xref: /linux/net/ipv6/af_inet6.c (revision 60e13231561b3a4c5269bfa1ef6c0569ad6f28ec)
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	= IPV6_DEFAULT_MCASTHOPS;
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 
276 	if (addr->sin6_family != AF_INET6)
277 		return -EAFNOSUPPORT;
278 
279 	addr_type = ipv6_addr_type(&addr->sin6_addr);
280 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
281 		return -EINVAL;
282 
283 	snum = ntohs(addr->sin6_port);
284 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
285 		return -EACCES;
286 
287 	lock_sock(sk);
288 
289 	/* Check these errors (active socket, double bind). */
290 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
291 		err = -EINVAL;
292 		goto out;
293 	}
294 
295 	/* Check if the address belongs to the host. */
296 	if (addr_type == IPV6_ADDR_MAPPED) {
297 		int chk_addr_ret;
298 
299 		/* Binding to v4-mapped address on a v6-only socket
300 		 * makes no sense
301 		 */
302 		if (np->ipv6only) {
303 			err = -EINVAL;
304 			goto out;
305 		}
306 
307 		/* Reproduce AF_INET checks to make the bindings consistent */
308 		v4addr = addr->sin6_addr.s6_addr32[3];
309 		chk_addr_ret = inet_addr_type(net, v4addr);
310 		if (!sysctl_ip_nonlocal_bind &&
311 		    !(inet->freebind || inet->transparent) &&
312 		    v4addr != htonl(INADDR_ANY) &&
313 		    chk_addr_ret != RTN_LOCAL &&
314 		    chk_addr_ret != RTN_MULTICAST &&
315 		    chk_addr_ret != RTN_BROADCAST) {
316 			err = -EADDRNOTAVAIL;
317 			goto out;
318 		}
319 	} else {
320 		if (addr_type != IPV6_ADDR_ANY) {
321 			struct net_device *dev = NULL;
322 
323 			rcu_read_lock();
324 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
325 				if (addr_len >= sizeof(struct sockaddr_in6) &&
326 				    addr->sin6_scope_id) {
327 					/* Override any existing binding, if another one
328 					 * is supplied by user.
329 					 */
330 					sk->sk_bound_dev_if = addr->sin6_scope_id;
331 				}
332 
333 				/* Binding to link-local address requires an interface */
334 				if (!sk->sk_bound_dev_if) {
335 					err = -EINVAL;
336 					goto out_unlock;
337 				}
338 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
339 				if (!dev) {
340 					err = -ENODEV;
341 					goto out_unlock;
342 				}
343 			}
344 
345 			/* ipv4 addr of the socket is invalid.  Only the
346 			 * unspecified and mapped address have a v4 equivalent.
347 			 */
348 			v4addr = LOOPBACK4_IPV6;
349 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
350 				if (!inet->transparent &&
351 				    !ipv6_chk_addr(net, &addr->sin6_addr,
352 						   dev, 0)) {
353 					err = -EADDRNOTAVAIL;
354 					goto out_unlock;
355 				}
356 			}
357 			rcu_read_unlock();
358 		}
359 	}
360 
361 	inet->inet_rcv_saddr = v4addr;
362 	inet->inet_saddr = v4addr;
363 
364 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
365 
366 	if (!(addr_type & IPV6_ADDR_MULTICAST))
367 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
368 
369 	/* Make sure we are allowed to bind here. */
370 	if (sk->sk_prot->get_port(sk, snum)) {
371 		inet_reset_saddr(sk);
372 		err = -EADDRINUSE;
373 		goto out;
374 	}
375 
376 	if (addr_type != IPV6_ADDR_ANY) {
377 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
378 		if (addr_type != IPV6_ADDR_MAPPED)
379 			np->ipv6only = 1;
380 	}
381 	if (snum)
382 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
383 	inet->inet_sport = htons(inet->inet_num);
384 	inet->inet_dport = 0;
385 	inet->inet_daddr = 0;
386 out:
387 	release_sock(sk);
388 	return err;
389 out_unlock:
390 	rcu_read_unlock();
391 	goto out;
392 }
393 
394 EXPORT_SYMBOL(inet6_bind);
395 
396 int inet6_release(struct socket *sock)
397 {
398 	struct sock *sk = sock->sk;
399 
400 	if (sk == NULL)
401 		return -EINVAL;
402 
403 	/* Free mc lists */
404 	ipv6_sock_mc_close(sk);
405 
406 	/* Free ac lists */
407 	ipv6_sock_ac_close(sk);
408 
409 	return inet_release(sock);
410 }
411 
412 EXPORT_SYMBOL(inet6_release);
413 
414 void inet6_destroy_sock(struct sock *sk)
415 {
416 	struct ipv6_pinfo *np = inet6_sk(sk);
417 	struct sk_buff *skb;
418 	struct ipv6_txoptions *opt;
419 
420 	/* Release rx options */
421 
422 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
423 		kfree_skb(skb);
424 
425 	if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
426 		kfree_skb(skb);
427 
428 	/* Free flowlabels */
429 	fl6_free_socklist(sk);
430 
431 	/* Free tx options */
432 
433 	if ((opt = xchg(&np->opt, NULL)) != NULL)
434 		sock_kfree_s(sk, opt, opt->tot_len);
435 }
436 
437 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
438 
439 /*
440  *	This does both peername and sockname.
