xref: /linux/net/ipv6/af_inet6.c (revision b233b28eac0cc37d07c2d007ea08c86c778c5af4)
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 __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
76 {
77 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
78 
79 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
80 }
81 
82 static int inet6_create(struct net *net, struct socket *sock, int protocol)
83 {
84 	struct inet_sock *inet;
85 	struct ipv6_pinfo *np;
86 	struct sock *sk;
87 	struct inet_protosw *answer;
88 	struct proto *answer_prot;
89 	unsigned char answer_flags;
90 	char answer_no_check;
91 	int try_loading_module = 0;
92 	int err;
93 
94 	if (sock->type != SOCK_RAW &&
95 	    sock->type != SOCK_DGRAM &&
96 	    !inet_ehash_secret)
97 		build_ehash_secret();
98 
99 	/* Look for the requested type/protocol pair. */
100 lookup_protocol:
101 	err = -ESOCKTNOSUPPORT;
102 	rcu_read_lock();
103 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
104 
105 		err = 0;
106 		/* Check the non-wild match. */
107 		if (protocol == answer->protocol) {
108 			if (protocol != IPPROTO_IP)
109 				break;
110 		} else {
111 			/* Check for the two wild cases. */
112 			if (IPPROTO_IP == protocol) {
113 				protocol = answer->protocol;
114 				break;
115 			}
116 			if (IPPROTO_IP == answer->protocol)
117 				break;
118 		}
119 		err = -EPROTONOSUPPORT;
120 	}
121 
122 	if (err) {
123 		if (try_loading_module < 2) {
124 			rcu_read_unlock();
125 			/*
126 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
127 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
128 			 */
129 			if (++try_loading_module == 1)
130 				request_module("net-pf-%d-proto-%d-type-%d",
131 						PF_INET6, protocol, sock->type);
132 			/*
133 			 * Fall back to generic, e.g. net-pf-10-proto-132
134 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
135 			 */
136 			else
137 				request_module("net-pf-%d-proto-%d",
138 						PF_INET6, protocol);
139 			goto lookup_protocol;
140 		} else
141 			goto out_rcu_unlock;
142 	}
143 
144 	err = -EPERM;
145 	if (answer->capability > 0 && !capable(answer->capability))
146 		goto out_rcu_unlock;
147 
148 	sock->ops = answer->ops;
149 	answer_prot = answer->prot;
150 	answer_no_check = answer->no_check;
151 	answer_flags = answer->flags;
152 	rcu_read_unlock();
153 
154 	WARN_ON(answer_prot->slab == NULL);
155 
156 	err = -ENOBUFS;
157 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
158 	if (sk == NULL)
159 		goto out;
160 
161 	sock_init_data(sock, sk);
162 
163 	err = 0;
164 	sk->sk_no_check = answer_no_check;
165 	if (INET_PROTOSW_REUSE & answer_flags)
166 		sk->sk_reuse = 1;
167 
168 	inet = inet_sk(sk);
169 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
170 
171 	if (SOCK_RAW == sock->type) {
172 		inet->num = protocol;
173 		if (IPPROTO_RAW == protocol)
174 			inet->hdrincl = 1;
175 	}
176 
177 	sk->sk_destruct		= inet_sock_destruct;
178 	sk->sk_family		= PF_INET6;
179 	sk->sk_protocol		= protocol;
180 
181 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
182 
183 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
184 	np->hop_limit	= -1;
185 	np->mcast_hops	= -1;
186 	np->mc_loop	= 1;
187 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
188 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
189 
190 	/* Init the ipv4 part of the socket since we can have sockets
191 	 * using v6 API for ipv4.
192 	 */
193 	inet->uc_ttl	= -1;
194 
195 	inet->mc_loop	= 1;
196 	inet->mc_ttl	= 1;
197 	inet->mc_index	= 0;
198 	inet->mc_list	= NULL;
199 
200 	if (ipv4_config.no_pmtu_disc)
201 		inet->pmtudisc = IP_PMTUDISC_DONT;
202 	else
203 		inet->pmtudisc = IP_PMTUDISC_WANT;
204 	/*
205 	 * Increment only the relevant sk_prot->socks debug field, this changes
206 	 * the previous behaviour of incrementing both the equivalent to
207 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
208 	 *
209 	 * This allows better debug granularity as we'll know exactly how many
210 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
211 	 * transport protocol socks. -acme
212 	 */
213 	sk_refcnt_debug_inc(sk);
214 
215 	if (inet->num) {
216 		/* It assumes that any protocol which allows
217 		 * the user to assign a number at socket
218 		 * creation time automatically shares.
