Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * PF_INET6 socket protocol family | |
3 | * Linux INET6 implementation | |
4 | * | |
5 | * Authors: | |
6 | * Pedro Roque <roque@di.fc.ul.pt> | |
7 | * | |
8 | * Adapted from linux/net/ipv4/af_inet.c | |
9 | * | |
10 | * $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $ | |
11 | * | |
12 | * Fixes: | |
13 | * piggy, Karl Knutson : Socket protocol table | |
14 | * Hideaki YOSHIFUJI : sin6_scope_id support | |
15 | * Arnaldo Melo : check proc_net_create return, cleanups | |
16 | * | |
17 | * This program is free software; you can redistribute it and/or | |
18 | * modify it under the terms of the GNU General Public License | |
19 | * as published by the Free Software Foundation; either version | |
20 | * 2 of the License, or (at your option) any later version. | |
21 | */ | |
22 | ||
23 | ||
24 | #include <linux/module.h> | |
25 | #include <linux/config.h> | |
26 | #include <linux/errno.h> | |
27 | #include <linux/types.h> | |
28 | #include <linux/socket.h> | |
29 | #include <linux/in.h> | |
30 | #include <linux/kernel.h> | |
1da177e4 LT |
31 | #include <linux/sched.h> |
32 | #include <linux/timer.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/sockios.h> | |
35 | #include <linux/net.h> | |
36 | #include <linux/fcntl.h> | |
37 | #include <linux/mm.h> | |
38 | #include <linux/interrupt.h> | |
39 | #include <linux/proc_fs.h> | |
40 | #include <linux/stat.h> | |
41 | #include <linux/init.h> | |
42 | ||
43 | #include <linux/inet.h> | |
44 | #include <linux/netdevice.h> | |
45 | #include <linux/icmpv6.h> | |
46 | #include <linux/smp_lock.h> | |
2cc7d573 | 47 | #include <linux/netfilter_ipv6.h> |
1da177e4 LT |
48 | |
49 | #include <net/ip.h> | |
50 | #include <net/ipv6.h> | |
51 | #include <net/udp.h> | |
52 | #include <net/tcp.h> | |
53 | #include <net/ipip.h> | |
54 | #include <net/protocol.h> | |
55 | #include <net/inet_common.h> | |
56 | #include <net/transp_v6.h> | |
57 | #include <net/ip6_route.h> | |
58 | #include <net/addrconf.h> | |
59 | #ifdef CONFIG_IPV6_TUNNEL | |
60 | #include <net/ip6_tunnel.h> | |
61 | #endif | |
62 | ||
63 | #include <asm/uaccess.h> | |
64 | #include <asm/system.h> | |
65 | ||
66 | MODULE_AUTHOR("Cast of dozens"); | |
67 | MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); | |
68 | MODULE_LICENSE("GPL"); | |
69 | ||
1da177e4 LT |
70 | int sysctl_ipv6_bindv6only; |
71 | ||
1da177e4 LT |
72 | /* The inetsw table contains everything that inet_create needs to |
73 | * build a new socket. | |
74 | */ | |
75 | static struct list_head inetsw6[SOCK_MAX]; | |
76 | static DEFINE_SPINLOCK(inetsw6_lock); | |
77 | ||
1da177e4 LT |
78 | static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) |
79 | { | |
80 | const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); | |
81 | ||
82 | return (struct ipv6_pinfo *)(((u8 *)sk) + offset); | |
83 | } | |
84 | ||
85 | static int inet6_create(struct socket *sock, int protocol) | |
86 | { | |
87 | struct inet_sock *inet; | |
88 | struct ipv6_pinfo *np; | |
89 | struct sock *sk; | |
90 | struct list_head *p; | |
91 | struct inet_protosw *answer; | |
92 | struct proto *answer_prot; | |
93 | unsigned char answer_flags; | |
94 | char answer_no_check; | |
af1afe86 YH |
95 | int try_loading_module = 0; |
96 | int err; | |
1da177e4 LT |
97 | |
98 | /* Look for the requested type/protocol pair. */ | |
99 | answer = NULL; | |
af1afe86 YH |
100 | lookup_protocol: |
101 | err = -ESOCKTNOSUPPORT; | |
1da177e4 LT |
102 | rcu_read_lock(); |
103 | list_for_each_rcu(p, &inetsw6[sock->type]) { | |
104 | answer = list_entry(p, struct inet_protosw, list); | |
105 | ||
106 | /* Check the non-wild match. */ | |
107 | if (protocol == answer->protocol) { | |
108 | if (protocol != IPPROTO_IP) | |
109 | break; | |
110 | } else { | |