441  */
442 
443 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
444 		 int *uaddr_len, int peer)
445 {
446 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
447 	struct sock *sk = sock->sk;
448 	struct inet_sock *inet = inet_sk(sk);
449 	struct ipv6_pinfo *np = inet6_sk(sk);
450 
451 	sin->sin6_family = AF_INET6;
452 	sin->sin6_flowinfo = 0;
453 	sin->sin6_scope_id = 0;
454 	if (peer) {
455 		if (!inet->inet_dport)
456 			return -ENOTCONN;
457 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
458 		    peer == 1)
459 			return -ENOTCONN;
460 		sin->sin6_port = inet->inet_dport;
461 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
462 		if (np->sndflow)
463 			sin->sin6_flowinfo = np->flow_label;
464 	} else {
465 		if (ipv6_addr_any(&np->rcv_saddr))
466 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
467 		else
468 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
469 
470 		sin->sin6_port = inet->inet_sport;
471 	}
472 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
473 		sin->sin6_scope_id = sk->sk_bound_dev_if;
474 	*uaddr_len = sizeof(*sin);
475 	return 0;
476 }
477 
478 EXPORT_SYMBOL(inet6_getname);
479 
480 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
481 {
482 	struct sock *sk = sock->sk;
483 	struct net *net = sock_net(sk);
484 
485 	switch(cmd)
486 	{
487 	case SIOCGSTAMP:
488 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
489 
490 	case SIOCGSTAMPNS:
491 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
492 
493 	case SIOCADDRT:
494 	case SIOCDELRT:
495 
496 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
497 
498 	case SIOCSIFADDR:
499 		return addrconf_add_ifaddr(net, (void __user *) arg);
500 	case SIOCDIFADDR:
501 		return addrconf_del_ifaddr(net, (void __user *) arg);
502 	case SIOCSIFDSTADDR:
503 		return addrconf_set_dstaddr(net, (void __user *) arg);
504 	default:
505 		if (!sk->sk_prot->ioctl)
506 			return -ENOIOCTLCMD;
507 		return sk->sk_prot->ioctl(sk, cmd, arg);
508 	}
509 	/*NOTREACHED*/
510 	return 0;
511 }
512 
513 EXPORT_SYMBOL(inet6_ioctl);
514 
515 const struct proto_ops inet6_stream_ops = {
516 	.family		   = PF_INET6,
517 	.owner		   = THIS_MODULE,
518 	.release	   = inet6_release,
519 	.bind		   = inet6_bind,
520 	.connect	   = inet_stream_connect,	/* ok		*/
521 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
522 	.accept		   = inet_accept,		/* ok		*/
523 	.getname	   = inet6_getname,
524 	.poll		   = tcp_poll,			/* ok		*/
525 	.ioctl		   = inet6_ioctl,		/* must change  */
526 	.listen		   = inet_listen,		/* ok		*/
527 	.shutdown	   = inet_shutdown,		/* ok		*/
528 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
529 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
530 	.sendmsg	   = inet_sendmsg,		/* ok		*/
531 	.recvmsg	   = inet_recvmsg,		/* ok		*/
532 	.mmap		   = sock_no_mmap,
533 	.sendpage	   = inet_sendpage,
534 	.splice_read	   = tcp_splice_read,
535 #ifdef CONFIG_COMPAT
536 	.compat_setsockopt = compat_sock_common_setsockopt,
537 	.compat_getsockopt = compat_sock_common_getsockopt,
538 #endif
539 };
540 
541 const struct proto_ops inet6_dgram_ops = {
542 	.family		   = PF_INET6,
543 	.owner		   = THIS_MODULE,
544 	.release	   = inet6_release,
545 	.bind		   = inet6_bind,
546 	.connect	   = inet_dgram_connect,	/* ok		*/
547 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
548 	.accept		   = sock_no_accept,		/* a do nothing	*/
549 	.getname	   = inet6_getname,
550 	.poll		   = udp_poll,			/* ok		*/
551 	.ioctl		   = inet6_ioctl,		/* must change  */
552 	.listen		   = sock_no_listen,		/* ok		*/
553 	.shutdown	   = inet_shutdown,		/* ok		*/
554 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
