af_inet6.c revision 94911fe31710e355e9fcac8b1a7eb36c4953f36b 
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 *	$Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
11 *
12 * 	Fixes:
13 *	piggy, Karl Knutson	:	Socket protocol table
14 * 	Hideaki YOSHIFUJI	:	sin6_scope_id support
15 * 	Arnaldo Melo		: 	check proc_net_create return, cleanups
16 *
17 *	This program is free software; you can redistribute it and/or
18 *      modify it under the terms of the GNU General Public License
19 *      as published by the Free Software Foundation; either version
20 *      2 of the License, or (at your option) any later version.
21 */
22
23
24#include <linux/module.h>
25#include <linux/capability.h>
26#include <linux/errno.h>
27#include <linux/types.h>
28#include <linux/socket.h>
29#include <linux/in.h>
30#include <linux/kernel.h>
31#include <linux/timer.h>
32#include <linux/string.h>
33#include <linux/sockios.h>
34#include <linux/net.h>
35#include <linux/fcntl.h>
36#include <linux/mm.h>
37#include <linux/interrupt.h>
38#include <linux/proc_fs.h>
39#include <linux/stat.h>
40#include <linux/init.h>
41
42#include <linux/inet.h>
43#include <linux/netdevice.h>
44#include <linux/icmpv6.h>
45#include <linux/netfilter_ipv6.h>
46
47#include <net/ip.h>
48#include <net/ipv6.h>
49#include <net/udp.h>
50#include <net/udplite.h>
51#include <net/tcp.h>
52#include <net/ipip.h>
53#include <net/protocol.h>
54#include <net/inet_common.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
65MODULE_AUTHOR("Cast of dozens");
66MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67MODULE_LICENSE("GPL");
68
69/* The inetsw6 table contains everything that inet6_create needs to
70 * build a new socket.
71 */
72static struct list_head inetsw6[SOCK_MAX];
73static DEFINE_SPINLOCK(inetsw6_lock);
74
75static __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
82static 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 list_head *p;
88	struct inet_protosw *answer;
89	struct proto *answer_prot;
90	unsigned char answer_flags;
91	char answer_no_check;
92	int try_loading_module = 0;
93	int err;
94
95	if (net != &init_net)
96		return -EAFNOSUPPORT;
97
98	if (sock->type != SOCK_RAW &&
99	    sock->type != SOCK_DGRAM &&
100	    !inet_ehash_secret)
101		build_ehash_secret();
102
103	/* Look for the requested type/protocol pair. */
104	answer = NULL;
105lookup_protocol:
106	err = -ESOCKTNOSUPPORT;
107	rcu_read_lock();
108	list_for_each_rcu(p, &inetsw6[sock->type]) {
109		answer = list_entry(p, struct inet_protosw, list);
110
111		/* Check the non-wild match. */
112		if (protocol == answer->protocol) {
113			if (protocol != IPPROTO_IP)
114				break;
115		} else {
116			/* Check for the two wild cases. */
117			if (IPPROTO_IP == protocol) {
118				protocol = answer->protocol;
119				break;
120			}
121			if (IPPROTO_IP == answer->protocol)
122				break;
123		}
124		err = -EPROTONOSUPPORT;
125		answer = NULL;
126	}
127
128	if (!answer) {
129		if (try_loading_module < 2) {
130			rcu_read_unlock();
131			/*
132			 * Be more specific, e.g. net-pf-10-proto-132-type-1
133			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
134			 */
135			if (++try_loading_module == 1)
136				request_module("net-pf-%d-proto-%d-type-%d",
137						PF_INET6, protocol, sock->type);
138			/*
139			 * Fall back to generic, e.g. net-pf-10-proto-132
140			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
141			 */
142			else
143				request_module("net-pf-%d-proto-%d",
144						PF_INET6, protocol);
145			goto lookup_protocol;
146		} else
147			goto out_rcu_unlock;
148	}
149
150	err = -EPERM;
151	if (answer->capability > 0 && !capable(answer->capability))
152		goto out_rcu_unlock;
153
154	sock->ops = answer->ops;
155	answer_prot = answer->prot;
156	answer_no_check = answer->no_check;
157	answer_flags = answer->flags;
158	rcu_read_unlock();
159
160	BUG_TRAP(answer_prot->slab != NULL);
161
162	err = -ENOBUFS;
163	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
164	if (sk == NULL)
165		goto out;
166
167	sock_init_data(sock, sk);
168
169	err = 0;
170	sk->sk_no_check = answer_no_check;
171	if (INET_PROTOSW_REUSE & answer_flags)
172		sk->sk_reuse = 1;
173
174	inet = inet_sk(sk);
175	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
176
177	if (SOCK_RAW == sock->type) {
178		inet->num = protocol;
179		if (IPPROTO_RAW == protocol)
180			inet->hdrincl = 1;
181	}
182
183	sk->sk_destruct		= inet_sock_destruct;
184	sk->sk_family		= PF_INET6;
185	sk->sk_protocol		= protocol;
186
187	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
188
189	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
190	np->hop_limit	= -1;
191	np->mcast_hops	= -1;
192	np->mc_loop	= 1;
193	np->pmtudisc	= IPV6_PMTUDISC_WANT;
194	np->ipv6only	= init_net.ipv6.sysctl.bindv6only;
195
196	/* Init the ipv4 part of the socket since we can have sockets
197	 * using v6 API for ipv4.
