Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * 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 | ||
afd46503 JP |
68 | #define pr_fmt(fmt) "IPv4: " fmt |
69 | ||
f4c50d99 | 70 | #include <linux/err.h> |
1da177e4 LT |
71 | #include <linux/errno.h> |
72 | #include <linux/types.h> | |
73 | #include <linux/socket.h> | |
74 | #include <linux/in.h> | |
75 | #include <linux/kernel.h> | |
1da177e4 LT |
76 | #include <linux/module.h> |
77 | #include <linux/sched.h> | |
78 | #include <linux/timer.h> | |
79 | #include <linux/string.h> | |
80 | #include <linux/sockios.h> | |
81 | #include <linux/net.h> | |
4fc268d2 | 82 | #include <linux/capability.h> |
1da177e4 LT |
83 | #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> | |
b3da2cf3 | 90 | #include <linux/random.h> |
5a0e3ad6 | 91 | #include <linux/slab.h> |
1da177e4 LT |
92 | |
93 | #include <asm/uaccess.h> | |
1da177e4 | 94 | |
1da177e4 LT |
95 | #include <linux/inet.h> |
96 | #include <linux/igmp.h> | |
14c85021 | 97 | #include <linux/inetdevice.h> |
1da177e4 | 98 | #include <linux/netdevice.h> |
73cc19f1 | 99 | #include <net/checksum.h> |
1da177e4 LT |
100 | #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> | |
295f7324 | 105 | #include <net/inet_connection_sock.h> |
1da177e4 LT |
106 | #include <net/tcp.h> |
107 | #include <net/udp.h> | |
ba4e58ec | 108 | #include <net/udplite.h> |
c319b4d7 | 109 | #include <net/ping.h> |
1da177e4 LT |
110 | #include <linux/skbuff.h> |
111 | #include <net/sock.h> | |
112 | #include <net/raw.h> | |
113 | #include <net/icmp.h> | |
1da177e4 | 114 | #include <net/inet_common.h> |
045a0fa0 | 115 | #include <net/ip_tunnels.h> |
1da177e4 | 116 | #include <net/xfrm.h> |
9b4661bd | 117 | #include <net/net_namespace.h> |
aebda156 | 118 | #include <net/secure_seq.h> |
1da177e4 LT |
119 | #ifdef CONFIG_IP_MROUTE |
120 | #include <linux/mroute.h> | |
121 | #endif | |
3236b004 | 122 | #include <net/l3mdev.h> |
1da177e4 | 123 | |
1da177e4 LT |
124 | |
125 | /* The inetsw table contains everything that inet_create needs to | |
126 | * build a new socket. | |
127 | */ | |
128 | static struct list_head inetsw[SOCK_MAX]; | |
129 | static DEFINE_SPINLOCK(inetsw_lock); | |
130 | ||
131 | /* New destruction routine */ | |
132 | ||
133 | void 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 | ||
95766fff HA |
140 | sk_mem_reclaim(sk); |
141 | ||
1da177e4 | 142 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 143 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
144 | sk->sk_state, sk); |
145 | return; | |
146 | } | |
147 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 148 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
149 | return; |
150 | } | |
151 | ||
547b792c IJ |
152 | 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); | |
1da177e4 | 156 | |
f6d8bd05 | 157 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
b6c6712a | 158 | dst_release(rcu_dereference_check(sk->sk_dst_cache, 1)); |
41063e9d | 159 | dst_release(sk->sk_rx_dst); |
e6848976 | 160 | sk_refcnt_debug_dec(sk); |
1da177e4 | 161 | } |
3d1427f8 | 162 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
163 | |
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 | ||
174 | static 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); | |
c720c7e8 | 180 | if (!inet->inet_num) { |
1da177e4 LT |
181 | if (sk->sk_prot->get_port(sk, 0)) { |
182 | release_sock(sk); | |
183 | return -EAGAIN; | |
184 | } | |
c720c7e8 | 185 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
186 | } |
187 | release_sock(sk); | |
188 | return 0; | |
189 | } | |
190 | ||
191 | /* | |
192 | * Move a socket into listening state. | |
193 | */ | |
194 | int 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) { | |
8336886f JC |
214 | /* Check special setups for testing purpose to enable TFO w/o |
215 | * requiring TCP_FASTOPEN sockopt. | |
216 | * Note that only TCP sockets (SOCK_STREAM) will reach here. | |
217 | * Also fastopenq may already been allocated because this | |
218 | * socket was in TCP_LISTEN state previously but was | |
219 | * shutdown() (rather than close()). | |
220 | */ | |
221 | if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 && | |
0536fcc0 | 222 | !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) { |
8336886f | 223 | if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0) |
0536fcc0 | 224 | fastopen_queue_tune(sk, backlog); |
8336886f JC |
225 | else if ((sysctl_tcp_fastopen & |
226 | TFO_SERVER_WO_SOCKOPT2) != 0) | |
0536fcc0 | 227 | fastopen_queue_tune(sk, |
8336886f | 228 | ((uint)sysctl_tcp_fastopen) >> 16); |
dfea2aa6 CP |
229 | |
230 | tcp_fastopen_init_key_once(true); | |
8336886f | 231 | } |
72a3effa | 232 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
233 | if (err) |
234 | goto out; | |
235 | } | |
236 | sk->sk_max_ack_backlog = backlog; | |
237 | err = 0; | |
238 | ||
239 | out: | |
240 | release_sock(sk); | |
241 | return err; | |
242 | } | |
3d1427f8 | 243 | EXPORT_SYMBOL(inet_listen); |
1da177e4 LT |
244 | |
245 | /* | |
246 | * Create an inet socket. | |
247 | */ | |
248 | ||
3f378b68 EP |
249 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
250 | int kern) | |
1da177e4 LT |
251 | { |
252 | struct sock *sk; | |
1da177e4 LT |
253 | struct inet_protosw *answer; |
254 | struct inet_sock *inet; | |
255 | struct proto *answer_prot; | |
256 | unsigned char answer_flags; | |
bb97d31f | 257 | int try_loading_module = 0; |
86c8f9d1 | 258 | int err; |
1da177e4 | 259 | |
79462ad0 HFS |
260 | if (protocol < 0 || protocol >= IPPROTO_MAX) |
261 | return -EINVAL; | |
262 | ||
1da177e4 LT |
263 | sock->state = SS_UNCONNECTED; |
264 | ||
265 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 266 | lookup_protocol: |
86c8f9d1 | 267 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 268 | rcu_read_lock(); |
696adfe8 | 269 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 270 | |
696adfe8 | 271 | err = 0; |
1da177e4 LT |
272 | /* Check the non-wild match. */ |
273 | if (protocol == answer->protocol) { | |
274 | if (protocol != IPPROTO_IP) | |
275 | break; | |
276 | } else { | |
277 | /* Check for the two wild cases. */ | |
278 | if (IPPROTO_IP == protocol) { | |
279 | protocol = answer->protocol; | |
280 | break; | |
281 | } | |
282 | if (IPPROTO_IP == answer->protocol) | |
283 | break; | |
284 | } | |
86c8f9d1 | 285 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
286 | } |
287 | ||
696adfe8 | 288 | if (unlikely(err)) { |
bb97d31f ACM |
289 | if (try_loading_module < 2) { |
290 | rcu_read_unlock(); | |
291 | /* | |
292 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
293 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
294 | */ | |
295 | if (++try_loading_module == 1) | |
296 | request_module("net-pf-%d-proto-%d-type-%d", | |
297 | PF_INET, protocol, sock->type); | |
298 | /* | |
299 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
300 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
301 | */ | |
302 | else | |
303 | request_module("net-pf-%d-proto-%d", | |
304 | PF_INET, protocol); | |
305 | goto lookup_protocol; | |
306 | } else | |
307 | goto out_rcu_unlock; | |
308 | } | |
309 | ||
1da177e4 | 310 | err = -EPERM; |
52e804c6 EB |
311 | if (sock->type == SOCK_RAW && !kern && |
312 | !ns_capable(net->user_ns, CAP_NET_RAW)) | |
1da177e4 | 313 | goto out_rcu_unlock; |
1da177e4 LT |
314 | |
315 | sock->ops = answer->ops; | |
316 | answer_prot = answer->prot; | |
1da177e4 LT |
317 | answer_flags = answer->flags; |
318 | rcu_read_unlock(); | |
319 | ||
51456b29 | 320 | WARN_ON(!answer_prot->slab); |
1da177e4 LT |
321 | |
322 | err = -ENOBUFS; | |
11aa9c28 | 323 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern); |
51456b29 | 324 | if (!sk) |
1da177e4 LT |
325 | goto out; |
326 | ||
327 | err = 0; | |
1da177e4 | 328 | if (INET_PROTOSW_REUSE & answer_flags) |
4a17fd52 | 329 | sk->sk_reuse = SK_CAN_REUSE; |
1da177e4 LT |
330 | |
331 | inet = inet_sk(sk); | |
469de9b9 | 332 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 | 333 | |
7b2ff18e JO |
334 | inet->nodefrag = 0; |
335 | ||
1da177e4 | 336 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 337 | inet->inet_num = protocol; |
1da177e4 LT |
338 | if (IPPROTO_RAW == protocol) |
339 | inet->hdrincl = 1; | |
340 | } | |