555 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
556 	.sendmsg	   = inet_sendmsg,		/* ok		*/
557 	.recvmsg	   = inet_recvmsg,		/* ok		*/
558 	.mmap		   = sock_no_mmap,
559 	.sendpage	   = sock_no_sendpage,
560 #ifdef CONFIG_COMPAT
561 	.compat_setsockopt = compat_sock_common_setsockopt,
562 	.compat_getsockopt = compat_sock_common_getsockopt,
563 #endif
564 };
565 
566 static const struct net_proto_family inet6_family_ops = {
567 	.family = PF_INET6,
568 	.create = inet6_create,
569 	.owner	= THIS_MODULE,
570 };
571 
572 int inet6_register_protosw(struct inet_protosw *p)
573 {
574 	struct list_head *lh;
575 	struct inet_protosw *answer;
576 	struct list_head *last_perm;
577 	int protocol = p->protocol;
578 	int ret;
579 
580 	spin_lock_bh(&inetsw6_lock);
581 
582 	ret = -EINVAL;
583 	if (p->type >= SOCK_MAX)
584 		goto out_illegal;
585 
586 	/* If we are trying to override a permanent protocol, bail. */
587 	answer = NULL;
588 	ret = -EPERM;
589 	last_perm = &inetsw6[p->type];
590 	list_for_each(lh, &inetsw6[p->type]) {
591 		answer = list_entry(lh, struct inet_protosw, list);
592 
593 		/* Check only the non-wild match. */
594 		if (INET_PROTOSW_PERMANENT & answer->flags) {
595 			if (protocol == answer->protocol)
596 				break;
597 			last_perm = lh;
598 		}
599 
600 		answer = NULL;
601 	}
602 	if (answer)
603 		goto out_permanent;
604 
605 	/* Add the new entry after the last permanent entry if any, so that
606 	 * the new entry does not override a permanent entry when matched with
607 	 * a wild-card protocol. But it is allowed to override any existing
608 	 * non-permanent entry.  This means that when we remove this entry, the
609 	 * system automatically returns to the old behavior.
610 	 */
611 	list_add_rcu(&p->list, last_perm);
612 	ret = 0;
613 out:
614 	spin_unlock_bh(&inetsw6_lock);
615 	return ret;
616 
617 out_permanent:
618 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
619 	       protocol);
620 	goto out;
621 
622 out_illegal:
623 	printk(KERN_ERR
624 	       "Ignoring attempt to register invalid socket type %d.\n",
625 	       p->type);
626 	goto out;
627 }
628 
629 EXPORT_SYMBOL(inet6_register_protosw);
630 
631 void
632 inet6_unregister_protosw(struct inet_protosw *p)
633 {
634 	if (INET_PROTOSW_PERMANENT & p->flags) {
635 		printk(KERN_ERR
636 		       "Attempt to unregister permanent protocol %d.\n",
637 		       p->protocol);
638 	} else {
639 		spin_lock_bh(&inetsw6_lock);
640 		list_del_rcu(&p->list);
641 		spin_unlock_bh(&inetsw6_lock);
642 
643 		synchronize_net();
644 	}
645 }
646 
647 EXPORT_SYMBOL(inet6_unregister_protosw);
648 
649 int inet6_sk_rebuild_header(struct sock *sk)
650 {
651 	struct ipv6_pinfo *np = inet6_sk(sk);
652 	struct dst_entry *dst;
653 
654 	dst = __sk_dst_check(sk, np->dst_cookie);
655 
656 	if (dst == NULL) {
657 		struct inet_sock *inet = inet_sk(sk);
658 		struct in6_addr *final_p, final;
659 		struct flowi6 fl6;
660 
661 		memset(&fl6, 0, sizeof(fl6));
662 		fl6.flowi6_proto = sk->sk_protocol;
663 		ipv6_addr_copy(&fl6.daddr, &np->daddr);
664 		ipv6_addr_copy(&fl6.saddr, &np->saddr);
665 		fl6.flowlabel = np->flow_label;
666 		fl6.flowi6_oif = sk->sk_bound_dev_if;
667 		fl6.flowi6_mark = sk->sk_mark;
668 		fl6.fl6_dport = inet->inet_dport;
669 		fl6.fl6_sport = inet->inet_sport;
670 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
671 
672 		final_p = fl6_update_dst(&fl6, np->opt, &final);
673 
674 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
675 		if (IS_ERR(dst)) {
676 			sk->sk_route_caps = 0;
677 			sk->sk_err_soft = -PTR_ERR(dst);
678 			return PTR_ERR(dst);
679 		}
680 
681 		__ip6_dst_store(sk, dst, NULL, NULL);
