blob: 08a8430f5647de15da425a81d900e8c76a037bf1 [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>
Paul Gortmakerd3fc0352016-07-11 16:37:51 -040076#include <linux/kmod.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#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
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080093#include <linux/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
Yafang Shao563e0bb2017-12-20 11:12:51 +0800124#include <trace/events/sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126/* The inetsw table contains everything that inet_create needs to
127 * build a new socket.
128 */
129static struct list_head inetsw[SOCK_MAX];
130static DEFINE_SPINLOCK(inetsw_lock);
131
132/* New destruction routine */
133
134void inet_sock_destruct(struct sock *sk)
135{
136 struct inet_sock *inet = inet_sk(sk);
137
138 __skb_queue_purge(&sk->sk_receive_queue);
Eric Dumazet8b27dae2019-03-22 08:56:40 -0700139 if (sk->sk_rx_skb_cache) {
140 __kfree_skb(sk->sk_rx_skb_cache);
141 sk->sk_rx_skb_cache = NULL;
142 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 __skb_queue_purge(&sk->sk_error_queue);
144
Hideo Aoki95766ff2007-12-31 00:29:24 -0800145 sk_mem_reclaim(sk);
146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700148 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 sk->sk_state, sk);
150 return;
151 }
152 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700153 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 return;
155 }
156
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700157 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
Reshetova, Elena14afee42017-06-30 13:08:00 +0300158 WARN_ON(refcount_read(&sk->sk_wmem_alloc));
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700159 WARN_ON(sk->sk_wmem_queued);
160 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000162 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Xin Long5869b8f2019-03-31 17:03:02 +0800163 dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700164 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700165 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700167EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168
169/*
170 * The routines beyond this point handle the behaviour of an AF_INET
171 * socket object. Mostly it punts to the subprotocols of IP to do
172 * the work.
173 */
174
175/*
176 * Automatically bind an unbound socket.
177 */
178
179static int inet_autobind(struct sock *sk)
180{
181 struct inet_sock *inet;
182 /* We may need to bind the socket. */
183 lock_sock(sk);
184 inet = inet_sk(sk);
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000185 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 if (sk->sk_prot->get_port(sk, 0)) {
187 release_sock(sk);
188 return -EAGAIN;
189 }
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000190 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 }
192 release_sock(sk);
193 return 0;
194}
195
196/*
197 * Move a socket into listening state.
198 */
199int inet_listen(struct socket *sock, int backlog)
200{
201 struct sock *sk = sock->sk;
202 unsigned char old_state;
Haishuang Yane1cfcbe2017-09-27 11:35:40 +0800203 int err, tcp_fastopen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205 lock_sock(sk);
206
207 err = -EINVAL;
208 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
209 goto out;
210
211 old_state = sk->sk_state;
212 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
213 goto out;
214
Yafang Shao1295e2c2018-11-07 19:20:16 +0800215 sk->sk_max_ack_backlog = backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 /* Really, if the socket is already in listen state
217 * we can only allow the backlog to be adjusted.
218 */
219 if (old_state != TCP_LISTEN) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700220 /* Enable TFO w/o requiring TCP_FASTOPEN socket option.
Jerry Chu83368862012-08-31 12:29:12 +0000221 * Note that only TCP sockets (SOCK_STREAM) will reach here.
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700222 * Also fastopen backlog may already been set via the option
223 * because the socket was in TCP_LISTEN state previously but
224 * was shutdown() rather than close().
Jerry Chu83368862012-08-31 12:29:12 +0000225 */
Haishuang Yane1cfcbe2017-09-27 11:35:40 +0800226 tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen;
227 if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
228 (tcp_fastopen & TFO_SERVER_ENABLE) &&
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700229 !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700230 fastopen_queue_tune(sk, backlog);
Haishuang Yan43713842017-09-27 11:35:42 +0800231 tcp_fastopen_init_key_once(sock_net(sk));
Jerry Chu83368862012-08-31 12:29:12 +0000232 }
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700233
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;
Andrey Ignatovf333ee02018-07-11 17:33:32 -0700237 tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 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
Hannes Frederic Sowa79462ad02015-12-14 22:03:39 +0100262 if (protocol < 0 || protocol >= IPPROTO_MAX)
263 return -EINVAL;
264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 sock->state = SS_UNCONNECTED;
266
267 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700268lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800269 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700271 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700273 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 /* Check the non-wild match. */
275 if (protocol == answer->protocol) {
276 if (protocol != IPPROTO_IP)
277 break;
278 } else {
279 /* Check for the two wild cases. */
280 if (IPPROTO_IP == protocol) {
281 protocol = answer->protocol;
282 break;
283 }
284 if (IPPROTO_IP == answer->protocol)
285 break;
286 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800287 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 }
289
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700290 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700291 if (try_loading_module < 2) {
292 rcu_read_unlock();
293 /*
294 * Be more specific, e.g. net-pf-2-proto-132-type-1
295 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
296 */
297 if (++try_loading_module == 1)
298 request_module("net-pf-%d-proto-%d-type-%d",
299 PF_INET, protocol, sock->type);
300 /*
301 * Fall back to generic, e.g. net-pf-2-proto-132
302 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
303 */
304 else
305 request_module("net-pf-%d-proto-%d",
306 PF_INET, protocol);
307 goto lookup_protocol;
308 } else
309 goto out_rcu_unlock;
310 }
311
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000313 if (sock->type == SOCK_RAW && !kern &&
314 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
317 sock->ops = answer->ops;
318 answer_prot = answer->prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 answer_flags = answer->flags;
320 rcu_read_unlock();
321
Ian Morris51456b22015-04-03 09:17:26 +0100322 WARN_ON(!answer_prot->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
324 err = -ENOBUFS;
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500325 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
Ian Morris51456b22015-04-03 09:17:26 +0100326 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 goto out;
328
329 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000331 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800334 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000336 inet->nodefrag = 0;
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000339 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 if (IPPROTO_RAW == protocol)
341 inet->hdrincl = 1;
342 }
343
Hannes Frederic Sowa974eda12013-12-14 05:13:38 +0100344 if (net->ipv4.sysctl_ip_no_pmtu_disc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 inet->pmtudisc = IP_PMTUDISC_DONT;
346 else
347 inet->pmtudisc = IP_PMTUDISC_WANT;
348
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000349 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351 sock_init_data(sock, sk);
352
353 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 sk->sk_protocol = protocol;
355 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
356
357 inet->uc_ttl = -1;
358 inet->mc_loop = 1;
359 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000360 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 inet->mc_index = 0;
362 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000363 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700365 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000367 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 /* It assumes that any protocol which allows
369 * the user to assign a number at socket
370 * creation time automatically
371 * shares.
372 */
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000373 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 /* Add to protocol hash chains. */
Craig Gallek086c6532016-02-10 11:50:35 -0500375 err = sk->sk_prot->hash(sk);
376 if (err) {
377 sk_common_release(sk);
378 goto out;
379 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 }
381
382 if (sk->sk_prot->init) {
383 err = sk->sk_prot->init(sk);
David Ahern610236582016-12-01 08:48:04 -0800384 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 sk_common_release(sk);
David Ahern610236582016-12-01 08:48:04 -0800386 goto out;
387 }
388 }
389
390 if (!kern) {
391 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
392 if (err) {
393 sk_common_release(sk);
394 goto out;
395 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 }
397out:
398 return err;
399out_rcu_unlock:
400 rcu_read_unlock();
401 goto out;
402}
403
404
405/*
406 * The peer socket should always be NULL (or else). When we call this
407 * function we are destroying the object and from then on nobody
408 * should refer to it.
409 */
410int inet_release(struct socket *sock)
411{
412 struct sock *sk = sock->sk;
413
414 if (sk) {
415 long timeout;
416
417 /* Applications forget to leave groups before exiting */
418 ip_mc_drop_socket(sk);
419
420 /* If linger is set, we don't return until the close
421 * is complete. Otherwise we return immediately. The
422 * actually closing is done the same either way.
