xref: /linux/net/ipv6/af_inet6.c (revision 5bdef865eb358b6f3760e25e591ae115e9eeddef)
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 
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/icmpv6.h>
43 #include <linux/netfilter_ipv6.h>
44 
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/tcp.h>
50 #include <net/ipip.h>
51 #include <net/protocol.h>
52 #include <net/inet_common.h>
53 #include <net/route.h>
54 #include <net/transp_v6.h>
55 #include <net/ip6_route.h>
56 #include <net/addrconf.h>
57 #ifdef CONFIG_IPV6_TUNNEL
58 #include <net/ip6_tunnel.h>
59 #endif
60 
61 #include <asm/uaccess.h>
62 #include <asm/system.h>
63 #include <linux/mroute6.h>
64 
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
68 
69 /* The inetsw6 table contains everything that inet6_create needs to
70  * build a new socket.
71  */
72 static struct list_head inetsw6[SOCK_MAX];
73 static DEFINE_SPINLOCK(inetsw6_lock);
74 
75 struct ipv6_params ipv6_defaults = {
76 	.disable_ipv6 = 0,
77 	.autoconf = 1,
78 };
79 
80 static int disable_ipv6_mod = 0;
81 
82 module_param_named(disable, disable_ipv6_mod, int, 0444);
83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
84 
85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
87 
88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
90 
91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
92 {
93 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
94 
95 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
96 }
97 
98 static int inet6_create(struct net *net, struct socket *sock, int protocol)
99 {
100 	struct inet_sock *inet;
101 	struct ipv6_pinfo *np;
102 	struct sock *sk;
103 	struct inet_protosw *answer;
104 	struct proto *answer_prot;
105 	unsigned char answer_flags;
106 	char answer_no_check;
107 	int try_loading_module = 0;
108 	int err;
109 
110 	if (sock->type != SOCK_RAW &&
111 	    sock->type != SOCK_DGRAM &&
112 	    !inet_ehash_secret)
113 		build_ehash_secret();
114 
115 	/* Look for the requested type/protocol pair. */
116 lookup_protocol:
117 	err = -ESOCKTNOSUPPORT;
118 	rcu_read_lock();
119 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
120 
121 		err = 0;
122 		/* Check the non-wild match. */
123 		if (protocol == answer->protocol) {
124 			if (protocol != IPPROTO_IP)
125 				break;
126 		} else {
127 			/* Check for the two wild cases. */
128 			if (IPPROTO_IP == protocol) {
129 				protocol = answer->protocol;
130 				break;
131 			}
132 			if (IPPROTO_IP == answer->protocol)
133 				break;
134 		}
135 		err = -EPROTONOSUPPORT;
136 	}
137 
138 	if (err) {
139 		if (try_loading_module < 2) {
140 			rcu_read_unlock();
141 			/*
142 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
143 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
144 			 */
145 			if (++try_loading_module == 1)
146 				request_module("net-pf-%d-proto-%d-type-%d",
147 						PF_INET6, protocol, sock->type);
148 			/*
149 			 * Fall back to generic, e.g. net-pf-10-proto-132
150 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
151 			 */
152 			else
153 				request_module("net-pf-%d-proto-%d",
154 						PF_INET6, protocol);
155 			goto lookup_protocol;
156 		} else
157 			goto out_rcu_unlock;
158 	}
159 
160 	err = -EPERM;
161 	if (answer->capability > 0 && !capable(answer->capability))
162 		goto out_rcu_unlock;
163 
164 	sock->ops = answer->ops;
165 	answer_prot = answer->prot;
166 	answer_no_check = answer->no_check;
167 	answer_flags = answer->flags;
168 	rcu_read_unlock();
169 
170 	WARN_ON(answer_prot->slab == NULL);
171 
172 	err = -ENOBUFS;
173 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
174 	if (sk == NULL)
175 		goto out;
176 
177 	sock_init_data(sock, sk);
178 
179 	err = 0;
180 	sk->sk_no_check = answer_no_check;
181 	if (INET_PROTOSW_REUSE & answer_flags)
182 		sk->sk_reuse = 1;
183 
184 	inet = inet_sk(sk);
185 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
186 
187 	if (SOCK_RAW == sock->type) {
188 		inet->num = protocol;
189 		if (IPPROTO_RAW == protocol)
190 			inet->hdrincl = 1;
191 	}
192 
193 	sk->sk_destruct		= inet_sock_destruct;
194 	sk->sk_family		= PF_INET6;
195 	sk->sk_protocol		= protocol;
196 
197 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
198 
199 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
200 	np->hop_limit	= -1;
201 	np->mcast_hops	= -1;
202 	np->mc_loop	= 1;
203 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
204 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
205 
206 	/* Init the ipv4 part of the socket since we can have sockets
207 	 * using v6 API for ipv4.
