xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 96de0e252cedffad61b3cb5e05662c591898e69a)
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/netfilter_ipv6.h>
46 
47 #include <net/ip.h>
48 #include <net/ipv6.h>
49 #include <net/udp.h>
50 #include <net/udplite.h>
51 #include <net/tcp.h>
52 #include <net/ipip.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/transp_v6.h>
56 #include <net/ip6_route.h>
57 #include <net/addrconf.h>
58 #ifdef CONFIG_IPV6_TUNNEL
59 #include <net/ip6_tunnel.h>
60 #endif
61 
62 #include <asm/uaccess.h>
63 #include <asm/system.h>
64 
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
68 
69 int sysctl_ipv6_bindv6only __read_mostly;
70 
71 /* The inetsw table contains everything that inet_create needs to
72  * build a new socket.
73  */
74 static struct list_head inetsw6[SOCK_MAX];
75 static DEFINE_SPINLOCK(inetsw6_lock);
76 
77 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
78 {
79 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
80 
81 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
82 }
83 
84 static int inet6_create(struct net *net, struct socket *sock, int protocol)
85 {
86 	struct inet_sock *inet;
87 	struct ipv6_pinfo *np;
88 	struct sock *sk;
89 	struct list_head *p;
90 	struct inet_protosw *answer;
91 	struct proto *answer_prot;
92 	unsigned char answer_flags;
93 	char answer_no_check;
94 	int try_loading_module = 0;
95 	int err;
96 
97 	if (net != &init_net)
98 		return -EAFNOSUPPORT;
99 
100 	if (sock->type != SOCK_RAW &&
101 	    sock->type != SOCK_DGRAM &&
102 	    !inet_ehash_secret)
103 		build_ehash_secret();
104 
105 	/* Look for the requested type/protocol pair. */
106 	answer = NULL;
107 lookup_protocol:
108 	err = -ESOCKTNOSUPPORT;
109 	rcu_read_lock();
110 	list_for_each_rcu(p, &inetsw6[sock->type]) {
111 		answer = list_entry(p, struct inet_protosw, list);
112 
113 		/* Check the non-wild match. */
114 		if (protocol == answer->protocol) {
115 			if (protocol != IPPROTO_IP)
116 				break;
117 		} else {
118 			/* Check for the two wild cases. */
119 			if (IPPROTO_IP == protocol) {
120 				protocol = answer->protocol;
121 				break;
122 			}
123 			if (IPPROTO_IP == answer->protocol)
124 				break;
125 		}
126 		err = -EPROTONOSUPPORT;
127 		answer = NULL;
128 	}
129 
130 	if (!answer) {
131 		if (try_loading_module < 2) {
132 			rcu_read_unlock();
133 			/*
134 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
135 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
136 			 */
137 			if (++try_loading_module == 1)
138 				request_module("net-pf-%d-proto-%d-type-%d",
139 						PF_INET6, protocol, sock->type);
140 			/*
141 			 * Fall back to generic, e.g. net-pf-10-proto-132
142 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
143 			 */
144 			else
145 				request_module("net-pf-%d-proto-%d",
146 						PF_INET6, protocol);
147 			goto lookup_protocol;
148 		} else
149 			goto out_rcu_unlock;
150 	}
151 
152 	err = -EPERM;
153 	if (answer->capability > 0 && !capable(answer->capability))
154 		goto out_rcu_unlock;
155 
156 	sock->ops = answer->ops;
157 	answer_prot = answer->prot;
158 	answer_no_check = answer->no_check;
159 	answer_flags = answer->flags;
160 	rcu_read_unlock();
161 
162 	BUG_TRAP(answer_prot->slab != NULL);
163 
164 	err = -ENOBUFS;
165 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, 1);
166 	if (sk == NULL)
167 		goto out;
168 
169 	sock_init_data(sock, sk);
170 
171 	err = 0;
172 	sk->sk_no_check = answer_no_check;
173 	if (INET_PROTOSW_REUSE & answer_flags)
174 		sk->sk_reuse = 1;
175 
176 	inet = inet_sk(sk);
177 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
178 
179 	if (SOCK_RAW == sock->type) {
180 		inet->num = protocol;
181 		if (IPPROTO_RAW == protocol)
182 			inet->hdrincl = 1;
183 	}
184 
185 	sk->sk_destruct		= inet_sock_destruct;
186 	sk->sk_family		= PF_INET6;
187 	sk->sk_protocol		= protocol;
188 
189 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
190 
191 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
192 	np->hop_limit	= -1;
193 	np->mcast_hops	= -1;
194 	np->mc_loop	= 1;
195 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
196 	np->ipv6only	= sysctl_ipv6_bindv6only;
197 
198 	/* Init the ipv4 part of the socket since we can have sockets
199 	 * using v6 API for ipv4.