423 *
424 * If the close is due to the process exiting, we never
425 * linger..
426 */
427 timeout = 0;
428 if (sock_flag(sk, SOCK_LINGER) &&
429 !(current->flags & PF_EXITING))
430 timeout = sk->sk_lingertime;
431 sock->sk = NULL;
432 sk->sk_prot->close(sk, timeout);
433 }
434 return 0;
435}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700436EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
439{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 int err;
442
443 /* If the socket has its own bind function then use it. (RAW) */
444 if (sk->sk_prot->bind) {
Andrey Ignatov3679d582018-03-30 15:08:04 -0700445 return sk->sk_prot->bind(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 if (addr_len < sizeof(struct sockaddr_in))
Andrey Ignatov3679d582018-03-30 15:08:04 -0700448 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700450 /* BPF prog is run before any checks are done so that if the prog
451 * changes context in a wrong way it will be caught.
452 */
453 err = BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr);
454 if (err)
Andrey Ignatov3679d582018-03-30 15:08:04 -0700455 return err;
456
457 return __inet_bind(sk, uaddr, addr_len, false, true);
458}
459EXPORT_SYMBOL(inet_bind);
460
461int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
462 bool force_bind_address_no_port, bool with_lock)
463{
464 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 struct inet_sock *inet = inet_sk(sk);
466 struct net *net = sock_net(sk);
467 unsigned short snum;
468 int chk_addr_ret;
469 u32 tb_id = RT_TABLE_LOCAL;
470 int err;
471
Marcus Meissnerc349a522011-07-04 01:30:29 +0000472 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400473 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
474 * only if s_addr is INADDR_ANY.
475 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000476 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400477 if (addr->sin_family != AF_UNSPEC ||
478 addr->sin_addr.s_addr != htonl(INADDR_ANY))
479 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000480 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700481
David Ahern3236b002015-09-29 20:07:14 -0700482 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
David Ahern30bbaa12015-08-13 14:59:05 -0600483 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485 /* Not specified by any standard per-se, however it breaks too
486 * many applications when removed. It is unfortunate since
487 * allowing applications to make a non-local bind solves
488 * several problems with systems using dynamic addressing.
489 * (ie. your servers still start up even if your ISDN link
490 * is temporarily down)
491 */
492 err = -EADDRNOTAVAIL;
Vincent Bernat83ba4642018-07-31 21:18:11 +0200493 if (!inet_can_nonlocal_bind(net, inet) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700494 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 chk_addr_ret != RTN_LOCAL &&
496 chk_addr_ret != RTN_MULTICAST &&
497 chk_addr_ret != RTN_BROADCAST)
498 goto out;
499
500 snum = ntohs(addr->sin_port);
501 err = -EACCES;
Krister Johansen4548b682017-01-20 17:49:11 -0800502 if (snum && snum < inet_prot_sock(net) &&
Eric W. Biederman35946982012-11-16 03:03:12 +0000503 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 goto out;
505
506 /* We keep a pair of addresses. rcv_saddr is the one
507 * used by hash lookups, and saddr is used for transmit.
508 *
509 * In the BSD API these are the same except where it
510 * would be illegal to use them (multicast/broadcast) in
511 * which case the sending device address is used.
512 */
Andrey Ignatov3679d582018-03-30 15:08:04 -0700513 if (with_lock)
514 lock_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
516 /* Check these errors (active socket, double bind). */
517 err = -EINVAL;
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000518 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 goto out_release_sock;
520
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000521 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000523 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
525 /* Make sure we are allowed to bind here. */
Andrey Ignatovaac3fc32018-03-30 15:08:07 -0700526 if (snum || !(inet->bind_address_no_port ||
527 force_bind_address_no_port)) {
528 if (sk->sk_prot->get_port(sk, snum)) {
529 inet->inet_saddr = inet->inet_rcv_saddr = 0;
530 err = -EADDRINUSE;
531 goto out_release_sock;
532 }
533 err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk);
534 if (err) {
535 inet->inet_saddr = inet->inet_rcv_saddr = 0;
536 goto out_release_sock;
537 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 }
539
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000540 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
542 if (snum)
543 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000544 inet->inet_sport = htons(inet->inet_num);
545 inet->inet_daddr = 0;
546 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 sk_dst_reset(sk);
548 err = 0;
549out_release_sock:
Andrey Ignatov3679d582018-03-30 15:08:04 -0700550 if (with_lock)
551 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552out:
553 return err;
554}
555
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000556int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 int addr_len, int flags)
558{
559 struct sock *sk = sock->sk;
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700560 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
Changli Gao6503d962010-03-31 22:58:26 +0000562 if (addr_len < sizeof(uaddr->sa_family))
563 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 if (uaddr->sa_family == AF_UNSPEC)
565 return sk->sk_prot->disconnect(sk, flags);
566
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700567 if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) {
568 err = sk->sk_prot->pre_connect(sk, uaddr, addr_len);
569 if (err)
570 return err;
571 }
572
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000573 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000575 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700577EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
Yuchung Cheng783237e2012-07-19 06:43:07 +0000579static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
WANG Cong14135f32016-11-01 16:04:36 -0700581 DEFINE_WAIT_FUNC(wait, woken_wake_function);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
WANG Cong14135f32016-11-01 16:04:36 -0700583 add_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000584 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 /* Basic assumption: if someone sets sk->sk_err, he _must_
587 * change state of the socket from TCP_SYN_*.
588 * Connect() does not allow to get error notifications
589 * without closing the socket.
590 */
591 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
592 release_sock(sk);
WANG Cong14135f32016-11-01 16:04:36 -0700593 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 lock_sock(sk);
595 if (signal_pending(current) || !timeo)
596 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 }
WANG Cong14135f32016-11-01 16:04:36 -0700598 remove_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000599 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 return timeo;
601}
602
603/*
604 * Connect to a remote host. There is regrettably still a little
605 * TCP 'magic' in here.
606 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000607int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
Willy Tarreau3979ad72017-01-25 14:42:46 +0100608 int addr_len, int flags, int is_sendmsg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609{
610 struct sock *sk = sock->sk;
611 int err;
612 long timeo;
613
Wei Wang19f6d3f2017-01-23 10:59:22 -0800614 /*
615 * uaddr can be NULL and addr_len can be 0 if:
616 * sk is a TCP fastopen active socket and
617 * TCP_FASTOPEN_CONNECT sockopt is set and
618 * we already have a valid cookie for this socket.
619 * In this case, user can call write() after connect().
620 * write() will invoke tcp_sendmsg_fastopen() which calls
621 * __inet_stream_connect().
622 */
623 if (uaddr) {
624 if (addr_len < sizeof(uaddr->sa_family))
625 return -EINVAL;
Changli Gao6503d962010-03-31 22:58:26 +0000626
Wei Wang19f6d3f2017-01-23 10:59:22 -0800627 if (uaddr->sa_family == AF_UNSPEC) {
628 err = sk->sk_prot->disconnect(sk, flags);
629 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
630 goto out;
631 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 }
633
634 switch (sock->state) {
635 default:
636 err = -EINVAL;
637 goto out;
638 case SS_CONNECTED:
639 err = -EISCONN;
640 goto out;
641 case SS_CONNECTING:
Wei Wang19f6d3f2017-01-23 10:59:22 -0800642 if (inet_sk(sk)->defer_connect)
Willy Tarreau3979ad72017-01-25 14:42:46 +0100643 err = is_sendmsg ? -EINPROGRESS : -EISCONN;
Wei Wang19f6d3f2017-01-23 10:59:22 -0800644 else
645 err = -EALREADY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 /* Fall out of switch with err, set for this state */
647 break;
648 case SS_UNCONNECTED:
649 err = -EISCONN;
650 if (sk->sk_state != TCP_CLOSE)
651 goto out;
652
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700653 if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) {
654 err = sk->sk_prot->pre_connect(sk, uaddr, addr_len);
655 if (err)
656 goto out;
657 }
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 err = sk->sk_prot->connect(sk, uaddr, addr_len);
660 if (err < 0)
661 goto out;
662
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900663 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Wei Wang19f6d3f2017-01-23 10:59:22 -0800665 if (!err && inet_sk(sk)->defer_connect)
666 goto out;
667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 /* Just entered SS_CONNECTING state; the only
669 * difference is that return value in non-blocking
670 * case is EINPROGRESS, rather than EALREADY.