341 | ||
974eda11 | 342 | if (net->ipv4.sysctl_ip_no_pmtu_disc) |
1da177e4 LT |
343 | inet->pmtudisc = IP_PMTUDISC_DONT; |
344 | else | |
345 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
346 | ||
c720c7e8 | 347 | inet->inet_id = 0; |
1da177e4 LT |
348 | |
349 | sock_init_data(sock, sk); | |
350 | ||
351 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
352 | sk->sk_protocol = protocol; |
353 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
354 | ||
355 | inet->uc_ttl = -1; | |
356 | inet->mc_loop = 1; | |
357 | inet->mc_ttl = 1; | |
f771bef9 | 358 | inet->mc_all = 1; |
1da177e4 LT |
359 | inet->mc_index = 0; |
360 | inet->mc_list = NULL; | |
4c507d28 | 361 | inet->rcv_tos = 0; |
1da177e4 | 362 | |
e6848976 | 363 | sk_refcnt_debug_inc(sk); |
1da177e4 | 364 | |
c720c7e8 | 365 | if (inet->inet_num) { |
1da177e4 LT |
366 | /* It assumes that any protocol which allows |
367 | * the user to assign a number at socket | |
368 | * creation time automatically | |
369 | * shares. | |
370 | */ | |
c720c7e8 | 371 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 | 372 | /* Add to protocol hash chains. */ |
086c653f CG |
373 | err = sk->sk_prot->hash(sk); |
374 | if (err) { | |
375 | sk_common_release(sk); | |
376 | goto out; | |
377 | } | |
1da177e4 LT |
378 | } |
379 | ||
380 | if (sk->sk_prot->init) { | |
381 | err = sk->sk_prot->init(sk); | |
382 | if (err) | |
383 | sk_common_release(sk); | |
384 | } | |
385 | out: | |
386 | return err; | |
387 | out_rcu_unlock: | |
388 | rcu_read_unlock(); | |
389 | goto out; | |
390 | } | |
391 | ||
392 | ||
393 | /* | |
394 | * The peer socket should always be NULL (or else). When we call this | |
395 | * function we are destroying the object and from then on nobody | |
396 | * should refer to it. | |
397 | */ | |
398 | int inet_release(struct socket *sock) | |
399 | { | |
400 | struct sock *sk = sock->sk; | |
401 | ||
402 | if (sk) { | |
403 | long timeout; | |
404 | ||
405 | /* Applications forget to leave groups before exiting */ | |
406 | ip_mc_drop_socket(sk); | |
407 | ||
408 | /* If linger is set, we don't return until the close | |
409 | * is complete. Otherwise we return immediately. The | |
410 | * actually closing is done the same either way. | |
411 | * | |
412 | * If the close is due to the process exiting, we never | |
413 | * linger.. | |
414 | */ | |
415 | timeout = 0; | |
416 | if (sock_flag(sk, SOCK_LINGER) && | |
417 | !(current->flags & PF_EXITING)) | |
418 | timeout = sk->sk_lingertime; | |
419 | sock->sk = NULL; | |
420 | sk->sk_prot->close(sk, timeout); | |
421 | } | |
422 | return 0; | |
423 | } | |
3d1427f8 | 424 | EXPORT_SYMBOL(inet_release); |
1da177e4 | 425 | |
1da177e4 LT |
426 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
427 | { | |
428 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
429 | struct sock *sk = sock->sk; | |
430 | struct inet_sock *inet = inet_sk(sk); | |
3594698a | 431 | struct net *net = sock_net(sk); |
1da177e4 LT |
432 | unsigned short snum; |
433 | int chk_addr_ret; | |
9b8ff518 | 434 | u32 tb_id = RT_TABLE_LOCAL; |
1da177e4 LT |
435 | int err; |
436 | ||
437 | /* If the socket has its own bind function then use it. (RAW) */ | |
438 | if (sk->sk_prot->bind) { | |
439 | err = sk->sk_prot->bind(sk, uaddr, addr_len); | |
440 | goto out; | |
441 | } | |
442 | err = -EINVAL; | |
443 | if (addr_len < sizeof(struct sockaddr_in)) | |
444 | goto out; | |
445 | ||
c349a528 | 446 | if (addr->sin_family != AF_INET) { |
29c486df ED |
447 | /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) |
448 | * only if s_addr is INADDR_ANY. | |
449 | */ | |
c349a528 | 450 | err = -EAFNOSUPPORT; |
29c486df ED |
451 | if (addr->sin_family != AF_UNSPEC || |
452 | addr->sin_addr.s_addr != htonl(INADDR_ANY)) | |
453 | goto out; | |
c349a528 | 454 | } |
d0733d2e | 455 | |
3236b004 | 456 | tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; |
30bbaa19 | 457 | chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); |
1da177e4 LT |
458 | |
459 | /* Not specified by any standard per-se, however it breaks too | |
460 | * many applications when removed. It is unfortunate since | |
461 | * allowing applications to make a non-local bind solves | |
462 | * several problems with systems using dynamic addressing. | |
463 | * (ie. your servers still start up even if your ISDN link | |
464 | * is temporarily down) | |
465 | */ | |
466 | err = -EADDRNOTAVAIL; | |
49a60158 | 467 | if (!net->ipv4.sysctl_ip_nonlocal_bind && |
b9fb1506 | 468 | !(inet->freebind || inet->transparent) && |
e6f1cebf | 469 | addr->sin_addr.s_addr != htonl(INADDR_ANY) && |
1da177e4 LT |
470 | chk_addr_ret != RTN_LOCAL && |
471 | chk_addr_ret != RTN_MULTICAST && | |
472 | chk_addr_ret != RTN_BROADCAST) | |
473 | goto out; | |
474 | ||
475 | snum = ntohs(addr->sin_port); | |
476 | err = -EACCES; | |
3594698a EB |
477 | if (snum && snum < PROT_SOCK && |
478 | !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) | |
1da177e4 LT |
479 | goto out; |
480 | ||
481 | /* We keep a pair of addresses. rcv_saddr is the one | |
482 | * used by hash lookups, and saddr is used for transmit. | |
483 | * | |
484 | * In the BSD API these are the same except where it | |
485 | * would be illegal to use them (multicast/broadcast) in | |
486 | * which case the sending device address is used. | |
487 | */ | |
488 | lock_sock(sk); | |
489 | ||
490 | /* Check these errors (active socket, double bind). */ | |
491 | err = -EINVAL; | |
c720c7e8 | 492 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
493 | goto out_release_sock; |
494 | ||
c720c7e8 | 495 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 496 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 497 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
498 | |
499 | /* Make sure we are allowed to bind here. */ | |
90c337da ED |
500 | if ((snum || !inet->bind_address_no_port) && |
501 | sk->sk_prot->get_port(sk, snum)) { | |
c720c7e8 | 502 | inet->inet_saddr = inet->inet_rcv_saddr = 0; |
1da177e4 LT |
503 | err = -EADDRINUSE; |
504 | goto out_release_sock; | |
505 | } | |
506 | ||
c720c7e8 | 507 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
508 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
509 | if (snum) | |
510 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
511 | inet->inet_sport = htons(inet->inet_num); |
512 | inet->inet_daddr = 0; | |
513 | inet->inet_dport = 0; | |
1da177e4 LT |
514 | sk_dst_reset(sk); |
515 | err = 0; | |
516 | out_release_sock: | |
517 | release_sock(sk); | |
518 | out: | |
519 | return err; | |
520 | } | |
3d1427f8 | 521 | EXPORT_SYMBOL(inet_bind); |
1da177e4 | 522 | |
5e73ea1a | 523 | int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, |
1da177e4 LT |
524 | int addr_len, int flags) |
525 | { | |
526 | struct sock *sk = sock->sk; | |
527 | ||
6503d961 CG |
528 | if (addr_len < sizeof(uaddr->sa_family)) |
529 | return -EINVAL; | |
1da177e4 LT |
530 | if (uaddr->sa_family == AF_UNSPEC) |
531 | return sk->sk_prot->disconnect(sk, flags); | |
532 | ||
c720c7e8 | 533 | if (!inet_sk(sk)->inet_num && inet_autobind(sk)) |
1da177e4 | 534 | return -EAGAIN; |
e3192690 | 535 | return sk->sk_prot->connect(sk, uaddr, addr_len); |
1da177e4 | 536 | } |
3d1427f8 | 537 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 | 538 | |
783237e8 | 539 | static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) |
1da177e4 LT |
540 | { |
541 | DEFINE_WAIT(wait); | |
542 | ||
aa395145 | 543 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
783237e8 | 544 | sk->sk_write_pending += writebias; |
1da177e4 LT |
545 | |
546 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
547 | * change state of the socket from TCP_SYN_*. | |
548 | * Connect() does not allow to get error notifications | |
549 | * without closing the socket. | |
550 | */ | |
551 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
552 | release_sock(sk); | |
553 | timeo = schedule_timeout(timeo); | |
554 | lock_sock(sk); | |
555 | if (signal_pending(current) || !timeo) | |
556 | break; | |
aa395145 | 557 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 558 | } |
aa395145 | 559 | finish_wait(sk_sleep(sk), &wait); |
783237e8 | 560 | sk->sk_write_pending -= writebias; |
1da177e4 LT |
561 | return timeo; |
