xref: /openbmc/linux/net/ipv6/af_inet6.c (revision b6dcefde)
1 /*
2  *	PF_INET6 socket protocol family
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Adapted from linux/net/ipv4/af_inet.c
9  *
10  * 	Fixes:
11  *	piggy, Karl Knutson	:	Socket protocol table
12  * 	Hideaki YOSHIFUJI	:	sin6_scope_id support
13  * 	Arnaldo Melo		: 	check proc_net_create return, cleanups
14  *
15  *	This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
20 
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/in.h>
28 #include <linux/kernel.h>
29 #include <linux/timer.h>
30 #include <linux/string.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/fcntl.h>
34 #include <linux/mm.h>
35 #include <linux/interrupt.h>
36 #include <linux/proc_fs.h>
37 #include <linux/stat.h>
38 #include <linux/init.h>
39 
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/icmpv6.h>
43 #include <linux/netfilter_ipv6.h>
44 
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/tcp.h>
50 #include <net/ipip.h>
51 #include <net/protocol.h>
52 #include <net/inet_common.h>
53 #include <net/route.h>
54 #include <net/transp_v6.h>
55 #include <net/ip6_route.h>
56 #include <net/addrconf.h>
57 #ifdef CONFIG_IPV6_TUNNEL
58 #include <net/ip6_tunnel.h>
59 #endif
60 
61 #include <asm/uaccess.h>
62 #include <asm/system.h>
63 #include <linux/mroute6.h>
64 
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
68 
69 /* The inetsw6 table contains everything that inet6_create needs to
70  * build a new socket.
71  */
72 static struct list_head inetsw6[SOCK_MAX];
73 static DEFINE_SPINLOCK(inetsw6_lock);
74 
75 struct ipv6_params ipv6_defaults = {
76 	.disable_ipv6 = 0,
77 	.autoconf = 1,
78 };
79 
80 static int disable_ipv6_mod = 0;
81 
82 module_param_named(disable, disable_ipv6_mod, int, 0444);
83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
84 
85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
87 
88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
90 
91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
92 {
93 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
94 
95 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
96 }
97 
98 static int inet6_create(struct net *net, struct socket *sock, int protocol,
99 			int kern)
100 {
101 	struct inet_sock *inet;
102 	struct ipv6_pinfo *np;
103 	struct sock *sk;
104 	struct inet_protosw *answer;
105 	struct proto *answer_prot;
106 	unsigned char answer_flags;
107 	char answer_no_check;
108 	int try_loading_module = 0;
109 	int err;
110 
111 	if (sock->type != SOCK_RAW &&
112 	    sock->type != SOCK_DGRAM &&
113 	    !inet_ehash_secret)
114 		build_ehash_secret();
115 
116 	/* Look for the requested type/protocol pair. */
117 lookup_protocol:
118 	err = -ESOCKTNOSUPPORT;
119 	rcu_read_lock();
120 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
121 
122 		err = 0;
123 		/* Check the non-wild match. */
124 		if (protocol == answer->protocol) {
125 			if (protocol != IPPROTO_IP)
126 				break;
127 		} else {
128 			/* Check for the two wild cases. */
129 			if (IPPROTO_IP == protocol) {
130 				protocol = answer->protocol;
131 				break;
132 			}
133 			if (IPPROTO_IP == answer->protocol)
134 				break;
135 		}
136 		err = -EPROTONOSUPPORT;
137 	}
138 
139 	if (err) {
140 		if (try_loading_module < 2) {
141 			rcu_read_unlock();
142 			/*
143 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
144 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
145 			 */
146 			if (++try_loading_module == 1)
147 				request_module("net-pf-%d-proto-%d-type-%d",
148 						PF_INET6, protocol, sock->type);
149 			/*
150 			 * Fall back to generic, e.g. net-pf-10-proto-132
151 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
152 			 */
153 			else
154 				request_module("net-pf-%d-proto-%d",
155 						PF_INET6, protocol);
156 			goto lookup_protocol;
157 		} else
158 			goto out_rcu_unlock;
159 	}
160 
161 	err = -EPERM;
162 	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
163 		goto out_rcu_unlock;
164 
165 	sock->ops = answer->ops;
166 	answer_prot = answer->prot;
167 	answer_no_check = answer->no_check;
168 	answer_flags = answer->flags;
169 	rcu_read_unlock();
170 
171 	WARN_ON(answer_prot->slab == NULL);
172 
173 	err = -ENOBUFS;
174 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
175 	if (sk == NULL)
176 		goto out;
177 
178 	sock_init_data(sock, sk);
179 
180 	err = 0;
181 	sk->sk_no_check = answer_no_check;
182 	if (INET_PROTOSW_REUSE & answer_flags)
183 		sk->sk_reuse = 1;
184 
185 	inet = inet_sk(sk);
186 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
187 
188 	if (SOCK_RAW == sock->type) {
189 		inet->inet_num = protocol;
190 		if (IPPROTO_RAW == protocol)
191 			inet->hdrincl = 1;
192 	}
193 
194 	sk->sk_destruct		= inet_sock_destruct;
195 	sk->sk_family		= PF_INET6;
196 	sk->sk_protocol		= protocol;
197 
198 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
199 
200 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
201 	np->hop_limit	= -1;
202 	np->mcast_hops	= -1;
203 	np->mc_loop	= 1;
204 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
205 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
206 
207 	/* Init the ipv4 part of the socket since we can have sockets
208 	 * using v6 API for ipv4.
