blob: 8217cd22f921d1eef025f2cda12bcb7f4b751384 [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>
Thomas Graf045a0fa2015-07-23 10:08:44 +0200115#include <net/ip_tunnels.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700117#include <net/net_namespace.h>
Eric Dumazetaebda152013-04-29 05:58:52 +0000118#include <net/secure_seq.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119#ifdef CONFIG_IP_MROUTE
120#include <linux/mroute.h>
121#endif
David Ahern3236b002015-09-29 20:07:14 -0700122#include <net/l3mdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125/* The inetsw table contains everything that inet_create needs to
126 * build a new socket.
127 */
128static struct list_head inetsw[SOCK_MAX];
129static DEFINE_SPINLOCK(inetsw_lock);
130
131/* New destruction routine */
132
133void inet_sock_destruct(struct sock *sk)
134{
135 struct inet_sock *inet = inet_sk(sk);
136
137 __skb_queue_purge(&sk->sk_receive_queue);
138 __skb_queue_purge(&sk->sk_error_queue);
139
Hideo Aoki95766ff2007-12-31 00:29:24 -0800140 sk_mem_reclaim(sk);
141
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700143 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 sk->sk_state, sk);
145 return;
146 }
147 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700148 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 return;
150 }
151
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700152 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
153 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
154 WARN_ON(sk->sk_wmem_queued);
155 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000157 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000158 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700159 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700160 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700162EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163
164/*
165 * The routines beyond this point handle the behaviour of an AF_INET
166 * socket object. Mostly it punts to the subprotocols of IP to do
167 * the work.
168 */
169
170/*
171 * Automatically bind an unbound socket.
172 */
173
174static int inet_autobind(struct sock *sk)
175{
176 struct inet_sock *inet;
177 /* We may need to bind the socket. */
178 lock_sock(sk);
179 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000180 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 if (sk->sk_prot->get_port(sk, 0)) {
182 release_sock(sk);
183 return -EAGAIN;
184 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000185 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 }
187 release_sock(sk);
188 return 0;
189}
190
191/*
192 * Move a socket into listening state.
193 */
194int inet_listen(struct socket *sock, int backlog)
195{
196 struct sock *sk = sock->sk;
197 unsigned char old_state;
198 int err;
199
200 lock_sock(sk);
201
202 err = -EINVAL;
203 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
204 goto out;
205
206 old_state = sk->sk_state;
207 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
208 goto out;
209
210 /* Really, if the socket is already in listen state
211 * we can only allow the backlog to be adjusted.
212 */
213 if (old_state != TCP_LISTEN) {
Jerry Chu83368862012-08-31 12:29:12 +0000214 /* Check special setups for testing purpose to enable TFO w/o
215 * requiring TCP_FASTOPEN sockopt.
216 * Note that only TCP sockets (SOCK_STREAM) will reach here.
217 * Also fastopenq may already been allocated because this
218 * socket was in TCP_LISTEN state previously but was
219 * shutdown() (rather than close()).
220 */
221 if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700222 !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
Jerry Chu83368862012-08-31 12:29:12 +0000223 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700224 fastopen_queue_tune(sk, backlog);
Jerry Chu83368862012-08-31 12:29:12 +0000225 else if ((sysctl_tcp_fastopen &
226 TFO_SERVER_WO_SOCKOPT2) != 0)
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700227 fastopen_queue_tune(sk,
Jerry Chu83368862012-08-31 12:29:12 +0000228 ((uint)sysctl_tcp_fastopen) >> 16);
Christoph Paaschdfea2aa2015-06-18 09:15:34 -0700229
230 tcp_fastopen_init_key_once(true);
Jerry Chu83368862012-08-31 12:29:12 +0000231 }
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800232 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 if (err)
234 goto out;
235 }
236 sk->sk_max_ack_backlog = backlog;
237 err = 0;
238
239out:
240 release_sock(sk);
241 return err;
242}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700243EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245/*
246 * Create an inet socket.
247 */
248
Eric Paris3f378b62009-11-05 22:18:14 -0800249static int inet_create(struct net *net, struct socket *sock, int protocol,
250 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251{
252 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 struct inet_protosw *answer;
254 struct inet_sock *inet;
255 struct proto *answer_prot;
256 unsigned char answer_flags;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700257 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800258 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259
Hannes Frederic Sowa79462ad02015-12-14 22:03:39 +0100260 if (protocol < 0 || protocol >= IPPROTO_MAX)
261 return -EINVAL;
262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 sock->state = SS_UNCONNECTED;
264
265 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700266lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800267 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700269 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700271 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 /* Check the non-wild match. */
273 if (protocol == answer->protocol) {
274 if (protocol != IPPROTO_IP)
275 break;
276 } else {
277 /* Check for the two wild cases. */
278 if (IPPROTO_IP == protocol) {
279 protocol = answer->protocol;
280 break;
281 }
282 if (IPPROTO_IP == answer->protocol)
283 break;
284 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800285 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 }
287
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700288 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700289 if (try_loading_module < 2) {
290 rcu_read_unlock();
291 /*
292 * Be more specific, e.g. net-pf-2-proto-132-type-1
293 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
294 */
295 if (++try_loading_module == 1)
296 request_module("net-pf-%d-proto-%d-type-%d",
297 PF_INET, protocol, sock->type);
298 /*
299 * Fall back to generic, e.g. net-pf-2-proto-132
300 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
301 */
302 else
303 request_module("net-pf-%d-proto-%d",
304 PF_INET, protocol);
305 goto lookup_protocol;
306 } else
307 goto out_rcu_unlock;
308 }
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000311 if (sock->type == SOCK_RAW && !kern &&
312 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
315 sock->ops = answer->ops;
316 answer_prot = answer->prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 answer_flags = answer->flags;
318 rcu_read_unlock();
319
Ian Morris51456b22015-04-03 09:17:26 +0100320 WARN_ON(!answer_prot->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
322 err = -ENOBUFS;
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500323 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
Ian Morris51456b22015-04-03 09:17:26 +0100324 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 goto out;
326
327 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000329 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
331 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800332 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000334 inet->nodefrag = 0;
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000337 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 if (IPPROTO_RAW == protocol)
339 inet->hdrincl = 1;
340 }
341
Hannes Frederic Sowa974eda12013-12-14 05:13:38 +0100342 if (net->ipv4.sysctl_ip_no_pmtu_disc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 inet->pmtudisc = IP_PMTUDISC_DONT;
344 else
345 inet->pmtudisc = IP_PMTUDISC_WANT;
346
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000347 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 sock_init_data(sock, sk);
350
351 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 sk->sk_protocol = protocol;
353 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
354
355 inet->uc_ttl = -1;
356 inet->mc_loop = 1;
357 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000358 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 inet->mc_index = 0;
360 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000361 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700363 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000365 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 /* It assumes that any protocol which allows
367 * the user to assign a number at socket
368 * creation time automatically
369 * shares.
