blob: c16634de706b993d560bbf91f1b47f2934a1b8a8 [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>
Chenbo Fengaa284cd2017-04-19 14:22:47 -070092#include <linux/netfilter/xt_qtaguid.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080094#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#include <linux/inet.h>
97#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020098#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/netdevice.h>
Herbert Xu73cc19f2008-12-15 23:41:09 -0800100#include <net/checksum.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <net/ip.h>
102#include <net/protocol.h>
103#include <net/arp.h>
104#include <net/route.h>
105#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700106#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/tcp.h>
108#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800109#include <net/udplite.h>
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +0000110#include <net/ping.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111#include <linux/skbuff.h>
112#include <net/sock.h>
113#include <net/raw.h>
114#include <net/icmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115#include <net/inet_common.h>
Thomas Graf045a0fa2015-07-23 10:08:44 +0200116#include <net/ip_tunnels.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700118#include <net/net_namespace.h>
Eric Dumazetaebda152013-04-29 05:58:52 +0000119#include <net/secure_seq.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#ifdef CONFIG_IP_MROUTE
121#include <linux/mroute.h>
122#endif
David Ahern3236b002015-09-29 20:07:14 -0700123#include <net/l3mdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
Robert Love7bafcbf2008-10-15 15:35:44 -0400125#ifdef CONFIG_ANDROID_PARANOID_NETWORK
126#include <linux/android_aid.h>
Chia-chi Yehd8460682009-06-30 11:23:04 +0800127
128static inline int current_has_network(void)
129{
130 return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
131}
132#else
133static inline int current_has_network(void)
134{
135 return 1;
136}
Robert Love7bafcbf2008-10-15 15:35:44 -0400137#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139/* The inetsw table contains everything that inet_create needs to
140 * build a new socket.
141 */
142static struct list_head inetsw[SOCK_MAX];
143static DEFINE_SPINLOCK(inetsw_lock);
144
145/* New destruction routine */
146
147void inet_sock_destruct(struct sock *sk)
148{
149 struct inet_sock *inet = inet_sk(sk);
150
151 __skb_queue_purge(&sk->sk_receive_queue);
152 __skb_queue_purge(&sk->sk_error_queue);
153
Hideo Aoki95766ff2007-12-31 00:29:24 -0800154 sk_mem_reclaim(sk);
155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700157 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 sk->sk_state, sk);
159 return;
160 }
161 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700162 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 return;
164 }
165
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700166 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
Reshetova, Elena14afee42017-06-30 13:08:00 +0300167 WARN_ON(refcount_read(&sk->sk_wmem_alloc));
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700168 WARN_ON(sk->sk_wmem_queued);
169 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000171 kfree(rcu_dereference_protected(inet->inet_opt, 1));
Eric Dumazetb6c67122010-04-08 23:03:29 +0000172 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
David S. Miller41063e92012-06-19 21:22:05 -0700173 dst_release(sk->sk_rx_dst);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700174 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700176EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178/*
179 * The routines beyond this point handle the behaviour of an AF_INET
180 * socket object. Mostly it punts to the subprotocols of IP to do
181 * the work.
182 */
183
184/*
185 * Automatically bind an unbound socket.
186 */
187
188static int inet_autobind(struct sock *sk)
189{
190 struct inet_sock *inet;
191 /* We may need to bind the socket. */
192 lock_sock(sk);
193 inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000194 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 if (sk->sk_prot->get_port(sk, 0)) {
196 release_sock(sk);
197 return -EAGAIN;
198 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000199 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 }
201 release_sock(sk);
202 return 0;
203}
204
205/*
206 * Move a socket into listening state.
207 */
208int inet_listen(struct socket *sock, int backlog)
209{
210 struct sock *sk = sock->sk;
211 unsigned char old_state;
212 int err;
213
214 lock_sock(sk);
215
216 err = -EINVAL;
217 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
218 goto out;
219
220 old_state = sk->sk_state;
221 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
222 goto out;
223
224 /* Really, if the socket is already in listen state
225 * we can only allow the backlog to be adjusted.
226 */
227 if (old_state != TCP_LISTEN) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700228 /* Enable TFO w/o requiring TCP_FASTOPEN socket option.
Jerry Chu83368862012-08-31 12:29:12 +0000229 * Note that only TCP sockets (SOCK_STREAM) will reach here.
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700230 * Also fastopen backlog may already been set via the option
231 * because the socket was in TCP_LISTEN state previously but
232 * was shutdown() rather than close().
Jerry Chu83368862012-08-31 12:29:12 +0000233 */
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700234 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
235 (sysctl_tcp_fastopen & TFO_SERVER_ENABLE) &&
Eric Dumazet0536fcc2015-09-29 07:42:52 -0700236 !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700237 fastopen_queue_tune(sk, backlog);
Christoph Paaschdfea2aa2015-06-18 09:15:34 -0700238 tcp_fastopen_init_key_once(true);
Jerry Chu83368862012-08-31 12:29:12 +0000239 }
Yuchung Chengcebc5cb2016-08-22 17:17:54 -0700240
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800241 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 if (err)
243 goto out;
244 }
245 sk->sk_max_ack_backlog = backlog;
246 err = 0;
247
248out:
249 release_sock(sk);
250 return err;
251}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700252EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
254/*
255 * Create an inet socket.
256 */
257
Eric Paris3f378b62009-11-05 22:18:14 -0800258static int inet_create(struct net *net, struct socket *sock, int protocol,
259 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
261 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 struct inet_protosw *answer;
263 struct inet_sock *inet;
264 struct proto *answer_prot;
265 unsigned char answer_flags;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700266 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800267 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Hannes Frederic Sowa79462ad02015-12-14 22:03:39 +0100269 if (protocol < 0 || protocol >= IPPROTO_MAX)
270 return -EINVAL;
271
Robert Love7bafcbf2008-10-15 15:35:44 -0400272 if (!current_has_network())
273 return -EACCES;
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 sock->state = SS_UNCONNECTED;
276
277 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700278lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800279 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700281 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700283 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 /* Check the non-wild match. */
285 if (protocol == answer->protocol) {
286 if (protocol != IPPROTO_IP)
287 break;
288 } else {
289 /* Check for the two wild cases. */
290 if (IPPROTO_IP == protocol) {
291 protocol = answer->protocol;
292 break;
293 }
294 if (IPPROTO_IP == answer->protocol)
295 break;
296 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800297 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 }
299
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700300 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700301 if (try_loading_module < 2) {
302 rcu_read_unlock();
303 /*
304 * Be more specific, e.g. net-pf-2-proto-132-type-1
305 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
306 */
307 if (++try_loading_module == 1)
308 request_module("net-pf-%d-proto-%d-type-%d",
309 PF_INET, protocol, sock->type);
310 /*
311 * Fall back to generic, e.g. net-pf-2-proto-132
312 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
313 */
314 else
315 request_module("net-pf-%d-proto-%d",
316 PF_INET, protocol);
317 goto lookup_protocol;
318 } else
319 goto out_rcu_unlock;
320 }
321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 err = -EPERM;
Chia-chi Yehd8460682009-06-30 11:23:04 +0800323 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
326 sock->ops = answer->ops;
327 answer_prot = answer->prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 answer_flags = answer->flags;
329 rcu_read_unlock();
330
Ian Morris51456b22015-04-03 09:17:26 +0100331 WARN_ON(!answer_prot->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333 err = -ENOBUFS;
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500334 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
Ian Morris51456b22015-04-03 09:17:26 +0100335 if (!sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 goto out;
337
338 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 if (INET_PROTOSW_REUSE & answer_flags)
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000340 sk->sk_reuse = SK_CAN_REUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
342 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800343 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344
Jiri Olsa7b2ff182010-06-15 01:07:31 +0000345 inet->nodefrag = 0;
346
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000348 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 if (IPPROTO_RAW == protocol)
350 inet->hdrincl = 1;
351 }
352
Hannes Frederic Sowa974eda12013-12-14 05:13:38 +0100353 if (net->ipv4.sysctl_ip_no_pmtu_disc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 inet->pmtudisc = IP_PMTUDISC_DONT;
355 else
356 inet->pmtudisc = IP_PMTUDISC_WANT;
357
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000358 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
360 sock_init_data(sock, sk);
361
362 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 sk->sk_protocol = protocol;
364 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
365
366 inet->uc_ttl = -1;
367 inet->mc_loop = 1;
368 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000369 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 inet->mc_index = 0;
371 inet->mc_list = NULL;
Jiri Benc4c507d22012-02-09 09:35:49 +0000372 inet->rcv_tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700374 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000376 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 /* It assumes that any protocol which allows
378 * the user to assign a number at socket
379 * creation time automatically
380 * shares.
