xref: /linux/net/ipv6/af_inet6.c (revision f3d9478b2ce468c3115b02ecae7e975990697f15)
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/config.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/in.h>
31 #include <linux/kernel.h>
32 #include <linux/sched.h>
33 #include <linux/timer.h>
34 #include <linux/string.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/fcntl.h>
38 #include <linux/mm.h>
39 #include <linux/interrupt.h>
40 #include <linux/proc_fs.h>
41 #include <linux/stat.h>
42 #include <linux/init.h>
43 
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <linux/icmpv6.h>
47 #include <linux/smp_lock.h>
48 #include <linux/netfilter_ipv6.h>
49 
50 #include <net/ip.h>
51 #include <net/ipv6.h>
52 #include <net/udp.h>
53 #include <net/tcp.h>
54 #include <net/ipip.h>
55 #include <net/protocol.h>
56 #include <net/inet_common.h>
57 #include <net/transp_v6.h>
58 #include <net/ip6_route.h>
59 #include <net/addrconf.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 
64 #include <asm/uaccess.h>
65 #include <asm/system.h>
66 
67 MODULE_AUTHOR("Cast of dozens");
68 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
69 MODULE_LICENSE("GPL");
70 
71 int sysctl_ipv6_bindv6only;
72 
73 /* The inetsw table contains everything that inet_create needs to
74  * build a new socket.
75  */
76 static struct list_head inetsw6[SOCK_MAX];
77 static DEFINE_SPINLOCK(inetsw6_lock);
78 
79 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
80 {
81 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
82 
83 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
84 }
85 
86 static int inet6_create(struct socket *sock, int protocol)
87 {
88 	struct inet_sock *inet;
89 	struct ipv6_pinfo *np;
90 	struct sock *sk;
91 	struct list_head *p;
92 	struct inet_protosw *answer;
93 	struct proto *answer_prot;
94 	unsigned char answer_flags;
95 	char answer_no_check;
96 	int try_loading_module = 0;
97 	int err;
98 
99 	/* Look for the requested type/protocol pair. */
100 	answer = NULL;
101 lookup_protocol:
102 	err = -ESOCKTNOSUPPORT;
103 	rcu_read_lock();
104 	list_for_each_rcu(p, &inetsw6[sock->type]) {
105 		answer = list_entry(p, struct inet_protosw, list);
106 
107 		/* Check the non-wild match. */
108 		if (protocol == answer->protocol) {
109 			if (protocol != IPPROTO_IP)
110 				break;
111 		} else {
112 			/* Check for the two wild cases. */
113 			if (IPPROTO_IP == protocol) {
114 				protocol = answer->protocol;
115 				break;
116 			}
117 			if (IPPROTO_IP == answer->protocol)
118 				break;
119 		}
120 		err = -EPROTONOSUPPORT;
121 		answer = NULL;
122 	}
123 
124 	if (!answer) {
125 		if (try_loading_module < 2) {
126 			rcu_read_unlock();
127 			/*
128 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
129 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
130 			 */
131 			if (++try_loading_module == 1)
132 				request_module("net-pf-%d-proto-%d-type-%d",
133 						PF_INET6, protocol, sock->type);
134 			/*
135 			 * Fall back to generic, e.g. net-pf-10-proto-132
136 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
137 			 */
138 			else
139 				request_module("net-pf-%d-proto-%d",
140 						PF_INET6, protocol);
141 			goto lookup_protocol;
142 		} else
143 			goto out_rcu_unlock;
144 	}
145 
146 	err = -EPERM;
147 	if (answer->capability > 0 && !capable(answer->capability))
148 		goto out_rcu_unlock;
149 
150 	sock->ops = answer->ops;
151 	answer_prot = answer->prot;
152 	answer_no_check = answer->no_check;
153 	answer_flags = answer->flags;
154 	rcu_read_unlock();
155 
156 	BUG_TRAP(answer_prot->slab != NULL);
157 
158 	err = -ENOBUFS;
159 	sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
160 	if (sk == NULL)
161 		goto out;
162 
163 	sock_init_data(sock, sk);
164 
165 	err = 0;
166 	sk->sk_no_check = answer_no_check;
167 	if (INET_PROTOSW_REUSE & answer_flags)
168 		sk->sk_reuse = 1;
169 
170 	inet = inet_sk(sk);
171 	inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
172 
173 	if (SOCK_RAW == sock->type) {
174 		inet->num = protocol;
175 		if (IPPROTO_RAW == protocol)
176 			inet->hdrincl = 1;
177 	}
178 
179 	sk->sk_destruct		= inet_sock_destruct;
180 	sk->sk_family		= PF_INET6;
181 	sk->sk_protocol		= protocol;
182 
183 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
184 
185 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
186 	np->hop_limit	= -1;
187 	np->mcast_hops	= -1;
188 	np->mc_loop	= 1;
189 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
190 	np->ipv6only	= sysctl_ipv6_bindv6only;
191 
192 	/* Init the ipv4 part of the socket since we can have sockets
193 	 * using v6 API for ipv4.
