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 | * | |
8 | * Version: $Id: af_inet.c,v 1.137 2002/02/01 22:01:03 davem Exp $ | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Florian La Roche, <flla@stud.uni-sb.de> | |
13 | * Alan Cox, <A.Cox@swansea.ac.uk> | |
14 | * | |
15 | * Changes (see also sock.c) | |
16 | * | |
17 | * piggy, | |
18 | * Karl Knutson : Socket protocol table | |
19 | * A.N.Kuznetsov : Socket death error in accept(). | |
20 | * John Richardson : Fix non blocking error in connect() | |
21 | * so sockets that fail to connect | |
22 | * don't return -EINPROGRESS. | |
23 | * Alan Cox : Asynchronous I/O support | |
24 | * Alan Cox : Keep correct socket pointer on sock | |
25 | * structures | |
26 | * when accept() ed | |
27 | * Alan Cox : Semantics of SO_LINGER aren't state | |
28 | * moved to close when you look carefully. | |
29 | * With this fixed and the accept bug fixed | |
30 | * some RPC stuff seems happier. | |
31 | * Niibe Yutaka : 4.4BSD style write async I/O | |
32 | * Alan Cox, | |
33 | * Tony Gale : Fixed reuse semantics. | |
34 | * Alan Cox : bind() shouldn't abort existing but dead | |
35 | * sockets. Stops FTP netin:.. I hope. | |
36 | * Alan Cox : bind() works correctly for RAW sockets. | |
37 | * Note that FreeBSD at least was broken | |
38 | * in this respect so be careful with | |
39 | * compatibility tests... | |
40 | * Alan Cox : routing cache support | |
41 | * Alan Cox : memzero the socket structure for | |
42 | * compactness. | |
43 | * Matt Day : nonblock connect error handler | |
44 | * Alan Cox : Allow large numbers of pending sockets | |
45 | * (eg for big web sites), but only if | |
46 | * specifically application requested. | |
47 | * Alan Cox : New buffering throughout IP. Used | |
48 | * dumbly. | |
49 | * Alan Cox : New buffering now used smartly. | |
50 | * Alan Cox : BSD rather than common sense | |
51 | * interpretation of listen. | |
52 | * Germano Caronni : Assorted small races. | |
53 | * Alan Cox : sendmsg/recvmsg basic support. | |
54 | * Alan Cox : Only sendmsg/recvmsg now supported. | |
55 | * Alan Cox : Locked down bind (see security list). | |
56 | * Alan Cox : Loosened bind a little. | |
57 | * Mike McLagan : ADD/DEL DLCI Ioctls | |
58 | * Willy Konynenberg : Transparent proxying support. | |
59 | * David S. Miller : New socket lookup architecture. | |
60 | * Some other random speedups. | |
61 | * Cyrus Durgin : Cleaned up file for kmod hacks. | |
62 | * Andi Kleen : Fix inet_stream_connect TCP race. | |
63 | * | |
64 | * This program is free software; you can redistribute it and/or | |
65 | * modify it under the terms of the GNU General Public License | |
66 | * as published by the Free Software Foundation; either version | |
67 | * 2 of the License, or (at your option) any later version. | |
68 | */ | |
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> |
1da177e4 LT |
91 | |
92 | #include <asm/uaccess.h> | |
93 | #include <asm/system.h> | |
94 | ||
1da177e4 LT |
95 | #include <linux/inet.h> |
96 | #include <linux/igmp.h> | |
14c85021 | 97 | #include <linux/inetdevice.h> |
1da177e4 LT |
98 | #include <linux/netdevice.h> |
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> | |
115 | #ifdef CONFIG_IP_MROUTE | |
116 | #include <linux/mroute.h> | |
117 | #endif | |
118 | ||
ba89966c | 119 | DEFINE_SNMP_STAT(struct linux_mib, net_statistics) __read_mostly; |
1da177e4 | 120 | |
1da177e4 LT |
121 | extern void ip_mc_drop_socket(struct sock *sk); |
122 | ||
123 | /* The inetsw table contains everything that inet_create needs to | |
124 | * build a new socket. | |
125 | */ | |
126 | static struct list_head inetsw[SOCK_MAX]; | |
127 | static DEFINE_SPINLOCK(inetsw_lock); | |
128 | ||
129 | /* New destruction routine */ | |
130 | ||
131 | void inet_sock_destruct(struct sock *sk) | |
132 | { | |
133 | struct inet_sock *inet = inet_sk(sk); | |
134 | ||
135 | __skb_queue_purge(&sk->sk_receive_queue); | |
136 | __skb_queue_purge(&sk->sk_error_queue); | |
137 | ||
138 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { | |
139 | printk("Attempt to release TCP socket in state %d %p\n", | |
140 | sk->sk_state, sk); | |
141 | return; | |
142 | } | |
143 | if (!sock_flag(sk, SOCK_DEAD)) { | |
144 | printk("Attempt to release alive inet socket %p\n", sk); | |
145 | return; | |
146 | } | |
147 | ||
148 | BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); | |
149 | BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); | |
150 | BUG_TRAP(!sk->sk_wmem_queued); | |
151 | BUG_TRAP(!sk->sk_forward_alloc); | |
152 | ||
a51482bd | 153 | kfree(inet->opt); |
1da177e4 | 154 | dst_release(sk->sk_dst_cache); |
e6848976 | 155 | sk_refcnt_debug_dec(sk); |
1da177e4 LT |
156 | } |
157 | ||
158 | /* | |
159 | * The routines beyond this point handle the behaviour of an AF_INET | |
160 | * socket object. Mostly it punts to the subprotocols of IP to do | |
161 | * the work. | |
162 | */ | |
163 | ||
164 | /* | |
165 | * Automatically bind an unbound socket. | |
166 | */ | |
167 | ||
168 | static int inet_autobind(struct sock *sk) | |
169 | { | |
170 | struct inet_sock *inet; | |
171 | /* We may need to bind the socket. */ | |
172 | lock_sock(sk); | |
173 | inet = inet_sk(sk); | |
174 | if (!inet->num) { | |
175 | if (sk->sk_prot->get_port(sk, 0)) { | |
176 | release_sock(sk); | |
177 | return -EAGAIN; | |
178 | } | |
179 | inet->sport = htons(inet->num); | |
180 | } | |
181 | release_sock(sk); | |
182 | return 0; | |
183 | } | |
184 | ||
185 | /* | |
186 | * Move a socket into listening state. | |
187 | */ | |
188 | int inet_listen(struct socket *sock, int backlog) | |
189 | { | |
190 | struct sock *sk = sock->sk; | |
191 | unsigned char old_state; | |
192 | int err; | |
193 | ||
194 | lock_sock(sk); | |
195 | ||
196 | err = -EINVAL; | |
197 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
198 | goto out; | |
199 | ||
200 | old_state = sk->sk_state; | |
201 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) | |
202 | goto out; | |
203 | ||
204 | /* Really, if the socket is already in listen state | |
205 | * we can only allow the backlog to be adjusted. | |
