xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 87c2ce3b)
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>
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>
47 #include <linux/netfilter_ipv6.h>
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 
70 int sysctl_ipv6_bindv6only;
71 
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 
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;
95 	int try_loading_module = 0;
96 	int err;
97 
98 	/* Look for the requested type/protocol pair. */
99 	answer = NULL;
100 lookup_protocol:
101 	err = -ESOCKTNOSUPPORT;
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 		}
119 		err = -EPROTONOSUPPORT;
120 		answer = NULL;
121 	}
122 
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;
146 	if (answer->capability > 0 && !capable(answer->capability))
147 		goto out_rcu_unlock;
148 
149 	sock->ops = answer->ops;
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 
157 	err = -ENOBUFS;
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 
164 	err = 0;
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 	inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
171 
172 	if (SOCK_RAW == sock->type) {
173 		inet->num = protocol;
174 		if (IPPROTO_RAW == protocol)
175 			inet->hdrincl = 1;
176 	}
177 
178 	sk->sk_destruct		= inet_sock_destruct;
179 	sk->sk_family		= PF_INET6;
180 	sk->sk_protocol		= protocol;
181 
182 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
183 
184 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
185 	np->hop_limit	= -1;
186 	np->mcast_hops	= -1;
187 	np->mc_loop	= 1;
188 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
189 	np->ipv6only	= sysctl_ipv6_bindv6only;
190 
191 	/* Init the ipv4 part of the socket since we can have sockets
192 	 * using v6 API for ipv4.
193 	 */
194 	inet->uc_ttl	= -1;
195 
196 	inet->mc_loop	= 1;
197 	inet->mc_ttl	= 1;
198 	inet->mc_index	= 0;
199 	inet->mc_list	= NULL;
200 
201 	if (ipv4_config.no_pmtu_disc)
202 		inet->pmtudisc = IP_PMTUDISC_DONT;
203 	else
204 		inet->pmtudisc = IP_PMTUDISC_WANT;
205 	/*
206 	 * Increment only the relevant sk_prot->socks debug field, this changes
207 	 * the previous behaviour of incrementing both the equivalent to
208 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
209 	 *
210 	 * This allows better debug granularity as we'll know exactly how many
211 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
212 	 * transport protocol socks. -acme
213 	 */
214 	sk_refcnt_debug_inc(sk);
215 
216 	if (inet->num) {
217 		/* It assumes that any protocol which allows
218 		 * the user to assign a number at socket
219 		 * creation time automatically shares.
220 		 */
221 		inet->sport = ntohs(inet->num);
222 		sk->sk_prot->hash(sk);
223 	}
224 	if (sk->sk_prot->init) {
225 		err = sk->sk_prot->init(sk);
226 		if (err) {
227 			sk_common_release(sk);
228 			goto out;
229 		}
230 	}
231 out:
232 	return err;
233 out_rcu_unlock:
234 	rcu_read_unlock();
235 	goto out;
236 }
237 
238 
239 /* bind for INET6 API */
240 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
241 {
242 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
243 	struct sock *sk = sock->sk;
244 	struct inet_sock *inet = inet_sk(sk);
245 	struct ipv6_pinfo *np = inet6_sk(sk);
246 	__u32 v4addr = 0;
247 	unsigned short snum;
248 	int addr_type = 0;
249 	int err = 0;
250 
251 	/* If the socket has its own bind function then use it. */
252 	if (sk->sk_prot->bind)
253 		return sk->sk_prot->bind(sk, uaddr, addr_len);
254 
255 	if (addr_len < SIN6_LEN_RFC2133)
256 		return -EINVAL;
257 	addr_type = ipv6_addr_type(&addr->sin6_addr);
258 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
259 		return -EINVAL;
260 
261 	snum = ntohs(addr->sin6_port);
262 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
263 		return -EACCES;
264 
265 	lock_sock(sk);
266 
267 	/* Check these errors (active socket, double bind). */
268 	if (sk->sk_state != TCP_CLOSE || inet->num) {
269 		err = -EINVAL;
270 		goto out;
271 	}
272 
273 	/* Check if the address belongs to the host. */
274 	if (addr_type == IPV6_ADDR_MAPPED) {
275 		v4addr = addr->sin6_addr.s6_addr32[3];
276 		if (inet_addr_type(v4addr) != RTN_LOCAL) {
277 			err = -EADDRNOTAVAIL;
278 			goto out;
279 		}
280 	} else {
281 		if (addr_type != IPV6_ADDR_ANY) {
282 			struct net_device *dev = NULL;
283 
284 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
285 				if (addr_len >= sizeof(struct sockaddr_in6) &&
286 				    addr->sin6_scope_id) {
287 					/* Override any existing binding, if another one
288 					 * is supplied by user.
