280 }
281 if (IPPROTO_IP == answer->protocol)
282 break;
283 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800284 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 }
286
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700287 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700288 if (try_loading_module < 2) {
289 rcu_read_unlock();
290 /*
291 * Be more specific, e.g. net-pf-2-proto-132-type-1
292 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
293 */
294 if (++try_loading_module == 1)
295 request_module("net-pf-%d-proto-%d-type-%d",
296 PF_INET, protocol, sock->type);
297 /*
298 * Fall back to generic, e.g. net-pf-2-proto-132
299 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
300 */
301 else
302 request_module("net-pf-%d-proto-%d",
303 PF_INET, protocol);
304 goto lookup_protocol;
305 } else
306 goto out_rcu_unlock;
307 }
308
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000310 if (sock->type == SOCK_RAW && !kern &&
311 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314 sock->ops = answer->ops;
315 answer_prot = answer->prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 answer_flags = answer->flags;
317 rcu_read_unlock();
318
Ian Morris51456b22015-04-03 09:17:26 +0100319 WARN_ON(!answer_prot->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
321 err = -ENOBUFS;
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500322 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
Ian Morris51456b22015-04-03 09:17:26 +0100323 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 goto out;
325
326 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000328 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
330 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800331 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000333 inet->nodefrag = 0;
334
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000336 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 if (IPPROTO_RAW == protocol)
338 inet->hdrincl = 1;
339 }
340
Hannes Frederic Sowa974eda12013-12-14 05:13:38 +0100341 if (net->ipv4.sysctl_ip_no_pmtu_disc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 inet->pmtudisc = IP_PMTUDISC_DONT;
343 else
344 inet->pmtudisc = IP_PMTUDISC_WANT;
345
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000346 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
348 sock_init_data(sock, sk);
349
350 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 sk->sk_protocol = protocol;
352 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
353
354 inet->uc_ttl = -1;
355 inet->mc_loop = 1;
356 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000357 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 inet->mc_index = 0;
359 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000360 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700362 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000364 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 /* It assumes that any protocol which allows
366 * the user to assign a number at socket
367 * creation time automatically
368 * shares.
369 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000370 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 /* Add to protocol hash chains. */
372 sk->sk_prot->hash(sk);
373 }
374
375 if (sk->sk_prot->init) {
376 err = sk->sk_prot->init(sk);
377 if (err)
378 sk_common_release(sk);
379 }
380out:
381 return err;
382out_rcu_unlock:
383 rcu_read_unlock();
384 goto out;
385}
386
387
388/*
389 * The peer socket should always be NULL (or else). When we call this
390 * function we are destroying the object and from then on nobody
391 * should refer to it.
392 */
393int inet_release(struct socket *sock)
394{
395 struct sock *sk = sock->sk;
396
397 if (sk) {
398 long timeout;
399
400 /* Applications forget to leave groups before exiting */
401 ip_mc_drop_socket(sk);
402
403 /* If linger is set, we don't return until the close
404 * is complete. Otherwise we return immediately. The
405 * actually closing is done the same either way.
406 *
407 * If the close is due to the process exiting, we never
408 * linger..
409 */
410 timeout = 0;
411 if (sock_flag(sk, SOCK_LINGER) &&
412 !(current->flags & PF_EXITING))
413 timeout = sk->sk_lingertime;
414 sock->sk = NULL;
415 sk->sk_prot->close(sk, timeout);
416 }
417 return 0;
418}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700419EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
422{
423 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
424 struct sock *sk = sock->sk;
425 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000426 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 unsigned short snum;
428 int chk_addr_ret;
429 int err;
430
431 /* If the socket has its own bind function then use it. (RAW) */
432 if (sk->sk_prot->bind) {
433 err = sk->sk_prot->bind(sk, uaddr, addr_len);
434 goto out;
435 }
436 err = -EINVAL;
437 if (addr_len < sizeof(struct sockaddr_in))
438 goto out;
439
Marcus Meissnerc349a522011-07-04 01:30:29 +0000440 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400441 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
442 * only if s_addr is INADDR_ANY.
443 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000444 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400445 if (addr->sin_family != AF_UNSPEC ||
446 addr->sin_addr.s_addr != htonl(INADDR_ANY))
447 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000448 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700449
Eric W. Biederman35946982012-11-16 03:03:12 +0000450 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
452 /* Not specified by any standard per-se, however it breaks too
453 * many applications when removed. It is unfortunate since
454 * allowing applications to make a non-local bind solves
455 * several problems with systems using dynamic addressing.
456 * (ie. your servers still start up even if your ISDN link
457 * is temporarily down)
458 */
459 err = -EADDRNOTAVAIL;
Vincent Bernat49a60152014-09-05 15:09:03 +0200460 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700461 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700462 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 chk_addr_ret != RTN_LOCAL &&
464 chk_addr_ret != RTN_MULTICAST &&
465 chk_addr_ret != RTN_BROADCAST)
466 goto out;
467
468 snum = ntohs(addr->sin_port);
469 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000470 if (snum && snum < PROT_SOCK &&
471 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 goto out;
473
474 /* We keep a pair of addresses. rcv_saddr is the one
475 * used by hash lookups, and saddr is used for transmit.
476 *
477 * In the BSD API these are the same except where it
478 * would be illegal to use them (multicast/broadcast) in
479 * which case the sending device address is used.
