xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 9cdb81c7)
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 	inet->rcv_tos	= 0;
218 
219 	if (ipv4_config.no_pmtu_disc)
220 		inet->pmtudisc = IP_PMTUDISC_DONT;
221 	else
222 		inet->pmtudisc = IP_PMTUDISC_WANT;
223 	/*
224 	 * Increment only the relevant sk_prot->socks debug field, this changes
225 	 * the previous behaviour of incrementing both the equivalent to
226 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
227 	 *
228 	 * This allows better debug granularity as we'll know exactly how many
229 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
230 	 * transport protocol socks. -acme
231 	 */
232 	sk_refcnt_debug_inc(sk);
233 
234 	if (inet->inet_num) {
235 		/* It assumes that any protocol which allows
236 		 * the user to assign a number at socket
237 		 * creation time automatically shares.
238 		 */
239 		inet->inet_sport = htons(inet->inet_num);
240 		sk->sk_prot->hash(sk);
241 	}
242 	if (sk->sk_prot->init) {
243 		err = sk->sk_prot->init(sk);
244 		if (err) {
245 			sk_common_release(sk);
246 			goto out;
247 		}
248 	}
249 out:
250 	return err;
251 out_rcu_unlock:
252 	rcu_read_unlock();
253 	goto out;
254 }
255 
256 
257 /* bind for INET6 API */
258 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
259 {
260 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
261 	struct sock *sk = sock->sk;
262 	struct inet_sock *inet = inet_sk(sk);
263 	struct ipv6_pinfo *np = inet6_sk(sk);
264 	struct net *net = sock_net(sk);
265 	__be32 v4addr = 0;
266 	unsigned short snum;
267 	int addr_type = 0;
268 	int err = 0;
269 
270 	/* If the socket has its own bind function then use it. */
271 	if (sk->sk_prot->bind)
272 		return sk->sk_prot->bind(sk, uaddr, addr_len);
273 
274 	if (addr_len < SIN6_LEN_RFC2133)
275 		return -EINVAL;
276 
277 	if (addr->sin6_family != AF_INET6)
278 		return -EAFNOSUPPORT;
279 
280 	addr_type = ipv6_addr_type(&addr->sin6_addr);
281 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
282 		return -EINVAL;
283 
284 	snum = ntohs(addr->sin6_port);
285 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
286 		return -EACCES;
287 
288 	lock_sock(sk);
289 
290 	/* Check these errors (active socket, double bind). */
291 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
292 		err = -EINVAL;
293 		goto out;
294 	}
295 
296 	/* Check if the address belongs to the host. */
297 	if (addr_type == IPV6_ADDR_MAPPED) {
298 		int chk_addr_ret;
299 
300 		/* Binding to v4-mapped address on a v6-only socket
301 		 * makes no sense
302 		 */
303 		if (np->ipv6only) {
304 			err = -EINVAL;
305 			goto out;
306 		}
307 
308 		/* Reproduce AF_INET checks to make the bindings consistent */
309 		v4addr = addr->sin6_addr.s6_addr32[3];
310 		chk_addr_ret = inet_addr_type(net, v4addr);
311 		if (!sysctl_ip_nonlocal_bind &&
312 		    !(inet->freebind || inet->transparent) &&
313 		    v4addr != htonl(INADDR_ANY) &&
314 		    chk_addr_ret != RTN_LOCAL &&
315 		    chk_addr_ret != RTN_MULTICAST &&
316 		    chk_addr_ret != RTN_BROADCAST) {
317 			err = -EADDRNOTAVAIL;
318 			goto out;
319 		}
320 	} else {
321 		if (addr_type != IPV6_ADDR_ANY) {
322 			struct net_device *dev = NULL;
323 
324 			rcu_read_lock();
325 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
326 				if (addr_len >= sizeof(struct sockaddr_in6) &&
327 				    addr->sin6_scope_id) {
328 					/* Override any existing binding, if another one
329 					 * is supplied by user.
330 					 */
331 					sk->sk_bound_dev_if = addr->sin6_scope_id;
332 				}
333 
334 				/* Binding to link-local address requires an interface */
335 				if (!sk->sk_bound_dev_if) {
336 					err = -EINVAL;
337 					goto out_unlock;
338 				}
339 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
340 				if (!dev) {
341 					err = -ENODEV;
342 					goto out_unlock;
343 				}
344 			}
345 
346 			/* ipv4 addr of the socket is invalid.  Only the
347 			 * unspecified and mapped address have a v4 equivalent.