671 */
672 err = -EINPROGRESS;
673 break;
674 }
675
676 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
677
678 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000679 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
680 tcp_sk(sk)->fastopen_req &&
681 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
682
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000684 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 goto out;
686
687 err = sock_intr_errno(timeo);
688 if (signal_pending(current))
689 goto out;
690 }
691
692 /* Connection was closed by RST, timeout, ICMP error
693 * or another process disconnected us.
694 */
695 if (sk->sk_state == TCP_CLOSE)
696 goto sock_error;
697
698 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
699 * and error was received after socket entered established state.
700 * Hence, it is handled normally after connect() return successfully.
701 */
702
703 sock->state = SS_CONNECTED;
704 err = 0;
705out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 return err;
707
708sock_error:
709 err = sock_error(sk) ? : -ECONNABORTED;
710 sock->state = SS_UNCONNECTED;
711 if (sk->sk_prot->disconnect(sk, flags))
712 sock->state = SS_DISCONNECTING;
713 goto out;
714}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000715EXPORT_SYMBOL(__inet_stream_connect);
716
717int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
718 int addr_len, int flags)
719{
720 int err;
721
722 lock_sock(sock->sk);
Willy Tarreau3979ad72017-01-25 14:42:46 +0100723 err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0);
Yuchung Chengcf60af02012-07-19 06:43:09 +0000724 release_sock(sock->sk);
725 return err;
726}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700727EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
729/*
730 * Accept a pending connection. The TCP layer now gives BSD semantics.
731 */
732
David Howellscdfbabf2017-03-09 08:09:05 +0000733int inet_accept(struct socket *sock, struct socket *newsock, int flags,
734 bool kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735{
736 struct sock *sk1 = sock->sk;
737 int err = -EINVAL;
David Howellscdfbabf2017-03-09 08:09:05 +0000738 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
740 if (!sk2)
741 goto do_err;
742
743 lock_sock(sk2);
744
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000745 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700746 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000747 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
748 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750 sock_graft(sk2, newsock);
751
752 newsock->state = SS_CONNECTED;
753 err = 0;
754 release_sock(sk2);
755do_err:
756 return err;
757}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700758EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760
761/*
762 * This does both peername and sockname.
763 */
764int inet_getname(struct socket *sock, struct sockaddr *uaddr,
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +0100765 int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
767 struct sock *sk = sock->sk;
768 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700769 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
771 sin->sin_family = AF_INET;
772 if (peer) {
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000773 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
775 peer == 1))
776 return -ENOTCONN;
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000777 sin->sin_port = inet->inet_dport;
778 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 } else {
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000780 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 if (!addr)
Eric Dumazetc720c7e82009-10-15 06:30:45 +0000782 addr = inet->inet_saddr;
783 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 sin->sin_addr.s_addr = addr;
785 }
786 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +0100787 return sizeof(*sin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700789EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Ying Xue1b784142015-03-02 15:37:48 +0800791int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792{
793 struct sock *sk = sock->sk;
794
David S. Millerc58dc012010-04-27 15:05:31 -0700795 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700796
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000798 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
799 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 return -EAGAIN;
801
Ying Xue1b784142015-03-02 15:37:48 +0800802 return sk->sk_prot->sendmsg(sk, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700804EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805
Changli Gao7ba42912010-07-10 20:41:55 +0000806ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
807 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 struct sock *sk = sock->sk;
810
David S. Millerc58dc012010-04-27 15:05:31 -0700811 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700812
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000814 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
815 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 return -EAGAIN;
817
818 if (sk->sk_prot->sendpage)
819 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
820 return sock_no_sendpage(sock, page, offset, size, flags);
821}
Changli Gao7ba42912010-07-10 20:41:55 +0000822EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
Ying Xue1b784142015-03-02 15:37:48 +0800824int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
825 int flags)
Tom Herbertfec5e652010-04-16 16:01:27 -0700826{
827 struct sock *sk = sock->sk;
828 int addr_len = 0;
829 int err;
830
Soheil Hassas Yeganehe3f2c4a2018-01-03 21:47:10 -0500831 if (likely(!(flags & MSG_ERRQUEUE)))
832 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700833
Ying Xue1b784142015-03-02 15:37:48 +0800834 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Tom Herbertfec5e652010-04-16 16:01:27 -0700835 flags & ~MSG_DONTWAIT, &addr_len);
836 if (err >= 0)
837 msg->msg_namelen = addr_len;
838 return err;
839}
840EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
842int inet_shutdown(struct socket *sock, int how)
843{
844 struct sock *sk = sock->sk;
845 int err = 0;
846
847 /* This should really check to make sure
848 * the socket is a TCP socket. (WHY AC...)
849 */
850 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
851 1->2 bit 2 snds.
852 2->3 */
853 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
854 return -EINVAL;
855
856 lock_sock(sk);
857 if (sock->state == SS_CONNECTING) {
858 if ((1 << sk->sk_state) &
859 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
860 sock->state = SS_DISCONNECTING;
861 else
862 sock->state = SS_CONNECTED;
863 }
864
865 switch (sk->sk_state) {
866 case TCP_CLOSE:
867 err = -ENOTCONN;
868 /* Hack to wake up other listeners, who can poll for
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800869 EPOLLHUP, even on eg. unconnected UDP sockets -- RR */
Gustavo A. R. Silvafcfd6df2017-10-16 15:48:55 -0500870 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 default:
872 sk->sk_shutdown |= how;
873 if (sk->sk_prot->shutdown)
874 sk->sk_prot->shutdown(sk, how);
875 break;
876
877 /* Remaining two branches are temporary solution for missing
878 * close() in multithreaded environment. It is _not_ a good idea,
879 * but we have no choice until close() is repaired at VFS level.
880 */
881 case TCP_LISTEN:
882 if (!(how & RCV_SHUTDOWN))
883 break;
Gustavo A. R. Silvafcfd6df2017-10-16 15:48:55 -0500884 /* fall through */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 case TCP_SYN_SENT:
886 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
887 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
888 break;
889 }
890
891 /* Wake up anyone sleeping in poll. */
892 sk->sk_state_change(sk);
893 release_sock(sk);
894 return err;
895}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700896EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897
898/*
899 * ioctl() calls you can issue on an INET socket. Most of these are
900 * device configuration and stuff and very rarely used. Some ioctls
901 * pass on to the socket itself.
902 *
903 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
904 * loads the devconfigure module does its configuring and unloads it.
905 * There's a good 20K of config code hanging around the kernel.
