xref: /linux/net/ipv6/af_inet6.c (revision a33f32244d8550da8b4a26e277ce07d5c6d158b5)
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  * 	Fixes:
11  *	piggy, Karl Knutson	:	Socket protocol table
12  * 	Hideaki YOSHIFUJI	:	sin6_scope_id support
13  * 	Arnaldo Melo		: 	check proc_net_create return, cleanups
14  *
15  *	This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
20 
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/in.h>
28 #include <linux/kernel.h>
29 #include <linux/timer.h>
30 #include <linux/string.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/fcntl.h>
34 #include <linux/mm.h>
35 #include <linux/interrupt.h>
36 #include <linux/proc_fs.h>
37 #include <linux/stat.h>
38 #include <linux/init.h>
39 #include <linux/slab.h>
40 
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/icmpv6.h>
44 #include <linux/netfilter_ipv6.h>
45 
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/udp.h>
49 #include <net/udplite.h>
50 #include <net/tcp.h>
51 #include <net/ipip.h>
52 #include <net/protocol.h>
53 #include <net/inet_common.h>
54 #include <net/route.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 #include <linux/mroute6.h>
65 
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
69 
70 /* The inetsw6 table contains everything that inet6_create needs to
71  * build a new socket.
72  */
73 static struct list_head inetsw6[SOCK_MAX];
74 static DEFINE_SPINLOCK(inetsw6_lock);
75 
76 struct ipv6_params ipv6_defaults = {
77 	.disable_ipv6 = 0,
78 	.autoconf = 1,
79 };
80 
81 static int disable_ipv6_mod = 0;
82 
83 module_param_named(disable, disable_ipv6_mod, int, 0444);
84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
85 
86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
88 
89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
91 
92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
93 {
94 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
95 
96 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
97 }
98 
99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
100 			int kern)
101 {
102 	struct inet_sock *inet;
103 	struct ipv6_pinfo *np;
104 	struct sock *sk;
105 	struct inet_protosw *answer;
106 	struct proto *answer_prot;
107 	unsigned char answer_flags;
108 	char answer_no_check;
109 	int try_loading_module = 0;
110 	int err;
111 
112 	if (sock->type != SOCK_RAW &&
113 	    sock->type != SOCK_DGRAM &&
114 	    !inet_ehash_secret)
115 		build_ehash_secret();
116 
117 	/* Look for the requested type/protocol pair. */
118 lookup_protocol:
119 	err = -ESOCKTNOSUPPORT;
120 	rcu_read_lock();
121 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122 
123 		err = 0;
124 		/* Check the non-wild match. */
125 		if (protocol == answer->protocol) {
126 			if (protocol != IPPROTO_IP)
127 				break;
128 		} else {
129 			/* Check for the two wild cases. */
130 			if (IPPROTO_IP == protocol) {
131 				protocol = answer->protocol;
132 				break;
133 			}
134 			if (IPPROTO_IP == answer->protocol)
135 				break;
136 		}
137 		err = -EPROTONOSUPPORT;
138 	}
139 
140 	if (err) {
141 		if (try_loading_module < 2) {
142 			rcu_read_unlock();
143 			/*
144 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146 			 */
147 			if (++try_loading_module == 1)
148 				request_module("net-pf-%d-proto-%d-type-%d",
149 						PF_INET6, protocol, sock->type);
150 			/*
151 			 * Fall back to generic, e.g. net-pf-10-proto-132
152 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 			 */
154 			else
155 				request_module("net-pf-%d-proto-%d",
156 						PF_INET6, protocol);
157 			goto lookup_protocol;
158 		} else
159 			goto out_rcu_unlock;
160 	}
161 
162 	err = -EPERM;
163 	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
164 		goto out_rcu_unlock;
165 
166 	sock->ops = answer->ops;
167 	answer_prot = answer->prot;
168 	answer_no_check = answer->no_check;
169 	answer_flags = answer->flags;
170 	rcu_read_unlock();
171 
172 	WARN_ON(answer_prot->slab == NULL);
173 
174 	err = -ENOBUFS;
175 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
176 	if (sk == NULL)
177 		goto out;
178 
179 	sock_init_data(sock, sk);
180 
181 	err = 0;
182 	sk->sk_no_check = answer_no_check;
183 	if (INET_PROTOSW_REUSE & answer_flags)
184 		sk->sk_reuse = 1;
185 
186 	inet = inet_sk(sk);
187 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
188 
189 	if (SOCK_RAW == sock->type) {
190 		inet->inet_num = protocol;
191 		if (IPPROTO_RAW == protocol)
192 			inet->hdrincl = 1;
193 	}
194 
195 	sk->sk_destruct		= inet_sock_destruct;
196 	sk->sk_family		= PF_INET6;
197 	sk->sk_protocol		= protocol;
198 
199 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
200 
201 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
202 	np->hop_limit	= -1;
203 	np->mcast_hops	= -1;
204 	np->mc_loop	= 1;
205 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
206 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
207 
208 	/* Init the ipv4 part of the socket since we can have sockets
209 	 * using v6 API for ipv4.