348 			 */
349 			v4addr = LOOPBACK4_IPV6;
350 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
351 				if (!(inet->freebind || inet->transparent) &&
352 				    !ipv6_chk_addr(net, &addr->sin6_addr,
353 						   dev, 0)) {
354 					err = -EADDRNOTAVAIL;
355 					goto out_unlock;
356 				}
357 			}
358 			rcu_read_unlock();
359 		}
360 	}
361 
362 	inet->inet_rcv_saddr = v4addr;
363 	inet->inet_saddr = v4addr;
364 
365 	np->rcv_saddr = addr->sin6_addr;
366 
367 	if (!(addr_type & IPV6_ADDR_MULTICAST))
368 		np->saddr = addr->sin6_addr;
369 
370 	/* Make sure we are allowed to bind here. */
371 	if (sk->sk_prot->get_port(sk, snum)) {
372 		inet_reset_saddr(sk);
373 		err = -EADDRINUSE;
374 		goto out;
375 	}
376 
377 	if (addr_type != IPV6_ADDR_ANY) {
378 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
379 		if (addr_type != IPV6_ADDR_MAPPED)
380 			np->ipv6only = 1;
381 	}
382 	if (snum)
383 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
384 	inet->inet_sport = htons(inet->inet_num);
385 	inet->inet_dport = 0;
386 	inet->inet_daddr = 0;
387 out:
388 	release_sock(sk);
389 	return err;
390 out_unlock:
391 	rcu_read_unlock();
392 	goto out;
393 }
394 
395 EXPORT_SYMBOL(inet6_bind);
396 
397 int inet6_release(struct socket *sock)
398 {
399 	struct sock *sk = sock->sk;
400 
401 	if (sk == NULL)
402 		return -EINVAL;
403 
404 	/* Free mc lists */
405 	ipv6_sock_mc_close(sk);
406 
407 	/* Free ac lists */
408 	ipv6_sock_ac_close(sk);
409 
410 	return inet_release(sock);
411 }
412 
413 EXPORT_SYMBOL(inet6_release);
414 
415 void inet6_destroy_sock(struct sock *sk)
416 {
417 	struct ipv6_pinfo *np = inet6_sk(sk);
418 	struct sk_buff *skb;
419 	struct ipv6_txoptions *opt;
420 
421 	/* Release rx options */
422 
423 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
424 		kfree_skb(skb);
425 
426 	if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
427 		kfree_skb(skb);
428 
429 	/* Free flowlabels */
430 	fl6_free_socklist(sk);
431 
432 	/* Free tx options */
433 
434 	if ((opt = xchg(&np->opt, NULL)) != NULL)
435 		sock_kfree_s(sk, opt, opt->tot_len);
436 }
437 
438 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
439 
440 /*
441  *	This does both peername and sockname.
442  */
443 
444 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
445 		 int *uaddr_len, int peer)
446 {
447 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
448 	struct sock *sk = sock->sk;
449 	struct inet_sock *inet = inet_sk(sk);
450 	struct ipv6_pinfo *np = inet6_sk(sk);
451 
452 	sin->sin6_family = AF_INET6;
453 	sin->sin6_flowinfo = 0;
454 	sin->sin6_scope_id = 0;
455 	if (peer) {
456 		if (!inet->inet_dport)
457 			return -ENOTCONN;
458 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
459 		    peer == 1)
460 			return -ENOTCONN;
461 		sin->sin6_port = inet->inet_dport;
462 		sin->sin6_addr = np->daddr;
463 		if (np->sndflow)
464 			sin->sin6_flowinfo = np->flow_label;
465 	} else {
466 		if (ipv6_addr_any(&np->rcv_saddr))
467 			sin->sin6_addr = np->saddr;
468 		else
469 			sin->sin6_addr = np->rcv_saddr;
470 
471 		sin->sin6_port = inet->inet_sport;
472 	}
473 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
474 		sin->sin6_scope_id = sk->sk_bound_dev_if;
475 	*uaddr_len = sizeof(*sin);
476 	return 0;
477 }
478 
479 EXPORT_SYMBOL(inet6_getname);
480 
481 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
482 {
483 	struct sock *sk = sock->sk;
484 	struct net *net = sock_net(sk);
485 
486 	switch(cmd)
487 	{
488 	case SIOCGSTAMP:
489 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
490 
491 	case SIOCGSTAMPNS:
492 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
493 
494 	case SIOCADDRT:
495 	case SIOCDELRT:
496 
497 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
498 
499 	case SIOCSIFADDR:
500 		return addrconf_add_ifaddr(net, (void __user *) arg);
501 	case SIOCDIFADDR:
502 		return addrconf_del_ifaddr(net, (void __user *) arg);
503 	case SIOCSIFDSTADDR:
504 		return addrconf_set_dstaddr(net, (void __user *) arg);
505 	default:
506 		if (!sk->sk_prot->ioctl)
507 			return -ENOIOCTLCMD;
508 		return sk->sk_prot->ioctl(sk, cmd, arg);
509 	}
510 	/*NOTREACHED*/
511 	return 0;
512 }
513 
514 EXPORT_SYMBOL(inet6_ioctl);
515 
516 const struct proto_ops inet6_stream_ops = {
517 	.family		   = PF_INET6,
518 	.owner		   = THIS_MODULE,
519 	.release	   = inet6_release,
520 	.bind		   = inet6_bind,
521 	.connect	   = inet_stream_connect,	/* ok		*/
522 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
523 	.accept		   = inet_accept,		/* ok		*/
524 	.getname	   = inet6_getname,
525 	.poll		   = tcp_poll,			/* ok		*/
526 	.ioctl		   = inet6_ioctl,		/* must change  */
527 	.listen		   = inet_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	   = inet_recvmsg,		/* ok		*/
533 	.mmap		   = sock_no_mmap,
534 	.sendpage	   = inet_sendpage,
535 	.splice_read	   = tcp_splice_read,
536 #ifdef CONFIG_COMPAT
537 	.compat_setsockopt = compat_sock_common_setsockopt,
538 	.compat_getsockopt = compat_sock_common_getsockopt,
539 #endif
540 };
541 
542 const struct proto_ops inet6_dgram_ops = {
543 	.family		   = PF_INET6,
544 	.owner		   = THIS_MODULE,
545 	.release	   = inet6_release,
546 	.bind		   = inet6_bind,
547 	.connect	   = inet_dgram_connect,	/* ok		*/
548 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
549 	.accept		   = sock_no_accept,		/* a do nothing	*/
550 	.getname	   = inet6_getname,
551 	.poll		   = udp_poll,			/* ok		*/
552 	.ioctl		   = inet6_ioctl,		/* must change  */
553 	.listen		   = sock_no_listen,		/* ok		*/
554 	.shutdown	   = inet_shutdown,		/* ok		*/
555 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
556 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
557 	.sendmsg	   = inet_sendmsg,		/* ok		*/
558 	.recvmsg	   = inet_recvmsg,		/* ok		*/
559 	.mmap		   = sock_no_mmap,
560 	.sendpage	   = sock_no_sendpage,
561 #ifdef CONFIG_COMPAT
562 	.compat_setsockopt = compat_sock_common_setsockopt,
563 	.compat_getsockopt = compat_sock_common_getsockopt,
564 #endif
565 };
566 
567 static const struct net_proto_family inet6_family_ops = {
568 	.family = PF_INET6,
569 	.create = inet6_create,
570 	.owner	= THIS_MODULE,
571 };
572 
573 int inet6_register_protosw(struct inet_protosw *p)
574 {
575 	struct list_head *lh;
576 	struct inet_protosw *answer;
577 	struct list_head *last_perm;
578 	int protocol = p->protocol;
579 	int ret;
580 
581 	spin_lock_bh(&inetsw6_lock);
582 
583 	ret = -EINVAL;
584 	if (p->type >= SOCK_MAX)
585 		goto out_illegal;
586 
587 	/* If we are trying to override a permanent protocol, bail. */
588 	answer = NULL;
589 	ret = -EPERM;
590 	last_perm = &inetsw6[p->type];
591 	list_for_each(lh, &inetsw6[p->type]) {
592 		answer = list_entry(lh, struct inet_protosw, list);
593 
594 		/* Check only the non-wild match. */
595 		if (INET_PROTOSW_PERMANENT & answer->flags) {
596 			if (protocol == answer->protocol)
597 				break;
598 			last_perm = lh;
599 		}
600 
601 		answer = NULL;
602 	}
603 	if (answer)
604 		goto out_permanent;
605 
606 	/* Add the new entry after the last permanent entry if any, so that
607 	 * the new entry does not override a permanent entry when matched with
608 	 * a wild-card protocol. But it is allowed to override any existing
609 	 * non-permanent entry.  This means that when we remove this entry, the
610 	 * system automatically returns to the old behavior.
