blob: 5ddf5cda07f4173b0c110c034390428fe2928097 [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
93#include <asm/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/inet.h>
96#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020097#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Herbert Xu73cc19f2008-12-15 23:41:09 -080099#include <net/checksum.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#include <net/ip.h>
101#include <net/protocol.h>
102#include <net/arp.h>
103#include <net/route.h>
104#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700105#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/tcp.h>
107#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800108#include <net/udplite.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000109#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110#include <linux/skbuff.h>
111#include <net/sock.h>
112#include <net/raw.h>
113#include <net/icmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#include <net/inet_common.h>
Thomas Graf045a0fa2015-07-23 10:08:44 +0200115#include <net/ip_tunnels.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700117#include <net/net_namespace.h>
Eric Dumazetaebda152013-04-29 05:58:52 +0000118#include <net/secure_seq.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119#ifdef CONFIG_IP_MROUTE
120#include <linux/mroute.h>
121#endif
David Ahern3236b002015-09-29 20:07:14 -0700122#include <net/l3mdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125/* The inetsw table contains everything that inet_create needs to
126 * build a new socket.
127 */
128static struct list_head inetsw[SOCK_MAX];
129static DEFINE_SPINLOCK(inetsw_lock);
130
131/* New destruction routine */
132
133void inet_sock_destruct(struct sock *sk)
134{
135 struct inet_sock *inet = inet_sk(sk);
136
137 __skb_queue_purge(&sk->sk_receive_queue);
138 __skb_queue_purge(&sk->sk_error_queue);
139
Hideo Aoki95766ff2007-12-31 00:29:24 -0800140 sk_mem_reclaim(sk);
141
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700143 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 sk->sk_state, sk);
145 return;
146 }
147 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700148 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 return;
150 }
151
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700152 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
153 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
154 WARN_ON(sk->sk_wmem_queued);
155 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000157 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000158 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700159 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700160 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700162EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163
164/*
165 * The routines beyond this point handle the behaviour of an AF_INET
166 * socket object. Mostly it punts to the subprotocols of IP to do
167 * the work.
168 */
169
170/*
171 * Automatically bind an unbound socket.
172 */
173
174static int inet_autobind(struct sock *sk)
175{
176 struct inet_sock *inet;
177 /* We may need to bind the socket. */
178 lock_sock(sk);
179 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000180 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 if (sk->sk_prot->get_port(sk, 0)) {
182 release_sock(sk);
183 return -EAGAIN;
184 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000185 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 }
187 release_sock(sk);
188 return 0;
189}
190
191/*
192 * Move a socket into listening state.
193 */
194int inet_listen(struct socket *sock, int backlog)
195{
196 struct sock *sk = sock->sk;
197 unsigned char old_state;
198 int err;
199
200 lock_sock(sk);
201
202 err = -EINVAL;
203 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
204 goto out;
205
206 old_state = sk->sk_state;
207 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
208 goto out;
209
210 /* Really, if the socket is already in listen state
211 * we can only allow the backlog to be adjusted.
212 */
213 if (old_state != TCP_LISTEN) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700214 /* Enable TFO w/o requiring TCP_FASTOPEN socket option.
Jerry Chu83368862012-08-31 12:29:12 +0000215 * Note that only TCP sockets (SOCK_STREAM) will reach here.
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700216 * Also fastopen backlog may already been set via the option
217 * because the socket was in TCP_LISTEN state previously but
218 * was shutdown() rather than close().
Jerry Chu83368862012-08-31 12:29:12 +0000219 */
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700220 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
221 (sysctl_tcp_fastopen & TFO_SERVER_ENABLE) &&
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700222 !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700223 fastopen_queue_tune(sk, backlog);
Christoph Paaschdfea2aa2015-06-18 09:15:34 -0700224 tcp_fastopen_init_key_once(true);
Jerry Chu83368862012-08-31 12:29:12 +0000225 }
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700226
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800227 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 if (err)
229 goto out;
230 }
231 sk->sk_max_ack_backlog = backlog;
232 err = 0;
233
234out:
235 release_sock(sk);
236 return err;
237}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700238EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240/*
241 * Create an inet socket.
242 */
243
Eric Paris3f378b62009-11-05 22:18:14 -0800244static int inet_create(struct net *net, struct socket *sock, int protocol,
245 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246{
247 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 struct inet_protosw *answer;
249 struct inet_sock *inet;
250 struct proto *answer_prot;
251 unsigned char answer_flags;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700252 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800253 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
Hannes Frederic Sowa79462ad02015-12-14 22:03:39 +0100255 if (protocol < 0 || protocol >= IPPROTO_MAX)
256 return -EINVAL;
257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 sock->state = SS_UNCONNECTED;
259
260 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700261lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800262 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700264 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700266 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 /* Check the non-wild match. */
268 if (protocol == answer->protocol) {
269 if (protocol != IPPROTO_IP)
270 break;
271 } else {
272 /* Check for the two wild cases. */
273 if (IPPROTO_IP == protocol) {
274 protocol = answer->protocol;
275 break;
276 }
277 if (IPPROTO_IP == answer->protocol)
278 break;
279 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800280 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 }
282
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700283 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700284 if (try_loading_module < 2) {
285 rcu_read_unlock();
286 /*
287 * Be more specific, e.g. net-pf-2-proto-132-type-1
288 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
289 */
290 if (++try_loading_module == 1)
291 request_module("net-pf-%d-proto-%d-type-%d",
292 PF_INET, protocol, sock->type);
293 /*
294 * Fall back to generic, e.g. net-pf-2-proto-132
295 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
296 */
297 else
298 request_module("net-pf-%d-proto-%d",
299 PF_INET, protocol);
300 goto lookup_protocol;
301 } else
302 goto out_rcu_unlock;
303 }
304
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000306 if (sock->type == SOCK_RAW && !kern &&
307 !ns_capable(net->user_ns, CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
310 sock->ops = answer->ops;
311 answer_prot = answer->prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 answer_flags = answer->flags;
313 rcu_read_unlock();
314
Ian Morris51456b22015-04-03 09:17:26 +0100315 WARN_ON(!answer_prot->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
317 err = -ENOBUFS;
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500318 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
Ian Morris51456b22015-04-03 09:17:26 +0100319 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 goto out;
321
322 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000324 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
326 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800327 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000329 inet->nodefrag = 0;
330
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000332 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 if (IPPROTO_RAW == protocol)
334 inet->hdrincl = 1;
335 }
336
Hannes Frederic Sowa974eda12013-12-14 05:13:38 +0100337 if (net->ipv4.sysctl_ip_no_pmtu_disc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 inet->pmtudisc = IP_PMTUDISC_DONT;
339 else
340 inet->pmtudisc = IP_PMTUDISC_WANT;
341
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000342 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
344 sock_init_data(sock, sk);
345
346 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 sk->sk_protocol = protocol;
348 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
349
350 inet->uc_ttl = -1;
351 inet->mc_loop = 1;
352 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000353 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 inet->mc_index = 0;
355 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000356 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700358 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000360 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 /* It assumes that any protocol which allows
362 * the user to assign a number at socket
363 * creation time automatically
364 * shares.