210 	 */
211 	inet->uc_ttl	= -1;
212 
213 	inet->mc_loop	= 1;
214 	inet->mc_ttl	= 1;
215 	inet->mc_index	= 0;
216 	inet->mc_list	= NULL;
217 
218 	if (ipv4_config.no_pmtu_disc)
219 		inet->pmtudisc = IP_PMTUDISC_DONT;
220 	else
221 		inet->pmtudisc = IP_PMTUDISC_WANT;
222 	/*
223 	 * Increment only the relevant sk_prot->socks debug field, this changes
224 	 * the previous behaviour of incrementing both the equivalent to
225 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
226 	 *
227 	 * This allows better debug granularity as we'll know exactly how many
228 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
229 	 * transport protocol socks. -acme
230 	 */
231 	sk_refcnt_debug_inc(sk);
232 
233 	if (inet->inet_num) {
234 		/* It assumes that any protocol which allows
235 		 * the user to assign a number at socket
236 		 * creation time automatically shares.
237 		 */
238 		inet->inet_sport = htons(inet->inet_num);
239 		sk->sk_prot->hash(sk);
240 	}
241 	if (sk->sk_prot->init) {
242 		err = sk->sk_prot->init(sk);
243 		if (err) {
244 			sk_common_release(sk);
245 			goto out;
246 		}
247 	}
248 out:
249 	return err;
250 out_rcu_unlock:
251 	rcu_read_unlock();
252 	goto out;
253 }
254 
255 
256 /* bind for INET6 API */
257 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
258 {
259 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
260 	struct sock *sk = sock->sk;
261 	struct inet_sock *inet = inet_sk(sk);
262 	struct ipv6_pinfo *np = inet6_sk(sk);
263 	struct net *net = sock_net(sk);
264 	__be32 v4addr = 0;
265 	unsigned short snum;
266 	int addr_type = 0;
267 	int err = 0;
268 
269 	/* If the socket has its own bind function then use it. */
270 	if (sk->sk_prot->bind)
271 		return sk->sk_prot->bind(sk, uaddr, addr_len);
272 
273 	if (addr_len < SIN6_LEN_RFC2133)
274 		return -EINVAL;
275 	addr_type = ipv6_addr_type(&addr->sin6_addr);
276 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
277 		return -EINVAL;
278 
279 	snum = ntohs(addr->sin6_port);
280 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
281 		return -EACCES;
282 
283 	lock_sock(sk);
284 
285 	/* Check these errors (active socket, double bind). */
286 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
287 		err = -EINVAL;
288 		goto out;
289 	}
290 
291 	/* Check if the address belongs to the host. */
292 	if (addr_type == IPV6_ADDR_MAPPED) {
293 		int chk_addr_ret;
294 
295 		/* Binding to v4-mapped address on a v6-only socket
296 		 * makes no sense
297 		 */
298 		if (np->ipv6only) {
299 			err = -EINVAL;
300 			goto out;
301 		}
302 
303 		/* Reproduce AF_INET checks to make the bindings consitant */
304 		v4addr = addr->sin6_addr.s6_addr32[3];
305 		chk_addr_ret = inet_addr_type(net, v4addr);
306 		if (!sysctl_ip_nonlocal_bind &&
307 		    !(inet->freebind || inet->transparent) &&
308 		    v4addr != htonl(INADDR_ANY) &&
309 		    chk_addr_ret != RTN_LOCAL &&
310 		    chk_addr_ret != RTN_MULTICAST &&
311 		    chk_addr_ret != RTN_BROADCAST) {
312 			err = -EADDRNOTAVAIL;
313 			goto out;
314 		}
315 	} else {
316 		if (addr_type != IPV6_ADDR_ANY) {
317 			struct net_device *dev = NULL;
318 
319 			rcu_read_lock();
320 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
321 				if (addr_len >= sizeof(struct sockaddr_in6) &&
322 				    addr->sin6_scope_id) {
323 					/* Override any existing binding, if another one
324 					 * is supplied by user.
