xref: /linux/net/ipv6/af_inet6.c (revision eb2bce7f5e7ac1ca6da434461217fadf3c688d2c)
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/capability.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>
31 #include <linux/timer.h>
32 #include <linux/string.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/fcntl.h>
36 #include <linux/mm.h>
37 #include <linux/interrupt.h>
38 #include <linux/proc_fs.h>
39 #include <linux/stat.h>
40 #include <linux/init.h>
41 
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/smp_lock.h>
46 #include <linux/netfilter_ipv6.h>
47 
48 #include <net/ip.h>
49 #include <net/ipv6.h>
50 #include <net/udp.h>
51 #include <net/udplite.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 #ifdef CONFIG_IPV6_MIP6
63 #include <net/mip6.h>
64 #endif
65 
66 #include <asm/uaccess.h>
67 #include <asm/system.h>
68 
69 MODULE_AUTHOR("Cast of dozens");
70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
71 MODULE_LICENSE("GPL");
72 
73 int sysctl_ipv6_bindv6only __read_mostly;
74 
75 /* The inetsw table contains everything that inet_create needs to
76  * build a new socket.
77  */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80 
81 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
82 {
83 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
84 
85 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
86 }
87 
88 static int inet6_create(struct socket *sock, int protocol)
89 {
90 	struct inet_sock *inet;
91 	struct ipv6_pinfo *np;
92 	struct sock *sk;
93 	struct list_head *p;
94 	struct inet_protosw *answer;
95 	struct proto *answer_prot;
96 	unsigned char answer_flags;
97 	char answer_no_check;
98 	int try_loading_module = 0;
99 	int err;
100 
101 	if (sock->type != SOCK_RAW &&
102 	    sock->type != SOCK_DGRAM &&
103 	    !inet_ehash_secret)
104 		build_ehash_secret();
105 
106 	/* Look for the requested type/protocol pair. */
107 	answer = NULL;
108 lookup_protocol:
109 	err = -ESOCKTNOSUPPORT;
110 	rcu_read_lock();
111 	list_for_each_rcu(p, &inetsw6[sock->type]) {
112 		answer = list_entry(p, struct inet_protosw, list);
113 
114 		/* Check the non-wild match. */
115 		if (protocol == answer->protocol) {
116 			if (protocol != IPPROTO_IP)
117 				break;
118 		} else {
119 			/* Check for the two wild cases. */
120 			if (IPPROTO_IP == protocol) {
121 				protocol = answer->protocol;
122 				break;
123 			}
124 			if (IPPROTO_IP == answer->protocol)
125 				break;
126 		}
127 		err = -EPROTONOSUPPORT;
128 		answer = NULL;
129 	}
130 
131 	if (!answer) {
132 		if (try_loading_module < 2) {
133 			rcu_read_unlock();
134 			/*
135 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
136 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
137 			 */
138 			if (++try_loading_module == 1)
139 				request_module("net-pf-%d-proto-%d-type-%d",
140 						PF_INET6, protocol, sock->type);
141 			/*
142 			 * Fall back to generic, e.g. net-pf-10-proto-132
143 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
144 			 */
145 			else
146 				request_module("net-pf-%d-proto-%d",
147 						PF_INET6, protocol);
148 			goto lookup_protocol;
149 		} else
150 			goto out_rcu_unlock;
151 	}
152 
153 	err = -EPERM;
154 	if (answer->capability > 0 && !capable(answer->capability))
155 		goto out_rcu_unlock;
156 
157 	sock->ops = answer->ops;
158 	answer_prot = answer->prot;
159 	answer_no_check = answer->no_check;
160 	answer_flags = answer->flags;
161 	rcu_read_unlock();
162 
163 	BUG_TRAP(answer_prot->slab != NULL);
164 
165 	err = -ENOBUFS;
166 	sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
167 	if (sk == NULL)
168 		goto out;
169 
170 	sock_init_data(sock, sk);
171 
172 	err = 0;
173 	sk->sk_no_check = answer_no_check;
174 	if (INET_PROTOSW_REUSE & answer_flags)
175 		sk->sk_reuse = 1;
176 
177 	inet = inet_sk(sk);
178 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
179 
180 	if (SOCK_RAW == sock->type) {
181 		inet->num = protocol;
182 		if (IPPROTO_RAW == protocol)
183 			inet->hdrincl = 1;
184 	}
185 
186 	sk->sk_destruct		= inet_sock_destruct;
187 	sk->sk_family		= PF_INET6;
188 	sk->sk_protocol		= protocol;
189 
190 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
191 
192 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
193 	np->hop_limit	= -1;
194 	np->mcast_hops	= -1;
195 	np->mc_loop	= 1;
196 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
197 	np->ipv6only	= sysctl_ipv6_bindv6only;
198 
199 	/* Init the ipv4 part of the socket since we can have sockets
200 	 * using v6 API for ipv4.