206 | */ | |
207 | if (old_state != TCP_LISTEN) { | |
72a3effa | 208 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
209 | if (err) |
210 | goto out; | |
211 | } | |
212 | sk->sk_max_ack_backlog = backlog; | |
213 | err = 0; | |
214 | ||
215 | out: | |
216 | release_sock(sk); | |
217 | return err; | |
218 | } | |
219 | ||
be776281 | 220 | u32 inet_ehash_secret __read_mostly; |
b3da2cf3 DM |
221 | EXPORT_SYMBOL(inet_ehash_secret); |
222 | ||
be776281 ED |
223 | /* |
224 | * inet_ehash_secret must be set exactly once | |
225 | * Instead of using a dedicated spinlock, we (ab)use inetsw_lock | |
226 | */ | |
b3da2cf3 DM |
227 | void build_ehash_secret(void) |
228 | { | |
be776281 ED |
229 | u32 rnd; |
230 | do { | |
231 | get_random_bytes(&rnd, sizeof(rnd)); | |
232 | } while (rnd == 0); | |
233 | spin_lock_bh(&inetsw_lock); | |
234 | if (!inet_ehash_secret) | |
235 | inet_ehash_secret = rnd; | |
236 | spin_unlock_bh(&inetsw_lock); | |
b3da2cf3 DM |
237 | } |
238 | EXPORT_SYMBOL(build_ehash_secret); | |
239 | ||
1da177e4 LT |
240 | /* |
241 | * Create an inet socket. | |
242 | */ | |
243 | ||
1b8d7ae4 | 244 | static int inet_create(struct net *net, struct socket *sock, int protocol) |
1da177e4 LT |
245 | { |
246 | struct sock *sk; | |
247 | struct list_head *p; | |
248 | struct inet_protosw *answer; | |
249 | struct inet_sock *inet; | |
250 | struct proto *answer_prot; | |
251 | unsigned char answer_flags; | |
252 | char answer_no_check; | |
bb97d31f | 253 | int try_loading_module = 0; |
86c8f9d1 | 254 | int err; |
1da177e4 | 255 | |
1b8d7ae4 EB |
256 | if (net != &init_net) |
257 | return -EAFNOSUPPORT; | |
258 | ||
b3da2cf3 DM |
259 | if (sock->type != SOCK_RAW && |
260 | sock->type != SOCK_DGRAM && | |
261 | !inet_ehash_secret) | |
262 | build_ehash_secret(); | |
263 | ||
1da177e4 LT |
264 | sock->state = SS_UNCONNECTED; |
265 | ||
266 | /* Look for the requested type/protocol pair. */ | |
267 | answer = NULL; | |
bb97d31f | 268 | lookup_protocol: |
86c8f9d1 | 269 | err = -ESOCKTNOSUPPORT; |
1da177e4 LT |
270 | rcu_read_lock(); |
271 | list_for_each_rcu(p, &inetsw[sock->type]) { | |
272 | answer = list_entry(p, struct inet_protosw, list); | |
273 | ||
274 | /* Check the non-wild match. */ | |
275 | if (protocol == answer->protocol) { | |
276 | if (protocol != IPPROTO_IP) | |
277 | break; | |
278 | } else { | |
279 | /* Check for the two wild cases. */ | |
280 | if (IPPROTO_IP == protocol) { | |
281 | protocol = answer->protocol; | |
282 | break; | |
283 | } | |
284 | if (IPPROTO_IP == answer->protocol) | |
285 | break; | |
286 | } | |
86c8f9d1 | 287 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
288 | answer = NULL; |
289 | } | |
290 | ||
bb97d31f ACM |
291 | if (unlikely(answer == NULL)) { |
292 | if (try_loading_module < 2) { | |
293 | rcu_read_unlock(); | |
294 | /* | |
295 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
296 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
297 | */ | |
298 | if (++try_loading_module == 1) | |
299 | request_module("net-pf-%d-proto-%d-type-%d", | |
300 | PF_INET, protocol, sock->type); | |
301 | /* | |
302 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
303 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
304 | */ | |
305 | else | |
306 | request_module("net-pf-%d-proto-%d", | |
307 | PF_INET, protocol); | |
308 | goto lookup_protocol; | |
309 | } else | |
310 | goto out_rcu_unlock; | |
311 | } | |
312 | ||
1da177e4 LT |
313 | err = -EPERM; |
314 | if (answer->capability > 0 && !capable(answer->capability)) | |
315 | goto out_rcu_unlock; | |
1da177e4 LT |
316 | |
317 | sock->ops = answer->ops; | |
318 | answer_prot = answer->prot; | |
319 | answer_no_check = answer->no_check; | |
320 | answer_flags = answer->flags; | |
321 | rcu_read_unlock(); | |
322 | ||
323 | BUG_TRAP(answer_prot->slab != NULL); | |
324 | ||
325 | err = -ENOBUFS; | |
1b8d7ae4 | 326 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, 1); |
1da177e4 LT |
327 | if (sk == NULL) |
328 | goto out; | |
329 | ||
330 | err = 0; | |
331 | sk->sk_no_check = answer_no_check; | |
332 | if (INET_PROTOSW_REUSE & answer_flags) | |
333 | sk->sk_reuse = 1; | |
334 | ||
335 | inet = inet_sk(sk); | |
469de9b9 | 336 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 LT |
337 | |
338 | if (SOCK_RAW == sock->type) { | |
339 | inet->num = protocol; | |
340 | if (IPPROTO_RAW == protocol) | |
341 | inet->hdrincl = 1; | |
342 | } | |
343 | ||
344 | if (ipv4_config.no_pmtu_disc) | |
345 | inet->pmtudisc = IP_PMTUDISC_DONT; | |
346 | else | |
347 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
348 | ||
349 | inet->id = 0; | |
350 | ||
351 | sock_init_data(sock, sk); | |
352 | ||
353 | sk->sk_destruct = inet_sock_destruct; | |
354 | sk->sk_family = PF_INET; | |
355 | sk->sk_protocol = protocol; | |
356 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
357 | ||
358 | inet->uc_ttl = -1; | |
359 | inet->mc_loop = 1; | |
360 | inet->mc_ttl = 1; | |
361 | inet->mc_index = 0; | |
362 | inet->mc_list = NULL; | |
363 | ||
e6848976 | 364 | sk_refcnt_debug_inc(sk); |
1da177e4 LT |
365 | |
366 | if (inet->num) { | |
367 | /* It assumes that any protocol which allows | |
368 | * the user to assign a number at socket | |
369 | * creation time automatically | |
370 | * shares. | |
371 | */ | |
372 | inet->sport = htons(inet->num); | |
373 | /* Add to protocol hash chains. */ | |
374 | sk->sk_prot->hash(sk); | |
375 | } | |
376 | ||
377 | if (sk->sk_prot->init) { | |
378 | err = sk->sk_prot->init(sk); | |
379 | if (err) | |
380 | sk_common_release(sk); | |
381 | } | |
382 | out: | |
383 | return err; | |
384 | out_rcu_unlock: | |
385 | rcu_read_unlock(); | |
386 | goto out; | |
387 | } | |
388 | ||
389 | ||
390 | /* | |
391 | * The peer socket should always be NULL (or else). When we call this | |
392 | * function we are destroying the object and from then on nobody | |
393 | * should refer to it. | |
394 | */ | |
395 | int inet_release(struct socket *sock) | |
396 | { | |
397 | struct sock *sk = sock->sk; | |
398 | ||
399 | if (sk) { | |
400 | long timeout; | |
401 | ||