200 	 */
201 	inet->uc_ttl	= -1;
202 
203 	inet->mc_loop	= 1;
204 	inet->mc_ttl	= 1;
205 	inet->mc_index	= 0;
206 	inet->mc_list	= NULL;
207 
208 	if (ipv4_config.no_pmtu_disc)
209 		inet->pmtudisc = IP_PMTUDISC_DONT;
210 	else
211 		inet->pmtudisc = IP_PMTUDISC_WANT;
212 	/*
213 	 * Increment only the relevant sk_prot->socks debug field, this changes
214 	 * the previous behaviour of incrementing both the equivalent to
215 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
216 	 *
217 	 * This allows better debug granularity as we'll know exactly how many
218 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
219 	 * transport protocol socks. -acme
220 	 */
221 	sk_refcnt_debug_inc(sk);
222 
223 	if (inet->num) {
224 		/* It assumes that any protocol which allows
225 		 * the user to assign a number at socket
226 		 * creation time automatically shares.
227 		 */
228 		inet->sport = htons(inet->num);
229 		sk->sk_prot->hash(sk);
230 	}
231 	if (sk->sk_prot->init) {
232 		err = sk->sk_prot->init(sk);
233 		if (err) {
234 			sk_common_release(sk);
235 			goto out;
236 		}
237 	}
238 out:
239 	return err;
240 out_rcu_unlock:
241 	rcu_read_unlock();
242 	goto out;
243 }
244 
245 
246 /* bind for INET6 API */
247 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
248 {
249 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
250 	struct sock *sk = sock->sk;
251 	struct inet_sock *inet = inet_sk(sk);
252 	struct ipv6_pinfo *np = inet6_sk(sk);
253 	__be32 v4addr = 0;
254 	unsigned short snum;
255 	int addr_type = 0;
256 	int err = 0;
257 
258 	/* If the socket has its own bind function then use it. */
259 	if (sk->sk_prot->bind)
260 		return sk->sk_prot->bind(sk, uaddr, addr_len);
261 
262 	if (addr_len < SIN6_LEN_RFC2133)
263 		return -EINVAL;
264 	addr_type = ipv6_addr_type(&addr->sin6_addr);
265 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
266 		return -EINVAL;
267 
268 	snum = ntohs(addr->sin6_port);
269 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
270 		return -EACCES;
271 
272 	lock_sock(sk);
273 
274 	/* Check these errors (active socket, double bind). */
275 	if (sk->sk_state != TCP_CLOSE || inet->num) {
276 		err = -EINVAL;
277 		goto out;
278 	}
279 
280 	/* Check if the address belongs to the host. */
281 	if (addr_type == IPV6_ADDR_MAPPED) {
282 		v4addr = addr->sin6_addr.s6_addr32[3];
283 		if (inet_addr_type(v4addr) != RTN_LOCAL) {
284 			err = -EADDRNOTAVAIL;
285 			goto out;
286 		}
287 	} else {
288 		if (addr_type != IPV6_ADDR_ANY) {
289 			struct net_device *dev = NULL;
290 
291 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
292 				if (addr_len >= sizeof(struct sockaddr_in6) &&
293 				    addr->sin6_scope_id) {
294 					/* Override any existing binding, if another one
295 					 * is supplied by user.
296 					 */
297 					sk->sk_bound_dev_if = addr->sin6_scope_id;
298 				}
299 
300 				/* Binding to link-local address requires an interface */
301 				if (!sk->sk_bound_dev_if) {
302 					err = -EINVAL;
303 					goto out;
304 				}
305 				dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
306 				if (!dev) {
307 					err = -ENODEV;
308 					goto out;
309 				}
310 			}
311 
312 			/* ipv4 addr of the socket is invalid.  Only the
313 			 * unspecified and mapped address have a v4 equivalent.