480 */
481 lock_sock(sk);
482
483 /* Check these errors (active socket, double bind). */
484 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000485 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 goto out_release_sock;
487
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000488 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000490 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 /* Make sure we are allowed to bind here. */
Eric Dumazet90c337d2015-06-06 21:17:57 -0700493 if ((snum || !inet->bind_address_no_port) &&
494 sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000495 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 err = -EADDRINUSE;
497 goto out_release_sock;
498 }
499
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000500 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
502 if (snum)
503 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000504 inet->inet_sport = htons(inet->inet_num);
505 inet->inet_daddr = 0;
506 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 sk_dst_reset(sk);
508 err = 0;
509out_release_sock:
510 release_sock(sk);
511out:
512 return err;
513}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700514EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000516int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 int addr_len, int flags)
518{
519 struct sock *sk = sock->sk;
520
Changli Gao6503d962010-03-31 22:58:26 +0000521 if (addr_len < sizeof(uaddr->sa_family))
522 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 if (uaddr->sa_family == AF_UNSPEC)
524 return sk->sk_prot->disconnect(sk, flags);
525
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000526 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000528 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700530EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Yuchung Cheng783237e2012-07-19 06:43:07 +0000532static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533{
534 DEFINE_WAIT(wait);
535
Eric Dumazetaa395142010-04-20 13:03:51 +0000536 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000537 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
539 /* Basic assumption: if someone sets sk->sk_err, he _must_
540 * change state of the socket from TCP_SYN_*.
541 * Connect() does not allow to get error notifications
542 * without closing the socket.
543 */
544 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
545 release_sock(sk);
546 timeo = schedule_timeout(timeo);
547 lock_sock(sk);
548 if (signal_pending(current) || !timeo)
549 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000550 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000552 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000553 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 return timeo;
555}
556
557/*
558 * Connect to a remote host. There is regrettably still a little
559 * TCP 'magic' in here.
560 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000561int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
562 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 struct sock *sk = sock->sk;
565 int err;
566 long timeo;
567
Changli Gao6503d962010-03-31 22:58:26 +0000568 if (addr_len < sizeof(uaddr->sa_family))
569 return -EINVAL;
570
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 if (uaddr->sa_family == AF_UNSPEC) {
572 err = sk->sk_prot->disconnect(sk, flags);
573 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
574 goto out;
575 }
576
577 switch (sock->state) {
578 default:
579 err = -EINVAL;
580 goto out;
581 case SS_CONNECTED:
582 err = -EISCONN;
583 goto out;
584 case SS_CONNECTING:
585 err = -EALREADY;
586 /* Fall out of switch with err, set for this state */
587 break;
588 case SS_UNCONNECTED:
589 err = -EISCONN;
590 if (sk->sk_state != TCP_CLOSE)
591 goto out;
592
593 err = sk->sk_prot->connect(sk, uaddr, addr_len);
594 if (err < 0)
595 goto out;
596
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900597 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
599 /* Just entered SS_CONNECTING state; the only
600 * difference is that return value in non-blocking
601 * case is EINPROGRESS, rather than EALREADY.
602 */
603 err = -EINPROGRESS;
604 break;
605 }
606
607 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
608
609 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000610 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
611 tcp_sk(sk)->fastopen_req &&
612 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
613
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000615 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 goto out;
617
618 err = sock_intr_errno(timeo);
619 if (signal_pending(current))
620 goto out;
621 }
622
623 /* Connection was closed by RST, timeout, ICMP error
624 * or another process disconnected us.
625 */
626 if (sk->sk_state == TCP_CLOSE)
627 goto sock_error;
628
629 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
630 * and error was received after socket entered established state.
631 * Hence, it is handled normally after connect() return successfully.
632 */
633
634 sock->state = SS_CONNECTED;
635 err = 0;
636out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 return err;
638
639sock_error:
640 err = sock_error(sk) ? : -ECONNABORTED;
641 sock->state = SS_UNCONNECTED;
642 if (sk->sk_prot->disconnect(sk, flags))
643 sock->state = SS_DISCONNECTING;
644 goto out;
645}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000646EXPORT_SYMBOL(__inet_stream_connect);
647
648int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
649 int addr_len, int flags)
650{
651 int err;
652
653 lock_sock(sock->sk);
654 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
655 release_sock(sock->sk);
656 return err;
657}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700658EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
660/*
661 * Accept a pending connection. The TCP layer now gives BSD semantics.
662 */
663
664int inet_accept(struct socket *sock, struct socket *newsock, int flags)
665{
666 struct sock *sk1 = sock->sk;
667 int err = -EINVAL;
668 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
669
670 if (!sk2)
671 goto do_err;
672
673 lock_sock(sk2);
674
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000675 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700676 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000677 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
678 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
680 sock_graft(sk2, newsock);
681
682 newsock->state = SS_CONNECTED;
683 err = 0;
684 release_sock(sk2);
685do_err:
686 return err;
687}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700688EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
690
691/*
692 * This does both peername and sockname.
