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 | ||
f4c50d99 | 68 | #include <linux/err.h> |
1da177e4 LT |
69 | #include <linux/errno.h> |
70 | #include <linux/types.h> | |
71 | #include <linux/socket.h> | |
72 | #include <linux/in.h> | |
73 | #include <linux/kernel.h> | |
1da177e4 LT |
74 | #include <linux/module.h> |
75 | #include <linux/sched.h> | |
76 | #include <linux/timer.h> | |
77 | #include <linux/string.h> | |
78 | #include <linux/sockios.h> | |
79 | #include <linux/net.h> | |
4fc268d2 | 80 | #include <linux/capability.h> |
1da177e4 LT |
81 | #include <linux/fcntl.h> |
82 | #include <linux/mm.h> | |
83 | #include <linux/interrupt.h> | |
84 | #include <linux/stat.h> | |
85 | #include <linux/init.h> | |
86 | #include <linux/poll.h> | |
87 | #include <linux/netfilter_ipv4.h> | |
b3da2cf3 | 88 | #include <linux/random.h> |
5a0e3ad6 | 89 | #include <linux/slab.h> |
1da177e4 LT |
90 | |
91 | #include <asm/uaccess.h> | |
92 | #include <asm/system.h> | |
93 | ||
1da177e4 LT |
94 | #include <linux/inet.h> |
95 | #include <linux/igmp.h> | |
14c85021 | 96 | #include <linux/inetdevice.h> |
1da177e4 | 97 | #include <linux/netdevice.h> |
73cc19f1 | 98 | #include <net/checksum.h> |
1da177e4 LT |
99 | #include <net/ip.h> |
100 | #include <net/protocol.h> | |
101 | #include <net/arp.h> | |
102 | #include <net/route.h> | |
103 | #include <net/ip_fib.h> | |
295f7324 | 104 | #include <net/inet_connection_sock.h> |
1da177e4 LT |
105 | #include <net/tcp.h> |
106 | #include <net/udp.h> | |
ba4e58ec | 107 | #include <net/udplite.h> |
1da177e4 LT |
108 | #include <linux/skbuff.h> |
109 | #include <net/sock.h> | |
110 | #include <net/raw.h> | |
111 | #include <net/icmp.h> | |
112 | #include <net/ipip.h> | |
113 | #include <net/inet_common.h> | |
114 | #include <net/xfrm.h> | |
9b4661bd | 115 | #include <net/net_namespace.h> |
1da177e4 LT |
116 | #ifdef CONFIG_IP_MROUTE |
117 | #include <linux/mroute.h> | |
118 | #endif | |
119 | ||
1da177e4 LT |
120 | |
121 | /* The inetsw table contains everything that inet_create needs to | |
122 | * build a new socket. | |
123 | */ | |
124 | static struct list_head inetsw[SOCK_MAX]; | |
125 | static DEFINE_SPINLOCK(inetsw_lock); | |
126 | ||
9ba63979 | 127 | struct ipv4_config ipv4_config; |
9ba63979 PE |
128 | EXPORT_SYMBOL(ipv4_config); |
129 | ||
1da177e4 LT |
130 | /* New destruction routine */ |
131 | ||
132 | void inet_sock_destruct(struct sock *sk) | |
133 | { | |
134 | struct inet_sock *inet = inet_sk(sk); | |
135 | ||
136 | __skb_queue_purge(&sk->sk_receive_queue); | |
137 | __skb_queue_purge(&sk->sk_error_queue); | |
138 | ||
95766fff HA |
139 | sk_mem_reclaim(sk); |
140 | ||
1da177e4 | 141 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 142 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
143 | sk->sk_state, sk); |
144 | return; | |
145 | } | |
146 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 147 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
148 | return; |
149 | } | |
150 | ||
547b792c IJ |
151 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
152 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
153 | WARN_ON(sk->sk_wmem_queued); | |
154 | WARN_ON(sk->sk_forward_alloc); | |
1da177e4 | 155 | |
f6d8bd05 | 156 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
b6c6712a | 157 | dst_release(rcu_dereference_check(sk->sk_dst_cache, 1)); |
e6848976 | 158 | sk_refcnt_debug_dec(sk); |
1da177e4 | 159 | } |
3d1427f8 | 160 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
161 | |
162 | /* | |
163 | * The routines beyond this point handle the behaviour of an AF_INET | |
164 | * socket object. Mostly it punts to the subprotocols of IP to do | |
165 | * the work. | |
166 | */ | |
167 | ||
168 | /* | |
169 | * Automatically bind an unbound socket. | |
170 | */ | |
171 | ||
172 | static int inet_autobind(struct sock *sk) | |
173 | { | |
174 | struct inet_sock *inet; | |
175 | /* We may need to bind the socket. */ | |
176 | lock_sock(sk); | |
177 | inet = inet_sk(sk); | |
c720c7e8 | 178 | if (!inet->inet_num) { |
1da177e4 LT |
179 | if (sk->sk_prot->get_port(sk, 0)) { |
180 | release_sock(sk); | |
181 | return -EAGAIN; | |
182 | } | |
c720c7e8 | 183 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
184 | } |
185 | release_sock(sk); | |
186 | return 0; | |
187 | } | |
188 | ||
189 | /* | |
190 | * Move a socket into listening state. | |
191 | */ | |
192 | int inet_listen(struct socket *sock, int backlog) | |
193 | { | |
194 | struct sock *sk = sock->sk; | |
195 | unsigned char old_state; | |
196 | int err; | |
197 | ||
198 | lock_sock(sk); | |
199 | ||
200 | err = -EINVAL; | |
201 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
202 | goto out; | |
203 | ||
204 | old_state = sk->sk_state; | |
205 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) | |
206 | goto out; | |
207 | ||
208 | /* Really, if the socket is already in listen state | |
209 | * we can only allow the backlog to be adjusted. | |
210 | */ | |
211 | if (old_state != TCP_LISTEN) { | |
72a3effa | 212 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
213 | if (err) |
214 | goto out; | |
215 | } | |
216 | sk->sk_max_ack_backlog = backlog; | |
217 | err = 0; | |
218 | ||
219 | out: | |
220 | release_sock(sk); | |
221 | return err; | |
222 | } | |
3d1427f8 | 223 | EXPORT_SYMBOL(inet_listen); |
1da177e4 | 224 | |
be776281 | 225 | u32 inet_ehash_secret __read_mostly; |
b3da2cf3 DM |
226 | EXPORT_SYMBOL(inet_ehash_secret); |
227 | ||
be776281 ED |
228 | /* |
229 | * inet_ehash_secret must be set exactly once | |
be776281 | 230 | */ |
b3da2cf3 DM |
231 | void build_ehash_secret(void) |
232 | { | |
be776281 | 233 | u32 rnd; |
49e8ab03 | 234 | |
be776281 ED |
235 | do { |
236 | get_random_bytes(&rnd, sizeof(rnd)); | |
237 | } while (rnd == 0); | |
49e8ab03 ED |
238 | |
239 | cmpxchg(&inet_ehash_secret, 0, rnd); | |
b3da2cf3 DM |
240 | } |
241 | EXPORT_SYMBOL(build_ehash_secret); | |
242 | ||
2342fd7e DL |
243 | static inline int inet_netns_ok(struct net *net, int protocol) |
244 | { | |
245 | int hash; | |
32613090 | 246 | const struct net_protocol *ipprot; |
2342fd7e | 247 | |
04f258ce | 248 | if (net_eq(net, &init_net)) |
2342fd7e DL |
249 | return 1; |
250 | ||
251 | hash = protocol & (MAX_INET_PROTOS - 1); | |
252 | ipprot = rcu_dereference(inet_protos[hash]); | |
253 | ||
254 | if (ipprot == NULL) | |
255 | /* raw IP is OK */ | |
256 | return 1; | |
257 | return ipprot->netns_ok; | |
258 | } | |
259 | ||
1da177e4 LT |
260 | /* |
261 | * Create an inet socket. | |
262 | */ | |
263 | ||
3f378b68 EP |
264 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
265 | int kern) | |
1da177e4 LT |
266 | { |
267 | struct sock *sk; | |
1da177e4 LT |
268 | struct inet_protosw *answer; |
269 | struct inet_sock *inet; | |
270 | struct proto *answer_prot; | |
271 | unsigned char answer_flags; | |
272 | char answer_no_check; | |
bb97d31f | 273 | int try_loading_module = 0; |
86c8f9d1 | 274 | int err; |
1da177e4 | 275 | |
04f258ce ED |
276 | if (unlikely(!inet_ehash_secret)) |
277 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
278 | build_ehash_secret(); | |
b3da2cf3 | 279 | |
1da177e4 LT |
280 | sock->state = SS_UNCONNECTED; |
281 | ||
282 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 283 | lookup_protocol: |
86c8f9d1 | 284 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 285 | rcu_read_lock(); |