289 					 */
290 					sk->sk_bound_dev_if = addr->sin6_scope_id;
291 				}
292 
293 				/* Binding to link-local address requires an interface */
294 				if (!sk->sk_bound_dev_if) {
295 					err = -EINVAL;
296 					goto out;
297 				}
298 				dev = dev_get_by_index(sk->sk_bound_dev_if);
299 				if (!dev) {
300 					err = -ENODEV;
301 					goto out;
302 				}
303 			}
304 
305 			/* ipv4 addr of the socket is invalid.  Only the
306 			 * unspecified and mapped address have a v4 equivalent.
307 			 */
308 			v4addr = LOOPBACK4_IPV6;
309 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
310 				if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
311 					if (dev)
312 						dev_put(dev);
313 					err = -EADDRNOTAVAIL;
314 					goto out;
315 				}
316 			}
317 			if (dev)
318 				dev_put(dev);
319 		}
320 	}
321 
322 	inet->rcv_saddr = v4addr;
323 	inet->saddr = v4addr;
324 
325 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
326 
327 	if (!(addr_type & IPV6_ADDR_MULTICAST))
328 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
329 
330 	/* Make sure we are allowed to bind here. */
331 	if (sk->sk_prot->get_port(sk, snum)) {
332 		inet_reset_saddr(sk);
333 		err = -EADDRINUSE;
334 		goto out;
335 	}
336 
337 	if (addr_type != IPV6_ADDR_ANY)
338 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
339 	if (snum)
340 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
341 	inet->sport = ntohs(inet->num);
342 	inet->dport = 0;
343 	inet->daddr = 0;
344 out:
345 	release_sock(sk);
346 	return err;
347 }
348 
349 int inet6_release(struct socket *sock)
350 {
351 	struct sock *sk = sock->sk;
352 
353 	if (sk == NULL)
354 		return -EINVAL;
355 
356 	/* Free mc lists */
357 	ipv6_sock_mc_close(sk);
358 
359 	/* Free ac lists */
360 	ipv6_sock_ac_close(sk);
361 
362 	return inet_release(sock);
363 }
364 
365 int inet6_destroy_sock(struct sock *sk)
366 {
367 	struct ipv6_pinfo *np = inet6_sk(sk);
368 	struct sk_buff *skb;
369 	struct ipv6_txoptions *opt;
370 
371 	/*
372 	 *	Release destination entry
373 	 */
374 
375 	sk_dst_reset(sk);
376 
377 	/* Release rx options */
378 
379 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
380 		kfree_skb(skb);
381 
382 	/* Free flowlabels */
383 	fl6_free_socklist(sk);
384 
385 	/* Free tx options */
386 
387 	if ((opt = xchg(&np->opt, NULL)) != NULL)
388 		sock_kfree_s(sk, opt, opt->tot_len);
389 
390 	return 0;
391 }
392 
393 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
394 
395 /*
396  *	This does both peername and sockname.