111 | /* Check for the two wild cases. */ | |
112 | if (IPPROTO_IP == protocol) { | |
113 | protocol = answer->protocol; | |
114 | break; | |
115 | } | |
116 | if (IPPROTO_IP == answer->protocol) | |
117 | break; | |
118 | } | |
af1afe86 | 119 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
120 | answer = NULL; |
121 | } | |
122 | ||
af1afe86 YH |
123 | if (!answer) { |
124 | if (try_loading_module < 2) { | |
125 | rcu_read_unlock(); | |
126 | /* | |
127 | * Be more specific, e.g. net-pf-10-proto-132-type-1 | |
128 | * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
129 | */ | |
130 | if (++try_loading_module == 1) | |
131 | request_module("net-pf-%d-proto-%d-type-%d", | |
132 | PF_INET6, protocol, sock->type); | |
133 | /* | |
134 | * Fall back to generic, e.g. net-pf-10-proto-132 | |
135 | * (net-pf-PF_INET6-proto-IPPROTO_SCTP) | |
136 | */ | |
137 | else | |
138 | request_module("net-pf-%d-proto-%d", | |
139 | PF_INET6, protocol); | |
140 | goto lookup_protocol; | |
141 | } else | |
142 | goto out_rcu_unlock; | |
143 | } | |
144 | ||
145 | err = -EPERM; | |
1da177e4 LT |
146 | if (answer->capability > 0 && !capable(answer->capability)) |
147 | goto out_rcu_unlock; | |
1da177e4 LT |
148 | |
149 | sock->ops = answer->ops; | |
1da177e4 LT |
150 | answer_prot = answer->prot; |
151 | answer_no_check = answer->no_check; | |
152 | answer_flags = answer->flags; | |
153 | rcu_read_unlock(); | |
154 | ||
155 | BUG_TRAP(answer_prot->slab != NULL); | |
156 | ||
af1afe86 | 157 | err = -ENOBUFS; |
1da177e4 LT |
158 | sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1); |
159 | if (sk == NULL) | |
160 | goto out; | |
161 | ||
162 | sock_init_data(sock, sk); | |
163 | ||
af1afe86 | 164 | err = 0; |
1da177e4 LT |
165 | sk->sk_no_check = answer_no_check; |
166 | if (INET_PROTOSW_REUSE & answer_flags) | |
167 | sk->sk_reuse = 1; | |
168 | ||
169 | inet = inet_sk(sk); | |
170 | ||
171 | if (SOCK_RAW == sock->type) { | |
172 | inet->num = protocol; | |
173 | if (IPPROTO_RAW == protocol) | |
174 | inet->hdrincl = 1; | |
175 | } | |
176 | ||
e6848976 | 177 | sk->sk_destruct = inet_sock_destruct; |
1da177e4 LT |
178 | sk->sk_family = PF_INET6; |
179 | sk->sk_protocol = protocol; | |
180 | ||
181 | sk->sk_backlog_rcv = answer->prot->backlog_rcv; | |
182 | ||
183 | inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); | |
184 | np->hop_limit = -1; | |
185 | np->mcast_hops = -1; | |
186 | np->mc_loop = 1; | |
187 | np->pmtudisc = IPV6_PMTUDISC_WANT; | |
188 | np->ipv6only = sysctl_ipv6_bindv6only; | |
189 | ||
190 | /* Init the ipv4 part of the socket since we can have sockets | |
191 | * using v6 API for ipv4. | |
192 | */ | |
193 | inet->uc_ttl = -1; | |
194 | ||
195 | inet->mc_loop = 1; | |
196 | inet->mc_ttl = 1; | |
197 | inet->mc_index = 0; | |
198 | inet->mc_list = NULL; | |
199 | ||
200 | if (ipv4_config.no_pmtu_disc) | |
201 | inet->pmtudisc = IP_PMTUDISC_DONT; | |
202 | else | |
203 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
e6848976 ACM |
204 | /* |
205 | * Increment only the relevant sk_prot->socks debug field, this changes | |
206 | * the previous behaviour of incrementing both the equivalent to | |
207 | * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. | |
208 | * | |
209 | * This allows better debug granularity as we'll know exactly how many | |
210 | * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 | |
211 | * transport protocol socks. -acme | |
212 | */ | |
213 | sk_refcnt_debug_inc(sk); | |
1da177e4 | 214 | |
1da177e4 LT |
215 | if (inet->num) { |
216 | /* It assumes that any protocol which allows | |
217 | * the user to assign a number at socket | |
218 | * creation time automatically shares. | |
219 | */ | |
220 | inet->sport = ntohs(inet->num); | |
221 | sk->sk_prot->hash(sk); | |
222 | } | |