906 */
907
908int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
909{
910 struct sock *sk = sock->sk;
911 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900912 struct net *net = sock_net(sk);
Al Viro03aef172017-07-01 07:53:12 -0400913 void __user *p = (void __user *)arg;
914 struct ifreq ifr;
Al Viroca25c302017-07-01 08:03:10 -0400915 struct rtentry rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
917 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700918 case SIOCGSTAMP:
919 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
920 break;
921 case SIOCGSTAMPNS:
922 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
923 break;
924 case SIOCADDRT:
925 case SIOCDELRT:
Al Viroca25c302017-07-01 08:03:10 -0400926 if (copy_from_user(&rt, p, sizeof(struct rtentry)))
927 return -EFAULT;
928 err = ip_rt_ioctl(net, cmd, &rt);
929 break;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700930 case SIOCRTMSG:
Al Viroca25c302017-07-01 08:03:10 -0400931 err = -EINVAL;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700932 break;
933 case SIOCDARP:
934 case SIOCGARP:
935 case SIOCSARP:
936 err = arp_ioctl(net, cmd, (void __user *)arg);
937 break;
938 case SIOCGIFADDR:
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700939 case SIOCGIFBRDADDR:
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700940 case SIOCGIFNETMASK:
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700941 case SIOCGIFDSTADDR:
Al Viro03aef172017-07-01 07:53:12 -0400942 case SIOCGIFPFLAGS:
943 if (copy_from_user(&ifr, p, sizeof(struct ifreq)))
944 return -EFAULT;
945 err = devinet_ioctl(net, cmd, &ifr);
946 if (!err && copy_to_user(p, &ifr, sizeof(struct ifreq)))
947 err = -EFAULT;
948 break;
949
950 case SIOCSIFADDR:
951 case SIOCSIFBRDADDR:
952 case SIOCSIFNETMASK:
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700953 case SIOCSIFDSTADDR:
954 case SIOCSIFPFLAGS:
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700955 case SIOCSIFFLAGS:
Al Viro03aef172017-07-01 07:53:12 -0400956 if (copy_from_user(&ifr, p, sizeof(struct ifreq)))
957 return -EFAULT;
958 err = devinet_ioctl(net, cmd, &ifr);
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700959 break;
960 default:
961 if (sk->sk_prot->ioctl)
962 err = sk->sk_prot->ioctl(sk, cmd, arg);
963 else
964 err = -ENOIOCTLCMD;
965 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 }
967 return err;
968}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700969EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000971#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000972static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000973{
974 struct sock *sk = sock->sk;
975 int err = -ENOIOCTLCMD;
976
977 if (sk->sk_prot->compat_ioctl)
978 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
979
980 return err;
981}
982#endif
983
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800984const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800985 .family = PF_INET,
986 .owner = THIS_MODULE,
987 .release = inet_release,
988 .bind = inet_bind,
989 .connect = inet_stream_connect,
990 .socketpair = sock_no_socketpair,
991 .accept = inet_accept,
992 .getname = inet_getname,
Linus Torvaldsa11e1d42018-06-28 09:43:44 -0700993 .poll = tcp_poll,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800994 .ioctl = inet_ioctl,
995 .listen = inet_listen,
996 .shutdown = inet_shutdown,
997 .setsockopt = sock_common_setsockopt,
998 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000999 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001000 .recvmsg = inet_recvmsg,
Eric Dumazet05255b82018-04-27 08:58:08 -07001001#ifdef CONFIG_MMU
Eric Dumazet93ab6cc2018-04-16 10:33:38 -07001002 .mmap = tcp_mmap,
Eric Dumazet05255b82018-04-27 08:58:08 -07001003#endif
Changli Gao7ba42912010-07-10 20:41:55 +00001004 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -08001005 .splice_read = tcp_splice_read,
Tom Herbert32035582016-08-28 14:43:18 -07001006 .read_sock = tcp_read_sock,
Tom Herbert306b13e2017-07-28 16:22:41 -07001007 .sendmsg_locked = tcp_sendmsg_locked,
1008 .sendpage_locked = tcp_sendpage_locked,
Tom Herbert32035582016-08-28 14:43:18 -07001009 .peek_len = tcp_peek_len,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001010#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001011 .compat_setsockopt = compat_sock_common_setsockopt,
1012 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001013 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001014#endif
Eric Dumazetd1361842018-04-16 10:33:35 -07001015 .set_rcvlowat = tcp_set_rcvlowat,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016};
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001017EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001019const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001020 .family = PF_INET,
1021 .owner = THIS_MODULE,
1022 .release = inet_release,
1023 .bind = inet_bind,
1024 .connect = inet_dgram_connect,
1025 .socketpair = sock_no_socketpair,
1026 .accept = sock_no_accept,
1027 .getname = inet_getname,
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07001028 .poll = udp_poll,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001029 .ioctl = inet_ioctl,
1030 .listen = sock_no_listen,
1031 .shutdown = inet_shutdown,
1032 .setsockopt = sock_common_setsockopt,
1033 .getsockopt = sock_common_getsockopt,
1034 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001035 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001036 .mmap = sock_no_mmap,
1037 .sendpage = inet_sendpage,
samanthakumar627d2d62016-04-05 12:41:16 -04001038 .set_peek_off = sk_set_peek_off,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001039#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001040 .compat_setsockopt = compat_sock_common_setsockopt,
1041 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001042 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001043#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044};
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001045EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
1047/*
1048 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07001049 * udp_poll
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001051static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001052 .family = PF_INET,
1053 .owner = THIS_MODULE,
1054 .release = inet_release,
1055 .bind = inet_bind,
1056 .connect = inet_dgram_connect,
1057 .socketpair = sock_no_socketpair,
1058 .accept = sock_no_accept,
1059 .getname = inet_getname,
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07001060 .poll = datagram_poll,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001061 .ioctl = inet_ioctl,
1062 .listen = sock_no_listen,
1063 .shutdown = inet_shutdown,
1064 .setsockopt = sock_common_setsockopt,
1065 .getsockopt = sock_common_getsockopt,
1066 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001067 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001068 .mmap = sock_no_mmap,
1069 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001070#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001071 .compat_setsockopt = compat_sock_common_setsockopt,
1072 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001073 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001074#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075};
1076
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001077static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 .family = PF_INET,
1079 .create = inet_create,
1080 .owner = THIS_MODULE,
1081};
1082
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083/* Upon startup we insert all the elements in inetsw_array[] into
1084 * the linked list inetsw.
1085 */
1086static struct inet_protosw inetsw_array[] =
1087{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001088 {
1089 .type = SOCK_STREAM,
1090 .protocol = IPPROTO_TCP,
1091 .prot = &tcp_prot,
1092 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001093 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001094 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001095 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001097 {
1098 .type = SOCK_DGRAM,
1099 .protocol = IPPROTO_UDP,
1100 .prot = &udp_prot,
1101 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001102 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001104
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001105 {
1106 .type = SOCK_DGRAM,
1107 .protocol = IPPROTO_ICMP,
1108 .prot = &ping_prot,
Eric Dumazet77d4b1d2017-06-03 09:29:25 -07001109 .ops = &inet_sockraw_ops,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001110 .flags = INET_PROTOSW_REUSE,
1111 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001114 .type = SOCK_RAW,
1115 .protocol = IPPROTO_IP, /* wild card */
1116 .prot = &raw_prot,
1117 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001118 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 }
1120};
1121
Denis Chengc40f6ff2007-09-16 16:39:25 -07001122#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
1124void inet_register_protosw(struct inet_protosw *p)
1125{
1126 struct list_head *lh;
1127 struct inet_protosw *answer;
1128 int protocol = p->protocol;
1129 struct list_head *last_perm;
1130
1131 spin_lock_bh(&inetsw_lock);
1132
1133 if (p->type >= SOCK_MAX)
1134 goto out_illegal;
1135
1136 /* If we are trying to override a permanent protocol, bail. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 last_perm = &inetsw[p->type];
1138 list_for_each(lh, &inetsw[p->type]) {
1139 answer = list_entry(lh, struct inet_protosw, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 /* Check only the non-wild match. */
Junwei Zhangf6c53332015-09-18 00:00:05 -04001141 if ((INET_PROTOSW_PERMANENT & answer->flags) == 0)
1142 break;
1143 if (protocol == answer->protocol)
1144 goto out_permanent;
1145 last_perm = lh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148 /* Add the new entry after the last permanent entry if any, so that
1149 * the new entry does not override a permanent entry when matched with
1150 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001151 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 * system automatically returns to the old behavior.