201 	 */
202 	inet->uc_ttl	= -1;
203 
204 	inet->mc_loop	= 1;
205 	inet->mc_ttl	= 1;
206 	inet->mc_index	= 0;
207 	inet->mc_list	= NULL;
208 
209 	if (ipv4_config.no_pmtu_disc)
210 		inet->pmtudisc = IP_PMTUDISC_DONT;
211 	else
212 		inet->pmtudisc = IP_PMTUDISC_WANT;
213 	/*
214 	 * Increment only the relevant sk_prot->socks debug field, this changes
215 	 * the previous behaviour of incrementing both the equivalent to
216 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
217 	 *
218 	 * This allows better debug granularity as we'll know exactly how many
219 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
220 	 * transport protocol socks. -acme
221 	 */
222 	sk_refcnt_debug_inc(sk);
223 
224 	if (inet->num) {
225 		/* It assumes that any protocol which allows
226 		 * the user to assign a number at socket
227 		 * creation time automatically shares.
228 		 */
229 		inet->sport = htons(inet->num);
230 		sk->sk_prot->hash(sk);
231 	}
232 	if (sk->sk_prot->init) {
233 		err = sk->sk_prot->init(sk);
234 		if (err) {
235 			sk_common_release(sk);
236 			goto out;
237 		}
238 	}
239 out:
240 	return err;
241 out_rcu_unlock:
242 	rcu_read_unlock();
243 	goto out;
244 }
245 
246 
247 /* bind for INET6 API */
248 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
249 {
250 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
251 	struct sock *sk = sock->sk;
252 	struct inet_sock *inet = inet_sk(sk);
253 	struct ipv6_pinfo *np = inet6_sk(sk);
254 	__be32 v4addr = 0;
255 	unsigned short snum;
256 	int addr_type = 0;
257 	int err = 0;
258 
259 	/* If the socket has its own bind function then use it. */
260 	if (sk->sk_prot->bind)
261 		return sk->sk_prot->bind(sk, uaddr, addr_len);
262 
263 	if (addr_len < SIN6_LEN_RFC2133)
264 		return -EINVAL;
265 	addr_type = ipv6_addr_type(&addr->sin6_addr);
266 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
267 		return -EINVAL;
268 
269 	snum = ntohs(addr->sin6_port);
270 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
271 		return -EACCES;
272 
273 	lock_sock(sk);
274 
275 	/* Check these errors (active socket, double bind). */
276 	if (sk->sk_state != TCP_CLOSE || inet->num) {
277 		err = -EINVAL;
278 		goto out;
279 	}
280 
281 	/* Check if the address belongs to the host. */
282 	if (addr_type == IPV6_ADDR_MAPPED) {
283 		v4addr = addr->sin6_addr.s6_addr32[3];
284 		if (inet_addr_type(v4addr) != RTN_LOCAL) {
285 			err = -EADDRNOTAVAIL;
286 			goto out;
287 		}
288 	} else {
289 		if (addr_type != IPV6_ADDR_ANY) {
290 			struct net_device *dev = NULL;
291 
292 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
293 				if (addr_len >= sizeof(struct sockaddr_in6) &&
294 				    addr->sin6_scope_id) {
295 					/* Override any existing binding, if another one
296 					 * is supplied by user.