219 		 */
220 		inet->sport = htons(inet->num);
221 		sk->sk_prot->hash(sk);
222 	}
223 	if (sk->sk_prot->init) {
224 		err = sk->sk_prot->init(sk);
225 		if (err) {
226 			sk_common_release(sk);
227 			goto out;
228 		}
229 	}
230 out:
231 	return err;
232 out_rcu_unlock:
233 	rcu_read_unlock();
234 	goto out;
235 }
236 
237 
238 /* bind for INET6 API */
239 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
240 {
241 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
242 	struct sock *sk = sock->sk;
243 	struct inet_sock *inet = inet_sk(sk);
244 	struct ipv6_pinfo *np = inet6_sk(sk);
245 	struct net *net = sock_net(sk);
246 	__be32 v4addr = 0;
247 	unsigned short snum;
248 	int addr_type = 0;
249 	int err = 0;
250 
251 	/* If the socket has its own bind function then use it. */
252 	if (sk->sk_prot->bind)
253 		return sk->sk_prot->bind(sk, uaddr, addr_len);
254 
255 	if (addr_len < SIN6_LEN_RFC2133)
256 		return -EINVAL;
257 	addr_type = ipv6_addr_type(&addr->sin6_addr);
258 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
259 		return -EINVAL;
260 
261 	snum = ntohs(addr->sin6_port);
262 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
263 		return -EACCES;
264 
265 	lock_sock(sk);
266 
267 	/* Check these errors (active socket, double bind). */
268 	if (sk->sk_state != TCP_CLOSE || inet->num) {
269 		err = -EINVAL;
270 		goto out;
271 	}
272 
273 	/* Check if the address belongs to the host. */
274 	if (addr_type == IPV6_ADDR_MAPPED) {
275 		v4addr = addr->sin6_addr.s6_addr32[3];
276 		if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
277 			err = -EADDRNOTAVAIL;
278 			goto out;
279 		}
280 	} else {
281 		if (addr_type != IPV6_ADDR_ANY) {
282 			struct net_device *dev = NULL;
283 
284 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
285 				if (addr_len >= sizeof(struct sockaddr_in6) &&
286 				    addr->sin6_scope_id) {
287 					/* Override any existing binding, if another one
288 					 * is supplied by user.
289 					 */
290 					sk->sk_bound_dev_if = addr->sin6_scope_id;
291 				}
292 
293 				/* Binding to link-local address requires an interface */
294 				if (!sk->sk_bound_dev_if) {
295 					err = -EINVAL;
296 					goto out;
297 				}
298 				dev = dev_get_by_index(net, sk->sk_bound_dev_if);
299 				if (!dev) {
300 					err = -ENODEV;
301 					goto out;
302 				}
303 			}
304 
305 			/* ipv4 addr of the socket is invalid.  Only the
306 			 * unspecified and mapped address have a v4 equivalent.
307 			 */
308 			v4addr = LOOPBACK4_IPV6;
309 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
310 				if (!ipv6_chk_addr(net, &addr->sin6_addr,
311 						   dev, 0)) {
312 					if (dev)
313 						dev_put(dev);
314 					err = -EADDRNOTAVAIL;
315 					goto out;
316 				}
317 			}
318 			if (dev)
319 				dev_put(dev);
320 		}
321 	}
322 
323 	inet->rcv_saddr = v4addr;
324 	inet->saddr = v4addr;
325 
326 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
327 
328 	if (!(addr_type & IPV6_ADDR_MULTICAST))
329 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
330 
331 	/* Make sure we are allowed to bind here. */
332 	if (sk->sk_prot->get_port(sk, snum)) {
333 		inet_reset_saddr(sk);
334 		err = -EADDRINUSE;
335 		goto out;
336 	}
337 
338 	if (addr_type != IPV6_ADDR_ANY)
339 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
340 	if (snum)
341 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
342 	inet->sport = htons(inet->num);
343 	inet->dport = 0;
344 	inet->daddr = 0;
345 out:
346 	release_sock(sk);
347 	return err;
348 }
349 
350 EXPORT_SYMBOL(inet6_bind);
351 
352 int inet6_release(struct socket *sock)
353 {
354 	struct sock *sk = sock->sk;
355 
356 	if (sk == NULL)
357 		return -EINVAL;
358 
359 	/* Free mc lists */
360 	ipv6_sock_mc_close(sk);
361 
362 	/* Free ac lists */
363 	ipv6_sock_ac_close(sk);
364 
365 	return inet_release(sock);
366 }
367 
368 EXPORT_SYMBOL(inet6_release);
369 
370 void inet6_destroy_sock(struct sock *sk)
371 {
372 	struct ipv6_pinfo *np = inet6_sk(sk);
373 	struct sk_buff *skb;
374 	struct ipv6_txoptions *opt;
375 
376 	/* Release rx options */
377 
378 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
379 		kfree_skb(skb);
380 
381 	/* Free flowlabels */
382 	fl6_free_socklist(sk);
383 
384 	/* Free tx options */
385 
386 	if ((opt = xchg(&np->opt, NULL)) != NULL)
387 		sock_kfree_s(sk, opt, opt->tot_len);
388 }
389 
390 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
391 
392 /*
393  *	This does both peername and sockname.