365 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000366 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 /* Add to protocol hash chains. */
Craig Gallek086c6532016-02-10 11:50:35 -0500368 err = sk->sk_prot->hash(sk);
369 if (err) {
370 sk_common_release(sk);
371 goto out;
372 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 }
374
375 if (sk->sk_prot->init) {
376 err = sk->sk_prot->init(sk);
377 if (err)
378 sk_common_release(sk);
379 }
380out:
381 return err;
382out_rcu_unlock:
383 rcu_read_unlock();
384 goto out;
385}
386
387
388/*
389 * The peer socket should always be NULL (or else). When we call this
390 * function we are destroying the object and from then on nobody
391 * should refer to it.
392 */
393int inet_release(struct socket *sock)
394{
395 struct sock *sk = sock->sk;
396
397 if (sk) {
398 long timeout;
399
400 /* Applications forget to leave groups before exiting */
401 ip_mc_drop_socket(sk);
402
403 /* If linger is set, we don't return until the close
404 * is complete. Otherwise we return immediately. The
405 * actually closing is done the same either way.
406 *
407 * If the close is due to the process exiting, we never
408 * linger..
409 */
410 timeout = 0;
411 if (sock_flag(sk, SOCK_LINGER) &&
412 !(current->flags & PF_EXITING))
413 timeout = sk->sk_lingertime;
414 sock->sk = NULL;
415 sk->sk_prot->close(sk, timeout);
416 }
417 return 0;
418}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700419EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
422{
423 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
424 struct sock *sk = sock->sk;
425 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000426 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 unsigned short snum;
428 int chk_addr_ret;
David Ahern9b8ff512015-09-01 14:26:35 -0600429 u32 tb_id = RT_TABLE_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 int err;
431
432 /* If the socket has its own bind function then use it. (RAW) */
433 if (sk->sk_prot->bind) {
434 err = sk->sk_prot->bind(sk, uaddr, addr_len);
435 goto out;
436 }
437 err = -EINVAL;
438 if (addr_len < sizeof(struct sockaddr_in))
439 goto out;
440
Marcus Meissnerc349a522011-07-04 01:30:29 +0000441 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400442 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
443 * only if s_addr is INADDR_ANY.
444 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000445 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400446 if (addr->sin_family != AF_UNSPEC ||
447 addr->sin_addr.s_addr != htonl(INADDR_ANY))
448 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000449 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700450
David Ahern3236b002015-09-29 20:07:14 -0700451 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
David Ahern30bbaa12015-08-13 14:59:05 -0600452 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454 /* Not specified by any standard per-se, however it breaks too
455 * many applications when removed. It is unfortunate since
456 * allowing applications to make a non-local bind solves
457 * several problems with systems using dynamic addressing.
458 * (ie. your servers still start up even if your ISDN link
459 * is temporarily down)
460 */
461 err = -EADDRNOTAVAIL;
Vincent Bernat49a60152014-09-05 15:09:03 +0200462 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700463 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700464 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 chk_addr_ret != RTN_LOCAL &&
466 chk_addr_ret != RTN_MULTICAST &&
467 chk_addr_ret != RTN_BROADCAST)
468 goto out;
469
470 snum = ntohs(addr->sin_port);
471 err = -EACCES;
Eric W. Biederman35946982012-11-16 03:03:12 +0000472 if (snum && snum < PROT_SOCK &&
473 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto out;
475
476 /* We keep a pair of addresses. rcv_saddr is the one
477 * used by hash lookups, and saddr is used for transmit.
478 *
479 * In the BSD API these are the same except where it
480 * would be illegal to use them (multicast/broadcast) in
481 * which case the sending device address is used.
482 */
483 lock_sock(sk);
484
485 /* Check these errors (active socket, double bind). */
486 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000487 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 goto out_release_sock;
489
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000490 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000492 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
494 /* Make sure we are allowed to bind here. */
Eric Dumazet90c337d2015-06-06 21:17:57 -0700495 if ((snum || !inet->bind_address_no_port) &&
496 sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000497 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 err = -EADDRINUSE;
499 goto out_release_sock;
500 }
501
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000502 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
504 if (snum)
505 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000506 inet->inet_sport = htons(inet->inet_num);
507 inet->inet_daddr = 0;
508 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 sk_dst_reset(sk);
510 err = 0;
511out_release_sock:
512 release_sock(sk);
513out:
514 return err;
515}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700516EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000518int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 int addr_len, int flags)
520{
521 struct sock *sk = sock->sk;
522
Changli Gao6503d962010-03-31 22:58:26 +0000523 if (addr_len < sizeof(uaddr->sa_family))
524 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (uaddr->sa_family == AF_UNSPEC)
526 return sk->sk_prot->disconnect(sk, flags);
527
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000528 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000530 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700532EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Yuchung Cheng783237e2012-07-19 06:43:07 +0000534static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535{
WANG Cong14135f32016-11-01 16:04:36 -0700536 DEFINE_WAIT_FUNC(wait, woken_wake_function);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
WANG Cong14135f32016-11-01 16:04:36 -0700538 add_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000539 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
541 /* Basic assumption: if someone sets sk->sk_err, he _must_
542 * change state of the socket from TCP_SYN_*.
543 * Connect() does not allow to get error notifications
544 * without closing the socket.
545 */
546 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
547 release_sock(sk);
WANG Cong14135f32016-11-01 16:04:36 -0700548 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 lock_sock(sk);
550 if (signal_pending(current) || !timeo)
551 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 }
WANG Cong14135f32016-11-01 16:04:36 -0700553 remove_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000554 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return timeo;
556}
557
558/*
559 * Connect to a remote host. There is regrettably still a little
560 * TCP 'magic' in here.
561 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000562int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
563 int addr_len, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564{
565 struct sock *sk = sock->sk;
566 int err;
567 long timeo;
568
Changli Gao6503d962010-03-31 22:58:26 +0000569 if (addr_len < sizeof(uaddr->sa_family))
570 return -EINVAL;
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 if (uaddr->sa_family == AF_UNSPEC) {
573 err = sk->sk_prot->disconnect(sk, flags);
574 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
575 goto out;
576 }
577
578 switch (sock->state) {
579 default:
580 err = -EINVAL;
581 goto out;
582 case SS_CONNECTED:
583 err = -EISCONN;
584 goto out;
585 case SS_CONNECTING:
586 err = -EALREADY;
587 /* Fall out of switch with err, set for this state */
588 break;
589 case SS_UNCONNECTED:
590 err = -EISCONN;
591 if (sk->sk_state != TCP_CLOSE)
592 goto out;
593
594 err = sk->sk_prot->connect(sk, uaddr, addr_len);
595 if (err < 0)
596 goto out;
597
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900598 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
600 /* Just entered SS_CONNECTING state; the only
601 * difference is that return value in non-blocking
602 * case is EINPROGRESS, rather than EALREADY.