370 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000371 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 /* Add to protocol hash chains. */
Craig Gallek086c6532016-02-10 11:50:35 -0500373 err = sk->sk_prot->hash(sk);
374 if (err) {
375 sk_common_release(sk);
376 goto out;
377 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 }
379
380 if (sk->sk_prot->init) {
381 err = sk->sk_prot->init(sk);
382 if (err)
383 sk_common_release(sk);
384 }
385out:
386 return err;
387out_rcu_unlock:
388 rcu_read_unlock();
389 goto out;
390}
391
392
393/*
394 * The peer socket should always be NULL (or else). When we call this
395 * function we are destroying the object and from then on nobody
396 * should refer to it.
397 */
398int inet_release(struct socket *sock)
399{
400 struct sock *sk = sock->sk;
401
402 if (sk) {
403 long timeout;
404
405 /* Applications forget to leave groups before exiting */
406 ip_mc_drop_socket(sk);
407
408 /* If linger is set, we don't return until the close
409 * is complete. Otherwise we return immediately. The
410 * actually closing is done the same either way.
411 *
412 * If the close is due to the process exiting, we never
413 * linger..
414 */
415 timeout = 0;
416 if (sock_flag(sk, SOCK_LINGER) &&
417 !(current->flags & PF_EXITING))
418 timeout = sk->sk_lingertime;
419 sock->sk = NULL;
420 sk->sk_prot->close(sk, timeout);
421 }
422 return 0;
423}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700424EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
427{
428 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
429 struct sock *sk = sock->sk;
430 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000431 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 unsigned short snum;
433 int chk_addr_ret;
David Ahern9b8ff512015-09-01 14:26:35 -0600434 u32 tb_id = RT_TABLE_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 int err;
436
437 /* If the socket has its own bind function then use it. (RAW) */
438 if (sk->sk_prot->bind) {
439 err = sk->sk_prot->bind(sk, uaddr, addr_len);
440 goto out;
441 }
442 err = -EINVAL;
443 if (addr_len < sizeof(struct sockaddr_in))
444 goto out;
445
Marcus Meissnerc349a522011-07-04 01:30:29 +0000446 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400447 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
448 * only if s_addr is INADDR_ANY.
449 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000450 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400451 if (addr->sin_family != AF_UNSPEC ||
452 addr->sin_addr.s_addr != htonl(INADDR_ANY))
453 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000454 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700455
David Ahern3236b002015-09-29 20:07:14 -0700456 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
David Ahern30bbaa12015-08-13 14:59:05 -0600457 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458
459 /* Not specified by any standard per-se, however it breaks too
460 * many applications when removed. It is unfortunate since
461 * allowing applications to make a non-local bind solves
462 * several problems with systems using dynamic addressing.
463 * (ie. your servers still start up even if your ISDN link
464 * is temporarily down)
465 */
466 err = -EADDRNOTAVAIL;
Vincent Bernat49a60152014-09-05 15:09:03 +0200467 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700468 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700469 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 chk_addr_ret != RTN_LOCAL &&
471 chk_addr_ret != RTN_MULTICAST &&
472 chk_addr_ret != RTN_BROADCAST)
473 goto out;
474
475 snum = ntohs(addr->sin_port);
476 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000477 if (snum && snum < PROT_SOCK &&
478 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 goto out;
480
481 /* We keep a pair of addresses. rcv_saddr is the one
482 * used by hash lookups, and saddr is used for transmit.
483 *
484 * In the BSD API these are the same except where it
485 * would be illegal to use them (multicast/broadcast) in
486 * which case the sending device address is used.
487 */
488 lock_sock(sk);
489
490 /* Check these errors (active socket, double bind). */
491 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000492 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 goto out_release_sock;
494
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000495 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000497 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
499 /* Make sure we are allowed to bind here. */
Eric Dumazet90c337d2015-06-06 21:17:57 -0700500 if ((snum || !inet->bind_address_no_port) &&
501 sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000502 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 err = -EADDRINUSE;
504 goto out_release_sock;
505 }
506
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000507 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
509 if (snum)
510 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000511 inet->inet_sport = htons(inet->inet_num);
512 inet->inet_daddr = 0;
513 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 sk_dst_reset(sk);
515 err = 0;
516out_release_sock:
517 release_sock(sk);
518out:
519 return err;
520}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700521EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000523int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 int addr_len, int flags)
525{
526 struct sock *sk = sock->sk;
527
Changli Gao6503d962010-03-31 22:58:26 +0000528 if (addr_len < sizeof(uaddr->sa_family))
529 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 if (uaddr->sa_family == AF_UNSPEC)
531 return sk->sk_prot->disconnect(sk, flags);
532
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000533 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000535 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700537EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Yuchung Cheng783237e2012-07-19 06:43:07 +0000539static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
541 DEFINE_WAIT(wait);
542
Eric Dumazetaa395142010-04-20 13:03:51 +0000543 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000544 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 /* Basic assumption: if someone sets sk->sk_err, he _must_
547 * change state of the socket from TCP_SYN_*.
548 * Connect() does not allow to get error notifications
549 * without closing the socket.
550 */
551 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
552 release_sock(sk);
553 timeo = schedule_timeout(timeo);
554 lock_sock(sk);
555 if (signal_pending(current) || !timeo)
556 break;
Eric Dumazetaa395142010-04-20 13:03:51 +0000557 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000559 finish_wait(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000560 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return timeo;
562}
563
564/*
565 * Connect to a remote host. There is regrettably still a little
566 * TCP 'magic' in here.