325 					 */
326 					sk->sk_bound_dev_if = addr->sin6_scope_id;
327 				}
328 
329 				/* Binding to link-local address requires an interface */
330 				if (!sk->sk_bound_dev_if) {
331 					err = -EINVAL;
332 					goto out_unlock;
333 				}
334 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
335 				if (!dev) {
336 					err = -ENODEV;
337 					goto out_unlock;
338 				}
339 			}
340 
341 			/* ipv4 addr of the socket is invalid.  Only the
342 			 * unspecified and mapped address have a v4 equivalent.
343 			 */
344 			v4addr = LOOPBACK4_IPV6;
345 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
346 				if (!ipv6_chk_addr(net, &addr->sin6_addr,
347 						   dev, 0)) {
348 					err = -EADDRNOTAVAIL;
349 					goto out_unlock;
350 				}
351 			}
352 			rcu_read_unlock();
353 		}
354 	}
355 
356 	inet->inet_rcv_saddr = v4addr;
357 	inet->inet_saddr = v4addr;
358 
359 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
360 
361 	if (!(addr_type & IPV6_ADDR_MULTICAST))
362 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
363 
364 	/* Make sure we are allowed to bind here. */
365 	if (sk->sk_prot->get_port(sk, snum)) {
366 		inet_reset_saddr(sk);
367 		err = -EADDRINUSE;
368 		goto out;
369 	}
370 
371 	if (addr_type != IPV6_ADDR_ANY) {
372 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
373 		if (addr_type != IPV6_ADDR_MAPPED)
374 			np->ipv6only = 1;
375 	}
376 	if (snum)
377 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
378 	inet->inet_sport = htons(inet->inet_num);
379 	inet->inet_dport = 0;
380 	inet->inet_daddr = 0;
381 out:
382 	release_sock(sk);
383 	return err;
384 out_unlock:
385 	rcu_read_unlock();
386 	goto out;
387 }
388 
389 EXPORT_SYMBOL(inet6_bind);
390 
391 int inet6_release(struct socket *sock)
392 {
393 	struct sock *sk = sock->sk;
394 
395 	if (sk == NULL)
396 		return -EINVAL;
397 
398 	/* Free mc lists */
399 	ipv6_sock_mc_close(sk);
400 
401 	/* Free ac lists */
402 	ipv6_sock_ac_close(sk);
403 
404 	return inet_release(sock);
405 }
406 
407 EXPORT_SYMBOL(inet6_release);
408 
409 void inet6_destroy_sock(struct sock *sk)
410 {
411 	struct ipv6_pinfo *np = inet6_sk(sk);
412 	struct sk_buff *skb;
413 	struct ipv6_txoptions *opt;
414 
415 	/* Release rx options */
416 
417 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
418 		kfree_skb(skb);
419 
420 	/* Free flowlabels */
421 	fl6_free_socklist(sk);
422 
423 	/* Free tx options */
424 
425 	if ((opt = xchg(&np->opt, NULL)) != NULL)
426 		sock_kfree_s(sk, opt, opt->tot_len);
427 }
428 
429 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
430 
431 /*
432  *	This does both peername and sockname.
433  */
434 
435 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
436 		 int *uaddr_len, int peer)
437 {
438 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
439 	struct sock *sk = sock->sk;
440 	struct inet_sock *inet = inet_sk(sk);
441 	struct ipv6_pinfo *np = inet6_sk(sk);
442 
443 	sin->sin6_family = AF_INET6;
444 	sin->sin6_flowinfo = 0;
445 	sin->sin6_scope_id = 0;
446 	if (peer) {
447 		if (!inet->inet_dport)
448 			return -ENOTCONN;
449 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
450 		    peer == 1)
451 			return -ENOTCONN;
452 		sin->sin6_port = inet->inet_dport;
453 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
454 		if (np->sndflow)
455 			sin->sin6_flowinfo = np->flow_label;
456 	} else {
457 		if (ipv6_addr_any(&np->rcv_saddr))
458 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
459 		else
460 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
461 
462 		sin->sin6_port = inet->inet_sport;
463 	}