402 | /* Applications forget to leave groups before exiting */ | |
403 | ip_mc_drop_socket(sk); | |
404 | ||
405 | /* If linger is set, we don't return until the close | |
406 | * is complete. Otherwise we return immediately. The | |
407 | * actually closing is done the same either way. | |
408 | * | |
409 | * If the close is due to the process exiting, we never | |
410 | * linger.. | |
411 | */ | |
412 | timeout = 0; | |
413 | if (sock_flag(sk, SOCK_LINGER) && | |
414 | !(current->flags & PF_EXITING)) | |
415 | timeout = sk->sk_lingertime; | |
416 | sock->sk = NULL; | |
417 | sk->sk_prot->close(sk, timeout); | |
418 | } | |
419 | return 0; | |
420 | } | |
421 | ||
422 | /* It is off by default, see below. */ | |
ab32ea5d | 423 | int sysctl_ip_nonlocal_bind __read_mostly; |
1da177e4 LT |
424 | |
425 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
426 | { | |
427 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
428 | struct sock *sk = sock->sk; | |
429 | struct inet_sock *inet = inet_sk(sk); | |
430 | unsigned short snum; | |
431 | int chk_addr_ret; | |
432 | int err; | |
433 | ||
434 | /* If the socket has its own bind function then use it. (RAW) */ | |
435 | if (sk->sk_prot->bind) { | |
436 | err = sk->sk_prot->bind(sk, uaddr, addr_len); | |
437 | goto out; | |
438 | } | |
439 | err = -EINVAL; | |
440 | if (addr_len < sizeof(struct sockaddr_in)) | |
441 | goto out; | |
442 | ||
443 | chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr); | |
444 | ||
445 | /* Not specified by any standard per-se, however it breaks too | |
446 | * many applications when removed. It is unfortunate since | |
447 | * allowing applications to make a non-local bind solves | |
448 | * several problems with systems using dynamic addressing. | |
449 | * (ie. your servers still start up even if your ISDN link | |
450 | * is temporarily down) | |
451 | */ | |
452 | err = -EADDRNOTAVAIL; | |
453 | if (!sysctl_ip_nonlocal_bind && | |
454 | !inet->freebind && | |
455 | addr->sin_addr.s_addr != INADDR_ANY && | |
456 | chk_addr_ret != RTN_LOCAL && | |
457 | chk_addr_ret != RTN_MULTICAST && | |
458 | chk_addr_ret != RTN_BROADCAST) | |
459 | goto out; | |
460 | ||
461 | snum = ntohs(addr->sin_port); | |
462 | err = -EACCES; | |
463 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) | |
464 | goto out; | |
465 | ||
466 | /* We keep a pair of addresses. rcv_saddr is the one | |
467 | * used by hash lookups, and saddr is used for transmit. | |
468 | * | |
469 | * In the BSD API these are the same except where it | |
470 | * would be illegal to use them (multicast/broadcast) in | |
471 | * which case the sending device address is used. | |
472 | */ | |
473 | lock_sock(sk); | |
474 | ||
475 | /* Check these errors (active socket, double bind). */ | |
476 | err = -EINVAL; | |
477 | if (sk->sk_state != TCP_CLOSE || inet->num) | |
478 | goto out_release_sock; | |
479 | ||
480 | inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr; | |
481 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) | |
482 | inet->saddr = 0; /* Use device */ | |
483 | ||
484 | /* Make sure we are allowed to bind here. */ | |
485 | if (sk->sk_prot->get_port(sk, snum)) { | |
486 | inet->saddr = inet->rcv_saddr = 0; | |
487 | err = -EADDRINUSE; | |
488 | goto out_release_sock; | |
489 | } | |
490 | ||
491 | if (inet->rcv_saddr) | |
492 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; | |
493 | if (snum) | |
494 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
495 | inet->sport = htons(inet->num); | |
496 | inet->daddr = 0; | |
497 | inet->dport = 0; | |
498 | sk_dst_reset(sk); | |
499 | err = 0; | |
500 | out_release_sock: | |
501 | release_sock(sk); | |
502 | out: | |
503 | return err; | |
504 | } | |
505 | ||
506 | int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr, | |
507 | int addr_len, int flags) | |
508 | { | |
509 | struct sock *sk = sock->sk; | |
510 | ||
511 | if (uaddr->sa_family == AF_UNSPEC) | |
512 | return sk->sk_prot->disconnect(sk, flags); | |
513 | ||
514 | if (!inet_sk(sk)->num && inet_autobind(sk)) | |
515 | return -EAGAIN; | |
516 | return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len); | |
517 | } | |
518 | ||
519 | static long inet_wait_for_connect(struct sock *sk, long timeo) | |
520 | { | |
521 | DEFINE_WAIT(wait); | |
522 | ||
523 | prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); | |
524 | ||
525 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
526 | * change state of the socket from TCP_SYN_*. | |
527 | * Connect() does not allow to get error notifications | |
528 | * without closing the socket. | |
529 | */ | |
530 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
531 | release_sock(sk); | |
532 | timeo = schedule_timeout(timeo); | |
533 | lock_sock(sk); | |
534 | if (signal_pending(current) || !timeo) | |
535 | break; | |
536 | prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE); | |
537 | } | |
538 | finish_wait(sk->sk_sleep, &wait); | |
539 | return timeo; | |
540 | } | |
541 | ||
542 | /* | |
543 | * Connect to a remote host. There is regrettably still a little | |
544 | * TCP 'magic' in here. | |
545 | */ | |
546 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
547 | int addr_len, int flags) | |
548 | { | |
549 | struct sock *sk = sock->sk; | |
550 | int err; | |
551 | long timeo; | |
552 | ||
553 | lock_sock(sk); | |
554 | ||
555 | if (uaddr->sa_family == AF_UNSPEC) { | |
556 | err = sk->sk_prot->disconnect(sk, flags); | |
557 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
558 | goto out; | |
559 | } | |
560 | ||
561 | switch (sock->state) { | |
562 | default: | |
563 | err = -EINVAL; | |
564 | goto out; | |
565 | case SS_CONNECTED: | |
566 | err = -EISCONN; | |
567 | goto out; | |
568 | case SS_CONNECTING: | |
569 | err = -EALREADY; | |
570 | /* Fall out of switch with err, set for this state */ | |
571 | break; | |
572 | case SS_UNCONNECTED: | |
573 | err = -EISCONN; | |
574 | if (sk->sk_state != TCP_CLOSE) | |
575 | goto out; | |
576 | ||
577 | err = sk->sk_prot->connect(sk, uaddr, addr_len); | |
578 | if (err < 0) | |
579 | goto out; | |
580 | ||