1153 */
1154 list_add_rcu(&p->list, last_perm);
1155out:
1156 spin_unlock_bh(&inetsw_lock);
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 return;
1159
1160out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001161 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 goto out;
1163
1164out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001165 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 p->type);
1167 goto out;
1168}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001169EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170
1171void inet_unregister_protosw(struct inet_protosw *p)
1172{
1173 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001174 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 p->protocol);
1176 } else {
1177 spin_lock_bh(&inetsw_lock);
1178 list_del_rcu(&p->list);
1179 spin_unlock_bh(&inetsw_lock);
1180
1181 synchronize_net();
1182 }
1183}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001184EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001186static int inet_sk_reselect_saddr(struct sock *sk)
1187{
1188 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e82009-10-15 06:30:45 +00001189 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e82009-10-15 06:30:45 +00001190 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001191 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001192 struct rtable *rt;
1193 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001194 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001195
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001196 inet_opt = rcu_dereference_protected(inet->inet_opt,
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001197 lockdep_sock_is_held(sk));
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001198 if (inet_opt && inet_opt->opt.srr)
1199 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001200
1201 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001202 fl4 = &inet->cork.fl.u.ip4;
1203 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001204 sk->sk_bound_dev_if, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001205 inet->inet_sport, inet->inet_dport, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001206 if (IS_ERR(rt))
1207 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001208
Changli Gaod8d1f302010-06-10 23:31:35 -07001209 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001210
David S. Miller6e869132011-05-06 16:18:04 -07001211 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001212
1213 if (new_saddr == old_saddr)
1214 return 0;
1215
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001216 if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001217 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1218 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001219 }
1220
Eric Dumazetc720c7e82009-10-15 06:30:45 +00001221 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001222
1223 /*
1224 * XXX The only one ugly spot where we need to
1225 * XXX really change the sockets identity after
1226 * XXX it has entered the hashes. -DaveM
1227 *
1228 * Besides that, it does not check for connection
1229 * uniqueness. Wait for troubles.
1230 */
Craig Gallek086c6532016-02-10 11:50:35 -05001231 return __sk_prot_rehash(sk);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001232}
1233
1234int inet_sk_rebuild_header(struct sock *sk)
1235{
1236 struct inet_sock *inet = inet_sk(sk);
1237 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001238 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001239 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001240 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001241 int err;
1242
1243 /* Route is OK, nothing to do. */
1244 if (rt)
1245 return 0;
1246
1247 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001248 rcu_read_lock();
1249 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e82009-10-15 06:30:45 +00001250 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001251 if (inet_opt && inet_opt->opt.srr)
1252 daddr = inet_opt->opt.faddr;
1253 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001254 fl4 = &inet->cork.fl.u.ip4;
1255 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001256 inet->inet_dport, inet->inet_sport,
1257 sk->sk_protocol, RT_CONN_FLAGS(sk),
1258 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001259 if (!IS_ERR(rt)) {
1260 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001261 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001262 } else {
1263 err = PTR_ERR(rt);
1264
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001265 /* Routing failed... */
1266 sk->sk_route_caps = 0;
1267 /*
1268 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1269 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1270 */
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001271 if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr ||
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001272 sk->sk_state != TCP_SYN_SENT ||
1273 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1274 (err = inet_sk_reselect_saddr(sk)) != 0)
1275 sk->sk_err_soft = -err;
1276 }
1277
1278 return err;
1279}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001280EXPORT_SYMBOL(inet_sk_rebuild_header);
1281
Yafang Shao563e0bb2017-12-20 11:12:51 +08001282void inet_sk_set_state(struct sock *sk, int state)
1283{
1284 trace_inet_sock_set_state(sk, sk->sk_state, state);
1285 sk->sk_state = state;
1286}
1287EXPORT_SYMBOL(inet_sk_set_state);
1288
1289void inet_sk_state_store(struct sock *sk, int newstate)
1290{
1291 trace_inet_sock_set_state(sk, sk->sk_state, newstate);
1292 smp_store_release(&sk->sk_state, newstate);
1293}
1294
Tom Herbertb8921ca2016-05-18 09:06:23 -07001295struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1296 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001297{
Willem de Bruijn0c19f8462017-11-21 10:22:25 -05001298 bool udpfrag = false, fixedid = false, gso_partial, encap;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001299 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001300 const struct net_offload *ops;
Willem de Bruijn0c19f8462017-11-21 10:22:25 -05001301 unsigned int offset = 0;
David S. Millerf9242b62012-06-19 18:56:21 -07001302 struct iphdr *iph;
Alexander Duyck802ab552016-04-10 21:45:03 -04001303 int proto, tot_len;
Eric Dumazet3347c962013-10-19 11:42:56 -07001304 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001305 int ihl;
1306 int id;
1307
Eric Dumazet3347c962013-10-19 11:42:56 -07001308 skb_reset_network_header(skb);
1309 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001310 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001311 goto out;
1312
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001313 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001314 ihl = iph->ihl * 4;
1315 if (ihl < sizeof(*iph))
1316 goto out;
1317
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001318 id = ntohs(iph->id);
1319 proto = iph->protocol;
1320
1321 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001322 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001323 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001324 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001325
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001326 encap = SKB_GSO_CB(skb)->encap_level > 0;
1327 if (encap)
Florian Westphal1e16aa32014-10-20 13:49:16 +02001328 features &= skb->dev->hw_enc_features;
Eric Dumazet3347c962013-10-19 11:42:56 -07001329 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001330
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001331 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001332
Herbert Xuf4c50d92006-06-22 03:02:40 -07001333 segs = ERR_PTR(-EPROTONOSUPPORT);
1334
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001335 if (!skb->encapsulation || encap) {
Willem de Bruijn0c19f8462017-11-21 10:22:25 -05001336 udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP);
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001337 fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID);
1338
1339 /* fixed ID is invalid if DF bit is not set */
Arnaldo Carvalho de Meloa5108872016-11-28 12:36:58 -03001340 if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF)))
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001341 goto out;
1342 }
Eric Dumazetdcd60772013-11-07 18:32:06 -08001343
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001344 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001345 if (likely(ops && ops->callbacks.gso_segment))
1346 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001347
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001348 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001349 goto out;
1350
Steffen Klassert07b26c92016-09-19 12:58:47 +02001351 gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL);
1352
Herbert Xuf4c50d92006-06-22 03:02:40 -07001353 skb = segs;
1354 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001355 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Willem de Bruijn0c19f8462017-11-21 10:22:25 -05001356 if (udpfrag) {
1357 iph->frag_off = htons(offset >> 3);
1358 if (skb->next)
1359 iph->frag_off |= htons(IP_MF);
1360 offset += skb->len - nhoff - ihl;
1361 tot_len = skb->len - nhoff;
1362 } else if (skb_is_gso(skb)) {
Alexander Duyck802ab552016-04-10 21:45:03 -04001363 if (!fixedid) {
1364 iph->id = htons(id);
1365 id += skb_shinfo(skb)->gso_segs;
1366 }
Steffen Klassert07b26c92016-09-19 12:58:47 +02001367
1368 if (gso_partial)
1369 tot_len = skb_shinfo(skb)->gso_size +
1370 SKB_GSO_CB(skb)->data_offset +
1371 skb->head - (unsigned char *)iph;
1372 else
1373 tot_len = skb->len - nhoff;
Alexander Duyck802ab552016-04-10 21:45:03 -04001374 } else {
1375 if (!fixedid)
1376 iph->id = htons(id++);
1377 tot_len = skb->len - nhoff;
Pravin B Shelar490ab082013-02-22 07:30:30 +00001378 }
Alexander Duyck802ab552016-04-10 21:45:03 -04001379 iph->tot_len = htons(tot_len);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001380 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001381 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001382 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001383 skb->network_header = (u8 *)iph - skb->head;
Toke Høiland-Jørgensenc56cae22018-09-13 16:43:07 +02001384 skb_reset_mac_len(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001385 } while ((skb = skb->next));
1386
1387out:
1388 return segs;
1389}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001390EXPORT_SYMBOL(inet_gso_segment);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001391
Willem de Bruijn418e8972019-02-20 10:52:12 -05001392static struct sk_buff *ipip_gso_segment(struct sk_buff *skb,
1393 netdev_features_t features)
1394{
1395 if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4))
1396 return ERR_PTR(-EINVAL);
1397
1398 return inet_gso_segment(skb, features);
1399}
1400
Paolo Abeni028e0a42018-12-14 11:51:59 +01001401INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *,
1402 struct sk_buff *));
1403INDIRECT_CALLABLE_DECLARE(struct sk_buff *udp4_gro_receive(struct list_head *,
1404 struct sk_buff *));
David Millerd4546c22018-06-24 14:13:49 +09001405struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001406{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001407 const struct net_offload *ops;
David Millerd4546c22018-06-24 14:13:49 +09001408 struct sk_buff *pp = NULL;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001409 const struct iphdr *iph;
David Millerd4546c22018-06-24 14:13:49 +09001410 struct sk_buff *p;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001411 unsigned int hlen;
1412 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001413 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001414 int flush = 1;
1415 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001416
Herbert Xua5b1cf22009-05-26 18:50:28 +00001417 off = skb_gro_offset(skb);
1418 hlen = off + sizeof(*iph);
1419 iph = skb_gro_header_fast(skb, off);
1420 if (skb_gro_header_hard(skb, hlen)) {
1421 iph = skb_gro_header_slow(skb, hlen, off);
1422 if (unlikely(!iph))
1423 goto out;
1424 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001425
David S. Millerf9242b62012-06-19 18:56:21 -07001426 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001427
1428 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001429 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001430 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001431 goto out_unlock;
1432
Herbert Xua5ad24b2009-02-08 18:00:39 +00001433 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001434 goto out_unlock;
1435
Steffen Klassert9b83e032017-04-28 10:54:32 +02001436 if (ip_is_fragment(iph))
1437 goto out_unlock;
1438
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001439 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001440 goto out_unlock;
1441
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001442 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001443 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001444 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001445
David Millerd4546c22018-06-24 14:13:49 +09001446 list_for_each_entry(p, head, list) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001447 struct iphdr *iph2;
Alexander Duyck15305452016-04-10 21:44:57 -04001448 u16 flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001449
1450 if (!NAPI_GRO_CB(p)->same_flow)
1451 continue;
1452
Jerry Chu299603e82013-12-11 20:53:45 -08001453 iph2 = (struct iphdr *)(p->data + off);
1454 /* The above works because, with the exception of the top
1455 * (inner most) layer, we only aggregate pkts with the same
1456 * hdr length so all the hdrs we'll need to verify will start
1457 * at the same offset.