693 */
694int inet_getname(struct socket *sock, struct sockaddr *uaddr,
695 int *uaddr_len, int peer)
696{
697 struct sock *sk = sock->sk;
698 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700699 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
701 sin->sin_family = AF_INET;
702 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000703 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
705 peer == 1))
706 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000707 sin->sin_port = inet->inet_dport;
708 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000710 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000712 addr = inet->inet_saddr;
713 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 sin->sin_addr.s_addr = addr;
715 }
716 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
717 *uaddr_len = sizeof(*sin);
718 return 0;
719}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700720EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Ying Xue1b784142015-03-02 15:37:48 +0800722int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723{
724 struct sock *sk = sock->sk;
725
David S. Millerc58dc012010-04-27 15:05:31 -0700726 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700727
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000729 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
730 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 return -EAGAIN;
732
Ying Xue1b784142015-03-02 15:37:48 +0800733 return sk->sk_prot->sendmsg(sk, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700735EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
Changli Gao7ba42912010-07-10 20:41:55 +0000737ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
738 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
740 struct sock *sk = sock->sk;
741
David S. Millerc58dc012010-04-27 15:05:31 -0700742 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000745 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
746 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 return -EAGAIN;
748
749 if (sk->sk_prot->sendpage)
750 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
751 return sock_no_sendpage(sock, page, offset, size, flags);
752}
Changli Gao7ba42912010-07-10 20:41:55 +0000753EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Ying Xue1b784142015-03-02 15:37:48 +0800755int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
756 int flags)
Tom Herbertfec5e652010-04-16 16:01:27 -0700757{
758 struct sock *sk = sock->sk;
759 int addr_len = 0;
760 int err;
761
David S. Millerc58dc012010-04-27 15:05:31 -0700762 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700763
Ying Xue1b784142015-03-02 15:37:48 +0800764 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Tom Herbertfec5e652010-04-16 16:01:27 -0700765 flags & ~MSG_DONTWAIT, &addr_len);
766 if (err >= 0)
767 msg->msg_namelen = addr_len;
768 return err;
769}
770EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
772int inet_shutdown(struct socket *sock, int how)
773{
774 struct sock *sk = sock->sk;
775 int err = 0;
776
777 /* This should really check to make sure
778 * the socket is a TCP socket. (WHY AC...)
779 */
780 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
781 1->2 bit 2 snds.
782 2->3 */
783 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
784 return -EINVAL;
785
786 lock_sock(sk);
787 if (sock->state == SS_CONNECTING) {
788 if ((1 << sk->sk_state) &
789 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
790 sock->state = SS_DISCONNECTING;
791 else
792 sock->state = SS_CONNECTED;
793 }
794
795 switch (sk->sk_state) {
796 case TCP_CLOSE:
797 err = -ENOTCONN;
798 /* Hack to wake up other listeners, who can poll for
799 POLLHUP, even on eg. unconnected UDP sockets -- RR */
800 default:
801 sk->sk_shutdown |= how;
802 if (sk->sk_prot->shutdown)
803 sk->sk_prot->shutdown(sk, how);
804 break;
805
806 /* Remaining two branches are temporary solution for missing
807 * close() in multithreaded environment. It is _not_ a good idea,
808 * but we have no choice until close() is repaired at VFS level.
809 */
810 case TCP_LISTEN:
811 if (!(how & RCV_SHUTDOWN))
812 break;
813 /* Fall through */
814 case TCP_SYN_SENT:
815 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
816 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
817 break;
818 }
819
820 /* Wake up anyone sleeping in poll. */
821 sk->sk_state_change(sk);
822 release_sock(sk);
823 return err;
824}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700825EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
827/*
828 * ioctl() calls you can issue on an INET socket. Most of these are
829 * device configuration and stuff and very rarely used. Some ioctls
830 * pass on to the socket itself.
831 *
832 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
833 * loads the devconfigure module does its configuring and unloads it.
834 * There's a good 20K of config code hanging around the kernel.
835 */
836
837int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
838{
839 struct sock *sk = sock->sk;
840 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900841 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842
843 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700844 case SIOCGSTAMP:
845 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
846 break;
847 case SIOCGSTAMPNS:
848 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
849 break;
850 case SIOCADDRT:
851 case SIOCDELRT:
852 case SIOCRTMSG:
853 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
854 break;
855 case SIOCDARP:
856 case SIOCGARP:
857 case SIOCSARP:
858 err = arp_ioctl(net, cmd, (void __user *)arg);
859 break;
860 case SIOCGIFADDR:
861 case SIOCSIFADDR:
862 case SIOCGIFBRDADDR:
863 case SIOCSIFBRDADDR:
864 case SIOCGIFNETMASK:
865 case SIOCSIFNETMASK:
866 case SIOCGIFDSTADDR:
867 case SIOCSIFDSTADDR:
868 case SIOCSIFPFLAGS:
869 case SIOCGIFPFLAGS:
870 case SIOCSIFFLAGS:
871 err = devinet_ioctl(net, cmd, (void __user *)arg);
872 break;
873 default:
874 if (sk->sk_prot->ioctl)
875 err = sk->sk_prot->ioctl(sk, cmd, arg);
876 else
877 err = -ENOIOCTLCMD;
878 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 }
880 return err;
881}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700882EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000884#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000885static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000886{
887 struct sock *sk = sock->sk;
888 int err = -ENOIOCTLCMD;
889
890 if (sk->sk_prot->compat_ioctl)
891 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
892
893 return err;
894}
895#endif
896
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800897const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800898 .family = PF_INET,
899 .owner = THIS_MODULE,
900 .release = inet_release,
901 .bind = inet_bind,
902 .connect = inet_stream_connect,
903 .socketpair = sock_no_socketpair,
904 .accept = inet_accept,
905 .getname = inet_getname,
906 .poll = tcp_poll,
907 .ioctl = inet_ioctl,
908 .listen = inet_listen,
909 .shutdown = inet_shutdown,
910 .setsockopt = sock_common_setsockopt,
911 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000912 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700913 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800914 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000915 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800916 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800917#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800918 .compat_setsockopt = compat_sock_common_setsockopt,
919 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000920 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800921#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700923EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800925const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800926 .family = PF_INET,
927 .owner = THIS_MODULE,
928 .release = inet_release,
929 .bind = inet_bind,
930 .connect = inet_dgram_connect,
931 .socketpair = sock_no_socketpair,
932 .accept = sock_no_accept,
933 .getname = inet_getname,
934 .poll = udp_poll,
935 .ioctl = inet_ioctl,
936 .listen = sock_no_listen,
937 .shutdown = inet_shutdown,
938 .setsockopt = sock_common_setsockopt,
939 .getsockopt = sock_common_getsockopt,
940 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700941 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800942 .mmap = sock_no_mmap,
943 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800944#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800945 .compat_setsockopt = compat_sock_common_setsockopt,
946 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000947 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800948#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700950EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
952/*
953 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
954 * udp_poll
955 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800956static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800957 .family = PF_INET,
958 .owner = THIS_MODULE,
959 .release = inet_release,
960 .bind = inet_bind,
961 .connect = inet_dgram_connect,
962 .socketpair = sock_no_socketpair,
963 .accept = sock_no_accept,
964 .getname = inet_getname,
965 .poll = datagram_poll,
966 .ioctl = inet_ioctl,
967 .listen = sock_no_listen,
968 .shutdown = inet_shutdown,
969 .setsockopt = sock_common_setsockopt,
970 .getsockopt = sock_common_getsockopt,
971 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700972 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800973 .mmap = sock_no_mmap,
974 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800975#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800976 .compat_setsockopt = compat_sock_common_setsockopt,
977 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000978 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800979#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980};
981
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +0000982static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 .family = PF_INET,
984 .create = inet_create,
985 .owner = THIS_MODULE,
986};
987
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988/* Upon startup we insert all the elements in inetsw_array[] into
989 * the linked list inetsw.