223 | if (sk->sk_prot->init) { | |
af1afe86 YH |
224 | err = sk->sk_prot->init(sk); |
225 | if (err) { | |
1da177e4 LT |
226 | sk_common_release(sk); |
227 | goto out; | |
228 | } | |
229 | } | |
230 | out: | |
af1afe86 | 231 | return err; |
1da177e4 LT |
232 | out_rcu_unlock: |
233 | rcu_read_unlock(); | |
234 | goto out; | |
235 | } | |
236 | ||
237 | ||
238 | /* bind for INET6 API */ | |
239 | int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
240 | { | |
241 | struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr; | |
242 | struct sock *sk = sock->sk; | |
243 | struct inet_sock *inet = inet_sk(sk); | |
244 | struct ipv6_pinfo *np = inet6_sk(sk); | |
245 | __u32 v4addr = 0; | |
246 | unsigned short snum; | |
247 | int addr_type = 0; | |
248 | int err = 0; | |
249 | ||
250 | /* If the socket has its own bind function then use it. */ | |
251 | if (sk->sk_prot->bind) | |
252 | return sk->sk_prot->bind(sk, uaddr, addr_len); | |
253 | ||
254 | if (addr_len < SIN6_LEN_RFC2133) | |
255 | return -EINVAL; | |
256 | addr_type = ipv6_addr_type(&addr->sin6_addr); | |
257 | if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM) | |
258 | return -EINVAL; | |
259 | ||
260 | snum = ntohs(addr->sin6_port); | |
261 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) | |
262 | return -EACCES; | |
263 | ||
264 | lock_sock(sk); | |
265 | ||
266 | /* Check these errors (active socket, double bind). */ | |
267 | if (sk->sk_state != TCP_CLOSE || inet->num) { | |
268 | err = -EINVAL; | |
269 | goto out; | |
270 | } | |
271 | ||
272 | /* Check if the address belongs to the host. */ | |
273 | if (addr_type == IPV6_ADDR_MAPPED) { | |
274 | v4addr = addr->sin6_addr.s6_addr32[3]; | |
275 | if (inet_addr_type(v4addr) != RTN_LOCAL) { | |
276 | err = -EADDRNOTAVAIL; | |
277 | goto out; | |
278 | } | |
279 | } else { | |
280 | if (addr_type != IPV6_ADDR_ANY) { | |
281 | struct net_device *dev = NULL; | |
282 | ||
283 | if (addr_type & IPV6_ADDR_LINKLOCAL) { | |
284 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
285 | addr->sin6_scope_id) { | |
286 | /* Override any existing binding, if another one | |
287 | * is supplied by user. | |
288 | */ | |
289 | sk->sk_bound_dev_if = addr->sin6_scope_id; | |
290 | } | |
291 | ||
292 | /* Binding to link-local address requires an interface */ | |
293 | if (!sk->sk_bound_dev_if) { | |
294 | err = -EINVAL; | |
295 | goto out; | |
296 | } | |
297 | dev = dev_get_by_index(sk->sk_bound_dev_if); | |
298 | if (!dev) { | |
299 | err = -ENODEV; | |
300 | goto out; | |
301 | } | |
302 | } | |
303 | ||
304 | /* ipv4 addr of the socket is invalid. Only the | |
305 | * unspecified and mapped address have a v4 equivalent. | |
306 | */ | |
307 | v4addr = LOOPBACK4_IPV6; | |
308 | if (!(addr_type & IPV6_ADDR_MULTICAST)) { | |
309 | if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) { | |
310 | if (dev) | |
311 | dev_put(dev); | |
312 | err = -EADDRNOTAVAIL; | |
313 | goto out; | |
314 | } | |
315 | } | |
316 | if (dev) | |
317 | dev_put(dev); | |
318 | } | |
319 | } | |
320 | ||
321 | inet->rcv_saddr = v4addr; | |
322 | inet->saddr = v4addr; | |
323 | ||
324 | ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); | |
325 | ||
326 | if (!(addr_type & IPV6_ADDR_MULTICAST)) | |
327 | ipv6_addr_copy(&np->saddr, &addr->sin6_addr); | |
328 | ||
329 | /* Make sure we are allowed to bind here. */ | |
330 | if (sk->sk_prot->get_port(sk, snum)) { | |
331 | inet_reset_saddr(sk); | |
332 | err = -EADDRINUSE; | |
333 | goto out; | |
334 | } | |
335 | ||
336 | if (addr_type != IPV6_ADDR_ANY) | |
337 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; | |
338 | if (snum) | |
339 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
340 | inet->sport = ntohs(inet->num); | |
341 | inet->dport = 0; | |
342 | inet->daddr = 0; | |
