xref: /openbmc/linux/net/ipv6/af_inet6.c (revision b627b4ed)
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 static int disable_ipv6 = 0;
76 module_param_named(disable, disable_ipv6, int, 0);
77 MODULE_PARM_DESC(disable, "Disable IPv6 such that it is non-functional");
78 
79 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
80 {
81 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
82 
83 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
84 }
85 
86 static int inet6_create(struct net *net, struct socket *sock, int protocol)
87 {
88 	struct inet_sock *inet;
89 	struct ipv6_pinfo *np;
90 	struct sock *sk;
91 	struct inet_protosw *answer;
92 	struct proto *answer_prot;
93 	unsigned char answer_flags;
94 	char answer_no_check;
95 	int try_loading_module = 0;
96 	int err;
97 
98 	if (sock->type != SOCK_RAW &&
99 	    sock->type != SOCK_DGRAM &&
100 	    !inet_ehash_secret)
101 		build_ehash_secret();
102 
103 	/* Look for the requested type/protocol pair. */
104 lookup_protocol:
105 	err = -ESOCKTNOSUPPORT;
106 	rcu_read_lock();
107 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
108 
109 		err = 0;
110 		/* Check the non-wild match. */
111 		if (protocol == answer->protocol) {
112 			if (protocol != IPPROTO_IP)
113 				break;
114 		} else {
115 			/* Check for the two wild cases. */
116 			if (IPPROTO_IP == protocol) {
117 				protocol = answer->protocol;
118 				break;
119 			}
120 			if (IPPROTO_IP == answer->protocol)
121 				break;
122 		}
123 		err = -EPROTONOSUPPORT;
124 	}
125 
126 	if (err) {
127 		if (try_loading_module < 2) {
128 			rcu_read_unlock();
129 			/*
130 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
131 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
132 			 */
133 			if (++try_loading_module == 1)
134 				request_module("net-pf-%d-proto-%d-type-%d",
135 						PF_INET6, protocol, sock->type);
136 			/*
137 			 * Fall back to generic, e.g. net-pf-10-proto-132
138 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
139 			 */
140 			else
141 				request_module("net-pf-%d-proto-%d",
142 						PF_INET6, protocol);
143 			goto lookup_protocol;
144 		} else
145 			goto out_rcu_unlock;
146 	}
147 
148 	err = -EPERM;
149 	if (answer->capability > 0 && !capable(answer->capability))
150 		goto out_rcu_unlock;
151 
152 	sock->ops = answer->ops;
153 	answer_prot = answer->prot;
154 	answer_no_check = answer->no_check;
155 	answer_flags = answer->flags;
156 	rcu_read_unlock();
157 
158 	WARN_ON(answer_prot->slab == NULL);
159 
160 	err = -ENOBUFS;
161 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
162 	if (sk == NULL)
163 		goto out;
164 
165 	sock_init_data(sock, sk);
166 
167 	err = 0;
168 	sk->sk_no_check = answer_no_check;
169 	if (INET_PROTOSW_REUSE & answer_flags)
170 		sk->sk_reuse = 1;
171 
172 	inet = inet_sk(sk);
173 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
174 
175 	if (SOCK_RAW == sock->type) {
176 		inet->num = protocol;
177 		if (IPPROTO_RAW == protocol)
178 			inet->hdrincl = 1;
179 	}
180 
181 	sk->sk_destruct		= inet_sock_destruct;
182 	sk->sk_family		= PF_INET6;
183 	sk->sk_protocol		= protocol;
184 
185 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
186 
187 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
188 	np->hop_limit	= -1;
189 	np->mcast_hops	= -1;
190 	np->mc_loop	= 1;
191 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
192 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
193 
194 	/* Init the ipv4 part of the socket since we can have sockets
195 	 * using v6 API for ipv4.
196 	 */
197 	inet->uc_ttl	= -1;
198 
199 	inet->mc_loop	= 1;
200 	inet->mc_ttl	= 1;
201 	inet->mc_index	= 0;
202 	inet->mc_list	= NULL;
203 
204 	if (ipv4_config.no_pmtu_disc)
205 		inet->pmtudisc = IP_PMTUDISC_DONT;
206 	else
207 		inet->pmtudisc = IP_PMTUDISC_WANT;
208 	/*
209 	 * Increment only the relevant sk_prot->socks debug field, this changes
210 	 * the previous behaviour of incrementing both the equivalent to
211 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
212 	 *
213 	 * This allows better debug granularity as we'll know exactly how many
214 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
215 	 * transport protocol socks. -acme
216 	 */
217 	sk_refcnt_debug_inc(sk);
218 
219 	if (inet->num) {
220 		/* It assumes that any protocol which allows
221 		 * the user to assign a number at socket
222 		 * creation time automatically shares.