394  */
395 
396 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
397 		 int *uaddr_len, int peer)
398 {
399 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
400 	struct sock *sk = sock->sk;
401 	struct inet_sock *inet = inet_sk(sk);
402 	struct ipv6_pinfo *np = inet6_sk(sk);
403 
404 	sin->sin6_family = AF_INET6;
405 	sin->sin6_flowinfo = 0;
406 	sin->sin6_scope_id = 0;
407 	if (peer) {
408 		if (!inet->dport)
409 			return -ENOTCONN;
410 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
411 		    peer == 1)
412 			return -ENOTCONN;
413 		sin->sin6_port = inet->dport;
414 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
415 		if (np->sndflow)
416 			sin->sin6_flowinfo = np->flow_label;
417 	} else {
418 		if (ipv6_addr_any(&np->rcv_saddr))
419 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
420 		else
421 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
422 
423 		sin->sin6_port = inet->sport;
424 	}
425 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
426 		sin->sin6_scope_id = sk->sk_bound_dev_if;
427 	*uaddr_len = sizeof(*sin);
428 	return(0);
429 }
430 
431 EXPORT_SYMBOL(inet6_getname);
432 
433 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
434 {
435 	struct sock *sk = sock->sk;
436 	struct net *net = sock_net(sk);
437 
438 	switch(cmd)
439 	{
440 	case SIOCGSTAMP:
441 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
442 
443 	case SIOCGSTAMPNS:
444 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
445 
446 	case SIOCADDRT:
447 	case SIOCDELRT:
448 
449 		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
450 
451 	case SIOCSIFADDR:
452 		return addrconf_add_ifaddr(net, (void __user *) arg);
453 	case SIOCDIFADDR:
454 		return addrconf_del_ifaddr(net, (void __user *) arg);
455 	case SIOCSIFDSTADDR:
456 		return addrconf_set_dstaddr(net, (void __user *) arg);
457 	default:
458 		if (!sk->sk_prot->ioctl)
459 			return -ENOIOCTLCMD;
460 		return sk->sk_prot->ioctl(sk, cmd, arg);
461 	}
462 	/*NOTREACHED*/
463 	return(0);
464 }
465 
466 EXPORT_SYMBOL(inet6_ioctl);
467 
468 const struct proto_ops inet6_stream_ops = {
469 	.family		   = PF_INET6,
470 	.owner		   = THIS_MODULE,
471 	.release	   = inet6_release,
472 	.bind		   = inet6_bind,
473 	.connect	   = inet_stream_connect,	/* ok		*/
474 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
475 	.accept		   = inet_accept,		/* ok		*/
476 	.getname	   = inet6_getname,
477 	.poll		   = tcp_poll,			/* ok		*/
478 	.ioctl		   = inet6_ioctl,		/* must change  */
479 	.listen		   = inet_listen,		/* ok		*/
480 	.shutdown	   = inet_shutdown,		/* ok		*/
481 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
482 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
483 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
484 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
485 	.mmap		   = sock_no_mmap,
486 	.sendpage	   = tcp_sendpage,
487 	.splice_read	   = tcp_splice_read,
488 #ifdef CONFIG_COMPAT
489 	.compat_setsockopt = compat_sock_common_setsockopt,
490 	.compat_getsockopt = compat_sock_common_getsockopt,
491 #endif
492 };
493 
494 const struct proto_ops inet6_dgram_ops = {
495 	.family		   = PF_INET6,
496 	.owner		   = THIS_MODULE,
497 	.release	   = inet6_release,
498 	.bind		   = inet6_bind,
499 	.connect	   = inet_dgram_connect,	/* ok		*/
500 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
501 	.accept		   = sock_no_accept,		/* a do nothing	*/
502 	.getname	   = inet6_getname,
503 	.poll		   = udp_poll,			/* ok		*/
504 	.ioctl		   = inet6_ioctl,		/* must change  */
505 	.listen		   = sock_no_listen,		/* ok		*/
506 	.shutdown	   = inet_shutdown,		/* ok		*/
507 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
508 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