343 | out: | |
344 | release_sock(sk); | |
345 | return err; | |
346 | } | |
347 | ||
348 | int inet6_release(struct socket *sock) | |
349 | { | |
350 | struct sock *sk = sock->sk; | |
351 | ||
352 | if (sk == NULL) | |
353 | return -EINVAL; | |
354 | ||
355 | /* Free mc lists */ | |
356 | ipv6_sock_mc_close(sk); | |
357 | ||
358 | /* Free ac lists */ | |
359 | ipv6_sock_ac_close(sk); | |
360 | ||
361 | return inet_release(sock); | |
362 | } | |
363 | ||
364 | int inet6_destroy_sock(struct sock *sk) | |
365 | { | |
366 | struct ipv6_pinfo *np = inet6_sk(sk); | |
367 | struct sk_buff *skb; | |
368 | struct ipv6_txoptions *opt; | |
369 | ||
370 | /* | |
371 | * Release destination entry | |
372 | */ | |
373 | ||
374 | sk_dst_reset(sk); | |
375 | ||
376 | /* Release rx options */ | |
377 | ||
378 | if ((skb = xchg(&np->pktoptions, NULL)) != NULL) | |
379 | kfree_skb(skb); | |
380 | ||
381 | /* Free flowlabels */ | |
382 | fl6_free_socklist(sk); | |
383 | ||
384 | /* Free tx options */ | |
385 | ||
386 | if ((opt = xchg(&np->opt, NULL)) != NULL) | |
387 | sock_kfree_s(sk, opt, opt->tot_len); | |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
3cf3dc6c ACM |
392 | EXPORT_SYMBOL_GPL(inet6_destroy_sock); |
393 | ||
1da177e4 LT |
394 | /* |
395 | * This does both peername and sockname. | |
396 | */ | |
397 | ||
398 | int inet6_getname(struct socket *sock, struct sockaddr *uaddr, | |
399 | int *uaddr_len, int peer) | |
400 | { | |
401 | struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr; | |
402 | struct sock *sk = sock->sk; | |
403 | struct inet_sock *inet = inet_sk(sk); | |
404 | struct ipv6_pinfo *np = inet6_sk(sk); | |
405 | ||
406 | sin->sin6_family = AF_INET6; | |
407 | sin->sin6_flowinfo = 0; | |
408 | sin->sin6_scope_id = 0; | |
409 | if (peer) { | |
410 | if (!inet->dport) | |
411 | return -ENOTCONN; | |
412 | if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && | |
413 | peer == 1) | |
414 | return -ENOTCONN; | |
415 | sin->sin6_port = inet->dport; | |
416 | ipv6_addr_copy(&sin->sin6_addr, &np->daddr); | |
417 | if (np->sndflow) | |
418 | sin->sin6_flowinfo = np->flow_label; | |
419 | } else { | |
420 | if (ipv6_addr_any(&np->rcv_saddr)) | |
421 | ipv6_addr_copy(&sin->sin6_addr, &np->saddr); | |
422 | else | |
423 | ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr); | |
424 | ||
425 | sin->sin6_port = inet->sport; | |
426 | } | |
427 | if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL) | |
428 | sin->sin6_scope_id = sk->sk_bound_dev_if; | |
429 | *uaddr_len = sizeof(*sin); | |
430 | return(0); | |
431 | } | |
432 | ||
433 | int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
434 | { | |
435 | struct sock *sk = sock->sk; | |
436 | int err = -EINVAL; | |
437 | ||
438 | switch(cmd) | |
439 | { | |
440 | case SIOCGSTAMP: | |
441 | return sock_get_timestamp(sk, (struct timeval __user *)arg); | |
442 | ||
443 | case SIOCADDRT: | |
444 | case SIOCDELRT: | |
445 | ||
446 | return(ipv6_route_ioctl(cmd,(void __user *)arg)); | |
447 | ||
448 | case SIOCSIFADDR: | |
449 | return addrconf_add_ifaddr((void __user *) arg); | |
450 | case SIOCDIFADDR: | |
451 | return addrconf_del_ifaddr((void __user *) arg); | |
452 | case SIOCSIFDSTADDR: | |
453 | return addrconf_set_dstaddr((void __user *) arg); | |
454 | default: | |
455 | if (!sk->sk_prot->ioctl || | |
456 | (err = sk->sk_prot->ioctl(sk, cmd, arg)) == -ENOIOCTLCMD) | |
457 | return(dev_ioctl(cmd,(void __user *) arg)); | |
458 | return err; | |
459 | } | |
460 | /*NOTREACHED*/ | |
461 | return(0); | |
462 | } | |
463 | ||
464 | struct proto_ops inet6_stream_ops = { | |
465 | .family = PF_INET6, | |
466 | .owner = THIS_MODULE, | |
467 | .release = inet6_release, | |
468 | .bind = inet6_bind, | |
469 | .connect = inet_stream_connect, /* ok */ | |