990 */
991static struct inet_protosw inetsw_array[] =
992{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900993 {
994 .type = SOCK_STREAM,
995 .protocol = IPPROTO_TCP,
996 .prot = &tcp_prot,
997 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900998 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800999 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001000 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001002 {
1003 .type = SOCK_DGRAM,
1004 .protocol = IPPROTO_UDP,
1005 .prot = &udp_prot,
1006 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001007 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001009
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001010 {
1011 .type = SOCK_DGRAM,
1012 .protocol = IPPROTO_ICMP,
1013 .prot = &ping_prot,
1014 .ops = &inet_dgram_ops,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001015 .flags = INET_PROTOSW_REUSE,
1016 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
1018 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001019 .type = SOCK_RAW,
1020 .protocol = IPPROTO_IP, /* wild card */
1021 .prot = &raw_prot,
1022 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001023 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 }
1025};
1026
Denis Chengc40f6ff2007-09-16 16:39:25 -07001027#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
1029void inet_register_protosw(struct inet_protosw *p)
1030{
1031 struct list_head *lh;
1032 struct inet_protosw *answer;
1033 int protocol = p->protocol;
1034 struct list_head *last_perm;
1035
1036 spin_lock_bh(&inetsw_lock);
1037
1038 if (p->type >= SOCK_MAX)
1039 goto out_illegal;
1040
1041 /* If we are trying to override a permanent protocol, bail. */
1042 answer = NULL;
1043 last_perm = &inetsw[p->type];
1044 list_for_each(lh, &inetsw[p->type]) {
1045 answer = list_entry(lh, struct inet_protosw, list);
1046
1047 /* Check only the non-wild match. */
1048 if (INET_PROTOSW_PERMANENT & answer->flags) {
1049 if (protocol == answer->protocol)
1050 break;
1051 last_perm = lh;
1052 }
1053
1054 answer = NULL;
1055 }
1056 if (answer)
1057 goto out_permanent;
1058
1059 /* Add the new entry after the last permanent entry if any, so that
1060 * the new entry does not override a permanent entry when matched with
1061 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001062 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 * system automatically returns to the old behavior.
1064 */
1065 list_add_rcu(&p->list, last_perm);
1066out:
1067 spin_unlock_bh(&inetsw_lock);
1068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 return;
1070
1071out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001072 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 goto out;
1074
1075out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001076 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 p->type);
1078 goto out;
1079}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001080EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
1082void inet_unregister_protosw(struct inet_protosw *p)
1083{
1084 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001085 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 p->protocol);
1087 } else {
1088 spin_lock_bh(&inetsw_lock);
1089 list_del_rcu(&p->list);
1090 spin_unlock_bh(&inetsw_lock);
1091
1092 synchronize_net();
1093 }
1094}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001095EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001097/*
1098 * Shall we try to damage output packets if routing dev changes?
1099 */
1100
Brian Haleyab32ea52006-09-22 14:15:41 -07001101int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001102
1103static int inet_sk_reselect_saddr(struct sock *sk)
1104{
1105 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001106 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001107 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001108 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001109 struct rtable *rt;
1110 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001111 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001112
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001113 inet_opt = rcu_dereference_protected(inet->inet_opt,
1114 sock_owned_by_user(sk));
1115 if (inet_opt && inet_opt->opt.srr)
1116 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001117
1118 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001119 fl4 = &inet->cork.fl.u.ip4;
1120 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001121 sk->sk_bound_dev_if, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001122 inet->inet_sport, inet->inet_dport, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001123 if (IS_ERR(rt))
1124 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001125
Changli Gaod8d1f302010-06-10 23:31:35 -07001126 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001127
David S. Miller6e869132011-05-06 16:18:04 -07001128 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001129
1130 if (new_saddr == old_saddr)
1131 return 0;
1132
1133 if (sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001134 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1135 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001136 }
1137
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001138 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001139
1140 /*
1141 * XXX The only one ugly spot where we need to
1142 * XXX really change the sockets identity after
1143 * XXX it has entered the hashes. -DaveM
1144 *
1145 * Besides that, it does not check for connection
1146 * uniqueness. Wait for troubles.