509 	.sendmsg	   = inet_sendmsg,		/* ok		*/
510 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
511 	.mmap		   = sock_no_mmap,
512 	.sendpage	   = sock_no_sendpage,
513 #ifdef CONFIG_COMPAT
514 	.compat_setsockopt = compat_sock_common_setsockopt,
515 	.compat_getsockopt = compat_sock_common_getsockopt,
516 #endif
517 };
518 
519 static struct net_proto_family inet6_family_ops = {
520 	.family = PF_INET6,
521 	.create = inet6_create,
522 	.owner	= THIS_MODULE,
523 };
524 
525 int inet6_register_protosw(struct inet_protosw *p)
526 {
527 	struct list_head *lh;
528 	struct inet_protosw *answer;
529 	struct list_head *last_perm;
530 	int protocol = p->protocol;
531 	int ret;
532 
533 	spin_lock_bh(&inetsw6_lock);
534 
535 	ret = -EINVAL;
536 	if (p->type >= SOCK_MAX)
537 		goto out_illegal;
538 
539 	/* If we are trying to override a permanent protocol, bail. */
540 	answer = NULL;
541 	ret = -EPERM;
542 	last_perm = &inetsw6[p->type];
543 	list_for_each(lh, &inetsw6[p->type]) {
544 		answer = list_entry(lh, struct inet_protosw, list);
545 
546 		/* Check only the non-wild match. */
547 		if (INET_PROTOSW_PERMANENT & answer->flags) {
548 			if (protocol == answer->protocol)
549 				break;
550 			last_perm = lh;
551 		}
552 
553 		answer = NULL;
554 	}
555 	if (answer)
556 		goto out_permanent;
557 
558 	/* Add the new entry after the last permanent entry if any, so that
559 	 * the new entry does not override a permanent entry when matched with
560 	 * a wild-card protocol. But it is allowed to override any existing
561 	 * non-permanent entry.  This means that when we remove this entry, the
562 	 * system automatically returns to the old behavior.
563 	 */
564 	list_add_rcu(&p->list, last_perm);
565 	ret = 0;
566 out:
567 	spin_unlock_bh(&inetsw6_lock);
568 	return ret;
569 
570 out_permanent:
571 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
572 	       protocol);
573 	goto out;
574 
575 out_illegal:
576 	printk(KERN_ERR
577 	       "Ignoring attempt to register invalid socket type %d.\n",
578 	       p->type);
579 	goto out;
580 }
581 
582 EXPORT_SYMBOL(inet6_register_protosw);
583 
584 void
585 inet6_unregister_protosw(struct inet_protosw *p)
586 {
587 	if (INET_PROTOSW_PERMANENT & p->flags) {
588 		printk(KERN_ERR
589 		       "Attempt to unregister permanent protocol %d.\n",
590 		       p->protocol);
591 	} else {
592 		spin_lock_bh(&inetsw6_lock);
593 		list_del_rcu(&p->list);
594 		spin_unlock_bh(&inetsw6_lock);
595 
596 		synchronize_net();
597 	}
598 }
599 
600 EXPORT_SYMBOL(inet6_unregister_protosw);
601 
602 int inet6_sk_rebuild_header(struct sock *sk)
603 {
604 	int err;
605 	struct dst_entry *dst;
606 	struct ipv6_pinfo *np = inet6_sk(sk);
607 
608 	dst = __sk_dst_check(sk, np->dst_cookie);
609 
610 	if (dst == NULL) {
611 		struct inet_sock *inet = inet_sk(sk);
612 		struct in6_addr *final_p = NULL, final;
613 		struct flowi fl;
614 
615 		memset(&fl, 0, sizeof(fl));
616 		fl.proto = sk->sk_protocol;
617 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
618 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
619 		fl.fl6_flowlabel = np->flow_label;
620 		fl.oif = sk->sk_bound_dev_if;
621 		fl.fl_ip_dport = inet->dport;
622 		fl.fl_ip_sport = inet->sport;
623 		security_sk_classify_flow(sk, &fl);
624 
625 		if (np->opt && np->opt->srcrt) {
626 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
627 			ipv6_addr_copy(&final, &fl.fl6_dst);
628 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
629 			final_p = &final;
630 		}
631 
632 		err = ip6_dst_lookup(sk, &dst, &fl);
633 		if (err) {
634 			sk->sk_route_caps = 0;
635 			return err;
636 		}
637 		if (final_p)
638 			ipv6_addr_copy(&fl.fl6_dst, final_p);
639 