682 	}
683 
684 	return 0;
685 }
686 
687 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
688 
689 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
690 {
691 	struct ipv6_pinfo *np = inet6_sk(sk);
692 	struct inet6_skb_parm *opt = IP6CB(skb);
693 
694 	if (np->rxopt.all) {
695 		if ((opt->hop && (np->rxopt.bits.hopopts ||
696 				  np->rxopt.bits.ohopopts)) ||
697 		    ((IPV6_FLOWINFO_MASK &
698 		      *(__be32 *)skb_network_header(skb)) &&
699 		     np->rxopt.bits.rxflow) ||
700 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
701 		     np->rxopt.bits.osrcrt)) ||
702 		    ((opt->dst1 || opt->dst0) &&
703 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
704 			return 1;
705 	}
706 	return 0;
707 }
708 
709 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
710 
711 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
712 {
713 	const struct inet6_protocol *ops = NULL;
714 
715 	for (;;) {
716 		struct ipv6_opt_hdr *opth;
717 		int len;
718 
719 		if (proto != NEXTHDR_HOP) {
720 			ops = rcu_dereference(inet6_protos[proto]);
721 
722 			if (unlikely(!ops))
723 				break;
724 
725 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
726 				break;
727 		}
728 
729 		if (unlikely(!pskb_may_pull(skb, 8)))
730 			break;
731 
732 		opth = (void *)skb->data;
733 		len = ipv6_optlen(opth);
734 
735 		if (unlikely(!pskb_may_pull(skb, len)))
736 			break;
737 
738 		proto = opth->nexthdr;
739 		__skb_pull(skb, len);
740 	}
741 
742 	return proto;
743 }
744 
745 static int ipv6_gso_send_check(struct sk_buff *skb)
746 {
747 	const struct ipv6hdr *ipv6h;
748 	const struct inet6_protocol *ops;
749 	int err = -EINVAL;
750 
751 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
752 		goto out;
753 
754 	ipv6h = ipv6_hdr(skb);
755 	__skb_pull(skb, sizeof(*ipv6h));
756 	err = -EPROTONOSUPPORT;
757 
758 	rcu_read_lock();
759 	ops = rcu_dereference(inet6_protos[
760 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
761 
762 	if (likely(ops && ops->gso_send_check)) {
763 		skb_reset_transport_header(skb);
764 		err = ops->gso_send_check(skb);
765 	}
766 	rcu_read_unlock();
767 
768 out:
769 	return err;
770 }
771 
772 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, u32 features)
773 {
774 	struct sk_buff *segs = ERR_PTR(-EINVAL);
775 	struct ipv6hdr *ipv6h;
776 	const struct inet6_protocol *ops;
777 	int proto;
778 	struct frag_hdr *fptr;
779 	unsigned int unfrag_ip6hlen;
780 	u8 *prevhdr;
781 	int offset = 0;
782 
783 	if (!(features & NETIF_F_V6_CSUM))
784 		features &= ~NETIF_F_SG;
785 
786 	if (unlikely(skb_shinfo(skb)->gso_type &
787 		     ~(SKB_GSO_UDP |
788 		       SKB_GSO_DODGY |
789 		       SKB_GSO_TCP_ECN |
790 		       SKB_GSO_TCPV6 |
791 		       0)))
792 		goto out;
793 
794 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
795 		goto out;
796 
797 	ipv6h = ipv6_hdr(skb);
798 	__skb_pull(skb, sizeof(*ipv6h));
799 	segs = ERR_PTR(-EPROTONOSUPPORT);
800 
801 	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
802 	rcu_read_lock();
803 	ops = rcu_dereference(inet6_protos[proto]);
804 	if (likely(ops && ops->gso_segment)) {
805 		skb_reset_transport_header(skb);
806 		segs = ops->gso_segment(skb, features);
807 	}
808 	rcu_read_unlock();
809 
810 	if (IS_ERR(segs))
811 		goto out;
812 
813 	for (skb = segs; skb; skb = skb->next) {
814 		ipv6h = ipv6_hdr(skb);
815 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
816 					   sizeof(*ipv6h));
817 		if (proto == IPPROTO_UDP) {
818 			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
819 			fptr = (struct frag_hdr *)(skb_network_header(skb) +
820 				unfrag_ip6hlen);
821 			fptr->frag_off = htons(offset);