381 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000382 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 /* Add to protocol hash chains. */
Craig Gallek086c6532016-02-10 11:50:35 -0500384 err = sk->sk_prot->hash(sk);
385 if (err) {
386 sk_common_release(sk);
387 goto out;
388 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 }
390
391 if (sk->sk_prot->init) {
392 err = sk->sk_prot->init(sk);
David Ahern61023652016-12-01 08:48:04 -0800393 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 sk_common_release(sk);
David Ahern61023652016-12-01 08:48:04 -0800395 goto out;
396 }
397 }
398
399 if (!kern) {
400 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
401 if (err) {
402 sk_common_release(sk);
403 goto out;
404 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 }
406out:
407 return err;
408out_rcu_unlock:
409 rcu_read_unlock();
410 goto out;
411}
412
413
414/*
415 * The peer socket should always be NULL (or else). When we call this
416 * function we are destroying the object and from then on nobody
417 * should refer to it.
418 */
419int inet_release(struct socket *sock)
420{
421 struct sock *sk = sock->sk;
422
423 if (sk) {
424 long timeout;
425
Chenbo Fengaa284cd2017-04-19 14:22:47 -0700426#ifdef CONFIG_NETFILTER_XT_MATCH_QTAGUID
427 qtaguid_untag(sock, true);
428#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 /* Applications forget to leave groups before exiting */
430 ip_mc_drop_socket(sk);
431
432 /* If linger is set, we don't return until the close
433 * is complete. Otherwise we return immediately. The
434 * actually closing is done the same either way.
435 *
436 * If the close is due to the process exiting, we never
437 * linger..
438 */
439 timeout = 0;
440 if (sock_flag(sk, SOCK_LINGER) &&
441 !(current->flags & PF_EXITING))
442 timeout = sk->sk_lingertime;
443 sock->sk = NULL;
444 sk->sk_prot->close(sk, timeout);
445 }
446 return 0;
447}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700448EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
451{
452 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
453 struct sock *sk = sock->sk;
454 struct inet_sock *inet = inet_sk(sk);
Eric W. Biederman35946982012-11-16 03:03:12 +0000455 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 unsigned short snum;
457 int chk_addr_ret;
David Ahern9b8ff512015-09-01 14:26:35 -0600458 u32 tb_id = RT_TABLE_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 int err;
460
461 /* If the socket has its own bind function then use it. (RAW) */
462 if (sk->sk_prot->bind) {
463 err = sk->sk_prot->bind(sk, uaddr, addr_len);
464 goto out;
465 }
466 err = -EINVAL;
467 if (addr_len < sizeof(struct sockaddr_in))
468 goto out;
469
Marcus Meissnerc349a522011-07-04 01:30:29 +0000470 if (addr->sin_family != AF_INET) {
Eric Dumazet29c486d2011-08-30 18:57:00 -0400471 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
472 * only if s_addr is INADDR_ANY.
473 */
Marcus Meissnerc349a522011-07-04 01:30:29 +0000474 err = -EAFNOSUPPORT;
Eric Dumazet29c486d2011-08-30 18:57:00 -0400475 if (addr->sin_family != AF_UNSPEC ||
476 addr->sin_addr.s_addr != htonl(INADDR_ANY))
477 goto out;
Marcus Meissnerc349a522011-07-04 01:30:29 +0000478 }
Marcus Meissnerd0733d22011-06-01 21:05:22 -0700479
David Ahern3236b002015-09-29 20:07:14 -0700480 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
David Ahern30bbaa12015-08-13 14:59:05 -0600481 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
483 /* Not specified by any standard per-se, however it breaks too
484 * many applications when removed. It is unfortunate since
485 * allowing applications to make a non-local bind solves
486 * several problems with systems using dynamic addressing.
487 * (ie. your servers still start up even if your ISDN link
488 * is temporarily down)
489 */
490 err = -EADDRNOTAVAIL;
Vincent Bernat49a60152014-09-05 15:09:03 +0200491 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700492 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700493 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 chk_addr_ret != RTN_LOCAL &&
495 chk_addr_ret != RTN_MULTICAST &&
496 chk_addr_ret != RTN_BROADCAST)
497 goto out;
498
499 snum = ntohs(addr->sin_port);
500 err = -EACCES;
Krister Johansen4548b682017-01-20 17:49:11 -0800501 if (snum && snum < inet_prot_sock(net) &&
Eric W. Biederman35946982012-11-16 03:03:12 +0000502 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 goto out;
504
505 /* We keep a pair of addresses. rcv_saddr is the one
506 * used by hash lookups, and saddr is used for transmit.
507 *
508 * In the BSD API these are the same except where it
509 * would be illegal to use them (multicast/broadcast) in
510 * which case the sending device address is used.
511 */
512 lock_sock(sk);
513
514 /* Check these errors (active socket, double bind). */
515 err = -EINVAL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000516 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 goto out_release_sock;
518
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000519 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000521 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
523 /* Make sure we are allowed to bind here. */
Eric Dumazet90c337d2015-06-06 21:17:57 -0700524 if ((snum || !inet->bind_address_no_port) &&
525 sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000526 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 err = -EADDRINUSE;
528 goto out_release_sock;
529 }
530
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000531 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
533 if (snum)
534 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000535 inet->inet_sport = htons(inet->inet_num);
536 inet->inet_daddr = 0;
537 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 sk_dst_reset(sk);
539 err = 0;
540out_release_sock:
541 release_sock(sk);
542out:
543 return err;
544}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700545EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000547int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 int addr_len, int flags)
549{
550 struct sock *sk = sock->sk;
551
Changli Gao6503d962010-03-31 22:58:26 +0000552 if (addr_len < sizeof(uaddr->sa_family))
553 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 if (uaddr->sa_family == AF_UNSPEC)
555 return sk->sk_prot->disconnect(sk, flags);
556
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000557 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 return -EAGAIN;
Joe Perchese3192692012-06-03 17:41:40 +0000559 return sk->sk_prot->connect(sk, uaddr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700561EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
Yuchung Cheng783237e2012-07-19 06:43:07 +0000563static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564{
WANG Cong14135f32016-11-01 16:04:36 -0700565 DEFINE_WAIT_FUNC(wait, woken_wake_function);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
WANG Cong14135f32016-11-01 16:04:36 -0700567 add_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000568 sk->sk_write_pending += writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
570 /* Basic assumption: if someone sets sk->sk_err, he _must_
571 * change state of the socket from TCP_SYN_*.
572 * Connect() does not allow to get error notifications
573 * without closing the socket.
574 */
575 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
576 release_sock(sk);
WANG Cong14135f32016-11-01 16:04:36 -0700577 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 lock_sock(sk);
579 if (signal_pending(current) || !timeo)
580 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 }
WANG Cong14135f32016-11-01 16:04:36 -0700582 remove_wait_queue(sk_sleep(sk), &wait);
Yuchung Cheng783237e2012-07-19 06:43:07 +0000583 sk->sk_write_pending -= writebias;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 return timeo;
585}
586
587/*
588 * Connect to a remote host. There is regrettably still a little
589 * TCP 'magic' in here.
590 */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000591int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
Willy Tarreau3979ad72017-01-25 14:42:46 +0100592 int addr_len, int flags, int is_sendmsg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593{
594 struct sock *sk = sock->sk;
595 int err;
596 long timeo;
597
Wei Wang19f6d3f2017-01-23 10:59:22 -0800598 /*
599 * uaddr can be NULL and addr_len can be 0 if:
600 * sk is a TCP fastopen active socket and
601 * TCP_FASTOPEN_CONNECT sockopt is set and
602 * we already have a valid cookie for this socket.
603 * In this case, user can call write() after connect().
604 * write() will invoke tcp_sendmsg_fastopen() which calls
605 * __inet_stream_connect().