562 | } | |
563 | ||
564 | /* | |
565 | * Connect to a remote host. There is regrettably still a little | |
566 | * TCP 'magic' in here. | |
567 | */ | |
cf60af03 YC |
568 | int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, |
569 | int addr_len, int flags) | |
1da177e4 LT |
570 | { |
571 | struct sock *sk = sock->sk; | |
572 | int err; | |
573 | long timeo; | |
574 | ||
6503d961 CG |
575 | if (addr_len < sizeof(uaddr->sa_family)) |
576 | return -EINVAL; | |
577 | ||
1da177e4 LT |
578 | if (uaddr->sa_family == AF_UNSPEC) { |
579 | err = sk->sk_prot->disconnect(sk, flags); | |
580 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
581 | goto out; | |
582 | } | |
583 | ||
584 | switch (sock->state) { | |
585 | default: | |
586 | err = -EINVAL; | |
587 | goto out; | |
588 | case SS_CONNECTED: | |
589 | err = -EISCONN; | |
590 | goto out; | |
591 | case SS_CONNECTING: | |
592 | err = -EALREADY; | |
593 | /* Fall out of switch with err, set for this state */ | |
594 | break; | |
595 | case SS_UNCONNECTED: | |
596 | err = -EISCONN; | |
597 | if (sk->sk_state != TCP_CLOSE) | |
598 | goto out; | |
599 | ||
600 | err = sk->sk_prot->connect(sk, uaddr, addr_len); | |
601 | if (err < 0) | |
602 | goto out; | |
603 | ||
e905a9ed | 604 | sock->state = SS_CONNECTING; |
1da177e4 LT |
605 | |
606 | /* Just entered SS_CONNECTING state; the only | |
607 | * difference is that return value in non-blocking | |
608 | * case is EINPROGRESS, rather than EALREADY. | |
609 | */ | |
610 | err = -EINPROGRESS; | |
611 | break; | |
612 | } | |
613 | ||
614 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
615 | ||
616 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
783237e8 YC |
617 | int writebias = (sk->sk_protocol == IPPROTO_TCP) && |
618 | tcp_sk(sk)->fastopen_req && | |
619 | tcp_sk(sk)->fastopen_req->data ? 1 : 0; | |
620 | ||
1da177e4 | 621 | /* Error code is set above */ |
783237e8 | 622 | if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) |
1da177e4 LT |
623 | goto out; |
624 | ||
625 | err = sock_intr_errno(timeo); | |
626 | if (signal_pending(current)) | |
627 | goto out; | |
628 | } | |
629 | ||
630 | /* Connection was closed by RST, timeout, ICMP error | |
631 | * or another process disconnected us. | |
632 | */ | |
633 | if (sk->sk_state == TCP_CLOSE) | |
634 | goto sock_error; | |
635 | ||
636 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
637 | * and error was received after socket entered established state. | |
638 | * Hence, it is handled normally after connect() return successfully. | |
639 | */ | |
640 | ||
641 | sock->state = SS_CONNECTED; | |
642 | err = 0; | |
643 | out: | |
1da177e4 LT |
644 | return err; |
645 | ||
646 | sock_error: | |
647 | err = sock_error(sk) ? : -ECONNABORTED; | |
648 | sock->state = SS_UNCONNECTED; | |
649 | if (sk->sk_prot->disconnect(sk, flags)) | |
650 | sock->state = SS_DISCONNECTING; | |
651 | goto out; | |
652 | } | |
cf60af03 YC |
653 | EXPORT_SYMBOL(__inet_stream_connect); |
654 | ||
655 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
656 | int addr_len, int flags) | |
657 | { | |
658 | int err; | |
659 | ||
660 | lock_sock(sock->sk); | |
661 | err = __inet_stream_connect(sock, uaddr, addr_len, flags); | |
662 | release_sock(sock->sk); | |
663 | return err; | |
664 | } | |
3d1427f8 | 665 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 LT |
666 | |
667 | /* | |
668 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
669 | */ | |
670 | ||
671 | int inet_accept(struct socket *sock, struct socket *newsock, int flags) | |
672 | { | |
673 | struct sock *sk1 = sock->sk; | |
674 | int err = -EINVAL; | |
675 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err); | |
676 | ||
677 | if (!sk2) | |
678 | goto do_err; | |
679 | ||
680 | lock_sock(sk2); | |
681 | ||
1eddcead | 682 | sock_rps_record_flow(sk2); |
547b792c | 683 | WARN_ON(!((1 << sk2->sk_state) & |
8336886f JC |
684 | (TCPF_ESTABLISHED | TCPF_SYN_RECV | |
685 | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
1da177e4 LT |
686 | |
687 | sock_graft(sk2, newsock); | |
688 | ||
689 | newsock->state = SS_CONNECTED; | |
690 | err = 0; | |
691 | release_sock(sk2); | |
692 | do_err: | |
693 | return err; | |
694 | } | |
3d1427f8 | 695 | EXPORT_SYMBOL(inet_accept); |
1da177e4 LT |
696 | |
697 | ||
698 | /* | |
699 | * This does both peername and sockname. | |
700 | */ | |
701 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
702 | int *uaddr_len, int peer) | |
703 | { | |
704 | struct sock *sk = sock->sk; | |
705 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 706 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
707 | |
708 | sin->sin_family = AF_INET; | |
709 | if (peer) { | |
c720c7e8 | 710 | if (!inet->inet_dport || |
1da177e4 LT |
711 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
712 | peer == 1)) | |
713 | return -ENOTCONN; | |
c720c7e8 ED |
714 | sin->sin_port = inet->inet_dport; |
715 | sin->sin_addr.s_addr = inet->inet_daddr; | |
1da177e4 | 716 | } else { |
c720c7e8 | 717 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 718 | if (!addr) |
c720c7e8 ED |
719 | addr = inet->inet_saddr; |
720 | sin->sin_port = inet->inet_sport; | |
1da177e4 LT |
721 | sin->sin_addr.s_addr = addr; |
722 | } | |
723 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
724 | *uaddr_len = sizeof(*sin); | |
725 | return 0; | |
726 | } | |
3d1427f8 | 727 | EXPORT_SYMBOL(inet_getname); |
1da177e4 | 728 | |
1b784140 | 729 | int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) |
1da177e4 LT |
730 | { |
731 | struct sock *sk = sock->sk; | |
732 | ||
c58dc01b | 733 | sock_rps_record_flow(sk); |
fec5e652 | 734 | |
1da177e4 | 735 | /* We may need to bind the socket. */ |
7ba42910 CG |
736 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
737 | inet_autobind(sk)) | |
1da177e4 LT |
738 | return -EAGAIN; |
739 | ||
1b784140 | 740 | return sk->sk_prot->sendmsg(sk, msg, size); |
1da177e4 | 741 | } |
3d1427f8 | 742 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 743 | |
7ba42910 CG |
744 | ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, |
745 | size_t size, int flags) | |
1da177e4 LT |
746 | { |
747 | struct sock *sk = sock->sk; | |
748 | ||
c58dc01b | 749 | sock_rps_record_flow(sk); |
fec5e652 | 750 | |
1da177e4 | 751 | /* We may need to bind the socket. */ |
7ba42910 CG |
752 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
753 | inet_autobind(sk)) | |
1da177e4 LT |
754 | return -EAGAIN; |
755 | ||
756 | if (sk->sk_prot->sendpage) | |
757 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
758 | return sock_no_sendpage(sock, page, offset, size, flags); | |
759 | } | |
7ba42910 | 760 | EXPORT_SYMBOL(inet_sendpage); |
1da177e4 | 761 | |
1b784140 YX |
762 | int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
763 | int flags) | |
fec5e652 TH |
764 | { |
765 | struct sock *sk = sock->sk; | |
766 | int addr_len = 0; | |
767 | int err; | |
768 | ||
c58dc01b | 769 | sock_rps_record_flow(sk); |
fec5e652 | 770 | |
1b784140 | 771 | err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT, |
fec5e652 TH |
772 | flags & ~MSG_DONTWAIT, &addr_len); |
773 | if (err >= 0) | |
774 | msg->msg_namelen = addr_len; | |
775 | return err; | |
776 | } | |
777 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
778 | |
779 | int inet_shutdown(struct socket *sock, int how) | |
780 | { | |
781 | struct sock *sk = sock->sk; | |
782 | int err = 0; | |
783 | ||
784 | /* This should really check to make sure | |
785 | * the socket is a TCP socket. (WHY AC...) | |
786 | */ | |
787 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
788 | 1->2 bit 2 snds. | |
789 | 2->3 */ | |
790 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
791 | return -EINVAL; | |
792 | ||
793 | lock_sock(sk); | |
794 | if (sock->state == SS_CONNECTING) { | |
795 | if ((1 << sk->sk_state) & | |
796 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
797 | sock->state = SS_DISCONNECTING; | |
798 | else | |
799 | sock->state = SS_CONNECTED; | |
800 | } | |
801 | ||
802 | switch (sk->sk_state) { | |
803 | case TCP_CLOSE: | |
804 | err = -ENOTCONN; | |
805 | /* Hack to wake up other listeners, who can poll for | |
806 | POLLHUP, even on eg. unconnected UDP sockets -- RR */ | |
807 | default: | |