314 			 */
315 			v4addr = LOOPBACK4_IPV6;
316 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
317 				if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
318 					if (dev)
319 						dev_put(dev);
320 					err = -EADDRNOTAVAIL;
321 					goto out;
322 				}
323 			}
324 			if (dev)
325 				dev_put(dev);
326 		}
327 	}
328 
329 	inet->rcv_saddr = v4addr;
330 	inet->saddr = v4addr;
331 
332 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
333 
334 	if (!(addr_type & IPV6_ADDR_MULTICAST))
335 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
336 
337 	/* Make sure we are allowed to bind here. */
338 	if (sk->sk_prot->get_port(sk, snum)) {
339 		inet_reset_saddr(sk);
340 		err = -EADDRINUSE;
341 		goto out;
342 	}
343 
344 	if (addr_type != IPV6_ADDR_ANY)
345 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
346 	if (snum)
347 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
348 	inet->sport = htons(inet->num);
349 	inet->dport = 0;
350 	inet->daddr = 0;
351 out:
352 	release_sock(sk);
353 	return err;
354 }
355 
356 EXPORT_SYMBOL(inet6_bind);
357 
358 int inet6_release(struct socket *sock)
359 {
360 	struct sock *sk = sock->sk;
361 
362 	if (sk == NULL)
363 		return -EINVAL;
364 
365 	/* Free mc lists */
366 	ipv6_sock_mc_close(sk);
367 
368 	/* Free ac lists */
369 	ipv6_sock_ac_close(sk);
370 
371 	return inet_release(sock);
372 }
373 
374 EXPORT_SYMBOL(inet6_release);
375 
376 int inet6_destroy_sock(struct sock *sk)
377 {
378 	struct ipv6_pinfo *np = inet6_sk(sk);
379 	struct sk_buff *skb;
380 	struct ipv6_txoptions *opt;
381 
382 	/* Release rx options */
383 
384 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
385 		kfree_skb(skb);
386 
387 	/* Free flowlabels */
388 	fl6_free_socklist(sk);
389 
390 	/* Free tx options */
391 
392 	if ((opt = xchg(&np->opt, NULL)) != NULL)
393 		sock_kfree_s(sk, opt, opt->tot_len);
394 
395 	return 0;
396 }
397 
398 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
399 
400 /*
401  *	This does both peername and sockname.
402  */
403 
404 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
405 		 int *uaddr_len, int peer)
406 {
407 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
408 	struct sock *sk = sock->sk;
409 	struct inet_sock *inet = inet_sk(sk);
410 	struct ipv6_pinfo *np = inet6_sk(sk);
411 
412 	sin->sin6_family = AF_INET6;
413 	sin->sin6_flowinfo = 0;
414 	sin->sin6_scope_id = 0;
415 	if (peer) {
416 		if (!inet->dport)
417 			return -ENOTCONN;
418 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
419 		    peer == 1)
420 			return -ENOTCONN;
421 		sin->sin6_port = inet->dport;
422 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
423 		if (np->sndflow)
424 			sin->sin6_flowinfo = np->flow_label;
425 	} else {
426 		if (ipv6_addr_any(&np->rcv_saddr))
427 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
428 		else
429 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
430 
431 		sin->sin6_port = inet->sport;
432 	}
433 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
434 		sin->sin6_scope_id = sk->sk_bound_dev_if;
435 	*uaddr_len = sizeof(*sin);
436 	return(0);
437 }
438 
439 EXPORT_SYMBOL(inet6_getname);
440 
441 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
442 {
443 	struct sock *sk = sock->sk;
444 
445 	switch(cmd)
446 	{
447 	case SIOCGSTAMP:
448 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
449 
450 	case SIOCGSTAMPNS:
451 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
452 
453 	case SIOCADDRT:
454 	case SIOCDELRT:
455 
456 		return(ipv6_route_ioctl(cmd,(void __user *)arg));