1147 */
1148 __sk_prot_rehash(sk);
1149 return 0;
1150}
1151
1152int inet_sk_rebuild_header(struct sock *sk)
1153{
1154 struct inet_sock *inet = inet_sk(sk);
1155 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001156 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001157 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001158 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001159 int err;
1160
1161 /* Route is OK, nothing to do. */
1162 if (rt)
1163 return 0;
1164
1165 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001166 rcu_read_lock();
1167 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001168 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001169 if (inet_opt && inet_opt->opt.srr)
1170 daddr = inet_opt->opt.faddr;
1171 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001172 fl4 = &inet->cork.fl.u.ip4;
1173 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001174 inet->inet_dport, inet->inet_sport,
1175 sk->sk_protocol, RT_CONN_FLAGS(sk),
1176 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001177 if (!IS_ERR(rt)) {
1178 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001179 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001180 } else {
1181 err = PTR_ERR(rt);
1182
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001183 /* Routing failed... */
1184 sk->sk_route_caps = 0;
1185 /*
1186 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1187 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1188 */
1189 if (!sysctl_ip_dynaddr ||
1190 sk->sk_state != TCP_SYN_SENT ||
1191 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1192 (err = inet_sk_reselect_saddr(sk)) != 0)
1193 sk->sk_err_soft = -err;
1194 }
1195
1196 return err;
1197}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001198EXPORT_SYMBOL(inet_sk_rebuild_header);
1199
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001200static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
Eric Dumazet3347c962013-10-19 11:42:56 -07001201 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001202{
1203 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001204 const struct net_offload *ops;
Eric Dumazet3347c962013-10-19 11:42:56 -07001205 unsigned int offset = 0;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001206 bool udpfrag, encap;
David S. Millerf9242b62012-06-19 18:56:21 -07001207 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001208 int proto;
Eric Dumazet3347c962013-10-19 11:42:56 -07001209 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001210 int ihl;
1211 int id;
1212
Herbert Xubbcf4672006-07-03 19:38:35 -07001213 if (unlikely(skb_shinfo(skb)->gso_type &
1214 ~(SKB_GSO_TCPV4 |
1215 SKB_GSO_UDP |
1216 SKB_GSO_DODGY |
1217 SKB_GSO_TCP_ECN |
Pravin B Shelar68c33162013-02-14 14:02:41 +00001218 SKB_GSO_GRE |
Tom Herbert4749c092014-06-04 17:20:23 -07001219 SKB_GSO_GRE_CSUM |
Eric Dumazetcb32f512013-10-19 11:42:57 -07001220 SKB_GSO_IPIP |
Eric Dumazet61c1db72013-10-20 20:47:30 -07001221 SKB_GSO_SIT |
Pravin B Shelar9b3eb5e2013-05-02 16:14:19 +00001222 SKB_GSO_TCPV6 |
Pravin B Shelar73136262013-03-07 13:21:51 +00001223 SKB_GSO_UDP_TUNNEL |
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07001224 SKB_GSO_UDP_TUNNEL_CSUM |
Tom Herberte585f232014-11-04 09:06:54 -08001225 SKB_GSO_TUNNEL_REMCSUM |
Herbert Xubbcf4672006-07-03 19:38:35 -07001226 0)))
1227 goto out;
1228
Eric Dumazet3347c962013-10-19 11:42:56 -07001229 skb_reset_network_header(skb);
1230 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001231 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001232 goto out;
1233
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001234 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001235 ihl = iph->ihl * 4;
1236 if (ihl < sizeof(*iph))
1237 goto out;
1238
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001239 id = ntohs(iph->id);
1240 proto = iph->protocol;
1241
1242 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001243 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001244 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001245 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001246
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001247 encap = SKB_GSO_CB(skb)->encap_level > 0;
1248 if (encap)
Florian Westphal1e16aa32014-10-20 13:49:16 +02001249 features &= skb->dev->hw_enc_features;
Eric Dumazet3347c962013-10-19 11:42:56 -07001250 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001251
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001252 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001253
Herbert Xuf4c50d92006-06-22 03:02:40 -07001254 segs = ERR_PTR(-EPROTONOSUPPORT);
1255
Hannes Frederic Sowa91a48a22014-02-24 00:48:05 +01001256 if (skb->encapsulation &&
1257 skb_shinfo(skb)->gso_type & (SKB_GSO_SIT|SKB_GSO_IPIP))
1258 udpfrag = proto == IPPROTO_UDP && encap;
1259 else
1260 udpfrag = proto == IPPROTO_UDP && !skb->encapsulation;
Eric Dumazetdcd60772013-11-07 18:32:06 -08001261
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001262 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001263 if (likely(ops && ops->callbacks.gso_segment))
1264 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001265
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001266 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001267 goto out;
1268
1269 skb = segs;
1270 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001271 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001272 if (udpfrag) {
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001273 iph->id = htons(id);
1274 iph->frag_off = htons(offset >> 3);
Ian Morris00db4122015-04-03 09:17:27 +01001275 if (skb->next)
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001276 iph->frag_off |= htons(IP_MF);
Eric Dumazet3347c962013-10-19 11:42:56 -07001277 offset += skb->len - nhoff - ihl;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001278 } else {
Pravin B Shelar25c77042013-03-24 17:36:29 +00001279 iph->id = htons(id++);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001280 }
Eric Dumazet3347c962013-10-19 11:42:56 -07001281 iph->tot_len = htons(skb->len - nhoff);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001282 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001283 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001284 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001285 skb->network_header = (u8 *)iph - skb->head;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001286 } while ((skb = skb->next));
1287
1288out:
1289 return segs;
1290}
1291
Herbert Xu73cc19f2008-12-15 23:41:09 -08001292static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1293 struct sk_buff *skb)
1294{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001295 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001296 struct sk_buff **pp = NULL;
1297 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001298 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001299 unsigned int hlen;
1300 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001301 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001302 int flush = 1;
1303 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001304
Herbert Xua5b1cf22009-05-26 18:50:28 +00001305 off = skb_gro_offset(skb);
1306 hlen = off + sizeof(*iph);
1307 iph = skb_gro_header_fast(skb, off);
1308 if (skb_gro_header_hard(skb, hlen)) {
1309 iph = skb_gro_header_slow(skb, hlen, off);
1310 if (unlikely(!iph))
1311 goto out;
1312 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001313
David S. Millerf9242b62012-06-19 18:56:21 -07001314 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001315
1316 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001317 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001318 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001319 goto out_unlock;
1320
Herbert Xua5ad24b2009-02-08 18:00:39 +00001321 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001322 goto out_unlock;
1323
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001324 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001325 goto out_unlock;
1326
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001327 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001328 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001329 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001330
1331 for (p = *head; p; p = p->next) {
1332 struct iphdr *iph2;
1333
1334 if (!NAPI_GRO_CB(p)->same_flow)
1335 continue;
1336
Jerry Chu299603e82013-12-11 20:53:45 -08001337 iph2 = (struct iphdr *)(p->data + off);
1338 /* The above works because, with the exception of the top
1339 * (inner most) layer, we only aggregate pkts with the same
1340 * hdr length so all the hdrs we'll need to verify will start
1341 * at the same offset.