208 	 */
209 	inet->uc_ttl	= -1;
210 
211 	inet->mc_loop	= 1;
212 	inet->mc_ttl	= 1;
213 	inet->mc_index	= 0;
214 	inet->mc_list	= NULL;
215 
216 	if (ipv4_config.no_pmtu_disc)
217 		inet->pmtudisc = IP_PMTUDISC_DONT;
218 	else
219 		inet->pmtudisc = IP_PMTUDISC_WANT;
220 	/*
221 	 * Increment only the relevant sk_prot->socks debug field, this changes
222 	 * the previous behaviour of incrementing both the equivalent to
223 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
224 	 *
225 	 * This allows better debug granularity as we'll know exactly how many
226 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
227 	 * transport protocol socks. -acme
228 	 */
229 	sk_refcnt_debug_inc(sk);
230 
231 	if (inet->num) {
232 		/* It assumes that any protocol which allows
233 		 * the user to assign a number at socket
234 		 * creation time automatically shares.
235 		 */
236 		inet->sport = htons(inet->num);
237 		sk->sk_prot->hash(sk);
238 	}
239 	if (sk->sk_prot->init) {
240 		err = sk->sk_prot->init(sk);
241 		if (err) {
242 			sk_common_release(sk);
243 			goto out;
244 		}
245 	}
246 out:
247 	return err;
248 out_rcu_unlock:
249 	rcu_read_unlock();
250 	goto out;
251 }
252 
253 
254 /* bind for INET6 API */
255 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
256 {
257 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
258 	struct sock *sk = sock->sk;
259 	struct inet_sock *inet = inet_sk(sk);
260 	struct ipv6_pinfo *np = inet6_sk(sk);
261 	struct net *net = sock_net(sk);
262 	__be32 v4addr = 0;
263 	unsigned short snum;
264 	int addr_type = 0;
265 	int err = 0;
266 
267 	/* If the socket has its own bind function then use it. */
268 	if (sk->sk_prot->bind)
269 		return sk->sk_prot->bind(sk, uaddr, addr_len);
270 
271 	if (addr_len < SIN6_LEN_RFC2133)
272 		return -EINVAL;
273 	addr_type = ipv6_addr_type(&addr->sin6_addr);
274 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
275 		return -EINVAL;
276 
277 	snum = ntohs(addr->sin6_port);
278 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
279 		return -EACCES;
280 
281 	lock_sock(sk);
282 
283 	/* Check these errors (active socket, double bind). */
284 	if (sk->sk_state != TCP_CLOSE || inet->num) {
285 		err = -EINVAL;
286 		goto out;
287 	}
288 
289 	/* Check if the address belongs to the host. */
290 	if (addr_type == IPV6_ADDR_MAPPED) {
291 		int chk_addr_ret;
292 
293 		/* Binding to v4-mapped address on a v6-only socket
294 		 * makes no sense
295 		 */
296 		if (np->ipv6only) {
297 			err = -EINVAL;
298 			goto out;
299 		}
300 
301 		/* Reproduce AF_INET checks to make the bindings consitant */
302 		v4addr = addr->sin6_addr.s6_addr32[3];
303 		chk_addr_ret = inet_addr_type(net, v4addr);
304 		if (!sysctl_ip_nonlocal_bind &&
305 		    !(inet->freebind || inet->transparent) &&
306 		    v4addr != htonl(INADDR_ANY) &&
307 		    chk_addr_ret != RTN_LOCAL &&
308 		    chk_addr_ret != RTN_MULTICAST &&
309 		    chk_addr_ret != RTN_BROADCAST)
310 			goto out;
311 	} else {
312 		if (addr_type != IPV6_ADDR_ANY) {
313 			struct net_device *dev = NULL;
314 
315 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
316 				if (addr_len >= sizeof(struct sockaddr_in6) &&
317 				    addr->sin6_scope_id) {
318 					/* Override any existing binding, if another one
319 					 * is supplied by user.
320 					 */
321 					sk->sk_bound_dev_if = addr->sin6_scope_id;
322 				}
323 
324 				/* Binding to link-local address requires an interface */
325 				if (!sk->sk_bound_dev_if) {
326 					err = -EINVAL;
327 					goto out;
328 				}
329 				dev = dev_get_by_index(net, sk->sk_bound_dev_if);
330 				if (!dev) {
331 					err = -ENODEV;
332 					goto out;
333 				}
334 			}
335 
336 			/* ipv4 addr of the socket is invalid.  Only the
337 			 * unspecified and mapped address have a v4 equivalent.