822 			if (skb->next != NULL)
823 				fptr->frag_off |= htons(IP6_MF);
824 			offset += (ntohs(ipv6h->payload_len) -
825 				   sizeof(struct frag_hdr));
826 		}
827 	}
828 
829 out:
830 	return segs;
831 }
832 
833 struct ipv6_gro_cb {
834 	struct napi_gro_cb napi;
835 	int proto;
836 };
837 
838 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
839 
840 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
841 					 struct sk_buff *skb)
842 {
843 	const struct inet6_protocol *ops;
844 	struct sk_buff **pp = NULL;
845 	struct sk_buff *p;
846 	struct ipv6hdr *iph;
847 	unsigned int nlen;
848 	unsigned int hlen;
849 	unsigned int off;
850 	int flush = 1;
851 	int proto;
852 	__wsum csum;
853 
854 	off = skb_gro_offset(skb);
855 	hlen = off + sizeof(*iph);
856 	iph = skb_gro_header_fast(skb, off);
857 	if (skb_gro_header_hard(skb, hlen)) {
858 		iph = skb_gro_header_slow(skb, hlen, off);
859 		if (unlikely(!iph))
860 			goto out;
861 	}
862 
863 	skb_gro_pull(skb, sizeof(*iph));
864 	skb_set_transport_header(skb, skb_gro_offset(skb));
865 
866 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
867 
868 	rcu_read_lock();
869 	proto = iph->nexthdr;
870 	ops = rcu_dereference(inet6_protos[proto]);
871 	if (!ops || !ops->gro_receive) {
872 		__pskb_pull(skb, skb_gro_offset(skb));
873 		proto = ipv6_gso_pull_exthdrs(skb, proto);
874 		skb_gro_pull(skb, -skb_transport_offset(skb));
875 		skb_reset_transport_header(skb);
876 		__skb_push(skb, skb_gro_offset(skb));
877 
878 		ops = rcu_dereference(inet6_protos[proto]);
879 		if (!ops || !ops->gro_receive)
880 			goto out_unlock;
881 
882 		iph = ipv6_hdr(skb);
883 	}
884 
885 	IPV6_GRO_CB(skb)->proto = proto;
886 
887 	flush--;
888 	nlen = skb_network_header_len(skb);
889 
890 	for (p = *head; p; p = p->next) {
891 		struct ipv6hdr *iph2;
892 
893 		if (!NAPI_GRO_CB(p)->same_flow)
894 			continue;
895 
896 		iph2 = ipv6_hdr(p);
897 
898 		/* All fields must match except length. */
899 		if (nlen != skb_network_header_len(p) ||
900 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
901 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
902 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
903 			NAPI_GRO_CB(p)->same_flow = 0;
904 			continue;
905 		}
906 
907 		NAPI_GRO_CB(p)->flush |= flush;
908 	}
909 
910 	NAPI_GRO_CB(skb)->flush |= flush;
911 
912 	csum = skb->csum;
913 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
914 
915 	pp = ops->gro_receive(head, skb);
916 
917 	skb->csum = csum;
918 
919 out_unlock:
920 	rcu_read_unlock();
921 
922 out:
923 	NAPI_GRO_CB(skb)->flush |= flush;
924 
925 	return pp;
926 }
927 
928 static int ipv6_gro_complete(struct sk_buff *skb)
929 {
930 	const struct inet6_protocol *ops;
931 	struct ipv6hdr *iph = ipv6_hdr(skb);
932 	int err = -ENOSYS;
933 
934 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
935 				 sizeof(*iph));
936 
937 	rcu_read_lock();
938 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
939 	if (WARN_ON(!ops || !ops->gro_complete))
940 		goto out_unlock;
941 
942 	err = ops->gro_complete(skb);
943 
944 out_unlock:
945 	rcu_read_unlock();
946 
947 	return err;
948 }
949 
950 static struct packet_type ipv6_packet_type __read_mostly = {
951 	.type = cpu_to_be16(ETH_P_IPV6),
952 	.func = ipv6_rcv,
953 	.gso_send_check = ipv6_gso_send_check,
954 	.gso_segment = ipv6_gso_segment,
955 	.gro_receive = ipv6_gro_receive,
956 	.gro_complete = ipv6_gro_complete,
957 };
958 
959 static int __init ipv6_packet_init(void)
960 {
961 	dev_add_pack(&ipv6_packet_type);
962 	return 0;
963 }
964 
965 static void ipv6_packet_cleanup(void)
966 {
967 	dev_remove_pack(&ipv6_packet_type);
968 }
969 