209 	 */
210 	inet->uc_ttl	= -1;
211 
212 	inet->mc_loop	= 1;
213 	inet->mc_ttl	= 1;
214 	inet->mc_index	= 0;
215 	inet->mc_list	= NULL;
216 
217 	if (ipv4_config.no_pmtu_disc)
218 		inet->pmtudisc = IP_PMTUDISC_DONT;
219 	else
220 		inet->pmtudisc = IP_PMTUDISC_WANT;
221 	/*
222 	 * Increment only the relevant sk_prot->socks debug field, this changes
223 	 * the previous behaviour of incrementing both the equivalent to
224 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
225 	 *
226 	 * This allows better debug granularity as we'll know exactly how many
227 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
228 	 * transport protocol socks. -acme
229 	 */
230 	sk_refcnt_debug_inc(sk);
231 
232 	if (inet->inet_num) {
233 		/* It assumes that any protocol which allows
234 		 * the user to assign a number at socket
235 		 * creation time automatically shares.
236 		 */
237 		inet->inet_sport = htons(inet->inet_num);
238 		sk->sk_prot->hash(sk);
239 	}
240 	if (sk->sk_prot->init) {
241 		err = sk->sk_prot->init(sk);
242 		if (err) {
243 			sk_common_release(sk);
244 			goto out;
245 		}
246 	}
247 out:
248 	return err;
249 out_rcu_unlock:
250 	rcu_read_unlock();
251 	goto out;
252 }
253 
254 
255 /* bind for INET6 API */
256 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
257 {
258 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
259 	struct sock *sk = sock->sk;
260 	struct inet_sock *inet = inet_sk(sk);
261 	struct ipv6_pinfo *np = inet6_sk(sk);
262 	struct net *net = sock_net(sk);
263 	__be32 v4addr = 0;
264 	unsigned short snum;
265 	int addr_type = 0;
266 	int err = 0;
267 
268 	/* If the socket has its own bind function then use it. */
269 	if (sk->sk_prot->bind)
270 		return sk->sk_prot->bind(sk, uaddr, addr_len);
271 
272 	if (addr_len < SIN6_LEN_RFC2133)
273 		return -EINVAL;
274 	addr_type = ipv6_addr_type(&addr->sin6_addr);
275 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
276 		return -EINVAL;
277 
278 	snum = ntohs(addr->sin6_port);
279 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
280 		return -EACCES;
281 
282 	lock_sock(sk);
283 
284 	/* Check these errors (active socket, double bind). */
285 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
286 		err = -EINVAL;
287 		goto out;
288 	}
289 
290 	/* Check if the address belongs to the host. */
291 	if (addr_type == IPV6_ADDR_MAPPED) {
292 		int chk_addr_ret;
293 
294 		/* Binding to v4-mapped address on a v6-only socket
295 		 * makes no sense
296 		 */
297 		if (np->ipv6only) {
298 			err = -EINVAL;
299 			goto out;
300 		}
301 
302 		/* Reproduce AF_INET checks to make the bindings consitant */
303 		v4addr = addr->sin6_addr.s6_addr32[3];
304 		chk_addr_ret = inet_addr_type(net, v4addr);
305 		if (!sysctl_ip_nonlocal_bind &&
306 		    !(inet->freebind || inet->transparent) &&
307 		    v4addr != htonl(INADDR_ANY) &&
308 		    chk_addr_ret != RTN_LOCAL &&
309 		    chk_addr_ret != RTN_MULTICAST &&
310 		    chk_addr_ret != RTN_BROADCAST) {
311 			err = -EADDRNOTAVAIL;
312 			goto out;
313 		}
314 	} else {
315 		if (addr_type != IPV6_ADDR_ANY) {
316 			struct net_device *dev = NULL;
317 
318 			rcu_read_lock();
319 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
320 				if (addr_len >= sizeof(struct sockaddr_in6) &&
321 				    addr->sin6_scope_id) {
322 					/* Override any existing binding, if another one
323 					 * is supplied by user.
324 					 */
325 					sk->sk_bound_dev_if = addr->sin6_scope_id;
326 				}
327 
328 				/* Binding to link-local address requires an interface */
329 				if (!sk->sk_bound_dev_if) {
330 					err = -EINVAL;
331 					goto out_unlock;
332 				}
333 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
334 				if (!dev) {
335 					err = -ENODEV;
336 					goto out_unlock;
337 				}
338 			}
339 
340 			/* ipv4 addr of the socket is invalid.  Only the
341 			 * unspecified and mapped address have a v4 equivalent.