464 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
465 		sin->sin6_scope_id = sk->sk_bound_dev_if;
466 	*uaddr_len = sizeof(*sin);
467 	return(0);
468 }
469 
470 EXPORT_SYMBOL(inet6_getname);
471 
472 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
473 {
474 	struct sock *sk = sock->sk;
475 	struct net *net = sock_net(sk);
476 
477 	switch(cmd)
478 	{
479 	case SIOCGSTAMP:
480 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
481 
482 	case SIOCGSTAMPNS:
483 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
484 
485 	case SIOCADDRT:
486 	case SIOCDELRT:
487 
488 		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
489 
490 	case SIOCSIFADDR:
491 		return addrconf_add_ifaddr(net, (void __user *) arg);
492 	case SIOCDIFADDR:
493 		return addrconf_del_ifaddr(net, (void __user *) arg);
494 	case SIOCSIFDSTADDR:
495 		return addrconf_set_dstaddr(net, (void __user *) arg);
496 	default:
497 		if (!sk->sk_prot->ioctl)
498 			return -ENOIOCTLCMD;
499 		return sk->sk_prot->ioctl(sk, cmd, arg);
500 	}
501 	/*NOTREACHED*/
502 	return(0);
503 }
504 
505 EXPORT_SYMBOL(inet6_ioctl);
506 
507 const struct proto_ops inet6_stream_ops = {
508 	.family		   = PF_INET6,
509 	.owner		   = THIS_MODULE,
510 	.release	   = inet6_release,
511 	.bind		   = inet6_bind,
512 	.connect	   = inet_stream_connect,	/* ok		*/
513 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
514 	.accept		   = inet_accept,		/* ok		*/
515 	.getname	   = inet6_getname,
516 	.poll		   = tcp_poll,			/* ok		*/
517 	.ioctl		   = inet6_ioctl,		/* must change  */
518 	.listen		   = inet_listen,		/* ok		*/
519 	.shutdown	   = inet_shutdown,		/* ok		*/
520 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
521 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
522 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
523 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
524 	.mmap		   = sock_no_mmap,
525 	.sendpage	   = tcp_sendpage,
526 	.splice_read	   = tcp_splice_read,
527 #ifdef CONFIG_COMPAT
528 	.compat_setsockopt = compat_sock_common_setsockopt,
529 	.compat_getsockopt = compat_sock_common_getsockopt,
530 #endif
531 };
532 
533 const struct proto_ops inet6_dgram_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		   = udp_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 const struct net_proto_family inet6_family_ops = {
559 	.family = PF_INET6,
560 	.create = inet6_create,
561 	.owner	= THIS_MODULE,
562 };
563 
564 int inet6_register_protosw(struct inet_protosw *p)
565 {
566 	struct list_head *lh;
567 	struct inet_protosw *answer;
568 	struct list_head *last_perm;
569 	int protocol = p->protocol;
570 	int ret;
571 
572 	spin_lock_bh(&inetsw6_lock);
573 
574 	ret = -EINVAL;
575 	if (p->type >= SOCK_MAX)
576 		goto out_illegal;
577 
578 	/* If we are trying to override a permanent protocol, bail. */
579 	answer = NULL;
580 	ret = -EPERM;
581 	last_perm = &inetsw6[p->type];
582 	list_for_each(lh, &inetsw6[p->type]) {
583 		answer = list_entry(lh, struct inet_protosw, list);
584 
585 		/* Check only the non-wild match. */
586 		if (INET_PROTOSW_PERMANENT & answer->flags) {
587 			if (protocol == answer->protocol)
588 				break;
589 			last_perm = lh;
590 		}
591 
592 		answer = NULL;
593 	}
594 	if (answer)
595 		goto out_permanent;
596 
597 	/* Add the new entry after the last permanent entry if any, so that
598 	 * the new entry does not override a permanent entry when matched with
599 	 * a wild-card protocol. But it is allowed to override any existing
600 	 * non-permanent entry.  This means that when we remove this entry, the
601 	 * system automatically returns to the old behavior.