808 | sk->sk_shutdown |= how; | |
809 | if (sk->sk_prot->shutdown) | |
810 | sk->sk_prot->shutdown(sk, how); | |
811 | break; | |
812 | ||
813 | /* Remaining two branches are temporary solution for missing | |
814 | * close() in multithreaded environment. It is _not_ a good idea, | |
815 | * but we have no choice until close() is repaired at VFS level. | |
816 | */ | |
817 | case TCP_LISTEN: | |
818 | if (!(how & RCV_SHUTDOWN)) | |
819 | break; | |
820 | /* Fall through */ | |
821 | case TCP_SYN_SENT: | |
822 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
823 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
824 | break; | |
825 | } | |
826 | ||
827 | /* Wake up anyone sleeping in poll. */ | |
828 | sk->sk_state_change(sk); | |
829 | release_sock(sk); | |
830 | return err; | |
831 | } | |
3d1427f8 | 832 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
833 | |
834 | /* | |
835 | * ioctl() calls you can issue on an INET socket. Most of these are | |
836 | * device configuration and stuff and very rarely used. Some ioctls | |
837 | * pass on to the socket itself. | |
838 | * | |
839 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
840 | * loads the devconfigure module does its configuring and unloads it. | |
841 | * There's a good 20K of config code hanging around the kernel. | |
842 | */ | |
843 | ||
844 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
845 | { | |
846 | struct sock *sk = sock->sk; | |
847 | int err = 0; | |
3b1e0a65 | 848 | struct net *net = sock_net(sk); |
1da177e4 LT |
849 | |
850 | switch (cmd) { | |
3d1427f8 ED |
851 | case SIOCGSTAMP: |
852 | err = sock_get_timestamp(sk, (struct timeval __user *)arg); | |
853 | break; | |
854 | case SIOCGSTAMPNS: | |
855 | err = sock_get_timestampns(sk, (struct timespec __user *)arg); | |
856 | break; | |
857 | case SIOCADDRT: | |
858 | case SIOCDELRT: | |
859 | case SIOCRTMSG: | |
860 | err = ip_rt_ioctl(net, cmd, (void __user *)arg); | |
861 | break; | |
862 | case SIOCDARP: | |
863 | case SIOCGARP: | |
864 | case SIOCSARP: | |
865 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
866 | break; | |
867 | case SIOCGIFADDR: | |
868 | case SIOCSIFADDR: | |
869 | case SIOCGIFBRDADDR: | |
870 | case SIOCSIFBRDADDR: | |
871 | case SIOCGIFNETMASK: | |
872 | case SIOCSIFNETMASK: | |
873 | case SIOCGIFDSTADDR: | |
874 | case SIOCSIFDSTADDR: | |
875 | case SIOCSIFPFLAGS: | |
876 | case SIOCGIFPFLAGS: | |
877 | case SIOCSIFFLAGS: | |
878 | err = devinet_ioctl(net, cmd, (void __user *)arg); | |
879 | break; | |
880 | default: | |
881 | if (sk->sk_prot->ioctl) | |
882 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
883 | else | |
884 | err = -ENOIOCTLCMD; | |
885 | break; | |
1da177e4 LT |
886 | } |
887 | return err; | |
888 | } | |
3d1427f8 | 889 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 890 | |
709b46e8 | 891 | #ifdef CONFIG_COMPAT |
686dc6b6 | 892 | static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
709b46e8 EB |
893 | { |
894 | struct sock *sk = sock->sk; | |
895 | int err = -ENOIOCTLCMD; | |
896 | ||
897 | if (sk->sk_prot->compat_ioctl) | |
898 | err = sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
899 | ||
900 | return err; | |
901 | } | |
902 | #endif | |
903 | ||
90ddc4f0 | 904 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
905 | .family = PF_INET, |
906 | .owner = THIS_MODULE, | |
907 | .release = inet_release, | |
908 | .bind = inet_bind, | |
909 | .connect = inet_stream_connect, | |
910 | .socketpair = sock_no_socketpair, | |
911 | .accept = inet_accept, | |
912 | .getname = inet_getname, | |
913 | .poll = tcp_poll, | |
914 | .ioctl = inet_ioctl, | |
915 | .listen = inet_listen, | |
916 | .shutdown = inet_shutdown, | |
917 | .setsockopt = sock_common_setsockopt, | |
918 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 919 | .sendmsg = inet_sendmsg, |
fec5e652 | 920 | .recvmsg = inet_recvmsg, |
543d9cfe | 921 | .mmap = sock_no_mmap, |
7ba42910 | 922 | .sendpage = inet_sendpage, |
9c55e01c | 923 | .splice_read = tcp_splice_read, |
3fdadf7d | 924 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
925 | .compat_setsockopt = compat_sock_common_setsockopt, |
926 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 927 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 928 | #endif |
1da177e4 | 929 | }; |
3d1427f8 | 930 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 931 | |
90ddc4f0 | 932 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
933 | .family = PF_INET, |
934 | .owner = THIS_MODULE, | |
935 | .release = inet_release, | |
936 | .bind = inet_bind, | |
937 | .connect = inet_dgram_connect, | |
938 | .socketpair = sock_no_socketpair, | |
939 | .accept = sock_no_accept, | |
940 | .getname = inet_getname, | |
941 | .poll = udp_poll, | |
942 | .ioctl = inet_ioctl, | |
943 | .listen = sock_no_listen, | |
944 | .shutdown = inet_shutdown, | |
945 | .setsockopt = sock_common_setsockopt, | |
946 | .getsockopt = sock_common_getsockopt, | |
947 | .sendmsg = inet_sendmsg, | |
fec5e652 | 948 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
949 | .mmap = sock_no_mmap, |
950 | .sendpage = inet_sendpage, | |
627d2d6b | 951 | .set_peek_off = sk_set_peek_off, |
3fdadf7d | 952 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
953 | .compat_setsockopt = compat_sock_common_setsockopt, |
954 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 955 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 956 | #endif |
1da177e4 | 957 | }; |
3d1427f8 | 958 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
959 | |
960 | /* | |
961 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
962 | * udp_poll | |
963 | */ | |
90ddc4f0 | 964 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
965 | .family = PF_INET, |
966 | .owner = THIS_MODULE, | |
967 | .release = inet_release, | |
968 | .bind = inet_bind, | |
969 | .connect = inet_dgram_connect, | |
970 | .socketpair = sock_no_socketpair, | |
971 | .accept = sock_no_accept, | |
972 | .getname = inet_getname, | |
973 | .poll = datagram_poll, | |
974 | .ioctl = inet_ioctl, | |
975 | .listen = sock_no_listen, | |
976 | .shutdown = inet_shutdown, | |
977 | .setsockopt = sock_common_setsockopt, | |
978 | .getsockopt = sock_common_getsockopt, | |
979 | .sendmsg = inet_sendmsg, | |
fec5e652 | 980 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
981 | .mmap = sock_no_mmap, |
982 | .sendpage = inet_sendpage, | |
3fdadf7d | 983 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
984 | .compat_setsockopt = compat_sock_common_setsockopt, |
985 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 986 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 987 | #endif |
1da177e4 LT |
988 | }; |
989 | ||
ec1b4cf7 | 990 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
991 | .family = PF_INET, |
992 | .create = inet_create, | |
993 | .owner = THIS_MODULE, | |
994 | }; | |
995 | ||
1da177e4 LT |
996 | /* Upon startup we insert all the elements in inetsw_array[] into |
997 | * the linked list inetsw. | |
998 | */ | |
999 | static struct inet_protosw inetsw_array[] = | |
1000 | { | |
e905a9ed YH |
1001 | { |
1002 | .type = SOCK_STREAM, | |
1003 | .protocol = IPPROTO_TCP, | |
1004 | .prot = &tcp_prot, | |
1005 | .ops = &inet_stream_ops, | |
e905a9ed | 1006 | .flags = INET_PROTOSW_PERMANENT | |
d83d8461 | 1007 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1008 | }, |
1009 | ||
1010 | { | |
1011 | .type = SOCK_DGRAM, | |
1012 | .protocol = IPPROTO_UDP, | |
1013 | .prot = &udp_prot, | |
1014 | .ops = &inet_dgram_ops, | |
e905a9ed | 1015 | .flags = INET_PROTOSW_PERMANENT, |
1da177e4 | 1016 | }, |
e905a9ed | 1017 | |
c319b4d7 VK |
1018 | { |
1019 | .type = SOCK_DGRAM, | |
1020 | .protocol = IPPROTO_ICMP, | |
1021 | .prot = &ping_prot, | |
1022 | .ops = &inet_dgram_ops, | |
c319b4d7 VK |
1023 | .flags = INET_PROTOSW_REUSE, |
1024 | }, | |
1da177e4 LT |
1025 | |
1026 | { | |
e905a9ed YH |
1027 | .type = SOCK_RAW, |
1028 | .protocol = IPPROTO_IP, /* wild card */ | |
1029 | .prot = &raw_prot, | |
1030 | .ops = &inet_sockraw_ops, | |
e905a9ed | 1031 | .flags = INET_PROTOSW_REUSE, |
1da177e4 LT |
1032 | } |
1033 | }; | |
1034 | ||
c40f6fff | 1035 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1036 | |
1037 | void inet_register_protosw(struct inet_protosw *p) | |