e905a9ed | 581 | sock->state = SS_CONNECTING; |
1da177e4 LT |
582 | |
583 | /* Just entered SS_CONNECTING state; the only | |
584 | * difference is that return value in non-blocking | |
585 | * case is EINPROGRESS, rather than EALREADY. | |
586 | */ | |
587 | err = -EINPROGRESS; | |
588 | break; | |
589 | } | |
590 | ||
591 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
592 | ||
593 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
594 | /* Error code is set above */ | |
595 | if (!timeo || !inet_wait_for_connect(sk, timeo)) | |
596 | goto out; | |
597 | ||
598 | err = sock_intr_errno(timeo); | |
599 | if (signal_pending(current)) | |
600 | goto out; | |
601 | } | |
602 | ||
603 | /* Connection was closed by RST, timeout, ICMP error | |
604 | * or another process disconnected us. | |
605 | */ | |
606 | if (sk->sk_state == TCP_CLOSE) | |
607 | goto sock_error; | |
608 | ||
609 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
610 | * and error was received after socket entered established state. | |
611 | * Hence, it is handled normally after connect() return successfully. | |
612 | */ | |
613 | ||
614 | sock->state = SS_CONNECTED; | |
615 | err = 0; | |
616 | out: | |
617 | release_sock(sk); | |
618 | return err; | |
619 | ||
620 | sock_error: | |
621 | err = sock_error(sk) ? : -ECONNABORTED; | |
622 | sock->state = SS_UNCONNECTED; | |
623 | if (sk->sk_prot->disconnect(sk, flags)) | |
624 | sock->state = SS_DISCONNECTING; | |
625 | goto out; | |
626 | } | |
627 | ||
628 | /* | |
629 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
630 | */ | |
631 | ||
632 | int inet_accept(struct socket *sock, struct socket *newsock, int flags) | |
633 | { | |
634 | struct sock *sk1 = sock->sk; | |
635 | int err = -EINVAL; | |
636 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err); | |
637 | ||
638 | if (!sk2) | |
639 | goto do_err; | |
640 | ||
641 | lock_sock(sk2); | |
642 | ||
643 | BUG_TRAP((1 << sk2->sk_state) & | |
644 | (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)); | |
645 | ||
646 | sock_graft(sk2, newsock); | |
647 | ||
648 | newsock->state = SS_CONNECTED; | |
649 | err = 0; | |
650 | release_sock(sk2); | |
651 | do_err: | |
652 | return err; | |
653 | } | |
654 | ||
655 | ||
656 | /* | |
657 | * This does both peername and sockname. | |
658 | */ | |
659 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
660 | int *uaddr_len, int peer) | |
661 | { | |
662 | struct sock *sk = sock->sk; | |
663 | struct inet_sock *inet = inet_sk(sk); | |
664 | struct sockaddr_in *sin = (struct sockaddr_in *)uaddr; | |
665 | ||
666 | sin->sin_family = AF_INET; | |
667 | if (peer) { | |
668 | if (!inet->dport || | |
669 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && | |
670 | peer == 1)) | |
671 | return -ENOTCONN; | |
672 | sin->sin_port = inet->dport; | |
673 | sin->sin_addr.s_addr = inet->daddr; | |
674 | } else { | |
714e85be | 675 | __be32 addr = inet->rcv_saddr; |
1da177e4 LT |
676 | if (!addr) |
677 | addr = inet->saddr; | |
678 | sin->sin_port = inet->sport; | |
679 | sin->sin_addr.s_addr = addr; | |
680 | } | |
681 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
682 | *uaddr_len = sizeof(*sin); | |
683 | return 0; | |
684 | } | |
685 | ||
686 | int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg, | |
687 | size_t size) | |
688 | { | |
689 | struct sock *sk = sock->sk; | |
690 | ||
691 | /* We may need to bind the socket. */ | |
692 | if (!inet_sk(sk)->num && inet_autobind(sk)) | |
693 | return -EAGAIN; | |
694 | ||
695 | return sk->sk_prot->sendmsg(iocb, sk, msg, size); | |
696 | } | |
697 | ||
698 | ||
699 | static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags) | |
700 | { | |
701 | struct sock *sk = sock->sk; | |
702 | ||
703 | /* We may need to bind the socket. */ | |
704 | if (!inet_sk(sk)->num && inet_autobind(sk)) | |
705 | return -EAGAIN; | |
706 | ||
707 | if (sk->sk_prot->sendpage) | |
708 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
709 | return sock_no_sendpage(sock, page, offset, size, flags); | |
710 | } | |
711 | ||
712 | ||
713 | int inet_shutdown(struct socket *sock, int how) | |
714 | { | |
715 | struct sock *sk = sock->sk; | |
716 | int err = 0; | |
717 | ||
718 | /* This should really check to make sure | |
719 | * the socket is a TCP socket. (WHY AC...) | |
720 | */ | |
721 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
722 | 1->2 bit 2 snds. | |
723 | 2->3 */ | |
724 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
725 | return -EINVAL; | |
726 | ||
727 | lock_sock(sk); | |
728 | if (sock->state == SS_CONNECTING) { | |
729 | if ((1 << sk->sk_state) & | |
730 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
731 | sock->state = SS_DISCONNECTING; | |
732 | else | |
733 | sock->state = SS_CONNECTED; | |
734 | } | |
735 | ||
736 | switch (sk->sk_state) { | |
737 | case TCP_CLOSE: | |
738 | err = -ENOTCONN; | |
739 | /* Hack to wake up other listeners, who can poll for | |
740 | POLLHUP, even on eg. unconnected UDP sockets -- RR */ | |
741 | default: | |
742 | sk->sk_shutdown |= how; | |
743 | if (sk->sk_prot->shutdown) | |
744 | sk->sk_prot->shutdown(sk, how); | |
745 | break; | |
746 | ||
747 | /* Remaining two branches are temporary solution for missing | |
748 | * close() in multithreaded environment. It is _not_ a good idea, | |
749 | * but we have no choice until close() is repaired at VFS level. | |
750 | */ | |
751 | case TCP_LISTEN: | |
752 | if (!(how & RCV_SHUTDOWN)) | |
753 | break; | |
754 | /* Fall through */ | |
755 | case TCP_SYN_SENT: | |
756 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
757 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
758 | break; | |
759 | } | |
760 | ||
761 | /* Wake up anyone sleeping in poll. */ | |
762 | sk->sk_state_change(sk); | |
763 | release_sock(sk); | |
764 | return err; | |
765 | } | |
766 | ||
767 | /* | |
768 | * ioctl() calls you can issue on an INET socket. Most of these are | |
769 | * device configuration and stuff and very rarely used. Some ioctls | |
770 | * pass on to the socket itself. | |
771 | * | |
772 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