567 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000568int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
569 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570{
571 struct sock *sk = sock->sk;
572 int err;
573 long timeo;
574
Changli Gao6503d962010-03-31 22:58:26 +0000575 if (addr_len < sizeof(uaddr->sa_family))
576 return -EINVAL;
577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 if (uaddr->sa_family == AF_UNSPEC) {
579 err = sk->sk_prot->disconnect(sk, flags);
580 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
581 goto out;
582 }
583
584 switch (sock->state) {
585 default:
586 err = -EINVAL;
587 goto out;
588 case SS_CONNECTED:
589 err = -EISCONN;
590 goto out;
591 case SS_CONNECTING:
592 err = -EALREADY;
593 /* Fall out of switch with err, set for this state */
594 break;
595 case SS_UNCONNECTED:
596 err = -EISCONN;
597 if (sk->sk_state != TCP_CLOSE)
598 goto out;
599
600 err = sk->sk_prot->connect(sk, uaddr, addr_len);
601 if (err < 0)
602 goto out;
603
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900604 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
606 /* Just entered SS_CONNECTING state; the only
607 * difference is that return value in non-blocking
608 * case is EINPROGRESS, rather than EALREADY.
609 */
610 err = -EINPROGRESS;
611 break;
612 }
613
614 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
615
616 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000617 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
618 tcp_sk(sk)->fastopen_req &&
619 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000622 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 goto out;
624
625 err = sock_intr_errno(timeo);
626 if (signal_pending(current))
627 goto out;
628 }
629
630 /* Connection was closed by RST, timeout, ICMP error
631 * or another process disconnected us.
632 */
633 if (sk->sk_state == TCP_CLOSE)
634 goto sock_error;
635
636 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
637 * and error was received after socket entered established state.
638 * Hence, it is handled normally after connect() return successfully.
639 */
640
641 sock->state = SS_CONNECTED;
642 err = 0;
643out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 return err;
645
646sock_error:
647 err = sock_error(sk) ? : -ECONNABORTED;
648 sock->state = SS_UNCONNECTED;
649 if (sk->sk_prot->disconnect(sk, flags))
650 sock->state = SS_DISCONNECTING;
651 goto out;
652}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000653EXPORT_SYMBOL(__inet_stream_connect);
654
655int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
656 int addr_len, int flags)
657{
658 int err;
659
660 lock_sock(sock->sk);
661 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
662 release_sock(sock->sk);
663 return err;
664}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700665EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
667/*
668 * Accept a pending connection. The TCP layer now gives BSD semantics.
669 */
670
671int inet_accept(struct socket *sock, struct socket *newsock, int flags)
672{
673 struct sock *sk1 = sock->sk;
674 int err = -EINVAL;
675 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
676
677 if (!sk2)
678 goto do_err;
679
680 lock_sock(sk2);
681
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000682 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700683 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000684 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
685 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
687 sock_graft(sk2, newsock);
688
689 newsock->state = SS_CONNECTED;
690 err = 0;
691 release_sock(sk2);
692do_err:
693 return err;
694}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700695EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
697
698/*
699 * This does both peername and sockname.
700 */
701int inet_getname(struct socket *sock, struct sockaddr *uaddr,
702 int *uaddr_len, int peer)
703{
704 struct sock *sk = sock->sk;
705 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700706 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
708 sin->sin_family = AF_INET;
709 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000710 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
712 peer == 1))
713 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000714 sin->sin_port = inet->inet_dport;
715 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000717 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000719 addr = inet->inet_saddr;
720 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 sin->sin_addr.s_addr = addr;
722 }
723 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
724 *uaddr_len = sizeof(*sin);
725 return 0;
726}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700727EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
Ying Xue1b784142015-03-02 15:37:48 +0800729int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730{
731 struct sock *sk = sock->sk;
732
David S. Millerc58dc012010-04-27 15:05:31 -0700733 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700734
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000736 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
737 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 return -EAGAIN;
739
Ying Xue1b784142015-03-02 15:37:48 +0800740 return sk->sk_prot->sendmsg(sk, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700742EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Changli Gao7ba42912010-07-10 20:41:55 +0000744ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
745 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746{
747 struct sock *sk = sock->sk;
748
David S. Millerc58dc012010-04-27 15:05:31 -0700749 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700750
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000752 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
753 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 return -EAGAIN;
755
756 if (sk->sk_prot->sendpage)
757 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
758 return sock_no_sendpage(sock, page, offset, size, flags);
759}
Changli Gao7ba42912010-07-10 20:41:55 +0000760EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Ying Xue1b784142015-03-02 15:37:48 +0800762int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
763 int flags)
Tom Herbertfec5e652010-04-16 16:01:27 -0700764{
765 struct sock *sk = sock->sk;
766 int addr_len = 0;
767 int err;
768
David S. Millerc58dc012010-04-27 15:05:31 -0700769 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700770
Ying Xue1b784142015-03-02 15:37:48 +0800771 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Tom Herbertfec5e652010-04-16 16:01:27 -0700772 flags & ~MSG_DONTWAIT, &addr_len);
773 if (err >= 0)
774 msg->msg_namelen = addr_len;
775 return err;
776}
777EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779int inet_shutdown(struct socket *sock, int how)
780{
781 struct sock *sk = sock->sk;
782 int err = 0;
783
784 /* This should really check to make sure
785 * the socket is a TCP socket. (WHY AC...)
786 */
787 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
788 1->2 bit 2 snds.
789 2->3 */
790 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
791 return -EINVAL;
792
793 lock_sock(sk);
794 if (sock->state == SS_CONNECTING) {
795 if ((1 << sk->sk_state) &
796 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
797 sock->state = SS_DISCONNECTING;
798 else
799 sock->state = SS_CONNECTED;
800 }
801
802 switch (sk->sk_state) {
803 case TCP_CLOSE:
804 err = -ENOTCONN;
805 /* Hack to wake up other listeners, who can poll for
806 POLLHUP, even on eg. unconnected UDP sockets -- RR */
807 default:
808 sk->sk_shutdown |= how;
809 if (sk->sk_prot->shutdown)
810 sk->sk_prot->shutdown(sk, how);
811 break;
812
813 /* Remaining two branches are temporary solution for missing
814 * close() in multithreaded environment. It is _not_ a good idea,
815 * but we have no choice until close() is repaired at VFS level.
816 */
817 case TCP_LISTEN:
818 if (!(how & RCV_SHUTDOWN))
819 break;
820 /* Fall through */
821 case TCP_SYN_SENT:
822 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
823 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
824 break;
825 }
826
827 /* Wake up anyone sleeping in poll. */
828 sk->sk_state_change(sk);
829 release_sock(sk);
830 return err;
831}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700832EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
834/*
835 * ioctl() calls you can issue on an INET socket. Most of these are
836 * device configuration and stuff and very rarely used. Some ioctls
837 * pass on to the socket itself.
838 *
839 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
840 * loads the devconfigure module does its configuring and unloads it.