1038 | { | |
1039 | struct list_head *lh; | |
1040 | struct inet_protosw *answer; | |
1041 | int protocol = p->protocol; | |
1042 | struct list_head *last_perm; | |
1043 | ||
1044 | spin_lock_bh(&inetsw_lock); | |
1045 | ||
1046 | if (p->type >= SOCK_MAX) | |
1047 | goto out_illegal; | |
1048 | ||
1049 | /* If we are trying to override a permanent protocol, bail. */ | |
1da177e4 LT |
1050 | last_perm = &inetsw[p->type]; |
1051 | list_for_each(lh, &inetsw[p->type]) { | |
1052 | answer = list_entry(lh, struct inet_protosw, list); | |
1da177e4 | 1053 | /* Check only the non-wild match. */ |
f6c53334 JZ |
1054 | if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) |
1055 | break; | |
1056 | if (protocol == answer->protocol) | |
1057 | goto out_permanent; | |
1058 | last_perm = lh; | |
1da177e4 | 1059 | } |
1da177e4 LT |
1060 | |
1061 | /* Add the new entry after the last permanent entry if any, so that | |
1062 | * the new entry does not override a permanent entry when matched with | |
1063 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1064 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1065 | * system automatically returns to the old behavior. |
1066 | */ | |
1067 | list_add_rcu(&p->list, last_perm); | |
1068 | out: | |
1069 | spin_unlock_bh(&inetsw_lock); | |
1070 | ||
1da177e4 LT |
1071 | return; |
1072 | ||
1073 | out_permanent: | |
058bd4d2 | 1074 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
1da177e4 LT |
1075 | goto out; |
1076 | ||
1077 | out_illegal: | |
058bd4d2 | 1078 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
1da177e4 LT |
1079 | p->type); |
1080 | goto out; | |
1081 | } | |
3d1427f8 | 1082 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1083 | |
1084 | void inet_unregister_protosw(struct inet_protosw *p) | |
1085 | { | |
1086 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
058bd4d2 | 1087 | pr_err("Attempt to unregister permanent protocol %d\n", |
1da177e4 LT |
1088 | p->protocol); |
1089 | } else { | |
1090 | spin_lock_bh(&inetsw_lock); | |
1091 | list_del_rcu(&p->list); | |
1092 | spin_unlock_bh(&inetsw_lock); | |
1093 | ||
1094 | synchronize_net(); | |
1095 | } | |
1096 | } | |
3d1427f8 | 1097 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1098 | |
32519f11 ACM |
1099 | static int inet_sk_reselect_saddr(struct sock *sk) |
1100 | { | |
1101 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1102 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1103 | __be32 daddr = inet->inet_daddr; |
6e869138 | 1104 | struct flowi4 *fl4; |
b23dd4fe DM |
1105 | struct rtable *rt; |
1106 | __be32 new_saddr; | |
f6d8bd05 | 1107 | struct ip_options_rcu *inet_opt; |
32519f11 | 1108 | |
f6d8bd05 | 1109 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 1110 | lockdep_sock_is_held(sk)); |
f6d8bd05 ED |
1111 | if (inet_opt && inet_opt->opt.srr) |
1112 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1113 | |
1114 | /* Query new route. */ | |
6e869138 DM |
1115 | fl4 = &inet->cork.fl.u.ip4; |
1116 | rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk), | |
b23dd4fe | 1117 | sk->sk_bound_dev_if, sk->sk_protocol, |
0e0d44ab | 1118 | inet->inet_sport, inet->inet_dport, sk); |
b23dd4fe DM |
1119 | if (IS_ERR(rt)) |
1120 | return PTR_ERR(rt); | |
32519f11 | 1121 | |
d8d1f30b | 1122 | sk_setup_caps(sk, &rt->dst); |
32519f11 | 1123 | |
6e869138 | 1124 | new_saddr = fl4->saddr; |
32519f11 ACM |
1125 | |
1126 | if (new_saddr == old_saddr) | |
1127 | return 0; | |
1128 | ||
287b7f38 | 1129 | if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) { |
058bd4d2 JP |
1130 | pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1131 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1132 | } |
1133 | ||
c720c7e8 | 1134 | inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; |
32519f11 ACM |
1135 | |
1136 | /* | |
1137 | * XXX The only one ugly spot where we need to | |
1138 | * XXX really change the sockets identity after | |
1139 | * XXX it has entered the hashes. -DaveM | |
1140 | * | |
1141 | * Besides that, it does not check for connection | |
1142 | * uniqueness. Wait for troubles. | |
1143 | */ | |
086c653f | 1144 | return __sk_prot_rehash(sk); |
32519f11 ACM |
1145 | } |
1146 | ||
1147 | int inet_sk_rebuild_header(struct sock *sk) | |
1148 | { | |
1149 | struct inet_sock *inet = inet_sk(sk); | |
1150 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1151 | __be32 daddr; |
f6d8bd05 | 1152 | struct ip_options_rcu *inet_opt; |
6e869138 | 1153 | struct flowi4 *fl4; |
32519f11 ACM |
1154 | int err; |
1155 | ||
1156 | /* Route is OK, nothing to do. */ | |
1157 | if (rt) | |
1158 | return 0; | |
1159 | ||
1160 | /* Reroute. */ | |
f6d8bd05 ED |
1161 | rcu_read_lock(); |
1162 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1163 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1164 | if (inet_opt && inet_opt->opt.srr) |
1165 | daddr = inet_opt->opt.faddr; | |
1166 | rcu_read_unlock(); | |
6e869138 DM |
1167 | fl4 = &inet->cork.fl.u.ip4; |
1168 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, | |
78fbfd8a DM |
1169 | inet->inet_dport, inet->inet_sport, |
1170 | sk->sk_protocol, RT_CONN_FLAGS(sk), | |
1171 | sk->sk_bound_dev_if); | |
b23dd4fe DM |
1172 | if (!IS_ERR(rt)) { |
1173 | err = 0; | |
d8d1f30b | 1174 | sk_setup_caps(sk, &rt->dst); |
b23dd4fe DM |
1175 | } else { |
1176 | err = PTR_ERR(rt); | |
1177 | ||
32519f11 ACM |
1178 | /* Routing failed... */ |
1179 | sk->sk_route_caps = 0; | |
1180 | /* | |
1181 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1182 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1183 | */ | |
287b7f38 | 1184 | if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr || |
32519f11 ACM |
1185 | sk->sk_state != TCP_SYN_SENT || |
1186 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1187 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
1188 | sk->sk_err_soft = -err; | |
1189 | } | |
1190 | ||
1191 | return err; | |
1192 | } | |
32519f11 ACM |
1193 | EXPORT_SYMBOL(inet_sk_rebuild_header); |
1194 | ||
c8f44aff | 1195 | static struct sk_buff *inet_gso_segment(struct sk_buff *skb, |
3347c960 | 1196 | netdev_features_t features) |
f4c50d99 | 1197 | { |
cbc53e08 | 1198 | bool udpfrag = false, fixedid = false, encap; |
f4c50d99 | 1199 | struct sk_buff *segs = ERR_PTR(-EINVAL); |
bca49f84 | 1200 | const struct net_offload *ops; |
3347c960 | 1201 | unsigned int offset = 0; |
f9242b6b | 1202 | struct iphdr *iph; |
802ab55a | 1203 | int proto, tot_len; |
3347c960 | 1204 | int nhoff; |
f4c50d99 HX |
1205 | int ihl; |
1206 | int id; | |
1207 | ||
bbcf467d HX |
1208 | if (unlikely(skb_shinfo(skb)->gso_type & |
1209 | ~(SKB_GSO_TCPV4 | | |
1210 | SKB_GSO_UDP | | |
1211 | SKB_GSO_DODGY | | |
1212 | SKB_GSO_TCP_ECN | | |
68c33163 | 1213 | SKB_GSO_GRE | |
4749c09c | 1214 | SKB_GSO_GRE_CSUM | |
cb32f511 | 1215 | SKB_GSO_IPIP | |
61c1db7f | 1216 | SKB_GSO_SIT | |
9b3eb5ed | 1217 | SKB_GSO_TCPV6 | |
73136267 | 1218 | SKB_GSO_UDP_TUNNEL | |
0f4f4ffa | 1219 | SKB_GSO_UDP_TUNNEL_CSUM | |
cbc53e08 | 1220 | SKB_GSO_TCP_FIXEDID | |
e585f236 | 1221 | SKB_GSO_TUNNEL_REMCSUM | |
802ab55a | 1222 | SKB_GSO_PARTIAL | |
bbcf467d HX |
1223 | 0))) |
1224 | goto out; | |
1225 | ||
3347c960 ED |
1226 | skb_reset_network_header(skb); |
1227 | nhoff = skb_network_header(skb) - skb_mac_header(skb); | |
bbcf467d | 1228 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) |
f4c50d99 HX |
1229 | goto out; |
1230 | ||
eddc9ec5 | 1231 | iph = ip_hdr(skb); |
f4c50d99 HX |
1232 | ihl = iph->ihl * 4; |
1233 | if (ihl < sizeof(*iph)) | |
1234 | goto out; | |
1235 | ||
47d27aad ED |
1236 | id = ntohs(iph->id); |
1237 | proto = iph->protocol; | |
1238 | ||
1239 | /* Warning: after this point, iph might be no longer valid */ | |
bbcf467d | 1240 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 | 1241 | goto out; |
47d27aad | 1242 | __skb_pull(skb, ihl); |
f4c50d99 | 1243 | |
8c3a897b ED |
1244 | encap = SKB_GSO_CB(skb)->encap_level > 0; |
1245 | if (encap) | |
1e16aa3d | 1246 | features &= skb->dev->hw_enc_features; |
3347c960 | 1247 | SKB_GSO_CB(skb)->encap_level += ihl; |
73136267 | 1248 | |
badff6d0 | 1249 | skb_reset_transport_header(skb); |
47d27aad | 1250 | |
f4c50d99 HX |
1251 | segs = ERR_PTR(-EPROTONOSUPPORT); |
1252 | ||