397  */
398 
399 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
400 		 int *uaddr_len, int peer)
401 {
402 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
403 	struct sock *sk = sock->sk;
404 	struct inet_sock *inet = inet_sk(sk);
405 	struct ipv6_pinfo *np = inet6_sk(sk);
406 
407 	sin->sin6_family = AF_INET6;
408 	sin->sin6_flowinfo = 0;
409 	sin->sin6_scope_id = 0;
410 	if (peer) {
411 		if (!inet->dport)
412 			return -ENOTCONN;
413 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
414 		    peer == 1)
415 			return -ENOTCONN;
416 		sin->sin6_port = inet->dport;
417 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
418 		if (np->sndflow)
419 			sin->sin6_flowinfo = np->flow_label;
420 	} else {
421 		if (ipv6_addr_any(&np->rcv_saddr))
422 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
423 		else
424 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
425 
426 		sin->sin6_port = inet->sport;
427 	}
428 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
429 		sin->sin6_scope_id = sk->sk_bound_dev_if;
430 	*uaddr_len = sizeof(*sin);
431 	return(0);
432 }
433 
434 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
435 {
436 	struct sock *sk = sock->sk;
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 			return -ENOIOCTLCMD;
457 		return sk->sk_prot->ioctl(sk, cmd, arg);
458 	}
459 	/*NOTREACHED*/
460 	return(0);
461 }
462 
463 const struct proto_ops inet6_stream_ops = {
464 	.family =	PF_INET6,
465 	.owner =	THIS_MODULE,
466 	.release =	inet6_release,
467 	.bind =		inet6_bind,
468 	.connect =	inet_stream_connect,		/* ok		*/
469 	.socketpair =	sock_no_socketpair,		/* a do nothing	*/
470 	.accept =	inet_accept,			/* ok		*/
471 	.getname =	inet6_getname,
472 	.poll =		tcp_poll,			/* ok		*/
473 	.ioctl =	inet6_ioctl,			/* must change  */
474 	.listen =	inet_listen,			/* ok		*/
475 	.shutdown =	inet_shutdown,			/* ok		*/
476 	.setsockopt =	sock_common_setsockopt,		/* ok		*/
477 	.getsockopt =	sock_common_getsockopt,		/* ok		*/
478 	.sendmsg =	inet_sendmsg,			/* ok		*/
479 	.recvmsg =	sock_common_recvmsg,		/* ok		*/
480 	.mmap =		sock_no_mmap,
481 	.sendpage =	tcp_sendpage
482 };
483 
484 const struct proto_ops inet6_dgram_ops = {
485 	.family =	PF_INET6,
486 	.owner =	THIS_MODULE,
487 	.release =	inet6_release,
488 	.bind =		inet6_bind,
489 	.connect =	inet_dgram_connect,		/* ok		*/
490 	.socketpair =	sock_no_socketpair,		/* a do nothing	*/
491 	.accept =	sock_no_accept,			/* a do nothing	*/
492 	.getname =	inet6_getname,
493 	.poll =		udp_poll,			/* ok		*/
494 	.ioctl =	inet6_ioctl,			/* must change  */
495 	.listen =	sock_no_listen,			/* ok		*/
496 	.shutdown =	inet_shutdown,			/* ok		*/
497 	.setsockopt =	sock_common_setsockopt,		/* ok		*/
498 	.getsockopt =	sock_common_getsockopt,		/* ok		*/
499 	.sendmsg =	inet_sendmsg,			/* ok		*/
500 	.recvmsg =	sock_common_recvmsg,		/* ok		*/
501 	.mmap =		sock_no_mmap,
502 	.sendpage =	sock_no_sendpage,
503 };
504 
505 static struct net_proto_family inet6_family_ops = {
506 	.family = PF_INET6,
507 	.create = inet6_create,
508 	.owner	= THIS_MODULE,
509 };
510 
511 /* Same as inet6_dgram_ops, sans udp_poll.  */
512 static const struct proto_ops inet6_sockraw_ops = {
513 	.family =	PF_INET6,
514 	.owner =	THIS_MODULE,
515 	.release =	inet6_release,
516 	.bind =		inet6_bind,
517 	.connect =	inet_dgram_connect,		/* ok		*/
518 	.socketpair =	sock_no_socketpair,		/* a do nothing	*/
519 	.accept =	sock_no_accept,			/* a do nothing	*/
520 	.getname =	inet6_getname,
521 	.poll =		datagram_poll,			/* ok		*/
522 	.ioctl =	inet6_ioctl,			/* must change  */
523 	.listen =	sock_no_listen,			/* ok		*/
524 	.shutdown =	inet_shutdown,			/* ok		*/
525 	.setsockopt =	sock_common_setsockopt,		/* ok		*/
526 	.getsockopt =	sock_common_getsockopt,		/* ok		*/