696adfe8 | 286 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 287 | |
696adfe8 | 288 | err = 0; |
1da177e4 LT |
289 | /* Check the non-wild match. */ |
290 | if (protocol == answer->protocol) { | |
291 | if (protocol != IPPROTO_IP) | |
292 | break; | |
293 | } else { | |
294 | /* Check for the two wild cases. */ | |
295 | if (IPPROTO_IP == protocol) { | |
296 | protocol = answer->protocol; | |
297 | break; | |
298 | } | |
299 | if (IPPROTO_IP == answer->protocol) | |
300 | break; | |
301 | } | |
86c8f9d1 | 302 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
303 | } |
304 | ||
696adfe8 | 305 | if (unlikely(err)) { |
bb97d31f ACM |
306 | if (try_loading_module < 2) { |
307 | rcu_read_unlock(); | |
308 | /* | |
309 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
310 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
311 | */ | |
312 | if (++try_loading_module == 1) | |
313 | request_module("net-pf-%d-proto-%d-type-%d", | |
314 | PF_INET, protocol, sock->type); | |
315 | /* | |
316 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
317 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
318 | */ | |
319 | else | |
320 | request_module("net-pf-%d-proto-%d", | |
321 | PF_INET, protocol); | |
322 | goto lookup_protocol; | |
323 | } else | |
324 | goto out_rcu_unlock; | |
325 | } | |
326 | ||
1da177e4 | 327 | err = -EPERM; |
c84b3268 | 328 | if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW)) |
1da177e4 | 329 | goto out_rcu_unlock; |
1da177e4 | 330 | |
2342fd7e DL |
331 | err = -EAFNOSUPPORT; |
332 | if (!inet_netns_ok(net, protocol)) | |
333 | goto out_rcu_unlock; | |
334 | ||
1da177e4 LT |
335 | sock->ops = answer->ops; |
336 | answer_prot = answer->prot; | |
337 | answer_no_check = answer->no_check; | |
338 | answer_flags = answer->flags; | |
339 | rcu_read_unlock(); | |
340 | ||
547b792c | 341 | WARN_ON(answer_prot->slab == NULL); |
1da177e4 LT |
342 | |
343 | err = -ENOBUFS; | |
6257ff21 | 344 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot); |
1da177e4 LT |
345 | if (sk == NULL) |
346 | goto out; | |
347 | ||
348 | err = 0; | |
349 | sk->sk_no_check = answer_no_check; | |
350 | if (INET_PROTOSW_REUSE & answer_flags) | |
351 | sk->sk_reuse = 1; | |
352 | ||
353 | inet = inet_sk(sk); | |
469de9b9 | 354 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 | 355 | |
7b2ff18e JO |
356 | inet->nodefrag = 0; |
357 | ||
1da177e4 | 358 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 359 | inet->inet_num = protocol; |
1da177e4 LT |
360 | if (IPPROTO_RAW == protocol) |
361 | inet->hdrincl = 1; | |
362 | } | |
363 | ||
364 | if (ipv4_config.no_pmtu_disc) | |
365 | inet->pmtudisc = IP_PMTUDISC_DONT; | |
366 | else | |
367 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
368 | ||
c720c7e8 | 369 | inet->inet_id = 0; |
1da177e4 LT |
370 | |
371 | sock_init_data(sock, sk); | |
372 | ||
373 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
374 | sk->sk_protocol = protocol; |
375 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
376 | ||
377 | inet->uc_ttl = -1; | |
378 | inet->mc_loop = 1; | |
379 | inet->mc_ttl = 1; | |
f771bef9 | 380 | inet->mc_all = 1; |
1da177e4 LT |
381 | inet->mc_index = 0; |
382 | inet->mc_list = NULL; | |
383 | ||
e6848976 | 384 | sk_refcnt_debug_inc(sk); |
1da177e4 | 385 | |
c720c7e8 | 386 | if (inet->inet_num) { |
1da177e4 LT |
387 | /* It assumes that any protocol which allows |
388 | * the user to assign a number at socket | |
389 | * creation time automatically | |
390 | * shares. | |
391 | */ | |
c720c7e8 | 392 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
393 | /* Add to protocol hash chains. */ |
394 | sk->sk_prot->hash(sk); | |
395 | } | |
396 | ||
397 | if (sk->sk_prot->init) { | |
398 | err = sk->sk_prot->init(sk); | |
399 | if (err) | |
400 | sk_common_release(sk); | |
401 | } | |
402 | out: | |
403 | return err; | |
404 | out_rcu_unlock: | |
405 | rcu_read_unlock(); | |
406 | goto out; | |
407 | } | |
408 | ||
409 | ||
410 | /* | |
411 | * The peer socket should always be NULL (or else). When we call this | |
412 | * function we are destroying the object and from then on nobody | |
413 | * should refer to it. | |
414 | */ | |
415 | int inet_release(struct socket *sock) | |
416 | { | |
417 | struct sock *sk = sock->sk; | |
418 | ||
419 | if (sk) { | |
420 | long timeout; | |
421 | ||
c58dc01b | 422 | sock_rps_reset_flow(sk); |
fec5e652 | 423 | |
1da177e4 LT |
424 | /* Applications forget to leave groups before exiting */ |
425 | ip_mc_drop_socket(sk); | |
426 | ||
427 | /* If linger is set, we don't return until the close | |
428 | * is complete. Otherwise we return immediately. The | |
429 | * actually closing is done the same either way. | |
430 | * | |
431 | * If the close is due to the process exiting, we never | |
432 | * linger.. | |
433 | */ | |
434 | timeout = 0; | |
435 | if (sock_flag(sk, SOCK_LINGER) && | |
436 | !(current->flags & PF_EXITING)) | |
437 | timeout = sk->sk_lingertime; | |
438 | sock->sk = NULL; | |
439 | sk->sk_prot->close(sk, timeout); | |
440 | } | |
441 | return 0; | |
442 | } | |
3d1427f8 | 443 | EXPORT_SYMBOL(inet_release); |
1da177e4 LT |
444 | |
445 | /* It is off by default, see below. */ | |
ab32ea5d | 446 | int sysctl_ip_nonlocal_bind __read_mostly; |
3d1427f8 | 447 | EXPORT_SYMBOL(sysctl_ip_nonlocal_bind); |
1da177e4 LT |
448 | |
449 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
450 | { | |
451 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
452 | struct sock *sk = sock->sk; | |
453 | struct inet_sock *inet = inet_sk(sk); | |
454 | unsigned short snum; | |
455 | int chk_addr_ret; | |
456 | int err; | |
457 | ||
458 | /* If the socket has its own bind function then use it. (RAW) */ | |
459 | if (sk->sk_prot->bind) { | |
460 | err = sk->sk_prot->bind(sk, uaddr, addr_len); | |
461 | goto out; | |
462 | } | |
463 | err = -EINVAL; | |
464 | if (addr_len < sizeof(struct sockaddr_in)) | |
465 | goto out; | |
466 | ||
3b1e0a65 | 467 | chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr); |
1da177e4 LT |
468 | |
469 | /* Not specified by any standard per-se, however it breaks too | |
470 | * many applications when removed. It is unfortunate since | |
471 | * allowing applications to make a non-local bind solves | |
472 | * several problems with systems using dynamic addressing. | |
473 | * (ie. your servers still start up even if your ISDN link | |
474 | * is temporarily down) | |
475 | */ | |
476 | err = -EADDRNOTAVAIL; | |
477 | if (!sysctl_ip_nonlocal_bind && | |
b9fb1506 | 478 | !(inet->freebind || inet->transparent) && |
e6f1cebf | 479 | addr->sin_addr.s_addr != htonl(INADDR_ANY) && |
1da177e4 LT |
480 | chk_addr_ret != RTN_LOCAL && |
481 | chk_addr_ret != RTN_MULTICAST && | |
482 | chk_addr_ret != RTN_BROADCAST) | |
483 | goto out; | |
484 | ||
485 | snum = ntohs(addr->sin_port); | |
486 | err = -EACCES; | |
487 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) | |
488 | goto out; | |
489 | ||
490 | /* We keep a pair of addresses. rcv_saddr is the one | |
491 | * used by hash lookups, and saddr is used for transmit. | |
492 | * | |
493 | * In the BSD API these are the same except where it | |
494 | * would be illegal to use them (multicast/broadcast) in | |
495 | * which case the sending device address is used. | |
496 | */ | |
497 | lock_sock(sk); | |
498 | ||
499 | /* Check these errors (active socket, double bind). */ | |
500 | err = -EINVAL; | |