611 	 */
612 	list_add_rcu(&p->list, last_perm);
613 	ret = 0;
614 out:
615 	spin_unlock_bh(&inetsw6_lock);
616 	return ret;
617 
618 out_permanent:
619 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
620 	       protocol);
621 	goto out;
622 
623 out_illegal:
624 	printk(KERN_ERR
625 	       "Ignoring attempt to register invalid socket type %d.\n",
626 	       p->type);
627 	goto out;
628 }
629 
630 EXPORT_SYMBOL(inet6_register_protosw);
631 
632 void
633 inet6_unregister_protosw(struct inet_protosw *p)
634 {
635 	if (INET_PROTOSW_PERMANENT & p->flags) {
636 		printk(KERN_ERR
637 		       "Attempt to unregister permanent protocol %d.\n",
638 		       p->protocol);
639 	} else {
640 		spin_lock_bh(&inetsw6_lock);
641 		list_del_rcu(&p->list);
642 		spin_unlock_bh(&inetsw6_lock);
643 
644 		synchronize_net();
645 	}
646 }
647 
648 EXPORT_SYMBOL(inet6_unregister_protosw);
649 
650 int inet6_sk_rebuild_header(struct sock *sk)
651 {
652 	struct ipv6_pinfo *np = inet6_sk(sk);
653 	struct dst_entry *dst;
654 
655 	dst = __sk_dst_check(sk, np->dst_cookie);
656 
657 	if (dst == NULL) {
658 		struct inet_sock *inet = inet_sk(sk);
659 		struct in6_addr *final_p, final;
660 		struct flowi6 fl6;
661 
662 		memset(&fl6, 0, sizeof(fl6));
663 		fl6.flowi6_proto = sk->sk_protocol;
664 		fl6.daddr = np->daddr;
665 		fl6.saddr = np->saddr;
666 		fl6.flowlabel = np->flow_label;
667 		fl6.flowi6_oif = sk->sk_bound_dev_if;
668 		fl6.flowi6_mark = sk->sk_mark;
669 		fl6.fl6_dport = inet->inet_dport;
670 		fl6.fl6_sport = inet->inet_sport;
671 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
672 
673 		final_p = fl6_update_dst(&fl6, np->opt, &final);
674 
675 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
676 		if (IS_ERR(dst)) {
677 			sk->sk_route_caps = 0;
678 			sk->sk_err_soft = -PTR_ERR(dst);
679 			return PTR_ERR(dst);
680 		}
681 
682 		__ip6_dst_store(sk, dst, NULL, NULL);
683 	}
684 
685 	return 0;
686 }
687 
688 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
689 
690 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
691 {
692 	struct ipv6_pinfo *np = inet6_sk(sk);
693 	struct inet6_skb_parm *opt = IP6CB(skb);
694 
695 	if (np->rxopt.all) {
696 		if ((opt->hop && (np->rxopt.bits.hopopts ||
697 				  np->rxopt.bits.ohopopts)) ||
698 		    ((IPV6_FLOWINFO_MASK &
699 		      *(__be32 *)skb_network_header(skb)) &&
700 		     np->rxopt.bits.rxflow) ||
701 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
702 		     np->rxopt.bits.osrcrt)) ||
703 		    ((opt->dst1 || opt->dst0) &&
704 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
705 			return 1;
706 	}
707 	return 0;
708 }
709 
710 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
711 
712 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
713 {
714 	const struct inet6_protocol *ops = NULL;
715 
716 	for (;;) {
717 		struct ipv6_opt_hdr *opth;
718 		int len;
719 
720 		if (proto != NEXTHDR_HOP) {
721 			ops = rcu_dereference(inet6_protos[proto]);
722 
723 			if (unlikely(!ops))
724 				break;
725 
726 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
727 				break;
728 		}
729 
730 		if (unlikely(!pskb_may_pull(skb, 8)))
731 			break;
732 
733 		opth = (void *)skb->data;
734 		len = ipv6_optlen(opth);
735 
736 		if (unlikely(!pskb_may_pull(skb, len)))
737 			break;
738 
739 		proto = opth->nexthdr;
740 		__skb_pull(skb, len);
741 	}
742 
743 	return proto;
744 }
745 
746 static int ipv6_gso_send_check(struct sk_buff *skb)
747 {
748 	const struct ipv6hdr *ipv6h;
749 	const struct inet6_protocol *ops;
750 	int err = -EINVAL;
751 
752 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
753 		goto out;
754 
755 	ipv6h = ipv6_hdr(skb);