1458 */
Herbert Xua5ad24b2009-02-08 18:00:39 +00001459 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001460 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1461 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001462 NAPI_GRO_CB(p)->same_flow = 0;
1463 continue;
1464 }
1465
1466 /* All fields must match except length and checksum. */
1467 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001468 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001469 (iph->tos ^ iph2->tos) |
Jerry Chubf5a7552014-01-07 10:23:19 -08001470 ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001471
1472 NAPI_GRO_CB(p)->flush |= flush;
Alexander Duyck15305452016-04-10 21:44:57 -04001473
1474 /* We need to store of the IP ID check to be included later
1475 * when we can verify that this packet does in fact belong
1476 * to a given flow.
1477 */
1478 flush_id = (u16)(id - ntohs(iph2->id));
1479
1480 /* This bit of code makes it much easier for us to identify
1481 * the cases where we are doing atomic vs non-atomic IP ID
1482 * checks. Specifically an atomic check can return IP ID
1483 * values 0 - 0xFFFF, while a non-atomic check can only
1484 * return 0 or 0xFFFF.
1485 */
1486 if (!NAPI_GRO_CB(p)->is_atomic ||
1487 !(iph->frag_off & htons(IP_DF))) {
1488 flush_id ^= NAPI_GRO_CB(p)->count;
1489 flush_id = flush_id ? 0xFFFF : 0;
1490 }
1491
1492 /* If the previous IP ID value was based on an atomic
1493 * datagram we can overwrite the value and ignore it.
1494 */
1495 if (NAPI_GRO_CB(skb)->is_atomic)
1496 NAPI_GRO_CB(p)->flush_id = flush_id;
1497 else
1498 NAPI_GRO_CB(p)->flush_id |= flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001499 }
1500
Alexander Duyck15305452016-04-10 21:44:57 -04001501 NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001502 NAPI_GRO_CB(skb)->flush |= flush;
Jerry Chu299603e82013-12-11 20:53:45 -08001503 skb_set_network_header(skb, off);
1504 /* The above will be needed by the transport layer if there is one
1505 * immediately following this IP hdr.
1506 */
1507
Eric Dumazet2c804d02014-09-30 22:12:05 -07001508 /* Note : No need to call skb_gro_postpull_rcsum() here,
1509 * as we already checked checksum over ipv4 header was 0
1510 */
Herbert Xu86911732009-01-29 14:19:50 +00001511 skb_gro_pull(skb, sizeof(*iph));
1512 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001513
Paolo Abeni028e0a42018-12-14 11:51:59 +01001514 pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive,
1515 ops->callbacks.gro_receive, head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001516
1517out_unlock:
1518 rcu_read_unlock();
1519
1520out:
Steffen Klassert5f114162017-02-15 09:39:39 +01001521 skb_gro_flush_final(skb, pp, flush);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001522
1523 return pp;
1524}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001525EXPORT_SYMBOL(inet_gro_receive);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001526
David Millerd4546c22018-06-24 14:13:49 +09001527static struct sk_buff *ipip_gro_receive(struct list_head *head,
1528 struct sk_buff *skb)
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001529{
1530 if (NAPI_GRO_CB(skb)->encap_mark) {
1531 NAPI_GRO_CB(skb)->flush = 1;
1532 return NULL;
1533 }
1534
1535 NAPI_GRO_CB(skb)->encap_mark = 1;
1536
1537 return inet_gro_receive(head, skb);
1538}
1539
Deepa Dinamani822c8682016-02-27 00:32:15 -08001540#define SECONDS_PER_DAY 86400
1541
1542/* inet_current_timestamp - Return IP network timestamp
1543 *
1544 * Return milliseconds since midnight in network byte order.
1545 */
1546__be32 inet_current_timestamp(void)
1547{
1548 u32 secs;
1549 u32 msecs;
1550 struct timespec64 ts;
1551
1552 ktime_get_real_ts64(&ts);
1553
1554 /* Get secs since midnight. */
1555 (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs);
1556 /* Convert to msecs. */
1557 msecs = secs * MSEC_PER_SEC;
1558 /* Convert nsec to msec. */
1559 msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC;
1560
1561 /* Convert to network byte order. */
Deepa Dinamani3ba9d302016-03-21 18:21:26 -07001562 return htonl(msecs);
Deepa Dinamani822c8682016-02-27 00:32:15 -08001563}
1564EXPORT_SYMBOL(inet_current_timestamp);
1565
Willem de Bruijnf4713a32014-11-26 14:53:02 -05001566int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1567{
1568 if (sk->sk_family == AF_INET)
1569 return ip_recv_error(sk, msg, len, addr_len);
1570#if IS_ENABLED(CONFIG_IPV6)
1571 if (sk->sk_family == AF_INET6)
1572 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1573#endif
1574 return -EINVAL;
1575}
1576
Paolo Abeni028e0a42018-12-14 11:51:59 +01001577INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *, int));
1578INDIRECT_CALLABLE_DECLARE(int udp4_gro_complete(struct sk_buff *, int));
Tom Herbertb8921ca2016-05-18 09:06:23 -07001579int inet_gro_complete(struct sk_buff *skb, int nhoff)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001580{
Jerry Chu299603e82013-12-11 20:53:45 -08001581 __be16 newlen = htons(skb->len - nhoff);
1582 struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001583 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001584 int proto = iph->protocol;
1585 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001586
Paolo Abeni294acf12017-03-07 18:33:31 +01001587 if (skb->encapsulation) {
1588 skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP));
Jerry Chuc3caf112014-07-14 15:54:46 -07001589 skb_set_inner_network_header(skb, nhoff);
Paolo Abeni294acf12017-03-07 18:33:31 +01001590 }
Jerry Chuc3caf112014-07-14 15:54:46 -07001591
Herbert Xu73cc19f2008-12-15 23:41:09 -08001592 csum_replace2(&iph->check, iph->tot_len, newlen);
1593 iph->tot_len = newlen;
1594
1595 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001596 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001597 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001598 goto out_unlock;
1599
Jerry Chu299603e82013-12-11 20:53:45 -08001600 /* Only need to add sizeof(*iph) to get to the next hdr below
1601 * because any hdr with option will have been flushed in
1602 * inet_gro_receive().