970 static int __net_init ipv6_init_mibs(struct net *net)
971 {
972 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
973 			  sizeof(struct udp_mib),
974 			  __alignof__(struct udp_mib)) < 0)
975 		return -ENOMEM;
976 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
977 			  sizeof(struct udp_mib),
978 			  __alignof__(struct udp_mib)) < 0)
979 		goto err_udplite_mib;
980 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
981 			  sizeof(struct ipstats_mib),
982 			  __alignof__(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),
986 			  __alignof__(struct icmpv6_mib)) < 0)
987 		goto err_icmp_mib;
988 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
989 			  sizeof(struct icmpv6msg_mib),
990 			  __alignof__(struct icmpv6msg_mib)) < 0)
991 		goto err_icmpmsg_mib;
992 	return 0;
993 
994 err_icmpmsg_mib:
995 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
996 err_icmp_mib:
997 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
998 err_ip_mib:
999 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1000 err_udplite_mib:
1001 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1002 	return -ENOMEM;
1003 }
1004 
1005 static void ipv6_cleanup_mibs(struct net *net)
1006 {
1007 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1008 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1009 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1010 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1011 	snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1012 }
1013 
1014 static int __net_init inet6_net_init(struct net *net)
1015 {
1016 	int err = 0;
1017 
1018 	net->ipv6.sysctl.bindv6only = 0;
1019 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
1020 
1021 	err = ipv6_init_mibs(net);
1022 	if (err)
1023 		return err;
1024 #ifdef CONFIG_PROC_FS
1025 	err = udp6_proc_init(net);
1026 	if (err)
1027 		goto out;
1028 	err = tcp6_proc_init(net);
1029 	if (err)
1030 		goto proc_tcp6_fail;
1031 	err = ac6_proc_init(net);
1032 	if (err)
1033 		goto proc_ac6_fail;
1034 #endif
1035 	return err;
1036 
1037 #ifdef CONFIG_PROC_FS
1038 proc_ac6_fail:
1039 	tcp6_proc_exit(net);
1040 proc_tcp6_fail:
1041 	udp6_proc_exit(net);
1042 out:
1043 	ipv6_cleanup_mibs(net);
1044 	return err;
1045 #endif
1046 }
1047 
1048 static void __net_exit inet6_net_exit(struct net *net)
1049 {
1050 #ifdef CONFIG_PROC_FS
1051 	udp6_proc_exit(net);
1052 	tcp6_proc_exit(net);
1053 	ac6_proc_exit(net);
1054 #endif
1055 	ipv6_cleanup_mibs(net);
1056 }
1057 
1058 static struct pernet_operations inet6_net_ops = {
1059 	.init = inet6_net_init,
1060 	.exit = inet6_net_exit,
1061 };
1062 
1063 static int __init inet6_init(void)
1064 {
1065 	struct sk_buff *dummy_skb;
1066 	struct list_head *r;
1067 	int err = 0;
1068 
1069 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1070 
1071 	/* Register the socket-side information for inet6_create.  */
1072 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1073 		INIT_LIST_HEAD(r);
1074 
1075 	if (disable_ipv6_mod) {
1076 		printk(KERN_INFO
1077 		       "IPv6: Loaded, but administratively disabled, "
1078 		       "reboot required to enable\n");
1079 		goto out;
1080 	}
1081 
1082 	err = proto_register(&tcpv6_prot, 1);
1083 	if (err)
1084 		goto out;
1085 
1086 	err = proto_register(&udpv6_prot, 1);
1087 	if (err)
1088 		goto out_unregister_tcp_proto;
1089 
1090 	err = proto_register(&udplitev6_prot, 1);
1091 	if (err)
1092 		goto out_unregister_udp_proto;
1093 
1094 	err = proto_register(&rawv6_prot, 1);
1095 	if (err)
1096 		goto out_unregister_udplite_proto;
1097 
1098 
1099 	/* We MUST register RAW sockets before we create the ICMP6,
1100 	 * IGMP6, or NDISC control sockets.