223 		 */
224 		inet->sport = htons(inet->num);
225 		sk->sk_prot->hash(sk);
226 	}
227 	if (sk->sk_prot->init) {
228 		err = sk->sk_prot->init(sk);
229 		if (err) {
230 			sk_common_release(sk);
231 			goto out;
232 		}
233 	}
234 out:
235 	return err;
236 out_rcu_unlock:
237 	rcu_read_unlock();
238 	goto out;
239 }
240 
241 
242 /* bind for INET6 API */
243 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
244 {
245 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
246 	struct sock *sk = sock->sk;
247 	struct inet_sock *inet = inet_sk(sk);
248 	struct ipv6_pinfo *np = inet6_sk(sk);
249 	struct net *net = sock_net(sk);
250 	__be32 v4addr = 0;
251 	unsigned short snum;
252 	int addr_type = 0;
253 	int err = 0;
254 
255 	/* If the socket has its own bind function then use it. */
256 	if (sk->sk_prot->bind)
257 		return sk->sk_prot->bind(sk, uaddr, addr_len);
258 
259 	if (addr_len < SIN6_LEN_RFC2133)
260 		return -EINVAL;
261 	addr_type = ipv6_addr_type(&addr->sin6_addr);
262 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
263 		return -EINVAL;
264 
265 	snum = ntohs(addr->sin6_port);
266 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
267 		return -EACCES;
268 
269 	lock_sock(sk);
270 
271 	/* Check these errors (active socket, double bind). */
272 	if (sk->sk_state != TCP_CLOSE || inet->num) {
273 		err = -EINVAL;
274 		goto out;
275 	}
276 
277 	/* Check if the address belongs to the host. */
278 	if (addr_type == IPV6_ADDR_MAPPED) {
279 		int chk_addr_ret;
280 
281 		/* Binding to v4-mapped address on a v6-only socket
282 		 * makes no sense
283 		 */
284 		if (np->ipv6only) {
285 			err = -EINVAL;
286 			goto out;
287 		}
288 
289 		/* Reproduce AF_INET checks to make the bindings consitant */
290 		v4addr = addr->sin6_addr.s6_addr32[3];
291 		chk_addr_ret = inet_addr_type(net, v4addr);
292 		if (!sysctl_ip_nonlocal_bind &&
293 		    !(inet->freebind || inet->transparent) &&
294 		    v4addr != htonl(INADDR_ANY) &&
295 		    chk_addr_ret != RTN_LOCAL &&
296 		    chk_addr_ret != RTN_MULTICAST &&
297 		    chk_addr_ret != RTN_BROADCAST)
298 			goto out;
299 	} else {
300 		if (addr_type != IPV6_ADDR_ANY) {
301 			struct net_device *dev = NULL;
302 
303 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
304 				if (addr_len >= sizeof(struct sockaddr_in6) &&
305 				    addr->sin6_scope_id) {
306 					/* Override any existing binding, if another one
307 					 * is supplied by user.
308 					 */
309 					sk->sk_bound_dev_if = addr->sin6_scope_id;
310 				}
311 
312 				/* Binding to link-local address requires an interface */
313 				if (!sk->sk_bound_dev_if) {
314 					err = -EINVAL;
315 					goto out;
316 				}
317 				dev = dev_get_by_index(net, sk->sk_bound_dev_if);
318 				if (!dev) {
319 					err = -ENODEV;
320 					goto out;
321 				}
322 			}
323 
324 			/* ipv4 addr of the socket is invalid.  Only the
325 			 * unspecified and mapped address have a v4 equivalent.
326 			 */
327 			v4addr = LOOPBACK4_IPV6;
328 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
329 				if (!ipv6_chk_addr(net, &addr->sin6_addr,
330 						   dev, 0)) {
331 					if (dev)
332 						dev_put(dev);
333 					err = -EADDRNOTAVAIL;
334 					goto out;
335 				}
336 			}
337 			if (dev)
338 				dev_put(dev);
339 		}
340 	}
341 
342 	inet->rcv_saddr = v4addr;
343 	inet->saddr = v4addr;
344 
345 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
346 
347 	if (!(addr_type & IPV6_ADDR_MULTICAST))
348 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
349 
350 	/* Make sure we are allowed to bind here. */
351 	if (sk->sk_prot->get_port(sk, snum)) {
352 		inet_reset_saddr(sk);
353 		err = -EADDRINUSE;
354 		goto out;
355 	}
356 
357 	if (addr_type != IPV6_ADDR_ANY) {
358 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
359 		if (addr_type != IPV6_ADDR_MAPPED)
360 			np->ipv6only = 1;
361 	}
362 	if (snum)
363 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
364 	inet->sport = htons(inet->num);
365 	inet->dport = 0;
366 	inet->daddr = 0;
367 out:
368 	release_sock(sk);
369 	return err;
370 }
371 
372 EXPORT_SYMBOL(inet6_bind);
373 
374 int inet6_release(struct socket *sock)
375 {
376 	struct sock *sk = sock->sk;
377 
378 	if (sk == NULL)
379 		return -EINVAL;
380 
381 	/* Free mc lists */
382 	ipv6_sock_mc_close(sk);
383 
384 	/* Free ac lists */
385 	ipv6_sock_ac_close(sk);
386 
387 	return inet_release(sock);
388 }
389 
390 EXPORT_SYMBOL(inet6_release);
391 
392 void inet6_destroy_sock(struct sock *sk)
393 {
394 	struct ipv6_pinfo *np = inet6_sk(sk);
395 	struct sk_buff *skb;
396 	struct ipv6_txoptions *opt;
397 
398 	/* Release rx options */
399 
400 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
401 		kfree_skb(skb);
402 
403 	/* Free flowlabels */
404 	fl6_free_socklist(sk);
405 
406 	/* Free tx options */
407 
408 	if ((opt = xchg(&np->opt, NULL)) != NULL)
409 		sock_kfree_s(sk, opt, opt->tot_len);
410 }
411 
412 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
413 
414 /*
415  *	This does both peername and sockname.