342 			 */
343 			v4addr = LOOPBACK4_IPV6;
344 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
345 				if (!ipv6_chk_addr(net, &addr->sin6_addr,
346 						   dev, 0)) {
347 					err = -EADDRNOTAVAIL;
348 					goto out_unlock;
349 				}
350 			}
351 			rcu_read_unlock();
352 		}
353 	}
354 
355 	inet->inet_rcv_saddr = v4addr;
356 	inet->inet_saddr = v4addr;
357 
358 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
359 
360 	if (!(addr_type & IPV6_ADDR_MULTICAST))
361 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
362 
363 	/* Make sure we are allowed to bind here. */
364 	if (sk->sk_prot->get_port(sk, snum)) {
365 		inet_reset_saddr(sk);
366 		err = -EADDRINUSE;
367 		goto out;
368 	}
369 
370 	if (addr_type != IPV6_ADDR_ANY) {
371 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
372 		if (addr_type != IPV6_ADDR_MAPPED)
373 			np->ipv6only = 1;
374 	}
375 	if (snum)
376 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
377 	inet->inet_sport = htons(inet->inet_num);
378 	inet->inet_dport = 0;
379 	inet->inet_daddr = 0;
380 out:
381 	release_sock(sk);
382 	return err;
383 out_unlock:
384 	rcu_read_unlock();
385 	goto out;
386 }
387 
388 EXPORT_SYMBOL(inet6_bind);
389 
390 int inet6_release(struct socket *sock)
391 {
392 	struct sock *sk = sock->sk;
393 
394 	if (sk == NULL)
395 		return -EINVAL;
396 
397 	/* Free mc lists */
398 	ipv6_sock_mc_close(sk);
399 
400 	/* Free ac lists */
401 	ipv6_sock_ac_close(sk);
402 
403 	return inet_release(sock);
404 }
405 
406 EXPORT_SYMBOL(inet6_release);
407 
408 void inet6_destroy_sock(struct sock *sk)
409 {
410 	struct ipv6_pinfo *np = inet6_sk(sk);
411 	struct sk_buff *skb;
412 	struct ipv6_txoptions *opt;
413 
414 	/* Release rx options */
415 
416 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
417 		kfree_skb(skb);
418 
419 	/* Free flowlabels */
420 	fl6_free_socklist(sk);
421 
422 	/* Free tx options */
423 
424 	if ((opt = xchg(&np->opt, NULL)) != NULL)
425 		sock_kfree_s(sk, opt, opt->tot_len);
426 }
427 
428 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
429 
430 /*
431  *	This does both peername and sockname.
432  */
433 
434 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
435 		 int *uaddr_len, int peer)
436 {
437 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
438 	struct sock *sk = sock->sk;
439 	struct inet_sock *inet = inet_sk(sk);
440 	struct ipv6_pinfo *np = inet6_sk(sk);
441 
442 	sin->sin6_family = AF_INET6;
443 	sin->sin6_flowinfo = 0;
444 	sin->sin6_scope_id = 0;
445 	if (peer) {
446 		if (!inet->inet_dport)
447 			return -ENOTCONN;
448 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
449 		    peer == 1)
450 			return -ENOTCONN;
451 		sin->sin6_port = inet->inet_dport;
452 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
453 		if (np->sndflow)
454 			sin->sin6_flowinfo = np->flow_label;
455 	} else {
456 		if (ipv6_addr_any(&np->rcv_saddr))
457 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
458 		else
459 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
460 
461 		sin->sin6_port = inet->inet_sport;
462 	}
463 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
464 		sin->sin6_scope_id = sk->sk_bound_dev_if;
465 	*uaddr_len = sizeof(*sin);
466 	return(0);
467 }
468 
469 EXPORT_SYMBOL(inet6_getname);
470 
471 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
472 {
473 	struct sock *sk = sock->sk;
474 	struct net *net = sock_net(sk);
475 
476 	switch(cmd)
477 	{
478 	case SIOCGSTAMP:
479 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
480 
481 	case SIOCGSTAMPNS:
482 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
483 
484 	case SIOCADDRT:
485 	case SIOCDELRT:
486 
487 		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
488 
489 	case SIOCSIFADDR:
490 		return addrconf_add_ifaddr(net, (void __user *) arg);
491 	case SIOCDIFADDR:
492 		return addrconf_del_ifaddr(net, (void __user *) arg);
493 	case SIOCSIFDSTADDR:
494 		return addrconf_set_dstaddr(net, (void __user *) arg);
495 	default:
496 		if (!sk->sk_prot->ioctl)
497 			return -ENOIOCTLCMD;
498 		return sk->sk_prot->ioctl(sk, cmd, arg);
499 	}
500 	/*NOTREACHED*/
501 	return(0);
502 }
503 
504 EXPORT_SYMBOL(inet6_ioctl);
505 
506 const struct proto_ops inet6_stream_ops = {
507 	.family		   = PF_INET6,
508 	.owner		   = THIS_MODULE,
509 	.release	   = inet6_release,
510 	.bind		   = inet6_bind,
511 	.connect	   = inet_stream_connect,	/* ok		*/
512 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
513 	.accept		   = inet_accept,		/* ok		*/
514 	.getname	   = inet6_getname,
515 	.poll		   = tcp_poll,			/* ok		*/
516 	.ioctl		   = inet6_ioctl,		/* must change  */
517 	.listen		   = inet_listen,		/* ok		*/
518 	.shutdown	   = inet_shutdown,		/* ok		*/
519 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
520 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
521 	.sendmsg	   = tcp_sendmsg,		/* ok		*/
522 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
523 	.mmap		   = sock_no_mmap,
524 	.sendpage	   = tcp_sendpage,
525 	.splice_read	   = tcp_splice_read,
526 #ifdef CONFIG_COMPAT
527 	.compat_setsockopt = compat_sock_common_setsockopt,
528 	.compat_getsockopt = compat_sock_common_getsockopt,
529 #endif
530 };
531 
532 const struct proto_ops inet6_dgram_ops = {
533 	.family		   = PF_INET6,
534 	.owner		   = THIS_MODULE,
535 	.release	   = inet6_release,
536 	.bind		   = inet6_bind,
537 	.connect	   = inet_dgram_connect,	/* ok		*/
538 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
539 	.accept		   = sock_no_accept,		/* a do nothing	*/
540 	.getname	   = inet6_getname,
541 	.poll		   = udp_poll,			/* ok		*/
542 	.ioctl		   = inet6_ioctl,		/* must change  */
543 	.listen		   = sock_no_listen,		/* ok		*/
544 	.shutdown	   = inet_shutdown,		/* ok		*/
545 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
546 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
547 	.sendmsg	   = inet_sendmsg,		/* ok		*/
548 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
549 	.mmap		   = sock_no_mmap,
550 	.sendpage	   = sock_no_sendpage,
551 #ifdef CONFIG_COMPAT
552 	.compat_setsockopt = compat_sock_common_setsockopt,
553 	.compat_getsockopt = compat_sock_common_getsockopt,
554 #endif
555 };
556 
557 static const struct net_proto_family inet6_family_ops = {
558 	.family = PF_INET6,
559 	.create = inet6_create,
560 	.owner	= THIS_MODULE,
561 };
562 
563 int inet6_register_protosw(struct inet_protosw *p)
564 {
565 	struct list_head *lh;
566 	struct inet_protosw *answer;
567 	struct list_head *last_perm;
568 	int protocol = p->protocol;
569 	int ret;
570 
571 	spin_lock_bh(&inetsw6_lock);
572 
573 	ret = -EINVAL;
574 	if (p->type >= SOCK_MAX)
575 		goto out_illegal;
576 
577 	/* If we are trying to override a permanent protocol, bail. */
578 	answer = NULL;
579 	ret = -EPERM;
580 	last_perm = &inetsw6[p->type];
581 	list_for_each(lh, &inetsw6[p->type]) {
582 		answer = list_entry(lh, struct inet_protosw, list);
583 
584 		/* Check only the non-wild match. */
585 		if (INET_PROTOSW_PERMANENT & answer->flags) {
586 			if (protocol == answer->protocol)
587 				break;
588 			last_perm = lh;
589 		}
590 
591 		answer = NULL;
592 	}
593 	if (answer)
594 		goto out_permanent;
595 
596 	/* Add the new entry after the last permanent entry if any, so that
597 	 * the new entry does not override a permanent entry when matched with
598 	 * a wild-card protocol. But it is allowed to override any existing
599 	 * non-permanent entry.  This means that when we remove this entry, the
600 	 * system automatically returns to the old behavior.