756 	__skb_pull(skb, sizeof(*ipv6h));
757 	err = -EPROTONOSUPPORT;
758 
759 	rcu_read_lock();
760 	ops = rcu_dereference(inet6_protos[
761 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
762 
763 	if (likely(ops && ops->gso_send_check)) {
764 		skb_reset_transport_header(skb);
765 		err = ops->gso_send_check(skb);
766 	}
767 	rcu_read_unlock();
768 
769 out:
770 	return err;
771 }
772 
773 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
774 	netdev_features_t features)
775 {
776 	struct sk_buff *segs = ERR_PTR(-EINVAL);
777 	struct ipv6hdr *ipv6h;
778 	const struct inet6_protocol *ops;
779 	int proto;
780 	struct frag_hdr *fptr;
781 	unsigned int unfrag_ip6hlen;
782 	u8 *prevhdr;
783 	int offset = 0;
784 
785 	if (!(features & NETIF_F_V6_CSUM))
786 		features &= ~NETIF_F_SG;
787 
788 	if (unlikely(skb_shinfo(skb)->gso_type &
789 		     ~(SKB_GSO_UDP |
790 		       SKB_GSO_DODGY |
791 		       SKB_GSO_TCP_ECN |
792 		       SKB_GSO_TCPV6 |
793 		       0)))
794 		goto out;
795 
796 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
797 		goto out;
798 
799 	ipv6h = ipv6_hdr(skb);
800 	__skb_pull(skb, sizeof(*ipv6h));
801 	segs = ERR_PTR(-EPROTONOSUPPORT);
802 
803 	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
804 	rcu_read_lock();
805 	ops = rcu_dereference(inet6_protos[proto]);
806 	if (likely(ops && ops->gso_segment)) {
807 		skb_reset_transport_header(skb);
808 		segs = ops->gso_segment(skb, features);
809 	}
810 	rcu_read_unlock();
811 
812 	if (IS_ERR(segs))
813 		goto out;
814 
815 	for (skb = segs; skb; skb = skb->next) {
816 		ipv6h = ipv6_hdr(skb);
817 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
818 					   sizeof(*ipv6h));
819 		if (proto == IPPROTO_UDP) {
820 			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
821 			fptr = (struct frag_hdr *)(skb_network_header(skb) +
822 				unfrag_ip6hlen);
823 			fptr->frag_off = htons(offset);
824 			if (skb->next != NULL)
825 				fptr->frag_off |= htons(IP6_MF);
826 			offset += (ntohs(ipv6h->payload_len) -
827 				   sizeof(struct frag_hdr));
828 		}
829 	}
830 
831 out:
832 	return segs;
833 }
834 
835 struct ipv6_gro_cb {
836 	struct napi_gro_cb napi;
837 	int proto;
838 };
839 
840 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
841 
842 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
843 					 struct sk_buff *skb)
844 {
845 	const struct inet6_protocol *ops;
846 	struct sk_buff **pp = NULL;
847 	struct sk_buff *p;
848 	struct ipv6hdr *iph;
849 	unsigned int nlen;
850 	unsigned int hlen;
851 	unsigned int off;
852 	int flush = 1;
853 	int proto;
854 	__wsum csum;
855 
856 	off = skb_gro_offset(skb);
857 	hlen = off + sizeof(*iph);
858 	iph = skb_gro_header_fast(skb, off);
859 	if (skb_gro_header_hard(skb, hlen)) {
860 		iph = skb_gro_header_slow(skb, hlen, off);
861 		if (unlikely(!iph))
862 			goto out;
863 	}
864 
865 	skb_gro_pull(skb, sizeof(*iph));
866 	skb_set_transport_header(skb, skb_gro_offset(skb));
867 
868 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
869 
870 	rcu_read_lock();
871 	proto = iph->nexthdr;
872 	ops = rcu_dereference(inet6_protos[proto]);
873 	if (!ops || !ops->gro_receive) {
874 		__pskb_pull(skb, skb_gro_offset(skb));
875 		proto = ipv6_gso_pull_exthdrs(skb, proto);
876 		skb_gro_pull(skb, -skb_transport_offset(skb));
877 		skb_reset_transport_header(skb);
878 		__skb_push(skb, skb_gro_offset(skb));
879 
880 		ops = rcu_dereference(inet6_protos[proto]);
881 		if (!ops || !ops->gro_receive)
882 			goto out_unlock;
883 
884 		iph = ipv6_hdr(skb);
885 	}
886 
887 	IPV6_GRO_CB(skb)->proto = proto;
888 
889 	flush--;
890 	nlen = skb_network_header_len(skb);
891 
892 	for (p = *head; p; p = p->next) {