416  */
417 
418 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
419 		 int *uaddr_len, int peer)
420 {
421 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
422 	struct sock *sk = sock->sk;
423 	struct inet_sock *inet = inet_sk(sk);
424 	struct ipv6_pinfo *np = inet6_sk(sk);
425 
426 	sin->sin6_family = AF_INET6;
427 	sin->sin6_flowinfo = 0;
428 	sin->sin6_scope_id = 0;
429 	if (peer) {
430 		if (!inet->dport)
431 			return -ENOTCONN;
432 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
433 		    peer == 1)
434 			return -ENOTCONN;
435 		sin->sin6_port = inet->dport;
436 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
437 		if (np->sndflow)
438 			sin->sin6_flowinfo = np->flow_label;
439 	} else {
440 		if (ipv6_addr_any(&np->rcv_saddr))
441 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
442 		else
443 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
444 
445 		sin->sin6_port = inet->sport;
446 	}
447 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
448 		sin->sin6_scope_id = sk->sk_bound_dev_if;
449 	*uaddr_len = sizeof(*sin);
450 	return(0);
451 }
452 
453 EXPORT_SYMBOL(inet6_getname);
454 
455 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
456 {
457 	struct sock *sk = sock->sk;
458 	struct net *net = sock_net(sk);
459 
460 	switch(cmd)
461 	{
462 	case SIOCGSTAMP:
463 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
464 
465 	case SIOCGSTAMPNS:
466 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
467 
468 	case SIOCADDRT:
469 	case SIOCDELRT:
470 
471 		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
472 
473 	case SIOCSIFADDR:
474 		return addrconf_add_ifaddr(net, (void __user *) arg);
475 	case SIOCDIFADDR:
476 		return addrconf_del_ifaddr(net, (void __user *) arg);
477 	case SIOCSIFDSTADDR:
478 		return addrconf_set_dstaddr(net, (void __user *) arg);
479 	default:
480 		if (!sk->sk_prot->ioctl)
481 			return -ENOIOCTLCMD;
482 		return sk->sk_prot->ioctl(sk, cmd, arg);
483 	}
484 	/*NOTREACHED*/
485 	return(0);
486 }
487 
488 EXPORT_SYMBOL(inet6_ioctl);
489 
490 const struct proto_ops inet6_stream_ops = {
491 	.family		   = PF_INET6,
492 	.owner		   = THIS_MODULE,
493 	.release	   = inet6_release,
494 	.bind		   = inet6_bind,
495 	.connect	   = inet_stream_connect,	/* ok		*/
496 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
497 	.accept		   = inet_accept,		/* ok		*/
498 	.getname	   = inet6_getname,
499 	.poll		   = tcp_poll,			/* ok		*/
500 	.ioctl		   = inet6_ioctl,		/* must change  */
501 	.listen		   = inet_listen,		/* ok		*/
502 	.shutdown	   = inet_shutdown,		/* ok		*/
503 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
504 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
505 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
506 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
507 	.mmap		   = sock_no_mmap,
508 	.sendpage	   = tcp_sendpage,
509 	.splice_read	   = tcp_splice_read,
510 #ifdef CONFIG_COMPAT
511 	.compat_setsockopt = compat_sock_common_setsockopt,
512 	.compat_getsockopt = compat_sock_common_getsockopt,
513 #endif
514 };
515 
516 const struct proto_ops inet6_dgram_ops = {
517 	.family		   = PF_INET6,
518 	.owner		   = THIS_MODULE,
519 	.release	   = inet6_release,
520 	.bind		   = inet6_bind,
521 	.connect	   = inet_dgram_connect,	/* ok		*/
522 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
523 	.accept		   = sock_no_accept,		/* a do nothing	*/
524 	.getname	   = inet6_getname,
525 	.poll		   = udp_poll,			/* ok		*/
526 	.ioctl		   = inet6_ioctl,		/* must change  */
527 	.listen		   = sock_no_listen,		/* ok		*/
528 	.shutdown	   = inet_shutdown,		/* ok		*/
529 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
530 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
531 	.sendmsg	   = inet_sendmsg,		/* ok		*/
532 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