603 */
604 err = -EINPROGRESS;
605 break;
606 }
607
608 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
609
610 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000611 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
612 tcp_sk(sk)->fastopen_req &&
613 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000616 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 goto out;
618
619 err = sock_intr_errno(timeo);
620 if (signal_pending(current))
621 goto out;
622 }
623
624 /* Connection was closed by RST, timeout, ICMP error
625 * or another process disconnected us.
626 */
627 if (sk->sk_state == TCP_CLOSE)
628 goto sock_error;
629
630 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
631 * and error was received after socket entered established state.
632 * Hence, it is handled normally after connect() return successfully.
633 */
634
635 sock->state = SS_CONNECTED;
636 err = 0;
637out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 return err;
639
640sock_error:
641 err = sock_error(sk) ? : -ECONNABORTED;
642 sock->state = SS_UNCONNECTED;
643 if (sk->sk_prot->disconnect(sk, flags))
644 sock->state = SS_DISCONNECTING;
645 goto out;
646}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000647EXPORT_SYMBOL(__inet_stream_connect);
648
649int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
650 int addr_len, int flags)
651{
652 int err;
653
654 lock_sock(sock->sk);
655 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
656 release_sock(sock->sk);
657 return err;
658}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700659EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
661/*
662 * Accept a pending connection. The TCP layer now gives BSD semantics.
663 */
664
665int inet_accept(struct socket *sock, struct socket *newsock, int flags)
666{
667 struct sock *sk1 = sock->sk;
668 int err = -EINVAL;
669 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
670
671 if (!sk2)
672 goto do_err;
673
674 lock_sock(sk2);
675
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000676 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700677 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000678 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
679 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
681 sock_graft(sk2, newsock);
682
683 newsock->state = SS_CONNECTED;
684 err = 0;
685 release_sock(sk2);
686do_err:
687 return err;
688}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700689EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
691
692/*
693 * This does both peername and sockname.
694 */
695int inet_getname(struct socket *sock, struct sockaddr *uaddr,
696 int *uaddr_len, int peer)
697{
698 struct sock *sk = sock->sk;
699 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700700 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
702 sin->sin_family = AF_INET;
703 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000704 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
706 peer == 1))
707 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000708 sin->sin_port = inet->inet_dport;
709 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000711 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000713 addr = inet->inet_saddr;
714 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 sin->sin_addr.s_addr = addr;
716 }
717 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
718 *uaddr_len = sizeof(*sin);
719 return 0;
720}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700721EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
Ying Xue1b784142015-03-02 15:37:48 +0800723int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724{
725 struct sock *sk = sock->sk;
726
David S. Millerc58dc012010-04-27 15:05:31 -0700727 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000730 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
731 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 return -EAGAIN;
733
Ying Xue1b784142015-03-02 15:37:48 +0800734 return sk->sk_prot->sendmsg(sk, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700736EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
Changli Gao7ba42912010-07-10 20:41:55 +0000738ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
739 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740{
741 struct sock *sk = sock->sk;
742
David S. Millerc58dc012010-04-27 15:05:31 -0700743 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000746 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
747 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 return -EAGAIN;
749
750 if (sk->sk_prot->sendpage)
751 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
752 return sock_no_sendpage(sock, page, offset, size, flags);
753}
Changli Gao7ba42912010-07-10 20:41:55 +0000754EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755
Ying Xue1b784142015-03-02 15:37:48 +0800756int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
757 int flags)
Tom Herbertfec5e652010-04-16 16:01:27 -0700758{
759 struct sock *sk = sock->sk;
760 int addr_len = 0;
761 int err;
762
David S. Millerc58dc012010-04-27 15:05:31 -0700763 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700764
Ying Xue1b784142015-03-02 15:37:48 +0800765 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Tom Herbertfec5e652010-04-16 16:01:27 -0700766 flags & ~MSG_DONTWAIT, &addr_len);
767 if (err >= 0)
768 msg->msg_namelen = addr_len;
769 return err;
770}
771EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
773int inet_shutdown(struct socket *sock, int how)
774{
775 struct sock *sk = sock->sk;
776 int err = 0;
777
778 /* This should really check to make sure
779 * the socket is a TCP socket. (WHY AC...)
780 */
781 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
782 1->2 bit 2 snds.
783 2->3 */
784 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
785 return -EINVAL;
786
787 lock_sock(sk);
788 if (sock->state == SS_CONNECTING) {
789 if ((1 << sk->sk_state) &
790 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
791 sock->state = SS_DISCONNECTING;
792 else
793 sock->state = SS_CONNECTED;
794 }
795
796 switch (sk->sk_state) {
797 case TCP_CLOSE:
798 err = -ENOTCONN;
799 /* Hack to wake up other listeners, who can poll for
800 POLLHUP, even on eg. unconnected UDP sockets -- RR */
801 default:
802 sk->sk_shutdown |= how;
803 if (sk->sk_prot->shutdown)
804 sk->sk_prot->shutdown(sk, how);
805 break;
806
807 /* Remaining two branches are temporary solution for missing
808 * close() in multithreaded environment. It is _not_ a good idea,
809 * but we have no choice until close() is repaired at VFS level.
810 */
811 case TCP_LISTEN:
812 if (!(how & RCV_SHUTDOWN))
813 break;
814 /* Fall through */
815 case TCP_SYN_SENT:
816 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
817 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
818 break;
819 }
820
821 /* Wake up anyone sleeping in poll. */
822 sk->sk_state_change(sk);
823 release_sock(sk);
824 return err;
825}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700826EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828/*
829 * ioctl() calls you can issue on an INET socket. Most of these are
830 * device configuration and stuff and very rarely used. Some ioctls
831 * pass on to the socket itself.
832 *
833 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
834 * loads the devconfigure module does its configuring and unloads it.
835 * There's a good 20K of config code hanging around the kernel.