470 | .socketpair = sock_no_socketpair, /* a do nothing */ | |
471 | .accept = inet_accept, /* ok */ | |
472 | .getname = inet6_getname, | |
473 | .poll = tcp_poll, /* ok */ | |
474 | .ioctl = inet6_ioctl, /* must change */ | |
475 | .listen = inet_listen, /* ok */ | |
476 | .shutdown = inet_shutdown, /* ok */ | |
477 | .setsockopt = sock_common_setsockopt, /* ok */ | |
478 | .getsockopt = sock_common_getsockopt, /* ok */ | |
479 | .sendmsg = inet_sendmsg, /* ok */ | |
480 | .recvmsg = sock_common_recvmsg, /* ok */ | |
481 | .mmap = sock_no_mmap, | |
482 | .sendpage = tcp_sendpage | |
483 | }; | |
484 | ||
485 | struct proto_ops inet6_dgram_ops = { | |
486 | .family = PF_INET6, | |
487 | .owner = THIS_MODULE, | |
488 | .release = inet6_release, | |
489 | .bind = inet6_bind, | |
490 | .connect = inet_dgram_connect, /* ok */ | |
491 | .socketpair = sock_no_socketpair, /* a do nothing */ | |
492 | .accept = sock_no_accept, /* a do nothing */ | |
493 | .getname = inet6_getname, | |
494 | .poll = udp_poll, /* ok */ | |
495 | .ioctl = inet6_ioctl, /* must change */ | |
496 | .listen = sock_no_listen, /* ok */ | |
497 | .shutdown = inet_shutdown, /* ok */ | |
498 | .setsockopt = sock_common_setsockopt, /* ok */ | |
499 | .getsockopt = sock_common_getsockopt, /* ok */ | |
500 | .sendmsg = inet_sendmsg, /* ok */ | |
501 | .recvmsg = sock_common_recvmsg, /* ok */ | |
502 | .mmap = sock_no_mmap, | |
503 | .sendpage = sock_no_sendpage, | |
504 | }; | |
505 | ||
506 | static struct net_proto_family inet6_family_ops = { | |
507 | .family = PF_INET6, | |
508 | .create = inet6_create, | |
509 | .owner = THIS_MODULE, | |
510 | }; | |
511 | ||
1da177e4 LT |
512 | /* Same as inet6_dgram_ops, sans udp_poll. */ |
513 | static struct proto_ops inet6_sockraw_ops = { | |
514 | .family = PF_INET6, | |
515 | .owner = THIS_MODULE, | |
516 | .release = inet6_release, | |
517 | .bind = inet6_bind, | |
518 | .connect = inet_dgram_connect, /* ok */ | |
519 | .socketpair = sock_no_socketpair, /* a do nothing */ | |
520 | .accept = sock_no_accept, /* a do nothing */ | |
521 | .getname = inet6_getname, | |
522 | .poll = datagram_poll, /* ok */ | |
523 | .ioctl = inet6_ioctl, /* must change */ | |
524 | .listen = sock_no_listen, /* ok */ | |
525 | .shutdown = inet_shutdown, /* ok */ | |
526 | .setsockopt = sock_common_setsockopt, /* ok */ | |
527 | .getsockopt = sock_common_getsockopt, /* ok */ | |
528 | .sendmsg = inet_sendmsg, /* ok */ | |
529 | .recvmsg = sock_common_recvmsg, /* ok */ | |
530 | .mmap = sock_no_mmap, | |
531 | .sendpage = sock_no_sendpage, | |
532 | }; | |
533 | ||
534 | static struct inet_protosw rawv6_protosw = { | |
535 | .type = SOCK_RAW, | |
536 | .protocol = IPPROTO_IP, /* wild card */ | |
537 | .prot = &rawv6_prot, | |
538 | .ops = &inet6_sockraw_ops, | |
539 | .capability = CAP_NET_RAW, | |
540 | .no_check = UDP_CSUM_DEFAULT, | |
541 | .flags = INET_PROTOSW_REUSE, | |
542 | }; | |
543 | ||
544 | void | |
545 | inet6_register_protosw(struct inet_protosw *p) | |
546 | { | |
547 | struct list_head *lh; | |
548 | struct inet_protosw *answer; | |
549 | int protocol = p->protocol; | |
550 | struct list_head *last_perm; | |
551 | ||
552 | spin_lock_bh(&inetsw6_lock); | |
553 | ||
554 | if (p->type >= SOCK_MAX) | |
555 | goto out_illegal; | |
556 | ||
557 | /* If we are trying to override a permanent protocol, bail. */ | |
558 | answer = NULL; | |
559 | last_perm = &inetsw6[p->type]; | |
560 | list_for_each(lh, &inetsw6[p->type]) { | |
561 | answer = list_entry(lh, struct inet_protosw, list); | |
562 | ||
563 | /* Check only the non-wild match. */ | |
564 | if (INET_PROTOSW_PERMANENT & answer->flags) { | |
565 | if (protocol == answer->protocol) | |
566 | break; | |
567 | last_perm = lh; | |
568 | } | |
569 | ||
570 | answer = NULL; | |
571 | } | |
572 | if (answer) | |