533 	.mmap		   = sock_no_mmap,
534 	.sendpage	   = sock_no_sendpage,
535 #ifdef CONFIG_COMPAT
536 	.compat_setsockopt = compat_sock_common_setsockopt,
537 	.compat_getsockopt = compat_sock_common_getsockopt,
538 #endif
539 };
540 
541 static struct net_proto_family inet6_family_ops = {
542 	.family = PF_INET6,
543 	.create = inet6_create,
544 	.owner	= THIS_MODULE,
545 };
546 
547 int inet6_register_protosw(struct inet_protosw *p)
548 {
549 	struct list_head *lh;
550 	struct inet_protosw *answer;
551 	struct list_head *last_perm;
552 	int protocol = p->protocol;
553 	int ret;
554 
555 	spin_lock_bh(&inetsw6_lock);
556 
557 	ret = -EINVAL;
558 	if (p->type >= SOCK_MAX)
559 		goto out_illegal;
560 
561 	/* If we are trying to override a permanent protocol, bail. */
562 	answer = NULL;
563 	ret = -EPERM;
564 	last_perm = &inetsw6[p->type];
565 	list_for_each(lh, &inetsw6[p->type]) {
566 		answer = list_entry(lh, struct inet_protosw, list);
567 
568 		/* Check only the non-wild match. */
569 		if (INET_PROTOSW_PERMANENT & answer->flags) {
570 			if (protocol == answer->protocol)
571 				break;
572 			last_perm = lh;
573 		}
574 
575 		answer = NULL;
576 	}
577 	if (answer)
578 		goto out_permanent;
579 
580 	/* Add the new entry after the last permanent entry if any, so that
581 	 * the new entry does not override a permanent entry when matched with
582 	 * a wild-card protocol. But it is allowed to override any existing
583 	 * non-permanent entry.  This means that when we remove this entry, the
584 	 * system automatically returns to the old behavior.
585 	 */
586 	list_add_rcu(&p->list, last_perm);
587 	ret = 0;
588 out:
589 	spin_unlock_bh(&inetsw6_lock);
590 	return ret;
591 
592 out_permanent:
593 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
594 	       protocol);
595 	goto out;
596 
597 out_illegal:
598 	printk(KERN_ERR
599 	       "Ignoring attempt to register invalid socket type %d.\n",
600 	       p->type);
601 	goto out;
602 }
603 
604 EXPORT_SYMBOL(inet6_register_protosw);
605 
606 void
607 inet6_unregister_protosw(struct inet_protosw *p)
608 {
609 	if (INET_PROTOSW_PERMANENT & p->flags) {
610 		printk(KERN_ERR
611 		       "Attempt to unregister permanent protocol %d.\n",
612 		       p->protocol);
613 	} else {
614 		spin_lock_bh(&inetsw6_lock);
615 		list_del_rcu(&p->list);
616 		spin_unlock_bh(&inetsw6_lock);
617 
618 		synchronize_net();
619 	}
620 }
621 
622 EXPORT_SYMBOL(inet6_unregister_protosw);
623 
624 int inet6_sk_rebuild_header(struct sock *sk)
625 {
626 	int err;
627 	struct dst_entry *dst;
628 	struct ipv6_pinfo *np = inet6_sk(sk);
629 
630 	dst = __sk_dst_check(sk, np->dst_cookie);
631 
632 	if (dst == NULL) {
633 		struct inet_sock *inet = inet_sk(sk);
634 		struct in6_addr *final_p = NULL, final;
635 		struct flowi fl;
636 
637 		memset(&fl, 0, sizeof(fl));
638 		fl.proto = sk->sk_protocol;
639 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
640 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
641 		fl.fl6_flowlabel = np->flow_label;
642 		fl.oif = sk->sk_bound_dev_if;
643 		fl.fl_ip_dport = inet->dport;
644 		fl.fl_ip_sport = inet->sport;
645 		security_sk_classify_flow(sk, &fl);
646 
647 		if (np->opt && np->opt->srcrt) {
648 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
649 			ipv6_addr_copy(&final, &fl.fl6_dst);
650 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
651 			final_p = &final;
652 		}
653 
654 		err = ip6_dst_lookup(sk, &dst, &fl);
655 		if (err) {
656 			sk->sk_route_caps = 0;
657 			return err;
658 		}
659 		if (final_p)
660 			ipv6_addr_copy(&fl.fl6_dst, final_p);
661 
662 		if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
663 			sk->sk_err_soft = -err;
664 			return err;
665 		}
666 
667 		__ip6_dst_store(sk, dst, NULL, NULL);
668 	}
669 
670 	return 0;