640 		if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
641 			sk->sk_err_soft = -err;
642 			return err;
643 		}
644 
645 		__ip6_dst_store(sk, dst, NULL, NULL);
646 	}
647 
648 	return 0;
649 }
650 
651 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
652 
653 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
654 {
655 	struct ipv6_pinfo *np = inet6_sk(sk);
656 	struct inet6_skb_parm *opt = IP6CB(skb);
657 
658 	if (np->rxopt.all) {
659 		if ((opt->hop && (np->rxopt.bits.hopopts ||
660 				  np->rxopt.bits.ohopopts)) ||
661 		    ((IPV6_FLOWINFO_MASK &
662 		      *(__be32 *)skb_network_header(skb)) &&
663 		     np->rxopt.bits.rxflow) ||
664 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
665 		     np->rxopt.bits.osrcrt)) ||
666 		    ((opt->dst1 || opt->dst0) &&
667 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
668 			return 1;
669 	}
670 	return 0;
671 }
672 
673 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
674 
675 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
676 {
677 	struct inet6_protocol *ops = NULL;
678 
679 	for (;;) {
680 		struct ipv6_opt_hdr *opth;
681 		int len;
682 
683 		if (proto != NEXTHDR_HOP) {
684 			ops = rcu_dereference(inet6_protos[proto]);
685 
686 			if (unlikely(!ops))
687 				break;
688 
689 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
690 				break;
691 		}
692 
693 		if (unlikely(!pskb_may_pull(skb, 8)))
694 			break;
695 
696 		opth = (void *)skb->data;
697 		len = ipv6_optlen(opth);
698 
699 		if (unlikely(!pskb_may_pull(skb, len)))
700 			break;
701 
702 		proto = opth->nexthdr;
703 		__skb_pull(skb, len);
704 	}
705 
706 	return proto;
707 }
708 
709 static int ipv6_gso_send_check(struct sk_buff *skb)
710 {
711 	struct ipv6hdr *ipv6h;
712 	struct inet6_protocol *ops;
713 	int err = -EINVAL;
714 
715 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
716 		goto out;
717 
718 	ipv6h = ipv6_hdr(skb);
719 	__skb_pull(skb, sizeof(*ipv6h));
720 	err = -EPROTONOSUPPORT;
721 
722 	rcu_read_lock();
723 	ops = rcu_dereference(inet6_protos[
724 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
725 
726 	if (likely(ops && ops->gso_send_check)) {
727 		skb_reset_transport_header(skb);
728 		err = ops->gso_send_check(skb);
729 	}
730 	rcu_read_unlock();
731 
732 out:
733 	return err;
734 }
735 
736 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
737 {
738 	struct sk_buff *segs = ERR_PTR(-EINVAL);
739 	struct ipv6hdr *ipv6h;
740 	struct inet6_protocol *ops;
741 
742 	if (!(features & NETIF_F_V6_CSUM))
743 		features &= ~NETIF_F_SG;
744 
745 	if (unlikely(skb_shinfo(skb)->gso_type &
746 		     ~(SKB_GSO_UDP |
747 		       SKB_GSO_DODGY |
748 		       SKB_GSO_TCP_ECN |
749 		       SKB_GSO_TCPV6 |
750 		       0)))
751 		goto out;
752 
753 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
754 		goto out;
755 
756 	ipv6h = ipv6_hdr(skb);
757 	__skb_pull(skb, sizeof(*ipv6h));
758 	segs = ERR_PTR(-EPROTONOSUPPORT);
759 
760 	rcu_read_lock();
761 	ops = rcu_dereference(inet6_protos[
762 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
763 
764 	if (likely(ops && ops->gso_segment)) {
765 		skb_reset_transport_header(skb);
766 		segs = ops->gso_segment(skb, features);
767 	}
768 	rcu_read_unlock();
769 
770 	if (unlikely(IS_ERR(segs)))
771 		goto out;
772 
773 	for (skb = segs; skb; skb = skb->next) {
774 		ipv6h = ipv6_hdr(skb);
775 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
776 					   sizeof(*ipv6h));
777 	}
778 
779 out:
780 	return segs;
781 }
782 
783 struct ipv6_gro_cb {
784 	struct napi_gro_cb napi;
785 	int proto;
786 };
787 
788 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
789 
790 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
791 					 struct sk_buff *skb)