573 | goto out_permanent; | |
574 | ||
575 | /* Add the new entry after the last permanent entry if any, so that | |
576 | * the new entry does not override a permanent entry when matched with | |
577 | * a wild-card protocol. But it is allowed to override any existing | |
578 | * non-permanent entry. This means that when we remove this entry, the | |
579 | * system automatically returns to the old behavior. | |
580 | */ | |
581 | list_add_rcu(&p->list, last_perm); | |
582 | out: | |
583 | spin_unlock_bh(&inetsw6_lock); | |
584 | return; | |
585 | ||
586 | out_permanent: | |
587 | printk(KERN_ERR "Attempt to override permanent protocol %d.\n", | |
588 | protocol); | |
589 | goto out; | |
590 | ||
591 | out_illegal: | |
592 | printk(KERN_ERR | |
593 | "Ignoring attempt to register invalid socket type %d.\n", | |
594 | p->type); | |
595 | goto out; | |
596 | } | |
597 | ||
598 | void | |
599 | inet6_unregister_protosw(struct inet_protosw *p) | |
600 | { | |
601 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
602 | printk(KERN_ERR | |
603 | "Attempt to unregister permanent protocol %d.\n", | |
604 | p->protocol); | |
605 | } else { | |
606 | spin_lock_bh(&inetsw6_lock); | |
607 | list_del_rcu(&p->list); | |
608 | spin_unlock_bh(&inetsw6_lock); | |
609 | ||
610 | synchronize_net(); | |
611 | } | |
612 | } | |
613 | ||
b9750ce1 ACM |
614 | int inet6_sk_rebuild_header(struct sock *sk) |
615 | { | |
616 | int err; | |
617 | struct dst_entry *dst; | |
618 | struct ipv6_pinfo *np = inet6_sk(sk); | |
619 | ||
620 | dst = __sk_dst_check(sk, np->dst_cookie); | |
621 | ||
622 | if (dst == NULL) { | |
623 | struct inet_sock *inet = inet_sk(sk); | |
624 | struct in6_addr *final_p = NULL, final; | |
625 | struct flowi fl; | |
626 | ||
627 | memset(&fl, 0, sizeof(fl)); | |
628 | fl.proto = sk->sk_protocol; | |
629 | ipv6_addr_copy(&fl.fl6_dst, &np->daddr); | |
630 | ipv6_addr_copy(&fl.fl6_src, &np->saddr); | |
631 | fl.fl6_flowlabel = np->flow_label; | |
632 | fl.oif = sk->sk_bound_dev_if; | |
633 | fl.fl_ip_dport = inet->dport; | |
634 | fl.fl_ip_sport = inet->sport; | |
635 | ||
636 | if (np->opt && np->opt->srcrt) { | |
637 | struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt; | |
638 | ipv6_addr_copy(&final, &fl.fl6_dst); | |
639 | ipv6_addr_copy(&fl.fl6_dst, rt0->addr); | |
640 | final_p = &final; | |
641 | } | |
642 | ||
643 | err = ip6_dst_lookup(sk, &dst, &fl); | |
644 | if (err) { | |
645 | sk->sk_route_caps = 0; | |
646 | return err; | |
647 | } | |
648 | if (final_p) | |
649 | ipv6_addr_copy(&fl.fl6_dst, final_p); | |
650 | ||
651 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) { | |
652 | sk->sk_err_soft = -err; | |
653 | return err; | |
654 | } | |
655 | ||
656 | ip6_dst_store(sk, dst, NULL); | |
657 | sk->sk_route_caps = dst->dev->features & | |
658 | ~(NETIF_F_IP_CSUM | NETIF_F_TSO); | |
659 | } | |
660 | ||
661 | return 0; | |
662 | } | |
663 | ||
664 | EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); | |
665 | ||
399c07de ACM |
666 | int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb) |
667 | { | |
668 | struct ipv6_pinfo *np = inet6_sk(sk); | |
669 | struct inet6_skb_parm *opt = IP6CB(skb); | |
670 | ||
671 | if (np->rxopt.all) { | |
672 | if ((opt->hop && (np->rxopt.bits.hopopts || | |
673 | np->rxopt.bits.ohopopts)) || | |
674 | ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) && | |
675 | np->rxopt.bits.rxflow) || | |
676 | (opt->srcrt && (np->rxopt.bits.srcrt || | |
677 | np->rxopt.bits.osrcrt)) || | |
678 | ((opt->dst1 || opt->dst0) && | |
679 | (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) | |
680 | return 1; | |
681 | } | |
682 | return 0; | |
683 | } | |
684 | ||
685 | EXPORT_SYMBOL_GPL(ipv6_opt_accepted); | |
686 | ||
1da177e4 LT |
687 | int |