606 */
607 if (uaddr) {
608 if (addr_len < sizeof(uaddr->sa_family))
609 return -EINVAL;
Changli Gao6503d962010-03-31 22:58:26 +0000610
Wei Wang19f6d3f2017-01-23 10:59:22 -0800611 if (uaddr->sa_family == AF_UNSPEC) {
612 err = sk->sk_prot->disconnect(sk, flags);
613 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
614 goto out;
615 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 }
617
618 switch (sock->state) {
619 default:
620 err = -EINVAL;
621 goto out;
622 case SS_CONNECTED:
623 err = -EISCONN;
624 goto out;
625 case SS_CONNECTING:
Wei Wang19f6d3f2017-01-23 10:59:22 -0800626 if (inet_sk(sk)->defer_connect)
Willy Tarreau3979ad72017-01-25 14:42:46 +0100627 err = is_sendmsg ? -EINPROGRESS : -EISCONN;
Wei Wang19f6d3f2017-01-23 10:59:22 -0800628 else
629 err = -EALREADY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 /* Fall out of switch with err, set for this state */
631 break;
632 case SS_UNCONNECTED:
633 err = -EISCONN;
634 if (sk->sk_state != TCP_CLOSE)
635 goto out;
636
637 err = sk->sk_prot->connect(sk, uaddr, addr_len);
638 if (err < 0)
639 goto out;
640
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900641 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
Wei Wang19f6d3f2017-01-23 10:59:22 -0800643 if (!err && inet_sk(sk)->defer_connect)
644 goto out;
645
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 /* Just entered SS_CONNECTING state; the only
647 * difference is that return value in non-blocking
648 * case is EINPROGRESS, rather than EALREADY.
649 */
650 err = -EINPROGRESS;
651 break;
652 }
653
654 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
655
656 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
Yuchung Cheng783237e2012-07-19 06:43:07 +0000657 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
658 tcp_sk(sk)->fastopen_req &&
659 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 /* Error code is set above */
Yuchung Cheng783237e2012-07-19 06:43:07 +0000662 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 goto out;
664
665 err = sock_intr_errno(timeo);
666 if (signal_pending(current))
667 goto out;
668 }
669
670 /* Connection was closed by RST, timeout, ICMP error
671 * or another process disconnected us.
672 */
673 if (sk->sk_state == TCP_CLOSE)
674 goto sock_error;
675
676 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
677 * and error was received after socket entered established state.
678 * Hence, it is handled normally after connect() return successfully.
679 */
680
681 sock->state = SS_CONNECTED;
682 err = 0;
683out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 return err;
685
686sock_error:
687 err = sock_error(sk) ? : -ECONNABORTED;
688 sock->state = SS_UNCONNECTED;
689 if (sk->sk_prot->disconnect(sk, flags))
690 sock->state = SS_DISCONNECTING;
691 goto out;
692}
Yuchung Chengcf60af02012-07-19 06:43:09 +0000693EXPORT_SYMBOL(__inet_stream_connect);
694
695int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
696 int addr_len, int flags)
697{
698 int err;
699
700 lock_sock(sock->sk);
Willy Tarreau3979ad72017-01-25 14:42:46 +0100701 err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0);
Yuchung Chengcf60af02012-07-19 06:43:09 +0000702 release_sock(sock->sk);
703 return err;
704}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700705EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
707/*
708 * Accept a pending connection. The TCP layer now gives BSD semantics.
709 */
710
David Howellscdfbabf2017-03-09 08:09:05 +0000711int inet_accept(struct socket *sock, struct socket *newsock, int flags,
712 bool kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713{
714 struct sock *sk1 = sock->sk;
715 int err = -EINVAL;
David Howellscdfbabf2017-03-09 08:09:05 +0000716 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
718 if (!sk2)
719 goto do_err;
720
721 lock_sock(sk2);
722
Eric Dumazet1eddcea2011-06-17 03:45:15 +0000723 sock_rps_record_flow(sk2);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700724 WARN_ON(!((1 << sk2->sk_state) &
Jerry Chu83368862012-08-31 12:29:12 +0000725 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
726 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728 sock_graft(sk2, newsock);
729
730 newsock->state = SS_CONNECTED;
731 err = 0;
732 release_sock(sk2);
733do_err:
734 return err;
735}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700736EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
738
739/*
740 * This does both peername and sockname.
741 */
742int inet_getname(struct socket *sock, struct sockaddr *uaddr,
743 int *uaddr_len, int peer)
744{
745 struct sock *sk = sock->sk;
746 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700747 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
749 sin->sin_family = AF_INET;
750 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000751 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
753 peer == 1))
754 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000755 sin->sin_port = inet->inet_dport;
756 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000758 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000760 addr = inet->inet_saddr;
761 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 sin->sin_addr.s_addr = addr;
763 }
764 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
765 *uaddr_len = sizeof(*sin);
766 return 0;
767}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700768EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
Ying Xue1b784142015-03-02 15:37:48 +0800770int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771{
772 struct sock *sk = sock->sk;
773
David S. Millerc58dc012010-04-27 15:05:31 -0700774 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000777 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
778 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 return -EAGAIN;
780
Ying Xue1b784142015-03-02 15:37:48 +0800781 return sk->sk_prot->sendmsg(sk, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700783EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Changli Gao7ba42912010-07-10 20:41:55 +0000785ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
786 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787{
788 struct sock *sk = sock->sk;
789
David S. Millerc58dc012010-04-27 15:05:31 -0700790 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700791
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 /* We may need to bind the socket. */
Changli Gao7ba42912010-07-10 20:41:55 +0000793 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
794 inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 return -EAGAIN;
796
797 if (sk->sk_prot->sendpage)
798 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
799 return sock_no_sendpage(sock, page, offset, size, flags);
800}
Changli Gao7ba42912010-07-10 20:41:55 +0000801EXPORT_SYMBOL(inet_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Ying Xue1b784142015-03-02 15:37:48 +0800803int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
804 int flags)
Tom Herbertfec5e652010-04-16 16:01:27 -0700805{
806 struct sock *sk = sock->sk;
807 int addr_len = 0;
808 int err;
809
David S. Millerc58dc012010-04-27 15:05:31 -0700810 sock_rps_record_flow(sk);
Tom Herbertfec5e652010-04-16 16:01:27 -0700811
Ying Xue1b784142015-03-02 15:37:48 +0800812 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Tom Herbertfec5e652010-04-16 16:01:27 -0700813 flags & ~MSG_DONTWAIT, &addr_len);
814 if (err >= 0)
815 msg->msg_namelen = addr_len;
816 return err;
817}
818EXPORT_SYMBOL(inet_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
820int inet_shutdown(struct socket *sock, int how)
821{
822 struct sock *sk = sock->sk;
823 int err = 0;
824
825 /* This should really check to make sure
826 * the socket is a TCP socket. (WHY AC...)
827 */
828 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
829 1->2 bit 2 snds.
830 2->3 */
831 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
832 return -EINVAL;
833
834 lock_sock(sk);
835 if (sock->state == SS_CONNECTING) {
836 if ((1 << sk->sk_state) &
837 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
838 sock->state = SS_DISCONNECTING;
839 else
840 sock->state = SS_CONNECTED;
841 }
842
843 switch (sk->sk_state) {
844 case TCP_CLOSE:
845 err = -ENOTCONN;
846 /* Hack to wake up other listeners, who can poll for
847 POLLHUP, even on eg. unconnected UDP sockets -- RR */
848 default:
849 sk->sk_shutdown |= how;
850 if (sk->sk_prot->shutdown)
851 sk->sk_prot->shutdown(sk, how);
852 break;
853
854 /* Remaining two branches are temporary solution for missing
855 * close() in multithreaded environment. It is _not_ a good idea,
856 * but we have no choice until close() is repaired at VFS level.
857 */
858 case TCP_LISTEN:
859 if (!(how & RCV_SHUTDOWN))
860 break;
861 /* Fall through */
862 case TCP_SYN_SENT:
863 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
864 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
865 break;
866 }
867
868 /* Wake up anyone sleeping in poll. */
869 sk->sk_state_change(sk);
870 release_sock(sk);
871 return err;
872}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700873EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875/*
876 * ioctl() calls you can issue on an INET socket. Most of these are
877 * device configuration and stuff and very rarely used. Some ioctls
878 * pass on to the socket itself.
879 *
880 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
881 * loads the devconfigure module does its configuring and unloads it.
882 * There's a good 20K of config code hanging around the kernel.