194 	 */
195 	inet->uc_ttl	= -1;
196 
197 	inet->mc_loop	= 1;
198 	inet->mc_ttl	= 1;
199 	inet->mc_index	= 0;
200 	inet->mc_list	= NULL;
201 
202 	if (ipv4_config.no_pmtu_disc)
203 		inet->pmtudisc = IP_PMTUDISC_DONT;
204 	else
205 		inet->pmtudisc = IP_PMTUDISC_WANT;
206 	/*
207 	 * Increment only the relevant sk_prot->socks debug field, this changes
208 	 * the previous behaviour of incrementing both the equivalent to
209 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
210 	 *
211 	 * This allows better debug granularity as we'll know exactly how many
212 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
213 	 * transport protocol socks. -acme
214 	 */
215 	sk_refcnt_debug_inc(sk);
216 
217 	if (inet->num) {
218 		/* It assumes that any protocol which allows
219 		 * the user to assign a number at socket
220 		 * creation time automatically shares.
221 		 */
222 		inet->sport = ntohs(inet->num);
223 		sk->sk_prot->hash(sk);
224 	}
225 	if (sk->sk_prot->init) {
226 		err = sk->sk_prot->init(sk);
227 		if (err) {
228 			sk_common_release(sk);
229 			goto out;
230 		}
231 	}
232 out:
233 	return err;
234 out_rcu_unlock:
235 	rcu_read_unlock();
236 	goto out;
237 }
238 
239 
240 /* bind for INET6 API */
241 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
242 {
243 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
244 	struct sock *sk = sock->sk;
245 	struct inet_sock *inet = inet_sk(sk);
246 	struct ipv6_pinfo *np = inet6_sk(sk);
247 	__u32 v4addr = 0;
248 	unsigned short snum;
249 	int addr_type = 0;
250 	int err = 0;
251 
252 	/* If the socket has its own bind function then use it. */
253 	if (sk->sk_prot->bind)
254 		return sk->sk_prot->bind(sk, uaddr, addr_len);
255 
256 	if (addr_len < SIN6_LEN_RFC2133)
257 		return -EINVAL;
258 	addr_type = ipv6_addr_type(&addr->sin6_addr);
259 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
260 		return -EINVAL;
261 
262 	snum = ntohs(addr->sin6_port);
263 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
264 		return -EACCES;
265 
266 	lock_sock(sk);
267 
268 	/* Check these errors (active socket, double bind). */
269 	if (sk->sk_state != TCP_CLOSE || inet->num) {
270 		err = -EINVAL;
271 		goto out;
272 	}
273 
274 	/* Check if the address belongs to the host. */
275 	if (addr_type == IPV6_ADDR_MAPPED) {
276 		v4addr = addr->sin6_addr.s6_addr32[3];
277 		if (inet_addr_type(v4addr) != RTN_LOCAL) {
278 			err = -EADDRNOTAVAIL;
279 			goto out;
280 		}
281 	} else {
282 		if (addr_type != IPV6_ADDR_ANY) {
283 			struct net_device *dev = NULL;
284 
285 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
286 				if (addr_len >= sizeof(struct sockaddr_in6) &&
287 				    addr->sin6_scope_id) {
288 					/* Override any existing binding, if another one
289 					 * is supplied by user.