cbc53e08 AD |
1253 | if (!skb->encapsulation || encap) { |
1254 | udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); | |
1255 | fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); | |
1256 | ||
1257 | /* fixed ID is invalid if DF bit is not set */ | |
1258 | if (fixedid && !(iph->frag_off & htons(IP_DF))) | |
1259 | goto out; | |
1260 | } | |
dcd60771 | 1261 | |
bca49f84 | 1262 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 VY |
1263 | if (likely(ops && ops->callbacks.gso_segment)) |
1264 | segs = ops->callbacks.gso_segment(skb, features); | |
f4c50d99 | 1265 | |
50c3a487 | 1266 | if (IS_ERR_OR_NULL(segs)) |
f4c50d99 HX |
1267 | goto out; |
1268 | ||
1269 | skb = segs; | |
1270 | do { | |
3347c960 | 1271 | iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); |
8c3a897b | 1272 | if (udpfrag) { |
d7ca4cc0 | 1273 | iph->frag_off = htons(offset >> 3); |
00db4124 | 1274 | if (skb->next) |
d7ca4cc0 | 1275 | iph->frag_off |= htons(IP_MF); |
3347c960 | 1276 | offset += skb->len - nhoff - ihl; |
802ab55a AD |
1277 | tot_len = skb->len - nhoff; |
1278 | } else if (skb_is_gso(skb)) { | |
1279 | if (!fixedid) { | |
1280 | iph->id = htons(id); | |
1281 | id += skb_shinfo(skb)->gso_segs; | |
1282 | } | |
1283 | tot_len = skb_shinfo(skb)->gso_size + | |
1284 | SKB_GSO_CB(skb)->data_offset + | |
1285 | skb->head - (unsigned char *)iph; | |
1286 | } else { | |
1287 | if (!fixedid) | |
1288 | iph->id = htons(id++); | |
1289 | tot_len = skb->len - nhoff; | |
490ab081 | 1290 | } |
802ab55a | 1291 | iph->tot_len = htons(tot_len); |
47d27aad | 1292 | ip_send_check(iph); |
8c3a897b | 1293 | if (encap) |
3347c960 | 1294 | skb_reset_inner_headers(skb); |
3347c960 | 1295 | skb->network_header = (u8 *)iph - skb->head; |
f4c50d99 HX |
1296 | } while ((skb = skb->next)); |
1297 | ||
1298 | out: | |
1299 | return segs; | |
1300 | } | |
1301 | ||
73cc19f1 HX |
1302 | static struct sk_buff **inet_gro_receive(struct sk_buff **head, |
1303 | struct sk_buff *skb) | |
1304 | { | |
bca49f84 | 1305 | const struct net_offload *ops; |
73cc19f1 HX |
1306 | struct sk_buff **pp = NULL; |
1307 | struct sk_buff *p; | |
b71d1d42 | 1308 | const struct iphdr *iph; |
a5b1cf28 HX |
1309 | unsigned int hlen; |
1310 | unsigned int off; | |
1075f3f6 | 1311 | unsigned int id; |
73cc19f1 HX |
1312 | int flush = 1; |
1313 | int proto; | |
73cc19f1 | 1314 | |
a5b1cf28 HX |
1315 | off = skb_gro_offset(skb); |
1316 | hlen = off + sizeof(*iph); | |
1317 | iph = skb_gro_header_fast(skb, off); | |
1318 | if (skb_gro_header_hard(skb, hlen)) { | |
1319 | iph = skb_gro_header_slow(skb, hlen, off); | |
1320 | if (unlikely(!iph)) | |
1321 | goto out; | |
1322 | } | |
73cc19f1 | 1323 | |
f9242b6b | 1324 | proto = iph->protocol; |
73cc19f1 HX |
1325 | |
1326 | rcu_read_lock(); | |
bca49f84 | 1327 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1328 | if (!ops || !ops->callbacks.gro_receive) |
73cc19f1 HX |
1329 | goto out_unlock; |
1330 | ||
a5ad24be | 1331 | if (*(u8 *)iph != 0x45) |
73cc19f1 HX |
1332 | goto out_unlock; |
1333 | ||
a9e050f4 | 1334 | if (unlikely(ip_fast_csum((u8 *)iph, 5))) |
73cc19f1 HX |
1335 | goto out_unlock; |
1336 | ||
0eae88f3 | 1337 | id = ntohl(*(__be32 *)&iph->id); |
db8caf3d | 1338 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); |
1075f3f6 | 1339 | id >>= 16; |
73cc19f1 HX |
1340 | |
1341 | for (p = *head; p; p = p->next) { | |
1342 | struct iphdr *iph2; | |
1530545e | 1343 | u16 flush_id; |
73cc19f1 HX |
1344 | |
1345 | if (!NAPI_GRO_CB(p)->same_flow) | |
1346 | continue; | |
1347 | ||
299603e8 JC |
1348 | iph2 = (struct iphdr *)(p->data + off); |
1349 | /* The above works because, with the exception of the top | |
1350 | * (inner most) layer, we only aggregate pkts with the same | |
1351 | * hdr length so all the hdrs we'll need to verify will start | |
1352 | * at the same offset. | |
1353 | */ | |
a5ad24be | 1354 | if ((iph->protocol ^ iph2->protocol) | |
0eae88f3 ED |
1355 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1356 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1357 | NAPI_GRO_CB(p)->same_flow = 0; |
1358 | continue; | |
1359 | } | |
1360 | ||
1361 | /* All fields must match except length and checksum. */ | |
1362 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be | 1363 | (iph->ttl ^ iph2->ttl) | |
a9e050f4 | 1364 | (iph->tos ^ iph2->tos) | |
bf5a755f | 1365 | ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); |
73cc19f1 HX |
1366 | |
1367 | NAPI_GRO_CB(p)->flush |= flush; | |
1530545e AD |
1368 | |
1369 | /* We need to store of the IP ID check to be included later | |
1370 | * when we can verify that this packet does in fact belong | |
1371 | * to a given flow. | |
1372 | */ | |
1373 | flush_id = (u16)(id - ntohs(iph2->id)); | |
1374 | ||
1375 | /* This bit of code makes it much easier for us to identify | |
1376 | * the cases where we are doing atomic vs non-atomic IP ID | |
1377 | * checks. Specifically an atomic check can return IP ID | |
1378 | * values 0 - 0xFFFF, while a non-atomic check can only | |
1379 | * return 0 or 0xFFFF. | |
1380 | */ | |
1381 | if (!NAPI_GRO_CB(p)->is_atomic || | |
1382 | !(iph->frag_off & htons(IP_DF))) { | |
1383 | flush_id ^= NAPI_GRO_CB(p)->count; | |
1384 | flush_id = flush_id ? 0xFFFF : 0; | |
1385 | } | |
1386 | ||
1387 | /* If the previous IP ID value was based on an atomic | |
1388 | * datagram we can overwrite the value and ignore it. | |
1389 | */ | |
1390 | if (NAPI_GRO_CB(skb)->is_atomic) | |
1391 | NAPI_GRO_CB(p)->flush_id = flush_id; | |
1392 | else | |
1393 | NAPI_GRO_CB(p)->flush_id |= flush_id; | |
73cc19f1 HX |
1394 | } |
1395 | ||
1530545e | 1396 | NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); |
73cc19f1 | 1397 | NAPI_GRO_CB(skb)->flush |= flush; |
299603e8 JC |
1398 | skb_set_network_header(skb, off); |
1399 | /* The above will be needed by the transport layer if there is one | |
1400 | * immediately following this IP hdr. | |
1401 | */ | |
1402 | ||
2c804d0f ED |
1403 | /* Note : No need to call skb_gro_postpull_rcsum() here, |
1404 | * as we already checked checksum over ipv4 header was 0 | |
1405 | */ | |
86911732 HX |
1406 | skb_gro_pull(skb, sizeof(*iph)); |
1407 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 | 1408 | |
f191a1d1 | 1409 | pp = ops->callbacks.gro_receive(head, skb); |
73cc19f1 HX |
1410 | |
1411 | out_unlock: | |
1412 | rcu_read_unlock(); | |
1413 | ||
1414 | out: | |
1415 | NAPI_GRO_CB(skb)->flush |= flush; | |
1416 | ||
1417 | return pp; | |
1418 | } | |
1419 | ||
fac8e0f5 JG |
1420 | static struct sk_buff **ipip_gro_receive(struct sk_buff **head, |
1421 | struct sk_buff *skb) | |
1422 | { | |
1423 | if (NAPI_GRO_CB(skb)->encap_mark) { | |
1424 | NAPI_GRO_CB(skb)->flush = 1; | |
1425 | return NULL; | |
1426 | } | |
1427 | ||
1428 | NAPI_GRO_CB(skb)->encap_mark = 1; | |
1429 | ||
1430 | return inet_gro_receive(head, skb); | |
1431 | } | |
1432 | ||
822c8685 DD |
1433 | #define SECONDS_PER_DAY 86400 |
1434 | ||
1435 | /* inet_current_timestamp - Return IP network timestamp | |
1436 | * | |
1437 | * Return milliseconds since midnight in network byte order. | |
1438 | */ | |
1439 | __be32 inet_current_timestamp(void) | |
1440 | { | |
1441 | u32 secs; | |
1442 | u32 msecs; | |
1443 | struct timespec64 ts; | |
1444 | ||
1445 | ktime_get_real_ts64(&ts); | |
1446 | ||
1447 | /* Get secs since midnight. */ | |
1448 | (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); | |
1449 | /* Convert to msecs. */ | |
1450 | msecs = secs * MSEC_PER_SEC; | |
1451 | /* Convert nsec to msec. */ | |
1452 | msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; | |
1453 | ||
1454 | /* Convert to network byte order. */ | |
3ba9d300 | 1455 | return htonl(msecs); |
822c8685 DD |
1456 | } |
1457 | EXPORT_SYMBOL(inet_current_timestamp); | |
1458 | ||
f4713a3d WB |
1459 | int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) |
1460 | { | |
1461 | if (sk->sk_family == AF_INET) | |
1462 | return ip_recv_error(sk, msg, len, addr_len); | |
1463 | #if IS_ENABLED(CONFIG_IPV6) | |
1464 | if (sk->sk_family == AF_INET6) | |
1465 | return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); | |
1466 | #endif | |
1467 | return -EINVAL; | |
1468 | } | |
1469 | ||
299603e8 | 1470 | static int inet_gro_complete(struct sk_buff *skb, int nhoff) |