1342 */
Herbert Xua5ad24b2009-02-08 18:00:39 +00001343 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001344 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1345 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001346 NAPI_GRO_CB(p)->same_flow = 0;
1347 continue;
1348 }
1349
1350 /* All fields must match except length and checksum. */
1351 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001352 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001353 (iph->tos ^ iph2->tos) |
Jerry Chubf5a7552014-01-07 10:23:19 -08001354 ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001355
Jerry Chubf5a7552014-01-07 10:23:19 -08001356 /* Save the IP ID check to be included later when we get to
1357 * the transport layer so only the inner most IP ID is checked.
1358 * This is because some GSO/TSO implementations do not
1359 * correctly increment the IP ID for the outer hdrs.
1360 */
1361 NAPI_GRO_CB(p)->flush_id =
1362 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001363 NAPI_GRO_CB(p)->flush |= flush;
1364 }
1365
1366 NAPI_GRO_CB(skb)->flush |= flush;
Jerry Chu299603e82013-12-11 20:53:45 -08001367 skb_set_network_header(skb, off);
1368 /* The above will be needed by the transport layer if there is one
1369 * immediately following this IP hdr.
1370 */
1371
Eric Dumazet2c804d02014-09-30 22:12:05 -07001372 /* Note : No need to call skb_gro_postpull_rcsum() here,
1373 * as we already checked checksum over ipv4 header was 0
1374 */
Herbert Xu86911732009-01-29 14:19:50 +00001375 skb_gro_pull(skb, sizeof(*iph));
1376 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001377
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001378 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001379
1380out_unlock:
1381 rcu_read_unlock();
1382
1383out:
1384 NAPI_GRO_CB(skb)->flush |= flush;
1385
1386 return pp;
1387}
1388
Willem de Bruijnf4713a32014-11-26 14:53:02 -05001389int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1390{
1391 if (sk->sk_family == AF_INET)
1392 return ip_recv_error(sk, msg, len, addr_len);
1393#if IS_ENABLED(CONFIG_IPV6)
1394 if (sk->sk_family == AF_INET6)
1395 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1396#endif
1397 return -EINVAL;
1398}
1399
Jerry Chu299603e82013-12-11 20:53:45 -08001400static int inet_gro_complete(struct sk_buff *skb, int nhoff)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001401{
Jerry Chu299603e82013-12-11 20:53:45 -08001402 __be16 newlen = htons(skb->len - nhoff);
1403 struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001404 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001405 int proto = iph->protocol;
1406 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001407
Jerry Chuc3caf112014-07-14 15:54:46 -07001408 if (skb->encapsulation)
1409 skb_set_inner_network_header(skb, nhoff);
1410
Herbert Xu73cc19f2008-12-15 23:41:09 -08001411 csum_replace2(&iph->check, iph->tot_len, newlen);
1412 iph->tot_len = newlen;
1413
1414 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001415 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001416 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001417 goto out_unlock;
1418
Jerry Chu299603e82013-12-11 20:53:45 -08001419 /* Only need to add sizeof(*iph) to get to the next hdr below
1420 * because any hdr with option will have been flushed in
1421 * inet_gro_receive().
1422 */
1423 err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001424
1425out_unlock:
1426 rcu_read_unlock();
1427
1428 return err;
1429}
1430
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001431int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001432 unsigned short type, unsigned char protocol,
1433 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001434{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001435 struct socket *sock;
Eric W. Biederman26abe142015-05-08 21:10:31 -05001436 int rc = sock_create_kern(net, family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001437
1438 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001439 *sk = sock->sk;
1440 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001441 /*
1442 * Unhash it so that IP input processing does not even see it,
1443 * we do not wish this socket to see incoming packets.