290 					 */
291 					sk->sk_bound_dev_if = addr->sin6_scope_id;
292 				}
293 
294 				/* Binding to link-local address requires an interface */
295 				if (!sk->sk_bound_dev_if) {
296 					err = -EINVAL;
297 					goto out;
298 				}
299 				dev = dev_get_by_index(sk->sk_bound_dev_if);
300 				if (!dev) {
301 					err = -ENODEV;
302 					goto out;
303 				}
304 			}
305 
306 			/* ipv4 addr of the socket is invalid.  Only the
307 			 * unspecified and mapped address have a v4 equivalent.
308 			 */
309 			v4addr = LOOPBACK4_IPV6;
310 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
311 				if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
312 					if (dev)
313 						dev_put(dev);
314 					err = -EADDRNOTAVAIL;
315 					goto out;
316 				}
317 			}
318 			if (dev)
319 				dev_put(dev);
320 		}
321 	}
322 
323 	inet->rcv_saddr = v4addr;
324 	inet->saddr = v4addr;
325 
326 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
327 
328 	if (!(addr_type & IPV6_ADDR_MULTICAST))
329 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
330 
331 	/* Make sure we are allowed to bind here. */
332 	if (sk->sk_prot->get_port(sk, snum)) {
333 		inet_reset_saddr(sk);
334 		err = -EADDRINUSE;
335 		goto out;
336 	}
337 
338 	if (addr_type != IPV6_ADDR_ANY)
339 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
340 	if (snum)
341 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
342 	inet->sport = ntohs(inet->num);
343 	inet->dport = 0;
344 	inet->daddr = 0;
345 out:
346 	release_sock(sk);
347 	return err;
348 }
349 
350 int inet6_release(struct socket *sock)
351 {
352 	struct sock *sk = sock->sk;
353 
354 	if (sk == NULL)
355 		return -EINVAL;
356 
357 	/* Free mc lists */
358 	ipv6_sock_mc_close(sk);
359 
360 	/* Free ac lists */
361 	ipv6_sock_ac_close(sk);
362 
363 	return inet_release(sock);
364 }
365 
366 int inet6_destroy_sock(struct sock *sk)
367 {
368 	struct ipv6_pinfo *np = inet6_sk(sk);
369 	struct sk_buff *skb;
370 	struct ipv6_txoptions *opt;
371 
372 	/* Release rx options */
373 
374 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
375 		kfree_skb(skb);
376 
377 	/* Free flowlabels */
378 	fl6_free_socklist(sk);
379 
380 	/* Free tx options */
381 
382 	if ((opt = xchg(&np->opt, NULL)) != NULL)
383 		sock_kfree_s(sk, opt, opt->tot_len);
384 
385 	return 0;
386 }
387 
388 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
389 
390 /*
391  *	This does both peername and sockname.
392  */
393 
394 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
395 		 int *uaddr_len, int peer)
396 {
397 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
398 	struct sock *sk = sock->sk;
399 	struct inet_sock *inet = inet_sk(sk);
400 	struct ipv6_pinfo *np = inet6_sk(sk);
401 
402 	sin->sin6_family = AF_INET6;
403 	sin->sin6_flowinfo = 0;
404 	sin->sin6_scope_id = 0;
405 	if (peer) {
406 		if (!inet->dport)
407 			return -ENOTCONN;
408 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
409 		    peer == 1)
410 			return -ENOTCONN;
411 		sin->sin6_port = inet->dport;
412 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
413 		if (np->sndflow)
414 			sin->sin6_flowinfo = np->flow_label;
415 	} else {
416 		if (ipv6_addr_any(&np->rcv_saddr))
417 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
418 		else
419 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
420 
421 		sin->sin6_port = inet->sport;
422 	}
423 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
424 		sin->sin6_scope_id = sk->sk_bound_dev_if;
425 	*uaddr_len = sizeof(*sin);
426 	return(0);
427 }
428 
429 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
430 {
431 	struct sock *sk = sock->sk;
432 
433 	switch(cmd)
434 	{
435 	case SIOCGSTAMP:
436 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
437 
438 	case SIOCADDRT:
439 	case SIOCDELRT:
440 
441 		return(ipv6_route_ioctl(cmd,(void __user *)arg));
442 
443 	case SIOCSIFADDR:
444 		return addrconf_add_ifaddr((void __user *) arg);
445 	case SIOCDIFADDR:
446 		return addrconf_del_ifaddr((void __user *) arg);
447 	case SIOCSIFDSTADDR:
448 		return addrconf_set_dstaddr((void __user *) arg);
449 	default:
450 		if (!sk->sk_prot->ioctl)
451 			return -ENOIOCTLCMD;
452 		return sk->sk_prot->ioctl(sk, cmd, arg);
453 	}
454 	/*NOTREACHED*/
455 	return(0);
456 }
457 
458 const struct proto_ops inet6_stream_ops = {
459 	.family		   = PF_INET6,
460 	.owner		   = THIS_MODULE,
461 	.release	   = inet6_release,
462 	.bind		   = inet6_bind,
463 	.connect	   = inet_stream_connect,	/* ok		*/
464 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
465 	.accept		   = inet_accept,		/* ok		*/
466 	.getname	   = inet6_getname,
467 	.poll		   = tcp_poll,			/* ok		*/
468 	.ioctl		   = inet6_ioctl,		/* must change  */
469 	.listen		   = inet_listen,		/* ok		*/
470 	.shutdown	   = inet_shutdown,		/* ok		*/
471 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
472 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
473 	.sendmsg	   = inet_sendmsg,		/* ok		*/
474 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
475 	.mmap		   = sock_no_mmap,
476 	.sendpage	   = tcp_sendpage,
477 #ifdef CONFIG_COMPAT
478 	.compat_setsockopt = compat_sock_common_setsockopt,
479 	.compat_getsockopt = compat_sock_common_getsockopt,
480 #endif
481 };
482 
483 const struct proto_ops inet6_dgram_ops = {
484 	.family		   = PF_INET6,
485 	.owner		   = THIS_MODULE,
486 	.release	   = inet6_release,
487 	.bind		   = inet6_bind,
488 	.connect	   = inet_dgram_connect,	/* ok		*/
489 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
490 	.accept		   = sock_no_accept,		/* a do nothing	*/
491 	.getname	   = inet6_getname,
492 	.poll		   = udp_poll,			/* ok		*/
493 	.ioctl		   = inet6_ioctl,		/* must change  */
494 	.listen		   = sock_no_listen,		/* ok		*/
495 	.shutdown	   = inet_shutdown,		/* ok		*/
496 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
497 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
498 	.sendmsg	   = inet_sendmsg,		/* ok		*/
499 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
500 	.mmap		   = sock_no_mmap,
501 	.sendpage	   = sock_no_sendpage,
502 #ifdef CONFIG_COMPAT
503 	.compat_setsockopt = compat_sock_common_setsockopt,
504 	.compat_getsockopt = compat_sock_common_getsockopt,
505 #endif
506 };
507 
508 static struct net_proto_family inet6_family_ops = {
509 	.family = PF_INET6,
510 	.create = inet6_create,
511 	.owner	= THIS_MODULE,
512 };
513 
514 /* Same as inet6_dgram_ops, sans udp_poll.  */
515 static const struct proto_ops inet6_sockraw_ops = {
516 	.family		   = PF_INET6,
517 	.owner		   = THIS_MODULE,
518 	.release	   = inet6_release,
519 	.bind		   = inet6_bind,
520 	.connect	   = inet_dgram_connect,	/* ok		*/
521 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
522 	.accept		   = sock_no_accept,		/* a do nothing	*/
523 	.getname	   = inet6_getname,
524 	.poll		   = datagram_poll,		/* ok		*/
525 	.ioctl		   = inet6_ioctl,		/* must change  */
526 	.listen		   = sock_no_listen,		/* ok		*/
527 	.shutdown	   = inet_shutdown,		/* ok		*/
528 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
529 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
530 	.sendmsg	   = inet_sendmsg,		/* ok		*/
531 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
532 	.mmap		   = sock_no_mmap,
533 	.sendpage	   = sock_no_sendpage,
534 #ifdef CONFIG_COMPAT
535 	.compat_setsockopt = compat_sock_common_setsockopt,