338 			 */
339 			v4addr = LOOPBACK4_IPV6;
340 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
341 				if (!ipv6_chk_addr(net, &addr->sin6_addr,
342 						   dev, 0)) {
343 					if (dev)
344 						dev_put(dev);
345 					err = -EADDRNOTAVAIL;
346 					goto out;
347 				}
348 			}
349 			if (dev)
350 				dev_put(dev);
351 		}
352 	}
353 
354 	inet->rcv_saddr = v4addr;
355 	inet->saddr = v4addr;
356 
357 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
358 
359 	if (!(addr_type & IPV6_ADDR_MULTICAST))
360 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
361 
362 	/* Make sure we are allowed to bind here. */
363 	if (sk->sk_prot->get_port(sk, snum)) {
364 		inet_reset_saddr(sk);
365 		err = -EADDRINUSE;
366 		goto out;
367 	}
368 
369 	if (addr_type != IPV6_ADDR_ANY) {
370 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
371 		if (addr_type != IPV6_ADDR_MAPPED)
372 			np->ipv6only = 1;
373 	}
374 	if (snum)
375 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
376 	inet->sport = htons(inet->num);
377 	inet->dport = 0;
378 	inet->daddr = 0;
379 out:
380 	release_sock(sk);
381 	return err;
382 }
383 
384 EXPORT_SYMBOL(inet6_bind);
385 
386 int inet6_release(struct socket *sock)
387 {
388 	struct sock *sk = sock->sk;
389 
390 	if (sk == NULL)
391 		return -EINVAL;
392 
393 	/* Free mc lists */
394 	ipv6_sock_mc_close(sk);
395 
396 	/* Free ac lists */
397 	ipv6_sock_ac_close(sk);
398 
399 	return inet_release(sock);
400 }
401 
402 EXPORT_SYMBOL(inet6_release);
403 
404 void inet6_destroy_sock(struct sock *sk)
405 {
406 	struct ipv6_pinfo *np = inet6_sk(sk);
407 	struct sk_buff *skb;
408 	struct ipv6_txoptions *opt;
409 
410 	/* Release rx options */
411 
412 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
413 		kfree_skb(skb);
414 
415 	/* Free flowlabels */
416 	fl6_free_socklist(sk);
417 
418 	/* Free tx options */
419 
420 	if ((opt = xchg(&np->opt, NULL)) != NULL)
421 		sock_kfree_s(sk, opt, opt->tot_len);
422 }
423 
424 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
425 
426 /*
427  *	This does both peername and sockname.
428  */
429 
430 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
431 		 int *uaddr_len, int peer)
432 {
433 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
434 	struct sock *sk = sock->sk;
435 	struct inet_sock *inet = inet_sk(sk);
436 	struct ipv6_pinfo *np = inet6_sk(sk);
437 
438 	sin->sin6_family = AF_INET6;
439 	sin->sin6_flowinfo = 0;
440 	sin->sin6_scope_id = 0;
441 	if (peer) {
442 		if (!inet->dport)
443 			return -ENOTCONN;
444 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
445 		    peer == 1)
446 			return -ENOTCONN;
447 		sin->sin6_port = inet->dport;
448 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
449 		if (np->sndflow)
450 			sin->sin6_flowinfo = np->flow_label;
451 	} else {
452 		if (ipv6_addr_any(&np->rcv_saddr))
453 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
454 		else
455 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
456 
457 		sin->sin6_port = inet->sport;
458 	}
459 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
460 		sin->sin6_scope_id = sk->sk_bound_dev_if;
461 	*uaddr_len = sizeof(*sin);
462 	return(0);
463 }
464 
465 EXPORT_SYMBOL(inet6_getname);
466 
467 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
468 {
469 	struct sock *sk = sock->sk;
470 	struct net *net = sock_net(sk);
471 
472 	switch(cmd)
473 	{
474 	case SIOCGSTAMP:
475 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
476 
477 	case SIOCGSTAMPNS:
478 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
479 
480 	case SIOCADDRT:
481 	case SIOCDELRT:
482 
483 		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
484 
485 	case SIOCSIFADDR:
486 		return addrconf_add_ifaddr(net, (void __user *) arg);
487 	case SIOCDIFADDR:
488 		return addrconf_del_ifaddr(net, (void __user *) arg);
489 	case SIOCSIFDSTADDR:
490 		return addrconf_set_dstaddr(net, (void __user *) arg);
491 	default:
492 		if (!sk->sk_prot->ioctl)
493 			return -ENOIOCTLCMD;
494 		return sk->sk_prot->ioctl(sk, cmd, arg);
495 	}
496 	/*NOTREACHED*/
497 	return(0);
498 }
499 
500 EXPORT_SYMBOL(inet6_ioctl);
501 
502 const struct proto_ops inet6_stream_ops = {
503 	.family		   = PF_INET6,