1101 	 */
1102 	err = rawv6_init();
1103 	if (err)
1104 		goto out_unregister_raw_proto;
1105 
1106 	/* Register the family here so that the init calls below will
1107 	 * be able to create sockets. (?? is this dangerous ??)
1108 	 */
1109 	err = sock_register(&inet6_family_ops);
1110 	if (err)
1111 		goto out_sock_register_fail;
1112 
1113 #ifdef CONFIG_SYSCTL
1114 	err = ipv6_static_sysctl_register();
1115 	if (err)
1116 		goto static_sysctl_fail;
1117 #endif
1118 	/*
1119 	 *	ipngwg API draft makes clear that the correct semantics
1120 	 *	for TCP and UDP is to consider one TCP and UDP instance
1121 	 *	in a host available by both INET and INET6 APIs and
1122 	 *	able to communicate via both network protocols.
1123 	 */
1124 
1125 	err = register_pernet_subsys(&inet6_net_ops);
1126 	if (err)
1127 		goto register_pernet_fail;
1128 	err = icmpv6_init();
1129 	if (err)
1130 		goto icmp_fail;
1131 	err = ip6_mr_init();
1132 	if (err)
1133 		goto ipmr_fail;
1134 	err = ndisc_init();
1135 	if (err)
1136 		goto ndisc_fail;
1137 	err = igmp6_init();
1138 	if (err)
1139 		goto igmp_fail;
1140 	err = ipv6_netfilter_init();
1141 	if (err)
1142 		goto netfilter_fail;
1143 	/* Create /proc/foo6 entries. */
1144 #ifdef CONFIG_PROC_FS
1145 	err = -ENOMEM;
1146 	if (raw6_proc_init())
1147 		goto proc_raw6_fail;
1148 	if (udplite6_proc_init())
1149 		goto proc_udplite6_fail;
1150 	if (ipv6_misc_proc_init())
1151 		goto proc_misc6_fail;
1152 	if (if6_proc_init())
1153 		goto proc_if6_fail;
1154 #endif
1155 	err = ip6_route_init();
1156 	if (err)
1157 		goto ip6_route_fail;
1158 	err = ip6_flowlabel_init();
1159 	if (err)
1160 		goto ip6_flowlabel_fail;
1161 	err = addrconf_init();
1162 	if (err)
1163 		goto addrconf_fail;
1164 
1165 	/* Init v6 extension headers. */
1166 	err = ipv6_exthdrs_init();
1167 	if (err)
1168 		goto ipv6_exthdrs_fail;
1169 
1170 	err = ipv6_frag_init();
1171 	if (err)
1172 		goto ipv6_frag_fail;
1173 
1174 	/* Init v6 transport protocols. */
1175 	err = udpv6_init();
1176 	if (err)
1177 		goto udpv6_fail;
1178 
1179 	err = udplitev6_init();
1180 	if (err)
1181 		goto udplitev6_fail;
1182 
1183 	err = tcpv6_init();
1184 	if (err)
1185 		goto tcpv6_fail;
1186 
1187 	err = ipv6_packet_init();
1188 	if (err)
1189 		goto ipv6_packet_fail;
1190 
1191 #ifdef CONFIG_SYSCTL
1192 	err = ipv6_sysctl_register();
1193 	if (err)
1194 		goto sysctl_fail;
1195 #endif
1196 out:
1197 	return err;
1198 
1199 #ifdef CONFIG_SYSCTL
1200 sysctl_fail:
1201 	ipv6_packet_cleanup();
1202 #endif
1203 ipv6_packet_fail:
1204 	tcpv6_exit();
1205 tcpv6_fail:
1206 	udplitev6_exit();
1207 udplitev6_fail:
1208 	udpv6_exit();
1209 udpv6_fail:
1210 	ipv6_frag_exit();