457 
458 	case SIOCSIFADDR:
459 		return addrconf_add_ifaddr((void __user *) arg);
460 	case SIOCDIFADDR:
461 		return addrconf_del_ifaddr((void __user *) arg);
462 	case SIOCSIFDSTADDR:
463 		return addrconf_set_dstaddr((void __user *) arg);
464 	default:
465 		if (!sk->sk_prot->ioctl)
466 			return -ENOIOCTLCMD;
467 		return sk->sk_prot->ioctl(sk, cmd, arg);
468 	}
469 	/*NOTREACHED*/
470 	return(0);
471 }
472 
473 EXPORT_SYMBOL(inet6_ioctl);
474 
475 const struct proto_ops inet6_stream_ops = {
476 	.family		   = PF_INET6,
477 	.owner		   = THIS_MODULE,
478 	.release	   = inet6_release,
479 	.bind		   = inet6_bind,
480 	.connect	   = inet_stream_connect,	/* ok		*/
481 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
482 	.accept		   = inet_accept,		/* ok		*/
483 	.getname	   = inet6_getname,
484 	.poll		   = tcp_poll,			/* ok		*/
485 	.ioctl		   = inet6_ioctl,		/* must change  */
486 	.listen		   = inet_listen,		/* ok		*/
487 	.shutdown	   = inet_shutdown,		/* ok		*/
488 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
489 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
490 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
491 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
492 	.mmap		   = sock_no_mmap,
493 	.sendpage	   = tcp_sendpage,
494 #ifdef CONFIG_COMPAT
495 	.compat_setsockopt = compat_sock_common_setsockopt,
496 	.compat_getsockopt = compat_sock_common_getsockopt,
497 #endif
498 };
499 
500 const struct proto_ops inet6_dgram_ops = {
501 	.family		   = PF_INET6,
502 	.owner		   = THIS_MODULE,
503 	.release	   = inet6_release,
504 	.bind		   = inet6_bind,
505 	.connect	   = inet_dgram_connect,	/* ok		*/
506 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
507 	.accept		   = sock_no_accept,		/* a do nothing	*/
508 	.getname	   = inet6_getname,
509 	.poll		   = udp_poll,			/* ok		*/
510 	.ioctl		   = inet6_ioctl,		/* must change  */
511 	.listen		   = sock_no_listen,		/* ok		*/
512 	.shutdown	   = inet_shutdown,		/* ok		*/
513 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
514 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
515 	.sendmsg	   = inet_sendmsg,		/* ok		*/
516 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
517 	.mmap		   = sock_no_mmap,
518 	.sendpage	   = sock_no_sendpage,
519 #ifdef CONFIG_COMPAT
520 	.compat_setsockopt = compat_sock_common_setsockopt,
521 	.compat_getsockopt = compat_sock_common_getsockopt,
522 #endif
523 };
524 
525 static struct net_proto_family inet6_family_ops = {
526 	.family = PF_INET6,
527 	.create = inet6_create,
528 	.owner	= THIS_MODULE,
529 };
530 
531 /* Same as inet6_dgram_ops, sans udp_poll.  */
532 static const struct proto_ops inet6_sockraw_ops = {
533 	.family		   = PF_INET6,
534 	.owner		   = THIS_MODULE,
535 	.release	   = inet6_release,
536 	.bind		   = inet6_bind,
537 	.connect	   = inet_dgram_connect,	/* ok		*/
538 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
539 	.accept		   = sock_no_accept,		/* a do nothing	*/
540 	.getname	   = inet6_getname,
541 	.poll		   = datagram_poll,		/* ok		*/
542 	.ioctl		   = inet6_ioctl,		/* must change  */
543 	.listen		   = sock_no_listen,		/* ok		*/
544 	.shutdown	   = inet_shutdown,		/* ok		*/
545 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
546 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
547 	.sendmsg	   = inet_sendmsg,		/* ok		*/
548 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
549 	.mmap		   = sock_no_mmap,
550 	.sendpage	   = sock_no_sendpage,
551 #ifdef CONFIG_COMPAT
552 	.compat_setsockopt = compat_sock_common_setsockopt,