c720c7e8 | 501 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
502 | goto out_release_sock; |
503 | ||
c720c7e8 | 504 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 505 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 506 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
507 | |
508 | /* Make sure we are allowed to bind here. */ | |
509 | if (sk->sk_prot->get_port(sk, snum)) { | |
c720c7e8 | 510 | inet->inet_saddr = inet->inet_rcv_saddr = 0; |
1da177e4 LT |
511 | err = -EADDRINUSE; |
512 | goto out_release_sock; | |
513 | } | |
514 | ||
c720c7e8 | 515 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
516 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
517 | if (snum) | |
518 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
519 | inet->inet_sport = htons(inet->inet_num); |
520 | inet->inet_daddr = 0; | |
521 | inet->inet_dport = 0; | |
1da177e4 LT |
522 | sk_dst_reset(sk); |
523 | err = 0; | |
524 | out_release_sock: | |
525 | release_sock(sk); | |
526 | out: | |
527 | return err; | |
528 | } | |
3d1427f8 | 529 | EXPORT_SYMBOL(inet_bind); |
1da177e4 LT |
530 | |
531 | int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr, | |
532 | int addr_len, int flags) | |
533 | { | |
534 | struct sock *sk = sock->sk; | |
535 | ||
6503d961 CG |
536 | if (addr_len < sizeof(uaddr->sa_family)) |
537 | return -EINVAL; | |
1da177e4 LT |
538 | if (uaddr->sa_family == AF_UNSPEC) |
539 | return sk->sk_prot->disconnect(sk, flags); | |
540 | ||
c720c7e8 | 541 | if (!inet_sk(sk)->inet_num && inet_autobind(sk)) |
1da177e4 LT |
542 | return -EAGAIN; |
543 | return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len); | |
544 | } | |
3d1427f8 | 545 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 LT |
546 | |
547 | static long inet_wait_for_connect(struct sock *sk, long timeo) | |
548 | { | |
549 | DEFINE_WAIT(wait); | |
550 | ||
aa395145 | 551 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 LT |
552 | |
553 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
554 | * change state of the socket from TCP_SYN_*. | |
555 | * Connect() does not allow to get error notifications | |
556 | * without closing the socket. | |
557 | */ | |
558 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
559 | release_sock(sk); | |
560 | timeo = schedule_timeout(timeo); | |
561 | lock_sock(sk); | |
562 | if (signal_pending(current) || !timeo) | |
563 | break; | |
aa395145 | 564 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 565 | } |
aa395145 | 566 | finish_wait(sk_sleep(sk), &wait); |
1da177e4 LT |
567 | return timeo; |
568 | } | |
569 | ||
570 | /* | |
571 | * Connect to a remote host. There is regrettably still a little | |
572 | * TCP 'magic' in here. | |
573 | */ | |
574 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
575 | int addr_len, int flags) | |
576 | { | |
577 | struct sock *sk = sock->sk; | |
578 | int err; | |
579 | long timeo; | |
580 | ||
6503d961 CG |
581 | if (addr_len < sizeof(uaddr->sa_family)) |
582 | return -EINVAL; | |
583 | ||
1da177e4 LT |
584 | lock_sock(sk); |
585 | ||
586 | if (uaddr->sa_family == AF_UNSPEC) { | |
587 | err = sk->sk_prot->disconnect(sk, flags); | |
588 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
589 | goto out; | |
590 | } | |
591 | ||
592 | switch (sock->state) { | |
593 | default: | |
594 | err = -EINVAL; | |
595 | goto out; | |
596 | case SS_CONNECTED: | |
597 | err = -EISCONN; | |
598 | goto out; | |
599 | case SS_CONNECTING: | |
600 | err = -EALREADY; | |
601 | /* Fall out of switch with err, set for this state */ | |
602 | break; | |
603 | case SS_UNCONNECTED: | |
604 | err = -EISCONN; | |
605 | if (sk->sk_state != TCP_CLOSE) | |
606 | goto out; | |
607 | ||
608 | err = sk->sk_prot->connect(sk, uaddr, addr_len); | |
609 | if (err < 0) | |
610 | goto out; | |
611 | ||
e905a9ed | 612 | sock->state = SS_CONNECTING; |
1da177e4 LT |
613 | |
614 | /* Just entered SS_CONNECTING state; the only | |
615 | * difference is that return value in non-blocking | |
616 | * case is EINPROGRESS, rather than EALREADY. | |
617 | */ | |
618 | err = -EINPROGRESS; | |
619 | break; | |
620 | } | |
621 | ||
622 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
623 | ||
624 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
625 | /* Error code is set above */ | |
626 | if (!timeo || !inet_wait_for_connect(sk, timeo)) | |
627 | goto out; | |
628 | ||
629 | err = sock_intr_errno(timeo); | |
630 | if (signal_pending(current)) | |
631 | goto out; | |
632 | } | |
633 | ||
634 | /* Connection was closed by RST, timeout, ICMP error | |
635 | * or another process disconnected us. | |
636 | */ | |
637 | if (sk->sk_state == TCP_CLOSE) | |
638 | goto sock_error; | |
639 | ||
640 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
641 | * and error was received after socket entered established state. | |
642 | * Hence, it is handled normally after connect() return successfully. | |
643 | */ | |
644 | ||
645 | sock->state = SS_CONNECTED; | |
646 | err = 0; | |
647 | out: | |
648 | release_sock(sk); | |
649 | return err; | |
650 | ||
651 | sock_error: | |
652 | err = sock_error(sk) ? : -ECONNABORTED; | |
653 | sock->state = SS_UNCONNECTED; | |
654 | if (sk->sk_prot->disconnect(sk, flags)) | |
655 | sock->state = SS_DISCONNECTING; | |
656 | goto out; | |
657 | } | |
3d1427f8 | 658 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 LT |
659 | |
660 | /* | |
661 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
662 | */ | |
663 | ||
664 | int inet_accept(struct socket *sock, struct socket *newsock, int flags) | |
665 | { | |
666 | struct sock *sk1 = sock->sk; | |
667 | int err = -EINVAL; | |
668 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err); | |
669 | ||
670 | if (!sk2) | |
671 | goto do_err; | |
672 | ||
673 | lock_sock(sk2); | |
674 | ||
547b792c IJ |
675 | WARN_ON(!((1 << sk2->sk_state) & |
676 | (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
1da177e4 LT |
677 | |
678 | sock_graft(sk2, newsock); | |
679 | ||
680 | newsock->state = SS_CONNECTED; | |
681 | err = 0; | |
682 | release_sock(sk2); | |
683 | do_err: | |
684 | return err; | |
685 | } | |
3d1427f8 | 686 | EXPORT_SYMBOL(inet_accept); |
1da177e4 LT |
687 | |
688 | ||
689 | /* | |
690 | * This does both peername and sockname. | |
691 | */ | |
692 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
693 | int *uaddr_len, int peer) | |
694 | { | |
695 | struct sock *sk = sock->sk; | |
696 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 697 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
698 | |
699 | sin->sin_family = AF_INET; | |
700 | if (peer) { | |
c720c7e8 | 701 | if (!inet->inet_dport || |
1da177e4 LT |
702 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
703 | peer == 1)) | |
704 | return -ENOTCONN; | |
c720c7e8 ED |
705 | sin->sin_port = inet->inet_dport; |
706 | sin->sin_addr.s_addr = inet->inet_daddr; | |
1da177e4 | 707 | } else { |
c720c7e8 | 708 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 709 | if (!addr) |
c720c7e8 ED |
710 | addr = inet->inet_saddr; |
711 | sin->sin_port = inet->inet_sport; | |
1da177e4 LT |
712 | sin->sin_addr.s_addr = addr; |
713 | } | |
714 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
715 | *uaddr_len = sizeof(*sin); | |
716 | return 0; | |
717 | } | |
3d1427f8 | 718 | EXPORT_SYMBOL(inet_getname); |
1da177e4 LT |
719 | |