893 		struct ipv6hdr *iph2;
894 
895 		if (!NAPI_GRO_CB(p)->same_flow)
896 			continue;
897 
898 		iph2 = ipv6_hdr(p);
899 
900 		/* All fields must match except length. */
901 		if (nlen != skb_network_header_len(p) ||
902 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
903 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
904 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
905 			NAPI_GRO_CB(p)->same_flow = 0;
906 			continue;
907 		}
908 
909 		NAPI_GRO_CB(p)->flush |= flush;
910 	}
911 
912 	NAPI_GRO_CB(skb)->flush |= flush;
913 
914 	csum = skb->csum;
915 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
916 
917 	pp = ops->gro_receive(head, skb);
918 
919 	skb->csum = csum;
920 
921 out_unlock:
922 	rcu_read_unlock();
923 
924 out:
925 	NAPI_GRO_CB(skb)->flush |= flush;
926 
927 	return pp;
928 }
929 
930 static int ipv6_gro_complete(struct sk_buff *skb)
931 {
932 	const struct inet6_protocol *ops;
933 	struct ipv6hdr *iph = ipv6_hdr(skb);
934 	int err = -ENOSYS;
935 
936 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
937 				 sizeof(*iph));
938 
939 	rcu_read_lock();
940 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
941 	if (WARN_ON(!ops || !ops->gro_complete))
942 		goto out_unlock;
943 
944 	err = ops->gro_complete(skb);
945 
946 out_unlock:
947 	rcu_read_unlock();
948 
949 	return err;
950 }
951 
952 static struct packet_type ipv6_packet_type __read_mostly = {
953 	.type = cpu_to_be16(ETH_P_IPV6),
954 	.func = ipv6_rcv,
955 	.gso_send_check = ipv6_gso_send_check,
956 	.gso_segment = ipv6_gso_segment,
957 	.gro_receive = ipv6_gro_receive,
958 	.gro_complete = ipv6_gro_complete,
959 };
960 
961 static int __init ipv6_packet_init(void)
962 {
963 	dev_add_pack(&ipv6_packet_type);
964 	return 0;
965 }
966 
967 static void ipv6_packet_cleanup(void)
968 {
969 	dev_remove_pack(&ipv6_packet_type);
970 }
971 
972 static int __net_init ipv6_init_mibs(struct net *net)
973 {
974 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
975 			  sizeof(struct udp_mib),
976 			  __alignof__(struct udp_mib)) < 0)
977 		return -ENOMEM;
978 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
979 			  sizeof(struct udp_mib),
980 			  __alignof__(struct udp_mib)) < 0)
981 		goto err_udplite_mib;
982 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
983 			  sizeof(struct ipstats_mib),
984 			  __alignof__(struct ipstats_mib)) < 0)
985 		goto err_ip_mib;
986 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
987 			  sizeof(struct icmpv6_mib),
988 			  __alignof__(struct icmpv6_mib)) < 0)
989 		goto err_icmp_mib;
990 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
991 						GFP_KERNEL);
992 	if (!net->mib.icmpv6msg_statistics)
993 		goto err_icmpmsg_mib;
994 	return 0;
995 
996 err_icmpmsg_mib:
997 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
998 err_icmp_mib:
999 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1000 err_ip_mib:
1001 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1002 err_udplite_mib:
1003 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1004 	return -ENOMEM;
1005 }
1006 
1007 static void ipv6_cleanup_mibs(struct net *net)
1008 {
1009 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1010 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1011 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1012 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1013 	kfree(net->mib.icmpv6msg_statistics);
1014 }
1015 
1016 static int __net_init inet6_net_init(struct net *net)
1017 {
1018 	int err = 0;
1019 
1020 	net->ipv6.sysctl.bindv6only = 0;
1021 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
1022 
1023 	err = ipv6_init_mibs(net);