527 	.sendmsg =	inet_sendmsg,			/* ok		*/
528 	.recvmsg =	sock_common_recvmsg,		/* ok		*/
529 	.mmap =		sock_no_mmap,
530 	.sendpage =	sock_no_sendpage,
531 };
532 
533 static struct inet_protosw rawv6_protosw = {
534 	.type		= SOCK_RAW,
535 	.protocol	= IPPROTO_IP,	/* wild card */
536 	.prot		= &rawv6_prot,
537 	.ops		= &inet6_sockraw_ops,
538 	.capability	= CAP_NET_RAW,
539 	.no_check	= UDP_CSUM_DEFAULT,
540 	.flags		= INET_PROTOSW_REUSE,
541 };
542 
543 void
544 inet6_register_protosw(struct inet_protosw *p)
545 {
546 	struct list_head *lh;
547 	struct inet_protosw *answer;
548 	int protocol = p->protocol;
549 	struct list_head *last_perm;
550 
551 	spin_lock_bh(&inetsw6_lock);
552 
553 	if (p->type >= SOCK_MAX)
554 		goto out_illegal;
555 
556 	/* If we are trying to override a permanent protocol, bail. */
557 	answer = NULL;
558 	last_perm = &inetsw6[p->type];
559 	list_for_each(lh, &inetsw6[p->type]) {
560 		answer = list_entry(lh, struct inet_protosw, list);
561 
562 		/* Check only the non-wild match. */
563 		if (INET_PROTOSW_PERMANENT & answer->flags) {
564 			if (protocol == answer->protocol)
565 				break;
566 			last_perm = lh;
567 		}
568 
569 		answer = NULL;
570 	}
571 	if (answer)
572 		goto out_permanent;
573 
574 	/* Add the new entry after the last permanent entry if any, so that
575 	 * the new entry does not override a permanent entry when matched with
576 	 * a wild-card protocol. But it is allowed to override any existing
577 	 * non-permanent entry.  This means that when we remove this entry, the
578 	 * system automatically returns to the old behavior.
579 	 */
580 	list_add_rcu(&p->list, last_perm);
581 out:
582 	spin_unlock_bh(&inetsw6_lock);
583 	return;
584 
585 out_permanent:
586 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
587 	       protocol);
588 	goto out;
589 
590 out_illegal:
591 	printk(KERN_ERR
592 	       "Ignoring attempt to register invalid socket type %d.\n",
593 	       p->type);
594 	goto out;
595 }
596 
597 void
598 inet6_unregister_protosw(struct inet_protosw *p)
599 {
600 	if (INET_PROTOSW_PERMANENT & p->flags) {
601 		printk(KERN_ERR
602 		       "Attempt to unregister permanent protocol %d.\n",
603 		       p->protocol);
604 	} else {
605 		spin_lock_bh(&inetsw6_lock);
606 		list_del_rcu(&p->list);
607 		spin_unlock_bh(&inetsw6_lock);
608 
609 		synchronize_net();
610 	}
611 }
612 
613 int inet6_sk_rebuild_header(struct sock *sk)
614 {
615 	int err;
616 	struct dst_entry *dst;
617 	struct ipv6_pinfo *np = inet6_sk(sk);
618 
619 	dst = __sk_dst_check(sk, np->dst_cookie);
620 
621 	if (dst == NULL) {
622 		struct inet_sock *inet = inet_sk(sk);
623 		struct in6_addr *final_p = NULL, final;
624 		struct flowi fl;
625 
626 		memset(&fl, 0, sizeof(fl));
627 		fl.proto = sk->sk_protocol;
628 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
629 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
630 		fl.fl6_flowlabel = np->flow_label;
631 		fl.oif = sk->sk_bound_dev_if;
632 		fl.fl_ip_dport = inet->dport;
633 		fl.fl_ip_sport = inet->sport;
634 
635 		if (np->opt && np->opt->srcrt) {
636 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
637 			ipv6_addr_copy(&final, &fl.fl6_dst);
638 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
639 			final_p = &final;
640 		}
641 
642 		err = ip6_dst_lookup(sk, &dst, &fl);
643 		if (err) {
644 			sk->sk_route_caps = 0;
645 			return err;
646 		}
647 		if (final_p)
648 			ipv6_addr_copy(&fl.fl6_dst, final_p);
649 
650 		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
651 			sk->sk_err_soft = -err;
652 			return err;
653 		}
654 
655 		ip6_dst_store(sk, dst, NULL);
656 		sk->sk_route_caps = dst->dev->features &
657 			~(NETIF_F_IP_CSUM | NETIF_F_TSO);
658 	}
659 
660 	return 0;
661 }
662 