536 	.compat_getsockopt = compat_sock_common_getsockopt,
537 #endif
538 };
539 
540 static struct inet_protosw rawv6_protosw = {
541 	.type		= SOCK_RAW,
542 	.protocol	= IPPROTO_IP,	/* wild card */
543 	.prot		= &rawv6_prot,
544 	.ops		= &inet6_sockraw_ops,
545 	.capability	= CAP_NET_RAW,
546 	.no_check	= UDP_CSUM_DEFAULT,
547 	.flags		= INET_PROTOSW_REUSE,
548 };
549 
550 void
551 inet6_register_protosw(struct inet_protosw *p)
552 {
553 	struct list_head *lh;
554 	struct inet_protosw *answer;
555 	int protocol = p->protocol;
556 	struct list_head *last_perm;
557 
558 	spin_lock_bh(&inetsw6_lock);
559 
560 	if (p->type >= SOCK_MAX)
561 		goto out_illegal;
562 
563 	/* If we are trying to override a permanent protocol, bail. */
564 	answer = NULL;
565 	last_perm = &inetsw6[p->type];
566 	list_for_each(lh, &inetsw6[p->type]) {
567 		answer = list_entry(lh, struct inet_protosw, list);
568 
569 		/* Check only the non-wild match. */
570 		if (INET_PROTOSW_PERMANENT & answer->flags) {
571 			if (protocol == answer->protocol)
572 				break;
573 			last_perm = lh;
574 		}
575 
576 		answer = NULL;
577 	}
578 	if (answer)
579 		goto out_permanent;
580 
581 	/* Add the new entry after the last permanent entry if any, so that
582 	 * the new entry does not override a permanent entry when matched with
583 	 * a wild-card protocol. But it is allowed to override any existing
584 	 * non-permanent entry.  This means that when we remove this entry, the
585 	 * system automatically returns to the old behavior.
586 	 */
587 	list_add_rcu(&p->list, last_perm);
588 out:
589 	spin_unlock_bh(&inetsw6_lock);
590 	return;
591 
592 out_permanent:
593 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
594 	       protocol);
595 	goto out;
596 
597 out_illegal:
598 	printk(KERN_ERR
599 	       "Ignoring attempt to register invalid socket type %d.\n",
600 	       p->type);
601 	goto out;
602 }
603 
604 void
605 inet6_unregister_protosw(struct inet_protosw *p)
606 {
607 	if (INET_PROTOSW_PERMANENT & p->flags) {
608 		printk(KERN_ERR
609 		       "Attempt to unregister permanent protocol %d.\n",
610 		       p->protocol);
611 	} else {
612 		spin_lock_bh(&inetsw6_lock);
613 		list_del_rcu(&p->list);
614 		spin_unlock_bh(&inetsw6_lock);
615 
616 		synchronize_net();
617 	}
618 }
619 
620 int inet6_sk_rebuild_header(struct sock *sk)
621 {
622 	int err;
623 	struct dst_entry *dst;
624 	struct ipv6_pinfo *np = inet6_sk(sk);
625 
626 	dst = __sk_dst_check(sk, np->dst_cookie);
627 
628 	if (dst == NULL) {
629 		struct inet_sock *inet = inet_sk(sk);
630 		struct in6_addr *final_p = NULL, final;
631 		struct flowi fl;
632 
633 		memset(&fl, 0, sizeof(fl));
634 		fl.proto = sk->sk_protocol;
635 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
636 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
637 		fl.fl6_flowlabel = np->flow_label;
638 		fl.oif = sk->sk_bound_dev_if;
639 		fl.fl_ip_dport = inet->dport;
640 		fl.fl_ip_sport = inet->sport;
641 
642 		if (np->opt && np->opt->srcrt) {
643 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
644 			ipv6_addr_copy(&final, &fl.fl6_dst);
645 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
646 			final_p = &final;
647 		}
648 
649 		err = ip6_dst_lookup(sk, &dst, &fl);
650 		if (err) {
651 			sk->sk_route_caps = 0;
652 			return err;
653 		}
654 		if (final_p)
655 			ipv6_addr_copy(&fl.fl6_dst, final_p);
656 
657 		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
658 			sk->sk_err_soft = -err;
659 			return err;
660 		}
661 
662 		ip6_dst_store(sk, dst, NULL);
663 		sk->sk_route_caps = dst->dev->features &
664 			~(NETIF_F_IP_CSUM | NETIF_F_TSO);
665 	}
666 
667 	return 0;
668 }
669 