504 	.owner		   = THIS_MODULE,
505 	.release	   = inet6_release,
506 	.bind		   = inet6_bind,
507 	.connect	   = inet_stream_connect,	/* ok		*/
508 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
509 	.accept		   = inet_accept,		/* ok		*/
510 	.getname	   = inet6_getname,
511 	.poll		   = tcp_poll,			/* ok		*/
512 	.ioctl		   = inet6_ioctl,		/* must change  */
513 	.listen		   = inet_listen,		/* ok		*/
514 	.shutdown	   = inet_shutdown,		/* ok		*/
515 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
516 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
517 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
518 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
519 	.mmap		   = sock_no_mmap,
520 	.sendpage	   = tcp_sendpage,
521 	.splice_read	   = tcp_splice_read,
522 #ifdef CONFIG_COMPAT
523 	.compat_setsockopt = compat_sock_common_setsockopt,
524 	.compat_getsockopt = compat_sock_common_getsockopt,
525 #endif
526 };
527 
528 const struct proto_ops inet6_dgram_ops = {
529 	.family		   = PF_INET6,
530 	.owner		   = THIS_MODULE,
531 	.release	   = inet6_release,
532 	.bind		   = inet6_bind,
533 	.connect	   = inet_dgram_connect,	/* ok		*/
534 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
535 	.accept		   = sock_no_accept,		/* a do nothing	*/
536 	.getname	   = inet6_getname,
537 	.poll		   = udp_poll,			/* ok		*/
538 	.ioctl		   = inet6_ioctl,		/* must change  */
539 	.listen		   = sock_no_listen,		/* ok		*/
540 	.shutdown	   = inet_shutdown,		/* ok		*/
541 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
542 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
543 	.sendmsg	   = inet_sendmsg,		/* ok		*/
544 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
545 	.mmap		   = sock_no_mmap,
546 	.sendpage	   = sock_no_sendpage,
547 #ifdef CONFIG_COMPAT
548 	.compat_setsockopt = compat_sock_common_setsockopt,
549 	.compat_getsockopt = compat_sock_common_getsockopt,
550 #endif
551 };
552 
553 static struct net_proto_family inet6_family_ops = {
554 	.family = PF_INET6,
555 	.create = inet6_create,
556 	.owner	= THIS_MODULE,
557 };
558 
559 int inet6_register_protosw(struct inet_protosw *p)
560 {
561 	struct list_head *lh;
562 	struct inet_protosw *answer;
563 	struct list_head *last_perm;
564 	int protocol = p->protocol;
565 	int ret;
566 
567 	spin_lock_bh(&inetsw6_lock);
568 
569 	ret = -EINVAL;
570 	if (p->type >= SOCK_MAX)
571 		goto out_illegal;
572 
573 	/* If we are trying to override a permanent protocol, bail. */
574 	answer = NULL;
575 	ret = -EPERM;
576 	last_perm = &inetsw6[p->type];
577 	list_for_each(lh, &inetsw6[p->type]) {
578 		answer = list_entry(lh, struct inet_protosw, list);
579 
580 		/* Check only the non-wild match. */
581 		if (INET_PROTOSW_PERMANENT & answer->flags) {
582 			if (protocol == answer->protocol)
583 				break;
584 			last_perm = lh;
585 		}
586 
587 		answer = NULL;
588 	}
589 	if (answer)
590 		goto out_permanent;
591 
592 	/* Add the new entry after the last permanent entry if any, so that
593 	 * the new entry does not override a permanent entry when matched with
594 	 * a wild-card protocol. But it is allowed to override any existing
595 	 * non-permanent entry.  This means that when we remove this entry, the
596 	 * system automatically returns to the old behavior.
597 	 */
598 	list_add_rcu(&p->list, last_perm);
599 	ret = 0;
600 out:
601 	spin_unlock_bh(&inetsw6_lock);
602 	return ret;
603 
604 out_permanent:
605 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
606 	       protocol);
607 	goto out;
608 
609 out_illegal:
610 	printk(KERN_ERR
611 	       "Ignoring attempt to register invalid socket type %d.\n",
612 	       p->type);
613 	goto out;
614 }
615 
616 EXPORT_SYMBOL(inet6_register_protosw);
617 
618 void
619 inet6_unregister_protosw(struct inet_protosw *p)
620 {
621 	if (INET_PROTOSW_PERMANENT & p->flags) {
622 		printk(KERN_ERR
623 		       "Attempt to unregister permanent protocol %d.\n",
624 		       p->protocol);
625 	} else {
626 		spin_lock_bh(&inetsw6_lock);
627 		list_del_rcu(&p->list);
628 		spin_unlock_bh(&inetsw6_lock);
629 
630 		synchronize_net();
631 	}
632 }
633 
634 EXPORT_SYMBOL(inet6_unregister_protosw);
635 
636 int inet6_sk_rebuild_header(struct sock *sk)