297 					 */
298 					sk->sk_bound_dev_if = addr->sin6_scope_id;
299 				}
300 
301 				/* Binding to link-local address requires an interface */
302 				if (!sk->sk_bound_dev_if) {
303 					err = -EINVAL;
304 					goto out;
305 				}
306 				dev = dev_get_by_index(sk->sk_bound_dev_if);
307 				if (!dev) {
308 					err = -ENODEV;
309 					goto out;
310 				}
311 			}
312 
313 			/* ipv4 addr of the socket is invalid.  Only the
314 			 * unspecified and mapped address have a v4 equivalent.
315 			 */
316 			v4addr = LOOPBACK4_IPV6;
317 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
318 				if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
319 					if (dev)
320 						dev_put(dev);
321 					err = -EADDRNOTAVAIL;
322 					goto out;
323 				}
324 			}
325 			if (dev)
326 				dev_put(dev);
327 		}
328 	}
329 
330 	inet->rcv_saddr = v4addr;
331 	inet->saddr = v4addr;
332 
333 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
334 
335 	if (!(addr_type & IPV6_ADDR_MULTICAST))
336 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
337 
338 	/* Make sure we are allowed to bind here. */
339 	if (sk->sk_prot->get_port(sk, snum)) {
340 		inet_reset_saddr(sk);
341 		err = -EADDRINUSE;
342 		goto out;
343 	}
344 
345 	if (addr_type != IPV6_ADDR_ANY)
346 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
347 	if (snum)
348 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
349 	inet->sport = htons(inet->num);
350 	inet->dport = 0;
351 	inet->daddr = 0;
352 out:
353 	release_sock(sk);
354 	return err;
355 }
356 
357 EXPORT_SYMBOL(inet6_bind);
358 
359 int inet6_release(struct socket *sock)
360 {
361 	struct sock *sk = sock->sk;
362 
363 	if (sk == NULL)
364 		return -EINVAL;
365 
366 	/* Free mc lists */
367 	ipv6_sock_mc_close(sk);
368 
369 	/* Free ac lists */
370 	ipv6_sock_ac_close(sk);
371 
372 	return inet_release(sock);
373 }
374 
375 EXPORT_SYMBOL(inet6_release);
376 
377 int inet6_destroy_sock(struct sock *sk)
378 {
379 	struct ipv6_pinfo *np = inet6_sk(sk);
380 	struct sk_buff *skb;
381 	struct ipv6_txoptions *opt;
382 
383 	/* Release rx options */
384 
385 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
386 		kfree_skb(skb);
387 
388 	/* Free flowlabels */
389 	fl6_free_socklist(sk);
390 
391 	/* Free tx options */
392 
393 	if ((opt = xchg(&np->opt, NULL)) != NULL)
394 		sock_kfree_s(sk, opt, opt->tot_len);
395 
396 	return 0;
397 }
398 
399 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
400 
401 /*
402  *	This does both peername and sockname.
403  */
404 
405 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
406 		 int *uaddr_len, int peer)
407 {
408 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
409 	struct sock *sk = sock->sk;
410 	struct inet_sock *inet = inet_sk(sk);
411 	struct ipv6_pinfo *np = inet6_sk(sk);
412 
413 	sin->sin6_family = AF_INET6;
414 	sin->sin6_flowinfo = 0;
415 	sin->sin6_scope_id = 0;
416 	if (peer) {
417 		if (!inet->dport)
418 			return -ENOTCONN;
419 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
420 		    peer == 1)
421 			return -ENOTCONN;
422 		sin->sin6_port = inet->dport;
423 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
424 		if (np->sndflow)
425 			sin->sin6_flowinfo = np->flow_label;
426 	} else {
427 		if (ipv6_addr_any(&np->rcv_saddr))
428 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
429 		else
430 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
431 
432 		sin->sin6_port = inet->sport;
433 	}