836 */
837
838int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
839{
840 struct sock *sk = sock->sk;
841 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900842 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
844 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700845 case SIOCGSTAMP:
846 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
847 break;
848 case SIOCGSTAMPNS:
849 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
850 break;
851 case SIOCADDRT:
852 case SIOCDELRT:
853 case SIOCRTMSG:
854 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
855 break;
856 case SIOCDARP:
857 case SIOCGARP:
858 case SIOCSARP:
859 err = arp_ioctl(net, cmd, (void __user *)arg);
860 break;
861 case SIOCGIFADDR:
862 case SIOCSIFADDR:
863 case SIOCGIFBRDADDR:
864 case SIOCSIFBRDADDR:
865 case SIOCGIFNETMASK:
866 case SIOCSIFNETMASK:
867 case SIOCGIFDSTADDR:
868 case SIOCSIFDSTADDR:
869 case SIOCSIFPFLAGS:
870 case SIOCGIFPFLAGS:
871 case SIOCSIFFLAGS:
872 err = devinet_ioctl(net, cmd, (void __user *)arg);
873 break;
874 default:
875 if (sk->sk_prot->ioctl)
876 err = sk->sk_prot->ioctl(sk, cmd, arg);
877 else
878 err = -ENOIOCTLCMD;
879 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 }
881 return err;
882}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700883EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000885#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000886static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000887{
888 struct sock *sk = sock->sk;
889 int err = -ENOIOCTLCMD;
890
891 if (sk->sk_prot->compat_ioctl)
892 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
893
894 return err;
895}
896#endif
897
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800898const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800899 .family = PF_INET,
900 .owner = THIS_MODULE,
901 .release = inet_release,
902 .bind = inet_bind,
903 .connect = inet_stream_connect,
904 .socketpair = sock_no_socketpair,
905 .accept = inet_accept,
906 .getname = inet_getname,
907 .poll = tcp_poll,
908 .ioctl = inet_ioctl,
909 .listen = inet_listen,
910 .shutdown = inet_shutdown,
911 .setsockopt = sock_common_setsockopt,
912 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000913 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700914 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800915 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000916 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800917 .splice_read = tcp_splice_read,
Tom Herbert32035582016-08-28 14:43:18 -0700918 .read_sock = tcp_read_sock,
919 .peek_len = tcp_peek_len,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800920#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800921 .compat_setsockopt = compat_sock_common_setsockopt,
922 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000923 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800924#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700926EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800928const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800929 .family = PF_INET,
930 .owner = THIS_MODULE,
931 .release = inet_release,
932 .bind = inet_bind,
933 .connect = inet_dgram_connect,
934 .socketpair = sock_no_socketpair,
935 .accept = sock_no_accept,
936 .getname = inet_getname,
937 .poll = udp_poll,
938 .ioctl = inet_ioctl,
939 .listen = sock_no_listen,
940 .shutdown = inet_shutdown,
941 .setsockopt = sock_common_setsockopt,
942 .getsockopt = sock_common_getsockopt,
943 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700944 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800945 .mmap = sock_no_mmap,
946 .sendpage = inet_sendpage,
samanthakumar627d2d62016-04-05 12:41:16 -0400947 .set_peek_off = sk_set_peek_off,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800948#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800949 .compat_setsockopt = compat_sock_common_setsockopt,
950 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000951 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800952#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700954EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
956/*
957 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
958 * udp_poll
959 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800960static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800961 .family = PF_INET,
962 .owner = THIS_MODULE,
963 .release = inet_release,
964 .bind = inet_bind,
965 .connect = inet_dgram_connect,
966 .socketpair = sock_no_socketpair,
967 .accept = sock_no_accept,
968 .getname = inet_getname,
969 .poll = datagram_poll,
970 .ioctl = inet_ioctl,
971 .listen = sock_no_listen,
972 .shutdown = inet_shutdown,
973 .setsockopt = sock_common_setsockopt,
974 .getsockopt = sock_common_getsockopt,
975 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700976 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800977 .mmap = sock_no_mmap,
978 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800979#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800980 .compat_setsockopt = compat_sock_common_setsockopt,
981 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000982 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800983#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984};
985
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +0000986static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 .family = PF_INET,
988 .create = inet_create,
989 .owner = THIS_MODULE,
990};
991
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992/* Upon startup we insert all the elements in inetsw_array[] into
993 * the linked list inetsw.
994 */
995static struct inet_protosw inetsw_array[] =
996{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900997 {
998 .type = SOCK_STREAM,
999 .protocol = IPPROTO_TCP,
1000 .prot = &tcp_prot,
1001 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001002 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001003 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001004 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001006 {
1007 .type = SOCK_DGRAM,
1008 .protocol = IPPROTO_UDP,
1009 .prot = &udp_prot,
1010 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001011 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001013
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001014 {
1015 .type = SOCK_DGRAM,
1016 .protocol = IPPROTO_ICMP,
1017 .prot = &ping_prot,
1018 .ops = &inet_dgram_ops,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001019 .flags = INET_PROTOSW_REUSE,
1020 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
1022 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001023 .type = SOCK_RAW,
1024 .protocol = IPPROTO_IP, /* wild card */
1025 .prot = &raw_prot,
1026 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001027 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 }
1029};
1030
Denis Chengc40f6ff2007-09-16 16:39:25 -07001031#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
1033void inet_register_protosw(struct inet_protosw *p)
1034{
1035 struct list_head *lh;
1036 struct inet_protosw *answer;
1037 int protocol = p->protocol;
1038 struct list_head *last_perm;
1039
1040 spin_lock_bh(&inetsw_lock);
1041
1042 if (p->type >= SOCK_MAX)
1043 goto out_illegal;
1044
1045 /* If we are trying to override a permanent protocol, bail. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 last_perm = &inetsw[p->type];
1047 list_for_each(lh, &inetsw[p->type]) {
1048 answer = list_entry(lh, struct inet_protosw, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 /* Check only the non-wild match. */
Junwei Zhangf6c53332015-09-18 00:00:05 -04001050 if ((INET_PROTOSW_PERMANENT & answer->flags) == 0)
1051 break;
1052 if (protocol == answer->protocol)
1053 goto out_permanent;
1054 last_perm = lh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
1057 /* Add the new entry after the last permanent entry if any, so that
1058 * the new entry does not override a permanent entry when matched with
1059 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001060 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 * system automatically returns to the old behavior.