73cc19f1 | 1471 | { |
299603e8 JC |
1472 | __be16 newlen = htons(skb->len - nhoff); |
1473 | struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); | |
bca49f84 | 1474 | const struct net_offload *ops; |
f9242b6b | 1475 | int proto = iph->protocol; |
73cc19f1 | 1476 | int err = -ENOSYS; |
73cc19f1 | 1477 | |
c3caf119 JC |
1478 | if (skb->encapsulation) |
1479 | skb_set_inner_network_header(skb, nhoff); | |
1480 | ||
73cc19f1 HX |
1481 | csum_replace2(&iph->check, iph->tot_len, newlen); |
1482 | iph->tot_len = newlen; | |
1483 | ||
1484 | rcu_read_lock(); | |
bca49f84 | 1485 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1486 | if (WARN_ON(!ops || !ops->callbacks.gro_complete)) |
73cc19f1 HX |
1487 | goto out_unlock; |
1488 | ||
299603e8 JC |
1489 | /* Only need to add sizeof(*iph) to get to the next hdr below |
1490 | * because any hdr with option will have been flushed in | |
1491 | * inet_gro_receive(). | |
1492 | */ | |
1493 | err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph)); | |
73cc19f1 HX |
1494 | |
1495 | out_unlock: | |
1496 | rcu_read_unlock(); | |
1497 | ||
1498 | return err; | |
1499 | } | |
1500 | ||
b8cba75b JG |
1501 | static int ipip_gro_complete(struct sk_buff *skb, int nhoff) |
1502 | { | |
1503 | skb->encapsulation = 1; | |
1504 | skb_shinfo(skb)->gso_type |= SKB_GSO_IPIP; | |
1505 | return inet_gro_complete(skb, nhoff); | |
1506 | } | |
1507 | ||
eee4fe4d | 1508 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1509 | unsigned short type, unsigned char protocol, |
1510 | struct net *net) | |
3d58b5fa | 1511 | { |
eee4fe4d | 1512 | struct socket *sock; |
26abe143 | 1513 | int rc = sock_create_kern(net, family, type, protocol, &sock); |
3d58b5fa DL |
1514 | |
1515 | if (rc == 0) { | |
eee4fe4d DL |
1516 | *sk = sock->sk; |
1517 | (*sk)->sk_allocation = GFP_ATOMIC; | |
3d58b5fa DL |
1518 | /* |
1519 | * Unhash it so that IP input processing does not even see it, | |
1520 | * we do not wish this socket to see incoming packets. | |
1521 | */ | |
eee4fe4d | 1522 | (*sk)->sk_prot->unhash(*sk); |
3d58b5fa DL |
1523 | } |
1524 | return rc; | |
1525 | } | |
3d58b5fa DL |
1526 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1527 | ||
c4c6bc31 R |
1528 | u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt) |
1529 | { | |
1530 | return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt); | |
1531 | } | |
1532 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field); | |
1533 | ||
698365fa | 1534 | unsigned long snmp_fold_field(void __percpu *mib, int offt) |
5e0f0435 HX |
1535 | { |
1536 | unsigned long res = 0; | |
698365fa | 1537 | int i; |
5e0f0435 | 1538 | |
698365fa | 1539 | for_each_possible_cpu(i) |
c4c6bc31 | 1540 | res += snmp_get_cpu_field(mib, i, offt); |
5e0f0435 HX |
1541 | return res; |
1542 | } | |
1543 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1544 | ||
4ce3c183 ED |
1545 | #if BITS_PER_LONG==32 |
1546 | ||
80ec1927 | 1547 | u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, |
c4c6bc31 R |
1548 | size_t syncp_offset) |
1549 | { | |
1550 | void *bhptr; | |
1551 | struct u64_stats_sync *syncp; | |
1552 | u64 v; | |
1553 | unsigned int start; | |
1554 | ||
1555 | bhptr = per_cpu_ptr(mib, cpu); | |
1556 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); | |
1557 | do { | |
1558 | start = u64_stats_fetch_begin_irq(syncp); | |
1559 | v = *(((u64 *)bhptr) + offt); | |
1560 | } while (u64_stats_fetch_retry_irq(syncp, start)); | |
1561 | ||
1562 | return v; | |
1563 | } | |
1564 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); | |
1565 | ||
698365fa | 1566 | u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) |
4ce3c183 ED |
1567 | { |
1568 | u64 res = 0; | |
1569 | int cpu; | |
1570 | ||
1571 | for_each_possible_cpu(cpu) { | |
d1bfc625 | 1572 | res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); |
4ce3c183 ED |
1573 | } |
1574 | return res; | |
1575 | } | |
1576 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1577 | #endif | |
1578 | ||
1da177e4 | 1579 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1580 | static const struct net_protocol igmp_protocol = { |
1da177e4 | 1581 | .handler = igmp_rcv, |
b4ee07df | 1582 | .netns_ok = 1, |
1da177e4 LT |
1583 | }; |
1584 | #endif | |
1585 | ||
32613090 | 1586 | static const struct net_protocol tcp_protocol = { |
41063e9d DM |
1587 | .early_demux = tcp_v4_early_demux, |
1588 | .handler = tcp_v4_rcv, | |
1589 | .err_handler = tcp_v4_err, | |
41063e9d DM |
1590 | .no_policy = 1, |
1591 | .netns_ok = 1, | |
8ed1dc44 | 1592 | .icmp_strict_tag_validation = 1, |
1da177e4 LT |
1593 | }; |
1594 | ||
32613090 | 1595 | static const struct net_protocol udp_protocol = { |
421b3885 | 1596 | .early_demux = udp_v4_early_demux, |
1da177e4 LT |
1597 | .handler = udp_rcv, |
1598 | .err_handler = udp_err, | |
1599 | .no_policy = 1, | |
92f1fecb | 1600 | .netns_ok = 1, |
1da177e4 LT |
1601 | }; |
1602 | ||
32613090 | 1603 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1604 | .handler = icmp_rcv, |
5b052042 | 1605 | .err_handler = icmp_err, |
8b7817f3 | 1606 | .no_policy = 1, |
92f1fecb | 1607 | .netns_ok = 1, |
1da177e4 LT |
1608 | }; |
1609 | ||
9b4661bd PE |
1610 | static __net_init int ipv4_mib_init_net(struct net *net) |
1611 | { | |
827da44c JS |
1612 | int i; |
1613 | ||
698365fa WC |
1614 | net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); |
1615 | if (!net->mib.tcp_statistics) | |
57ef42d5 | 1616 | goto err_tcp_mib; |
698365fa WC |
1617 | net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); |
1618 | if (!net->mib.ip_statistics) | |
a20f5799 | 1619 | goto err_ip_mib; |
827da44c JS |
1620 | |
1621 | for_each_possible_cpu(i) { | |
1622 | struct ipstats_mib *af_inet_stats; | |
698365fa | 1623 | af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); |
827da44c | 1624 | u64_stats_init(&af_inet_stats->syncp); |
827da44c JS |
1625 | } |
1626 | ||
698365fa WC |
1627 | net->mib.net_statistics = alloc_percpu(struct linux_mib); |
1628 | if (!net->mib.net_statistics) | |
61a7e260 | 1629 | goto err_net_mib; |
698365fa WC |
1630 | net->mib.udp_statistics = alloc_percpu(struct udp_mib); |
1631 | if (!net->mib.udp_statistics) | |
2f275f91 | 1632 | goto err_udp_mib; |
698365fa WC |
1633 | net->mib.udplite_statistics = alloc_percpu(struct udp_mib); |
1634 | if (!net->mib.udplite_statistics) | |
386019d3 | 1635 | goto err_udplite_mib; |
698365fa WC |
1636 | net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); |
1637 | if (!net->mib.icmp_statistics) | |
b60538a0 | 1638 | goto err_icmp_mib; |
acb32ba3 ED |
1639 | net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), |
1640 | GFP_KERNEL); | |
1641 | if (!net->mib.icmpmsg_statistics) | |
923c6586 | 1642 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1643 | |
1644 | tcp_mib_init(net); | |
9b4661bd | 1645 | return 0; |
57ef42d5 | 1646 | |
923c6586 | 1647 | err_icmpmsg_mib: |
698365fa | 1648 | free_percpu(net->mib.icmp_statistics); |
b60538a0 | 1649 | err_icmp_mib: |
698365fa | 1650 | free_percpu(net->mib.udplite_statistics); |
386019d3 | 1651 | err_udplite_mib: |
698365fa | 1652 | free_percpu(net->mib.udp_statistics); |
2f275f91 | 1653 | err_udp_mib: |
698365fa | 1654 | free_percpu(net->mib.net_statistics); |
61a7e260 | 1655 | err_net_mib: |
698365fa | 1656 | free_percpu(net->mib.ip_statistics); |
a20f5799 | 1657 | err_ip_mib: |
698365fa | 1658 | free_percpu(net->mib.tcp_statistics); |
57ef42d5 PE |
1659 | err_tcp_mib: |
1660 | return -ENOMEM; | |
9b4661bd PE |
1661 | } |
1662 | ||
1663 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1664 | { | |
acb32ba3 | 1665 | kfree(net->mib.icmpmsg_statistics); |
698365fa WC |
1666 | free_percpu(net->mib.icmp_statistics); |
1667 | free_percpu(net->mib.udplite_statistics); | |
1668 | free_percpu(net->mib.udp_statistics); | |
1669 | free_percpu(net->mib.net_statistics); | |
1670 | free_percpu(net->mib.ip_statistics); | |
1671 | free_percpu(net->mib.tcp_statistics); | |
9b4661bd PE |
1672 | } |
1673 | ||
1674 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1675 | .init = ipv4_mib_init_net, | |
1676 | .exit = ipv4_mib_exit_net, | |
1677 | }; | |
1678 | ||
1da177e4 LT |
1679 | static int __init init_ipv4_mibs(void) |
1680 | { | |