434 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
435 		sin->sin6_scope_id = sk->sk_bound_dev_if;
436 	*uaddr_len = sizeof(*sin);
437 	return(0);
438 }
439 
440 EXPORT_SYMBOL(inet6_getname);
441 
442 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
443 {
444 	struct sock *sk = sock->sk;
445 
446 	switch(cmd)
447 	{
448 	case SIOCGSTAMP:
449 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
450 
451 	case SIOCGSTAMPNS:
452 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
453 
454 	case SIOCADDRT:
455 	case SIOCDELRT:
456 
457 		return(ipv6_route_ioctl(cmd,(void __user *)arg));
458 
459 	case SIOCSIFADDR:
460 		return addrconf_add_ifaddr((void __user *) arg);
461 	case SIOCDIFADDR:
462 		return addrconf_del_ifaddr((void __user *) arg);
463 	case SIOCSIFDSTADDR:
464 		return addrconf_set_dstaddr((void __user *) arg);
465 	default:
466 		if (!sk->sk_prot->ioctl)
467 			return -ENOIOCTLCMD;
468 		return sk->sk_prot->ioctl(sk, cmd, arg);
469 	}
470 	/*NOTREACHED*/
471 	return(0);
472 }
473 
474 EXPORT_SYMBOL(inet6_ioctl);
475 
476 const struct proto_ops inet6_stream_ops = {
477 	.family		   = PF_INET6,
478 	.owner		   = THIS_MODULE,
479 	.release	   = inet6_release,
480 	.bind		   = inet6_bind,
481 	.connect	   = inet_stream_connect,	/* ok		*/
482 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
483 	.accept		   = inet_accept,		/* ok		*/
484 	.getname	   = inet6_getname,
485 	.poll		   = tcp_poll,			/* ok		*/
486 	.ioctl		   = inet6_ioctl,		/* must change  */
487 	.listen		   = inet_listen,		/* ok		*/
488 	.shutdown	   = inet_shutdown,		/* ok		*/
489 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
490 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
491 	.sendmsg	   = inet_sendmsg,		/* ok		*/
492 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
493 	.mmap		   = sock_no_mmap,
494 	.sendpage	   = tcp_sendpage,
495 #ifdef CONFIG_COMPAT
496 	.compat_setsockopt = compat_sock_common_setsockopt,
497 	.compat_getsockopt = compat_sock_common_getsockopt,
498 #endif
499 };
500 
501 const struct proto_ops inet6_dgram_ops = {
502 	.family		   = PF_INET6,
503 	.owner		   = THIS_MODULE,
504 	.release	   = inet6_release,
505 	.bind		   = inet6_bind,
506 	.connect	   = inet_dgram_connect,	/* ok		*/
507 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
508 	.accept		   = sock_no_accept,		/* a do nothing	*/
509 	.getname	   = inet6_getname,
510 	.poll		   = udp_poll,			/* ok		*/
511 	.ioctl		   = inet6_ioctl,		/* must change  */
512 	.listen		   = sock_no_listen,		/* ok		*/
513 	.shutdown	   = inet_shutdown,		/* ok		*/
514 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
515 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
516 	.sendmsg	   = inet_sendmsg,		/* ok		*/
517 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
518 	.mmap		   = sock_no_mmap,
519 	.sendpage	   = sock_no_sendpage,
520 #ifdef CONFIG_COMPAT
521 	.compat_setsockopt = compat_sock_common_setsockopt,
522 	.compat_getsockopt = compat_sock_common_getsockopt,
523 #endif
524 };
525 
526 static struct net_proto_family inet6_family_ops = {
527 	.family = PF_INET6,
528 	.create = inet6_create,
529 	.owner	= THIS_MODULE,
530 };
531 
532 /* Same as inet6_dgram_ops, sans udp_poll.  */
533 static const struct proto_ops inet6_sockraw_ops = {
534 	.family		   = PF_INET6,
535 	.owner		   = THIS_MODULE,
536 	.release	   = inet6_release,
537 	.bind		   = inet6_bind,
538 	.connect	   = inet_dgram_connect,	/* ok		*/