637 {
638 	int err;
639 	struct dst_entry *dst;
640 	struct ipv6_pinfo *np = inet6_sk(sk);
641 
642 	dst = __sk_dst_check(sk, np->dst_cookie);
643 
644 	if (dst == NULL) {
645 		struct inet_sock *inet = inet_sk(sk);
646 		struct in6_addr *final_p = NULL, final;
647 		struct flowi fl;
648 
649 		memset(&fl, 0, sizeof(fl));
650 		fl.proto = sk->sk_protocol;
651 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
652 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
653 		fl.fl6_flowlabel = np->flow_label;
654 		fl.oif = sk->sk_bound_dev_if;
655 		fl.fl_ip_dport = inet->dport;
656 		fl.fl_ip_sport = inet->sport;
657 		security_sk_classify_flow(sk, &fl);
658 
659 		if (np->opt && np->opt->srcrt) {
660 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
661 			ipv6_addr_copy(&final, &fl.fl6_dst);
662 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
663 			final_p = &final;
664 		}
665 
666 		err = ip6_dst_lookup(sk, &dst, &fl);
667 		if (err) {
668 			sk->sk_route_caps = 0;
669 			return err;
670 		}
671 		if (final_p)
672 			ipv6_addr_copy(&fl.fl6_dst, final_p);
673 
674 		if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
675 			sk->sk_err_soft = -err;
676 			return err;
677 		}
678 
679 		__ip6_dst_store(sk, dst, NULL, NULL);
680 	}
681 
682 	return 0;
683 }
684 
685 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
686 
687 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
688 {
689 	struct ipv6_pinfo *np = inet6_sk(sk);
690 	struct inet6_skb_parm *opt = IP6CB(skb);
691 
692 	if (np->rxopt.all) {
693 		if ((opt->hop && (np->rxopt.bits.hopopts ||
694 				  np->rxopt.bits.ohopopts)) ||
695 		    ((IPV6_FLOWINFO_MASK &
696 		      *(__be32 *)skb_network_header(skb)) &&
697 		     np->rxopt.bits.rxflow) ||
698 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
699 		     np->rxopt.bits.osrcrt)) ||
700 		    ((opt->dst1 || opt->dst0) &&
701 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
702 			return 1;
703 	}
704 	return 0;
705 }
706 
707 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
708 
709 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
710 {
711 	struct inet6_protocol *ops = NULL;
712 
713 	for (;;) {
714 		struct ipv6_opt_hdr *opth;
715 		int len;
716 
717 		if (proto != NEXTHDR_HOP) {
718 			ops = rcu_dereference(inet6_protos[proto]);
719 
720 			if (unlikely(!ops))
721 				break;
722 
723 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
724 				break;
725 		}
726 
727 		if (unlikely(!pskb_may_pull(skb, 8)))
728 			break;
729 
730 		opth = (void *)skb->data;
731 		len = ipv6_optlen(opth);
732 
733 		if (unlikely(!pskb_may_pull(skb, len)))
734 			break;
735 
736 		proto = opth->nexthdr;
737 		__skb_pull(skb, len);
738 	}
739 
740 	return proto;
741 }
742 
743 static int ipv6_gso_send_check(struct sk_buff *skb)
744 {
745 	struct ipv6hdr *ipv6h;
746 	struct inet6_protocol *ops;
747 	int err = -EINVAL;
748 
749 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
750 		goto out;
751 
752 	ipv6h = ipv6_hdr(skb);
753 	__skb_pull(skb, sizeof(*ipv6h));
754 	err = -EPROTONOSUPPORT;
755 
756 	rcu_read_lock();
757 	ops = rcu_dereference(inet6_protos[
758 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
759 
760 	if (likely(ops && ops->gso_send_check)) {
761 		skb_reset_transport_header(skb);
762 		err = ops->gso_send_check(skb);
763 	}
764 	rcu_read_unlock();
765 
766 out:
767 	return err;
768 }
769 
770 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
771 {
772 	struct sk_buff *segs = ERR_PTR(-EINVAL);
773 	struct ipv6hdr *ipv6h;
774 	struct inet6_protocol *ops;
775 
776 	if (!(features & NETIF_F_V6_CSUM))
777 		features &= ~NETIF_F_SG;
778 
779 	if (unlikely(skb_shinfo(skb)->gso_type &
780 		     ~(SKB_GSO_UDP |
781 		       SKB_GSO_DODGY |
782 		       SKB_GSO_TCP_ECN |
783 		       SKB_GSO_TCPV6 |
784 		       0)))
785 		goto out;
786 
787 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
788 		goto out;
789 
790 	ipv6h = ipv6_hdr(skb);
791 	__skb_pull(skb, sizeof(*ipv6h));
792 	segs = ERR_PTR(-EPROTONOSUPPORT);
793 
794 	rcu_read_lock();
795 	ops = rcu_dereference(inet6_protos[
796 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
797 
798 	if (likely(ops && ops->gso_segment)) {