841 * There's a good 20K of config code hanging around the kernel.
842 */
843
844int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
845{
846 struct sock *sk = sock->sk;
847 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900848 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
850 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700851 case SIOCGSTAMP:
852 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
853 break;
854 case SIOCGSTAMPNS:
855 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
856 break;
857 case SIOCADDRT:
858 case SIOCDELRT:
859 case SIOCRTMSG:
860 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
861 break;
862 case SIOCDARP:
863 case SIOCGARP:
864 case SIOCSARP:
865 err = arp_ioctl(net, cmd, (void __user *)arg);
866 break;
867 case SIOCGIFADDR:
868 case SIOCSIFADDR:
869 case SIOCGIFBRDADDR:
870 case SIOCSIFBRDADDR:
871 case SIOCGIFNETMASK:
872 case SIOCSIFNETMASK:
873 case SIOCGIFDSTADDR:
874 case SIOCSIFDSTADDR:
875 case SIOCSIFPFLAGS:
876 case SIOCGIFPFLAGS:
877 case SIOCSIFFLAGS:
878 err = devinet_ioctl(net, cmd, (void __user *)arg);
879 break;
880 default:
881 if (sk->sk_prot->ioctl)
882 err = sk->sk_prot->ioctl(sk, cmd, arg);
883 else
884 err = -ENOIOCTLCMD;
885 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 }
887 return err;
888}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700889EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000891#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000892static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000893{
894 struct sock *sk = sock->sk;
895 int err = -ENOIOCTLCMD;
896
897 if (sk->sk_prot->compat_ioctl)
898 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
899
900 return err;
901}
902#endif
903
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800904const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800905 .family = PF_INET,
906 .owner = THIS_MODULE,
907 .release = inet_release,
908 .bind = inet_bind,
909 .connect = inet_stream_connect,
910 .socketpair = sock_no_socketpair,
911 .accept = inet_accept,
912 .getname = inet_getname,
913 .poll = tcp_poll,
914 .ioctl = inet_ioctl,
915 .listen = inet_listen,
916 .shutdown = inet_shutdown,
917 .setsockopt = sock_common_setsockopt,
918 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000919 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700920 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800921 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000922 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800923 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800924#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800925 .compat_setsockopt = compat_sock_common_setsockopt,
926 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000927 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800928#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700930EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800932const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800933 .family = PF_INET,
934 .owner = THIS_MODULE,
935 .release = inet_release,
936 .bind = inet_bind,
937 .connect = inet_dgram_connect,
938 .socketpair = sock_no_socketpair,
939 .accept = sock_no_accept,
940 .getname = inet_getname,
941 .poll = udp_poll,
942 .ioctl = inet_ioctl,
943 .listen = sock_no_listen,
944 .shutdown = inet_shutdown,
945 .setsockopt = sock_common_setsockopt,
946 .getsockopt = sock_common_getsockopt,
947 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700948 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800949 .mmap = sock_no_mmap,
950 .sendpage = inet_sendpage,
samanthakumar627d2d62016-04-05 12:41:16 -0400951 .set_peek_off = sk_set_peek_off,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800952#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800953 .compat_setsockopt = compat_sock_common_setsockopt,
954 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000955 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800956#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700958EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960/*
961 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
962 * udp_poll
963 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800964static const struct proto_ops inet_sockraw_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 = datagram_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};
989
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +0000990static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 .family = PF_INET,
992 .create = inet_create,
993 .owner = THIS_MODULE,
994};
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996/* Upon startup we insert all the elements in inetsw_array[] into
997 * the linked list inetsw.
998 */
999static struct inet_protosw inetsw_array[] =
1000{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001001 {
1002 .type = SOCK_STREAM,
1003 .protocol = IPPROTO_TCP,
1004 .prot = &tcp_prot,
1005 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001006 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001007 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001008 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001010 {
1011 .type = SOCK_DGRAM,
1012 .protocol = IPPROTO_UDP,
1013 .prot = &udp_prot,
1014 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001015 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001017
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001018 {
1019 .type = SOCK_DGRAM,
1020 .protocol = IPPROTO_ICMP,
1021 .prot = &ping_prot,
1022 .ops = &inet_dgram_ops,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001023 .flags = INET_PROTOSW_REUSE,
1024 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
1026 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001027 .type = SOCK_RAW,
1028 .protocol = IPPROTO_IP, /* wild card */
1029 .prot = &raw_prot,
1030 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001031 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 }
1033};
1034
Denis Chengc40f6ff2007-09-16 16:39:25 -07001035#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
1037void inet_register_protosw(struct inet_protosw *p)
1038{
1039 struct list_head *lh;
1040 struct inet_protosw *answer;
1041 int protocol = p->protocol;
1042 struct list_head *last_perm;
1043
1044 spin_lock_bh(&inetsw_lock);
1045
1046 if (p->type >= SOCK_MAX)
1047 goto out_illegal;
1048
1049 /* If we are trying to override a permanent protocol, bail. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 last_perm = &inetsw[p->type];
1051 list_for_each(lh, &inetsw[p->type]) {
1052 answer = list_entry(lh, struct inet_protosw, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 /* Check only the non-wild match. */
Junwei Zhangf6c53332015-09-18 00:00:05 -04001054 if ((INET_PROTOSW_PERMANENT & answer->flags) == 0)
1055 break;
1056 if (protocol == answer->protocol)
1057 goto out_permanent;
1058 last_perm = lh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
1061 /* Add the new entry after the last permanent entry if any, so that
1062 * the new entry does not override a permanent entry when matched with
1063 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001064 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 * system automatically returns to the old behavior.