539 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
540 	.accept		   = sock_no_accept,		/* a do nothing	*/
541 	.getname	   = inet6_getname,
542 	.poll		   = datagram_poll,		/* ok		*/
543 	.ioctl		   = inet6_ioctl,		/* must change  */
544 	.listen		   = sock_no_listen,		/* ok		*/
545 	.shutdown	   = inet_shutdown,		/* ok		*/
546 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
547 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
548 	.sendmsg	   = inet_sendmsg,		/* ok		*/
549 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
550 	.mmap		   = sock_no_mmap,
551 	.sendpage	   = sock_no_sendpage,
552 #ifdef CONFIG_COMPAT
553 	.compat_setsockopt = compat_sock_common_setsockopt,
554 	.compat_getsockopt = compat_sock_common_getsockopt,
555 #endif
556 };
557 
558 static struct inet_protosw rawv6_protosw = {
559 	.type		= SOCK_RAW,
560 	.protocol	= IPPROTO_IP,	/* wild card */
561 	.prot		= &rawv6_prot,
562 	.ops		= &inet6_sockraw_ops,
563 	.capability	= CAP_NET_RAW,
564 	.no_check	= UDP_CSUM_DEFAULT,
565 	.flags		= INET_PROTOSW_REUSE,
566 };
567 
568 void
569 inet6_register_protosw(struct inet_protosw *p)
570 {
571 	struct list_head *lh;
572 	struct inet_protosw *answer;
573 	int protocol = p->protocol;
574 	struct list_head *last_perm;
575 
576 	spin_lock_bh(&inetsw6_lock);
577 
578 	if (p->type >= SOCK_MAX)
579 		goto out_illegal;
580 
581 	/* If we are trying to override a permanent protocol, bail. */
582 	answer = NULL;
583 	last_perm = &inetsw6[p->type];
584 	list_for_each(lh, &inetsw6[p->type]) {
585 		answer = list_entry(lh, struct inet_protosw, list);
586 
587 		/* Check only the non-wild match. */
588 		if (INET_PROTOSW_PERMANENT & answer->flags) {
589 			if (protocol == answer->protocol)
590 				break;
591 			last_perm = lh;
592 		}
593 
594 		answer = NULL;
595 	}
596 	if (answer)
597 		goto out_permanent;
598 
599 	/* Add the new entry after the last permanent entry if any, so that
600 	 * the new entry does not override a permanent entry when matched with
601 	 * a wild-card protocol. But it is allowed to override any existing
602 	 * non-permanent entry.  This means that when we remove this entry, the
603 	 * system automatically returns to the old behavior.
604 	 */
605 	list_add_rcu(&p->list, last_perm);
606 out:
607 	spin_unlock_bh(&inetsw6_lock);
608 	return;
609 
610 out_permanent:
611 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
612 	       protocol);
613 	goto out;
614 
615 out_illegal:
616 	printk(KERN_ERR
617 	       "Ignoring attempt to register invalid socket type %d.\n",
618 	       p->type);
619 	goto out;
620 }
621 
622 EXPORT_SYMBOL(inet6_register_protosw);
623 
624 void
625 inet6_unregister_protosw(struct inet_protosw *p)
626 {
627 	if (INET_PROTOSW_PERMANENT & p->flags) {
628 		printk(KERN_ERR
629 		       "Attempt to unregister permanent protocol %d.\n",
630 		       p->protocol);
631 	} else {
632 		spin_lock_bh(&inetsw6_lock);
633 		list_del_rcu(&p->list);
634 		spin_unlock_bh(&inetsw6_lock);
635 
636 		synchronize_net();
637 	}
638 }
639 
640 EXPORT_SYMBOL(inet6_unregister_protosw);
641 
642 int inet6_sk_rebuild_header(struct sock *sk)
643 {
644 	int err;
645 	struct dst_entry *dst;
646 	struct ipv6_pinfo *np = inet6_sk(sk);
647 
648 	dst = __sk_dst_check(sk, np->dst_cookie);
649 
650 	if (dst == NULL) {
651 		struct inet_sock *inet = inet_sk(sk);
652 		struct in6_addr *final_p = NULL, final;
653 		struct flowi fl;
654 
655 		memset(&fl, 0, sizeof(fl));