601 	 */
602 	list_add_rcu(&p->list, last_perm);
603 	ret = 0;
604 out:
605 	spin_unlock_bh(&inetsw6_lock);
606 	return ret;
607 
608 out_permanent:
609 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
610 	       protocol);
611 	goto out;
612 
613 out_illegal:
614 	printk(KERN_ERR
615 	       "Ignoring attempt to register invalid socket type %d.\n",
616 	       p->type);
617 	goto out;
618 }
619 
620 EXPORT_SYMBOL(inet6_register_protosw);
621 
622 void
623 inet6_unregister_protosw(struct inet_protosw *p)
624 {
625 	if (INET_PROTOSW_PERMANENT & p->flags) {
626 		printk(KERN_ERR
627 		       "Attempt to unregister permanent protocol %d.\n",
628 		       p->protocol);
629 	} else {
630 		spin_lock_bh(&inetsw6_lock);
631 		list_del_rcu(&p->list);
632 		spin_unlock_bh(&inetsw6_lock);
633 
634 		synchronize_net();
635 	}
636 }
637 
638 EXPORT_SYMBOL(inet6_unregister_protosw);
639 
640 int inet6_sk_rebuild_header(struct sock *sk)
641 {
642 	int err;
643 	struct dst_entry *dst;
644 	struct ipv6_pinfo *np = inet6_sk(sk);
645 
646 	dst = __sk_dst_check(sk, np->dst_cookie);
647 
648 	if (dst == NULL) {
649 		struct inet_sock *inet = inet_sk(sk);
650 		struct in6_addr *final_p = NULL, final;
651 		struct flowi fl;
652 
653 		memset(&fl, 0, sizeof(fl));
654 		fl.proto = sk->sk_protocol;
655 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
656 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
657 		fl.fl6_flowlabel = np->flow_label;
658 		fl.oif = sk->sk_bound_dev_if;
659 		fl.mark = sk->sk_mark;
660 		fl.fl_ip_dport = inet->inet_dport;
661 		fl.fl_ip_sport = inet->inet_sport;
662 		security_sk_classify_flow(sk, &fl);
663 
664 		if (np->opt && np->opt->srcrt) {
665 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
666 			ipv6_addr_copy(&final, &fl.fl6_dst);
667 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
668 			final_p = &final;
669 		}
670 
671 		err = ip6_dst_lookup(sk, &dst, &fl);
672 		if (err) {
673 			sk->sk_route_caps = 0;
674 			return err;
675 		}
676 		if (final_p)
677 			ipv6_addr_copy(&fl.fl6_dst, final_p);
678 
679 		if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
680 			sk->sk_err_soft = -err;
681 			return err;
682 		}
683 
684 		__ip6_dst_store(sk, dst, NULL, NULL);
685 	}
686 
687 	return 0;
688 }
689 
690 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
691 
692 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
693 {
694 	struct ipv6_pinfo *np = inet6_sk(sk);
695 	struct inet6_skb_parm *opt = IP6CB(skb);
696 
697 	if (np->rxopt.all) {
698 		if ((opt->hop && (np->rxopt.bits.hopopts ||
699 				  np->rxopt.bits.ohopopts)) ||
700 		    ((IPV6_FLOWINFO_MASK &
701 		      *(__be32 *)skb_network_header(skb)) &&
702 		     np->rxopt.bits.rxflow) ||
703 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
704 		     np->rxopt.bits.osrcrt)) ||
705 		    ((opt->dst1 || opt->dst0) &&
706 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
707 			return 1;
708 	}
709 	return 0;
710 }
711 
712 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
713 
714 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
715 {
716 	const struct inet6_protocol *ops = NULL;
717 
718 	for (;;) {
719 		struct ipv6_opt_hdr *opth;
720 		int len;
721 
722 		if (proto != NEXTHDR_HOP) {
723 			ops = rcu_dereference(inet6_protos[proto]);
724 
725 			if (unlikely(!ops))
726 				break;
727 
728 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
729 				break;
730 		}
731 
732 		if (unlikely(!pskb_may_pull(skb, 8)))
733 			break;
734 
735 		opth = (void *)skb->data;
736 		len = ipv6_optlen(opth);
737 
738 		if (unlikely(!pskb_may_pull(skb, len)))
739 			break;
740 
741 		proto = opth->nexthdr;
742 		__skb_pull(skb, len);
743 	}
744 
745 	return proto;
746 }
747 
748 static int ipv6_gso_send_check(struct sk_buff *skb)
749 {
750 	struct ipv6hdr *ipv6h;
751 	const struct inet6_protocol *ops;
752 	int err = -EINVAL;
753 
754 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
755 		goto out;
756 
757 	ipv6h = ipv6_hdr(skb);
758 	__skb_pull(skb, sizeof(*ipv6h));
759 	err = -EPROTONOSUPPORT;