602 	 */
603 	list_add_rcu(&p->list, last_perm);
604 	ret = 0;
605 out:
606 	spin_unlock_bh(&inetsw6_lock);
607 	return ret;
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.mark = sk->sk_mark;
661 		fl.fl_ip_dport = inet->inet_dport;
662 		fl.fl_ip_sport = inet->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(sock_net(sk), &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 ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
716 {
717 	const struct inet6_protocol *ops = NULL;
718 
719 	for (;;) {
720 		struct ipv6_opt_hdr *opth;
721 		int len;
722 
723 		if (proto != NEXTHDR_HOP) {
724 			ops = rcu_dereference(inet6_protos[proto]);
725 
726 			if (unlikely(!ops))
727 				break;
728 
729 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
730 				break;
731 		}
732 
733 		if (unlikely(!pskb_may_pull(skb, 8)))
734 			break;
735 
736 		opth = (void *)skb->data;
737 		len = ipv6_optlen(opth);
738 
739 		if (unlikely(!pskb_may_pull(skb, len)))
740 			break;
741 
742 		proto = opth->nexthdr;
743 		__skb_pull(skb, len);
744 	}
745 
746 	return proto;
747 }
748 
749 static int ipv6_gso_send_check(struct sk_buff *skb)
750 {
751 	struct ipv6hdr *ipv6h;
752 	const struct inet6_protocol *ops;
753 	int err = -EINVAL;
754 
755 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
756 		goto out;
757 
758 	ipv6h = ipv6_hdr(skb);
759 	__skb_pull(skb, sizeof(*ipv6h));
760 	err = -EPROTONOSUPPORT;
761 
762 	rcu_read_lock();
763 	ops = rcu_dereference(inet6_protos[
764 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
765 
766 	if (likely(ops && ops->gso_send_check)) {
767 		skb_reset_transport_header(skb);
768 		err = ops->gso_send_check(skb);
769 	}
770 	rcu_read_unlock();
771 
772 out:
773 	return err;
774 }
775 
776 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
777 {
778 	struct sk_buff *segs = ERR_PTR(-EINVAL);
779 	struct ipv6hdr *ipv6h;
780 	const struct inet6_protocol *ops;
781 	int proto;
782 	struct frag_hdr *fptr;
783 	unsigned int unfrag_ip6hlen;
784 	u8 *prevhdr;
785 	int offset = 0;
786 
787 	if (!(features & NETIF_F_V6_CSUM))
788 		features &= ~NETIF_F_SG;
789 
790 	if (unlikely(skb_shinfo(skb)->gso_type &
791 		     ~(SKB_GSO_UDP |
792 		       SKB_GSO_DODGY |
793 		       SKB_GSO_TCP_ECN |
794 		       SKB_GSO_TCPV6 |
795 		       0)))
796 		goto out;
797 
798 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
799 		goto out;
800 
801 	ipv6h = ipv6_hdr(skb);
802 	__skb_pull(skb, sizeof(*ipv6h));
803 	segs = ERR_PTR(-EPROTONOSUPPORT);
804 
805 	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
806 	rcu_read_lock();
807 	ops = rcu_dereference(inet6_protos[proto]);
808 	if (likely(ops && ops->gso_segment)) {
809 		skb_reset_transport_header(skb);
810 		segs = ops->gso_segment(skb, features);
811 	}
812 	rcu_read_unlock();
813 
814 	if (unlikely(IS_ERR(segs)))
815 		goto out;
816 
817 	for (skb = segs; skb; skb = skb->next) {
818 		ipv6h = ipv6_hdr(skb);
819 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
820 					   sizeof(*ipv6h));
821 		if (proto == IPPROTO_UDP) {
822 			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
823 			fptr = (struct frag_hdr *)(skb_network_header(skb) +
824 				unfrag_ip6hlen);
825 			fptr->frag_off = htons(offset);
826 			if (skb->next != NULL)
827 				fptr->frag_off |= htons(IP6_MF);
828 			offset += (ntohs(ipv6h->payload_len) -
829 				   sizeof(struct frag_hdr));
830 		}
831 	}
832 
833 out:
834 	return segs;
835 }
836 
837 struct ipv6_gro_cb {
838 	struct napi_gro_cb napi;
839 	int proto;
840 };
841 
842 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
843 
844 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
845 					 struct sk_buff *skb)
846 {
847 	const struct inet6_protocol *ops;
848 	struct sk_buff **pp = NULL;
849 	struct sk_buff *p;
850 	struct ipv6hdr *iph;
851 	unsigned int nlen;
852 	unsigned int hlen;
853 	unsigned int off;
854 	int flush = 1;
855 	int proto;
856 	__wsum csum;
857 