773 | * loads the devconfigure module does its configuring and unloads it. | |
774 | * There's a good 20K of config code hanging around the kernel. | |
775 | */ | |
776 | ||
777 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
778 | { | |
779 | struct sock *sk = sock->sk; | |
780 | int err = 0; | |
781 | ||
782 | switch (cmd) { | |
783 | case SIOCGSTAMP: | |
784 | err = sock_get_timestamp(sk, (struct timeval __user *)arg); | |
785 | break; | |
ae40eb1e ED |
786 | case SIOCGSTAMPNS: |
787 | err = sock_get_timestampns(sk, (struct timespec __user *)arg); | |
788 | break; | |
1da177e4 LT |
789 | case SIOCADDRT: |
790 | case SIOCDELRT: | |
791 | case SIOCRTMSG: | |
792 | err = ip_rt_ioctl(cmd, (void __user *)arg); | |
793 | break; | |
794 | case SIOCDARP: | |
795 | case SIOCGARP: | |
796 | case SIOCSARP: | |
797 | err = arp_ioctl(cmd, (void __user *)arg); | |
798 | break; | |
799 | case SIOCGIFADDR: | |
800 | case SIOCSIFADDR: | |
801 | case SIOCGIFBRDADDR: | |
802 | case SIOCSIFBRDADDR: | |
803 | case SIOCGIFNETMASK: | |
804 | case SIOCSIFNETMASK: | |
805 | case SIOCGIFDSTADDR: | |
806 | case SIOCSIFDSTADDR: | |
807 | case SIOCSIFPFLAGS: | |
808 | case SIOCGIFPFLAGS: | |
809 | case SIOCSIFFLAGS: | |
810 | err = devinet_ioctl(cmd, (void __user *)arg); | |
811 | break; | |
812 | default: | |
b5e5fa5e CH |
813 | if (sk->sk_prot->ioctl) |
814 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
815 | else | |
816 | err = -ENOIOCTLCMD; | |
1da177e4 LT |
817 | break; |
818 | } | |
819 | return err; | |
820 | } | |
821 | ||
90ddc4f0 | 822 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
823 | .family = PF_INET, |
824 | .owner = THIS_MODULE, | |
825 | .release = inet_release, | |
826 | .bind = inet_bind, | |
827 | .connect = inet_stream_connect, | |
828 | .socketpair = sock_no_socketpair, | |
829 | .accept = inet_accept, | |
830 | .getname = inet_getname, | |
831 | .poll = tcp_poll, | |
832 | .ioctl = inet_ioctl, | |
833 | .listen = inet_listen, | |
834 | .shutdown = inet_shutdown, | |
835 | .setsockopt = sock_common_setsockopt, | |
836 | .getsockopt = sock_common_getsockopt, | |
3516ffb0 | 837 | .sendmsg = tcp_sendmsg, |
543d9cfe ACM |
838 | .recvmsg = sock_common_recvmsg, |
839 | .mmap = sock_no_mmap, | |
840 | .sendpage = tcp_sendpage, | |
3fdadf7d | 841 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
842 | .compat_setsockopt = compat_sock_common_setsockopt, |
843 | .compat_getsockopt = compat_sock_common_getsockopt, | |
3fdadf7d | 844 | #endif |
1da177e4 LT |
845 | }; |
846 | ||
90ddc4f0 | 847 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
848 | .family = PF_INET, |
849 | .owner = THIS_MODULE, | |
850 | .release = inet_release, | |
851 | .bind = inet_bind, | |
852 | .connect = inet_dgram_connect, | |
853 | .socketpair = sock_no_socketpair, | |
854 | .accept = sock_no_accept, | |
855 | .getname = inet_getname, | |
856 | .poll = udp_poll, | |
857 | .ioctl = inet_ioctl, | |
858 | .listen = sock_no_listen, | |
859 | .shutdown = inet_shutdown, | |
860 | .setsockopt = sock_common_setsockopt, | |
861 | .getsockopt = sock_common_getsockopt, | |
862 | .sendmsg = inet_sendmsg, | |
863 | .recvmsg = sock_common_recvmsg, | |
864 | .mmap = sock_no_mmap, | |
865 | .sendpage = inet_sendpage, | |
3fdadf7d | 866 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
867 | .compat_setsockopt = compat_sock_common_setsockopt, |
868 | .compat_getsockopt = compat_sock_common_getsockopt, | |
3fdadf7d | 869 | #endif |
1da177e4 LT |
870 | }; |
871 | ||
872 | /* | |
873 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
874 | * udp_poll | |
875 | */ | |
90ddc4f0 | 876 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
877 | .family = PF_INET, |
878 | .owner = THIS_MODULE, | |
879 | .release = inet_release, | |
880 | .bind = inet_bind, | |
881 | .connect = inet_dgram_connect, | |
882 | .socketpair = sock_no_socketpair, | |
883 | .accept = sock_no_accept, | |
884 | .getname = inet_getname, | |
885 | .poll = datagram_poll, | |
886 | .ioctl = inet_ioctl, | |
887 | .listen = sock_no_listen, | |
888 | .shutdown = inet_shutdown, | |
889 | .setsockopt = sock_common_setsockopt, | |
890 | .getsockopt = sock_common_getsockopt, | |
891 | .sendmsg = inet_sendmsg, | |
892 | .recvmsg = sock_common_recvmsg, | |
893 | .mmap = sock_no_mmap, | |
894 | .sendpage = inet_sendpage, | |
3fdadf7d | 895 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
896 | .compat_setsockopt = compat_sock_common_setsockopt, |
897 | .compat_getsockopt = compat_sock_common_getsockopt, | |
3fdadf7d | 898 | #endif |
1da177e4 LT |
899 | }; |
900 | ||
901 | static struct net_proto_family inet_family_ops = { | |
902 | .family = PF_INET, | |
903 | .create = inet_create, | |
904 | .owner = THIS_MODULE, | |
905 | }; | |
906 | ||
1da177e4 LT |
907 | /* Upon startup we insert all the elements in inetsw_array[] into |
908 | * the linked list inetsw. | |
909 | */ | |
910 | static struct inet_protosw inetsw_array[] = | |
911 | { | |
e905a9ed YH |
912 | { |
913 | .type = SOCK_STREAM, | |
914 | .protocol = IPPROTO_TCP, | |
915 | .prot = &tcp_prot, | |
916 | .ops = &inet_stream_ops, | |
917 | .capability = -1, | |
918 | .no_check = 0, | |
919 | .flags = INET_PROTOSW_PERMANENT | | |
d83d8461 | 920 | INET_PROTOSW_ICSK, |
e905a9ed YH |
921 | }, |
922 | ||
923 | { | |
924 | .type = SOCK_DGRAM, | |
925 | .protocol = IPPROTO_UDP, | |
926 | .prot = &udp_prot, | |
927 | .ops = &inet_dgram_ops, | |
928 | .capability = -1, | |
929 | .no_check = UDP_CSUM_DEFAULT, | |
930 | .flags = INET_PROTOSW_PERMANENT, | |
1da177e4 | 931 | }, |
e905a9ed | 932 | |
1da177e4 LT |
933 | |
934 | { | |
e905a9ed YH |
935 | .type = SOCK_RAW, |
936 | .protocol = IPPROTO_IP, /* wild card */ | |
937 | .prot = &raw_prot, | |
938 | .ops = &inet_sockraw_ops, | |
939 | .capability = CAP_NET_RAW, | |
940 | .no_check = UDP_CSUM_DEFAULT, | |
941 | .flags = INET_PROTOSW_REUSE, | |
1da177e4 LT |
942 | } |
943 | }; | |
944 | ||
c40f6fff | 945 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
946 | |
947 | void inet_register_protosw(struct inet_protosw *p) | |
948 | { | |