799 		skb_reset_transport_header(skb);
800 		segs = ops->gso_segment(skb, features);
801 	}
802 	rcu_read_unlock();
803 
804 	if (unlikely(IS_ERR(segs)))
805 		goto out;
806 
807 	for (skb = segs; skb; skb = skb->next) {
808 		ipv6h = ipv6_hdr(skb);
809 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
810 					   sizeof(*ipv6h));
811 	}
812 
813 out:
814 	return segs;
815 }
816 
817 struct ipv6_gro_cb {
818 	struct napi_gro_cb napi;
819 	int proto;
820 };
821 
822 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
823 
824 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
825 					 struct sk_buff *skb)
826 {
827 	struct inet6_protocol *ops;
828 	struct sk_buff **pp = NULL;
829 	struct sk_buff *p;
830 	struct ipv6hdr *iph;
831 	unsigned int nlen;
832 	unsigned int hlen;
833 	unsigned int off;
834 	int flush = 1;
835 	int proto;
836 	__wsum csum;
837 
838 	off = skb_gro_offset(skb);
839 	hlen = off + sizeof(*iph);
840 	iph = skb_gro_header_fast(skb, off);
841 	if (skb_gro_header_hard(skb, hlen)) {
842 		iph = skb_gro_header_slow(skb, hlen, off);
843 		if (unlikely(!iph))
844 			goto out;
845 	}
846 
847 	skb_gro_pull(skb, sizeof(*iph));
848 	skb_set_transport_header(skb, skb_gro_offset(skb));
849 
850 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
851 
852 	rcu_read_lock();
853 	proto = iph->nexthdr;
854 	ops = rcu_dereference(inet6_protos[proto]);
855 	if (!ops || !ops->gro_receive) {
856 		__pskb_pull(skb, skb_gro_offset(skb));
857 		proto = ipv6_gso_pull_exthdrs(skb, proto);
858 		skb_gro_pull(skb, -skb_transport_offset(skb));
859 		skb_reset_transport_header(skb);
860 		__skb_push(skb, skb_gro_offset(skb));
861 
862 		if (!ops || !ops->gro_receive)
863 			goto out_unlock;
864 
865 		iph = ipv6_hdr(skb);
866 	}
867 
868 	IPV6_GRO_CB(skb)->proto = proto;
869 
870 	flush--;
871 	nlen = skb_network_header_len(skb);
872 
873 	for (p = *head; p; p = p->next) {
874 		struct ipv6hdr *iph2;
875 
876 		if (!NAPI_GRO_CB(p)->same_flow)
877 			continue;
878 
879 		iph2 = ipv6_hdr(p);
880 
881 		/* All fields must match except length. */
882 		if (nlen != skb_network_header_len(p) ||
883 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
884 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
885 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
886 			NAPI_GRO_CB(p)->same_flow = 0;
887 			continue;
888 		}
889 
890 		NAPI_GRO_CB(p)->flush |= flush;
891 	}
892 
893 	NAPI_GRO_CB(skb)->flush |= flush;
894 
895 	csum = skb->csum;
896 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
897 
898 	pp = ops->gro_receive(head, skb);
899 
900 	skb->csum = csum;
901 
902 out_unlock:
903 	rcu_read_unlock();
904 
905 out:
906 	NAPI_GRO_CB(skb)->flush |= flush;
907 
908 	return pp;
909 }
910 
911 static int ipv6_gro_complete(struct sk_buff *skb)
912 {
913 	struct inet6_protocol *ops;
914 	struct ipv6hdr *iph = ipv6_hdr(skb);
915 	int err = -ENOSYS;
916 
917 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
918 				 sizeof(*iph));
919 
920 	rcu_read_lock();
921 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
922 	if (WARN_ON(!ops || !ops->gro_complete))
923 		goto out_unlock;
924 
925 	err = ops->gro_complete(skb);
926 
927 out_unlock:
928 	rcu_read_unlock();
929 
930 	return err;
931 }
932 
933 static struct packet_type ipv6_packet_type __read_mostly = {
934 	.type = cpu_to_be16(ETH_P_IPV6),
935 	.func = ipv6_rcv,
936 	.gso_send_check = ipv6_gso_send_check,
937 	.gso_segment = ipv6_gso_segment,
938 	.gro_receive = ipv6_gro_receive,
939 	.gro_complete = ipv6_gro_complete,
940 };
941 
942 static int __init ipv6_packet_init(void)
943 {
944 	dev_add_pack(&ipv6_packet_type);
945 	return 0;
946 }
947 
948 static void ipv6_packet_cleanup(void)
949 {
950 	dev_remove_pack(&ipv6_packet_type);
951 }
952 
953 static int __net_init ipv6_init_mibs(struct net *net)
954 {
955 	if (snmp_mib_init((void **)net->mib.udp_stats_in6,
956 			  sizeof (struct udp_mib)) < 0)
957 		return -ENOMEM;
958 	if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
959 			  sizeof (struct udp_mib)) < 0)
960 		goto err_udplite_mib;
961 	if (snmp_mib_init((void **)net->mib.ipv6_statistics,
962 			  sizeof(struct ipstats_mib)) < 0)
963 		goto err_ip_mib;