1062 */
1063 list_add_rcu(&p->list, last_perm);
1064out:
1065 spin_unlock_bh(&inetsw_lock);
1066
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 return;
1068
1069out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001070 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 goto out;
1072
1073out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001074 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 p->type);
1076 goto out;
1077}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001078EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079
1080void inet_unregister_protosw(struct inet_protosw *p)
1081{
1082 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001083 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 p->protocol);
1085 } else {
1086 spin_lock_bh(&inetsw_lock);
1087 list_del_rcu(&p->list);
1088 spin_unlock_bh(&inetsw_lock);
1089
1090 synchronize_net();
1091 }
1092}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001093EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001095static int inet_sk_reselect_saddr(struct sock *sk)
1096{
1097 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001098 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001099 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001100 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001101 struct rtable *rt;
1102 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001103 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001104
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001105 inet_opt = rcu_dereference_protected(inet->inet_opt,
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001106 lockdep_sock_is_held(sk));
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001107 if (inet_opt && inet_opt->opt.srr)
1108 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001109
1110 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001111 fl4 = &inet->cork.fl.u.ip4;
1112 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001113 sk->sk_bound_dev_if, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001114 inet->inet_sport, inet->inet_dport, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001115 if (IS_ERR(rt))
1116 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001117
Changli Gaod8d1f302010-06-10 23:31:35 -07001118 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001119
David S. Miller6e869132011-05-06 16:18:04 -07001120 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001121
1122 if (new_saddr == old_saddr)
1123 return 0;
1124
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001125 if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001126 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1127 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001128 }
1129
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001130 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001131
1132 /*
1133 * XXX The only one ugly spot where we need to
1134 * XXX really change the sockets identity after
1135 * XXX it has entered the hashes. -DaveM
1136 *
1137 * Besides that, it does not check for connection
1138 * uniqueness. Wait for troubles.
1139 */
Craig Gallek086c6532016-02-10 11:50:35 -05001140 return __sk_prot_rehash(sk);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001141}
1142
1143int inet_sk_rebuild_header(struct sock *sk)
1144{
1145 struct inet_sock *inet = inet_sk(sk);
1146 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001147 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001148 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001149 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001150 int err;
1151
1152 /* Route is OK, nothing to do. */
1153 if (rt)
1154 return 0;
1155
1156 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001157 rcu_read_lock();
1158 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001159 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001160 if (inet_opt && inet_opt->opt.srr)
1161 daddr = inet_opt->opt.faddr;
1162 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001163 fl4 = &inet->cork.fl.u.ip4;
1164 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001165 inet->inet_dport, inet->inet_sport,
1166 sk->sk_protocol, RT_CONN_FLAGS(sk),
1167 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001168 if (!IS_ERR(rt)) {
1169 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001170 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001171 } else {
1172 err = PTR_ERR(rt);
1173
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001174 /* Routing failed... */
1175 sk->sk_route_caps = 0;
1176 /*
1177 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1178 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1179 */
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001180 if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr ||
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001181 sk->sk_state != TCP_SYN_SENT ||
1182 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1183 (err = inet_sk_reselect_saddr(sk)) != 0)
1184 sk->sk_err_soft = -err;
1185 }
1186
1187 return err;
1188}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001189EXPORT_SYMBOL(inet_sk_rebuild_header);
1190
Tom Herbertb8921ca2016-05-18 09:06:23 -07001191struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1192 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001193{
Steffen Klassert07b26c92016-09-19 12:58:47 +02001194 bool udpfrag = false, fixedid = false, gso_partial, encap;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001195 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001196 const struct net_offload *ops;
Eric Dumazet3347c962013-10-19 11:42:56 -07001197 unsigned int offset = 0;
David S. Millerf9242b62012-06-19 18:56:21 -07001198 struct iphdr *iph;
Alexander Duyck802ab552016-04-10 21:45:03 -04001199 int proto, tot_len;
Eric Dumazet3347c962013-10-19 11:42:56 -07001200 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001201 int ihl;
1202 int id;
1203
Eric Dumazet3347c962013-10-19 11:42:56 -07001204 skb_reset_network_header(skb);
1205 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001206 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001207 goto out;
1208
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001209 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001210 ihl = iph->ihl * 4;
1211 if (ihl < sizeof(*iph))
1212 goto out;
1213
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001214 id = ntohs(iph->id);
1215 proto = iph->protocol;
1216
1217 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001218 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001219 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001220 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001221
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001222 encap = SKB_GSO_CB(skb)->encap_level > 0;
1223 if (encap)
Florian Westphal1e16aa32014-10-20 13:49:16 +02001224 features &= skb->dev->hw_enc_features;
Eric Dumazet3347c962013-10-19 11:42:56 -07001225 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001226
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001227 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001228
Herbert Xuf4c50d92006-06-22 03:02:40 -07001229 segs = ERR_PTR(-EPROTONOSUPPORT);
1230
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001231 if (!skb->encapsulation || encap) {
1232 udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP);
1233 fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID);
1234
1235 /* fixed ID is invalid if DF bit is not set */
1236 if (fixedid && !(iph->frag_off & htons(IP_DF)))
1237 goto out;
1238 }
Eric Dumazetdcd60772013-11-07 18:32:06 -08001239
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001240 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001241 if (likely(ops && ops->callbacks.gso_segment))
1242 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001243
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001244 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001245 goto out;
1246
Steffen Klassert07b26c92016-09-19 12:58:47 +02001247 gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL);
1248
Herbert Xuf4c50d92006-06-22 03:02:40 -07001249 skb = segs;
1250 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001251 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001252 if (udpfrag) {
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001253 iph->frag_off = htons(offset >> 3);
Ian Morris00db4122015-04-03 09:17:27 +01001254 if (skb->next)
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001255 iph->frag_off |= htons(IP_MF);
Eric Dumazet3347c962013-10-19 11:42:56 -07001256 offset += skb->len - nhoff - ihl;
Alexander Duyck802ab552016-04-10 21:45:03 -04001257 tot_len = skb->len - nhoff;
1258 } else if (skb_is_gso(skb)) {
1259 if (!fixedid) {
1260 iph->id = htons(id);
1261 id += skb_shinfo(skb)->gso_segs;
1262 }
Steffen Klassert07b26c92016-09-19 12:58:47 +02001263
1264 if (gso_partial)
1265 tot_len = skb_shinfo(skb)->gso_size +
1266 SKB_GSO_CB(skb)->data_offset +
1267 skb->head - (unsigned char *)iph;
1268 else
1269 tot_len = skb->len - nhoff;
Alexander Duyck802ab552016-04-10 21:45:03 -04001270 } else {
1271 if (!fixedid)
1272 iph->id = htons(id++);
1273 tot_len = skb->len - nhoff;
Pravin B Shelar490ab082013-02-22 07:30:30 +00001274 }
Alexander Duyck802ab552016-04-10 21:45:03 -04001275 iph->tot_len = htons(tot_len);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001276 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001277 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001278 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001279 skb->network_header = (u8 *)iph - skb->head;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001280 } while ((skb = skb->next));
1281
1282out:
1283 return segs;
1284}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001285EXPORT_SYMBOL(inet_gso_segment);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001286
Tom Herbertb8921ca2016-05-18 09:06:23 -07001287struct sk_buff **inet_gro_receive(struct sk_buff **head, struct sk_buff *skb)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001288{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001289 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001290 struct sk_buff **pp = NULL;
1291 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001292 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001293 unsigned int hlen;
1294 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001295 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001296 int flush = 1;
1297 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001298
Herbert Xua5b1cf22009-05-26 18:50:28 +00001299 off = skb_gro_offset(skb);
1300 hlen = off + sizeof(*iph);
1301 iph = skb_gro_header_fast(skb, off);
1302 if (skb_gro_header_hard(skb, hlen)) {
1303 iph = skb_gro_header_slow(skb, hlen, off);
1304 if (unlikely(!iph))
1305 goto out;
1306 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001307
David S. Millerf9242b62012-06-19 18:56:21 -07001308 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001309
1310 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001311 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001312 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001313 goto out_unlock;
1314
Herbert Xua5ad24b2009-02-08 18:00:39 +00001315 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001316 goto out_unlock;
1317
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001318 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001319 goto out_unlock;
1320
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001321 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001322 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001323 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001324
1325 for (p = *head; p; p = p->next) {
1326 struct iphdr *iph2;
Alexander Duyck15305452016-04-10 21:44:57 -04001327 u16 flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001328
1329 if (!NAPI_GRO_CB(p)->same_flow)
1330 continue;
1331
Jerry Chu299603e82013-12-11 20:53:45 -08001332 iph2 = (struct iphdr *)(p->data + off);
1333 /* The above works because, with the exception of the top
1334 * (inner most) layer, we only aggregate pkts with the same
1335 * hdr length so all the hdrs we'll need to verify will start
1336 * at the same offset.