792 {
793 	struct inet6_protocol *ops;
794 	struct sk_buff **pp = NULL;
795 	struct sk_buff *p;
796 	struct ipv6hdr *iph;
797 	unsigned int nlen;
798 	int flush = 1;
799 	int proto;
800 	__wsum csum;
801 
802 	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
803 		goto out;
804 
805 	iph = ipv6_hdr(skb);
806 	__skb_pull(skb, sizeof(*iph));
807 
808 	flush += ntohs(iph->payload_len) != skb->len;
809 
810 	rcu_read_lock();
811 	proto = ipv6_gso_pull_exthdrs(skb, iph->nexthdr);
812 	iph = ipv6_hdr(skb);
813 	IPV6_GRO_CB(skb)->proto = proto;
814 	ops = rcu_dereference(inet6_protos[proto]);
815 	if (!ops || !ops->gro_receive)
816 		goto out_unlock;
817 
818 	flush--;
819 	skb_reset_transport_header(skb);
820 	nlen = skb_network_header_len(skb);
821 
822 	for (p = *head; p; p = p->next) {
823 		struct ipv6hdr *iph2;
824 
825 		if (!NAPI_GRO_CB(p)->same_flow)
826 			continue;
827 
828 		iph2 = ipv6_hdr(p);
829 
830 		/* All fields must match except length. */
831 		if (nlen != skb_network_header_len(p) ||
832 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
833 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
834 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
835 			NAPI_GRO_CB(p)->same_flow = 0;
836 			continue;
837 		}
838 
839 		NAPI_GRO_CB(p)->flush |= flush;
840 	}
841 
842 	NAPI_GRO_CB(skb)->flush |= flush;
843 
844 	csum = skb->csum;
845 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
846 
847 	pp = ops->gro_receive(head, skb);
848 
849 	skb->csum = csum;
850 
851 out_unlock:
852 	rcu_read_unlock();
853 
854 out:
855 	NAPI_GRO_CB(skb)->flush |= flush;
856 
857 	return pp;
858 }
859 
860 static int ipv6_gro_complete(struct sk_buff *skb)
861 {
862 	struct inet6_protocol *ops;
863 	struct ipv6hdr *iph = ipv6_hdr(skb);
864 	int err = -ENOSYS;
865 
866 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
867 				 sizeof(*iph));
868 
869 	rcu_read_lock();
870 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
871 	if (WARN_ON(!ops || !ops->gro_complete))
872 		goto out_unlock;
873 
874 	err = ops->gro_complete(skb);
875 
876 out_unlock:
877 	rcu_read_unlock();
878 
879 	return err;
880 }
881 
882 static struct packet_type ipv6_packet_type = {
883 	.type = __constant_htons(ETH_P_IPV6),
884 	.func = ipv6_rcv,
885 	.gso_send_check = ipv6_gso_send_check,
886 	.gso_segment = ipv6_gso_segment,
887 	.gro_receive = ipv6_gro_receive,
888 	.gro_complete = ipv6_gro_complete,
889 };
890 
891 static int __init ipv6_packet_init(void)
892 {
893 	dev_add_pack(&ipv6_packet_type);
894 	return 0;
895 }
896 
897 static void ipv6_packet_cleanup(void)
898 {
899 	dev_remove_pack(&ipv6_packet_type);
900 }
901 
902 static int __net_init ipv6_init_mibs(struct net *net)
903 {
904 	if (snmp_mib_init((void **)net->mib.udp_stats_in6,
905 			  sizeof (struct udp_mib)) < 0)
906 		return -ENOMEM;
907 	if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
908 			  sizeof (struct udp_mib)) < 0)
909 		goto err_udplite_mib;
910 	if (snmp_mib_init((void **)net->mib.ipv6_statistics,
911 			  sizeof(struct ipstats_mib)) < 0)
912 		goto err_ip_mib;
913 	if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
914 			  sizeof(struct icmpv6_mib)) < 0)
915 		goto err_icmp_mib;
916 	if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
917 			  sizeof(struct icmpv6msg_mib)) < 0)
918 		goto err_icmpmsg_mib;
919 	return 0;
920 
921 err_icmpmsg_mib:
922 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
923 err_icmp_mib:
924 	snmp_mib_free((void **)net->mib.ipv6_statistics);
925 err_ip_mib:
926 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
927 err_udplite_mib:
928 	snmp_mib_free((void **)net->mib.udp_stats_in6);
929 	return -ENOMEM;
930 }
931 