553 	.compat_getsockopt = compat_sock_common_getsockopt,
554 #endif
555 };
556 
557 static struct inet_protosw rawv6_protosw = {
558 	.type		= SOCK_RAW,
559 	.protocol	= IPPROTO_IP,	/* wild card */
560 	.prot		= &rawv6_prot,
561 	.ops		= &inet6_sockraw_ops,
562 	.capability	= CAP_NET_RAW,
563 	.no_check	= UDP_CSUM_DEFAULT,
564 	.flags		= INET_PROTOSW_REUSE,
565 };
566 
567 void
568 inet6_register_protosw(struct inet_protosw *p)
569 {
570 	struct list_head *lh;
571 	struct inet_protosw *answer;
572 	int protocol = p->protocol;
573 	struct list_head *last_perm;
574 
575 	spin_lock_bh(&inetsw6_lock);
576 
577 	if (p->type >= SOCK_MAX)
578 		goto out_illegal;
579 
580 	/* If we are trying to override a permanent protocol, bail. */
581 	answer = NULL;
582 	last_perm = &inetsw6[p->type];
583 	list_for_each(lh, &inetsw6[p->type]) {
584 		answer = list_entry(lh, struct inet_protosw, list);
585 
586 		/* Check only the non-wild match. */
587 		if (INET_PROTOSW_PERMANENT & answer->flags) {
588 			if (protocol == answer->protocol)
589 				break;
590 			last_perm = lh;
591 		}
592 
593 		answer = NULL;
594 	}
595 	if (answer)
596 		goto out_permanent;
597 
598 	/* Add the new entry after the last permanent entry if any, so that
599 	 * the new entry does not override a permanent entry when matched with
600 	 * a wild-card protocol. But it is allowed to override any existing
601 	 * non-permanent entry.  This means that when we remove this entry, the
602 	 * system automatically returns to the old behavior.
603 	 */
604 	list_add_rcu(&p->list, last_perm);
605 out:
606 	spin_unlock_bh(&inetsw6_lock);
607 	return;
608 
609 out_permanent:
610 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
611 	       protocol);
612 	goto out;
613 
614 out_illegal:
615 	printk(KERN_ERR
616 	       "Ignoring attempt to register invalid socket type %d.\n",
617 	       p->type);
618 	goto out;
619 }
620 
621 EXPORT_SYMBOL(inet6_register_protosw);
622 
623 void
624 inet6_unregister_protosw(struct inet_protosw *p)
625 {
626 	if (INET_PROTOSW_PERMANENT & p->flags) {
627 		printk(KERN_ERR
628 		       "Attempt to unregister permanent protocol %d.\n",
629 		       p->protocol);
630 	} else {
631 		spin_lock_bh(&inetsw6_lock);
632 		list_del_rcu(&p->list);
633 		spin_unlock_bh(&inetsw6_lock);
634 
635 		synchronize_net();
636 	}
637 }
638 
639 EXPORT_SYMBOL(inet6_unregister_protosw);
640 
641 int inet6_sk_rebuild_header(struct sock *sk)
642 {
643 	int err;
644 	struct dst_entry *dst;
645 	struct ipv6_pinfo *np = inet6_sk(sk);
646 
647 	dst = __sk_dst_check(sk, np->dst_cookie);
648 
649 	if (dst == NULL) {
650 		struct inet_sock *inet = inet_sk(sk);
651 		struct in6_addr *final_p = NULL, final;
652 		struct flowi fl;
653 
654 		memset(&fl, 0, sizeof(fl));
655 		fl.proto = sk->sk_protocol;
656 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
657 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
658 		fl.fl6_flowlabel = np->flow_label;
659 		fl.oif = sk->sk_bound_dev_if;
660 		fl.fl_ip_dport = inet->dport;
661 		fl.fl_ip_sport = inet->sport;
662 		security_sk_classify_flow(sk, &fl);
663 
664 		if (np->opt && np->opt->srcrt) {
665 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
666 			ipv6_addr_copy(&final, &fl.fl6_dst);
667 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
668 			final_p = &final;
669 		}
670 
671 		err = ip6_dst_lookup(sk, &dst, &fl);
672 		if (err) {
673 			sk->sk_route_caps = 0;
674 			return err;
675 		}
676 		if (final_p)
677 			ipv6_addr_copy(&fl.fl6_dst, final_p);
678 