964 	if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
965 			  sizeof(struct icmpv6_mib)) < 0)
966 		goto err_icmp_mib;
967 	if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
968 			  sizeof(struct icmpv6msg_mib)) < 0)
969 		goto err_icmpmsg_mib;
970 	return 0;
971 
972 err_icmpmsg_mib:
973 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
974 err_icmp_mib:
975 	snmp_mib_free((void **)net->mib.ipv6_statistics);
976 err_ip_mib:
977 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
978 err_udplite_mib:
979 	snmp_mib_free((void **)net->mib.udp_stats_in6);
980 	return -ENOMEM;
981 }
982 
983 static void __net_exit ipv6_cleanup_mibs(struct net *net)
984 {
985 	snmp_mib_free((void **)net->mib.udp_stats_in6);
986 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
987 	snmp_mib_free((void **)net->mib.ipv6_statistics);
988 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
989 	snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
990 }
991 
992 static int __net_init inet6_net_init(struct net *net)
993 {
994 	int err = 0;
995 
996 	net->ipv6.sysctl.bindv6only = 0;
997 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
998 
999 	err = ipv6_init_mibs(net);
1000 	if (err)
1001 		return err;
1002 #ifdef CONFIG_PROC_FS
1003 	err = udp6_proc_init(net);
1004 	if (err)
1005 		goto out;
1006 	err = tcp6_proc_init(net);
1007 	if (err)
1008 		goto proc_tcp6_fail;
1009 	err = ac6_proc_init(net);
1010 	if (err)
1011 		goto proc_ac6_fail;
1012 #endif
1013 	return err;
1014 
1015 #ifdef CONFIG_PROC_FS
1016 proc_ac6_fail:
1017 	tcp6_proc_exit(net);
1018 proc_tcp6_fail:
1019 	udp6_proc_exit(net);
1020 out:
1021 	ipv6_cleanup_mibs(net);
1022 	return err;
1023 #endif
1024 }
1025 
1026 static void inet6_net_exit(struct net *net)
1027 {
1028 #ifdef CONFIG_PROC_FS
1029 	udp6_proc_exit(net);
1030 	tcp6_proc_exit(net);
1031 	ac6_proc_exit(net);
1032 #endif
1033 	ipv6_cleanup_mibs(net);
1034 }
1035 
1036 static struct pernet_operations inet6_net_ops = {
1037 	.init = inet6_net_init,
1038 	.exit = inet6_net_exit,
1039 };
1040 
1041 static int __init inet6_init(void)
1042 {
1043 	struct sk_buff *dummy_skb;
1044 	struct list_head *r;
1045 	int err = 0;
1046 
1047 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1048 
1049 	/* Register the socket-side information for inet6_create.  */
1050 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1051 		INIT_LIST_HEAD(r);
1052 
1053 	if (disable_ipv6_mod) {
1054 		printk(KERN_INFO
1055 		       "IPv6: Loaded, but administratively disabled, "
1056 		       "reboot required to enable\n");
1057 		goto out;
1058 	}
1059 
1060 	err = proto_register(&tcpv6_prot, 1);
1061 	if (err)
1062 		goto out;
1063 
1064 	err = proto_register(&udpv6_prot, 1);
1065 	if (err)
1066 		goto out_unregister_tcp_proto;
1067 
1068 	err = proto_register(&udplitev6_prot, 1);
1069 	if (err)
1070 		goto out_unregister_udp_proto;
1071 
1072 	err = proto_register(&rawv6_prot, 1);
1073 	if (err)
1074 		goto out_unregister_udplite_proto;
1075 
1076 
1077 	/* We MUST register RAW sockets before we create the ICMP6,
1078 	 * IGMP6, or NDISC control sockets.
1079 	 */
1080 	err = rawv6_init();
1081 	if (err)
1082 		goto out_unregister_raw_proto;
1083 
1084 	/* Register the family here so that the init calls below will
1085 	 * be able to create sockets. (?? is this dangerous ??)
1086 	 */
1087 	err = sock_register(&inet6_family_ops);
1088 	if (err)
1089 		goto out_sock_register_fail;
1090 
1091 #ifdef CONFIG_SYSCTL
1092 	err = ipv6_static_sysctl_register();
1093 	if (err)
1094 		goto static_sysctl_fail;
1095 #endif
1096 	/*
1097 	 *	ipngwg API draft makes clear that the correct semantics
1098 	 *	for TCP and UDP is to consider one TCP and UDP instance
1099 	 *	in a host availiable by both INET and INET6 APIs and
1100 	 *	able to communicate via both network protocols.
1101 	 */
1102 
1103 	err = register_pernet_subsys(&inet6_net_ops);
1104 	if (err)
1105 		goto register_pernet_fail;
1106 	err = icmpv6_init();
1107 	if (err)
1108 		goto icmp_fail;
1109 	err = ip6_mr_init();
1110 	if (err)
1111 		goto ipmr_fail;
1112 	err = ndisc_init();
1113 	if (err)
1114 		goto ndisc_fail;
1115 	err = igmp6_init();
1116 	if (err)
1117 		goto igmp_fail;
1118 	err = ipv6_netfilter_init();