720 | int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg, | |
721 | size_t size) | |
722 | { | |
723 | struct sock *sk = sock->sk; | |
724 | ||
c58dc01b | 725 | sock_rps_record_flow(sk); |
fec5e652 | 726 | |
1da177e4 | 727 | /* We may need to bind the socket. */ |
7ba42910 CG |
728 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
729 | inet_autobind(sk)) | |
1da177e4 LT |
730 | return -EAGAIN; |
731 | ||
732 | return sk->sk_prot->sendmsg(iocb, sk, msg, size); | |
733 | } | |
3d1427f8 | 734 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 735 | |
7ba42910 CG |
736 | ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, |
737 | size_t size, int flags) | |
1da177e4 LT |
738 | { |
739 | struct sock *sk = sock->sk; | |
740 | ||
c58dc01b | 741 | sock_rps_record_flow(sk); |
fec5e652 | 742 | |
1da177e4 | 743 | /* We may need to bind the socket. */ |
7ba42910 CG |
744 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
745 | inet_autobind(sk)) | |
1da177e4 LT |
746 | return -EAGAIN; |
747 | ||
748 | if (sk->sk_prot->sendpage) | |
749 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
750 | return sock_no_sendpage(sock, page, offset, size, flags); | |
751 | } | |
7ba42910 | 752 | EXPORT_SYMBOL(inet_sendpage); |
1da177e4 | 753 | |
fec5e652 TH |
754 | int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg, |
755 | size_t size, int flags) | |
756 | { | |
757 | struct sock *sk = sock->sk; | |
758 | int addr_len = 0; | |
759 | int err; | |
760 | ||
c58dc01b | 761 | sock_rps_record_flow(sk); |
fec5e652 TH |
762 | |
763 | err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT, | |
764 | flags & ~MSG_DONTWAIT, &addr_len); | |
765 | if (err >= 0) | |
766 | msg->msg_namelen = addr_len; | |
767 | return err; | |
768 | } | |
769 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
770 | |
771 | int inet_shutdown(struct socket *sock, int how) | |
772 | { | |
773 | struct sock *sk = sock->sk; | |
774 | int err = 0; | |
775 | ||
776 | /* This should really check to make sure | |
777 | * the socket is a TCP socket. (WHY AC...) | |
778 | */ | |
779 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
780 | 1->2 bit 2 snds. | |
781 | 2->3 */ | |
782 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
783 | return -EINVAL; | |
784 | ||
785 | lock_sock(sk); | |
786 | if (sock->state == SS_CONNECTING) { | |
787 | if ((1 << sk->sk_state) & | |
788 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
789 | sock->state = SS_DISCONNECTING; | |
790 | else | |
791 | sock->state = SS_CONNECTED; | |
792 | } | |
793 | ||
794 | switch (sk->sk_state) { | |
795 | case TCP_CLOSE: | |
796 | err = -ENOTCONN; | |
797 | /* Hack to wake up other listeners, who can poll for | |
798 | POLLHUP, even on eg. unconnected UDP sockets -- RR */ | |
799 | default: | |
800 | sk->sk_shutdown |= how; | |
801 | if (sk->sk_prot->shutdown) | |
802 | sk->sk_prot->shutdown(sk, how); | |
803 | break; | |
804 | ||
805 | /* Remaining two branches are temporary solution for missing | |
806 | * close() in multithreaded environment. It is _not_ a good idea, | |
807 | * but we have no choice until close() is repaired at VFS level. | |
808 | */ | |
809 | case TCP_LISTEN: | |
810 | if (!(how & RCV_SHUTDOWN)) | |
811 | break; | |
812 | /* Fall through */ | |
813 | case TCP_SYN_SENT: | |
814 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
815 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
816 | break; | |
817 | } | |
818 | ||
819 | /* Wake up anyone sleeping in poll. */ | |
820 | sk->sk_state_change(sk); | |
821 | release_sock(sk); | |
822 | return err; | |
823 | } | |
3d1427f8 | 824 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
825 | |
826 | /* | |
827 | * ioctl() calls you can issue on an INET socket. Most of these are | |
828 | * device configuration and stuff and very rarely used. Some ioctls | |
829 | * pass on to the socket itself. | |
830 | * | |
831 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
832 | * loads the devconfigure module does its configuring and unloads it. | |
833 | * There's a good 20K of config code hanging around the kernel. | |
834 | */ | |
835 | ||
836 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
837 | { | |
838 | struct sock *sk = sock->sk; | |
839 | int err = 0; | |
3b1e0a65 | 840 | struct net *net = sock_net(sk); |
1da177e4 LT |
841 | |
842 | switch (cmd) { | |
3d1427f8 ED |
843 | case SIOCGSTAMP: |
844 | err = sock_get_timestamp(sk, (struct timeval __user *)arg); | |
845 | break; | |
846 | case SIOCGSTAMPNS: | |
847 | err = sock_get_timestampns(sk, (struct timespec __user *)arg); | |
848 | break; | |
849 | case SIOCADDRT: | |
850 | case SIOCDELRT: | |
851 | case SIOCRTMSG: | |
852 | err = ip_rt_ioctl(net, cmd, (void __user *)arg); | |
853 | break; | |
854 | case SIOCDARP: | |
855 | case SIOCGARP: | |
856 | case SIOCSARP: | |
857 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
858 | break; | |
859 | case SIOCGIFADDR: | |
860 | case SIOCSIFADDR: | |
861 | case SIOCGIFBRDADDR: | |
862 | case SIOCSIFBRDADDR: | |
863 | case SIOCGIFNETMASK: | |
864 | case SIOCSIFNETMASK: | |
865 | case SIOCGIFDSTADDR: | |
866 | case SIOCSIFDSTADDR: | |
867 | case SIOCSIFPFLAGS: | |
868 | case SIOCGIFPFLAGS: | |
869 | case SIOCSIFFLAGS: | |
870 | err = devinet_ioctl(net, cmd, (void __user *)arg); | |
871 | break; | |
872 | default: | |
873 | if (sk->sk_prot->ioctl) | |
874 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
875 | else | |
876 | err = -ENOIOCTLCMD; | |
877 | break; | |
1da177e4 LT |
878 | } |
879 | return err; | |
880 | } | |
3d1427f8 | 881 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 882 | |
709b46e8 EB |
883 | #ifdef CONFIG_COMPAT |
884 | int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
885 | { | |
886 | struct sock *sk = sock->sk; | |
887 | int err = -ENOIOCTLCMD; | |
888 | ||
889 | if (sk->sk_prot->compat_ioctl) | |
890 | err = sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
891 | ||
892 | return err; | |
893 | } | |
894 | #endif | |
895 | ||
90ddc4f0 | 896 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
897 | .family = PF_INET, |
898 | .owner = THIS_MODULE, | |
899 | .release = inet_release, | |
900 | .bind = inet_bind, | |
901 | .connect = inet_stream_connect, | |
902 | .socketpair = sock_no_socketpair, | |
903 | .accept = inet_accept, | |
904 | .getname = inet_getname, | |
905 | .poll = tcp_poll, | |
906 | .ioctl = inet_ioctl, | |
907 | .listen = inet_listen, | |
908 | .shutdown = inet_shutdown, | |
909 | .setsockopt = sock_common_setsockopt, | |
910 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 911 | .sendmsg = inet_sendmsg, |
fec5e652 | 912 | .recvmsg = inet_recvmsg, |
543d9cfe | 913 | .mmap = sock_no_mmap, |
7ba42910 | 914 | .sendpage = inet_sendpage, |
9c55e01c | 915 | .splice_read = tcp_splice_read, |
3fdadf7d | 916 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
917 | .compat_setsockopt = compat_sock_common_setsockopt, |
918 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 919 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 920 | #endif |
1da177e4 | 921 | }; |
3d1427f8 | 922 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 923 | |
90ddc4f0 | 924 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
925 | .family = PF_INET, |
926 | .owner = THIS_MODULE, | |
927 | .release = inet_release, | |
928 | .bind = inet_bind, | |
929 | .connect = inet_dgram_connect, | |
930 | .socketpair = sock_no_socketpair, | |