949 | struct list_head *lh; | |
950 | struct inet_protosw *answer; | |
951 | int protocol = p->protocol; | |
952 | struct list_head *last_perm; | |
953 | ||
954 | spin_lock_bh(&inetsw_lock); | |
955 | ||
956 | if (p->type >= SOCK_MAX) | |
957 | goto out_illegal; | |
958 | ||
959 | /* If we are trying to override a permanent protocol, bail. */ | |
960 | answer = NULL; | |
961 | last_perm = &inetsw[p->type]; | |
962 | list_for_each(lh, &inetsw[p->type]) { | |
963 | answer = list_entry(lh, struct inet_protosw, list); | |
964 | ||
965 | /* Check only the non-wild match. */ | |
966 | if (INET_PROTOSW_PERMANENT & answer->flags) { | |
967 | if (protocol == answer->protocol) | |
968 | break; | |
969 | last_perm = lh; | |
970 | } | |
971 | ||
972 | answer = NULL; | |
973 | } | |
974 | if (answer) | |
975 | goto out_permanent; | |
976 | ||
977 | /* Add the new entry after the last permanent entry if any, so that | |
978 | * the new entry does not override a permanent entry when matched with | |
979 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 980 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
981 | * system automatically returns to the old behavior. |
982 | */ | |
983 | list_add_rcu(&p->list, last_perm); | |
984 | out: | |
985 | spin_unlock_bh(&inetsw_lock); | |
986 | ||
987 | synchronize_net(); | |
988 | ||
989 | return; | |
990 | ||
991 | out_permanent: | |
992 | printk(KERN_ERR "Attempt to override permanent protocol %d.\n", | |
993 | protocol); | |
994 | goto out; | |
995 | ||
996 | out_illegal: | |
997 | printk(KERN_ERR | |
998 | "Ignoring attempt to register invalid socket type %d.\n", | |
999 | p->type); | |
1000 | goto out; | |
1001 | } | |
1002 | ||
1003 | void inet_unregister_protosw(struct inet_protosw *p) | |
1004 | { | |
1005 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
1006 | printk(KERN_ERR | |
1007 | "Attempt to unregister permanent protocol %d.\n", | |
1008 | p->protocol); | |
1009 | } else { | |
1010 | spin_lock_bh(&inetsw_lock); | |
1011 | list_del_rcu(&p->list); | |
1012 | spin_unlock_bh(&inetsw_lock); | |
1013 | ||
1014 | synchronize_net(); | |
1015 | } | |
1016 | } | |
1017 | ||
32519f11 ACM |
1018 | /* |
1019 | * Shall we try to damage output packets if routing dev changes? | |
1020 | */ | |
1021 | ||
ab32ea5d | 1022 | int sysctl_ip_dynaddr __read_mostly; |
32519f11 ACM |
1023 | |
1024 | static int inet_sk_reselect_saddr(struct sock *sk) | |
1025 | { | |
1026 | struct inet_sock *inet = inet_sk(sk); | |
1027 | int err; | |
1028 | struct rtable *rt; | |
714e85be AV |
1029 | __be32 old_saddr = inet->saddr; |
1030 | __be32 new_saddr; | |
bada8adc | 1031 | __be32 daddr = inet->daddr; |
32519f11 ACM |
1032 | |
1033 | if (inet->opt && inet->opt->srr) | |
1034 | daddr = inet->opt->faddr; | |
1035 | ||
1036 | /* Query new route. */ | |
1037 | err = ip_route_connect(&rt, daddr, 0, | |
1038 | RT_CONN_FLAGS(sk), | |
1039 | sk->sk_bound_dev_if, | |
1040 | sk->sk_protocol, | |
8eb9086f | 1041 | inet->sport, inet->dport, sk, 0); |
32519f11 ACM |
1042 | if (err) |
1043 | return err; | |
1044 | ||
1045 | sk_setup_caps(sk, &rt->u.dst); | |
1046 | ||
1047 | new_saddr = rt->rt_src; | |
1048 | ||
1049 | if (new_saddr == old_saddr) | |
1050 | return 0; | |
1051 | ||
1052 | if (sysctl_ip_dynaddr > 1) { | |
1053 | printk(KERN_INFO "%s(): shifting inet->" | |
1054 | "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n", | |
1055 | __FUNCTION__, | |
1056 | NIPQUAD(old_saddr), | |
1057 | NIPQUAD(new_saddr)); | |
1058 | } | |
1059 | ||
1060 | inet->saddr = inet->rcv_saddr = new_saddr; | |
1061 | ||
1062 | /* | |
1063 | * XXX The only one ugly spot where we need to | |
1064 | * XXX really change the sockets identity after | |
1065 | * XXX it has entered the hashes. -DaveM | |
1066 | * | |
1067 | * Besides that, it does not check for connection | |
1068 | * uniqueness. Wait for troubles. | |
1069 | */ | |
1070 | __sk_prot_rehash(sk); | |
1071 | return 0; | |
1072 | } | |
1073 | ||
1074 | int inet_sk_rebuild_header(struct sock *sk) | |
1075 | { | |
1076 | struct inet_sock *inet = inet_sk(sk); | |
1077 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1078 | __be32 daddr; |
32519f11 ACM |
1079 | int err; |
1080 | ||
1081 | /* Route is OK, nothing to do. */ | |
1082 | if (rt) | |
1083 | return 0; | |
1084 | ||
1085 | /* Reroute. */ | |
1086 | daddr = inet->daddr; | |
1087 | if (inet->opt && inet->opt->srr) | |
1088 | daddr = inet->opt->faddr; | |
1089 | { | |
1090 | struct flowi fl = { | |
1091 | .oif = sk->sk_bound_dev_if, | |
1092 | .nl_u = { | |
1093 | .ip4_u = { | |
1094 | .daddr = daddr, | |
1095 | .saddr = inet->saddr, | |
1096 | .tos = RT_CONN_FLAGS(sk), | |
1097 | }, | |
1098 | }, | |
1099 | .proto = sk->sk_protocol, | |
1100 | .uli_u = { | |
1101 | .ports = { | |
1102 | .sport = inet->sport, | |
1103 | .dport = inet->dport, | |
1104 | }, | |
1105 | }, | |
1106 | }; | |
e905a9ed | 1107 | |
beb8d13b | 1108 | security_sk_classify_flow(sk, &fl); |
32519f11 ACM |
1109 | err = ip_route_output_flow(&rt, &fl, sk, 0); |
1110 | } | |
1111 | if (!err) | |
1112 | sk_setup_caps(sk, &rt->u.dst); | |
1113 | else { | |
1114 | /* Routing failed... */ | |
1115 | sk->sk_route_caps = 0; | |
1116 | /* | |
1117 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1118 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1119 | */ | |
1120 | if (!sysctl_ip_dynaddr || | |
1121 | sk->sk_state != TCP_SYN_SENT || | |
1122 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1123 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
1124 | sk->sk_err_soft = -err; | |
1125 | } | |
1126 | ||
1127 | return err; | |
1128 | } | |
1129 | ||
1130 | EXPORT_SYMBOL(inet_sk_rebuild_header); | |
1131 | ||
a430a43d HX |
1132 | static int inet_gso_send_check(struct sk_buff *skb) |
1133 | { | |
1134 | struct iphdr *iph; | |
1135 | struct net_protocol *ops; | |
1136 | int proto; | |
1137 | int ihl; | |
1138 | int err = -EINVAL; | |
1139 | ||
1140 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) | |
1141 | goto out; | |
1142 | ||
eddc9ec5 | 1143 | iph = ip_hdr(skb); |
a430a43d HX |
1144 | ihl = iph->ihl * 4; |
1145 | if (ihl < sizeof(*iph)) | |