d89cbbb1 | 1681 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1682 | } |
1683 | ||
c9d8f1a6 CW |
1684 | static __net_init int inet_init_net(struct net *net) |
1685 | { | |
1686 | /* | |
1687 | * Set defaults for local port range | |
1688 | */ | |
1689 | seqlock_init(&net->ipv4.ip_local_ports.lock); | |
1690 | net->ipv4.ip_local_ports.range[0] = 32768; | |
07f4c900 | 1691 | net->ipv4.ip_local_ports.range[1] = 60999; |
ba6b918a CW |
1692 | |
1693 | seqlock_init(&net->ipv4.ping_group_range.lock); | |
1694 | /* | |
1695 | * Sane defaults - nobody may create ping sockets. | |
1696 | * Boot scripts should set this to distro-specific group. | |
1697 | */ | |
1698 | net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); | |
1699 | net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); | |
c9d8f1a6 CW |
1700 | return 0; |
1701 | } | |
1702 | ||
1703 | static __net_exit void inet_exit_net(struct net *net) | |
1704 | { | |
1705 | } | |
1706 | ||
1707 | static __net_initdata struct pernet_operations af_inet_ops = { | |
1708 | .init = inet_init_net, | |
1709 | .exit = inet_exit_net, | |
1710 | }; | |
1711 | ||
1712 | static int __init init_inet_pernet_ops(void) | |
1713 | { | |
1714 | return register_pernet_subsys(&af_inet_ops); | |
1715 | } | |
1716 | ||
1da177e4 | 1717 | static int ipv4_proc_init(void); |
1da177e4 | 1718 | |
30e224d7 HX |
1719 | /* |
1720 | * IP protocol layer initialiser | |
1721 | */ | |
1722 | ||
22061d80 VY |
1723 | static struct packet_offload ip_packet_offload __read_mostly = { |
1724 | .type = cpu_to_be16(ETH_P_IP), | |
f191a1d1 | 1725 | .callbacks = { |
f191a1d1 VY |
1726 | .gso_segment = inet_gso_segment, |
1727 | .gro_receive = inet_gro_receive, | |
1728 | .gro_complete = inet_gro_complete, | |
1729 | }, | |
30e224d7 HX |
1730 | }; |
1731 | ||
cb32f511 ED |
1732 | static const struct net_offload ipip_offload = { |
1733 | .callbacks = { | |
cb32f511 | 1734 | .gso_segment = inet_gso_segment, |
fac8e0f5 | 1735 | .gro_receive = ipip_gro_receive, |
b8cba75b | 1736 | .gro_complete = ipip_gro_complete, |
cb32f511 ED |
1737 | }, |
1738 | }; | |
1739 | ||
808a8f88 VY |
1740 | static int __init ipv4_offload_init(void) |
1741 | { | |
1742 | /* | |
1743 | * Add offloads | |
1744 | */ | |
da5bab07 | 1745 | if (udpv4_offload_init() < 0) |
808a8f88 | 1746 | pr_crit("%s: Cannot add UDP protocol offload\n", __func__); |
28850dc7 DB |
1747 | if (tcpv4_offload_init() < 0) |
1748 | pr_crit("%s: Cannot add TCP protocol offload\n", __func__); | |
808a8f88 VY |
1749 | |
1750 | dev_add_offload(&ip_packet_offload); | |
cb32f511 | 1751 | inet_add_offload(&ipip_offload, IPPROTO_IPIP); |
808a8f88 VY |
1752 | return 0; |
1753 | } | |
1754 | ||
1755 | fs_initcall(ipv4_offload_init); | |
1756 | ||
1757 | static struct packet_type ip_packet_type __read_mostly = { | |
1758 | .type = cpu_to_be16(ETH_P_IP), | |
1759 | .func = ip_rcv, | |
1760 | }; | |
1761 | ||
1da177e4 LT |
1762 | static int __init inet_init(void) |
1763 | { | |
1da177e4 LT |
1764 | struct inet_protosw *q; |
1765 | struct list_head *r; | |
1766 | int rc = -EINVAL; | |
1767 | ||
b4772ef8 | 1768 | sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); |
1da177e4 LT |
1769 | |
1770 | rc = proto_register(&tcp_prot, 1); | |
1771 | if (rc) | |
122ff243 | 1772 | goto out; |
1da177e4 LT |
1773 | |
1774 | rc = proto_register(&udp_prot, 1); | |
1775 | if (rc) | |
1776 | goto out_unregister_tcp_proto; | |
1777 | ||
1778 | rc = proto_register(&raw_prot, 1); | |
1779 | if (rc) | |
1780 | goto out_unregister_udp_proto; | |
1781 | ||
c319b4d7 VK |
1782 | rc = proto_register(&ping_prot, 1); |
1783 | if (rc) | |
1784 | goto out_unregister_raw_proto; | |
1785 | ||
1da177e4 | 1786 | /* |
e905a9ed | 1787 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1788 | */ |
1789 | ||
e905a9ed | 1790 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1791 | |
bd7b1533 AV |
1792 | #ifdef CONFIG_SYSCTL |
1793 | ip_static_sysctl_init(); | |
1794 | #endif | |
1795 | ||
1da177e4 LT |
1796 | /* |
1797 | * Add all the base protocols. | |
1798 | */ | |
1799 | ||
1800 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
058bd4d2 | 1801 | pr_crit("%s: Cannot add ICMP protocol\n", __func__); |
1da177e4 | 1802 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) |
058bd4d2 | 1803 | pr_crit("%s: Cannot add UDP protocol\n", __func__); |
1da177e4 | 1804 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) |
058bd4d2 | 1805 | pr_crit("%s: Cannot add TCP protocol\n", __func__); |
1da177e4 LT |
1806 | #ifdef CONFIG_IP_MULTICAST |
1807 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
058bd4d2 | 1808 | pr_crit("%s: Cannot add IGMP protocol\n", __func__); |
1da177e4 LT |
1809 | #endif |
1810 | ||
1811 | /* Register the socket-side information for inet_create. */ | |
1812 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1813 | INIT_LIST_HEAD(r); | |
1814 | ||
1815 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1816 | inet_register_protosw(q); | |
1817 | ||
1818 | /* | |
1819 | * Set the ARP module up | |
1820 | */ | |
1821 | ||
1822 | arp_init(); | |
1823 | ||
e905a9ed YH |
1824 | /* |
1825 | * Set the IP module up | |
1826 | */ | |
1da177e4 LT |
1827 | |
1828 | ip_init(); | |
1829 | ||
9b0f976f | 1830 | tcp_v4_init(); |
1da177e4 LT |
1831 | |
1832 | /* Setup TCP slab cache for open requests. */ | |
1833 | tcp_init(); | |
1834 | ||
95766fff HA |
1835 | /* Setup UDP memory threshold */ |
1836 | udp_init(); | |
1837 | ||
ba4e58ec GR |
1838 | /* Add UDP-Lite (RFC 3828) */ |
1839 | udplite4_register(); | |
1da177e4 | 1840 | |
c319b4d7 VK |
1841 | ping_init(); |
1842 | ||
1da177e4 LT |
1843 | /* |
1844 | * Set the ICMP layer up | |
1845 | */ | |
1846 | ||
a5710d65 DL |
1847 | if (icmp_init() < 0) |
1848 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
1849 | |
1850 | /* | |
1851 | * Initialise the multicast router | |
1852 | */ | |
1853 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 | 1854 | if (ip_mr_init()) |
058bd4d2 | 1855 | pr_crit("%s: Cannot init ipv4 mroute\n", __func__); |
1da177e4 | 1856 | #endif |
c9d8f1a6 CW |
1857 | |
1858 | if (init_inet_pernet_ops()) | |
1859 | pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); | |
1da177e4 LT |
1860 | /* |
1861 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 1862 | */ |
1da177e4 | 1863 | |
132adf54 | 1864 | if (init_ipv4_mibs()) |
058bd4d2 | 1865 | pr_crit("%s: Cannot init ipv4 mibs\n", __func__); |
e905a9ed | 1866 | |
1da177e4 LT |
1867 | ipv4_proc_init(); |
1868 | ||
1869 | ipfrag_init(); | |
1870 | ||
30e224d7 HX |
1871 | dev_add_pack(&ip_packet_type); |
1872 | ||
045a0fa0 TG |
1873 | ip_tunnel_core_init(); |
1874 | ||
1da177e4 LT |
1875 | rc = 0; |
1876 | out: | |
1877 | return rc; | |
c319b4d7 VK |
1878 | out_unregister_raw_proto: |
1879 | proto_unregister(&raw_prot); | |
1da177e4 LT |
1880 | out_unregister_udp_proto: |
1881 | proto_unregister(&udp_prot); | |
321efff7 OK |
1882 | out_unregister_tcp_proto: |
1883 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
1884 | goto out; |
1885 | } | |
1886 | ||
a536e077 | 1887 | fs_initcall(inet_init); |
1da177e4 LT |
1888 | |
1889 | /* ------------------------------------------------------------------------ */ | |
1890 | ||
1891 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
1892 | static int __init ipv4_proc_init(void) |
1893 | { | |
1894 | int rc = 0; | |
1895 | ||
1896 | if (raw_proc_init()) | |
1897 | goto out_raw; | |
1898 | if (tcp4_proc_init()) | |
1899 | goto out_tcp; | |
1900 | if (udp4_proc_init()) | |
1901 | goto out_udp; | |
c319b4d7 VK |
1902 | if (ping_proc_init()) |
1903 | goto out_ping; | |
1da177e4 LT |
1904 | if (ip_misc_proc_init()) |
1905 | goto out_misc; | |
1906 | out: | |
1907 | return rc; | |
1908 | out_misc: | |
c319b4d7 VK |
1909 | ping_proc_exit(); |
1910 | out_ping: | |
1da177e4 LT |
1911 | udp4_proc_exit(); |
1912 | out_udp: | |
1913 | tcp4_proc_exit(); | |
1914 | out_tcp: | |
1915 | raw_proc_exit(); | |
1916 | out_raw: | |
1917 | rc = -ENOMEM; | |
1918 | goto out; | |
1919 | } | |
1920 | ||
1921 | #else /* CONFIG_PROC_FS */ | |
1922 | static int __init ipv4_proc_init(void) | |
1923 | { | |
1924 | return 0; | |
1925 | } | |
1926 | #endif /* CONFIG_PROC_FS */ | |
1927 | ||
1928 | MODULE_ALIAS_NETPROTO(PF_INET); | |
1929 |