931 | .accept = sock_no_accept, | |
932 | .getname = inet_getname, | |
933 | .poll = udp_poll, | |
934 | .ioctl = inet_ioctl, | |
935 | .listen = sock_no_listen, | |
936 | .shutdown = inet_shutdown, | |
937 | .setsockopt = sock_common_setsockopt, | |
938 | .getsockopt = sock_common_getsockopt, | |
939 | .sendmsg = inet_sendmsg, | |
fec5e652 | 940 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
941 | .mmap = sock_no_mmap, |
942 | .sendpage = inet_sendpage, | |
3fdadf7d | 943 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
944 | .compat_setsockopt = compat_sock_common_setsockopt, |
945 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 946 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 947 | #endif |
1da177e4 | 948 | }; |
3d1427f8 | 949 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
950 | |
951 | /* | |
952 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
953 | * udp_poll | |
954 | */ | |
90ddc4f0 | 955 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
956 | .family = PF_INET, |
957 | .owner = THIS_MODULE, | |
958 | .release = inet_release, | |
959 | .bind = inet_bind, | |
960 | .connect = inet_dgram_connect, | |
961 | .socketpair = sock_no_socketpair, | |
962 | .accept = sock_no_accept, | |
963 | .getname = inet_getname, | |
964 | .poll = datagram_poll, | |
965 | .ioctl = inet_ioctl, | |
966 | .listen = sock_no_listen, | |
967 | .shutdown = inet_shutdown, | |
968 | .setsockopt = sock_common_setsockopt, | |
969 | .getsockopt = sock_common_getsockopt, | |
970 | .sendmsg = inet_sendmsg, | |
fec5e652 | 971 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
972 | .mmap = sock_no_mmap, |
973 | .sendpage = inet_sendpage, | |
3fdadf7d | 974 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
975 | .compat_setsockopt = compat_sock_common_setsockopt, |
976 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 977 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 978 | #endif |
1da177e4 LT |
979 | }; |
980 | ||
ec1b4cf7 | 981 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
982 | .family = PF_INET, |
983 | .create = inet_create, | |
984 | .owner = THIS_MODULE, | |
985 | }; | |
986 | ||
1da177e4 LT |
987 | /* Upon startup we insert all the elements in inetsw_array[] into |
988 | * the linked list inetsw. | |
989 | */ | |
990 | static struct inet_protosw inetsw_array[] = | |
991 | { | |
e905a9ed YH |
992 | { |
993 | .type = SOCK_STREAM, | |
994 | .protocol = IPPROTO_TCP, | |
995 | .prot = &tcp_prot, | |
996 | .ops = &inet_stream_ops, | |
e905a9ed YH |
997 | .no_check = 0, |
998 | .flags = INET_PROTOSW_PERMANENT | | |
d83d8461 | 999 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1000 | }, |
1001 | ||
1002 | { | |
1003 | .type = SOCK_DGRAM, | |
1004 | .protocol = IPPROTO_UDP, | |
1005 | .prot = &udp_prot, | |
1006 | .ops = &inet_dgram_ops, | |
e905a9ed YH |
1007 | .no_check = UDP_CSUM_DEFAULT, |
1008 | .flags = INET_PROTOSW_PERMANENT, | |
1da177e4 | 1009 | }, |
e905a9ed | 1010 | |
1da177e4 LT |
1011 | |
1012 | { | |
e905a9ed YH |
1013 | .type = SOCK_RAW, |
1014 | .protocol = IPPROTO_IP, /* wild card */ | |
1015 | .prot = &raw_prot, | |
1016 | .ops = &inet_sockraw_ops, | |
e905a9ed YH |
1017 | .no_check = UDP_CSUM_DEFAULT, |
1018 | .flags = INET_PROTOSW_REUSE, | |
1da177e4 LT |
1019 | } |
1020 | }; | |
1021 | ||
c40f6fff | 1022 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1023 | |
1024 | void inet_register_protosw(struct inet_protosw *p) | |
1025 | { | |
1026 | struct list_head *lh; | |
1027 | struct inet_protosw *answer; | |
1028 | int protocol = p->protocol; | |
1029 | struct list_head *last_perm; | |
1030 | ||
1031 | spin_lock_bh(&inetsw_lock); | |
1032 | ||
1033 | if (p->type >= SOCK_MAX) | |
1034 | goto out_illegal; | |
1035 | ||
1036 | /* If we are trying to override a permanent protocol, bail. */ | |
1037 | answer = NULL; | |
1038 | last_perm = &inetsw[p->type]; | |
1039 | list_for_each(lh, &inetsw[p->type]) { | |
1040 | answer = list_entry(lh, struct inet_protosw, list); | |
1041 | ||
1042 | /* Check only the non-wild match. */ | |
1043 | if (INET_PROTOSW_PERMANENT & answer->flags) { | |
1044 | if (protocol == answer->protocol) | |
1045 | break; | |
1046 | last_perm = lh; | |
1047 | } | |
1048 | ||
1049 | answer = NULL; | |
1050 | } | |
1051 | if (answer) | |
1052 | goto out_permanent; | |
1053 | ||
1054 | /* Add the new entry after the last permanent entry if any, so that | |
1055 | * the new entry does not override a permanent entry when matched with | |
1056 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1057 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1058 | * system automatically returns to the old behavior. |
1059 | */ | |
1060 | list_add_rcu(&p->list, last_perm); | |
1061 | out: | |
1062 | spin_unlock_bh(&inetsw_lock); | |
1063 | ||
1da177e4 LT |
1064 | return; |
1065 | ||
1066 | out_permanent: | |
1067 | printk(KERN_ERR "Attempt to override permanent protocol %d.\n", | |
1068 | protocol); | |
1069 | goto out; | |
1070 | ||
1071 | out_illegal: | |
1072 | printk(KERN_ERR | |
1073 | "Ignoring attempt to register invalid socket type %d.\n", | |
1074 | p->type); | |
1075 | goto out; | |
1076 | } | |
3d1427f8 | 1077 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1078 | |
1079 | void inet_unregister_protosw(struct inet_protosw *p) | |
1080 | { | |
1081 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
1082 | printk(KERN_ERR | |
1083 | "Attempt to unregister permanent protocol %d.\n", | |
1084 | p->protocol); | |
1085 | } else { | |
1086 | spin_lock_bh(&inetsw_lock); | |
1087 | list_del_rcu(&p->list); | |
1088 | spin_unlock_bh(&inetsw_lock); | |
1089 | ||
1090 | synchronize_net(); | |
1091 | } | |
1092 | } | |
3d1427f8 | 1093 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1094 | |
32519f11 ACM |
1095 | /* |
1096 | * Shall we try to damage output packets if routing dev changes? | |
1097 | */ | |
1098 | ||
ab32ea5d | 1099 | int sysctl_ip_dynaddr __read_mostly; |
32519f11 ACM |
1100 | |
1101 | static int inet_sk_reselect_saddr(struct sock *sk) | |
1102 | { | |
1103 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1104 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1105 | __be32 daddr = inet->inet_daddr; |
2d7192d6 | 1106 | struct flowi4 fl4; |
b23dd4fe DM |
1107 | struct rtable *rt; |
1108 | __be32 new_saddr; | |
f6d8bd05 | 1109 | struct ip_options_rcu *inet_opt; |
32519f11 | 1110 | |
f6d8bd05 ED |
1111 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1112 | sock_owned_by_user(sk)); | |
1113 | if (inet_opt && inet_opt->opt.srr) | |
1114 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1115 | |
1116 | /* Query new route. */ | |
2d7192d6 | 1117 | rt = ip_route_connect(&fl4, daddr, 0, RT_CONN_FLAGS(sk), |
b23dd4fe DM |
1118 | sk->sk_bound_dev_if, sk->sk_protocol, |
1119 | inet->inet_sport, inet->inet_dport, sk, false); | |
1120 | if (IS_ERR(rt)) | |
1121 | return PTR_ERR(rt); | |
32519f11 | 1122 | |
d8d1f30b | 1123 | sk_setup_caps(sk, &rt->dst); |
32519f11 | 1124 | |
b8831877 | 1125 | new_saddr = fl4.saddr; |
32519f11 ACM |
1126 | |
1127 | if (new_saddr == old_saddr) | |
1128 | return 0; | |
1129 | ||
1130 | if (sysctl_ip_dynaddr > 1) { | |
673d57e7 HH |
1131 | printk(KERN_INFO "%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1132 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1133 | } |
1134 | ||
c720c7e8 | 1135 | inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; |