932 static void __net_exit ipv6_cleanup_mibs(struct net *net)
933 {
934 	snmp_mib_free((void **)net->mib.udp_stats_in6);
935 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
936 	snmp_mib_free((void **)net->mib.ipv6_statistics);
937 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
938 	snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
939 }
940 
941 static int __net_init inet6_net_init(struct net *net)
942 {
943 	int err = 0;
944 
945 	net->ipv6.sysctl.bindv6only = 0;
946 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
947 
948 	err = ipv6_init_mibs(net);
949 	if (err)
950 		return err;
951 #ifdef CONFIG_PROC_FS
952 	err = udp6_proc_init(net);
953 	if (err)
954 		goto out;
955 	err = tcp6_proc_init(net);
956 	if (err)
957 		goto proc_tcp6_fail;
958 	err = ac6_proc_init(net);
959 	if (err)
960 		goto proc_ac6_fail;
961 #endif
962 	return err;
963 
964 #ifdef CONFIG_PROC_FS
965 proc_ac6_fail:
966 	tcp6_proc_exit(net);
967 proc_tcp6_fail:
968 	udp6_proc_exit(net);
969 out:
970 	ipv6_cleanup_mibs(net);
971 	return err;
972 #endif
973 }
974 
975 static void inet6_net_exit(struct net *net)
976 {
977 #ifdef CONFIG_PROC_FS
978 	udp6_proc_exit(net);
979 	tcp6_proc_exit(net);
980 	ac6_proc_exit(net);
981 #endif
982 	ipv6_cleanup_mibs(net);
983 }
984 
985 static struct pernet_operations inet6_net_ops = {
986 	.init = inet6_net_init,
987 	.exit = inet6_net_exit,
988 };
989 
990 static int __init inet6_init(void)
991 {
992 	struct sk_buff *dummy_skb;
993 	struct list_head *r;
994 	int err;
995 
996 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
997 
998 	err = proto_register(&tcpv6_prot, 1);
999 	if (err)
1000 		goto out;
1001 
1002 	err = proto_register(&udpv6_prot, 1);
1003 	if (err)
1004 		goto out_unregister_tcp_proto;
1005 
1006 	err = proto_register(&udplitev6_prot, 1);
1007 	if (err)
1008 		goto out_unregister_udp_proto;
1009 
1010 	err = proto_register(&rawv6_prot, 1);
1011 	if (err)
1012 		goto out_unregister_udplite_proto;
1013 
1014 
1015 	/* Register the socket-side information for inet6_create.  */
1016 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1017 		INIT_LIST_HEAD(r);
1018 
1019 	/* We MUST register RAW sockets before we create the ICMP6,
1020 	 * IGMP6, or NDISC control sockets.
1021 	 */
1022 	err = rawv6_init();
1023 	if (err)
1024 		goto out_unregister_raw_proto;
1025 
1026 	/* Register the family here so that the init calls below will
1027 	 * be able to create sockets. (?? is this dangerous ??)
1028 	 */
1029 	err = sock_register(&inet6_family_ops);
1030 	if (err)
1031 		goto out_sock_register_fail;
1032 
1033 #ifdef CONFIG_SYSCTL
1034 	err = ipv6_static_sysctl_register();
1035 	if (err)
1036 		goto static_sysctl_fail;
1037 #endif
1038 	/*
1039 	 *	ipngwg API draft makes clear that the correct semantics
1040 	 *	for TCP and UDP is to consider one TCP and UDP instance
1041 	 *	in a host availiable by both INET and INET6 APIs and
1042 	 *	able to communicate via both network protocols.
1043 	 */
1044 
1045 	err = register_pernet_subsys(&inet6_net_ops);
1046 	if (err)
1047 		goto register_pernet_fail;
1048 	err = icmpv6_init();
1049 	if (err)
1050 		goto icmp_fail;
1051 	err = ip6_mr_init();
1052 	if (err)
1053 		goto ipmr_fail;
1054 	err = ndisc_init();
1055 	if (err)
1056 		goto ndisc_fail;
1057 	err = igmp6_init();
1058 	if (err)
1059 		goto igmp_fail;
1060 	err = ipv6_netfilter_init();
1061 	if (err)
1062 		goto netfilter_fail;
1063 	/* Create /proc/foo6 entries. */
1064 #ifdef CONFIG_PROC_FS
1065 	err = -ENOMEM;
1066 	if (raw6_proc_init())
1067 		goto proc_raw6_fail;
1068 	if (udplite6_proc_init())
1069 		goto proc_udplite6_fail;
1070 	if (ipv6_misc_proc_init())
1071 		goto proc_misc6_fail;
1072 	if (if6_proc_init())
1073 		goto proc_if6_fail;