1024 	if (err)
1025 		return err;
1026 #ifdef CONFIG_PROC_FS
1027 	err = udp6_proc_init(net);
1028 	if (err)
1029 		goto out;
1030 	err = tcp6_proc_init(net);
1031 	if (err)
1032 		goto proc_tcp6_fail;
1033 	err = ac6_proc_init(net);
1034 	if (err)
1035 		goto proc_ac6_fail;
1036 #endif
1037 	return err;
1038 
1039 #ifdef CONFIG_PROC_FS
1040 proc_ac6_fail:
1041 	tcp6_proc_exit(net);
1042 proc_tcp6_fail:
1043 	udp6_proc_exit(net);
1044 out:
1045 	ipv6_cleanup_mibs(net);
1046 	return err;
1047 #endif
1048 }
1049 
1050 static void __net_exit inet6_net_exit(struct net *net)
1051 {
1052 #ifdef CONFIG_PROC_FS
1053 	udp6_proc_exit(net);
1054 	tcp6_proc_exit(net);
1055 	ac6_proc_exit(net);
1056 #endif
1057 	ipv6_cleanup_mibs(net);
1058 }
1059 
1060 static struct pernet_operations inet6_net_ops = {
1061 	.init = inet6_net_init,
1062 	.exit = inet6_net_exit,
1063 };
1064 
1065 static int __init inet6_init(void)
1066 {
1067 	struct sk_buff *dummy_skb;
1068 	struct list_head *r;
1069 	int err = 0;
1070 
1071 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1072 
1073 	/* Register the socket-side information for inet6_create.  */
1074 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1075 		INIT_LIST_HEAD(r);
1076 
1077 	if (disable_ipv6_mod) {
1078 		printk(KERN_INFO
1079 		       "IPv6: Loaded, but administratively disabled, "
1080 		       "reboot required to enable\n");
1081 		goto out;
1082 	}
1083 
1084 	err = proto_register(&tcpv6_prot, 1);
1085 	if (err)
1086 		goto out;
1087 
1088 	err = proto_register(&udpv6_prot, 1);
1089 	if (err)
1090 		goto out_unregister_tcp_proto;
1091 
1092 	err = proto_register(&udplitev6_prot, 1);
1093 	if (err)
1094 		goto out_unregister_udp_proto;
1095 
1096 	err = proto_register(&rawv6_prot, 1);
1097 	if (err)
1098 		goto out_unregister_udplite_proto;
1099 
1100 
1101 	/* We MUST register RAW sockets before we create the ICMP6,
1102 	 * IGMP6, or NDISC control sockets.
1103 	 */
1104 	err = rawv6_init();
1105 	if (err)
1106 		goto out_unregister_raw_proto;
1107 
1108 	/* Register the family here so that the init calls below will
1109 	 * be able to create sockets. (?? is this dangerous ??)
1110 	 */
1111 	err = sock_register(&inet6_family_ops);
1112 	if (err)
1113 		goto out_sock_register_fail;
1114 
1115 #ifdef CONFIG_SYSCTL
1116 	err = ipv6_static_sysctl_register();
1117 	if (err)
1118 		goto static_sysctl_fail;
1119 #endif
1120 	tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1121 
1122 	/*
1123 	 *	ipngwg API draft makes clear that the correct semantics
1124 	 *	for TCP and UDP is to consider one TCP and UDP instance
1125 	 *	in a host available by both INET and INET6 APIs and
1126 	 *	able to communicate via both network protocols.
1127 	 */
1128 
1129 	err = register_pernet_subsys(&inet6_net_ops);
1130 	if (err)
1131 		goto register_pernet_fail;
1132 	err = icmpv6_init();
1133 	if (err)
1134 		goto icmp_fail;
1135 	err = ip6_mr_init();
1136 	if (err)
1137 		goto ipmr_fail;
1138 	err = ndisc_init();
1139 	if (err)
1140 		goto ndisc_fail;
1141 	err = igmp6_init();
1142 	if (err)
1143 		goto igmp_fail;
1144 	err = ipv6_netfilter_init();
1145 	if (err)
1146 		goto netfilter_fail;
1147 	/* Create /proc/foo6 entries. */
1148 #ifdef CONFIG_PROC_FS
1149 	err = -ENOMEM;
1150 	if (raw6_proc_init())
1151 		goto proc_raw6_fail;
1152 	if (udplite6_proc_init())
1153 		goto proc_udplite6_fail;
1154 	if (ipv6_misc_proc_init())
1155 		goto proc_misc6_fail;
1156 	if (if6_proc_init())
1157 		goto proc_if6_fail;
1158 #endif
1159 	err = ip6_route_init();
1160 	if (err)
1161 		goto ip6_route_fail;
1162 	err = ip6_flowlabel_init();
1163 	if (err)
1164 		goto ip6_flowlabel_fail;