858 	off = skb_gro_offset(skb);
859 	hlen = off + sizeof(*iph);
860 	iph = skb_gro_header_fast(skb, off);
861 	if (skb_gro_header_hard(skb, hlen)) {
862 		iph = skb_gro_header_slow(skb, hlen, off);
863 		if (unlikely(!iph))
864 			goto out;
865 	}
866 
867 	skb_gro_pull(skb, sizeof(*iph));
868 	skb_set_transport_header(skb, skb_gro_offset(skb));
869 
870 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
871 
872 	rcu_read_lock();
873 	proto = iph->nexthdr;
874 	ops = rcu_dereference(inet6_protos[proto]);
875 	if (!ops || !ops->gro_receive) {
876 		__pskb_pull(skb, skb_gro_offset(skb));
877 		proto = ipv6_gso_pull_exthdrs(skb, proto);
878 		skb_gro_pull(skb, -skb_transport_offset(skb));
879 		skb_reset_transport_header(skb);
880 		__skb_push(skb, skb_gro_offset(skb));
881 
882 		if (!ops || !ops->gro_receive)
883 			goto out_unlock;
884 
885 		iph = ipv6_hdr(skb);
886 	}
887 
888 	IPV6_GRO_CB(skb)->proto = proto;
889 
890 	flush--;
891 	nlen = skb_network_header_len(skb);
892 
893 	for (p = *head; p; p = p->next) {
894 		struct ipv6hdr *iph2;
895 
896 		if (!NAPI_GRO_CB(p)->same_flow)
897 			continue;
898 
899 		iph2 = ipv6_hdr(p);
900 
901 		/* All fields must match except length. */
902 		if (nlen != skb_network_header_len(p) ||
903 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
904 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
905 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
906 			NAPI_GRO_CB(p)->same_flow = 0;
907 			continue;
908 		}
909 
910 		NAPI_GRO_CB(p)->flush |= flush;
911 	}
912 
913 	NAPI_GRO_CB(skb)->flush |= flush;
914 
915 	csum = skb->csum;
916 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
917 
918 	pp = ops->gro_receive(head, skb);
919 
920 	skb->csum = csum;
921 
922 out_unlock:
923 	rcu_read_unlock();
924 
925 out:
926 	NAPI_GRO_CB(skb)->flush |= flush;
927 
928 	return pp;
929 }
930 
931 static int ipv6_gro_complete(struct sk_buff *skb)
932 {
933 	const struct inet6_protocol *ops;
934 	struct ipv6hdr *iph = ipv6_hdr(skb);
935 	int err = -ENOSYS;
936 
937 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
938 				 sizeof(*iph));
939 
940 	rcu_read_lock();
941 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
942 	if (WARN_ON(!ops || !ops->gro_complete))
943 		goto out_unlock;
944 
945 	err = ops->gro_complete(skb);
946 
947 out_unlock:
948 	rcu_read_unlock();
949 
950 	return err;
951 }
952 
953 static struct packet_type ipv6_packet_type __read_mostly = {
954 	.type = cpu_to_be16(ETH_P_IPV6),
955 	.func = ipv6_rcv,
956 	.gso_send_check = ipv6_gso_send_check,
957 	.gso_segment = ipv6_gso_segment,
958 	.gro_receive = ipv6_gro_receive,
959 	.gro_complete = ipv6_gro_complete,
960 };
961 
962 static int __init ipv6_packet_init(void)
963 {
964 	dev_add_pack(&ipv6_packet_type);
965 	return 0;
966 }
967 
968 static void ipv6_packet_cleanup(void)
969 {
970 	dev_remove_pack(&ipv6_packet_type);
971 }
972 
973 static int __net_init ipv6_init_mibs(struct net *net)
974 {
975 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
976 			  sizeof (struct udp_mib)) < 0)
977 		return -ENOMEM;
978 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
979 			  sizeof (struct udp_mib)) < 0)
980 		goto err_udplite_mib;
981 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
982 			  sizeof(struct ipstats_mib)) < 0)
983 		goto err_ip_mib;
984 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
985 			  sizeof(struct icmpv6_mib)) < 0)
986 		goto err_icmp_mib;
987 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
988 			  sizeof(struct icmpv6msg_mib)) < 0)
989 		goto err_icmpmsg_mib;
990 	return 0;
991 
992 err_icmpmsg_mib:
993 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
994 err_icmp_mib:
995 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
996 err_ip_mib:
997 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
998 err_udplite_mib:
999 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1000 	return -ENOMEM;