1146 | goto out; | |
1147 | ||
1148 | if (unlikely(!pskb_may_pull(skb, ihl))) | |
1149 | goto out; | |
1150 | ||
badff6d0 ACM |
1151 | __skb_pull(skb, ihl); |
1152 | skb_reset_transport_header(skb); | |
eddc9ec5 | 1153 | iph = ip_hdr(skb); |
a430a43d HX |
1154 | proto = iph->protocol & (MAX_INET_PROTOS - 1); |
1155 | err = -EPROTONOSUPPORT; | |
1156 | ||
1157 | rcu_read_lock(); | |
1158 | ops = rcu_dereference(inet_protos[proto]); | |
1159 | if (likely(ops && ops->gso_send_check)) | |
1160 | err = ops->gso_send_check(skb); | |
1161 | rcu_read_unlock(); | |
1162 | ||
1163 | out: | |
1164 | return err; | |
1165 | } | |
1166 | ||
576a30eb | 1167 | static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features) |
f4c50d99 HX |
1168 | { |
1169 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
1170 | struct iphdr *iph; | |
1171 | struct net_protocol *ops; | |
1172 | int proto; | |
1173 | int ihl; | |
1174 | int id; | |
1175 | ||
d212f87b SH |
1176 | if (!(features & NETIF_F_V4_CSUM)) |
1177 | features &= ~NETIF_F_SG; | |
1178 | ||
bbcf467d HX |
1179 | if (unlikely(skb_shinfo(skb)->gso_type & |
1180 | ~(SKB_GSO_TCPV4 | | |
1181 | SKB_GSO_UDP | | |
1182 | SKB_GSO_DODGY | | |
1183 | SKB_GSO_TCP_ECN | | |
1184 | 0))) | |
1185 | goto out; | |
1186 | ||
1187 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) | |
f4c50d99 HX |
1188 | goto out; |
1189 | ||
eddc9ec5 | 1190 | iph = ip_hdr(skb); |
f4c50d99 HX |
1191 | ihl = iph->ihl * 4; |
1192 | if (ihl < sizeof(*iph)) | |
1193 | goto out; | |
1194 | ||
bbcf467d | 1195 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 HX |
1196 | goto out; |
1197 | ||
badff6d0 ACM |
1198 | __skb_pull(skb, ihl); |
1199 | skb_reset_transport_header(skb); | |
eddc9ec5 | 1200 | iph = ip_hdr(skb); |
f4c50d99 HX |
1201 | id = ntohs(iph->id); |
1202 | proto = iph->protocol & (MAX_INET_PROTOS - 1); | |
1203 | segs = ERR_PTR(-EPROTONOSUPPORT); | |
1204 | ||
1205 | rcu_read_lock(); | |
1206 | ops = rcu_dereference(inet_protos[proto]); | |
bbcf467d | 1207 | if (likely(ops && ops->gso_segment)) |
576a30eb | 1208 | segs = ops->gso_segment(skb, features); |
f4c50d99 HX |
1209 | rcu_read_unlock(); |
1210 | ||
576a30eb | 1211 | if (!segs || unlikely(IS_ERR(segs))) |
f4c50d99 HX |
1212 | goto out; |
1213 | ||
1214 | skb = segs; | |
1215 | do { | |
eddc9ec5 | 1216 | iph = ip_hdr(skb); |
f4c50d99 HX |
1217 | iph->id = htons(id++); |
1218 | iph->tot_len = htons(skb->len - skb->mac_len); | |
1219 | iph->check = 0; | |
d56f90a7 | 1220 | iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl); |
f4c50d99 HX |
1221 | } while ((skb = skb->next)); |
1222 | ||
1223 | out: | |
1224 | return segs; | |
1225 | } | |
1226 | ||
5e0f0435 HX |
1227 | unsigned long snmp_fold_field(void *mib[], int offt) |
1228 | { | |
1229 | unsigned long res = 0; | |
1230 | int i; | |
1231 | ||
1232 | for_each_possible_cpu(i) { | |
1233 | res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt); | |
1234 | res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt); | |
1235 | } | |
1236 | return res; | |
1237 | } | |
1238 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1239 | ||
1240 | int snmp_mib_init(void *ptr[2], size_t mibsize, size_t mibalign) | |
1241 | { | |
1242 | BUG_ON(ptr == NULL); | |
1243 | ptr[0] = __alloc_percpu(mibsize); | |
1244 | if (!ptr[0]) | |
1245 | goto err0; | |
1246 | ptr[1] = __alloc_percpu(mibsize); | |
1247 | if (!ptr[1]) | |
1248 | goto err1; | |
1249 | return 0; | |
1250 | err1: | |
1251 | free_percpu(ptr[0]); | |
1252 | ptr[0] = NULL; | |
1253 | err0: | |
1254 | return -ENOMEM; | |
1255 | } | |
1256 | EXPORT_SYMBOL_GPL(snmp_mib_init); | |
1257 | ||
1258 | void snmp_mib_free(void *ptr[2]) | |
1259 | { | |
1260 | BUG_ON(ptr == NULL); | |
1261 | free_percpu(ptr[0]); | |
1262 | free_percpu(ptr[1]); | |
1263 | ptr[0] = ptr[1] = NULL; | |
1264 | } | |
1265 | EXPORT_SYMBOL_GPL(snmp_mib_free); | |
1266 | ||
1da177e4 LT |
1267 | #ifdef CONFIG_IP_MULTICAST |
1268 | static struct net_protocol igmp_protocol = { | |
1269 | .handler = igmp_rcv, | |
1270 | }; | |
1271 | #endif | |
1272 | ||
1273 | static struct net_protocol tcp_protocol = { | |
1274 | .handler = tcp_v4_rcv, | |
1275 | .err_handler = tcp_v4_err, | |
a430a43d | 1276 | .gso_send_check = tcp_v4_gso_send_check, |
f4c50d99 | 1277 | .gso_segment = tcp_tso_segment, |
1da177e4 LT |
1278 | .no_policy = 1, |
1279 | }; | |
1280 | ||
1281 | static struct net_protocol udp_protocol = { | |
1282 | .handler = udp_rcv, | |
1283 | .err_handler = udp_err, | |
1284 | .no_policy = 1, | |
1285 | }; | |
1286 | ||
1287 | static struct net_protocol icmp_protocol = { | |
1288 | .handler = icmp_rcv, | |
1289 | }; | |
1290 | ||
1291 | static int __init init_ipv4_mibs(void) | |
1292 | { | |
33490170 YH |
1293 | if (snmp_mib_init((void **)net_statistics, |
1294 | sizeof(struct linux_mib), | |
1295 | __alignof__(struct linux_mib)) < 0) | |
1296 | goto err_net_mib; | |
1297 | if (snmp_mib_init((void **)ip_statistics, | |
1298 | sizeof(struct ipstats_mib), | |
1299 | __alignof__(struct ipstats_mib)) < 0) | |
1300 | goto err_ip_mib; | |
1301 | if (snmp_mib_init((void **)icmp_statistics, | |
1302 | sizeof(struct icmp_mib), | |
1303 | __alignof__(struct icmp_mib)) < 0) | |
1304 | goto err_icmp_mib; | |
96793b48 DS |
1305 | if (snmp_mib_init((void **)icmpmsg_statistics, |
1306 | sizeof(struct icmpmsg_mib), | |
1307 | __alignof__(struct icmpmsg_mib)) < 0) | |
1308 | goto err_icmpmsg_mib; | |
33490170 YH |
1309 | if (snmp_mib_init((void **)tcp_statistics, |
1310 | sizeof(struct tcp_mib), | |
1311 | __alignof__(struct tcp_mib)) < 0) | |
1312 | goto err_tcp_mib; | |
1313 | if (snmp_mib_init((void **)udp_statistics, | |
1314 | sizeof(struct udp_mib), | |
1315 | __alignof__(struct udp_mib)) < 0) | |
1316 | goto err_udp_mib; | |
1317 | if (snmp_mib_init((void **)udplite_statistics, | |
1318 | sizeof(struct udp_mib), | |
1319 | __alignof__(struct udp_mib)) < 0) | |
1320 | goto err_udplite_mib; | |
1321 | ||
1322 | tcp_mib_init(); | |
1da177e4 LT |
1323 | |
1324 | return 0; | |
33490170 YH |
1325 | |
1326 | err_udplite_mib: | |
1327 | snmp_mib_free((void **)udp_statistics); | |