1066 */
1067 list_add_rcu(&p->list, last_perm);
1068out:
1069 spin_unlock_bh(&inetsw_lock);
1070
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 return;
1072
1073out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001074 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 goto out;
1076
1077out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001078 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 p->type);
1080 goto out;
1081}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001082EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
1084void inet_unregister_protosw(struct inet_protosw *p)
1085{
1086 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001087 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 p->protocol);
1089 } else {
1090 spin_lock_bh(&inetsw_lock);
1091 list_del_rcu(&p->list);
1092 spin_unlock_bh(&inetsw_lock);
1093
1094 synchronize_net();
1095 }
1096}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001097EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001099static int inet_sk_reselect_saddr(struct sock *sk)
1100{
1101 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001102 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001103 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001104 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001105 struct rtable *rt;
1106 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001107 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001108
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001109 inet_opt = rcu_dereference_protected(inet->inet_opt,
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001110 lockdep_sock_is_held(sk));
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001111 if (inet_opt && inet_opt->opt.srr)
1112 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001113
1114 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001115 fl4 = &inet->cork.fl.u.ip4;
1116 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001117 sk->sk_bound_dev_if, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001118 inet->inet_sport, inet->inet_dport, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001119 if (IS_ERR(rt))
1120 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001121
Changli Gaod8d1f302010-06-10 23:31:35 -07001122 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001123
David S. Miller6e869132011-05-06 16:18:04 -07001124 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001125
1126 if (new_saddr == old_saddr)
1127 return 0;
1128
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001129 if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001130 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1131 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001132 }
1133
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001134 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001135
1136 /*
1137 * XXX The only one ugly spot where we need to
1138 * XXX really change the sockets identity after
1139 * XXX it has entered the hashes. -DaveM
1140 *
1141 * Besides that, it does not check for connection
1142 * uniqueness. Wait for troubles.
1143 */
Craig Gallek086c6532016-02-10 11:50:35 -05001144 return __sk_prot_rehash(sk);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001145}
1146
1147int inet_sk_rebuild_header(struct sock *sk)
1148{
1149 struct inet_sock *inet = inet_sk(sk);
1150 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001151 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001152 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001153 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001154 int err;
1155
1156 /* Route is OK, nothing to do. */
1157 if (rt)
1158 return 0;
1159
1160 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001161 rcu_read_lock();
1162 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001163 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001164 if (inet_opt && inet_opt->opt.srr)
1165 daddr = inet_opt->opt.faddr;
1166 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001167 fl4 = &inet->cork.fl.u.ip4;
1168 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001169 inet->inet_dport, inet->inet_sport,
1170 sk->sk_protocol, RT_CONN_FLAGS(sk),
1171 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001172 if (!IS_ERR(rt)) {
1173 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001174 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001175 } else {
1176 err = PTR_ERR(rt);
1177
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001178 /* Routing failed... */
1179 sk->sk_route_caps = 0;
1180 /*
1181 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1182 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1183 */
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001184 if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr ||
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001185 sk->sk_state != TCP_SYN_SENT ||
1186 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1187 (err = inet_sk_reselect_saddr(sk)) != 0)
1188 sk->sk_err_soft = -err;
1189 }
1190
1191 return err;
1192}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001193EXPORT_SYMBOL(inet_sk_rebuild_header);
1194
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001195static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
Eric Dumazet3347c962013-10-19 11:42:56 -07001196 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001197{
1198 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001199 const struct net_offload *ops;
Eric Dumazet3347c962013-10-19 11:42:56 -07001200 unsigned int offset = 0;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001201 bool udpfrag, encap;
David S. Millerf9242b62012-06-19 18:56:21 -07001202 struct iphdr *iph;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001203 int proto;
Eric Dumazet3347c962013-10-19 11:42:56 -07001204 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001205 int ihl;
1206 int id;
1207
Herbert Xubbcf4672006-07-03 19:38:35 -07001208 if (unlikely(skb_shinfo(skb)->gso_type &
1209 ~(SKB_GSO_TCPV4 |
1210 SKB_GSO_UDP |
1211 SKB_GSO_DODGY |
1212 SKB_GSO_TCP_ECN |
Pravin B Shelar68c33162013-02-14 14:02:41 +00001213 SKB_GSO_GRE |
Tom Herbert4749c092014-06-04 17:20:23 -07001214 SKB_GSO_GRE_CSUM |
Eric Dumazetcb32f512013-10-19 11:42:57 -07001215 SKB_GSO_IPIP |
Eric Dumazet61c1db72013-10-20 20:47:30 -07001216 SKB_GSO_SIT |
Pravin B Shelar9b3eb5e2013-05-02 16:14:19 +00001217 SKB_GSO_TCPV6 |
Pravin B Shelar73136262013-03-07 13:21:51 +00001218 SKB_GSO_UDP_TUNNEL |
Tom Herbert0f4f4ff2014-06-04 17:20:16 -07001219 SKB_GSO_UDP_TUNNEL_CSUM |
Tom Herberte585f232014-11-04 09:06:54 -08001220 SKB_GSO_TUNNEL_REMCSUM |
Herbert Xubbcf4672006-07-03 19:38:35 -07001221 0)))
1222 goto out;
1223
Eric Dumazet3347c962013-10-19 11:42:56 -07001224 skb_reset_network_header(skb);
1225 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001226 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001227 goto out;
1228
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001229 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001230 ihl = iph->ihl * 4;
1231 if (ihl < sizeof(*iph))
1232 goto out;
1233
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001234 id = ntohs(iph->id);
1235 proto = iph->protocol;
1236
1237 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001238 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001239 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001240 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001241
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001242 encap = SKB_GSO_CB(skb)->encap_level > 0;
1243 if (encap)
Florian Westphal1e16aa32014-10-20 13:49:16 +02001244 features &= skb->dev->hw_enc_features;
Eric Dumazet3347c962013-10-19 11:42:56 -07001245 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001246
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001247 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001248
Herbert Xuf4c50d92006-06-22 03:02:40 -07001249 segs = ERR_PTR(-EPROTONOSUPPORT);
1250
Hannes Frederic Sowa91a48a22014-02-24 00:48:05 +01001251 if (skb->encapsulation &&
1252 skb_shinfo(skb)->gso_type & (SKB_GSO_SIT|SKB_GSO_IPIP))
1253 udpfrag = proto == IPPROTO_UDP && encap;
1254 else
1255 udpfrag = proto == IPPROTO_UDP && !skb->encapsulation;
Eric Dumazetdcd60772013-11-07 18:32:06 -08001256
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001257 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001258 if (likely(ops && ops->callbacks.gso_segment))
1259 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001260
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001261 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001262 goto out;
1263
1264 skb = segs;
1265 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001266 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001267 if (udpfrag) {
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001268 iph->id = htons(id);
1269 iph->frag_off = htons(offset >> 3);
Ian Morris00db4122015-04-03 09:17:27 +01001270 if (skb->next)
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001271 iph->frag_off |= htons(IP_MF);
Eric Dumazet3347c962013-10-19 11:42:56 -07001272 offset += skb->len - nhoff - ihl;
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001273 } else {
Pravin B Shelar25c77042013-03-24 17:36:29 +00001274 iph->id = htons(id++);
Pravin B Shelar490ab082013-02-22 07:30:30 +00001275 }
Eric Dumazet3347c962013-10-19 11:42:56 -07001276 iph->tot_len = htons(skb->len - nhoff);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001277 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001278 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001279 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001280 skb->network_header = (u8 *)iph - skb->head;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001281 } while ((skb = skb->next));
1282
1283out:
1284 return segs;
1285}
1286
Herbert Xu73cc19f2008-12-15 23:41:09 -08001287static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1288 struct sk_buff *skb)
1289{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001290 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001291 struct sk_buff **pp = NULL;
1292 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001293 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001294 unsigned int hlen;
1295 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001296 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001297 int flush = 1;
1298 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001299
Herbert Xua5b1cf22009-05-26 18:50:28 +00001300 off = skb_gro_offset(skb);
1301 hlen = off + sizeof(*iph);
1302 iph = skb_gro_header_fast(skb, off);
1303 if (skb_gro_header_hard(skb, hlen)) {
1304 iph = skb_gro_header_slow(skb, hlen, off);
1305 if (unlikely(!iph))
1306 goto out;
1307 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001308
David S. Millerf9242b62012-06-19 18:56:21 -07001309 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001310
1311 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001312 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001313 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001314 goto out_unlock;
1315
Herbert Xua5ad24b2009-02-08 18:00:39 +00001316 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001317 goto out_unlock;
1318
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001319 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001320 goto out_unlock;
1321
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001322 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001323 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001324 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001325
1326 for (p = *head; p; p = p->next) {
1327 struct iphdr *iph2;
1328
1329 if (!NAPI_GRO_CB(p)->same_flow)
1330 continue;
1331
Jerry Chu299603e82013-12-11 20:53:45 -08001332 iph2 = (struct iphdr *)(p->data + off);
1333 /* The above works because, with the exception of the top
1334 * (inner most) layer, we only aggregate pkts with the same
1335 * hdr length so all the hdrs we'll need to verify will start
1336 * at the same offset.