656 		fl.proto = sk->sk_protocol;
657 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
658 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
659 		fl.fl6_flowlabel = np->flow_label;
660 		fl.oif = sk->sk_bound_dev_if;
661 		fl.fl_ip_dport = inet->dport;
662 		fl.fl_ip_sport = inet->sport;
663 		security_sk_classify_flow(sk, &fl);
664 
665 		if (np->opt && np->opt->srcrt) {
666 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
667 			ipv6_addr_copy(&final, &fl.fl6_dst);
668 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
669 			final_p = &final;
670 		}
671 
672 		err = ip6_dst_lookup(sk, &dst, &fl);
673 		if (err) {
674 			sk->sk_route_caps = 0;
675 			return err;
676 		}
677 		if (final_p)
678 			ipv6_addr_copy(&fl.fl6_dst, final_p);
679 
680 		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
681 			sk->sk_err_soft = -err;
682 			return err;
683 		}
684 
685 		__ip6_dst_store(sk, dst, NULL, NULL);
686 	}
687 
688 	return 0;
689 }
690 
691 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
692 
693 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
694 {
695 	struct ipv6_pinfo *np = inet6_sk(sk);
696 	struct inet6_skb_parm *opt = IP6CB(skb);
697 
698 	if (np->rxopt.all) {
699 		if ((opt->hop && (np->rxopt.bits.hopopts ||
700 				  np->rxopt.bits.ohopopts)) ||
701 		    ((IPV6_FLOWINFO_MASK &
702 		      *(__be32 *)skb_network_header(skb)) &&
703 		     np->rxopt.bits.rxflow) ||
704 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
705 		     np->rxopt.bits.osrcrt)) ||
706 		    ((opt->dst1 || opt->dst0) &&
707 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
708 			return 1;
709 	}
710 	return 0;
711 }
712 
713 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
714 
715 static int __init init_ipv6_mibs(void)
716 {
717 	if (snmp_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
718 			  __alignof__(struct ipstats_mib)) < 0)
719 		goto err_ip_mib;
720 	if (snmp_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
721 			  __alignof__(struct icmpv6_mib)) < 0)
722 		goto err_icmp_mib;
723 	if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
724 			  __alignof__(struct udp_mib)) < 0)
725 		goto err_udp_mib;
726 	if (snmp_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
727 			  __alignof__(struct udp_mib)) < 0)
728 		goto err_udplite_mib;
729 	return 0;
730 
731 err_udplite_mib:
732 	snmp_mib_free((void **)udp_stats_in6);
733 err_udp_mib:
734 	snmp_mib_free((void **)icmpv6_statistics);
735 err_icmp_mib:
736 	snmp_mib_free((void **)ipv6_statistics);
737 err_ip_mib:
738 	return -ENOMEM;
739 
740 }
741 
742 static void cleanup_ipv6_mibs(void)
743 {
744 	snmp_mib_free((void **)ipv6_statistics);
745 	snmp_mib_free((void **)icmpv6_statistics);
746 	snmp_mib_free((void **)udp_stats_in6);
747 	snmp_mib_free((void **)udplite_stats_in6);
748 }
749 
750 static int __init inet6_init(void)
751 {
752 	struct sk_buff *dummy_skb;
753 	struct list_head *r;
754 	int err;
755 
756 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
757 
758 #ifdef MODULE
759 #if 0 /* FIXME --RR */
760 	if (!mod_member_present(&__this_module, can_unload))
761 	  return -EINVAL;
762 
763 	__this_module.can_unload = &ipv6_unload;
764 #endif
765 #endif
766 
767 	err = proto_register(&tcpv6_prot, 1);
768 	if (err)
769 		goto out;
770 
771 	err = proto_register(&udpv6_prot, 1);
772 	if (err)
773 		goto out_unregister_tcp_proto;
774 
775 	err = proto_register(&udplitev6_prot, 1);
776 	if (err)