1603 */
Paolo Abeni028e0a42018-12-14 11:51:59 +01001604 err = INDIRECT_CALL_2(ops->callbacks.gro_complete,
1605 tcp4_gro_complete, udp4_gro_complete,
1606 skb, nhoff + sizeof(*iph));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001607
1608out_unlock:
1609 rcu_read_unlock();
1610
1611 return err;
1612}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001613EXPORT_SYMBOL(inet_gro_complete);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001614
Jesse Grossb8cba752016-03-19 09:32:00 -07001615static int ipip_gro_complete(struct sk_buff *skb, int nhoff)
1616{
1617 skb->encapsulation = 1;
Tom Herbert7e133182016-05-18 09:06:10 -07001618 skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4;
Jesse Grossb8cba752016-03-19 09:32:00 -07001619 return inet_gro_complete(skb, nhoff);
1620}
1621
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001622int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001623 unsigned short type, unsigned char protocol,
1624 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001625{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001626 struct socket *sock;
Eric W. Biederman26abe142015-05-08 21:10:31 -05001627 int rc = sock_create_kern(net, family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001628
1629 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001630 *sk = sock->sk;
1631 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001632 /*
1633 * Unhash it so that IP input processing does not even see it,
1634 * we do not wish this socket to see incoming packets.
1635 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001636 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001637 }
1638 return rc;
1639}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001640EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1641
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301642u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
1643{
1644 return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
1645}
1646EXPORT_SYMBOL_GPL(snmp_get_cpu_field);
1647
WANG Cong698365f2014-05-05 15:55:55 -07001648unsigned long snmp_fold_field(void __percpu *mib, int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001649{
1650 unsigned long res = 0;
WANG Cong698365f2014-05-05 15:55:55 -07001651 int i;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001652
WANG Cong698365f2014-05-05 15:55:55 -07001653 for_each_possible_cpu(i)
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301654 res += snmp_get_cpu_field(mib, i, offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001655 return res;
1656}
1657EXPORT_SYMBOL_GPL(snmp_fold_field);
1658
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001659#if BITS_PER_LONG==32
1660
David S. Miller80ec1922015-08-30 22:40:44 -07001661u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt,
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301662 size_t syncp_offset)
1663{
1664 void *bhptr;
1665 struct u64_stats_sync *syncp;
1666 u64 v;
1667 unsigned int start;
1668
1669 bhptr = per_cpu_ptr(mib, cpu);
1670 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1671 do {
1672 start = u64_stats_fetch_begin_irq(syncp);
1673 v = *(((u64 *)bhptr) + offt);
1674 } while (u64_stats_fetch_retry_irq(syncp, start));
1675
1676 return v;
1677}
1678EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);
1679
WANG Cong698365f2014-05-05 15:55:55 -07001680u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001681{
1682 u64 res = 0;
1683 int cpu;
1684
1685 for_each_possible_cpu(cpu) {
Madalin Bucurd1bfc622015-08-31 15:46:07 +03001686 res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001687 }
1688 return res;
1689}
1690EXPORT_SYMBOL_GPL(snmp_fold_field64);
1691#endif
1692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001694static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001696 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697};
1698#endif
1699
David Aherna8e3bb32017-08-28 15:14:20 -07001700/* thinking of making this const? Don't.
1701 * early_demux can change based on sysctl.
1702 */
David Ahern03157932017-08-28 13:23:09 -07001703static struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001704 .early_demux = tcp_v4_early_demux,
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001705 .early_demux_handler = tcp_v4_early_demux,
David S. Miller41063e92012-06-19 21:22:05 -07001706 .handler = tcp_v4_rcv,
1707 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001708 .no_policy = 1,
1709 .netns_ok = 1,
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +01001710 .icmp_strict_tag_validation = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711};
1712
David Aherna8e3bb32017-08-28 15:14:20 -07001713/* thinking of making this const? Don't.
1714 * early_demux can change based on sysctl.
1715 */
David Ahern03157932017-08-28 13:23:09 -07001716static struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001717 .early_demux = udp_v4_early_demux,
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001718 .early_demux_handler = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 .handler = udp_rcv,
1720 .err_handler = udp_err,
1721 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001722 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723};
1724
Alexey Dobriyan32613092009-09-14 12:21:47 +00001725static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001727 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001728 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001729 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730};
1731
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001732static __net_init int ipv4_mib_init_net(struct net *net)
1733{
John Stultz827da442013-10-07 15:51:58 -07001734 int i;
1735
WANG Cong698365f2014-05-05 15:55:55 -07001736 net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1737 if (!net->mib.tcp_statistics)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001738 goto err_tcp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001739 net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1740 if (!net->mib.ip_statistics)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001741 goto err_ip_mib;
John Stultz827da442013-10-07 15:51:58 -07001742
1743 for_each_possible_cpu(i) {
1744 struct ipstats_mib *af_inet_stats;
WANG Cong698365f2014-05-05 15:55:55 -07001745 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
John Stultz827da442013-10-07 15:51:58 -07001746 u64_stats_init(&af_inet_stats->syncp);
John Stultz827da442013-10-07 15:51:58 -07001747 }
1748
WANG Cong698365f2014-05-05 15:55:55 -07001749 net->mib.net_statistics = alloc_percpu(struct linux_mib);
1750 if (!net->mib.net_statistics)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001751 goto err_net_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001752 net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1753 if (!net->mib.udp_statistics)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001754 goto err_udp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001755 net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1756 if (!net->mib.udplite_statistics)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001757 goto err_udplite_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001758 net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1759 if (!net->mib.icmp_statistics)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001760 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001761 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1762 GFP_KERNEL);
1763 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001764 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001765
1766 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001767 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001768
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001769err_icmpmsg_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001770 free_percpu(net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001771err_icmp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001772 free_percpu(net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001773err_udplite_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001774 free_percpu(net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001775err_udp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001776 free_percpu(net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001777err_net_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001778 free_percpu(net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001779err_ip_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001780 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001781err_tcp_mib:
1782 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001783}
1784
1785static __net_exit void ipv4_mib_exit_net(struct net *net)
1786{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001787 kfree(net->mib.icmpmsg_statistics);
WANG Cong698365f2014-05-05 15:55:55 -07001788 free_percpu(net->mib.icmp_statistics);
1789 free_percpu(net->mib.udplite_statistics);
1790 free_percpu(net->mib.udp_statistics);
1791 free_percpu(net->mib.net_statistics);
1792 free_percpu(net->mib.ip_statistics);
1793 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001794}
1795
1796static __net_initdata struct pernet_operations ipv4_mib_ops = {
1797 .init = ipv4_mib_init_net,
1798 .exit = ipv4_mib_exit_net,
1799};
1800
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801static int __init init_ipv4_mibs(void)
1802{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001803 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804}
1805
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001806static __net_init int inet_init_net(struct net *net)
1807{
1808 /*
1809 * Set defaults for local port range
1810 */
1811 seqlock_init(&net->ipv4.ip_local_ports.lock);
1812 net->ipv4.ip_local_ports.range[0] = 32768;
Eric Dumazet07f4c902015-05-24 14:49:35 -07001813 net->ipv4.ip_local_ports.range[1] = 60999;
Cong Wangba6b9182014-05-06 11:02:50 -07001814
1815 seqlock_init(&net->ipv4.ping_group_range.lock);
1816 /*
1817 * Sane defaults - nobody may create ping sockets.
1818 * Boot scripts should set this to distro-specific group.