1337 */
Herbert Xua5ad24b2009-02-08 18:00:39 +00001338 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001339 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1340 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001341 NAPI_GRO_CB(p)->same_flow = 0;
1342 continue;
1343 }
1344
1345 /* All fields must match except length and checksum. */
1346 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001347 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001348 (iph->tos ^ iph2->tos) |
Jerry Chubf5a7552014-01-07 10:23:19 -08001349 ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001350
1351 NAPI_GRO_CB(p)->flush |= flush;
Alexander Duyck15305452016-04-10 21:44:57 -04001352
1353 /* We need to store of the IP ID check to be included later
1354 * when we can verify that this packet does in fact belong
1355 * to a given flow.
1356 */
1357 flush_id = (u16)(id - ntohs(iph2->id));
1358
1359 /* This bit of code makes it much easier for us to identify
1360 * the cases where we are doing atomic vs non-atomic IP ID
1361 * checks. Specifically an atomic check can return IP ID
1362 * values 0 - 0xFFFF, while a non-atomic check can only
1363 * return 0 or 0xFFFF.
1364 */
1365 if (!NAPI_GRO_CB(p)->is_atomic ||
1366 !(iph->frag_off & htons(IP_DF))) {
1367 flush_id ^= NAPI_GRO_CB(p)->count;
1368 flush_id = flush_id ? 0xFFFF : 0;
1369 }
1370
1371 /* If the previous IP ID value was based on an atomic
1372 * datagram we can overwrite the value and ignore it.
1373 */
1374 if (NAPI_GRO_CB(skb)->is_atomic)
1375 NAPI_GRO_CB(p)->flush_id = flush_id;
1376 else
1377 NAPI_GRO_CB(p)->flush_id |= flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001378 }
1379
Alexander Duyck15305452016-04-10 21:44:57 -04001380 NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001381 NAPI_GRO_CB(skb)->flush |= flush;
Jerry Chu299603e82013-12-11 20:53:45 -08001382 skb_set_network_header(skb, off);
1383 /* The above will be needed by the transport layer if there is one
1384 * immediately following this IP hdr.
1385 */
1386
Eric Dumazet2c804d02014-09-30 22:12:05 -07001387 /* Note : No need to call skb_gro_postpull_rcsum() here,
1388 * as we already checked checksum over ipv4 header was 0
1389 */
Herbert Xu86911732009-01-29 14:19:50 +00001390 skb_gro_pull(skb, sizeof(*iph));
1391 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001392
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02001393 pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001394
1395out_unlock:
1396 rcu_read_unlock();
1397
1398out:
1399 NAPI_GRO_CB(skb)->flush |= flush;
1400
1401 return pp;
1402}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001403EXPORT_SYMBOL(inet_gro_receive);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001404
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001405static struct sk_buff **ipip_gro_receive(struct sk_buff **head,
1406 struct sk_buff *skb)
1407{
1408 if (NAPI_GRO_CB(skb)->encap_mark) {
1409 NAPI_GRO_CB(skb)->flush = 1;
1410 return NULL;
1411 }
1412
1413 NAPI_GRO_CB(skb)->encap_mark = 1;
1414
1415 return inet_gro_receive(head, skb);
1416}
1417
Deepa Dinamani822c8682016-02-27 00:32:15 -08001418#define SECONDS_PER_DAY 86400
1419
1420/* inet_current_timestamp - Return IP network timestamp
1421 *
1422 * Return milliseconds since midnight in network byte order.
1423 */
1424__be32 inet_current_timestamp(void)
1425{
1426 u32 secs;
1427 u32 msecs;
1428 struct timespec64 ts;
1429
1430 ktime_get_real_ts64(&ts);
1431
1432 /* Get secs since midnight. */
1433 (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs);
1434 /* Convert to msecs. */
1435 msecs = secs * MSEC_PER_SEC;
1436 /* Convert nsec to msec. */
1437 msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC;
1438
1439 /* Convert to network byte order. */
Deepa Dinamani3ba9d302016-03-21 18:21:26 -07001440 return htonl(msecs);
Deepa Dinamani822c8682016-02-27 00:32:15 -08001441}
1442EXPORT_SYMBOL(inet_current_timestamp);
1443
Willem de Bruijnf4713a32014-11-26 14:53:02 -05001444int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1445{
1446 if (sk->sk_family == AF_INET)
1447 return ip_recv_error(sk, msg, len, addr_len);
1448#if IS_ENABLED(CONFIG_IPV6)
1449 if (sk->sk_family == AF_INET6)
1450 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1451#endif
1452 return -EINVAL;
1453}
1454
Tom Herbertb8921ca2016-05-18 09:06:23 -07001455int inet_gro_complete(struct sk_buff *skb, int nhoff)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001456{
Jerry Chu299603e82013-12-11 20:53:45 -08001457 __be16 newlen = htons(skb->len - nhoff);
1458 struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001459 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001460 int proto = iph->protocol;
1461 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001462
Jerry Chuc3caf112014-07-14 15:54:46 -07001463 if (skb->encapsulation)
1464 skb_set_inner_network_header(skb, nhoff);
1465
Herbert Xu73cc19f2008-12-15 23:41:09 -08001466 csum_replace2(&iph->check, iph->tot_len, newlen);
1467 iph->tot_len = newlen;
1468
1469 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001470 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001471 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001472 goto out_unlock;
1473
Jerry Chu299603e82013-12-11 20:53:45 -08001474 /* Only need to add sizeof(*iph) to get to the next hdr below
1475 * because any hdr with option will have been flushed in
1476 * inet_gro_receive().