1001 }
1002 
1003 static void ipv6_cleanup_mibs(struct net *net)
1004 {
1005 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1006 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1007 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1008 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1009 	snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1010 }
1011 
1012 static int __net_init inet6_net_init(struct net *net)
1013 {
1014 	int err = 0;
1015 
1016 	net->ipv6.sysctl.bindv6only = 0;
1017 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
1018 
1019 	err = ipv6_init_mibs(net);
1020 	if (err)
1021 		return err;
1022 #ifdef CONFIG_PROC_FS
1023 	err = udp6_proc_init(net);
1024 	if (err)
1025 		goto out;
1026 	err = tcp6_proc_init(net);
1027 	if (err)
1028 		goto proc_tcp6_fail;
1029 	err = ac6_proc_init(net);
1030 	if (err)
1031 		goto proc_ac6_fail;
1032 #endif
1033 	return err;
1034 
1035 #ifdef CONFIG_PROC_FS
1036 proc_ac6_fail:
1037 	tcp6_proc_exit(net);
1038 proc_tcp6_fail:
1039 	udp6_proc_exit(net);
1040 out:
1041 	ipv6_cleanup_mibs(net);
1042 	return err;
1043 #endif
1044 }
1045 
1046 static void __net_exit inet6_net_exit(struct net *net)
1047 {
1048 #ifdef CONFIG_PROC_FS
1049 	udp6_proc_exit(net);
1050 	tcp6_proc_exit(net);
1051 	ac6_proc_exit(net);
1052 #endif
1053 	ipv6_cleanup_mibs(net);
1054 }
1055 
1056 static struct pernet_operations inet6_net_ops = {
1057 	.init = inet6_net_init,
1058 	.exit = inet6_net_exit,
1059 };
1060 
1061 static int __init inet6_init(void)
1062 {
1063 	struct sk_buff *dummy_skb;
1064 	struct list_head *r;
1065 	int err = 0;
1066 
1067 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1068 
1069 	/* Register the socket-side information for inet6_create.  */
1070 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1071 		INIT_LIST_HEAD(r);
1072 
1073 	if (disable_ipv6_mod) {
1074 		printk(KERN_INFO
1075 		       "IPv6: Loaded, but administratively disabled, "
1076 		       "reboot required to enable\n");
1077 		goto out;
1078 	}
1079 
1080 	err = proto_register(&tcpv6_prot, 1);
1081 	if (err)
1082 		goto out;
1083 
1084 	err = proto_register(&udpv6_prot, 1);
1085 	if (err)
1086 		goto out_unregister_tcp_proto;
1087 
1088 	err = proto_register(&udplitev6_prot, 1);
1089 	if (err)
1090 		goto out_unregister_udp_proto;
1091 
1092 	err = proto_register(&rawv6_prot, 1);
1093 	if (err)
1094 		goto out_unregister_udplite_proto;
1095 
1096 
1097 	/* We MUST register RAW sockets before we create the ICMP6,
1098 	 * IGMP6, or NDISC control sockets.
1099 	 */
1100 	err = rawv6_init();
1101 	if (err)
1102 		goto out_unregister_raw_proto;
1103 
1104 	/* Register the family here so that the init calls below will
1105 	 * be able to create sockets. (?? is this dangerous ??)
1106 	 */
1107 	err = sock_register(&inet6_family_ops);
1108 	if (err)
1109 		goto out_sock_register_fail;
1110 
1111 #ifdef CONFIG_SYSCTL
1112 	err = ipv6_static_sysctl_register();
1113 	if (err)
1114 		goto static_sysctl_fail;
1115 #endif
1116 	/*
1117 	 *	ipngwg API draft makes clear that the correct semantics
1118 	 *	for TCP and UDP is to consider one TCP and UDP instance
1119 	 *	in a host availiable by both INET and INET6 APIs and
1120 	 *	able to communicate via both network protocols.
1121 	 */
1122 
1123 	err = register_pernet_subsys(&inet6_net_ops);
1124 	if (err)
1125 		goto register_pernet_fail;
1126 	err = icmpv6_init();
1127 	if (err)
1128 		goto icmp_fail;
1129 	err = ip6_mr_init();
1130 	if (err)
1131 		goto ipmr_fail;
1132 	err = ndisc_init();
1133 	if (err)
1134 		goto ndisc_fail;
1135 	err = igmp6_init();
1136 	if (err)
1137 		goto igmp_fail;
1138 	err = ipv6_netfilter_init();
1139 	if (err)
1140 		goto netfilter_fail;
1141 	/* Create /proc/foo6 entries. */
1142 #ifdef CONFIG_PROC_FS
1143 	err = -ENOMEM;
1144 	if (raw6_proc_init())
1145 		goto proc_raw6_fail;
1146 	if (udplite6_proc_init())
1147 		goto proc_udplite6_fail;
1148 	if (ipv6_misc_proc_init())