883 */
884
885int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
886{
887 struct sock *sk = sock->sk;
888 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900889 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
891 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700892 case SIOCGSTAMP:
893 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
894 break;
895 case SIOCGSTAMPNS:
896 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
897 break;
898 case SIOCADDRT:
899 case SIOCDELRT:
900 case SIOCRTMSG:
901 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
902 break;
903 case SIOCDARP:
904 case SIOCGARP:
905 case SIOCSARP:
906 err = arp_ioctl(net, cmd, (void __user *)arg);
907 break;
908 case SIOCGIFADDR:
909 case SIOCSIFADDR:
910 case SIOCGIFBRDADDR:
911 case SIOCSIFBRDADDR:
912 case SIOCGIFNETMASK:
913 case SIOCSIFNETMASK:
914 case SIOCGIFDSTADDR:
915 case SIOCSIFDSTADDR:
916 case SIOCSIFPFLAGS:
917 case SIOCGIFPFLAGS:
918 case SIOCSIFFLAGS:
919 err = devinet_ioctl(net, cmd, (void __user *)arg);
920 break;
921 default:
922 if (sk->sk_prot->ioctl)
923 err = sk->sk_prot->ioctl(sk, cmd, arg);
924 else
925 err = -ENOIOCTLCMD;
926 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 }
928 return err;
929}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700930EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000932#ifdef CONFIG_COMPAT
Gerrit Renker686dc6b2011-10-15 09:26:56 +0000933static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000934{
935 struct sock *sk = sock->sk;
936 int err = -ENOIOCTLCMD;
937
938 if (sk->sk_prot->compat_ioctl)
939 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
940
941 return err;
942}
943#endif
944
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800945const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800946 .family = PF_INET,
947 .owner = THIS_MODULE,
948 .release = inet_release,
949 .bind = inet_bind,
950 .connect = inet_stream_connect,
951 .socketpair = sock_no_socketpair,
952 .accept = inet_accept,
953 .getname = inet_getname,
954 .poll = tcp_poll,
955 .ioctl = inet_ioctl,
956 .listen = inet_listen,
957 .shutdown = inet_shutdown,
958 .setsockopt = sock_common_setsockopt,
959 .getsockopt = sock_common_getsockopt,
Changli Gao7ba42912010-07-10 20:41:55 +0000960 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700961 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800962 .mmap = sock_no_mmap,
Changli Gao7ba42912010-07-10 20:41:55 +0000963 .sendpage = inet_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800964 .splice_read = tcp_splice_read,
Tom Herbert32035582016-08-28 14:43:18 -0700965 .read_sock = tcp_read_sock,
Tom Herbert306b13e2017-07-28 16:22:41 -0700966 .sendmsg_locked = tcp_sendmsg_locked,
967 .sendpage_locked = tcp_sendpage_locked,
Tom Herbert32035582016-08-28 14:43:18 -0700968 .peek_len = tcp_peek_len,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800969#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800970 .compat_setsockopt = compat_sock_common_setsockopt,
971 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000972 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800973#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700975EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800977const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800978 .family = PF_INET,
979 .owner = THIS_MODULE,
980 .release = inet_release,
981 .bind = inet_bind,
982 .connect = inet_dgram_connect,
983 .socketpair = sock_no_socketpair,
984 .accept = sock_no_accept,
985 .getname = inet_getname,
986 .poll = udp_poll,
987 .ioctl = inet_ioctl,
988 .listen = sock_no_listen,
989 .shutdown = inet_shutdown,
990 .setsockopt = sock_common_setsockopt,
991 .getsockopt = sock_common_getsockopt,
992 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -0700993 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800994 .mmap = sock_no_mmap,
995 .sendpage = inet_sendpage,
samanthakumar627d2d62016-04-05 12:41:16 -0400996 .set_peek_off = sk_set_peek_off,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800997#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800998 .compat_setsockopt = compat_sock_common_setsockopt,
999 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001000 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001001#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002};
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001003EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004
1005/*
1006 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
1007 * udp_poll
1008 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001009static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001010 .family = PF_INET,
1011 .owner = THIS_MODULE,
1012 .release = inet_release,
1013 .bind = inet_bind,
1014 .connect = inet_dgram_connect,
1015 .socketpair = sock_no_socketpair,
1016 .accept = sock_no_accept,
1017 .getname = inet_getname,
1018 .poll = datagram_poll,
1019 .ioctl = inet_ioctl,
1020 .listen = sock_no_listen,
1021 .shutdown = inet_shutdown,
1022 .setsockopt = sock_common_setsockopt,
1023 .getsockopt = sock_common_getsockopt,
1024 .sendmsg = inet_sendmsg,
Tom Herbertfec5e652010-04-16 16:01:27 -07001025 .recvmsg = inet_recvmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001026 .mmap = sock_no_mmap,
1027 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001028#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001029 .compat_setsockopt = compat_sock_common_setsockopt,
1030 .compat_getsockopt = compat_sock_common_getsockopt,
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001031 .compat_ioctl = inet_compat_ioctl,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001032#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033};
1034
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00001035static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 .family = PF_INET,
1037 .create = inet_create,
1038 .owner = THIS_MODULE,
1039};
1040
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041/* Upon startup we insert all the elements in inetsw_array[] into
1042 * the linked list inetsw.
1043 */
1044static struct inet_protosw inetsw_array[] =
1045{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001046 {
1047 .type = SOCK_STREAM,
1048 .protocol = IPPROTO_TCP,
1049 .prot = &tcp_prot,
1050 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001051 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -08001052 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001053 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001055 {
1056 .type = SOCK_DGRAM,
1057 .protocol = IPPROTO_UDP,
1058 .prot = &udp_prot,
1059 .ops = &inet_dgram_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001060 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001062
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001063 {
1064 .type = SOCK_DGRAM,
1065 .protocol = IPPROTO_ICMP,
1066 .prot = &ping_prot,
Eric Dumazet77d4b1d2017-06-03 09:29:25 -07001067 .ops = &inet_sockraw_ops,
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001068 .flags = INET_PROTOSW_REUSE,
1069 },
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
1071 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001072 .type = SOCK_RAW,
1073 .protocol = IPPROTO_IP, /* wild card */
1074 .prot = &raw_prot,
1075 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001076 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 }
1078};
1079
Denis Chengc40f6ff2007-09-16 16:39:25 -07001080#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
1082void inet_register_protosw(struct inet_protosw *p)
1083{
1084 struct list_head *lh;
1085 struct inet_protosw *answer;
1086 int protocol = p->protocol;
1087 struct list_head *last_perm;
1088
1089 spin_lock_bh(&inetsw_lock);
1090
1091 if (p->type >= SOCK_MAX)
1092 goto out_illegal;
1093
1094 /* If we are trying to override a permanent protocol, bail. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 last_perm = &inetsw[p->type];
1096 list_for_each(lh, &inetsw[p->type]) {
1097 answer = list_entry(lh, struct inet_protosw, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 /* Check only the non-wild match. */
Junwei Zhangf6c53332015-09-18 00:00:05 -04001099 if ((INET_PROTOSW_PERMANENT & answer->flags) == 0)
1100 break;
1101 if (protocol == answer->protocol)
1102 goto out_permanent;
1103 last_perm = lh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
1106 /* Add the new entry after the last permanent entry if any, so that
1107 * the new entry does not override a permanent entry when matched with
1108 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001109 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 * system automatically returns to the old behavior.
1111 */
1112 list_add_rcu(&p->list, last_perm);
1113out:
1114 spin_unlock_bh(&inetsw_lock);
1115
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 return;
1117
1118out_permanent:
Joe Perches058bd4d2012-03-11 18:36:11 +00001119 pr_err("Attempt to override permanent protocol %d\n", protocol);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 goto out;
1121
1122out_illegal:
Joe Perches058bd4d2012-03-11 18:36:11 +00001123 pr_err("Ignoring attempt to register invalid socket type %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 p->type);
1125 goto out;
1126}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001127EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
1129void inet_unregister_protosw(struct inet_protosw *p)
1130{
1131 if (INET_PROTOSW_PERMANENT & p->flags) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001132 pr_err("Attempt to unregister permanent protocol %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 p->protocol);
1134 } else {
1135 spin_lock_bh(&inetsw_lock);
1136 list_del_rcu(&p->list);
1137 spin_unlock_bh(&inetsw_lock);
1138
1139 synchronize_net();
1140 }
1141}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001142EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001144static int inet_sk_reselect_saddr(struct sock *sk)
1145{
1146 struct inet_sock *inet = inet_sk(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001147 __be32 old_saddr = inet->inet_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001148 __be32 daddr = inet->inet_daddr;
David S. Miller6e869132011-05-06 16:18:04 -07001149 struct flowi4 *fl4;
David S. Millerb23dd4f2011-03-02 14:31:35 -08001150 struct rtable *rt;
1151 __be32 new_saddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001152 struct ip_options_rcu *inet_opt;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001153
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001154 inet_opt = rcu_dereference_protected(inet->inet_opt,
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001155 lockdep_sock_is_held(sk));
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001156 if (inet_opt && inet_opt->opt.srr)
1157 daddr = inet_opt->opt.faddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001158
1159 /* Query new route. */
David S. Miller6e869132011-05-06 16:18:04 -07001160 fl4 = &inet->cork.fl.u.ip4;
1161 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
David S. Millerb23dd4f2011-03-02 14:31:35 -08001162 sk->sk_bound_dev_if, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001163 inet->inet_sport, inet->inet_dport, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001164 if (IS_ERR(rt))
1165 return PTR_ERR(rt);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001166
Changli Gaod8d1f302010-06-10 23:31:35 -07001167 sk_setup_caps(sk, &rt->dst);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001168
David S. Miller6e869132011-05-06 16:18:04 -07001169 new_saddr = fl4->saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001170
1171 if (new_saddr == old_saddr)
1172 return 0;
1173
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001174 if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) {
Joe Perches058bd4d2012-03-11 18:36:11 +00001175 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1176 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001177 }
1178
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001179 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001180
1181 /*
1182 * XXX The only one ugly spot where we need to
1183 * XXX really change the sockets identity after
1184 * XXX it has entered the hashes. -DaveM
1185 *
1186 * Besides that, it does not check for connection
1187 * uniqueness. Wait for troubles.