671 }
672 
673 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
674 
675 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
676 {
677 	struct ipv6_pinfo *np = inet6_sk(sk);
678 	struct inet6_skb_parm *opt = IP6CB(skb);
679 
680 	if (np->rxopt.all) {
681 		if ((opt->hop && (np->rxopt.bits.hopopts ||
682 				  np->rxopt.bits.ohopopts)) ||
683 		    ((IPV6_FLOWINFO_MASK &
684 		      *(__be32 *)skb_network_header(skb)) &&
685 		     np->rxopt.bits.rxflow) ||
686 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
687 		     np->rxopt.bits.osrcrt)) ||
688 		    ((opt->dst1 || opt->dst0) &&
689 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
690 			return 1;
691 	}
692 	return 0;
693 }
694 
695 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
696 
697 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
698 {
699 	struct inet6_protocol *ops = NULL;
700 
701 	for (;;) {
702 		struct ipv6_opt_hdr *opth;
703 		int len;
704 
705 		if (proto != NEXTHDR_HOP) {
706 			ops = rcu_dereference(inet6_protos[proto]);
707 
708 			if (unlikely(!ops))
709 				break;
710 
711 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
712 				break;
713 		}
714 
715 		if (unlikely(!pskb_may_pull(skb, 8)))
716 			break;
717 
718 		opth = (void *)skb->data;
719 		len = ipv6_optlen(opth);
720 
721 		if (unlikely(!pskb_may_pull(skb, len)))
722 			break;
723 
724 		proto = opth->nexthdr;
725 		__skb_pull(skb, len);
726 	}
727 
728 	return proto;
729 }
730 
731 static int ipv6_gso_send_check(struct sk_buff *skb)
732 {
733 	struct ipv6hdr *ipv6h;
734 	struct inet6_protocol *ops;
735 	int err = -EINVAL;
736 
737 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
738 		goto out;
739 
740 	ipv6h = ipv6_hdr(skb);
741 	__skb_pull(skb, sizeof(*ipv6h));
742 	err = -EPROTONOSUPPORT;
743 
744 	rcu_read_lock();
745 	ops = rcu_dereference(inet6_protos[
746 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
747 
748 	if (likely(ops && ops->gso_send_check)) {
749 		skb_reset_transport_header(skb);
750 		err = ops->gso_send_check(skb);
751 	}
752 	rcu_read_unlock();
753 
754 out:
755 	return err;
756 }
757 
758 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
759 {
760 	struct sk_buff *segs = ERR_PTR(-EINVAL);
761 	struct ipv6hdr *ipv6h;
762 	struct inet6_protocol *ops;
763 
764 	if (!(features & NETIF_F_V6_CSUM))
765 		features &= ~NETIF_F_SG;
766 
767 	if (unlikely(skb_shinfo(skb)->gso_type &
768 		     ~(SKB_GSO_UDP |
769 		       SKB_GSO_DODGY |
770 		       SKB_GSO_TCP_ECN |
771 		       SKB_GSO_TCPV6 |
772 		       0)))
773 		goto out;
774 
775 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
776 		goto out;
777 
778 	ipv6h = ipv6_hdr(skb);
779 	__skb_pull(skb, sizeof(*ipv6h));
780 	segs = ERR_PTR(-EPROTONOSUPPORT);
781 
782 	rcu_read_lock();
783 	ops = rcu_dereference(inet6_protos[
784 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
785 
786 	if (likely(ops && ops->gso_segment)) {
787 		skb_reset_transport_header(skb);
788 		segs = ops->gso_segment(skb, features);
789 	}
790 	rcu_read_unlock();
791 
792 	if (unlikely(IS_ERR(segs)))
793 		goto out;
794 
795 	for (skb = segs; skb; skb = skb->next) {
796 		ipv6h = ipv6_hdr(skb);
797 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
798 					   sizeof(*ipv6h));
799 	}
800 
801 out:
802 	return segs;
803 }
804 
805 struct ipv6_gro_cb {
806 	struct napi_gro_cb napi;
807 	int proto;
808 };
809 
810 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
811 
812 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
813 					 struct sk_buff *skb)
814 {
815 	struct inet6_protocol *ops;
816 	struct sk_buff **pp = NULL;
817 	struct sk_buff *p;
818 	struct ipv6hdr *iph;
819 	unsigned int nlen;
820 	int flush = 1;
821 	int proto;
822 	__wsum csum;
823 
824 	iph = skb_gro_header(skb, sizeof(*iph));
825 	if (unlikely(!iph))