1328 | err_udp_mib: | |
1329 | snmp_mib_free((void **)tcp_statistics); | |
1330 | err_tcp_mib: | |
96793b48 DS |
1331 | snmp_mib_free((void **)icmpmsg_statistics); |
1332 | err_icmpmsg_mib: | |
33490170 YH |
1333 | snmp_mib_free((void **)icmp_statistics); |
1334 | err_icmp_mib: | |
1335 | snmp_mib_free((void **)ip_statistics); | |
1336 | err_ip_mib: | |
1337 | snmp_mib_free((void **)net_statistics); | |
1338 | err_net_mib: | |
1339 | return -ENOMEM; | |
1da177e4 LT |
1340 | } |
1341 | ||
1342 | static int ipv4_proc_init(void); | |
1da177e4 | 1343 | |
30e224d7 HX |
1344 | /* |
1345 | * IP protocol layer initialiser | |
1346 | */ | |
1347 | ||
1348 | static struct packet_type ip_packet_type = { | |
1349 | .type = __constant_htons(ETH_P_IP), | |
1350 | .func = ip_rcv, | |
a430a43d | 1351 | .gso_send_check = inet_gso_send_check, |
f4c50d99 | 1352 | .gso_segment = inet_gso_segment, |
30e224d7 HX |
1353 | }; |
1354 | ||
1da177e4 LT |
1355 | static int __init inet_init(void) |
1356 | { | |
1357 | struct sk_buff *dummy_skb; | |
1358 | struct inet_protosw *q; | |
1359 | struct list_head *r; | |
1360 | int rc = -EINVAL; | |
1361 | ||
ef047f5e | 1362 | BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb)); |
1da177e4 LT |
1363 | |
1364 | rc = proto_register(&tcp_prot, 1); | |
1365 | if (rc) | |
1366 | goto out; | |
1367 | ||
1368 | rc = proto_register(&udp_prot, 1); | |
1369 | if (rc) | |
1370 | goto out_unregister_tcp_proto; | |
1371 | ||
1372 | rc = proto_register(&raw_prot, 1); | |
1373 | if (rc) | |
1374 | goto out_unregister_udp_proto; | |
1375 | ||
1376 | /* | |
e905a9ed | 1377 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1378 | */ |
1379 | ||
e905a9ed | 1380 | (void)sock_register(&inet_family_ops); |
1da177e4 LT |
1381 | |
1382 | /* | |
1383 | * Add all the base protocols. | |
1384 | */ | |
1385 | ||
1386 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
1387 | printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n"); | |
1388 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) | |
1389 | printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n"); | |
1390 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) | |
1391 | printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n"); | |
1392 | #ifdef CONFIG_IP_MULTICAST | |
1393 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
1394 | printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n"); | |
1395 | #endif | |
1396 | ||
1397 | /* Register the socket-side information for inet_create. */ | |
1398 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1399 | INIT_LIST_HEAD(r); | |
1400 | ||
1401 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1402 | inet_register_protosw(q); | |
1403 | ||
1404 | /* | |
1405 | * Set the ARP module up | |
1406 | */ | |
1407 | ||
1408 | arp_init(); | |
1409 | ||
e905a9ed YH |
1410 | /* |
1411 | * Set the IP module up | |
1412 | */ | |
1da177e4 LT |
1413 | |
1414 | ip_init(); | |
1415 | ||
1416 | tcp_v4_init(&inet_family_ops); | |
1417 | ||
1418 | /* Setup TCP slab cache for open requests. */ | |
1419 | tcp_init(); | |
1420 | ||
ba4e58ec GR |
1421 | /* Add UDP-Lite (RFC 3828) */ |
1422 | udplite4_register(); | |
1da177e4 LT |
1423 | |
1424 | /* | |
1425 | * Set the ICMP layer up | |
1426 | */ | |
1427 | ||
1428 | icmp_init(&inet_family_ops); | |
1429 | ||
1430 | /* | |
1431 | * Initialise the multicast router | |
1432 | */ | |
1433 | #if defined(CONFIG_IP_MROUTE) | |
1434 | ip_mr_init(); | |
1435 | #endif | |
1436 | /* | |
1437 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 1438 | */ |
1da177e4 | 1439 | |
132adf54 | 1440 | if (init_ipv4_mibs()) |
1da177e4 | 1441 | printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ; |
e905a9ed | 1442 | |
1da177e4 LT |
1443 | ipv4_proc_init(); |
1444 | ||
1445 | ipfrag_init(); | |
1446 | ||
30e224d7 HX |
1447 | dev_add_pack(&ip_packet_type); |
1448 | ||
1da177e4 LT |
1449 | rc = 0; |
1450 | out: | |
1451 | return rc; | |
1da177e4 LT |
1452 | out_unregister_udp_proto: |
1453 | proto_unregister(&udp_prot); | |
321efff7 OK |
1454 | out_unregister_tcp_proto: |
1455 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
1456 | goto out; |
1457 | } | |
1458 | ||
a536e077 | 1459 | fs_initcall(inet_init); |
1da177e4 LT |
1460 | |
1461 | /* ------------------------------------------------------------------------ */ | |
1462 | ||
1463 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
1464 | static int __init ipv4_proc_init(void) |
1465 | { | |
1466 | int rc = 0; | |
1467 | ||
1468 | if (raw_proc_init()) | |
1469 | goto out_raw; | |
1470 | if (tcp4_proc_init()) | |
1471 | goto out_tcp; | |
1472 | if (udp4_proc_init()) | |
1473 | goto out_udp; | |
1474 | if (fib_proc_init()) | |
1475 | goto out_fib; | |
1476 | if (ip_misc_proc_init()) | |
1477 | goto out_misc; | |
1478 | out: | |
1479 | return rc; | |
1480 | out_misc: | |
1481 | fib_proc_exit(); | |
1482 | out_fib: | |
1483 | udp4_proc_exit(); | |
1484 | out_udp: | |
1485 | tcp4_proc_exit(); | |
1486 | out_tcp: | |
1487 | raw_proc_exit(); | |
1488 | out_raw: | |
1489 | rc = -ENOMEM; | |
1490 | goto out; | |
1491 | } | |
1492 | ||
1493 | #else /* CONFIG_PROC_FS */ | |
1494 | static int __init ipv4_proc_init(void) | |
1495 | { | |
1496 | return 0; | |
1497 | } | |
1498 | #endif /* CONFIG_PROC_FS */ | |
1499 | ||
1500 | MODULE_ALIAS_NETPROTO(PF_INET); | |
1501 | ||
1502 | EXPORT_SYMBOL(inet_accept); | |
1503 | EXPORT_SYMBOL(inet_bind); | |
1504 | EXPORT_SYMBOL(inet_dgram_connect); | |
1505 | EXPORT_SYMBOL(inet_dgram_ops); | |
1506 | EXPORT_SYMBOL(inet_getname); | |
1507 | EXPORT_SYMBOL(inet_ioctl); | |
1508 | EXPORT_SYMBOL(inet_listen); | |
1509 | EXPORT_SYMBOL(inet_register_protosw); | |
1510 | EXPORT_SYMBOL(inet_release); | |
1511 | EXPORT_SYMBOL(inet_sendmsg); | |
1512 | EXPORT_SYMBOL(inet_shutdown); | |
1513 | EXPORT_SYMBOL(inet_sock_destruct); | |
1514 | EXPORT_SYMBOL(inet_stream_connect); | |
1515 | EXPORT_SYMBOL(inet_stream_ops); | |
1516 | EXPORT_SYMBOL(inet_unregister_protosw); | |
1517 | EXPORT_SYMBOL(net_statistics); | |
cdac4e07 | 1518 | EXPORT_SYMBOL(sysctl_ip_nonlocal_bind); |