1165 	err = addrconf_init();
1166 	if (err)
1167 		goto addrconf_fail;
1168 
1169 	/* Init v6 extension headers. */
1170 	err = ipv6_exthdrs_init();
1171 	if (err)
1172 		goto ipv6_exthdrs_fail;
1173 
1174 	err = ipv6_frag_init();
1175 	if (err)
1176 		goto ipv6_frag_fail;
1177 
1178 	/* Init v6 transport protocols. */
1179 	err = udpv6_init();
1180 	if (err)
1181 		goto udpv6_fail;
1182 
1183 	err = udplitev6_init();
1184 	if (err)
1185 		goto udplitev6_fail;
1186 
1187 	err = tcpv6_init();
1188 	if (err)
1189 		goto tcpv6_fail;
1190 
1191 	err = ipv6_packet_init();
1192 	if (err)
1193 		goto ipv6_packet_fail;
1194 
1195 #ifdef CONFIG_SYSCTL
1196 	err = ipv6_sysctl_register();
1197 	if (err)
1198 		goto sysctl_fail;
1199 #endif
1200 out:
1201 	return err;
1202 
1203 #ifdef CONFIG_SYSCTL
1204 sysctl_fail:
1205 	ipv6_packet_cleanup();
1206 #endif
1207 ipv6_packet_fail:
1208 	tcpv6_exit();
1209 tcpv6_fail:
1210 	udplitev6_exit();
1211 udplitev6_fail:
1212 	udpv6_exit();
1213 udpv6_fail:
1214 	ipv6_frag_exit();
1215 ipv6_frag_fail:
1216 	ipv6_exthdrs_exit();
1217 ipv6_exthdrs_fail:
1218 	addrconf_cleanup();
1219 addrconf_fail:
1220 	ip6_flowlabel_cleanup();
1221 ip6_flowlabel_fail:
1222 	ip6_route_cleanup();
1223 ip6_route_fail:
1224 #ifdef CONFIG_PROC_FS
1225 	if6_proc_exit();
1226 proc_if6_fail:
1227 	ipv6_misc_proc_exit();
1228 proc_misc6_fail:
1229 	udplite6_proc_exit();
1230 proc_udplite6_fail:
1231 	raw6_proc_exit();
1232 proc_raw6_fail:
1233 #endif
1234 	ipv6_netfilter_fini();
1235 netfilter_fail:
1236 	igmp6_cleanup();
1237 igmp_fail:
1238 	ndisc_cleanup();
1239 ndisc_fail:
1240 	ip6_mr_cleanup();
1241 ipmr_fail:
1242 	icmpv6_cleanup();
1243 icmp_fail:
1244 	unregister_pernet_subsys(&inet6_net_ops);
1245 register_pernet_fail:
1246 #ifdef CONFIG_SYSCTL
1247 	ipv6_static_sysctl_unregister();
1248 static_sysctl_fail:
1249 #endif
1250 	sock_unregister(PF_INET6);
1251 	rtnl_unregister_all(PF_INET6);
1252 out_sock_register_fail:
1253 	rawv6_exit();
1254 out_unregister_raw_proto:
1255 	proto_unregister(&rawv6_prot);
1256 out_unregister_udplite_proto:
1257 	proto_unregister(&udplitev6_prot);
1258 out_unregister_udp_proto:
1259 	proto_unregister(&udpv6_prot);
1260 out_unregister_tcp_proto:
1261 	proto_unregister(&tcpv6_prot);
1262 	goto out;
1263 }
1264 module_init(inet6_init);
1265 
1266 static void __exit inet6_exit(void)
1267 {
1268 	if (disable_ipv6_mod)
1269 		return;
1270 
1271 	/* First of all disallow new sockets creation. */
1272 	sock_unregister(PF_INET6);
1273 	/* Disallow any further netlink messages */
1274 	rtnl_unregister_all(PF_INET6);
1275 
1276 #ifdef CONFIG_SYSCTL
1277 	ipv6_sysctl_unregister();
1278 #endif
1279 	udpv6_exit();
1280 	udplitev6_exit();
1281 	tcpv6_exit();
1282 
1283 	/* Cleanup code parts. */
1284 	ipv6_packet_cleanup();
1285 	ipv6_frag_exit();
1286 	ipv6_exthdrs_exit();
1287 	addrconf_cleanup();
1288 	ip6_flowlabel_cleanup();
1289 	ip6_route_cleanup();
1290 #ifdef CONFIG_PROC_FS
1291 
1292 	/* Cleanup code parts. */
1293 	if6_proc_exit();
1294 	ipv6_misc_proc_exit();
1295 	udplite6_proc_exit();
1296 	raw6_proc_exit();
1297 #endif
1298 	ipv6_netfilter_fini();
1299 	igmp6_cleanup();
1300 	ndisc_cleanup();
1301 	ip6_mr_cleanup();
1302 	icmpv6_cleanup();
1303 	rawv6_exit();
1304 
1305 	unregister_pernet_subsys(&inet6_net_ops);
1306 #ifdef CONFIG_SYSCTL
1307 	ipv6_static_sysctl_unregister();
1308 #endif
1309 	proto_unregister(&rawv6_prot);
1310 	proto_unregister(&udplitev6_prot);
1311 	proto_unregister(&udpv6_prot);
1312 	proto_unregister(&tcpv6_prot);
1313 
1314 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1315 }
1316 module_exit(inet6_exit);
1317 
1318 MODULE_ALIAS_NETPROTO(PF_INET6);
1319