826 		goto out;
827 
828 	skb_gro_pull(skb, sizeof(*iph));
829 	skb_set_transport_header(skb, skb_gro_offset(skb));
830 
831 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
832 
833 	rcu_read_lock();
834 	proto = iph->nexthdr;
835 	ops = rcu_dereference(inet6_protos[proto]);
836 	if (!ops || !ops->gro_receive) {
837 		__pskb_pull(skb, skb_gro_offset(skb));
838 		proto = ipv6_gso_pull_exthdrs(skb, proto);
839 		skb_gro_pull(skb, -skb_transport_offset(skb));
840 		skb_reset_transport_header(skb);
841 		__skb_push(skb, skb_gro_offset(skb));
842 
843 		if (!ops || !ops->gro_receive)
844 			goto out_unlock;
845 
846 		iph = ipv6_hdr(skb);
847 	}
848 
849 	IPV6_GRO_CB(skb)->proto = proto;
850 
851 	flush--;
852 	nlen = skb_network_header_len(skb);
853 
854 	for (p = *head; p; p = p->next) {
855 		struct ipv6hdr *iph2;
856 
857 		if (!NAPI_GRO_CB(p)->same_flow)
858 			continue;
859 
860 		iph2 = ipv6_hdr(p);
861 
862 		/* All fields must match except length. */
863 		if (nlen != skb_network_header_len(p) ||
864 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
865 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
866 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
867 			NAPI_GRO_CB(p)->same_flow = 0;
868 			continue;
869 		}
870 
871 		NAPI_GRO_CB(p)->flush |= flush;
872 	}
873 
874 	NAPI_GRO_CB(skb)->flush |= flush;
875 
876 	csum = skb->csum;
877 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
878 
879 	pp = ops->gro_receive(head, skb);
880 
881 	skb->csum = csum;
882 
883 out_unlock:
884 	rcu_read_unlock();
885 
886 out:
887 	NAPI_GRO_CB(skb)->flush |= flush;
888 
889 	return pp;
890 }
891 
892 static int ipv6_gro_complete(struct sk_buff *skb)
893 {
894 	struct inet6_protocol *ops;
895 	struct ipv6hdr *iph = ipv6_hdr(skb);
896 	int err = -ENOSYS;
897 
898 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
899 				 sizeof(*iph));
900 
901 	rcu_read_lock();
902 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
903 	if (WARN_ON(!ops || !ops->gro_complete))
904 		goto out_unlock;
905 
906 	err = ops->gro_complete(skb);
907 
908 out_unlock:
909 	rcu_read_unlock();
910 
911 	return err;
912 }
913 
914 static struct packet_type ipv6_packet_type __read_mostly = {
915 	.type = cpu_to_be16(ETH_P_IPV6),
916 	.func = ipv6_rcv,
917 	.gso_send_check = ipv6_gso_send_check,
918 	.gso_segment = ipv6_gso_segment,
919 	.gro_receive = ipv6_gro_receive,
920 	.gro_complete = ipv6_gro_complete,
921 };
922 
923 static int __init ipv6_packet_init(void)
924 {
925 	dev_add_pack(&ipv6_packet_type);
926 	return 0;
927 }
928 
929 static void ipv6_packet_cleanup(void)
930 {
931 	dev_remove_pack(&ipv6_packet_type);
932 }
933 
934 static int __net_init ipv6_init_mibs(struct net *net)
935 {
936 	if (snmp_mib_init((void **)net->mib.udp_stats_in6,
937 			  sizeof (struct udp_mib)) < 0)
938 		return -ENOMEM;
939 	if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
940 			  sizeof (struct udp_mib)) < 0)
941 		goto err_udplite_mib;
942 	if (snmp_mib_init((void **)net->mib.ipv6_statistics,
943 			  sizeof(struct ipstats_mib)) < 0)
944 		goto err_ip_mib;
945 	if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
946 			  sizeof(struct icmpv6_mib)) < 0)
947 		goto err_icmp_mib;
948 	if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
949 			  sizeof(struct icmpv6msg_mib)) < 0)
950 		goto err_icmpmsg_mib;
951 	return 0;
952 
953 err_icmpmsg_mib:
954 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
955 err_icmp_mib:
956 	snmp_mib_free((void **)net->mib.ipv6_statistics);
957 err_ip_mib:
958 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
959 err_udplite_mib:
960 	snmp_mib_free((void **)net->mib.udp_stats_in6);
961 	return -ENOMEM;
962 }
963 
964 static void __net_exit ipv6_cleanup_mibs(struct net *net)
965 {
966 	snmp_mib_free((void **)net->mib.udp_stats_in6);