1074 #endif
1075 	err = ip6_route_init();
1076 	if (err)
1077 		goto ip6_route_fail;
1078 	err = ip6_flowlabel_init();
1079 	if (err)
1080 		goto ip6_flowlabel_fail;
1081 	err = addrconf_init();
1082 	if (err)
1083 		goto addrconf_fail;
1084 
1085 	/* Init v6 extension headers. */
1086 	err = ipv6_exthdrs_init();
1087 	if (err)
1088 		goto ipv6_exthdrs_fail;
1089 
1090 	err = ipv6_frag_init();
1091 	if (err)
1092 		goto ipv6_frag_fail;
1093 
1094 	/* Init v6 transport protocols. */
1095 	err = udpv6_init();
1096 	if (err)
1097 		goto udpv6_fail;
1098 
1099 	err = udplitev6_init();
1100 	if (err)
1101 		goto udplitev6_fail;
1102 
1103 	err = tcpv6_init();
1104 	if (err)
1105 		goto tcpv6_fail;
1106 
1107 	err = ipv6_packet_init();
1108 	if (err)
1109 		goto ipv6_packet_fail;
1110 
1111 #ifdef CONFIG_SYSCTL
1112 	err = ipv6_sysctl_register();
1113 	if (err)
1114 		goto sysctl_fail;
1115 #endif
1116 out:
1117 	return err;
1118 
1119 #ifdef CONFIG_SYSCTL
1120 sysctl_fail:
1121 	ipv6_packet_cleanup();
1122 #endif
1123 ipv6_packet_fail:
1124 	tcpv6_exit();
1125 tcpv6_fail:
1126 	udplitev6_exit();
1127 udplitev6_fail:
1128 	udpv6_exit();
1129 udpv6_fail:
1130 	ipv6_frag_exit();
1131 ipv6_frag_fail:
1132 	ipv6_exthdrs_exit();
1133 ipv6_exthdrs_fail:
1134 	addrconf_cleanup();
1135 addrconf_fail:
1136 	ip6_flowlabel_cleanup();
1137 ip6_flowlabel_fail:
1138 	ip6_route_cleanup();
1139 ip6_route_fail:
1140 #ifdef CONFIG_PROC_FS
1141 	if6_proc_exit();
1142 proc_if6_fail:
1143 	ipv6_misc_proc_exit();
1144 proc_misc6_fail:
1145 	udplite6_proc_exit();
1146 proc_udplite6_fail:
1147 	raw6_proc_exit();
1148 proc_raw6_fail:
1149 #endif
1150 	ipv6_netfilter_fini();
1151 netfilter_fail:
1152 	igmp6_cleanup();
1153 igmp_fail:
1154 	ndisc_cleanup();
1155 ndisc_fail:
1156 	ip6_mr_cleanup();
1157 ipmr_fail:
1158 	icmpv6_cleanup();
1159 icmp_fail:
1160 	unregister_pernet_subsys(&inet6_net_ops);
1161 register_pernet_fail:
1162 #ifdef CONFIG_SYSCTL
1163 	ipv6_static_sysctl_unregister();
1164 static_sysctl_fail:
1165 #endif
1166 	sock_unregister(PF_INET6);
1167 	rtnl_unregister_all(PF_INET6);
1168 out_sock_register_fail:
1169 	rawv6_exit();
1170 out_unregister_raw_proto:
1171 	proto_unregister(&rawv6_prot);
1172 out_unregister_udplite_proto:
1173 	proto_unregister(&udplitev6_prot);
1174 out_unregister_udp_proto:
1175 	proto_unregister(&udpv6_prot);
1176 out_unregister_tcp_proto:
1177 	proto_unregister(&tcpv6_prot);
1178 	goto out;
1179 }
1180 module_init(inet6_init);
1181 
1182 static void __exit inet6_exit(void)
1183 {
1184 	/* First of all disallow new sockets creation. */
1185 	sock_unregister(PF_INET6);
1186 	/* Disallow any further netlink messages */
1187 	rtnl_unregister_all(PF_INET6);
1188 
1189 #ifdef CONFIG_SYSCTL
1190 	ipv6_sysctl_unregister();
1191 #endif
1192 	udpv6_exit();
1193 	udplitev6_exit();
1194 	tcpv6_exit();
1195 
1196 	/* Cleanup code parts. */
1197 	ipv6_packet_cleanup();
1198 	ipv6_frag_exit();
1199 	ipv6_exthdrs_exit();
1200 	addrconf_cleanup();
1201 	ip6_flowlabel_cleanup();
1202 	ip6_route_cleanup();
1203 #ifdef CONFIG_PROC_FS
1204 
1205 	/* Cleanup code parts. */
1206 	if6_proc_exit();
1207 	ipv6_misc_proc_exit();
1208 	udplite6_proc_exit();
1209 	raw6_proc_exit();
1210 #endif
1211 	ipv6_netfilter_fini();
1212 	igmp6_cleanup();
1213 	ndisc_cleanup();
1214 	ip6_mr_cleanup();
1215 	icmpv6_cleanup();
1216 	rawv6_exit();
1217 
1218 	unregister_pernet_subsys(&inet6_net_ops);
1219 #ifdef CONFIG_SYSCTL
1220 	ipv6_static_sysctl_unregister();
1221 #endif
1222 	proto_unregister(&rawv6_prot);
1223 	proto_unregister(&udplitev6_prot);
1224 	proto_unregister(&udpv6_prot);
1225 	proto_unregister(&tcpv6_prot);
1226 }
1227 module_exit(inet6_exit);
1228 
1229 MODULE_ALIAS_NETPROTO(PF_INET6);
1230