679 		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
680 			sk->sk_err_soft = -err;
681 			return err;
682 		}
683 
684 		__ip6_dst_store(sk, dst, NULL, NULL);
685 	}
686 
687 	return 0;
688 }
689 
690 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
691 
692 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
693 {
694 	struct ipv6_pinfo *np = inet6_sk(sk);
695 	struct inet6_skb_parm *opt = IP6CB(skb);
696 
697 	if (np->rxopt.all) {
698 		if ((opt->hop && (np->rxopt.bits.hopopts ||
699 				  np->rxopt.bits.ohopopts)) ||
700 		    ((IPV6_FLOWINFO_MASK &
701 		      *(__be32 *)skb_network_header(skb)) &&
702 		     np->rxopt.bits.rxflow) ||
703 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
704 		     np->rxopt.bits.osrcrt)) ||
705 		    ((opt->dst1 || opt->dst0) &&
706 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
707 			return 1;
708 	}
709 	return 0;
710 }
711 
712 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
713 
714 static int __init init_ipv6_mibs(void)
715 {
716 	if (snmp_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
717 			  __alignof__(struct ipstats_mib)) < 0)
718 		goto err_ip_mib;
719 	if (snmp_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
720 			  __alignof__(struct icmpv6_mib)) < 0)
721 		goto err_icmp_mib;
722 	if (snmp_mib_init((void **)icmpv6msg_statistics,
723 	    sizeof (struct icmpv6msg_mib), __alignof__(struct icmpv6_mib)) < 0)
724 		goto err_icmpmsg_mib;
725 	if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
726 			  __alignof__(struct udp_mib)) < 0)
727 		goto err_udp_mib;
728 	if (snmp_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
729 			  __alignof__(struct udp_mib)) < 0)
730 		goto err_udplite_mib;
731 	return 0;
732 
733 err_udplite_mib:
734 	snmp_mib_free((void **)udp_stats_in6);
735 err_udp_mib:
736 	snmp_mib_free((void **)icmpv6msg_statistics);
737 err_icmpmsg_mib:
738 	snmp_mib_free((void **)icmpv6_statistics);
739 err_icmp_mib:
740 	snmp_mib_free((void **)ipv6_statistics);
741 err_ip_mib:
742 	return -ENOMEM;
743 
744 }
745 
746 static void cleanup_ipv6_mibs(void)
747 {
748 	snmp_mib_free((void **)ipv6_statistics);
749 	snmp_mib_free((void **)icmpv6_statistics);
750 	snmp_mib_free((void **)icmpv6msg_statistics);
751 	snmp_mib_free((void **)udp_stats_in6);
752 	snmp_mib_free((void **)udplite_stats_in6);
753 }
754 
755 static int __init inet6_init(void)
756 {
757 	struct sk_buff *dummy_skb;
758 	struct list_head *r;
759 	int err;
760 
761 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
762 
763 #ifdef MODULE
764 #if 0 /* FIXME --RR */
765 	if (!mod_member_present(&__this_module, can_unload))
766 	  return -EINVAL;
767 
768 	__this_module.can_unload = &ipv6_unload;
769 #endif
770 #endif
771 
772 	err = proto_register(&tcpv6_prot, 1);
773 	if (err)
774 		goto out;
775 
776 	err = proto_register(&udpv6_prot, 1);
777 	if (err)
778 		goto out_unregister_tcp_proto;
779 
780 	err = proto_register(&udplitev6_prot, 1);
781 	if (err)
782 		goto out_unregister_udp_proto;
783 
784 	err = proto_register(&rawv6_prot, 1);
785 	if (err)
786 		goto out_unregister_udplite_proto;
787 
788 
789 	/* Register the socket-side information for inet6_create.  */
790 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
791 		INIT_LIST_HEAD(r);
792 
793 	/* We MUST register RAW sockets before we create the ICMP6,
794 	 * IGMP6, or NDISC control sockets.
795 	 */
796 	inet6_register_protosw(&rawv6_protosw);
797 
798 	/* Register the family here so that the init calls below will
799 	 * be able to create sockets. (?? is this dangerous ??)