663 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
664 
665 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
666 {
667 	struct ipv6_pinfo *np = inet6_sk(sk);
668 	struct inet6_skb_parm *opt = IP6CB(skb);
669 
670 	if (np->rxopt.all) {
671 		if ((opt->hop && (np->rxopt.bits.hopopts ||
672 				  np->rxopt.bits.ohopopts)) ||
673 		    ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
674 		     np->rxopt.bits.rxflow) ||
675 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
676 		     np->rxopt.bits.osrcrt)) ||
677 		    ((opt->dst1 || opt->dst0) &&
678 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
679 			return 1;
680 	}
681 	return 0;
682 }
683 
684 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
685 
686 int
687 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
688 {
689 	if (ptr == NULL)
690 		return -EINVAL;
691 
692 	ptr[0] = __alloc_percpu(mibsize);
693 	if (!ptr[0])
694 		goto err0;
695 
696 	ptr[1] = __alloc_percpu(mibsize);
697 	if (!ptr[1])
698 		goto err1;
699 
700 	return 0;
701 
702 err1:
703 	free_percpu(ptr[0]);
704 	ptr[0] = NULL;
705 err0:
706 	return -ENOMEM;
707 }
708 
709 void
710 snmp6_mib_free(void *ptr[2])
711 {
712 	if (ptr == NULL)
713 		return;
714 	if (ptr[0])
715 		free_percpu(ptr[0]);
716 	if (ptr[1])
717 		free_percpu(ptr[1]);
718 	ptr[0] = ptr[1] = NULL;
719 }
720 
721 static int __init init_ipv6_mibs(void)
722 {
723 	if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
724 			   __alignof__(struct ipstats_mib)) < 0)
725 		goto err_ip_mib;
726 	if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
727 			   __alignof__(struct icmpv6_mib)) < 0)
728 		goto err_icmp_mib;
729 	if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
730 			   __alignof__(struct udp_mib)) < 0)
731 		goto err_udp_mib;
732 	return 0;
733 
734 err_udp_mib:
735 	snmp6_mib_free((void **)icmpv6_statistics);
736 err_icmp_mib:
737 	snmp6_mib_free((void **)ipv6_statistics);
738 err_ip_mib:
739 	return -ENOMEM;
740 
741 }
742 
743 static void cleanup_ipv6_mibs(void)
744 {
745 	snmp6_mib_free((void **)ipv6_statistics);
746 	snmp6_mib_free((void **)icmpv6_statistics);
747 	snmp6_mib_free((void **)udp_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 #ifdef MODULE
757 #if 0 /* FIXME --RR */
758 	if (!mod_member_present(&__this_module, can_unload))
759 	  return -EINVAL;
760 
761 	__this_module.can_unload = &ipv6_unload;
762 #endif
763 #endif
764 
765 	if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
766 		printk(KERN_CRIT "inet6_proto_init: size fault\n");
767 		return -EINVAL;
768 	}
769 
770 	err = proto_register(&tcpv6_prot, 1);
771 	if (err)
772 		goto out;
773 
774 	err = proto_register(&udpv6_prot, 1);
775 	if (err)
776 		goto out_unregister_tcp_proto;
777 
778 	err = proto_register(&rawv6_prot, 1);
779 	if (err)
780 		goto out_unregister_udp_proto;
781 
782 
783 	/* Register the socket-side information for inet6_create.  */
784 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
785 		INIT_LIST_HEAD(r);
786 
787 	/* We MUST register RAW sockets before we create the ICMP6,
788 	 * IGMP6, or NDISC control sockets.
789 	 */
790 	inet6_register_protosw(&rawv6_protosw);
791 
792 	/* Register the family here so that the init calls below will
793 	 * be able to create sockets. (?? is this dangerous ??)
794 	 */
795 	err = sock_register(&inet6_family_ops);
796 	if (err)
797 		goto out_unregister_raw_proto;
798 
799 	/* Initialise ipv6 mibs */
800 	err = init_ipv6_mibs();
801 	if (err)
802 		goto out_unregister_sock;
803 
804 	/*
805 	 *	ipngwg API draft makes clear that the correct semantics
806 	 *	for TCP and UDP is to consider one TCP and UDP instance
807 	 *	in a host availiable by both INET and INET6 APIs and
808 	 *	able to communicate via both network protocols.