967 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
968 	snmp_mib_free((void **)net->mib.ipv6_statistics);
969 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
970 	snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
971 }
972 
973 static int __net_init inet6_net_init(struct net *net)
974 {
975 	int err = 0;
976 
977 	net->ipv6.sysctl.bindv6only = 0;
978 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
979 
980 	err = ipv6_init_mibs(net);
981 	if (err)
982 		return err;
983 #ifdef CONFIG_PROC_FS
984 	err = udp6_proc_init(net);
985 	if (err)
986 		goto out;
987 	err = tcp6_proc_init(net);
988 	if (err)
989 		goto proc_tcp6_fail;
990 	err = ac6_proc_init(net);
991 	if (err)
992 		goto proc_ac6_fail;
993 #endif
994 	return err;
995 
996 #ifdef CONFIG_PROC_FS
997 proc_ac6_fail:
998 	tcp6_proc_exit(net);
999 proc_tcp6_fail:
1000 	udp6_proc_exit(net);
1001 out:
1002 	ipv6_cleanup_mibs(net);
1003 	return err;
1004 #endif
1005 }
1006 
1007 static void inet6_net_exit(struct net *net)
1008 {
1009 #ifdef CONFIG_PROC_FS
1010 	udp6_proc_exit(net);
1011 	tcp6_proc_exit(net);
1012 	ac6_proc_exit(net);
1013 #endif
1014 	ipv6_cleanup_mibs(net);
1015 }
1016 
1017 static struct pernet_operations inet6_net_ops = {
1018 	.init = inet6_net_init,
1019 	.exit = inet6_net_exit,
1020 };
1021 
1022 static int __init inet6_init(void)
1023 {
1024 	struct sk_buff *dummy_skb;
1025 	struct list_head *r;
1026 	int err = 0;
1027 
1028 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1029 
1030 	/* Register the socket-side information for inet6_create.  */
1031 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1032 		INIT_LIST_HEAD(r);
1033 
1034 	if (disable_ipv6) {
1035 		printk(KERN_INFO
1036 		       "IPv6: Loaded, but administratively disabled, "
1037 		       "reboot required to enable\n");
1038 		goto out;
1039 	}
1040 
1041 	err = proto_register(&tcpv6_prot, 1);
1042 	if (err)
1043 		goto out;
1044 
1045 	err = proto_register(&udpv6_prot, 1);
1046 	if (err)
1047 		goto out_unregister_tcp_proto;
1048 
1049 	err = proto_register(&udplitev6_prot, 1);
1050 	if (err)
1051 		goto out_unregister_udp_proto;
1052 
1053 	err = proto_register(&rawv6_prot, 1);
1054 	if (err)
1055 		goto out_unregister_udplite_proto;
1056 
1057 
1058 	/* We MUST register RAW sockets before we create the ICMP6,
1059 	 * IGMP6, or NDISC control sockets.
1060 	 */
1061 	err = rawv6_init();
1062 	if (err)
1063 		goto out_unregister_raw_proto;
1064 
1065 	/* Register the family here so that the init calls below will
1066 	 * be able to create sockets. (?? is this dangerous ??)
1067 	 */
1068 	err = sock_register(&inet6_family_ops);
1069 	if (err)
1070 		goto out_sock_register_fail;
1071 
1072 #ifdef CONFIG_SYSCTL
1073 	err = ipv6_static_sysctl_register();
1074 	if (err)
1075 		goto static_sysctl_fail;
1076 #endif
1077 	/*
1078 	 *	ipngwg API draft makes clear that the correct semantics
1079 	 *	for TCP and UDP is to consider one TCP and UDP instance
1080 	 *	in a host availiable by both INET and INET6 APIs and
1081 	 *	able to communicate via both network protocols.
1082 	 */
1083 
1084 	err = register_pernet_subsys(&inet6_net_ops);
1085 	if (err)
1086 		goto register_pernet_fail;
1087 	err = icmpv6_init();
1088 	if (err)
1089 		goto icmp_fail;
1090 	err = ip6_mr_init();
1091 	if (err)
1092 		goto ipmr_fail;
1093 	err = ndisc_init();
1094 	if (err)
1095 		goto ndisc_fail;
1096 	err = igmp6_init();
1097 	if (err)
1098 		goto igmp_fail;
1099 	err = ipv6_netfilter_init();
1100 	if (err)
1101 		goto netfilter_fail;
1102 	/* Create /proc/foo6 entries. */
1103 #ifdef CONFIG_PROC_FS
1104 	err = -ENOMEM;
1105 	if (raw6_proc_init())
1106 		goto proc_raw6_fail;
1107 	if (udplite6_proc_init())
1108 		goto proc_udplite6_fail;
1109 	if (ipv6_misc_proc_init())
1110 		goto proc_misc6_fail;