1119 	if (err)
1120 		goto netfilter_fail;
1121 	/* Create /proc/foo6 entries. */
1122 #ifdef CONFIG_PROC_FS
1123 	err = -ENOMEM;
1124 	if (raw6_proc_init())
1125 		goto proc_raw6_fail;
1126 	if (udplite6_proc_init())
1127 		goto proc_udplite6_fail;
1128 	if (ipv6_misc_proc_init())
1129 		goto proc_misc6_fail;
1130 	if (if6_proc_init())
1131 		goto proc_if6_fail;
1132 #endif
1133 	err = ip6_route_init();
1134 	if (err)
1135 		goto ip6_route_fail;
1136 	err = ip6_flowlabel_init();
1137 	if (err)
1138 		goto ip6_flowlabel_fail;
1139 	err = addrconf_init();
1140 	if (err)
1141 		goto addrconf_fail;
1142 
1143 	/* Init v6 extension headers. */
1144 	err = ipv6_exthdrs_init();
1145 	if (err)
1146 		goto ipv6_exthdrs_fail;
1147 
1148 	err = ipv6_frag_init();
1149 	if (err)
1150 		goto ipv6_frag_fail;
1151 
1152 	/* Init v6 transport protocols. */
1153 	err = udpv6_init();
1154 	if (err)
1155 		goto udpv6_fail;
1156 
1157 	err = udplitev6_init();
1158 	if (err)
1159 		goto udplitev6_fail;
1160 
1161 	err = tcpv6_init();
1162 	if (err)
1163 		goto tcpv6_fail;
1164 
1165 	err = ipv6_packet_init();
1166 	if (err)
1167 		goto ipv6_packet_fail;
1168 
1169 #ifdef CONFIG_SYSCTL
1170 	err = ipv6_sysctl_register();
1171 	if (err)
1172 		goto sysctl_fail;
1173 #endif
1174 out:
1175 	return err;
1176 
1177 #ifdef CONFIG_SYSCTL
1178 sysctl_fail:
1179 	ipv6_packet_cleanup();
1180 #endif
1181 ipv6_packet_fail:
1182 	tcpv6_exit();
1183 tcpv6_fail:
1184 	udplitev6_exit();
1185 udplitev6_fail:
1186 	udpv6_exit();
1187 udpv6_fail:
1188 	ipv6_frag_exit();
1189 ipv6_frag_fail:
1190 	ipv6_exthdrs_exit();
1191 ipv6_exthdrs_fail:
1192 	addrconf_cleanup();
1193 addrconf_fail:
1194 	ip6_flowlabel_cleanup();
1195 ip6_flowlabel_fail:
1196 	ip6_route_cleanup();
1197 ip6_route_fail:
1198 #ifdef CONFIG_PROC_FS
1199 	if6_proc_exit();
1200 proc_if6_fail:
1201 	ipv6_misc_proc_exit();
1202 proc_misc6_fail:
1203 	udplite6_proc_exit();
1204 proc_udplite6_fail:
1205 	raw6_proc_exit();
1206 proc_raw6_fail:
1207 #endif
1208 	ipv6_netfilter_fini();
1209 netfilter_fail:
1210 	igmp6_cleanup();
1211 igmp_fail:
1212 	ndisc_cleanup();
1213 ndisc_fail:
1214 	ip6_mr_cleanup();
1215 ipmr_fail:
1216 	icmpv6_cleanup();
1217 icmp_fail:
1218 	unregister_pernet_subsys(&inet6_net_ops);
1219 register_pernet_fail:
1220 #ifdef CONFIG_SYSCTL
1221 	ipv6_static_sysctl_unregister();
1222 static_sysctl_fail:
1223 #endif
1224 	sock_unregister(PF_INET6);
1225 	rtnl_unregister_all(PF_INET6);
1226 out_sock_register_fail:
1227 	rawv6_exit();
1228 out_unregister_raw_proto:
1229 	proto_unregister(&rawv6_prot);
1230 out_unregister_udplite_proto:
1231 	proto_unregister(&udplitev6_prot);
1232 out_unregister_udp_proto:
1233 	proto_unregister(&udpv6_prot);
1234 out_unregister_tcp_proto:
1235 	proto_unregister(&tcpv6_prot);
1236 	goto out;
1237 }
1238 module_init(inet6_init);
1239 
1240 static void __exit inet6_exit(void)
1241 {
1242 	if (disable_ipv6_mod)
1243 		return;
1244 
1245 	/* First of all disallow new sockets creation. */
1246 	sock_unregister(PF_INET6);
1247 	/* Disallow any further netlink messages */
1248 	rtnl_unregister_all(PF_INET6);
1249 
1250 #ifdef CONFIG_SYSCTL
1251 	ipv6_sysctl_unregister();
1252 #endif
1253 	udpv6_exit();
1254 	udplitev6_exit();
1255 	tcpv6_exit();
1256 
1257 	/* Cleanup code parts. */
1258 	ipv6_packet_cleanup();
1259 	ipv6_frag_exit();
1260 	ipv6_exthdrs_exit();
1261 	addrconf_cleanup();
1262 	ip6_flowlabel_cleanup();
1263 	ip6_route_cleanup();
1264 #ifdef CONFIG_PROC_FS
1265 
1266 	/* Cleanup code parts. */
1267 	if6_proc_exit();
1268 	ipv6_misc_proc_exit();
1269 	udplite6_proc_exit();
1270 	raw6_proc_exit();
1271 #endif
1272 	ipv6_netfilter_fini();
1273 	igmp6_cleanup();
1274 	ndisc_cleanup();
1275 	ip6_mr_cleanup();
1276 	icmpv6_cleanup();
1277 	rawv6_exit();
1278 
1279 	unregister_pernet_subsys(&inet6_net_ops);
1280 #ifdef CONFIG_SYSCTL
1281 	ipv6_static_sysctl_unregister();
1282 #endif
1283 	proto_unregister(&rawv6_prot);
1284 	proto_unregister(&udplitev6_prot);
1285 	proto_unregister(&udpv6_prot);
1286 	proto_unregister(&tcpv6_prot);
1287 
1288 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1289 }
1290 module_exit(inet6_exit);
1291 
1292 MODULE_ALIAS_NETPROTO(PF_INET6);
1293