1819 */
1820 net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1821 net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001822
1823 /* Default values for sysctl-controlled parameters.
1824 * We set them here, in case sysctl is not compiled.
1825 */
1826 net->ipv4.sysctl_ip_default_ttl = IPDEFTTL;
Petr Machata432e05d2018-08-01 00:36:03 +02001827 net->ipv4.sysctl_ip_fwd_update_priority = 1;
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001828 net->ipv4.sysctl_ip_dynaddr = 0;
1829 net->ipv4.sysctl_ip_early_demux = 1;
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001830 net->ipv4.sysctl_udp_early_demux = 1;
1831 net->ipv4.sysctl_tcp_early_demux = 1;
Krister Johansen4548b682017-01-20 17:49:11 -08001832#ifdef CONFIG_SYSCTL
1833 net->ipv4.sysctl_ip_prot_sock = PROT_SOCK;
1834#endif
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001835
Nikolay Borisov17140202017-08-09 14:38:04 +03001836 /* Some igmp sysctl, whose values are always used */
1837 net->ipv4.sysctl_igmp_max_memberships = 20;
1838 net->ipv4.sysctl_igmp_max_msf = 10;
1839 /* IGMP reports for link-local multicast groups are enabled by default */
1840 net->ipv4.sysctl_igmp_llm_reports = 1;
1841 net->ipv4.sysctl_igmp_qrv = 2;
1842
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001843 return 0;
1844}
1845
1846static __net_exit void inet_exit_net(struct net *net)
1847{
1848}
1849
1850static __net_initdata struct pernet_operations af_inet_ops = {
1851 .init = inet_init_net,
1852 .exit = inet_exit_net,
1853};
1854
1855static int __init init_inet_pernet_ops(void)
1856{
1857 return register_pernet_subsys(&af_inet_ops);
1858}
1859
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Herbert Xu30e224d2005-07-05 14:40:10 -07001862/*
1863 * IP protocol layer initialiser
1864 */
1865
Vlad Yasevich22061d82012-11-15 08:49:11 +00001866static struct packet_offload ip_packet_offload __read_mostly = {
1867 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001868 .callbacks = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001869 .gso_segment = inet_gso_segment,
1870 .gro_receive = inet_gro_receive,
1871 .gro_complete = inet_gro_complete,
1872 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001873};
1874
Eric Dumazetcb32f512013-10-19 11:42:57 -07001875static const struct net_offload ipip_offload = {
1876 .callbacks = {
Willem de Bruijn418e8972019-02-20 10:52:12 -05001877 .gso_segment = ipip_gso_segment,
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001878 .gro_receive = ipip_gro_receive,
Jesse Grossb8cba752016-03-19 09:32:00 -07001879 .gro_complete = ipip_gro_complete,
Eric Dumazetcb32f512013-10-19 11:42:57 -07001880 },
1881};
1882
Tonghao Zhang93b1b312017-08-02 09:34:15 -07001883static int __init ipip_offload_init(void)
1884{
1885 return inet_add_offload(&ipip_offload, IPPROTO_IPIP);
1886}
1887
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001888static int __init ipv4_offload_init(void)
1889{
1890 /*
1891 * Add offloads
1892 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001893 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001894 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001895 if (tcpv4_offload_init() < 0)
1896 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Tonghao Zhang93b1b312017-08-02 09:34:15 -07001897 if (ipip_offload_init() < 0)
1898 pr_crit("%s: Cannot add IPIP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001899
1900 dev_add_offload(&ip_packet_offload);
1901 return 0;
1902}
1903
1904fs_initcall(ipv4_offload_init);
1905
1906static struct packet_type ip_packet_type __read_mostly = {
1907 .type = cpu_to_be16(ETH_P_IP),
1908 .func = ip_rcv,
Edward Cree17266ee2018-07-02 16:14:12 +01001909 .list_func = ip_list_rcv,
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001910};
1911
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912static int __init inet_init(void)
1913{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 struct inet_protosw *q;
1915 struct list_head *r;
1916 int rc = -EINVAL;
1917
Eyal Birgerb4772ef2015-03-01 14:58:29 +02001918 sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
1920 rc = proto_register(&tcp_prot, 1);
1921 if (rc)
WANG Cong122ff242014-05-12 16:04:53 -07001922 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
1924 rc = proto_register(&udp_prot, 1);
1925 if (rc)
1926 goto out_unregister_tcp_proto;
1927
1928 rc = proto_register(&raw_prot, 1);
1929 if (rc)
1930 goto out_unregister_udp_proto;
1931
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001932 rc = proto_register(&ping_prot, 1);
1933 if (rc)
1934 goto out_unregister_raw_proto;
1935
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001937 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 */
1939
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001940 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941
Al Virobd7b1532008-07-15 16:00:59 -04001942#ifdef CONFIG_SYSCTL
1943 ip_static_sysctl_init();
1944#endif
1945
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 /*
1947 * Add all the base protocols.
1948 */
1949
1950 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001951 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001953 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001955 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956#ifdef CONFIG_IP_MULTICAST
1957 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001958 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959#endif
1960
1961 /* Register the socket-side information for inet_create. */
1962 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1963 INIT_LIST_HEAD(r);
1964
1965 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1966 inet_register_protosw(q);
1967
1968 /*
1969 * Set the ARP module up
1970 */
1971
1972 arp_init();
1973
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001974 /*
1975 * Set the IP module up
1976 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977
1978 ip_init();
1979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 /* Setup TCP slab cache for open requests. */
1981 tcp_init();
1982
Hideo Aoki95766ff2007-12-31 00:29:24 -08001983 /* Setup UDP memory threshold */
1984 udp_init();
1985
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001986 /* Add UDP-Lite (RFC 3828) */
1987 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
Mike Manning68974452018-11-07 15:36:05 +00001989 raw_init();
1990
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001991 ping_init();
1992
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 /*
1994 * Set the ICMP layer up
1995 */
1996
Denis V. Luneva5710d62008-02-29 11:14:50 -08001997 if (icmp_init() < 0)
1998 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
2000 /*
2001 * Initialise the multicast router
2002 */
2003#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08002004 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00002005 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006#endif
Cong Wangc9d8f1a2014-05-06 11:02:49 -07002007
2008 if (init_inet_pernet_ops())
2009 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010 /*
2011 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002012 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013
Stephen Hemminger132adf52007-03-08 20:44:43 -08002014 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00002015 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002016
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 ipv4_proc_init();
2018
2019 ipfrag_init();
2020
Herbert Xu30e224d2005-07-05 14:40:10 -07002021 dev_add_pack(&ip_packet_type);
2022
Thomas Graf045a0fa2015-07-23 10:08:44 +02002023 ip_tunnel_core_init();
2024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 rc = 0;
2026out:
2027 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00002028out_unregister_raw_proto:
2029 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030out_unregister_udp_proto:
2031 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07002032out_unregister_tcp_proto:
2033 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 goto out;
2035}
2036
Heiko Carstensa536e072006-04-28 15:19:17 -07002037fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
2039/* ------------------------------------------------------------------------ */
2040
2041#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042static int __init ipv4_proc_init(void)
2043{
2044 int rc = 0;
2045
2046 if (raw_proc_init())
2047 goto out_raw;
2048 if (tcp4_proc_init())
2049 goto out_tcp;
2050 if (udp4_proc_init())
2051 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00002052 if (ping_proc_init())
2053 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 if (ip_misc_proc_init())
2055 goto out_misc;
2056out:
2057 return rc;
2058out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00002059 ping_proc_exit();
2060out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 udp4_proc_exit();
2062out_udp:
2063 tcp4_proc_exit();
2064out_tcp:
2065 raw_proc_exit();
2066out_raw:
2067 rc = -ENOMEM;
2068 goto out;
2069}
2070
2071#else /* CONFIG_PROC_FS */
2072static int __init ipv4_proc_init(void)
2073{
2074 return 0;
2075}
2076#endif /* CONFIG_PROC_FS */