1111 	if (if6_proc_init())
1112 		goto proc_if6_fail;
1113 #endif
1114 	err = ip6_route_init();
1115 	if (err)
1116 		goto ip6_route_fail;
1117 	err = ip6_flowlabel_init();
1118 	if (err)
1119 		goto ip6_flowlabel_fail;
1120 	err = addrconf_init();
1121 	if (err)
1122 		goto addrconf_fail;
1123 
1124 	/* Init v6 extension headers. */
1125 	err = ipv6_exthdrs_init();
1126 	if (err)
1127 		goto ipv6_exthdrs_fail;
1128 
1129 	err = ipv6_frag_init();
1130 	if (err)
1131 		goto ipv6_frag_fail;
1132 
1133 	/* Init v6 transport protocols. */
1134 	err = udpv6_init();
1135 	if (err)
1136 		goto udpv6_fail;
1137 
1138 	err = udplitev6_init();
1139 	if (err)
1140 		goto udplitev6_fail;
1141 
1142 	err = tcpv6_init();
1143 	if (err)
1144 		goto tcpv6_fail;
1145 
1146 	err = ipv6_packet_init();
1147 	if (err)
1148 		goto ipv6_packet_fail;
1149 
1150 #ifdef CONFIG_SYSCTL
1151 	err = ipv6_sysctl_register();
1152 	if (err)
1153 		goto sysctl_fail;
1154 #endif
1155 out:
1156 	return err;
1157 
1158 #ifdef CONFIG_SYSCTL
1159 sysctl_fail:
1160 	ipv6_packet_cleanup();
1161 #endif
1162 ipv6_packet_fail:
1163 	tcpv6_exit();
1164 tcpv6_fail:
1165 	udplitev6_exit();
1166 udplitev6_fail:
1167 	udpv6_exit();
1168 udpv6_fail:
1169 	ipv6_frag_exit();
1170 ipv6_frag_fail:
1171 	ipv6_exthdrs_exit();
1172 ipv6_exthdrs_fail:
1173 	addrconf_cleanup();
1174 addrconf_fail:
1175 	ip6_flowlabel_cleanup();
1176 ip6_flowlabel_fail:
1177 	ip6_route_cleanup();
1178 ip6_route_fail:
1179 #ifdef CONFIG_PROC_FS
1180 	if6_proc_exit();
1181 proc_if6_fail:
1182 	ipv6_misc_proc_exit();
1183 proc_misc6_fail:
1184 	udplite6_proc_exit();
1185 proc_udplite6_fail:
1186 	raw6_proc_exit();
1187 proc_raw6_fail:
1188 #endif
1189 	ipv6_netfilter_fini();
1190 netfilter_fail:
1191 	igmp6_cleanup();
1192 igmp_fail:
1193 	ndisc_cleanup();
1194 ndisc_fail:
1195 	ip6_mr_cleanup();
1196 ipmr_fail:
1197 	icmpv6_cleanup();
1198 icmp_fail:
1199 	unregister_pernet_subsys(&inet6_net_ops);
1200 register_pernet_fail:
1201 #ifdef CONFIG_SYSCTL
1202 	ipv6_static_sysctl_unregister();
1203 static_sysctl_fail:
1204 #endif
1205 	sock_unregister(PF_INET6);
1206 	rtnl_unregister_all(PF_INET6);
1207 out_sock_register_fail:
1208 	rawv6_exit();
1209 out_unregister_raw_proto:
1210 	proto_unregister(&rawv6_prot);
1211 out_unregister_udplite_proto:
1212 	proto_unregister(&udplitev6_prot);
1213 out_unregister_udp_proto:
1214 	proto_unregister(&udpv6_prot);
1215 out_unregister_tcp_proto:
1216 	proto_unregister(&tcpv6_prot);
1217 	goto out;
1218 }
1219 module_init(inet6_init);
1220 
1221 static void __exit inet6_exit(void)
1222 {
1223 	if (disable_ipv6)
1224 		return;
1225 
1226 	/* First of all disallow new sockets creation. */
1227 	sock_unregister(PF_INET6);
1228 	/* Disallow any further netlink messages */
1229 	rtnl_unregister_all(PF_INET6);
1230 
1231 #ifdef CONFIG_SYSCTL
1232 	ipv6_sysctl_unregister();
1233 #endif
1234 	udpv6_exit();
1235 	udplitev6_exit();
1236 	tcpv6_exit();
1237 
1238 	/* Cleanup code parts. */
1239 	ipv6_packet_cleanup();
1240 	ipv6_frag_exit();
1241 	ipv6_exthdrs_exit();
1242 	addrconf_cleanup();
1243 	ip6_flowlabel_cleanup();
1244 	ip6_route_cleanup();
1245 #ifdef CONFIG_PROC_FS
1246 
1247 	/* Cleanup code parts. */
1248 	if6_proc_exit();
1249 	ipv6_misc_proc_exit();
1250 	udplite6_proc_exit();
1251 	raw6_proc_exit();
1252 #endif
1253 	ipv6_netfilter_fini();
1254 	igmp6_cleanup();
1255 	ndisc_cleanup();
1256 	ip6_mr_cleanup();
1257 	icmpv6_cleanup();
1258 	rawv6_exit();
1259 
1260 	unregister_pernet_subsys(&inet6_net_ops);
1261 #ifdef CONFIG_SYSCTL
1262 	ipv6_static_sysctl_unregister();
1263 #endif
1264 	proto_unregister(&rawv6_prot);
1265 	proto_unregister(&udplitev6_prot);
1266 	proto_unregister(&udpv6_prot);
1267 	proto_unregister(&tcpv6_prot);
1268 }
1269 module_exit(inet6_exit);
1270 
1271 MODULE_ALIAS_NETPROTO(PF_INET6);
1272