688 | snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign) | |
689 | { | |
690 | if (ptr == NULL) | |
691 | return -EINVAL; | |
692 | ||
693 | ptr[0] = __alloc_percpu(mibsize, mibalign); | |
694 | if (!ptr[0]) | |
695 | goto err0; | |
696 | ||
697 | ptr[1] = __alloc_percpu(mibsize, mibalign); | |
698 | if (!ptr[1]) | |
699 | goto err1; | |
700 | ||
701 | return 0; | |
702 | ||
703 | err1: | |
704 | free_percpu(ptr[0]); | |
705 | ptr[0] = NULL; | |
706 | err0: | |
707 | return -ENOMEM; | |
708 | } | |
709 | ||
710 | void | |
711 | snmp6_mib_free(void *ptr[2]) | |
712 | { | |
713 | if (ptr == NULL) | |
714 | return; | |
715 | if (ptr[0]) | |
716 | free_percpu(ptr[0]); | |
717 | if (ptr[1]) | |
718 | free_percpu(ptr[1]); | |
719 | ptr[0] = ptr[1] = NULL; | |
720 | } | |
721 | ||
722 | static int __init init_ipv6_mibs(void) | |
723 | { | |
724 | if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib), | |
725 | __alignof__(struct ipstats_mib)) < 0) | |
726 | goto err_ip_mib; | |
727 | if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib), | |
728 | __alignof__(struct icmpv6_mib)) < 0) | |
729 | goto err_icmp_mib; | |
730 | if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib), | |
731 | __alignof__(struct udp_mib)) < 0) | |
732 | goto err_udp_mib; | |
733 | return 0; | |
734 | ||
735 | err_udp_mib: | |
736 | snmp6_mib_free((void **)icmpv6_statistics); | |
737 | err_icmp_mib: | |
738 | snmp6_mib_free((void **)ipv6_statistics); | |
739 | err_ip_mib: | |
740 | return -ENOMEM; | |
741 | ||
742 | } | |
743 | ||
744 | static void cleanup_ipv6_mibs(void) | |
745 | { | |
746 | snmp6_mib_free((void **)ipv6_statistics); | |
747 | snmp6_mib_free((void **)icmpv6_statistics); | |
748 | snmp6_mib_free((void **)udp_stats_in6); | |
749 | } | |
750 | ||
1da177e4 LT |
751 | static int __init inet6_init(void) |
752 | { | |
753 | struct sk_buff *dummy_skb; | |
754 | struct list_head *r; | |
755 | int err; | |
756 | ||
757 | #ifdef MODULE | |
758 | #if 0 /* FIXME --RR */ | |
759 | if (!mod_member_present(&__this_module, can_unload)) | |
760 | return -EINVAL; | |
761 | ||
762 | __this_module.can_unload = &ipv6_unload; | |
763 | #endif | |
764 | #endif | |
765 | ||
766 | if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) { | |
767 | printk(KERN_CRIT "inet6_proto_init: size fault\n"); | |
768 | return -EINVAL; | |
769 | } | |
770 | ||
771 | err = proto_register(&tcpv6_prot, 1); | |
772 | if (err) | |
773 | goto out; | |
774 | ||
775 | err = proto_register(&udpv6_prot, 1); | |
776 | if (err) | |
777 | goto out_unregister_tcp_proto; | |
778 | ||
779 | err = proto_register(&rawv6_prot, 1); | |
780 | if (err) | |
781 | goto out_unregister_udp_proto; | |
782 | ||
783 | ||
784 | /* Register the socket-side information for inet6_create. */ | |
785 | for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) | |
786 | INIT_LIST_HEAD(r); | |
787 | ||
788 | /* We MUST register RAW sockets before we create the ICMP6, | |
789 | * IGMP6, or NDISC control sockets. | |
790 | */ | |
791 | inet6_register_protosw(&rawv6_protosw); | |
792 | ||
793 | /* Register the family here so that the init calls below will | |
794 | * be able to create sockets. (?? is this dangerous ??) | |
795 | */ | |
8eb55910 DM |
796 | err = sock_register(&inet6_family_ops); |
797 | if (err) | |
798 | goto out_unregister_raw_proto; | |
1da177e4 LT |
799 | |
800 | /* Initialise ipv6 mibs */ | |
801 | err = init_ipv6_mibs(); | |
802 | if (err) | |
8eb55910 | 803 | goto out_unregister_sock; |
1da177e4 LT |
804 | |
805 | /* | |
806 | * ipngwg API draft makes clear that the correct semantics | |
807 | * for TCP and UDP is to consider one TCP and UDP instance | |
808 | * in a host availiable by both INET and INET6 APIs and | |
809 | * able to communicate via both network protocols. | |
810 | */ | |
811 | ||