1188 */
Craig Gallek086c6532016-02-10 11:50:35 -05001189 return __sk_prot_rehash(sk);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001190}
1191
1192int inet_sk_rebuild_header(struct sock *sk)
1193{
1194 struct inet_sock *inet = inet_sk(sk);
1195 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001196 __be32 daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001197 struct ip_options_rcu *inet_opt;
David S. Miller6e869132011-05-06 16:18:04 -07001198 struct flowi4 *fl4;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001199 int err;
1200
1201 /* Route is OK, nothing to do. */
1202 if (rt)
1203 return 0;
1204
1205 /* Reroute. */
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001206 rcu_read_lock();
1207 inet_opt = rcu_dereference(inet->inet_opt);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001208 daddr = inet->inet_daddr;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001209 if (inet_opt && inet_opt->opt.srr)
1210 daddr = inet_opt->opt.faddr;
1211 rcu_read_unlock();
David S. Miller6e869132011-05-06 16:18:04 -07001212 fl4 = &inet->cork.fl.u.ip4;
1213 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -05001214 inet->inet_dport, inet->inet_sport,
1215 sk->sk_protocol, RT_CONN_FLAGS(sk),
1216 sk->sk_bound_dev_if);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001217 if (!IS_ERR(rt)) {
1218 err = 0;
Changli Gaod8d1f302010-06-10 23:31:35 -07001219 sk_setup_caps(sk, &rt->dst);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001220 } else {
1221 err = PTR_ERR(rt);
1222
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001223 /* Routing failed... */
1224 sk->sk_route_caps = 0;
1225 /*
1226 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1227 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1228 */
Nikolay Borisov287b7f32016-02-15 12:11:29 +02001229 if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr ||
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001230 sk->sk_state != TCP_SYN_SENT ||
1231 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1232 (err = inet_sk_reselect_saddr(sk)) != 0)
1233 sk->sk_err_soft = -err;
1234 }
1235
1236 return err;
1237}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001238EXPORT_SYMBOL(inet_sk_rebuild_header);
1239
Tom Herbertb8921ca2016-05-18 09:06:23 -07001240struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1241 netdev_features_t features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001242{
Willem de Bruijn60335602017-11-21 10:22:25 -05001243 bool udpfrag = false, fixedid = false, gso_partial, encap;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001244 struct sk_buff *segs = ERR_PTR(-EINVAL);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001245 const struct net_offload *ops;
Willem de Bruijn60335602017-11-21 10:22:25 -05001246 unsigned int offset = 0;
David S. Millerf9242b62012-06-19 18:56:21 -07001247 struct iphdr *iph;
Alexander Duyck802ab552016-04-10 21:45:03 -04001248 int proto, tot_len;
Eric Dumazet3347c962013-10-19 11:42:56 -07001249 int nhoff;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001250 int ihl;
1251 int id;
1252
Eric Dumazet3347c962013-10-19 11:42:56 -07001253 skb_reset_network_header(skb);
1254 nhoff = skb_network_header(skb) - skb_mac_header(skb);
Herbert Xubbcf4672006-07-03 19:38:35 -07001255 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001256 goto out;
1257
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001258 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001259 ihl = iph->ihl * 4;
1260 if (ihl < sizeof(*iph))
1261 goto out;
1262
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001263 id = ntohs(iph->id);
1264 proto = iph->protocol;
1265
1266 /* Warning: after this point, iph might be no longer valid */
Herbert Xubbcf4672006-07-03 19:38:35 -07001267 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001268 goto out;
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001269 __skb_pull(skb, ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001270
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001271 encap = SKB_GSO_CB(skb)->encap_level > 0;
1272 if (encap)
Florian Westphal1e16aa32014-10-20 13:49:16 +02001273 features &= skb->dev->hw_enc_features;
Eric Dumazet3347c962013-10-19 11:42:56 -07001274 SKB_GSO_CB(skb)->encap_level += ihl;
Pravin B Shelar73136262013-03-07 13:21:51 +00001275
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001276 skb_reset_transport_header(skb);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001277
Herbert Xuf4c50d92006-06-22 03:02:40 -07001278 segs = ERR_PTR(-EPROTONOSUPPORT);
1279
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001280 if (!skb->encapsulation || encap) {
Willem de Bruijn60335602017-11-21 10:22:25 -05001281 udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP);
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001282 fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID);
1283
1284 /* fixed ID is invalid if DF bit is not set */
Arnaldo Carvalho de Meloa5108872016-11-28 12:36:58 -03001285 if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF)))
Alexander Duyckcbc53e02016-04-10 21:44:51 -04001286 goto out;
1287 }
Eric Dumazetdcd60772013-11-07 18:32:06 -08001288
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001289 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001290 if (likely(ops && ops->callbacks.gso_segment))
1291 segs = ops->callbacks.gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001292
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001293 if (IS_ERR_OR_NULL(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001294 goto out;
1295
Steffen Klassert07b26c92016-09-19 12:58:47 +02001296 gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL);
1297
Herbert Xuf4c50d92006-06-22 03:02:40 -07001298 skb = segs;
1299 do {
Eric Dumazet3347c962013-10-19 11:42:56 -07001300 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
Willem de Bruijn60335602017-11-21 10:22:25 -05001301 if (udpfrag) {
1302 iph->frag_off = htons(offset >> 3);
1303 if (skb->next)
1304 iph->frag_off |= htons(IP_MF);
1305 offset += skb->len - nhoff - ihl;
1306 tot_len = skb->len - nhoff;
1307 } else if (skb_is_gso(skb)) {
Alexander Duyck802ab552016-04-10 21:45:03 -04001308 if (!fixedid) {
1309 iph->id = htons(id);
1310 id += skb_shinfo(skb)->gso_segs;
1311 }
Steffen Klassert07b26c92016-09-19 12:58:47 +02001312
1313 if (gso_partial)
1314 tot_len = skb_shinfo(skb)->gso_size +
1315 SKB_GSO_CB(skb)->data_offset +
1316 skb->head - (unsigned char *)iph;
1317 else
1318 tot_len = skb->len - nhoff;
Alexander Duyck802ab552016-04-10 21:45:03 -04001319 } else {
1320 if (!fixedid)
1321 iph->id = htons(id++);
1322 tot_len = skb->len - nhoff;
Pravin B Shelar490ab082013-02-22 07:30:30 +00001323 }
Alexander Duyck802ab552016-04-10 21:45:03 -04001324 iph->tot_len = htons(tot_len);
Eric Dumazet47d27aa2013-10-18 13:13:27 -07001325 ip_send_check(iph);
Eric Dumazet8c3a8972013-10-27 18:18:16 -07001326 if (encap)
Eric Dumazet3347c962013-10-19 11:42:56 -07001327 skb_reset_inner_headers(skb);
Eric Dumazet3347c962013-10-19 11:42:56 -07001328 skb->network_header = (u8 *)iph - skb->head;
Toke Høiland-Jørgensen13a47052018-09-13 16:43:07 +02001329 skb_reset_mac_len(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001330 } while ((skb = skb->next));
1331
1332out:
1333 return segs;
1334}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001335EXPORT_SYMBOL(inet_gso_segment);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001336
Tom Herbertb8921ca2016-05-18 09:06:23 -07001337struct sk_buff **inet_gro_receive(struct sk_buff **head, struct sk_buff *skb)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001338{
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001339 const struct net_offload *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001340 struct sk_buff **pp = NULL;
1341 struct sk_buff *p;
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001342 const struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001343 unsigned int hlen;
1344 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001345 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001346 int flush = 1;
1347 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001348
Herbert Xua5b1cf22009-05-26 18:50:28 +00001349 off = skb_gro_offset(skb);
1350 hlen = off + sizeof(*iph);
1351 iph = skb_gro_header_fast(skb, off);
1352 if (skb_gro_header_hard(skb, hlen)) {
1353 iph = skb_gro_header_slow(skb, hlen, off);
1354 if (unlikely(!iph))
1355 goto out;
1356 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001357
David S. Millerf9242b62012-06-19 18:56:21 -07001358 proto = iph->protocol;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001359
1360 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001361 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001362 if (!ops || !ops->callbacks.gro_receive)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001363 goto out_unlock;
1364
Herbert Xua5ad24b2009-02-08 18:00:39 +00001365 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001366 goto out_unlock;
1367
Steffen Klassert9b83e032017-04-28 10:54:32 +02001368 if (ip_is_fragment(iph))
1369 goto out_unlock;
1370
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001371 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001372 goto out_unlock;
1373
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001374 id = ntohl(*(__be32 *)&iph->id);
Eric Dumazetdb8caf32013-05-31 11:18:10 +00001375 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
Herbert Xu1075f3f2009-05-26 18:50:29 +00001376 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001377
1378 for (p = *head; p; p = p->next) {
1379 struct iphdr *iph2;
Alexander Duyck15305452016-04-10 21:44:57 -04001380 u16 flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001381
1382 if (!NAPI_GRO_CB(p)->same_flow)
1383 continue;
1384
Jerry Chu299603e82013-12-11 20:53:45 -08001385 iph2 = (struct iphdr *)(p->data + off);
1386 /* The above works because, with the exception of the top
1387 * (inner most) layer, we only aggregate pkts with the same
1388 * hdr length so all the hdrs we'll need to verify will start
1389 * at the same offset.