1337 */
Herbert Xua5ad24b2009-02-08 18:00:39 +00001338 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001339 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1340 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001341 NAPI_GRO_CB(p)->same_flow = 0;
1342 continue;
1343 }
1344
1345 /* All fields must match except length and checksum. */
1346 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001347 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001348 (iph->tos ^ iph2->tos) |
Jerry Chubf5a7552014-01-07 10:23:19 -08001349 ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001350
Jerry Chubf5a7552014-01-07 10:23:19 -08001351 /* Save the IP ID check to be included later when we get to
1352 * the transport layer so only the inner most IP ID is checked.
1353 * This is because some GSO/TSO implementations do not
1354 * correctly increment the IP ID for the outer hdrs.
1355 */
1356 NAPI_GRO_CB(p)->flush_id =
1357 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001358 NAPI_GRO_CB(p)->flush |= flush;
1359 }
1360
1361 NAPI_GRO_CB(skb)->flush |= flush;
Jerry Chu299603e82013-12-11 20:53:45 -08001362 skb_set_network_header(skb, off);
1363 /* The above will be needed by the transport layer if there is one
1364 * immediately following this IP hdr.
1365 */
1366
Eric Dumazet2c804d02014-09-30 22:12:05 -07001367 /* Note : No need to call skb_gro_postpull_rcsum() here,
1368 * as we already checked checksum over ipv4 header was 0
1369 */
Herbert Xu86911732009-01-29 14:19:50 +00001370 skb_gro_pull(skb, sizeof(*iph));
1371 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001372
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001373 pp = ops->callbacks.gro_receive(head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001374
1375out_unlock:
1376 rcu_read_unlock();
1377
1378out:
1379 NAPI_GRO_CB(skb)->flush |= flush;
1380
1381 return pp;
1382}
1383
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001384static struct sk_buff **ipip_gro_receive(struct sk_buff **head,
1385 struct sk_buff *skb)
1386{
1387 if (NAPI_GRO_CB(skb)->encap_mark) {
1388 NAPI_GRO_CB(skb)->flush = 1;
1389 return NULL;
1390 }
1391
1392 NAPI_GRO_CB(skb)->encap_mark = 1;
1393
1394 return inet_gro_receive(head, skb);
1395}
1396
Deepa Dinamani822c8682016-02-27 00:32:15 -08001397#define SECONDS_PER_DAY 86400
1398
1399/* inet_current_timestamp - Return IP network timestamp
1400 *
1401 * Return milliseconds since midnight in network byte order.
1402 */
1403__be32 inet_current_timestamp(void)
1404{
1405 u32 secs;
1406 u32 msecs;
1407 struct timespec64 ts;
1408
1409 ktime_get_real_ts64(&ts);
1410
1411 /* Get secs since midnight. */
1412 (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs);
1413 /* Convert to msecs. */
1414 msecs = secs * MSEC_PER_SEC;
1415 /* Convert nsec to msec. */
1416 msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC;
1417
1418 /* Convert to network byte order. */
Deepa Dinamani3ba9d302016-03-21 18:21:26 -07001419 return htonl(msecs);
Deepa Dinamani822c8682016-02-27 00:32:15 -08001420}
1421EXPORT_SYMBOL(inet_current_timestamp);
1422
Willem de Bruijnf4713a32014-11-26 14:53:02 -05001423int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1424{
1425 if (sk->sk_family == AF_INET)
1426 return ip_recv_error(sk, msg, len, addr_len);
1427#if IS_ENABLED(CONFIG_IPV6)
1428 if (sk->sk_family == AF_INET6)
1429 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1430#endif
1431 return -EINVAL;
1432}
1433
Jerry Chu299603e82013-12-11 20:53:45 -08001434static int inet_gro_complete(struct sk_buff *skb, int nhoff)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001435{
Jerry Chu299603e82013-12-11 20:53:45 -08001436 __be16 newlen = htons(skb->len - nhoff);
1437 struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001438 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001439 int proto = iph->protocol;
1440 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001441
Jerry Chuc3caf112014-07-14 15:54:46 -07001442 if (skb->encapsulation)
1443 skb_set_inner_network_header(skb, nhoff);
1444
Herbert Xu73cc19f2008-12-15 23:41:09 -08001445 csum_replace2(&iph->check, iph->tot_len, newlen);
1446 iph->tot_len = newlen;
1447
1448 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001449 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001450 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001451 goto out_unlock;
1452
Jerry Chu299603e82013-12-11 20:53:45 -08001453 /* Only need to add sizeof(*iph) to get to the next hdr below
1454 * because any hdr with option will have been flushed in
1455 * inet_gro_receive().