1444 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001445 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001446 }
1447 return rc;
1448}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001449EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1450
WANG Cong698365f2014-05-05 15:55:55 -07001451unsigned long snmp_fold_field(void __percpu *mib, int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001452{
1453 unsigned long res = 0;
WANG Cong698365f2014-05-05 15:55:55 -07001454 int i;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001455
WANG Cong698365f2014-05-05 15:55:55 -07001456 for_each_possible_cpu(i)
1457 res += *(((unsigned long *) per_cpu_ptr(mib, i)) + offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001458 return res;
1459}
1460EXPORT_SYMBOL_GPL(snmp_fold_field);
1461
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001462#if BITS_PER_LONG==32
1463
WANG Cong698365f2014-05-05 15:55:55 -07001464u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001465{
1466 u64 res = 0;
1467 int cpu;
1468
1469 for_each_possible_cpu(cpu) {
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001470 void *bhptr;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001471 struct u64_stats_sync *syncp;
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001472 u64 v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001473 unsigned int start;
1474
WANG Cong698365f2014-05-05 15:55:55 -07001475 bhptr = per_cpu_ptr(mib, cpu);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001476 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1477 do {
Eric W. Biederman57a77442014-03-13 21:26:42 -07001478 start = u64_stats_fetch_begin_irq(syncp);
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001479 v = *(((u64 *) bhptr) + offt);
Eric W. Biederman57a77442014-03-13 21:26:42 -07001480 } while (u64_stats_fetch_retry_irq(syncp, start));
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001481
Eric Dumazet8f0ea0f2011-06-10 19:45:51 +00001482 res += v;
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001483 }
1484 return res;
1485}
1486EXPORT_SYMBOL_GPL(snmp_fold_field64);
1487#endif
1488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001490static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001492 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493};
1494#endif
1495
Alexey Dobriyan32613092009-09-14 12:21:47 +00001496static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001497 .early_demux = tcp_v4_early_demux,
1498 .handler = tcp_v4_rcv,
1499 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001500 .no_policy = 1,
1501 .netns_ok = 1,
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +01001502 .icmp_strict_tag_validation = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503};
1504
Alexey Dobriyan32613092009-09-14 12:21:47 +00001505static const struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001506 .early_demux = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 .handler = udp_rcv,
1508 .err_handler = udp_err,
1509 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001510 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511};
1512
Alexey Dobriyan32613092009-09-14 12:21:47 +00001513static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001515 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001516 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001517 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518};
1519
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001520static __net_init int ipv4_mib_init_net(struct net *net)
1521{
John Stultz827da442013-10-07 15:51:58 -07001522 int i;
1523
WANG Cong698365f2014-05-05 15:55:55 -07001524 net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1525 if (!net->mib.tcp_statistics)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001526 goto err_tcp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001527 net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1528 if (!net->mib.ip_statistics)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001529 goto err_ip_mib;
John Stultz827da442013-10-07 15:51:58 -07001530
1531 for_each_possible_cpu(i) {
1532 struct ipstats_mib *af_inet_stats;
WANG Cong698365f2014-05-05 15:55:55 -07001533 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
John Stultz827da442013-10-07 15:51:58 -07001534 u64_stats_init(&af_inet_stats->syncp);
John Stultz827da442013-10-07 15:51:58 -07001535 }
1536
WANG Cong698365f2014-05-05 15:55:55 -07001537 net->mib.net_statistics = alloc_percpu(struct linux_mib);
1538 if (!net->mib.net_statistics)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001539 goto err_net_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001540 net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1541 if (!net->mib.udp_statistics)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001542 goto err_udp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001543 net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1544 if (!net->mib.udplite_statistics)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001545 goto err_udplite_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001546 net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1547 if (!net->mib.icmp_statistics)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001548 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001549 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1550 GFP_KERNEL);
1551 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001552 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001553
1554 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001555 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001556
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001557err_icmpmsg_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001558 free_percpu(net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001559err_icmp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001560 free_percpu(net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001561err_udplite_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001562 free_percpu(net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001563err_udp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001564 free_percpu(net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001565err_net_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001566 free_percpu(net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001567err_ip_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001568 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001569err_tcp_mib:
1570 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001571}
1572
1573static __net_exit void ipv4_mib_exit_net(struct net *net)
1574{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001575 kfree(net->mib.icmpmsg_statistics);
WANG Cong698365f2014-05-05 15:55:55 -07001576 free_percpu(net->mib.icmp_statistics);
1577 free_percpu(net->mib.udplite_statistics);
1578 free_percpu(net->mib.udp_statistics);
1579 free_percpu(net->mib.net_statistics);
1580 free_percpu(net->mib.ip_statistics);
1581 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001582}
1583
1584static __net_initdata struct pernet_operations ipv4_mib_ops = {
1585 .init = ipv4_mib_init_net,
1586 .exit = ipv4_mib_exit_net,
1587};
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589static int __init init_ipv4_mibs(void)
1590{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001591 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592}
1593
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001594static __net_init int inet_init_net(struct net *net)
1595{
1596 /*
1597 * Set defaults for local port range
1598 */
1599 seqlock_init(&net->ipv4.ip_local_ports.lock);
1600 net->ipv4.ip_local_ports.range[0] = 32768;
Eric Dumazet07f4c902015-05-24 14:49:35 -07001601 net->ipv4.ip_local_ports.range[1] = 60999;
Cong Wangba6b9182014-05-06 11:02:50 -07001602
1603 seqlock_init(&net->ipv4.ping_group_range.lock);
1604 /*
1605 * Sane defaults - nobody may create ping sockets.
1606 * Boot scripts should set this to distro-specific group.