32519f11 ACM |
1136 | |
1137 | /* | |
1138 | * XXX The only one ugly spot where we need to | |
1139 | * XXX really change the sockets identity after | |
1140 | * XXX it has entered the hashes. -DaveM | |
1141 | * | |
1142 | * Besides that, it does not check for connection | |
1143 | * uniqueness. Wait for troubles. | |
1144 | */ | |
1145 | __sk_prot_rehash(sk); | |
1146 | return 0; | |
1147 | } | |
1148 | ||
1149 | int inet_sk_rebuild_header(struct sock *sk) | |
1150 | { | |
1151 | struct inet_sock *inet = inet_sk(sk); | |
1152 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1153 | __be32 daddr; |
f6d8bd05 | 1154 | struct ip_options_rcu *inet_opt; |
32519f11 ACM |
1155 | int err; |
1156 | ||
1157 | /* Route is OK, nothing to do. */ | |
1158 | if (rt) | |
1159 | return 0; | |
1160 | ||
1161 | /* Reroute. */ | |
f6d8bd05 ED |
1162 | rcu_read_lock(); |
1163 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1164 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1165 | if (inet_opt && inet_opt->opt.srr) |
1166 | daddr = inet_opt->opt.faddr; | |
1167 | rcu_read_unlock(); | |
78fbfd8a DM |
1168 | rt = ip_route_output_ports(sock_net(sk), sk, daddr, inet->inet_saddr, |
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 | */ | |
1184 | if (!sysctl_ip_dynaddr || | |
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 | ||
a430a43d HX |
1195 | static int inet_gso_send_check(struct sk_buff *skb) |
1196 | { | |
b71d1d42 | 1197 | const struct iphdr *iph; |
32613090 | 1198 | const struct net_protocol *ops; |
a430a43d HX |
1199 | int proto; |
1200 | int ihl; | |
1201 | int err = -EINVAL; | |
1202 | ||
1203 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) | |
1204 | goto out; | |
1205 | ||
eddc9ec5 | 1206 | iph = ip_hdr(skb); |
a430a43d HX |
1207 | ihl = iph->ihl * 4; |
1208 | if (ihl < sizeof(*iph)) | |
1209 | goto out; | |
1210 | ||
1211 | if (unlikely(!pskb_may_pull(skb, ihl))) | |
1212 | goto out; | |
1213 | ||
badff6d0 ACM |
1214 | __skb_pull(skb, ihl); |
1215 | skb_reset_transport_header(skb); | |
eddc9ec5 | 1216 | iph = ip_hdr(skb); |
a430a43d HX |
1217 | proto = iph->protocol & (MAX_INET_PROTOS - 1); |
1218 | err = -EPROTONOSUPPORT; | |
1219 | ||
1220 | rcu_read_lock(); | |
1221 | ops = rcu_dereference(inet_protos[proto]); | |
1222 | if (likely(ops && ops->gso_send_check)) | |
1223 | err = ops->gso_send_check(skb); | |
1224 | rcu_read_unlock(); | |
1225 | ||
1226 | out: | |
1227 | return err; | |
1228 | } | |
1229 | ||
04ed3e74 | 1230 | static struct sk_buff *inet_gso_segment(struct sk_buff *skb, u32 features) |
f4c50d99 HX |
1231 | { |
1232 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
1233 | struct iphdr *iph; | |
32613090 | 1234 | const struct net_protocol *ops; |
f4c50d99 HX |
1235 | int proto; |
1236 | int ihl; | |
1237 | int id; | |
d7ca4cc0 | 1238 | unsigned int offset = 0; |
f4c50d99 | 1239 | |
d212f87b SH |
1240 | if (!(features & NETIF_F_V4_CSUM)) |
1241 | features &= ~NETIF_F_SG; | |
1242 | ||
bbcf467d HX |
1243 | if (unlikely(skb_shinfo(skb)->gso_type & |
1244 | ~(SKB_GSO_TCPV4 | | |
1245 | SKB_GSO_UDP | | |
1246 | SKB_GSO_DODGY | | |
1247 | SKB_GSO_TCP_ECN | | |
1248 | 0))) | |
1249 | goto out; | |
1250 | ||
1251 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) | |
f4c50d99 HX |
1252 | goto out; |
1253 | ||
eddc9ec5 | 1254 | iph = ip_hdr(skb); |
f4c50d99 HX |
1255 | ihl = iph->ihl * 4; |
1256 | if (ihl < sizeof(*iph)) | |
1257 | goto out; | |
1258 | ||
bbcf467d | 1259 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 HX |
1260 | goto out; |
1261 | ||
badff6d0 ACM |
1262 | __skb_pull(skb, ihl); |
1263 | skb_reset_transport_header(skb); | |
eddc9ec5 | 1264 | iph = ip_hdr(skb); |
f4c50d99 HX |
1265 | id = ntohs(iph->id); |
1266 | proto = iph->protocol & (MAX_INET_PROTOS - 1); | |
1267 | segs = ERR_PTR(-EPROTONOSUPPORT); | |
1268 | ||
1269 | rcu_read_lock(); | |
1270 | ops = rcu_dereference(inet_protos[proto]); | |
bbcf467d | 1271 | if (likely(ops && ops->gso_segment)) |
576a30eb | 1272 | segs = ops->gso_segment(skb, features); |
f4c50d99 HX |
1273 | rcu_read_unlock(); |
1274 | ||
801678c5 | 1275 | if (!segs || IS_ERR(segs)) |
f4c50d99 HX |
1276 | goto out; |
1277 | ||
1278 | skb = segs; | |
1279 | do { | |
eddc9ec5 | 1280 | iph = ip_hdr(skb); |
d7ca4cc0 SS |
1281 | if (proto == IPPROTO_UDP) { |
1282 | iph->id = htons(id); | |
1283 | iph->frag_off = htons(offset >> 3); | |
1284 | if (skb->next != NULL) | |
1285 | iph->frag_off |= htons(IP_MF); | |
1286 | offset += (skb->len - skb->mac_len - iph->ihl * 4); | |
1287 | } else | |
1288 | iph->id = htons(id++); | |
f4c50d99 HX |
1289 | iph->tot_len = htons(skb->len - skb->mac_len); |
1290 | iph->check = 0; | |
d56f90a7 | 1291 | iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl); |
f4c50d99 HX |
1292 | } while ((skb = skb->next)); |
1293 | ||
1294 | out: | |
1295 | return segs; | |
1296 | } | |
1297 | ||
73cc19f1 HX |
1298 | static struct sk_buff **inet_gro_receive(struct sk_buff **head, |
1299 | struct sk_buff *skb) | |
1300 | { | |
32613090 | 1301 | const struct net_protocol *ops; |
73cc19f1 HX |
1302 | struct sk_buff **pp = NULL; |
1303 | struct sk_buff *p; | |
b71d1d42 | 1304 | const struct iphdr *iph; |
a5b1cf28 HX |
1305 | unsigned int hlen; |
1306 | unsigned int off; | |
1075f3f6 | 1307 | unsigned int id; |
73cc19f1 HX |
1308 | int flush = 1; |
1309 | int proto; | |
73cc19f1 | 1310 | |
a5b1cf28 HX |
1311 | off = skb_gro_offset(skb); |
1312 | hlen = off + sizeof(*iph); | |
1313 | iph = skb_gro_header_fast(skb, off); | |
1314 | if (skb_gro_header_hard(skb, hlen)) { | |
1315 | iph = skb_gro_header_slow(skb, hlen, off); | |
1316 | if (unlikely(!iph)) | |
1317 | goto out; | |
1318 | } | |
73cc19f1 | 1319 | |
73cc19f1 HX |
1320 | proto = iph->protocol & (MAX_INET_PROTOS - 1); |
1321 | ||
1322 | rcu_read_lock(); | |
1323 | ops = rcu_dereference(inet_protos[proto]); | |
1324 | if (!ops || !ops->gro_receive) | |
1325 | goto out_unlock; | |
1326 | ||
a5ad24be | 1327 | if (*(u8 *)iph != 0x45) |
73cc19f1 HX |
1328 | goto out_unlock; |
1329 | ||
1330 | if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) | |
1331 | goto out_unlock; | |
1332 | ||
0eae88f3 ED |
1333 | id = ntohl(*(__be32 *)&iph->id); |
1334 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF)); | |
1075f3f6 | 1335 | id >>= 16; |
73cc19f1 HX |
1336 | |
1337 | for (p = *head; p; p = p->next) { | |
1338 | struct iphdr *iph2; | |
1339 | ||
1340 | if (!NAPI_GRO_CB(p)->same_flow) | |
1341 | continue; | |
1342 | ||
1343 | iph2 = ip_hdr(p); | |
1344 | ||
a5ad24be HX |
1345 | if ((iph->protocol ^ iph2->protocol) | |
1346 | (iph->tos ^ iph2->tos) | | |
0eae88f3 ED |
1347 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1348 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1349 | NAPI_GRO_CB(p)->same_flow = 0; |
1350 | continue; | |
1351 | } | |
1352 | ||
1353 | /* All fields must match except length and checksum. */ | |
1354 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be HX |
1355 | (iph->ttl ^ iph2->ttl) | |
1356 | ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id); | |
73cc19f1 HX |
1357 | |
1358 | NAPI_GRO_CB(p)->flush |= flush; | |
1359 | } | |
1360 | ||
1361 | NAPI_GRO_CB(skb)->flush |= flush; | |
86911732 HX |
1362 | skb_gro_pull(skb, sizeof(*iph)); |
1363 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 HX |
1364 | |
1365 | pp = ops->gro_receive(head, skb); | |