1149 		goto proc_misc6_fail;
1150 	if (if6_proc_init())
1151 		goto proc_if6_fail;
1152 #endif
1153 	err = ip6_route_init();
1154 	if (err)
1155 		goto ip6_route_fail;
1156 	err = ip6_flowlabel_init();
1157 	if (err)
1158 		goto ip6_flowlabel_fail;
1159 	err = addrconf_init();
1160 	if (err)
1161 		goto addrconf_fail;
1162 
1163 	/* Init v6 extension headers. */
1164 	err = ipv6_exthdrs_init();
1165 	if (err)
1166 		goto ipv6_exthdrs_fail;
1167 
1168 	err = ipv6_frag_init();
1169 	if (err)
1170 		goto ipv6_frag_fail;
1171 
1172 	/* Init v6 transport protocols. */
1173 	err = udpv6_init();
1174 	if (err)
1175 		goto udpv6_fail;
1176 
1177 	err = udplitev6_init();
1178 	if (err)
1179 		goto udplitev6_fail;
1180 
1181 	err = tcpv6_init();
1182 	if (err)
1183 		goto tcpv6_fail;
1184 
1185 	err = ipv6_packet_init();
1186 	if (err)
1187 		goto ipv6_packet_fail;
1188 
1189 #ifdef CONFIG_SYSCTL
1190 	err = ipv6_sysctl_register();
1191 	if (err)
1192 		goto sysctl_fail;
1193 #endif
1194 out:
1195 	return err;
1196 
1197 #ifdef CONFIG_SYSCTL
1198 sysctl_fail:
1199 	ipv6_packet_cleanup();
1200 #endif
1201 ipv6_packet_fail:
1202 	tcpv6_exit();
1203 tcpv6_fail:
1204 	udplitev6_exit();
1205 udplitev6_fail:
1206 	udpv6_exit();
1207 udpv6_fail:
1208 	ipv6_frag_exit();
1209 ipv6_frag_fail:
1210 	ipv6_exthdrs_exit();
1211 ipv6_exthdrs_fail:
1212 	addrconf_cleanup();
1213 addrconf_fail:
1214 	ip6_flowlabel_cleanup();
1215 ip6_flowlabel_fail:
1216 	ip6_route_cleanup();
1217 ip6_route_fail:
1218 #ifdef CONFIG_PROC_FS
1219 	if6_proc_exit();
1220 proc_if6_fail:
1221 	ipv6_misc_proc_exit();
1222 proc_misc6_fail:
1223 	udplite6_proc_exit();
1224 proc_udplite6_fail:
1225 	raw6_proc_exit();
1226 proc_raw6_fail:
1227 #endif
1228 	ipv6_netfilter_fini();
1229 netfilter_fail:
1230 	igmp6_cleanup();
1231 igmp_fail:
1232 	ndisc_cleanup();
1233 ndisc_fail:
1234 	ip6_mr_cleanup();
1235 ipmr_fail:
1236 	icmpv6_cleanup();
1237 icmp_fail:
1238 	unregister_pernet_subsys(&inet6_net_ops);
1239 register_pernet_fail:
1240 #ifdef CONFIG_SYSCTL
1241 	ipv6_static_sysctl_unregister();
1242 static_sysctl_fail:
1243 #endif
1244 	sock_unregister(PF_INET6);
1245 	rtnl_unregister_all(PF_INET6);
1246 out_sock_register_fail:
1247 	rawv6_exit();
1248 out_unregister_raw_proto:
1249 	proto_unregister(&rawv6_prot);
1250 out_unregister_udplite_proto:
1251 	proto_unregister(&udplitev6_prot);
1252 out_unregister_udp_proto:
1253 	proto_unregister(&udpv6_prot);
1254 out_unregister_tcp_proto:
1255 	proto_unregister(&tcpv6_prot);
1256 	goto out;
1257 }
1258 module_init(inet6_init);
1259 
1260 static void __exit inet6_exit(void)
1261 {
1262 	if (disable_ipv6_mod)
1263 		return;
1264 
1265 	/* First of all disallow new sockets creation. */
1266 	sock_unregister(PF_INET6);
1267 	/* Disallow any further netlink messages */
1268 	rtnl_unregister_all(PF_INET6);
1269 
1270 #ifdef CONFIG_SYSCTL
1271 	ipv6_sysctl_unregister();
1272 #endif
1273 	udpv6_exit();
1274 	udplitev6_exit();
1275 	tcpv6_exit();
1276 
1277 	/* Cleanup code parts. */
1278 	ipv6_packet_cleanup();
1279 	ipv6_frag_exit();
1280 	ipv6_exthdrs_exit();
1281 	addrconf_cleanup();
1282 	ip6_flowlabel_cleanup();
1283 	ip6_route_cleanup();
1284 #ifdef CONFIG_PROC_FS
1285 
1286 	/* Cleanup code parts. */
1287 	if6_proc_exit();
1288 	ipv6_misc_proc_exit();
1289 	udplite6_proc_exit();
1290 	raw6_proc_exit();
1291 #endif
1292 	ipv6_netfilter_fini();
1293 	igmp6_cleanup();
1294 	ndisc_cleanup();
1295 	ip6_mr_cleanup();
1296 	icmpv6_cleanup();
1297 	rawv6_exit();
1298 
1299 	unregister_pernet_subsys(&inet6_net_ops);
1300 #ifdef CONFIG_SYSCTL
1301 	ipv6_static_sysctl_unregister();
1302 #endif
1303 	proto_unregister(&rawv6_prot);
1304 	proto_unregister(&udplitev6_prot);
1305 	proto_unregister(&udpv6_prot);
1306 	proto_unregister(&tcpv6_prot);
1307 
1308 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1309 }
1310 module_exit(inet6_exit);
1311 
1312 MODULE_ALIAS_NETPROTO(PF_INET6);
1313