1390 */
Herbert Xua5ad24b2009-02-08 18:00:39 +00001391 if ((iph->protocol ^ iph2->protocol) |
Eric Dumazet0eae88f2010-04-20 19:06:52 -07001392 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1393 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001394 NAPI_GRO_CB(p)->same_flow = 0;
1395 continue;
1396 }
1397
1398 /* All fields must match except length and checksum. */
1399 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001400 (iph->ttl ^ iph2->ttl) |
Eric Dumazeta9e050f2012-08-05 22:34:50 +00001401 (iph->tos ^ iph2->tos) |
Jerry Chubf5a7552014-01-07 10:23:19 -08001402 ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001403
1404 NAPI_GRO_CB(p)->flush |= flush;
Alexander Duyck15305452016-04-10 21:44:57 -04001405
1406 /* We need to store of the IP ID check to be included later
1407 * when we can verify that this packet does in fact belong
1408 * to a given flow.
1409 */
1410 flush_id = (u16)(id - ntohs(iph2->id));
1411
1412 /* This bit of code makes it much easier for us to identify
1413 * the cases where we are doing atomic vs non-atomic IP ID
1414 * checks. Specifically an atomic check can return IP ID
1415 * values 0 - 0xFFFF, while a non-atomic check can only
1416 * return 0 or 0xFFFF.
1417 */
1418 if (!NAPI_GRO_CB(p)->is_atomic ||
1419 !(iph->frag_off & htons(IP_DF))) {
1420 flush_id ^= NAPI_GRO_CB(p)->count;
1421 flush_id = flush_id ? 0xFFFF : 0;
1422 }
1423
1424 /* If the previous IP ID value was based on an atomic
1425 * datagram we can overwrite the value and ignore it.
1426 */
1427 if (NAPI_GRO_CB(skb)->is_atomic)
1428 NAPI_GRO_CB(p)->flush_id = flush_id;
1429 else
1430 NAPI_GRO_CB(p)->flush_id |= flush_id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001431 }
1432
Alexander Duyck15305452016-04-10 21:44:57 -04001433 NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001434 NAPI_GRO_CB(skb)->flush |= flush;
Jerry Chu299603e82013-12-11 20:53:45 -08001435 skb_set_network_header(skb, off);
1436 /* The above will be needed by the transport layer if there is one
1437 * immediately following this IP hdr.
1438 */
1439
Eric Dumazet2c804d02014-09-30 22:12:05 -07001440 /* Note : No need to call skb_gro_postpull_rcsum() here,
1441 * as we already checked checksum over ipv4 header was 0
1442 */
Herbert Xu86911732009-01-29 14:19:50 +00001443 skb_gro_pull(skb, sizeof(*iph));
1444 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001445
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02001446 pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001447
1448out_unlock:
1449 rcu_read_unlock();
1450
1451out:
Steffen Klassert5f114162017-02-15 09:39:39 +01001452 skb_gro_flush_final(skb, pp, flush);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001453
1454 return pp;
1455}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001456EXPORT_SYMBOL(inet_gro_receive);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001457
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001458static struct sk_buff **ipip_gro_receive(struct sk_buff **head,
1459 struct sk_buff *skb)
1460{
1461 if (NAPI_GRO_CB(skb)->encap_mark) {
1462 NAPI_GRO_CB(skb)->flush = 1;
1463 return NULL;
1464 }
1465
1466 NAPI_GRO_CB(skb)->encap_mark = 1;
1467
1468 return inet_gro_receive(head, skb);
1469}
1470
Deepa Dinamani822c8682016-02-27 00:32:15 -08001471#define SECONDS_PER_DAY 86400
1472
1473/* inet_current_timestamp - Return IP network timestamp
1474 *
1475 * Return milliseconds since midnight in network byte order.
1476 */
1477__be32 inet_current_timestamp(void)
1478{
1479 u32 secs;
1480 u32 msecs;
1481 struct timespec64 ts;
1482
1483 ktime_get_real_ts64(&ts);
1484
1485 /* Get secs since midnight. */
1486 (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs);
1487 /* Convert to msecs. */
1488 msecs = secs * MSEC_PER_SEC;
1489 /* Convert nsec to msec. */
1490 msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC;
1491
1492 /* Convert to network byte order. */
Deepa Dinamani3ba9d302016-03-21 18:21:26 -07001493 return htonl(msecs);
Deepa Dinamani822c8682016-02-27 00:32:15 -08001494}
1495EXPORT_SYMBOL(inet_current_timestamp);
1496
Willem de Bruijnf4713a32014-11-26 14:53:02 -05001497int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1498{
1499 if (sk->sk_family == AF_INET)
1500 return ip_recv_error(sk, msg, len, addr_len);
1501#if IS_ENABLED(CONFIG_IPV6)
1502 if (sk->sk_family == AF_INET6)
1503 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1504#endif
1505 return -EINVAL;
1506}
1507
Tom Herbertb8921ca2016-05-18 09:06:23 -07001508int inet_gro_complete(struct sk_buff *skb, int nhoff)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001509{
Jerry Chu299603e82013-12-11 20:53:45 -08001510 __be16 newlen = htons(skb->len - nhoff);
1511 struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001512 const struct net_offload *ops;
David S. Millerf9242b62012-06-19 18:56:21 -07001513 int proto = iph->protocol;
1514 int err = -ENOSYS;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001515
Paolo Abeni294acf12017-03-07 18:33:31 +01001516 if (skb->encapsulation) {
1517 skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP));
Jerry Chuc3caf112014-07-14 15:54:46 -07001518 skb_set_inner_network_header(skb, nhoff);
Paolo Abeni294acf12017-03-07 18:33:31 +01001519 }
Jerry Chuc3caf112014-07-14 15:54:46 -07001520
Herbert Xu73cc19f2008-12-15 23:41:09 -08001521 csum_replace2(&iph->check, iph->tot_len, newlen);
1522 iph->tot_len = newlen;
1523
1524 rcu_read_lock();
Vlad Yasevichbca49f82012-11-15 08:49:14 +00001525 ops = rcu_dereference(inet_offloads[proto]);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001526 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
Herbert Xu73cc19f2008-12-15 23:41:09 -08001527 goto out_unlock;
1528
Jerry Chu299603e82013-12-11 20:53:45 -08001529 /* Only need to add sizeof(*iph) to get to the next hdr below
1530 * because any hdr with option will have been flushed in
1531 * inet_gro_receive().