1366 | ||
1367 | out_unlock: | |
1368 | rcu_read_unlock(); | |
1369 | ||
1370 | out: | |
1371 | NAPI_GRO_CB(skb)->flush |= flush; | |
1372 | ||
1373 | return pp; | |
1374 | } | |
1375 | ||
1376 | static int inet_gro_complete(struct sk_buff *skb) | |
1377 | { | |
32613090 | 1378 | const struct net_protocol *ops; |
73cc19f1 HX |
1379 | struct iphdr *iph = ip_hdr(skb); |
1380 | int proto = iph->protocol & (MAX_INET_PROTOS - 1); | |
1381 | int err = -ENOSYS; | |
1382 | __be16 newlen = htons(skb->len - skb_network_offset(skb)); | |
1383 | ||
1384 | csum_replace2(&iph->check, iph->tot_len, newlen); | |
1385 | iph->tot_len = newlen; | |
1386 | ||
1387 | rcu_read_lock(); | |
1388 | ops = rcu_dereference(inet_protos[proto]); | |
1389 | if (WARN_ON(!ops || !ops->gro_complete)) | |
1390 | goto out_unlock; | |
1391 | ||
1392 | err = ops->gro_complete(skb); | |
1393 | ||
1394 | out_unlock: | |
1395 | rcu_read_unlock(); | |
1396 | ||
1397 | return err; | |
1398 | } | |
1399 | ||
eee4fe4d | 1400 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1401 | unsigned short type, unsigned char protocol, |
1402 | struct net *net) | |
3d58b5fa | 1403 | { |
eee4fe4d DL |
1404 | struct socket *sock; |
1405 | int rc = sock_create_kern(family, type, protocol, &sock); | |
3d58b5fa DL |
1406 | |
1407 | if (rc == 0) { | |
eee4fe4d DL |
1408 | *sk = sock->sk; |
1409 | (*sk)->sk_allocation = GFP_ATOMIC; | |
3d58b5fa DL |
1410 | /* |
1411 | * Unhash it so that IP input processing does not even see it, | |
1412 | * we do not wish this socket to see incoming packets. | |
1413 | */ | |
eee4fe4d | 1414 | (*sk)->sk_prot->unhash(*sk); |
5677242f DL |
1415 | |
1416 | sk_change_net(*sk, net); | |
3d58b5fa DL |
1417 | } |
1418 | return rc; | |
1419 | } | |
3d58b5fa DL |
1420 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1421 | ||
7d720c3e | 1422 | unsigned long snmp_fold_field(void __percpu *mib[], int offt) |
5e0f0435 HX |
1423 | { |
1424 | unsigned long res = 0; | |
1425 | int i; | |
1426 | ||
1427 | for_each_possible_cpu(i) { | |
1428 | res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt); | |
1429 | res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt); | |
1430 | } | |
1431 | return res; | |
1432 | } | |
1433 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1434 | ||
4ce3c183 ED |
1435 | #if BITS_PER_LONG==32 |
1436 | ||
1437 | u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset) | |
1438 | { | |
1439 | u64 res = 0; | |
1440 | int cpu; | |
1441 | ||
1442 | for_each_possible_cpu(cpu) { | |
1443 | void *bhptr, *userptr; | |
1444 | struct u64_stats_sync *syncp; | |
1445 | u64 v_bh, v_user; | |
1446 | unsigned int start; | |
1447 | ||
1448 | /* first mib used by softirq context, we must use _bh() accessors */ | |
1449 | bhptr = per_cpu_ptr(SNMP_STAT_BHPTR(mib), cpu); | |
1450 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); | |
1451 | do { | |
1452 | start = u64_stats_fetch_begin_bh(syncp); | |
1453 | v_bh = *(((u64 *) bhptr) + offt); | |
1454 | } while (u64_stats_fetch_retry_bh(syncp, start)); | |
1455 | ||
1456 | /* second mib used in USER context */ | |
1457 | userptr = per_cpu_ptr(SNMP_STAT_USRPTR(mib), cpu); | |
1458 | syncp = (struct u64_stats_sync *)(userptr + syncp_offset); | |
1459 | do { | |
1460 | start = u64_stats_fetch_begin(syncp); | |
1461 | v_user = *(((u64 *) userptr) + offt); | |
1462 | } while (u64_stats_fetch_retry(syncp, start)); | |
1463 | ||
1464 | res += v_bh + v_user; | |
1465 | } | |
1466 | return res; | |
1467 | } | |
1468 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1469 | #endif | |
1470 | ||
1823e4c8 | 1471 | int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align) |
5e0f0435 HX |
1472 | { |
1473 | BUG_ON(ptr == NULL); | |
1823e4c8 | 1474 | ptr[0] = __alloc_percpu(mibsize, align); |
5e0f0435 HX |
1475 | if (!ptr[0]) |
1476 | goto err0; | |
1823e4c8 | 1477 | ptr[1] = __alloc_percpu(mibsize, align); |
5e0f0435 HX |
1478 | if (!ptr[1]) |
1479 | goto err1; | |
1480 | return 0; | |
1481 | err1: | |
1482 | free_percpu(ptr[0]); | |
1483 | ptr[0] = NULL; | |
1484 | err0: | |
1485 | return -ENOMEM; | |
1486 | } | |
1487 | EXPORT_SYMBOL_GPL(snmp_mib_init); | |
1488 | ||
7d720c3e | 1489 | void snmp_mib_free(void __percpu *ptr[2]) |
5e0f0435 HX |
1490 | { |
1491 | BUG_ON(ptr == NULL); | |
1492 | free_percpu(ptr[0]); | |
1493 | free_percpu(ptr[1]); | |
1494 | ptr[0] = ptr[1] = NULL; | |
1495 | } | |
1496 | EXPORT_SYMBOL_GPL(snmp_mib_free); | |
1497 | ||
1da177e4 | 1498 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1499 | static const struct net_protocol igmp_protocol = { |
1da177e4 | 1500 | .handler = igmp_rcv, |
b4ee07df | 1501 | .netns_ok = 1, |
1da177e4 LT |
1502 | }; |
1503 | #endif | |
1504 | ||
32613090 | 1505 | static const struct net_protocol tcp_protocol = { |
1da177e4 LT |
1506 | .handler = tcp_v4_rcv, |
1507 | .err_handler = tcp_v4_err, | |
a430a43d | 1508 | .gso_send_check = tcp_v4_gso_send_check, |
f4c50d99 | 1509 | .gso_segment = tcp_tso_segment, |
bf296b12 HX |
1510 | .gro_receive = tcp4_gro_receive, |
1511 | .gro_complete = tcp4_gro_complete, | |
1da177e4 | 1512 | .no_policy = 1, |
92f1fecb | 1513 | .netns_ok = 1, |
1da177e4 LT |
1514 | }; |
1515 | ||
32613090 | 1516 | static const struct net_protocol udp_protocol = { |
1da177e4 LT |
1517 | .handler = udp_rcv, |
1518 | .err_handler = udp_err, | |
d7ca4cc0 SS |
1519 | .gso_send_check = udp4_ufo_send_check, |
1520 | .gso_segment = udp4_ufo_fragment, | |
1da177e4 | 1521 | .no_policy = 1, |
92f1fecb | 1522 | .netns_ok = 1, |
1da177e4 LT |
1523 | }; |
1524 | ||
32613090 | 1525 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1526 | .handler = icmp_rcv, |
8b7817f3 | 1527 | .no_policy = 1, |
92f1fecb | 1528 | .netns_ok = 1, |
1da177e4 LT |
1529 | }; |
1530 | ||
9b4661bd PE |
1531 | static __net_init int ipv4_mib_init_net(struct net *net) |
1532 | { | |
7d720c3e | 1533 | if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics, |
1823e4c8 ED |
1534 | sizeof(struct tcp_mib), |
1535 | __alignof__(struct tcp_mib)) < 0) | |
57ef42d5 | 1536 | goto err_tcp_mib; |
7d720c3e | 1537 | if (snmp_mib_init((void __percpu **)net->mib.ip_statistics, |
1823e4c8 ED |
1538 | sizeof(struct ipstats_mib), |
1539 | __alignof__(struct ipstats_mib)) < 0) | |
a20f5799 | 1540 | goto err_ip_mib; |
7d720c3e | 1541 | if (snmp_mib_init((void __percpu **)net->mib.net_statistics, |
1823e4c8 ED |
1542 | sizeof(struct linux_mib), |
1543 | __alignof__(struct linux_mib)) < 0) | |
61a7e260 | 1544 | goto err_net_mib; |
7d720c3e | 1545 | if (snmp_mib_init((void __percpu **)net->mib.udp_statistics, |
1823e4c8 ED |
1546 | sizeof(struct udp_mib), |
1547 | __alignof__(struct udp_mib)) < 0) | |
2f275f91 | 1548 | goto err_udp_mib; |
7d720c3e | 1549 | if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics, |
1823e4c8 ED |
1550 | sizeof(struct udp_mib), |
1551 | __alignof__(struct udp_mib)) < 0) | |
386019d3 | 1552 | goto err_udplite_mib; |
7d720c3e | 1553 | if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics, |
1823e4c8 ED |
1554 | sizeof(struct icmp_mib), |
1555 | __alignof__(struct icmp_mib)) < 0) | |
b60538a0 | 1556 | goto err_icmp_mib; |
7d720c3e | 1557 | if (snmp_mib_init((void __percpu **)net->mib.icmpmsg_statistics, |
1823e4c8 ED |
1558 | sizeof(struct icmpmsg_mib), |
1559 | __alignof__(struct icmpmsg_mib)) < 0) | |
923c6586 | 1560 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1561 | |