809 	 */
810 
811 #ifdef CONFIG_SYSCTL
812 	ipv6_sysctl_register();
813 #endif
814 	err = icmpv6_init(&inet6_family_ops);
815 	if (err)
816 		goto icmp_fail;
817 	err = ndisc_init(&inet6_family_ops);
818 	if (err)
819 		goto ndisc_fail;
820 	err = igmp6_init(&inet6_family_ops);
821 	if (err)
822 		goto igmp_fail;
823 	err = ipv6_netfilter_init();
824 	if (err)
825 		goto netfilter_fail;
826 	/* Create /proc/foo6 entries. */
827 #ifdef CONFIG_PROC_FS
828 	err = -ENOMEM;
829 	if (raw6_proc_init())
830 		goto proc_raw6_fail;
831 	if (tcp6_proc_init())
832 		goto proc_tcp6_fail;
833 	if (udp6_proc_init())
834 		goto proc_udp6_fail;
835 	if (ipv6_misc_proc_init())
836 		goto proc_misc6_fail;
837 
838 	if (ac6_proc_init())
839 		goto proc_anycast6_fail;
840 	if (if6_proc_init())
841 		goto proc_if6_fail;
842 #endif
843 	ip6_route_init();
844 	ip6_flowlabel_init();
845 	err = addrconf_init();
846 	if (err)
847 		goto addrconf_fail;
848 	sit_init();
849 
850 	/* Init v6 extension headers. */
851 	ipv6_rthdr_init();
852 	ipv6_frag_init();
853 	ipv6_nodata_init();
854 	ipv6_destopt_init();
855 
856 	/* Init v6 transport protocols. */
857 	udpv6_init();
858 	tcpv6_init();
859 
860 	ipv6_packet_init();
861 	err = 0;
862 out:
863 	return err;
864 
865 addrconf_fail:
866 	ip6_flowlabel_cleanup();
867 	ip6_route_cleanup();
868 #ifdef CONFIG_PROC_FS
869 	if6_proc_exit();
870 proc_if6_fail:
871 	ac6_proc_exit();
872 proc_anycast6_fail:
873 	ipv6_misc_proc_exit();
874 proc_misc6_fail:
875 	udp6_proc_exit();
876 proc_udp6_fail:
877 	tcp6_proc_exit();
878 proc_tcp6_fail:
879 	raw6_proc_exit();
880 proc_raw6_fail:
881 #endif
882 	ipv6_netfilter_fini();
883 netfilter_fail:
884 	igmp6_cleanup();
885 igmp_fail:
886 	ndisc_cleanup();
887 ndisc_fail:
888 	icmpv6_cleanup();
889 icmp_fail:
890 #ifdef CONFIG_SYSCTL
891 	ipv6_sysctl_unregister();
892 #endif
893 	cleanup_ipv6_mibs();
894 out_unregister_sock:
895 	sock_unregister(PF_INET6);
896 out_unregister_raw_proto:
897 	proto_unregister(&rawv6_prot);
898 out_unregister_udp_proto:
899 	proto_unregister(&udpv6_prot);
900 out_unregister_tcp_proto:
901 	proto_unregister(&tcpv6_prot);
902 	goto out;
903 }
904 module_init(inet6_init);
905 
906 static void __exit inet6_exit(void)
907 {
908 	/* First of all disallow new sockets creation. */
909 	sock_unregister(PF_INET6);
910 #ifdef CONFIG_PROC_FS
911 	if6_proc_exit();
912 	ac6_proc_exit();
913  	ipv6_misc_proc_exit();
914  	udp6_proc_exit();
915  	tcp6_proc_exit();
916  	raw6_proc_exit();
917 #endif
918 	/* Cleanup code parts. */
919 	sit_cleanup();
920 	ip6_flowlabel_cleanup();
921 	addrconf_cleanup();
922 	ip6_route_cleanup();
923 	ipv6_packet_cleanup();
924 	igmp6_cleanup();
925 	ipv6_netfilter_fini();
926 	ndisc_cleanup();
927 	icmpv6_cleanup();
928 #ifdef CONFIG_SYSCTL
929 	ipv6_sysctl_unregister();
930 #endif
931 	cleanup_ipv6_mibs();
932 	proto_unregister(&rawv6_prot);
933 	proto_unregister(&udpv6_prot);
934 	proto_unregister(&tcpv6_prot);
935 }
936 module_exit(inet6_exit);
937 
938 MODULE_ALIAS_NETPROTO(PF_INET6);
939