1477 */
1478 err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001479
1480out_unlock:
1481 rcu_read_unlock();
1482
1483 return err;
1484}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001485EXPORT_SYMBOL(inet_gro_complete);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001486
Jesse Grossb8cba752016-03-19 09:32:00 -07001487static int ipip_gro_complete(struct sk_buff *skb, int nhoff)
1488{
1489 skb->encapsulation = 1;
Tom Herbert7e133182016-05-18 09:06:10 -07001490 skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4;
Jesse Grossb8cba752016-03-19 09:32:00 -07001491 return inet_gro_complete(skb, nhoff);
1492}
1493
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001494int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001495 unsigned short type, unsigned char protocol,
1496 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001497{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001498 struct socket *sock;
Eric W. Biederman26abe142015-05-08 21:10:31 -05001499 int rc = sock_create_kern(net, family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001500
1501 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001502 *sk = sock->sk;
1503 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001504 /*
1505 * Unhash it so that IP input processing does not even see it,
1506 * we do not wish this socket to see incoming packets.
1507 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001508 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001509 }
1510 return rc;
1511}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001512EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1513
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301514u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
1515{
1516 return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
1517}
1518EXPORT_SYMBOL_GPL(snmp_get_cpu_field);
1519
WANG Cong698365f2014-05-05 15:55:55 -07001520unsigned long snmp_fold_field(void __percpu *mib, int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001521{
1522 unsigned long res = 0;
WANG Cong698365f2014-05-05 15:55:55 -07001523 int i;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001524
WANG Cong698365f2014-05-05 15:55:55 -07001525 for_each_possible_cpu(i)
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301526 res += snmp_get_cpu_field(mib, i, offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001527 return res;
1528}
1529EXPORT_SYMBOL_GPL(snmp_fold_field);
1530
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001531#if BITS_PER_LONG==32
1532
David S. Miller80ec1922015-08-30 22:40:44 -07001533u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt,
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301534 size_t syncp_offset)
1535{
1536 void *bhptr;
1537 struct u64_stats_sync *syncp;
1538 u64 v;
1539 unsigned int start;
1540
1541 bhptr = per_cpu_ptr(mib, cpu);
1542 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1543 do {
1544 start = u64_stats_fetch_begin_irq(syncp);
1545 v = *(((u64 *)bhptr) + offt);
1546 } while (u64_stats_fetch_retry_irq(syncp, start));
1547
1548 return v;
1549}
1550EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);
1551
WANG Cong698365f2014-05-05 15:55:55 -07001552u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001553{
1554 u64 res = 0;
1555 int cpu;
1556
1557 for_each_possible_cpu(cpu) {
Madalin Bucurd1bfc622015-08-31 15:46:07 +03001558 res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001559 }
1560 return res;
1561}
1562EXPORT_SYMBOL_GPL(snmp_fold_field64);
1563#endif
1564
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001566static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001568 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569};
1570#endif
1571
Alexey Dobriyan32613092009-09-14 12:21:47 +00001572static const struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001573 .early_demux = tcp_v4_early_demux,
1574 .handler = tcp_v4_rcv,
1575 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001576 .no_policy = 1,
1577 .netns_ok = 1,
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +01001578 .icmp_strict_tag_validation = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579};
1580
Alexey Dobriyan32613092009-09-14 12:21:47 +00001581static const struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001582 .early_demux = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 .handler = udp_rcv,
1584 .err_handler = udp_err,
1585 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001586 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587};
1588
Alexey Dobriyan32613092009-09-14 12:21:47 +00001589static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001591 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001592 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001593 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594};
1595
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001596static __net_init int ipv4_mib_init_net(struct net *net)
1597{
John Stultz827da442013-10-07 15:51:58 -07001598 int i;
1599
WANG Cong698365f2014-05-05 15:55:55 -07001600 net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1601 if (!net->mib.tcp_statistics)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001602 goto err_tcp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001603 net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1604 if (!net->mib.ip_statistics)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001605 goto err_ip_mib;
John Stultz827da442013-10-07 15:51:58 -07001606
1607 for_each_possible_cpu(i) {
1608 struct ipstats_mib *af_inet_stats;
WANG Cong698365f2014-05-05 15:55:55 -07001609 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
John Stultz827da442013-10-07 15:51:58 -07001610 u64_stats_init(&af_inet_stats->syncp);
John Stultz827da442013-10-07 15:51:58 -07001611 }
1612
WANG Cong698365f2014-05-05 15:55:55 -07001613 net->mib.net_statistics = alloc_percpu(struct linux_mib);
1614 if (!net->mib.net_statistics)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001615 goto err_net_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001616 net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1617 if (!net->mib.udp_statistics)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001618 goto err_udp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001619 net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1620 if (!net->mib.udplite_statistics)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001621 goto err_udplite_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001622 net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1623 if (!net->mib.icmp_statistics)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001624 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001625 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1626 GFP_KERNEL);
1627 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001628 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001629
1630 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001631 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001632
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001633err_icmpmsg_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001634 free_percpu(net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001635err_icmp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001636 free_percpu(net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001637err_udplite_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001638 free_percpu(net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001639err_udp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001640 free_percpu(net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001641err_net_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001642 free_percpu(net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001643err_ip_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001644 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001645err_tcp_mib:
1646 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001647}
1648
1649static __net_exit void ipv4_mib_exit_net(struct net *net)
1650{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001651 kfree(net->mib.icmpmsg_statistics);
WANG Cong698365f2014-05-05 15:55:55 -07001652 free_percpu(net->mib.icmp_statistics);
1653 free_percpu(net->mib.udplite_statistics);
1654 free_percpu(net->mib.udp_statistics);
1655 free_percpu(net->mib.net_statistics);
1656 free_percpu(net->mib.ip_statistics);
1657 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001658}
1659
1660static __net_initdata struct pernet_operations ipv4_mib_ops = {
1661 .init = ipv4_mib_init_net,
1662 .exit = ipv4_mib_exit_net,
1663};
1664
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665static int __init init_ipv4_mibs(void)
1666{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001667 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668}
1669
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001670static __net_init int inet_init_net(struct net *net)
1671{
1672 /*
1673 * Set defaults for local port range
1674 */
1675 seqlock_init(&net->ipv4.ip_local_ports.lock);
1676 net->ipv4.ip_local_ports.range[0] = 32768;
Eric Dumazet07f4c902015-05-24 14:49:35 -07001677 net->ipv4.ip_local_ports.range[1] = 60999;
Cong Wangba6b9182014-05-06 11:02:50 -07001678
1679 seqlock_init(&net->ipv4.ping_group_range.lock);
1680 /*
1681 * Sane defaults - nobody may create ping sockets.