1456 */
1457 err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001458
1459out_unlock:
1460 rcu_read_unlock();
1461
1462 return err;
1463}
1464
Jesse Grossb8cba752016-03-19 09:32:00 -07001465static int ipip_gro_complete(struct sk_buff *skb, int nhoff)
1466{
1467 skb->encapsulation = 1;
1468 skb_shinfo(skb)->gso_type |= SKB_GSO_IPIP;
1469 return inet_gro_complete(skb, nhoff);
1470}
1471
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001472int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001473 unsigned short type, unsigned char protocol,
1474 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001475{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001476 struct socket *sock;
Eric W. Biederman26abe142015-05-08 21:10:31 -05001477 int rc = sock_create_kern(net, family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001478
1479 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001480 *sk = sock->sk;
1481 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001482 /*
1483 * Unhash it so that IP input processing does not even see it,
1484 * we do not wish this socket to see incoming packets.
1485 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001486 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001487 }
1488 return rc;
1489}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001490EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1491
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301492u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
1493{
1494 return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
1495}
1496EXPORT_SYMBOL_GPL(snmp_get_cpu_field);
1497
WANG Cong698365f2014-05-05 15:55:55 -07001498unsigned long snmp_fold_field(void __percpu *mib, int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001499{
1500 unsigned long res = 0;
WANG Cong698365f2014-05-05 15:55:55 -07001501 int i;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001502
WANG Cong698365f2014-05-05 15:55:55 -07001503 for_each_possible_cpu(i)
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301504 res += snmp_get_cpu_field(mib, i, offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001505 return res;
1506}
1507EXPORT_SYMBOL_GPL(snmp_fold_field);
1508
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001509#if BITS_PER_LONG==32
1510
David S. Miller80ec1922015-08-30 22:40:44 -07001511u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt,
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301512 size_t syncp_offset)
1513{
1514 void *bhptr;
1515 struct u64_stats_sync *syncp;
1516 u64 v;
1517 unsigned int start;
1518
1519 bhptr = per_cpu_ptr(mib, cpu);
1520 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1521 do {
1522 start = u64_stats_fetch_begin_irq(syncp);
1523 v = *(((u64 *)bhptr) + offt);
1524 } while (u64_stats_fetch_retry_irq(syncp, start));
1525
1526 return v;
1527}
1528EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);
1529
WANG Cong698365f2014-05-05 15:55:55 -07001530u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001531{
1532 u64 res = 0;
1533 int cpu;
1534
1535 for_each_possible_cpu(cpu) {
Madalin Bucurd1bfc622015-08-31 15:46:07 +03001536 res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001537 }
1538 return res;
1539}
1540EXPORT_SYMBOL_GPL(snmp_fold_field64);
1541#endif
1542
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001544static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001546 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547};
1548#endif
1549
Alexey Dobriyan32613092009-09-14 12:21:47 +00001550static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001551 .early_demux = tcp_v4_early_demux,
1552 .handler = tcp_v4_rcv,
1553 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001554 .no_policy = 1,
1555 .netns_ok = 1,
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +01001556 .icmp_strict_tag_validation = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557};
1558
Alexey Dobriyan32613092009-09-14 12:21:47 +00001559static const struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001560 .early_demux = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 .handler = udp_rcv,
1562 .err_handler = udp_err,
1563 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001564 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565};
1566
Alexey Dobriyan32613092009-09-14 12:21:47 +00001567static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001569 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001570 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001571 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572};
1573
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001574static __net_init int ipv4_mib_init_net(struct net *net)
1575{
John Stultz827da442013-10-07 15:51:58 -07001576 int i;
1577
WANG Cong698365f2014-05-05 15:55:55 -07001578 net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1579 if (!net->mib.tcp_statistics)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001580 goto err_tcp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001581 net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1582 if (!net->mib.ip_statistics)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001583 goto err_ip_mib;
John Stultz827da442013-10-07 15:51:58 -07001584
1585 for_each_possible_cpu(i) {
1586 struct ipstats_mib *af_inet_stats;
WANG Cong698365f2014-05-05 15:55:55 -07001587 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
John Stultz827da442013-10-07 15:51:58 -07001588 u64_stats_init(&af_inet_stats->syncp);
John Stultz827da442013-10-07 15:51:58 -07001589 }
1590
WANG Cong698365f2014-05-05 15:55:55 -07001591 net->mib.net_statistics = alloc_percpu(struct linux_mib);
1592 if (!net->mib.net_statistics)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001593 goto err_net_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001594 net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1595 if (!net->mib.udp_statistics)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001596 goto err_udp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001597 net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1598 if (!net->mib.udplite_statistics)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001599 goto err_udplite_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001600 net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1601 if (!net->mib.icmp_statistics)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001602 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001603 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1604 GFP_KERNEL);
1605 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001606 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001607
1608 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001609 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001610
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001611err_icmpmsg_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001612 free_percpu(net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001613err_icmp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001614 free_percpu(net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001615err_udplite_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001616 free_percpu(net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001617err_udp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001618 free_percpu(net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001619err_net_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001620 free_percpu(net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001621err_ip_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001622 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001623err_tcp_mib:
1624 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001625}
1626
1627static __net_exit void ipv4_mib_exit_net(struct net *net)
1628{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001629 kfree(net->mib.icmpmsg_statistics);
WANG Cong698365f2014-05-05 15:55:55 -07001630 free_percpu(net->mib.icmp_statistics);
1631 free_percpu(net->mib.udplite_statistics);
1632 free_percpu(net->mib.udp_statistics);
1633 free_percpu(net->mib.net_statistics);
1634 free_percpu(net->mib.ip_statistics);
1635 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001636}
1637
1638static __net_initdata struct pernet_operations ipv4_mib_ops = {
1639 .init = ipv4_mib_init_net,
1640 .exit = ipv4_mib_exit_net,
1641};
1642
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643static int __init init_ipv4_mibs(void)
1644{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001645 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646}
1647
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001648static __net_init int inet_init_net(struct net *net)
1649{
1650 /*
1651 * Set defaults for local port range
1652 */
1653 seqlock_init(&net->ipv4.ip_local_ports.lock);
1654 net->ipv4.ip_local_ports.range[0] = 32768;
Eric Dumazet07f4c902015-05-24 14:49:35 -07001655 net->ipv4.ip_local_ports.range[1] = 60999;
Cong Wangba6b9182014-05-06 11:02:50 -07001656
1657 seqlock_init(&net->ipv4.ping_group_range.lock);
1658 /*
1659 * Sane defaults - nobody may create ping sockets.