1211 ipv6_frag_fail:
1212 	ipv6_exthdrs_exit();
1213 ipv6_exthdrs_fail:
1214 	addrconf_cleanup();
1215 addrconf_fail:
1216 	ip6_flowlabel_cleanup();
1217 ip6_flowlabel_fail:
1218 	ip6_route_cleanup();
1219 ip6_route_fail:
1220 #ifdef CONFIG_PROC_FS
1221 	if6_proc_exit();
1222 proc_if6_fail:
1223 	ipv6_misc_proc_exit();
1224 proc_misc6_fail:
1225 	udplite6_proc_exit();
1226 proc_udplite6_fail:
1227 	raw6_proc_exit();
1228 proc_raw6_fail:
1229 #endif
1230 	ipv6_netfilter_fini();
1231 netfilter_fail:
1232 	igmp6_cleanup();
1233 igmp_fail:
1234 	ndisc_cleanup();
1235 ndisc_fail:
1236 	ip6_mr_cleanup();
1237 ipmr_fail:
1238 	icmpv6_cleanup();
1239 icmp_fail:
1240 	unregister_pernet_subsys(&inet6_net_ops);
1241 register_pernet_fail:
1242 #ifdef CONFIG_SYSCTL
1243 	ipv6_static_sysctl_unregister();
1244 static_sysctl_fail:
1245 #endif
1246 	sock_unregister(PF_INET6);
1247 	rtnl_unregister_all(PF_INET6);
1248 out_sock_register_fail:
1249 	rawv6_exit();
1250 out_unregister_raw_proto:
1251 	proto_unregister(&rawv6_prot);
1252 out_unregister_udplite_proto:
1253 	proto_unregister(&udplitev6_prot);
1254 out_unregister_udp_proto:
1255 	proto_unregister(&udpv6_prot);
1256 out_unregister_tcp_proto:
1257 	proto_unregister(&tcpv6_prot);
1258 	goto out;
1259 }
1260 module_init(inet6_init);
1261 
1262 static void __exit inet6_exit(void)
1263 {
1264 	if (disable_ipv6_mod)
1265 		return;
1266 
1267 	/* First of all disallow new sockets creation. */
1268 	sock_unregister(PF_INET6);
1269 	/* Disallow any further netlink messages */
1270 	rtnl_unregister_all(PF_INET6);
1271 
1272 #ifdef CONFIG_SYSCTL
1273 	ipv6_sysctl_unregister();
1274 #endif
1275 	udpv6_exit();
1276 	udplitev6_exit();
1277 	tcpv6_exit();
1278 
1279 	/* Cleanup code parts. */
1280 	ipv6_packet_cleanup();
1281 	ipv6_frag_exit();
1282 	ipv6_exthdrs_exit();
1283 	addrconf_cleanup();
1284 	ip6_flowlabel_cleanup();
1285 	ip6_route_cleanup();
1286 #ifdef CONFIG_PROC_FS
1287 
1288 	/* Cleanup code parts. */
1289 	if6_proc_exit();
1290 	ipv6_misc_proc_exit();
1291 	udplite6_proc_exit();
1292 	raw6_proc_exit();
1293 #endif
1294 	ipv6_netfilter_fini();
1295 	igmp6_cleanup();
1296 	ndisc_cleanup();
1297 	ip6_mr_cleanup();
1298 	icmpv6_cleanup();
1299 	rawv6_exit();
1300 
1301 	unregister_pernet_subsys(&inet6_net_ops);
1302 #ifdef CONFIG_SYSCTL
1303 	ipv6_static_sysctl_unregister();
1304 #endif
1305 	proto_unregister(&rawv6_prot);
1306 	proto_unregister(&udplitev6_prot);
1307 	proto_unregister(&udpv6_prot);
1308 	proto_unregister(&tcpv6_prot);
1309 
1310 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1311 }
1312 module_exit(inet6_exit);
1313 
1314 MODULE_ALIAS_NETPROTO(PF_INET6);
1315