1682 * Boot scripts should set this to distro-specific group.
1683 */
1684 net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1685 net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001686
1687 /* Default values for sysctl-controlled parameters.
1688 * We set them here, in case sysctl is not compiled.
1689 */
1690 net->ipv4.sysctl_ip_default_ttl = IPDEFTTL;
1691 net->ipv4.sysctl_ip_dynaddr = 0;
1692 net->ipv4.sysctl_ip_early_demux = 1;
1693
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001694 return 0;
1695}
1696
1697static __net_exit void inet_exit_net(struct net *net)
1698{
1699}
1700
1701static __net_initdata struct pernet_operations af_inet_ops = {
1702 .init = inet_init_net,
1703 .exit = inet_exit_net,
1704};
1705
1706static int __init init_inet_pernet_ops(void)
1707{
1708 return register_pernet_subsys(&af_inet_ops);
1709}
1710
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712
Herbert Xu30e224d2005-07-05 14:40:10 -07001713/*
1714 * IP protocol layer initialiser
1715 */
1716
Vlad Yasevich22061d82012-11-15 08:49:11 +00001717static struct packet_offload ip_packet_offload __read_mostly = {
1718 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001719 .callbacks = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001720 .gso_segment = inet_gso_segment,
1721 .gro_receive = inet_gro_receive,
1722 .gro_complete = inet_gro_complete,
1723 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001724};
1725
Eric Dumazetcb32f512013-10-19 11:42:57 -07001726static const struct net_offload ipip_offload = {
1727 .callbacks = {
Eric Dumazetcb32f512013-10-19 11:42:57 -07001728 .gso_segment = inet_gso_segment,
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001729 .gro_receive = ipip_gro_receive,
Jesse Grossb8cba752016-03-19 09:32:00 -07001730 .gro_complete = ipip_gro_complete,
Eric Dumazetcb32f512013-10-19 11:42:57 -07001731 },
1732};
1733
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001734static int __init ipv4_offload_init(void)
1735{
1736 /*
1737 * Add offloads
1738 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001739 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001740 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001741 if (tcpv4_offload_init() < 0)
1742 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001743
1744 dev_add_offload(&ip_packet_offload);
Eric Dumazetcb32f512013-10-19 11:42:57 -07001745 inet_add_offload(&ipip_offload, IPPROTO_IPIP);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001746 return 0;
1747}
1748
1749fs_initcall(ipv4_offload_init);
1750
1751static struct packet_type ip_packet_type __read_mostly = {
1752 .type = cpu_to_be16(ETH_P_IP),
1753 .func = ip_rcv,
1754};
1755
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756static int __init inet_init(void)
1757{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 struct inet_protosw *q;
1759 struct list_head *r;
1760 int rc = -EINVAL;
1761
Eyal Birgerb4772ef2015-03-01 14:58:29 +02001762 sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763
1764 rc = proto_register(&tcp_prot, 1);
1765 if (rc)
WANG Cong122ff242014-05-12 16:04:53 -07001766 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767
1768 rc = proto_register(&udp_prot, 1);
1769 if (rc)
1770 goto out_unregister_tcp_proto;
1771
1772 rc = proto_register(&raw_prot, 1);
1773 if (rc)
1774 goto out_unregister_udp_proto;
1775
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001776 rc = proto_register(&ping_prot, 1);
1777 if (rc)
1778 goto out_unregister_raw_proto;
1779
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001781 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 */
1783
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001784 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785
Al Virobd7b1532008-07-15 16:00:59 -04001786#ifdef CONFIG_SYSCTL
1787 ip_static_sysctl_init();
1788#endif
1789
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 /*
1791 * Add all the base protocols.
1792 */
1793
1794 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001795 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001797 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001799 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800#ifdef CONFIG_IP_MULTICAST
1801 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001802 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803#endif
1804
1805 /* Register the socket-side information for inet_create. */
1806 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1807 INIT_LIST_HEAD(r);
1808
1809 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1810 inet_register_protosw(q);
1811
1812 /*
1813 * Set the ARP module up
1814 */
1815
1816 arp_init();
1817
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001818 /*
1819 * Set the IP module up
1820 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
1822 ip_init();
1823
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001824 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825
1826 /* Setup TCP slab cache for open requests. */
1827 tcp_init();
1828
Hideo Aoki95766ff2007-12-31 00:29:24 -08001829 /* Setup UDP memory threshold */
1830 udp_init();
1831
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001832 /* Add UDP-Lite (RFC 3828) */
1833 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001835 ping_init();
1836
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 /*
1838 * Set the ICMP layer up
1839 */
1840
Denis V. Luneva5710d62008-02-29 11:14:50 -08001841 if (icmp_init() < 0)
1842 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843
1844 /*
1845 * Initialise the multicast router
1846 */
1847#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001848 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001849 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850#endif
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001851
1852 if (init_inet_pernet_ops())
1853 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 /*
1855 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001856 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857
Stephen Hemminger132adf52007-03-08 20:44:43 -08001858 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001859 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 ipv4_proc_init();
1862
1863 ipfrag_init();
1864
Herbert Xu30e224d2005-07-05 14:40:10 -07001865 dev_add_pack(&ip_packet_type);
1866
Thomas Graf045a0fa2015-07-23 10:08:44 +02001867 ip_tunnel_core_init();
1868
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 rc = 0;
1870out:
1871 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001872out_unregister_raw_proto:
1873 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874out_unregister_udp_proto:
1875 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001876out_unregister_tcp_proto:
1877 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 goto out;
1879}
1880
Heiko Carstensa536e072006-04-28 15:19:17 -07001881fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882
1883/* ------------------------------------------------------------------------ */
1884
1885#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886static int __init ipv4_proc_init(void)
1887{
1888 int rc = 0;
1889
1890 if (raw_proc_init())
1891 goto out_raw;
1892 if (tcp4_proc_init())
1893 goto out_tcp;
1894 if (udp4_proc_init())
1895 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001896 if (ping_proc_init())
1897 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 if (ip_misc_proc_init())
1899 goto out_misc;
1900out:
1901 return rc;
1902out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001903 ping_proc_exit();
1904out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 udp4_proc_exit();
1906out_udp:
1907 tcp4_proc_exit();
1908out_tcp:
1909 raw_proc_exit();
1910out_raw:
1911 rc = -ENOMEM;
1912 goto out;
1913}
1914
1915#else /* CONFIG_PROC_FS */
1916static int __init ipv4_proc_init(void)
1917{
1918 return 0;
1919}
1920#endif /* CONFIG_PROC_FS */