760 
761 	rcu_read_lock();
762 	ops = rcu_dereference(inet6_protos[
763 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
764 
765 	if (likely(ops && ops->gso_send_check)) {
766 		skb_reset_transport_header(skb);
767 		err = ops->gso_send_check(skb);
768 	}
769 	rcu_read_unlock();
770 
771 out:
772 	return err;
773 }
774 
775 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
776 {
777 	struct sk_buff *segs = ERR_PTR(-EINVAL);
778 	struct ipv6hdr *ipv6h;
779 	const struct inet6_protocol *ops;
780 	int proto;
781 	struct frag_hdr *fptr;
782 	unsigned int unfrag_ip6hlen;
783 	u8 *prevhdr;
784 	int offset = 0;
785 
786 	if (!(features & NETIF_F_V6_CSUM))
787 		features &= ~NETIF_F_SG;
788 
789 	if (unlikely(skb_shinfo(skb)->gso_type &
790 		     ~(SKB_GSO_UDP |
791 		       SKB_GSO_DODGY |
792 		       SKB_GSO_TCP_ECN |
793 		       SKB_GSO_TCPV6 |
794 		       0)))
795 		goto out;
796 
797 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
798 		goto out;
799 
800 	ipv6h = ipv6_hdr(skb);
801 	__skb_pull(skb, sizeof(*ipv6h));
802 	segs = ERR_PTR(-EPROTONOSUPPORT);
803 
804 	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
805 	rcu_read_lock();
806 	ops = rcu_dereference(inet6_protos[proto]);
807 	if (likely(ops && ops->gso_segment)) {
808 		skb_reset_transport_header(skb);
809 		segs = ops->gso_segment(skb, features);
810 	}
811 	rcu_read_unlock();
812 
813 	if (unlikely(IS_ERR(segs)))
814 		goto out;
815 
816 	for (skb = segs; skb; skb = skb->next) {
817 		ipv6h = ipv6_hdr(skb);
818 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
819 					   sizeof(*ipv6h));
820 		if (proto == IPPROTO_UDP) {
821 			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
822 			fptr = (struct frag_hdr *)(skb_network_header(skb) +
823 				unfrag_ip6hlen);
824 			fptr->frag_off = htons(offset);
825 			if (skb->next != NULL)
826 				fptr->frag_off |= htons(IP6_MF);
827 			offset += (ntohs(ipv6h->payload_len) -
828 				   sizeof(struct frag_hdr));
829 		}
830 	}
831 
832 out:
833 	return segs;
834 }
835 
836 struct ipv6_gro_cb {
837 	struct napi_gro_cb napi;
838 	int proto;
839 };
840 
841 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
842 
843 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
844 					 struct sk_buff *skb)
845 {
846 	const struct inet6_protocol *ops;
847 	struct sk_buff **pp = NULL;
848 	struct sk_buff *p;
849 	struct ipv6hdr *iph;
850 	unsigned int nlen;
851 	unsigned int hlen;
852 	unsigned int off;
853 	int flush = 1;
854 	int proto;
855 	__wsum csum;
856 
857 	off = skb_gro_offset(skb);
858 	hlen = off + sizeof(*iph);
859 	iph = skb_gro_header_fast(skb, off);
860 	if (skb_gro_header_hard(skb, hlen)) {
861 		iph = skb_gro_header_slow(skb, hlen, off);
862 		if (unlikely(!iph))
863 			goto out;
864 	}
865 
866 	skb_gro_pull(skb, sizeof(*iph));
867 	skb_set_transport_header(skb, skb_gro_offset(skb));
868 
869 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
870 
871 	rcu_read_lock();
872 	proto = iph->nexthdr;
873 	ops = rcu_dereference(inet6_protos[proto]);
874 	if (!ops || !ops->gro_receive) {
875 		__pskb_pull(skb, skb_gro_offset(skb));
876 		proto = ipv6_gso_pull_exthdrs(skb, proto);
877 		skb_gro_pull(skb, -skb_transport_offset(skb));
878 		skb_reset_transport_header(skb);
879 		__skb_push(skb, skb_gro_offset(skb));
880 
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 **)net->mib.udp_stats_in6,
975 			  sizeof (struct udp_mib)) < 0)
976 		return -ENOMEM;
977 	if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
978 			  sizeof (struct udp_mib)) < 0)
979 		goto err_udplite_mib;
980 	if (snmp_mib_init((void **)net->mib.ipv6_statistics,
981 			  sizeof(struct ipstats_mib)) < 0)
982 		goto err_ip_mib;
983 	if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
984 			  sizeof(struct icmpv6_mib)) < 0)
985 		goto err_icmp_mib;
986 	if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
987 			  sizeof(struct icmpv6msg_mib)) < 0)
988 		goto err_icmpmsg_mib;
989 	return 0;
990 
991 err_icmpmsg_mib:
992 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