670 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
671 
672 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
673 {
674 	struct ipv6_pinfo *np = inet6_sk(sk);
675 	struct inet6_skb_parm *opt = IP6CB(skb);
676 
677 	if (np->rxopt.all) {
678 		if ((opt->hop && (np->rxopt.bits.hopopts ||
679 				  np->rxopt.bits.ohopopts)) ||
680 		    ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
681 		     np->rxopt.bits.rxflow) ||
682 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
683 		     np->rxopt.bits.osrcrt)) ||
684 		    ((opt->dst1 || opt->dst0) &&
685 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
686 			return 1;
687 	}
688 	return 0;
689 }
690 
691 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
692 
693 int
694 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
695 {
696 	if (ptr == NULL)
697 		return -EINVAL;
698 
699 	ptr[0] = __alloc_percpu(mibsize);
700 	if (!ptr[0])
701 		goto err0;
702 
703 	ptr[1] = __alloc_percpu(mibsize);
704 	if (!ptr[1])
705 		goto err1;
706 
707 	return 0;
708 
709 err1:
710 	free_percpu(ptr[0]);
711 	ptr[0] = NULL;
712 err0:
713 	return -ENOMEM;
714 }
715 
716 void
717 snmp6_mib_free(void *ptr[2])
718 {
719 	if (ptr == NULL)
720 		return;
721 	if (ptr[0])
722 		free_percpu(ptr[0]);
723 	if (ptr[1])
724 		free_percpu(ptr[1]);
725 	ptr[0] = ptr[1] = NULL;
726 }
727 
728 static int __init init_ipv6_mibs(void)
729 {
730 	if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
731 			   __alignof__(struct ipstats_mib)) < 0)
732 		goto err_ip_mib;
733 	if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
734 			   __alignof__(struct icmpv6_mib)) < 0)
735 		goto err_icmp_mib;
736 	if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
737 			   __alignof__(struct udp_mib)) < 0)
738 		goto err_udp_mib;
739 	return 0;
740 
741 err_udp_mib:
742 	snmp6_mib_free((void **)icmpv6_statistics);
743 err_icmp_mib:
744 	snmp6_mib_free((void **)ipv6_statistics);
745 err_ip_mib:
746 	return -ENOMEM;
747 
748 }
749 
750 static void cleanup_ipv6_mibs(void)
751 {
752 	snmp6_mib_free((void **)ipv6_statistics);
753 	snmp6_mib_free((void **)icmpv6_statistics);
754 	snmp6_mib_free((void **)udp_stats_in6);
755 }
756 
757 static int __init inet6_init(void)
758 {
759 	struct sk_buff *dummy_skb;
760         struct list_head *r;
761 	int err;
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 	if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
773 		printk(KERN_CRIT "inet6_proto_init: size fault\n");
774 		return -EINVAL;
775 	}
776 
777 	err = proto_register(&tcpv6_prot, 1);
778 	if (err)
779 		goto out;
780 
781 	err = proto_register(&udpv6_prot, 1);
782 	if (err)
783 		goto out_unregister_tcp_proto;
784 
785 	err = proto_register(&rawv6_prot, 1);
786 	if (err)
787 		goto out_unregister_udp_proto;
788 
789 
790 	/* Register the socket-side information for inet6_create.  */
791 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
792 		INIT_LIST_HEAD(r);
793 
794 	/* We MUST register RAW sockets before we create the ICMP6,
795 	 * IGMP6, or NDISC control sockets.
796 	 */
797 	inet6_register_protosw(&rawv6_protosw);
798 
799 	/* Register the family here so that the init calls below will
800 	 * be able to create sockets. (?? is this dangerous ??)
801 	 */
802 	err = sock_register(&inet6_family_ops);
803 	if (err)
804 		goto out_unregister_raw_proto;
805 
806 	/* Initialise ipv6 mibs */
807 	err = init_ipv6_mibs();
808 	if (err)
809 		goto out_unregister_sock;
810 
811 	/*
812 	 *	ipngwg API draft makes clear that the correct semantics
813 	 *	for TCP and UDP is to consider one TCP and UDP instance
814 	 *	in a host availiable by both INET and INET6 APIs and
815 	 *	able to communicate via both network protocols.