1660 * Boot scripts should set this to distro-specific group.
1661 */
1662 net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1663 net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001664 return 0;
1665}
1666
1667static __net_exit void inet_exit_net(struct net *net)
1668{
1669}
1670
1671static __net_initdata struct pernet_operations af_inet_ops = {
1672 .init = inet_init_net,
1673 .exit = inet_exit_net,
1674};
1675
1676static int __init init_inet_pernet_ops(void)
1677{
1678 return register_pernet_subsys(&af_inet_ops);
1679}
1680
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682
Herbert Xu30e224d2005-07-05 14:40:10 -07001683/*
1684 * IP protocol layer initialiser
1685 */
1686
Vlad Yasevich22061d82012-11-15 08:49:11 +00001687static struct packet_offload ip_packet_offload __read_mostly = {
1688 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001689 .callbacks = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001690 .gso_segment = inet_gso_segment,
1691 .gro_receive = inet_gro_receive,
1692 .gro_complete = inet_gro_complete,
1693 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001694};
1695
Eric Dumazetcb32f512013-10-19 11:42:57 -07001696static const struct net_offload ipip_offload = {
1697 .callbacks = {
Eric Dumazetcb32f512013-10-19 11:42:57 -07001698 .gso_segment = inet_gso_segment,
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001699 .gro_receive = ipip_gro_receive,
Jesse Grossb8cba752016-03-19 09:32:00 -07001700 .gro_complete = ipip_gro_complete,
Eric Dumazetcb32f512013-10-19 11:42:57 -07001701 },
1702};
1703
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001704static int __init ipv4_offload_init(void)
1705{
1706 /*
1707 * Add offloads
1708 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001709 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001710 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001711 if (tcpv4_offload_init() < 0)
1712 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001713
1714 dev_add_offload(&ip_packet_offload);
Eric Dumazetcb32f512013-10-19 11:42:57 -07001715 inet_add_offload(&ipip_offload, IPPROTO_IPIP);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001716 return 0;
1717}
1718
1719fs_initcall(ipv4_offload_init);
1720
1721static struct packet_type ip_packet_type __read_mostly = {
1722 .type = cpu_to_be16(ETH_P_IP),
1723 .func = ip_rcv,
1724};
1725
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726static int __init inet_init(void)
1727{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 struct inet_protosw *q;
1729 struct list_head *r;
1730 int rc = -EINVAL;
1731
Eyal Birgerb4772ef2015-03-01 14:58:29 +02001732 sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
1734 rc = proto_register(&tcp_prot, 1);
1735 if (rc)
WANG Cong122ff242014-05-12 16:04:53 -07001736 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
1738 rc = proto_register(&udp_prot, 1);
1739 if (rc)
1740 goto out_unregister_tcp_proto;
1741
1742 rc = proto_register(&raw_prot, 1);
1743 if (rc)
1744 goto out_unregister_udp_proto;
1745
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001746 rc = proto_register(&ping_prot, 1);
1747 if (rc)
1748 goto out_unregister_raw_proto;
1749
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001751 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 */
1753
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001754 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755
Al Virobd7b1532008-07-15 16:00:59 -04001756#ifdef CONFIG_SYSCTL
1757 ip_static_sysctl_init();
1758#endif
1759
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 /*
1761 * Add all the base protocols.
1762 */
1763
1764 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001765 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001767 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001769 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770#ifdef CONFIG_IP_MULTICAST
1771 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001772 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773#endif
1774
1775 /* Register the socket-side information for inet_create. */
1776 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1777 INIT_LIST_HEAD(r);
1778
1779 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1780 inet_register_protosw(q);
1781
1782 /*
1783 * Set the ARP module up
1784 */
1785
1786 arp_init();
1787
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001788 /*
1789 * Set the IP module up
1790 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 ip_init();
1793
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001794 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795
1796 /* Setup TCP slab cache for open requests. */
1797 tcp_init();
1798
Hideo Aoki95766ff2007-12-31 00:29:24 -08001799 /* Setup UDP memory threshold */
1800 udp_init();
1801
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001802 /* Add UDP-Lite (RFC 3828) */
1803 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001805 ping_init();
1806
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 /*
1808 * Set the ICMP layer up
1809 */
1810
Denis V. Luneva5710d62008-02-29 11:14:50 -08001811 if (icmp_init() < 0)
1812 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813
1814 /*
1815 * Initialise the multicast router
1816 */
1817#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001818 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001819 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820#endif
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001821
1822 if (init_inet_pernet_ops())
1823 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 /*
1825 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001826 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827
Stephen Hemminger132adf52007-03-08 20:44:43 -08001828 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001829 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001830
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 ipv4_proc_init();
1832
1833 ipfrag_init();
1834
Herbert Xu30e224d2005-07-05 14:40:10 -07001835 dev_add_pack(&ip_packet_type);
1836
Thomas Graf045a0fa2015-07-23 10:08:44 +02001837 ip_tunnel_core_init();
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 rc = 0;
1840out:
1841 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001842out_unregister_raw_proto:
1843 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844out_unregister_udp_proto:
1845 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001846out_unregister_tcp_proto:
1847 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 goto out;
1849}
1850
Heiko Carstensa536e072006-04-28 15:19:17 -07001851fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
1853/* ------------------------------------------------------------------------ */
1854
1855#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856static int __init ipv4_proc_init(void)
1857{
1858 int rc = 0;
1859
1860 if (raw_proc_init())
1861 goto out_raw;
1862 if (tcp4_proc_init())
1863 goto out_tcp;
1864 if (udp4_proc_init())
1865 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001866 if (ping_proc_init())
1867 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 if (ip_misc_proc_init())
1869 goto out_misc;
1870out:
1871 return rc;
1872out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001873 ping_proc_exit();
1874out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 udp4_proc_exit();
1876out_udp:
1877 tcp4_proc_exit();
1878out_tcp:
1879 raw_proc_exit();
1880out_raw:
1881 rc = -ENOMEM;
1882 goto out;
1883}
1884
1885#else /* CONFIG_PROC_FS */
1886static int __init ipv4_proc_init(void)
1887{
1888 return 0;
1889}
1890#endif /* CONFIG_PROC_FS */
1891
1892MODULE_ALIAS_NETPROTO(PF_INET);
1893