993 err_icmp_mib:
994 	snmp_mib_free((void **)net->mib.ipv6_statistics);
995 err_ip_mib:
996 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
997 err_udplite_mib:
998 	snmp_mib_free((void **)net->mib.udp_stats_in6);
999 	return -ENOMEM;
1000 }
1001 
1002 static void __net_exit ipv6_cleanup_mibs(struct net *net)
1003 {
1004 	snmp_mib_free((void **)net->mib.udp_stats_in6);
1005 	snmp_mib_free((void **)net->mib.udplite_stats_in6);
1006 	snmp_mib_free((void **)net->mib.ipv6_statistics);
1007 	snmp_mib_free((void **)net->mib.icmpv6_statistics);
1008 	snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
1009 }
1010 
1011 static int __net_init inet6_net_init(struct net *net)
1012 {
1013 	int err = 0;
1014 
1015 	net->ipv6.sysctl.bindv6only = 0;
1016 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
1017 
1018 	err = ipv6_init_mibs(net);
1019 	if (err)
1020 		return err;
1021 #ifdef CONFIG_PROC_FS
1022 	err = udp6_proc_init(net);
1023 	if (err)
1024 		goto out;
1025 	err = tcp6_proc_init(net);
1026 	if (err)
1027 		goto proc_tcp6_fail;
1028 	err = ac6_proc_init(net);
1029 	if (err)
1030 		goto proc_ac6_fail;
1031 #endif
1032 	return err;
1033 
1034 #ifdef CONFIG_PROC_FS
1035 proc_ac6_fail:
1036 	tcp6_proc_exit(net);
1037 proc_tcp6_fail:
1038 	udp6_proc_exit(net);
1039 out:
1040 	ipv6_cleanup_mibs(net);
1041 	return err;
1042 #endif
1043 }
1044 
1045 static void inet6_net_exit(struct net *net)
1046 {
1047 #ifdef CONFIG_PROC_FS
1048 	udp6_proc_exit(net);
1049 	tcp6_proc_exit(net);
1050 	ac6_proc_exit(net);
1051 #endif
1052 	ipv6_cleanup_mibs(net);
1053 }
1054 
1055 static struct pernet_operations inet6_net_ops = {
1056 	.init = inet6_net_init,
1057 	.exit = inet6_net_exit,
1058 };
1059 
1060 static int __init inet6_init(void)
1061 {
1062 	struct sk_buff *dummy_skb;
1063 	struct list_head *r;
1064 	int err = 0;
1065 
1066 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1067 
1068 	/* Register the socket-side information for inet6_create.  */
1069 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1070 		INIT_LIST_HEAD(r);
1071 
1072 	if (disable_ipv6_mod) {
1073 		printk(KERN_INFO
1074 		       "IPv6: Loaded, but administratively disabled, "
1075 		       "reboot required to enable\n");
1076 		goto out;
1077 	}
1078 
1079 	err = proto_register(&tcpv6_prot, 1);
1080 	if (err)
1081 		goto out;
1082 
1083 	err = proto_register(&udpv6_prot, 1);
1084 	if (err)
1085 		goto out_unregister_tcp_proto;
1086 
1087 	err = proto_register(&udplitev6_prot, 1);
1088 	if (err)
1089 		goto out_unregister_udp_proto;
1090 
1091 	err = proto_register(&rawv6_prot, 1);
1092 	if (err)
1093 		goto out_unregister_udplite_proto;
1094 
1095 
1096 	/* We MUST register RAW sockets before we create the ICMP6,
1097 	 * IGMP6, or NDISC control sockets.
1098 	 */
1099 	err = rawv6_init();
1100 	if (err)
1101 		goto out_unregister_raw_proto;
1102 
1103 	/* Register the family here so that the init calls below will
1104 	 * be able to create sockets. (?? is this dangerous ??)
1105 	 */
1106 	err = sock_register(&inet6_family_ops);
1107 	if (err)
1108 		goto out_sock_register_fail;
1109 
1110 #ifdef CONFIG_SYSCTL
1111 	err = ipv6_static_sysctl_register();
1112 	if (err)
1113 		goto static_sysctl_fail;
1114 #endif
1115 	/*
1116 	 *	ipngwg API draft makes clear that the correct semantics
1117 	 *	for TCP and UDP is to consider one TCP and UDP instance
1118 	 *	in a host availiable by both INET and INET6 APIs and
1119 	 *	able to communicate via both network protocols.
1120 	 */
1121 
1122 	err = register_pernet_subsys(&inet6_net_ops);
1123 	if (err)
1124 		goto register_pernet_fail;
1125 	err = icmpv6_init();
1126 	if (err)
1127 		goto icmp_fail;
1128 	err = ip6_mr_init();
1129 	if (err)
1130 		goto ipmr_fail;
1131 	err = ndisc_init();
1132 	if (err)
1133 		goto ndisc_fail;
1134 	err = igmp6_init();
1135 	if (err)
1136 		goto igmp_fail;
1137 	err = ipv6_netfilter_init();
1138 	if (err)
1139 		goto netfilter_fail;
1140 	/* Create /proc/foo6 entries. */
1141 #ifdef CONFIG_PROC_FS
1142 	err = -ENOMEM;
1143 	if (raw6_proc_init())