1562 | tcp_mib_init(net); | |
9b4661bd | 1563 | return 0; |
57ef42d5 | 1564 | |
923c6586 | 1565 | err_icmpmsg_mib: |
7d720c3e | 1566 | snmp_mib_free((void __percpu **)net->mib.icmp_statistics); |
b60538a0 | 1567 | err_icmp_mib: |
7d720c3e | 1568 | snmp_mib_free((void __percpu **)net->mib.udplite_statistics); |
386019d3 | 1569 | err_udplite_mib: |
7d720c3e | 1570 | snmp_mib_free((void __percpu **)net->mib.udp_statistics); |
2f275f91 | 1571 | err_udp_mib: |
7d720c3e | 1572 | snmp_mib_free((void __percpu **)net->mib.net_statistics); |
61a7e260 | 1573 | err_net_mib: |
7d720c3e | 1574 | snmp_mib_free((void __percpu **)net->mib.ip_statistics); |
a20f5799 | 1575 | err_ip_mib: |
7d720c3e | 1576 | snmp_mib_free((void __percpu **)net->mib.tcp_statistics); |
57ef42d5 PE |
1577 | err_tcp_mib: |
1578 | return -ENOMEM; | |
9b4661bd PE |
1579 | } |
1580 | ||
1581 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1582 | { | |
7d720c3e TH |
1583 | snmp_mib_free((void __percpu **)net->mib.icmpmsg_statistics); |
1584 | snmp_mib_free((void __percpu **)net->mib.icmp_statistics); | |
1585 | snmp_mib_free((void __percpu **)net->mib.udplite_statistics); | |
1586 | snmp_mib_free((void __percpu **)net->mib.udp_statistics); | |
1587 | snmp_mib_free((void __percpu **)net->mib.net_statistics); | |
1588 | snmp_mib_free((void __percpu **)net->mib.ip_statistics); | |
1589 | snmp_mib_free((void __percpu **)net->mib.tcp_statistics); | |
9b4661bd PE |
1590 | } |
1591 | ||
1592 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1593 | .init = ipv4_mib_init_net, | |
1594 | .exit = ipv4_mib_exit_net, | |
1595 | }; | |
1596 | ||
1da177e4 LT |
1597 | static int __init init_ipv4_mibs(void) |
1598 | { | |
d89cbbb1 | 1599 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1600 | } |
1601 | ||
1602 | static int ipv4_proc_init(void); | |
1da177e4 | 1603 | |
30e224d7 HX |
1604 | /* |
1605 | * IP protocol layer initialiser | |
1606 | */ | |
1607 | ||
7546dd97 | 1608 | static struct packet_type ip_packet_type __read_mostly = { |
09640e63 | 1609 | .type = cpu_to_be16(ETH_P_IP), |
30e224d7 | 1610 | .func = ip_rcv, |
a430a43d | 1611 | .gso_send_check = inet_gso_send_check, |
f4c50d99 | 1612 | .gso_segment = inet_gso_segment, |
73cc19f1 HX |
1613 | .gro_receive = inet_gro_receive, |
1614 | .gro_complete = inet_gro_complete, | |
30e224d7 HX |
1615 | }; |
1616 | ||
1da177e4 LT |
1617 | static int __init inet_init(void) |
1618 | { | |
1619 | struct sk_buff *dummy_skb; | |
1620 | struct inet_protosw *q; | |
1621 | struct list_head *r; | |
1622 | int rc = -EINVAL; | |
1623 | ||
ef047f5e | 1624 | BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb)); |
1da177e4 | 1625 | |
e3826f1e AW |
1626 | sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL); |
1627 | if (!sysctl_local_reserved_ports) | |
1628 | goto out; | |
1629 | ||
1da177e4 LT |
1630 | rc = proto_register(&tcp_prot, 1); |
1631 | if (rc) | |
e3826f1e | 1632 | goto out_free_reserved_ports; |
1da177e4 LT |
1633 | |
1634 | rc = proto_register(&udp_prot, 1); | |
1635 | if (rc) | |
1636 | goto out_unregister_tcp_proto; | |
1637 | ||
1638 | rc = proto_register(&raw_prot, 1); | |
1639 | if (rc) | |
1640 | goto out_unregister_udp_proto; | |
1641 | ||
1642 | /* | |
e905a9ed | 1643 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1644 | */ |
1645 | ||
e905a9ed | 1646 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1647 | |
bd7b1533 AV |
1648 | #ifdef CONFIG_SYSCTL |
1649 | ip_static_sysctl_init(); | |
1650 | #endif | |
1651 | ||
1da177e4 LT |
1652 | /* |
1653 | * Add all the base protocols. | |
1654 | */ | |
1655 | ||
1656 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
1657 | printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n"); | |
1658 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) | |
1659 | printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n"); | |
1660 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) | |
1661 | printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n"); | |
1662 | #ifdef CONFIG_IP_MULTICAST | |
1663 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
1664 | printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n"); | |
1665 | #endif | |
1666 | ||
1667 | /* Register the socket-side information for inet_create. */ | |
1668 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1669 | INIT_LIST_HEAD(r); | |
1670 | ||
1671 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1672 | inet_register_protosw(q); | |
1673 | ||
1674 | /* | |
1675 | * Set the ARP module up | |
1676 | */ | |
1677 | ||
1678 | arp_init(); | |
1679 | ||
e905a9ed YH |
1680 | /* |
1681 | * Set the IP module up | |
1682 | */ | |
1da177e4 LT |
1683 | |
1684 | ip_init(); | |
1685 | ||
9b0f976f | 1686 | tcp_v4_init(); |
1da177e4 LT |
1687 | |
1688 | /* Setup TCP slab cache for open requests. */ | |
1689 | tcp_init(); | |
1690 | ||
95766fff HA |
1691 | /* Setup UDP memory threshold */ |
1692 | udp_init(); | |
1693 | ||
ba4e58ec GR |
1694 | /* Add UDP-Lite (RFC 3828) */ |
1695 | udplite4_register(); | |
1da177e4 LT |
1696 | |
1697 | /* | |
1698 | * Set the ICMP layer up | |
1699 | */ | |
1700 | ||
a5710d65 DL |
1701 | if (icmp_init() < 0) |
1702 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
1703 | |
1704 | /* | |
1705 | * Initialise the multicast router | |
1706 | */ | |
1707 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 WC |
1708 | if (ip_mr_init()) |
1709 | printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n"); | |
1da177e4 LT |
1710 | #endif |
1711 | /* | |
1712 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 1713 | */ |
1da177e4 | 1714 | |
132adf54 | 1715 | if (init_ipv4_mibs()) |
03d2f897 | 1716 | printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); |
e905a9ed | 1717 | |
1da177e4 LT |
1718 | ipv4_proc_init(); |
1719 | ||
1720 | ipfrag_init(); | |
1721 | ||
30e224d7 HX |
1722 | dev_add_pack(&ip_packet_type); |
1723 | ||
1da177e4 LT |
1724 | rc = 0; |
1725 | out: | |
1726 | return rc; | |
1da177e4 LT |
1727 | out_unregister_udp_proto: |
1728 | proto_unregister(&udp_prot); | |
321efff7 OK |
1729 | out_unregister_tcp_proto: |
1730 | proto_unregister(&tcp_prot); | |
e3826f1e AW |
1731 | out_free_reserved_ports: |
1732 | kfree(sysctl_local_reserved_ports); | |
1da177e4 LT |
1733 | goto out; |
1734 | } | |
1735 | ||
a536e077 | 1736 | fs_initcall(inet_init); |
1da177e4 LT |
1737 | |
1738 | /* ------------------------------------------------------------------------ */ | |
1739 | ||
1740 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
1741 | static int __init ipv4_proc_init(void) |
1742 | { | |
1743 | int rc = 0; | |
1744 | ||
1745 | if (raw_proc_init()) | |
1746 | goto out_raw; | |
1747 | if (tcp4_proc_init()) | |
1748 | goto out_tcp; | |
1749 | if (udp4_proc_init()) | |
1750 | goto out_udp; | |
1da177e4 LT |
1751 | if (ip_misc_proc_init()) |
1752 | goto out_misc; | |
1753 | out: | |
1754 | return rc; | |
1755 | out_misc: | |
1da177e4 LT |
1756 | udp4_proc_exit(); |
1757 | out_udp: | |
1758 | tcp4_proc_exit(); | |
1759 | out_tcp: | |
1760 | raw_proc_exit(); | |
1761 | out_raw: | |
1762 | rc = -ENOMEM; | |
1763 | goto out; | |
1764 | } | |
1765 | ||
1766 | #else /* CONFIG_PROC_FS */ | |
1767 | static int __init ipv4_proc_init(void) | |
1768 | { | |
1769 | return 0; | |
1770 | } | |
1771 | #endif /* CONFIG_PROC_FS */ | |
1772 | ||
1773 | MODULE_ALIAS_NETPROTO(PF_INET); | |
1774 |