777 		goto out_unregister_udp_proto;
778 
779 	err = proto_register(&rawv6_prot, 1);
780 	if (err)
781 		goto out_unregister_udplite_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 	 */
796 	err = sock_register(&inet6_family_ops);
797 	if (err)
798 		goto out_unregister_raw_proto;
799 
800 	/* Initialise ipv6 mibs */
801 	err = init_ipv6_mibs();
802 	if (err)
803 		goto out_unregister_sock;
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;
824 	err = ipv6_netfilter_init();
825 	if (err)
826 		goto netfilter_fail;
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 (udplite6_proc_init())
837 		goto proc_udplite6_fail;
838 	if (ipv6_misc_proc_init())
839 		goto proc_misc6_fail;
840 
841 	if (ac6_proc_init())
842 		goto proc_anycast6_fail;
843 	if (if6_proc_init())
844 		goto proc_if6_fail;
845 #endif
846 	ip6_route_init();
847 	ip6_flowlabel_init();
848 	err = addrconf_init();
849 	if (err)
850 		goto addrconf_fail;
851 
852 	/* Init v6 extension headers. */
853 	ipv6_rthdr_init();
854 	ipv6_frag_init();
855 	ipv6_nodata_init();
856 	ipv6_destopt_init();
857 #ifdef CONFIG_IPV6_MIP6
858 	mip6_init();
859 #endif
860 
861 	/* Init v6 transport protocols. */
862 	udpv6_init();
863 	udplitev6_init();
864 	tcpv6_init();
865 
866 	ipv6_packet_init();
867 	err = 0;
868 out:
869 	return err;
870 
871 addrconf_fail:
872 	ip6_flowlabel_cleanup();
873 	ip6_route_cleanup();
874 #ifdef CONFIG_PROC_FS
875 	if6_proc_exit();
876 proc_if6_fail:
877 	ac6_proc_exit();
878 proc_anycast6_fail:
879 	ipv6_misc_proc_exit();
880 proc_misc6_fail:
881 	udplite6_proc_exit();
882 proc_udplite6_fail:
883 	udp6_proc_exit();
884 proc_udp6_fail:
885 	tcp6_proc_exit();
886 proc_tcp6_fail:
887 	raw6_proc_exit();
888 proc_raw6_fail:
889 #endif
890 	ipv6_netfilter_fini();
891 netfilter_fail:
892 	igmp6_cleanup();
893 igmp_fail:
894 	ndisc_cleanup();
895 ndisc_fail:
896 	icmpv6_cleanup();
897 icmp_fail:
898 #ifdef CONFIG_SYSCTL
899 	ipv6_sysctl_unregister();
900 #endif
901 	cleanup_ipv6_mibs();
902 out_unregister_sock:
903 	sock_unregister(PF_INET6);
904 out_unregister_raw_proto:
905 	proto_unregister(&rawv6_prot);
906 out_unregister_udplite_proto:
907 	proto_unregister(&udplitev6_prot);
908 out_unregister_udp_proto:
909 	proto_unregister(&udpv6_prot);
910 out_unregister_tcp_proto:
911 	proto_unregister(&tcpv6_prot);
912 	goto out;
913 }
914 module_init(inet6_init);
915 
916 static void __exit inet6_exit(void)
917 {
918 	/* First of all disallow new sockets creation. */
919 	sock_unregister(PF_INET6);
920 	/* Disallow any further netlink messages */
921 	rtnl_unregister_all(PF_INET6);
922 
923 	/* Cleanup code parts. */
924 	ipv6_packet_cleanup();
925 #ifdef CONFIG_IPV6_MIP6
926 	mip6_fini();
927 #endif
928 	addrconf_cleanup();
929 	ip6_flowlabel_cleanup();
930 	ip6_route_cleanup();
931 #ifdef CONFIG_PROC_FS
932 
933 	/* Cleanup code parts. */
934 	if6_proc_exit();
935 	ac6_proc_exit();
936 	ipv6_misc_proc_exit();
937 	udplite6_proc_exit();
938 	udp6_proc_exit();
939 	tcp6_proc_exit();
940 	raw6_proc_exit();
941 #endif
942 	ipv6_netfilter_fini();
943 	igmp6_cleanup();
944 	ndisc_cleanup();
945 	icmpv6_cleanup();
946 #ifdef CONFIG_SYSCTL
947 	ipv6_sysctl_unregister();
948 #endif
949 	cleanup_ipv6_mibs();
950 	proto_unregister(&rawv6_prot);
951 	proto_unregister(&udplitev6_prot);
952 	proto_unregister(&udpv6_prot);
953 	proto_unregister(&tcpv6_prot);
954 }
955 module_exit(inet6_exit);
956 
957 MODULE_ALIAS_NETPROTO(PF_INET6);
958