800 	 */
801 	err = sock_register(&inet6_family_ops);
802 	if (err)
803 		goto out_unregister_raw_proto;
804 
805 	/* Initialise ipv6 mibs */
806 	err = init_ipv6_mibs();
807 	if (err)
808 		goto out_unregister_sock;
809 
810 	/*
811 	 *	ipngwg API draft makes clear that the correct semantics
812 	 *	for TCP and UDP is to consider one TCP and UDP instance
813 	 *	in a host availiable by both INET and INET6 APIs and
814 	 *	able to communicate via both network protocols.
815 	 */
816 
817 #ifdef CONFIG_SYSCTL
818 	ipv6_sysctl_register();
819 #endif
820 	err = icmpv6_init(&inet6_family_ops);
821 	if (err)
822 		goto icmp_fail;
823 	err = ndisc_init(&inet6_family_ops);
824 	if (err)
825 		goto ndisc_fail;
826 	err = igmp6_init(&inet6_family_ops);
827 	if (err)
828 		goto igmp_fail;
829 	err = ipv6_netfilter_init();
830 	if (err)
831 		goto netfilter_fail;
832 	/* Create /proc/foo6 entries. */
833 #ifdef CONFIG_PROC_FS
834 	err = -ENOMEM;
835 	if (raw6_proc_init())
836 		goto proc_raw6_fail;
837 	if (tcp6_proc_init())
838 		goto proc_tcp6_fail;
839 	if (udp6_proc_init())
840 		goto proc_udp6_fail;
841 	if (udplite6_proc_init())
842 		goto proc_udplite6_fail;
843 	if (ipv6_misc_proc_init())
844 		goto proc_misc6_fail;
845 
846 	if (ac6_proc_init())
847 		goto proc_anycast6_fail;
848 	if (if6_proc_init())
849 		goto proc_if6_fail;
850 #endif
851 	ip6_route_init();
852 	ip6_flowlabel_init();
853 	err = addrconf_init();
854 	if (err)
855 		goto addrconf_fail;
856 
857 	/* Init v6 extension headers. */
858 	ipv6_rthdr_init();
859 	ipv6_frag_init();
860 	ipv6_nodata_init();
861 	ipv6_destopt_init();
862 
863 	/* Init v6 transport protocols. */
864 	udpv6_init();
865 	udplitev6_init();
866 	tcpv6_init();
867 
868 	ipv6_packet_init();
869 	err = 0;
870 out:
871 	return err;
872 
873 addrconf_fail:
874 	ip6_flowlabel_cleanup();
875 	ip6_route_cleanup();
876 #ifdef CONFIG_PROC_FS
877 	if6_proc_exit();
878 proc_if6_fail:
879 	ac6_proc_exit();
880 proc_anycast6_fail:
881 	ipv6_misc_proc_exit();
882 proc_misc6_fail:
883 	udplite6_proc_exit();
884 proc_udplite6_fail:
885 	udp6_proc_exit();
886 proc_udp6_fail:
887 	tcp6_proc_exit();
888 proc_tcp6_fail:
889 	raw6_proc_exit();
890 proc_raw6_fail:
891 #endif
892 	ipv6_netfilter_fini();
893 netfilter_fail:
894 	igmp6_cleanup();
895 igmp_fail:
896 	ndisc_cleanup();
897 ndisc_fail:
898 	icmpv6_cleanup();
899 icmp_fail:
900 #ifdef CONFIG_SYSCTL
901 	ipv6_sysctl_unregister();
902 #endif
903 	cleanup_ipv6_mibs();
904 out_unregister_sock:
905 	sock_unregister(PF_INET6);
906 out_unregister_raw_proto:
907 	proto_unregister(&rawv6_prot);
908 out_unregister_udplite_proto:
909 	proto_unregister(&udplitev6_prot);
910 out_unregister_udp_proto:
911 	proto_unregister(&udpv6_prot);
912 out_unregister_tcp_proto:
913 	proto_unregister(&tcpv6_prot);
914 	goto out;
915 }
916 module_init(inet6_init);
917 
918 static void __exit inet6_exit(void)
919 {
920 	/* First of all disallow new sockets creation. */
921 	sock_unregister(PF_INET6);
922 	/* Disallow any further netlink messages */
923 	rtnl_unregister_all(PF_INET6);
924 
925 	/* Cleanup code parts. */
926 	ipv6_packet_cleanup();
927 
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