1144 		goto proc_raw6_fail;
1145 	if (udplite6_proc_init())
1146 		goto proc_udplite6_fail;
1147 	if (ipv6_misc_proc_init())
1148 		goto proc_misc6_fail;
1149 	if (if6_proc_init())
1150 		goto proc_if6_fail;
1151 #endif
1152 	err = ip6_route_init();
1153 	if (err)
1154 		goto ip6_route_fail;
1155 	err = ip6_flowlabel_init();
1156 	if (err)
1157 		goto ip6_flowlabel_fail;
1158 	err = addrconf_init();
1159 	if (err)
1160 		goto addrconf_fail;
1161 
1162 	/* Init v6 extension headers. */
1163 	err = ipv6_exthdrs_init();
1164 	if (err)
1165 		goto ipv6_exthdrs_fail;
1166 
1167 	err = ipv6_frag_init();
1168 	if (err)
1169 		goto ipv6_frag_fail;
1170 
1171 	/* Init v6 transport protocols. */
1172 	err = udpv6_init();
1173 	if (err)
1174 		goto udpv6_fail;
1175 
1176 	err = udplitev6_init();
1177 	if (err)
1178 		goto udplitev6_fail;
1179 
1180 	err = tcpv6_init();
1181 	if (err)
1182 		goto tcpv6_fail;
1183 
1184 	err = ipv6_packet_init();
1185 	if (err)
1186 		goto ipv6_packet_fail;
1187 
1188 #ifdef CONFIG_SYSCTL
1189 	err = ipv6_sysctl_register();
1190 	if (err)
1191 		goto sysctl_fail;
1192 #endif
1193 out:
1194 	return err;
1195 
1196 #ifdef CONFIG_SYSCTL
1197 sysctl_fail:
1198 	ipv6_packet_cleanup();
1199 #endif
1200 ipv6_packet_fail:
1201 	tcpv6_exit();
1202 tcpv6_fail:
1203 	udplitev6_exit();
1204 udplitev6_fail:
1205 	udpv6_exit();
1206 udpv6_fail:
1207 	ipv6_frag_exit();
1208 ipv6_frag_fail:
1209 	ipv6_exthdrs_exit();
1210 ipv6_exthdrs_fail:
1211 	addrconf_cleanup();
1212 addrconf_fail:
1213 	ip6_flowlabel_cleanup();
1214 ip6_flowlabel_fail:
1215 	ip6_route_cleanup();
1216 ip6_route_fail:
1217 #ifdef CONFIG_PROC_FS
1218 	if6_proc_exit();
1219 proc_if6_fail:
1220 	ipv6_misc_proc_exit();
1221 proc_misc6_fail:
1222 	udplite6_proc_exit();
1223 proc_udplite6_fail:
1224 	raw6_proc_exit();
1225 proc_raw6_fail:
1226 #endif
1227 	ipv6_netfilter_fini();
1228 netfilter_fail:
1229 	igmp6_cleanup();
1230 igmp_fail:
1231 	ndisc_cleanup();
1232 ndisc_fail:
1233 	ip6_mr_cleanup();
1234 ipmr_fail:
1235 	icmpv6_cleanup();
1236 icmp_fail:
1237 	unregister_pernet_subsys(&inet6_net_ops);
1238 register_pernet_fail:
1239 #ifdef CONFIG_SYSCTL
1240 	ipv6_static_sysctl_unregister();
1241 static_sysctl_fail:
1242 #endif
1243 	sock_unregister(PF_INET6);
1244 	rtnl_unregister_all(PF_INET6);
1245 out_sock_register_fail:
1246 	rawv6_exit();
1247 out_unregister_raw_proto:
1248 	proto_unregister(&rawv6_prot);
1249 out_unregister_udplite_proto:
1250 	proto_unregister(&udplitev6_prot);
1251 out_unregister_udp_proto:
1252 	proto_unregister(&udpv6_prot);
1253 out_unregister_tcp_proto:
1254 	proto_unregister(&tcpv6_prot);
1255 	goto out;
1256 }
1257 module_init(inet6_init);
1258 
1259 static void __exit inet6_exit(void)
1260 {
1261 	if (disable_ipv6_mod)
1262 		return;
1263 
1264 	/* First of all disallow new sockets creation. */
1265 	sock_unregister(PF_INET6);
1266 	/* Disallow any further netlink messages */
1267 	rtnl_unregister_all(PF_INET6);
1268 
1269 #ifdef CONFIG_SYSCTL
1270 	ipv6_sysctl_unregister();
1271 #endif
1272 	udpv6_exit();
1273 	udplitev6_exit();
1274 	tcpv6_exit();
1275 
1276 	/* Cleanup code parts. */
1277 	ipv6_packet_cleanup();
1278 	ipv6_frag_exit();
1279 	ipv6_exthdrs_exit();
1280 	addrconf_cleanup();
1281 	ip6_flowlabel_cleanup();
1282 	ip6_route_cleanup();
1283 #ifdef CONFIG_PROC_FS
1284 
1285 	/* Cleanup code parts. */
1286 	if6_proc_exit();
1287 	ipv6_misc_proc_exit();
1288 	udplite6_proc_exit();
1289 	raw6_proc_exit();
1290 #endif
1291 	ipv6_netfilter_fini();
1292 	igmp6_cleanup();
1293 	ndisc_cleanup();
1294 	ip6_mr_cleanup();
1295 	icmpv6_cleanup();
1296 	rawv6_exit();
1297 
1298 	unregister_pernet_subsys(&inet6_net_ops);
1299 #ifdef CONFIG_SYSCTL
1300 	ipv6_static_sysctl_unregister();
1301 #endif
1302 	proto_unregister(&rawv6_prot);
1303 	proto_unregister(&udplitev6_prot);
1304 	proto_unregister(&udpv6_prot);
1305 	proto_unregister(&tcpv6_prot);
1306 
1307 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1308 }
1309 module_exit(inet6_exit);
1310 
1311 MODULE_ALIAS_NETPROTO(PF_INET6);
1312