1532 */
1533 err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001534
1535out_unlock:
1536 rcu_read_unlock();
1537
1538 return err;
1539}
Tom Herbertb8921ca2016-05-18 09:06:23 -07001540EXPORT_SYMBOL(inet_gro_complete);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001541
Jesse Grossb8cba752016-03-19 09:32:00 -07001542static int ipip_gro_complete(struct sk_buff *skb, int nhoff)
1543{
1544 skb->encapsulation = 1;
Tom Herbert7e133182016-05-18 09:06:10 -07001545 skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4;
Jesse Grossb8cba752016-03-19 09:32:00 -07001546 return inet_gro_complete(skb, nhoff);
1547}
1548
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001549int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001550 unsigned short type, unsigned char protocol,
1551 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001552{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001553 struct socket *sock;
Eric W. Biederman26abe142015-05-08 21:10:31 -05001554 int rc = sock_create_kern(net, family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001555
1556 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001557 *sk = sock->sk;
1558 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001559 /*
1560 * Unhash it so that IP input processing does not even see it,
1561 * we do not wish this socket to see incoming packets.
1562 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001563 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001564 }
1565 return rc;
1566}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001567EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1568
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301569u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
1570{
1571 return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
1572}
1573EXPORT_SYMBOL_GPL(snmp_get_cpu_field);
1574
WANG Cong698365f2014-05-05 15:55:55 -07001575unsigned long snmp_fold_field(void __percpu *mib, int offt)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001576{
1577 unsigned long res = 0;
WANG Cong698365f2014-05-05 15:55:55 -07001578 int i;
Herbert Xu5e0f0432007-04-24 21:53:35 -07001579
WANG Cong698365f2014-05-05 15:55:55 -07001580 for_each_possible_cpu(i)
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301581 res += snmp_get_cpu_field(mib, i, offt);
Herbert Xu5e0f0432007-04-24 21:53:35 -07001582 return res;
1583}
1584EXPORT_SYMBOL_GPL(snmp_fold_field);
1585
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001586#if BITS_PER_LONG==32
1587
David S. Miller80ec1922015-08-30 22:40:44 -07001588u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt,
Raghavendra K Tc4c6bc32015-08-30 11:29:41 +05301589 size_t syncp_offset)
1590{
1591 void *bhptr;
1592 struct u64_stats_sync *syncp;
1593 u64 v;
1594 unsigned int start;
1595
1596 bhptr = per_cpu_ptr(mib, cpu);
1597 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1598 do {
1599 start = u64_stats_fetch_begin_irq(syncp);
1600 v = *(((u64 *)bhptr) + offt);
1601 } while (u64_stats_fetch_retry_irq(syncp, start));
1602
1603 return v;
1604}
1605EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);
1606
WANG Cong698365f2014-05-05 15:55:55 -07001607u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001608{
1609 u64 res = 0;
1610 int cpu;
1611
1612 for_each_possible_cpu(cpu) {
Madalin Bucurd1bfc622015-08-31 15:46:07 +03001613 res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset);
Eric Dumazet4ce3c182010-06-30 13:31:19 -07001614 }
1615 return res;
1616}
1617EXPORT_SYMBOL_GPL(snmp_fold_field64);
1618#endif
1619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001621static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001623 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624};
1625#endif
1626
David Aherna8e3bb32017-08-28 15:14:20 -07001627/* thinking of making this const? Don't.
1628 * early_demux can change based on sysctl.
1629 */
David Ahern03157932017-08-28 13:23:09 -07001630static struct net_protocol tcp_protocol = {
David S. Miller41063e92012-06-19 21:22:05 -07001631 .early_demux = tcp_v4_early_demux,
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001632 .early_demux_handler = tcp_v4_early_demux,
David S. Miller41063e92012-06-19 21:22:05 -07001633 .handler = tcp_v4_rcv,
1634 .err_handler = tcp_v4_err,
David S. Miller41063e92012-06-19 21:22:05 -07001635 .no_policy = 1,
1636 .netns_ok = 1,
Hannes Frederic Sowa8ed1dc42014-01-09 10:01:17 +01001637 .icmp_strict_tag_validation = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638};
1639
David Aherna8e3bb32017-08-28 15:14:20 -07001640/* thinking of making this const? Don't.
1641 * early_demux can change based on sysctl.
1642 */
David Ahern03157932017-08-28 13:23:09 -07001643static struct net_protocol udp_protocol = {
Shawn Bohrer421b3882013-10-07 11:01:39 -05001644 .early_demux = udp_v4_early_demux,
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001645 .early_demux_handler = udp_v4_early_demux,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 .handler = udp_rcv,
1647 .err_handler = udp_err,
1648 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001649 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650};
1651
Alexey Dobriyan32613092009-09-14 12:21:47 +00001652static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 .handler = icmp_rcv,
Li Wei5b052042013-02-21 22:18:44 +00001654 .err_handler = icmp_err,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001655 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001656 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657};
1658
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001659static __net_init int ipv4_mib_init_net(struct net *net)
1660{
John Stultz827da442013-10-07 15:51:58 -07001661 int i;
1662
WANG Cong698365f2014-05-05 15:55:55 -07001663 net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1664 if (!net->mib.tcp_statistics)
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001665 goto err_tcp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001666 net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1667 if (!net->mib.ip_statistics)
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001668 goto err_ip_mib;
John Stultz827da442013-10-07 15:51:58 -07001669
1670 for_each_possible_cpu(i) {
1671 struct ipstats_mib *af_inet_stats;
WANG Cong698365f2014-05-05 15:55:55 -07001672 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
John Stultz827da442013-10-07 15:51:58 -07001673 u64_stats_init(&af_inet_stats->syncp);
John Stultz827da442013-10-07 15:51:58 -07001674 }
1675
WANG Cong698365f2014-05-05 15:55:55 -07001676 net->mib.net_statistics = alloc_percpu(struct linux_mib);
1677 if (!net->mib.net_statistics)
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001678 goto err_net_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001679 net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1680 if (!net->mib.udp_statistics)
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001681 goto err_udp_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001682 net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1683 if (!net->mib.udplite_statistics)
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001684 goto err_udplite_mib;
WANG Cong698365f2014-05-05 15:55:55 -07001685 net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1686 if (!net->mib.icmp_statistics)
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001687 goto err_icmp_mib;
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001688 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1689 GFP_KERNEL);
1690 if (!net->mib.icmpmsg_statistics)
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001691 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001692
1693 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001694 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001695
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001696err_icmpmsg_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001697 free_percpu(net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001698err_icmp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001699 free_percpu(net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001700err_udplite_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001701 free_percpu(net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001702err_udp_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001703 free_percpu(net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001704err_net_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001705 free_percpu(net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001706err_ip_mib:
WANG Cong698365f2014-05-05 15:55:55 -07001707 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001708err_tcp_mib:
1709 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001710}
1711
1712static __net_exit void ipv4_mib_exit_net(struct net *net)
1713{
Eric Dumazetacb32ba2011-11-08 13:04:43 +00001714 kfree(net->mib.icmpmsg_statistics);
WANG Cong698365f2014-05-05 15:55:55 -07001715 free_percpu(net->mib.icmp_statistics);
1716 free_percpu(net->mib.udplite_statistics);
1717 free_percpu(net->mib.udp_statistics);
1718 free_percpu(net->mib.net_statistics);
1719 free_percpu(net->mib.ip_statistics);
1720 free_percpu(net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001721}
1722
1723static __net_initdata struct pernet_operations ipv4_mib_ops = {
1724 .init = ipv4_mib_init_net,
1725 .exit = ipv4_mib_exit_net,
1726};
1727
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728static int __init init_ipv4_mibs(void)
1729{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001730 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731}
1732
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001733static __net_init int inet_init_net(struct net *net)
1734{
1735 /*
1736 * Set defaults for local port range
1737 */
1738 seqlock_init(&net->ipv4.ip_local_ports.lock);
1739 net->ipv4.ip_local_ports.range[0] = 32768;
Eric Dumazet07f4c902015-05-24 14:49:35 -07001740 net->ipv4.ip_local_ports.range[1] = 60999;
Cong Wangba6b9182014-05-06 11:02:50 -07001741
1742 seqlock_init(&net->ipv4.ping_group_range.lock);
1743 /*
1744 * Sane defaults - nobody may create ping sockets.
1745 * Boot scripts should set this to distro-specific group.
1746 */
1747 net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1748 net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001749
1750 /* Default values for sysctl-controlled parameters.
1751 * We set them here, in case sysctl is not compiled.