1607 */
1608 net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1609 net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001610 return 0;
1611}
1612
1613static __net_exit void inet_exit_net(struct net *net)
1614{
1615}
1616
1617static __net_initdata struct pernet_operations af_inet_ops = {
1618 .init = inet_init_net,
1619 .exit = inet_exit_net,
1620};
1621
1622static int __init init_inet_pernet_ops(void)
1623{
1624 return register_pernet_subsys(&af_inet_ops);
1625}
1626
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628
Herbert Xu30e224d2005-07-05 14:40:10 -07001629/*
1630 * IP protocol layer initialiser
1631 */
1632
Vlad Yasevich22061d82012-11-15 08:49:11 +00001633static struct packet_offload ip_packet_offload __read_mostly = {
1634 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001635 .callbacks = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001636 .gso_segment = inet_gso_segment,
1637 .gro_receive = inet_gro_receive,
1638 .gro_complete = inet_gro_complete,
1639 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001640};
1641
Eric Dumazetcb32f512013-10-19 11:42:57 -07001642static const struct net_offload ipip_offload = {
1643 .callbacks = {
Eric Dumazetcb32f512013-10-19 11:42:57 -07001644 .gso_segment = inet_gso_segment,
Tom Herbert9667e9b2014-09-09 11:23:15 -07001645 .gro_receive = inet_gro_receive,
1646 .gro_complete = inet_gro_complete,
Eric Dumazetcb32f512013-10-19 11:42:57 -07001647 },
1648};
1649
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001650static int __init ipv4_offload_init(void)
1651{
1652 /*
1653 * Add offloads
1654 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001655 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001656 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001657 if (tcpv4_offload_init() < 0)
1658 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001659
1660 dev_add_offload(&ip_packet_offload);
Eric Dumazetcb32f512013-10-19 11:42:57 -07001661 inet_add_offload(&ipip_offload, IPPROTO_IPIP);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001662 return 0;
1663}
1664
1665fs_initcall(ipv4_offload_init);
1666
1667static struct packet_type ip_packet_type __read_mostly = {
1668 .type = cpu_to_be16(ETH_P_IP),
1669 .func = ip_rcv,
1670};
1671
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672static int __init inet_init(void)
1673{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 struct inet_protosw *q;
1675 struct list_head *r;
1676 int rc = -EINVAL;
1677
Eyal Birgerb4772ef2015-03-01 14:58:29 +02001678 sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
1680 rc = proto_register(&tcp_prot, 1);
1681 if (rc)
WANG Cong122ff242014-05-12 16:04:53 -07001682 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683
1684 rc = proto_register(&udp_prot, 1);
1685 if (rc)
1686 goto out_unregister_tcp_proto;
1687
1688 rc = proto_register(&raw_prot, 1);
1689 if (rc)
1690 goto out_unregister_udp_proto;
1691
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001692 rc = proto_register(&ping_prot, 1);
1693 if (rc)
1694 goto out_unregister_raw_proto;
1695
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001697 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 */
1699
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001700 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
Al Virobd7b1532008-07-15 16:00:59 -04001702#ifdef CONFIG_SYSCTL
1703 ip_static_sysctl_init();
1704#endif
1705
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 /*
1707 * Add all the base protocols.
1708 */
1709
1710 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001711 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001713 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001715 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716#ifdef CONFIG_IP_MULTICAST
1717 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001718 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719#endif
1720
1721 /* Register the socket-side information for inet_create. */
1722 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1723 INIT_LIST_HEAD(r);
1724
1725 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1726 inet_register_protosw(q);
1727
1728 /*
1729 * Set the ARP module up
1730 */
1731
1732 arp_init();
1733
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001734 /*
1735 * Set the IP module up
1736 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
1738 ip_init();
1739
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001740 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741
1742 /* Setup TCP slab cache for open requests. */
1743 tcp_init();
1744
Hideo Aoki95766ff2007-12-31 00:29:24 -08001745 /* Setup UDP memory threshold */
1746 udp_init();
1747
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001748 /* Add UDP-Lite (RFC 3828) */
1749 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001751 ping_init();
1752
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 /*
1754 * Set the ICMP layer up
1755 */
1756
Denis V. Luneva5710d62008-02-29 11:14:50 -08001757 if (icmp_init() < 0)
1758 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
1760 /*
1761 * Initialise the multicast router
1762 */
1763#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001764 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001765 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766#endif
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001767
1768 if (init_inet_pernet_ops())
1769 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 /*
1771 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001772 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
Stephen Hemminger132adf52007-03-08 20:44:43 -08001774 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001775 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001776
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 ipv4_proc_init();
1778
1779 ipfrag_init();
1780
Herbert Xu30e224d2005-07-05 14:40:10 -07001781 dev_add_pack(&ip_packet_type);
1782
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 rc = 0;
1784out:
1785 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001786out_unregister_raw_proto:
1787 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788out_unregister_udp_proto:
1789 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001790out_unregister_tcp_proto:
1791 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 goto out;
1793}
1794
Heiko Carstensa536e072006-04-28 15:19:17 -07001795fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
1797/* ------------------------------------------------------------------------ */
1798
1799#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800static int __init ipv4_proc_init(void)
1801{
1802 int rc = 0;
1803
1804 if (raw_proc_init())
1805 goto out_raw;
1806 if (tcp4_proc_init())
1807 goto out_tcp;
1808 if (udp4_proc_init())
1809 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001810 if (ping_proc_init())
1811 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 if (ip_misc_proc_init())
1813 goto out_misc;
1814out:
1815 return rc;
1816out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001817 ping_proc_exit();
1818out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 udp4_proc_exit();
1820out_udp:
1821 tcp4_proc_exit();
1822out_tcp:
1823 raw_proc_exit();
1824out_raw:
1825 rc = -ENOMEM;
1826 goto out;
1827}
1828
1829#else /* CONFIG_PROC_FS */
1830static int __init ipv4_proc_init(void)
1831{
1832 return 0;
1833}
1834#endif /* CONFIG_PROC_FS */
1835
1836MODULE_ALIAS_NETPROTO(PF_INET);
1837