816 	 */
817 
818 #ifdef CONFIG_SYSCTL
819 	ipv6_sysctl_register();
820 #endif
821 	err = icmpv6_init(&inet6_family_ops);
822 	if (err)
823 		goto icmp_fail;
824 	err = ndisc_init(&inet6_family_ops);
825 	if (err)
826 		goto ndisc_fail;
827 	err = igmp6_init(&inet6_family_ops);
828 	if (err)
829 		goto igmp_fail;
830 	err = ipv6_netfilter_init();
831 	if (err)
832 		goto netfilter_fail;
833 	/* Create /proc/foo6 entries. */
834 #ifdef CONFIG_PROC_FS
835 	err = -ENOMEM;
836 	if (raw6_proc_init())
837 		goto proc_raw6_fail;
838 	if (tcp6_proc_init())
839 		goto proc_tcp6_fail;
840 	if (udp6_proc_init())
841 		goto proc_udp6_fail;
842 	if (ipv6_misc_proc_init())
843 		goto proc_misc6_fail;
844 
845 	if (ac6_proc_init())
846 		goto proc_anycast6_fail;
847 	if (if6_proc_init())
848 		goto proc_if6_fail;
849 #endif
850 	ip6_route_init();
851 	ip6_flowlabel_init();
852 	err = addrconf_init();
853 	if (err)
854 		goto addrconf_fail;
855 	sit_init();
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 	tcpv6_init();
866 
867 	ipv6_packet_init();
868 	err = 0;
869 out:
870 	return err;
871 
872 addrconf_fail:
873 	ip6_flowlabel_cleanup();
874 	ip6_route_cleanup();
875 #ifdef CONFIG_PROC_FS
876 	if6_proc_exit();
877 proc_if6_fail:
878 	ac6_proc_exit();
879 proc_anycast6_fail:
880 	ipv6_misc_proc_exit();
881 proc_misc6_fail:
882 	udp6_proc_exit();
883 proc_udp6_fail:
884 	tcp6_proc_exit();
885 proc_tcp6_fail:
886 	raw6_proc_exit();
887 proc_raw6_fail:
888 #endif
889 	ipv6_netfilter_fini();
890 netfilter_fail:
891 	igmp6_cleanup();
892 igmp_fail:
893 	ndisc_cleanup();
894 ndisc_fail:
895 	icmpv6_cleanup();
896 icmp_fail:
897 #ifdef CONFIG_SYSCTL
898 	ipv6_sysctl_unregister();
899 #endif
900 	cleanup_ipv6_mibs();
901 out_unregister_sock:
902 	sock_unregister(PF_INET6);
903 out_unregister_raw_proto:
904 	proto_unregister(&rawv6_prot);
905 out_unregister_udp_proto:
906 	proto_unregister(&udpv6_prot);
907 out_unregister_tcp_proto:
908 	proto_unregister(&tcpv6_prot);
909 	goto out;
910 }
911 module_init(inet6_init);
912 
913 static void __exit inet6_exit(void)
914 {
915 	/* First of all disallow new sockets creation. */
916 	sock_unregister(PF_INET6);
917 #ifdef CONFIG_PROC_FS
918 	if6_proc_exit();
919 	ac6_proc_exit();
920  	ipv6_misc_proc_exit();
921  	udp6_proc_exit();
922  	tcp6_proc_exit();
923  	raw6_proc_exit();
924 #endif
925 	/* Cleanup code parts. */
926 	sit_cleanup();
927 	ip6_flowlabel_cleanup();
928 	addrconf_cleanup();
929 	ip6_route_cleanup();
930 	ipv6_packet_cleanup();
931 	igmp6_cleanup();
932 	ipv6_netfilter_fini();
933 	ndisc_cleanup();
934 	icmpv6_cleanup();
935 #ifdef CONFIG_SYSCTL
936 	ipv6_sysctl_unregister();
937 #endif
938 	cleanup_ipv6_mibs();
939 	proto_unregister(&rawv6_prot);
940 	proto_unregister(&udpv6_prot);
941 	proto_unregister(&tcpv6_prot);
942 }
943 module_exit(inet6_exit);
944 
945 MODULE_ALIAS_NETPROTO(PF_INET6);
946