812 | #ifdef CONFIG_SYSCTL | |
813 | ipv6_sysctl_register(); | |
814 | #endif | |
815 | err = icmpv6_init(&inet6_family_ops); | |
816 | if (err) | |
817 | goto icmp_fail; | |
818 | err = ndisc_init(&inet6_family_ops); | |
819 | if (err) | |
820 | goto ndisc_fail; | |
821 | err = igmp6_init(&inet6_family_ops); | |
822 | if (err) | |
823 | goto igmp_fail; | |
2cc7d573 HW |
824 | err = ipv6_netfilter_init(); |
825 | if (err) | |
826 | goto netfilter_fail; | |
1da177e4 LT |
827 | /* Create /proc/foo6 entries. */ |
828 | #ifdef CONFIG_PROC_FS | |
829 | err = -ENOMEM; | |
830 | if (raw6_proc_init()) | |
831 | goto proc_raw6_fail; | |
832 | if (tcp6_proc_init()) | |
833 | goto proc_tcp6_fail; | |
834 | if (udp6_proc_init()) | |
835 | goto proc_udp6_fail; | |
836 | if (ipv6_misc_proc_init()) | |
837 | goto proc_misc6_fail; | |
838 | ||
839 | if (ac6_proc_init()) | |
840 | goto proc_anycast6_fail; | |
841 | if (if6_proc_init()) | |
842 | goto proc_if6_fail; | |
843 | #endif | |
1da177e4 LT |
844 | ip6_route_init(); |
845 | ip6_flowlabel_init(); | |
846 | err = addrconf_init(); | |
847 | if (err) | |
848 | goto addrconf_fail; | |
849 | sit_init(); | |
850 | ||
851 | /* Init v6 extension headers. */ | |
852 | ipv6_rthdr_init(); | |
853 | ipv6_frag_init(); | |
854 | ipv6_nodata_init(); | |
855 | ipv6_destopt_init(); | |
856 | ||
857 | /* Init v6 transport protocols. */ | |
858 | udpv6_init(); | |
859 | tcpv6_init(); | |
e2ed4052 HX |
860 | |
861 | ipv6_packet_init(); | |
1da177e4 LT |
862 | err = 0; |
863 | out: | |
864 | return err; | |
865 | ||
866 | addrconf_fail: | |
867 | ip6_flowlabel_cleanup(); | |
868 | ip6_route_cleanup(); | |
1da177e4 LT |
869 | #ifdef CONFIG_PROC_FS |
870 | if6_proc_exit(); | |
871 | proc_if6_fail: | |
872 | ac6_proc_exit(); | |
873 | proc_anycast6_fail: | |
874 | ipv6_misc_proc_exit(); | |
875 | proc_misc6_fail: | |
876 | udp6_proc_exit(); | |
877 | proc_udp6_fail: | |
878 | tcp6_proc_exit(); | |
879 | proc_tcp6_fail: | |
880 | raw6_proc_exit(); | |
881 | proc_raw6_fail: | |
882 | #endif | |
2cc7d573 HW |
883 | ipv6_netfilter_fini(); |
884 | netfilter_fail: | |
1da177e4 LT |
885 | igmp6_cleanup(); |
886 | igmp_fail: | |
887 | ndisc_cleanup(); | |
888 | ndisc_fail: | |
889 | icmpv6_cleanup(); | |
890 | icmp_fail: | |
891 | #ifdef CONFIG_SYSCTL | |
892 | ipv6_sysctl_unregister(); | |
893 | #endif | |
894 | cleanup_ipv6_mibs(); | |
8eb55910 DM |
895 | out_unregister_sock: |
896 | sock_unregister(PF_INET6); | |
1da177e4 LT |
897 | out_unregister_raw_proto: |
898 | proto_unregister(&rawv6_prot); | |
899 | out_unregister_udp_proto: | |
900 | proto_unregister(&udpv6_prot); | |
901 | out_unregister_tcp_proto: | |
902 | proto_unregister(&tcpv6_prot); | |
903 | goto out; | |
904 | } | |
905 | module_init(inet6_init); | |
906 | ||
907 | static void __exit inet6_exit(void) | |
908 | { | |
909 | /* First of all disallow new sockets creation. */ | |
910 | sock_unregister(PF_INET6); | |
911 | #ifdef CONFIG_PROC_FS | |
912 | if6_proc_exit(); | |
913 | ac6_proc_exit(); | |
914 | ipv6_misc_proc_exit(); | |
915 | udp6_proc_exit(); | |
916 | tcp6_proc_exit(); | |
917 | raw6_proc_exit(); | |
918 | #endif | |
919 | /* Cleanup code parts. */ | |
920 | sit_cleanup(); | |
921 | ip6_flowlabel_cleanup(); | |
922 | addrconf_cleanup(); | |
923 | ip6_route_cleanup(); | |
924 | ipv6_packet_cleanup(); | |
925 | igmp6_cleanup(); | |
2cc7d573 | 926 | ipv6_netfilter_fini(); |
1da177e4 LT |
927 | ndisc_cleanup(); |
928 | icmpv6_cleanup(); | |
929 | #ifdef CONFIG_SYSCTL | |
930 | ipv6_sysctl_unregister(); | |
931 | #endif | |
932 | cleanup_ipv6_mibs(); | |
933 | proto_unregister(&rawv6_prot); | |
934 | proto_unregister(&udpv6_prot); | |
935 | proto_unregister(&tcpv6_prot); | |
936 | } | |
937 | module_exit(inet6_exit); | |
938 | ||
939 | MODULE_ALIAS_NETPROTO(PF_INET6); |