1752 */
1753 net->ipv4.sysctl_ip_default_ttl = IPDEFTTL;
1754 net->ipv4.sysctl_ip_dynaddr = 0;
1755 net->ipv4.sysctl_ip_early_demux = 1;
subashab@codeaurora.orgdddb64b2017-03-23 13:34:16 -06001756 net->ipv4.sysctl_udp_early_demux = 1;
1757 net->ipv4.sysctl_tcp_early_demux = 1;
Krister Johansen4548b682017-01-20 17:49:11 -08001758#ifdef CONFIG_SYSCTL
1759 net->ipv4.sysctl_ip_prot_sock = PROT_SOCK;
1760#endif
Ezequiel Garcia049bbf52016-05-20 13:21:10 -03001761
Nikolay Borisov17140202017-08-09 14:38:04 +03001762 /* Some igmp sysctl, whose values are always used */
1763 net->ipv4.sysctl_igmp_max_memberships = 20;
1764 net->ipv4.sysctl_igmp_max_msf = 10;
1765 /* IGMP reports for link-local multicast groups are enabled by default */
1766 net->ipv4.sysctl_igmp_llm_reports = 1;
1767 net->ipv4.sysctl_igmp_qrv = 2;
1768
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001769 return 0;
1770}
1771
1772static __net_exit void inet_exit_net(struct net *net)
1773{
1774}
1775
1776static __net_initdata struct pernet_operations af_inet_ops = {
1777 .init = inet_init_net,
1778 .exit = inet_exit_net,
1779};
1780
1781static int __init init_inet_pernet_ops(void)
1782{
1783 return register_pernet_subsys(&af_inet_ops);
1784}
1785
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
Herbert Xu30e224d2005-07-05 14:40:10 -07001788/*
1789 * IP protocol layer initialiser
1790 */
1791
Vlad Yasevich22061d82012-11-15 08:49:11 +00001792static struct packet_offload ip_packet_offload __read_mostly = {
1793 .type = cpu_to_be16(ETH_P_IP),
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001794 .callbacks = {
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00001795 .gso_segment = inet_gso_segment,
1796 .gro_receive = inet_gro_receive,
1797 .gro_complete = inet_gro_complete,
1798 },
Herbert Xu30e224d2005-07-05 14:40:10 -07001799};
1800
Eric Dumazetcb32f512013-10-19 11:42:57 -07001801static const struct net_offload ipip_offload = {
1802 .callbacks = {
Eric Dumazetcb32f512013-10-19 11:42:57 -07001803 .gso_segment = inet_gso_segment,
Jesse Grossfac8e0f2016-03-19 09:32:01 -07001804 .gro_receive = ipip_gro_receive,
Jesse Grossb8cba752016-03-19 09:32:00 -07001805 .gro_complete = ipip_gro_complete,
Eric Dumazetcb32f512013-10-19 11:42:57 -07001806 },
1807};
1808
Tonghao Zhang93b1b312017-08-02 09:34:15 -07001809static int __init ipip_offload_init(void)
1810{
1811 return inet_add_offload(&ipip_offload, IPPROTO_IPIP);
1812}
1813
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001814static int __init ipv4_offload_init(void)
1815{
1816 /*
1817 * Add offloads
1818 */
Daniel Borkmannda5bab02013-06-08 12:56:03 +02001819 if (udpv4_offload_init() < 0)
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001820 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001821 if (tcpv4_offload_init() < 0)
1822 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
Tonghao Zhang93b1b312017-08-02 09:34:15 -07001823 if (ipip_offload_init() < 0)
1824 pr_crit("%s: Cannot add IPIP protocol offload\n", __func__);
Vlad Yasevich808a8f82012-11-15 08:49:21 +00001825
1826 dev_add_offload(&ip_packet_offload);
1827 return 0;
1828}
1829
1830fs_initcall(ipv4_offload_init);
1831
1832static struct packet_type ip_packet_type __read_mostly = {
1833 .type = cpu_to_be16(ETH_P_IP),
1834 .func = ip_rcv,
1835};
1836
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837static int __init inet_init(void)
1838{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 struct inet_protosw *q;
1840 struct list_head *r;
1841 int rc = -EINVAL;
1842
Eyal Birgerb4772ef2015-03-01 14:58:29 +02001843 sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844
1845 rc = proto_register(&tcp_prot, 1);
1846 if (rc)
WANG Cong122ff242014-05-12 16:04:53 -07001847 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848
1849 rc = proto_register(&udp_prot, 1);
1850 if (rc)
1851 goto out_unregister_tcp_proto;
1852
1853 rc = proto_register(&raw_prot, 1);
1854 if (rc)
1855 goto out_unregister_udp_proto;
1856
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001857 rc = proto_register(&ping_prot, 1);
1858 if (rc)
1859 goto out_unregister_raw_proto;
1860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001862 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 */
1864
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001865 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866
Al Virobd7b1532008-07-15 16:00:59 -04001867#ifdef CONFIG_SYSCTL
1868 ip_static_sysctl_init();
1869#endif
1870
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 /*
1872 * Add all the base protocols.
1873 */
1874
1875 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001876 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001878 pr_crit("%s: Cannot add UDP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001880 pr_crit("%s: Cannot add TCP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881#ifdef CONFIG_IP_MULTICAST
1882 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
Joe Perches058bd4d2012-03-11 18:36:11 +00001883 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884#endif
1885
1886 /* Register the socket-side information for inet_create. */
1887 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1888 INIT_LIST_HEAD(r);
1889
1890 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1891 inet_register_protosw(q);
1892
1893 /*
1894 * Set the ARP module up
1895 */
1896
1897 arp_init();
1898
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001899 /*
1900 * Set the IP module up
1901 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902
1903 ip_init();
1904
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 /* Setup TCP slab cache for open requests. */
1906 tcp_init();
1907
Hideo Aoki95766ff2007-12-31 00:29:24 -08001908 /* Setup UDP memory threshold */
1909 udp_init();
1910
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001911 /* Add UDP-Lite (RFC 3828) */
1912 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001914 ping_init();
1915
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 /*
1917 * Set the ICMP layer up
1918 */
1919
Denis V. Luneva5710d62008-02-29 11:14:50 -08001920 if (icmp_init() < 0)
1921 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922
1923 /*
1924 * Initialise the multicast router
1925 */
1926#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001927 if (ip_mr_init())
Joe Perches058bd4d2012-03-11 18:36:11 +00001928 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929#endif
Cong Wangc9d8f1a2014-05-06 11:02:49 -07001930
1931 if (init_inet_pernet_ops())
1932 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 /*
1934 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001935 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
Stephen Hemminger132adf52007-03-08 20:44:43 -08001937 if (init_ipv4_mibs())
Joe Perches058bd4d2012-03-11 18:36:11 +00001938 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001939
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 ipv4_proc_init();
1941
1942 ipfrag_init();
1943
Herbert Xu30e224d2005-07-05 14:40:10 -07001944 dev_add_pack(&ip_packet_type);
1945
Thomas Graf045a0fa2015-07-23 10:08:44 +02001946 ip_tunnel_core_init();
1947
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 rc = 0;
1949out:
1950 return rc;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001951out_unregister_raw_proto:
1952 proto_unregister(&raw_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953out_unregister_udp_proto:
1954 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001955out_unregister_tcp_proto:
1956 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 goto out;
1958}
1959
Heiko Carstensa536e072006-04-28 15:19:17 -07001960fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961
1962/* ------------------------------------------------------------------------ */
1963
1964#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965static int __init ipv4_proc_init(void)
1966{
1967 int rc = 0;
1968
1969 if (raw_proc_init())
1970 goto out_raw;
1971 if (tcp4_proc_init())
1972 goto out_tcp;
1973 if (udp4_proc_init())
1974 goto out_udp;
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001975 if (ping_proc_init())
1976 goto out_ping;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 if (ip_misc_proc_init())
1978 goto out_misc;
1979out:
1980 return rc;
1981out_misc:
Vasiliy Kulikovc319b4d2011-05-13 10:01:00 +00001982 ping_proc_exit();
1983out_ping:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 udp4_proc_exit();
1985out_udp:
1986 tcp4_proc_exit();
1987out_tcp:
1988 raw_proc_exit();
1989out_raw:
1990 rc = -ENOMEM;
1991 goto out;
1992}
1993
1994#else /* CONFIG_PROC_FS */
1995static int __init ipv4_proc_init(void)
1996{
1997 return 0;
1998}
1999#endif /* CONFIG_PROC_FS */