198	 */
199	inet->uc_ttl	= -1;
200
201	inet->mc_loop	= 1;
202	inet->mc_ttl	= 1;
203	inet->mc_index	= 0;
204	inet->mc_list	= NULL;
205
206	if (ipv4_config.no_pmtu_disc)
207		inet->pmtudisc = IP_PMTUDISC_DONT;
208	else
209		inet->pmtudisc = IP_PMTUDISC_WANT;
210	/*
211	 * Increment only the relevant sk_prot->socks debug field, this changes
212	 * the previous behaviour of incrementing both the equivalent to
213	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
214	 *
215	 * This allows better debug granularity as we'll know exactly how many
216	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
217	 * transport protocol socks. -acme
218	 */
219	sk_refcnt_debug_inc(sk);
220
221	if (inet->num) {
222		/* It assumes that any protocol which allows
223		 * the user to assign a number at socket
224		 * creation time automatically shares.
225		 */
226		inet->sport = htons(inet->num);
227		sk->sk_prot->hash(sk);
228	}
229	if (sk->sk_prot->init) {
230		err = sk->sk_prot->init(sk);
231		if (err) {
232			sk_common_release(sk);
233			goto out;
234		}
235	}
236out:
237	return err;
238out_rcu_unlock:
239	rcu_read_unlock();
240	goto out;
241}
242
243
244/* bind for INET6 API */
245int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
246{
247	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
248	struct sock *sk = sock->sk;
249	struct inet_sock *inet = inet_sk(sk);
250	struct ipv6_pinfo *np = inet6_sk(sk);
251	__be32 v4addr = 0;
252	unsigned short snum;
253	int addr_type = 0;
254	int err = 0;
255
256	/* If the socket has its own bind function then use it. */
257	if (sk->sk_prot->bind)
258		return sk->sk_prot->bind(sk, uaddr, addr_len);
259
260	if (addr_len < SIN6_LEN_RFC2133)
261		return -EINVAL;
262	addr_type = ipv6_addr_type(&addr->sin6_addr);
263	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
264		return -EINVAL;
265
266	snum = ntohs(addr->sin6_port);
267	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
268		return -EACCES;
269
270	lock_sock(sk);
271
272	/* Check these errors (active socket, double bind). */
273	if (sk->sk_state != TCP_CLOSE || inet->num) {
274		err = -EINVAL;
275		goto out;
276	}
277
278	/* Check if the address belongs to the host. */
279	if (addr_type == IPV6_ADDR_MAPPED) {
280		v4addr = addr->sin6_addr.s6_addr32[3];
281		if (inet_addr_type(&init_net, v4addr) != RTN_LOCAL) {
282			err = -EADDRNOTAVAIL;
283			goto out;
284		}
285	} else {
286		if (addr_type != IPV6_ADDR_ANY) {
287			struct net_device *dev = NULL;
288
289			if (addr_type & IPV6_ADDR_LINKLOCAL) {
290				if (addr_len >= sizeof(struct sockaddr_in6) &&
291				    addr->sin6_scope_id) {
292					/* Override any existing binding, if another one
293					 * is supplied by user.
294					 */
295					sk->sk_bound_dev_if = addr->sin6_scope_id;
296				}
297
298				/* Binding to link-local address requires an interface */
299				if (!sk->sk_bound_dev_if) {
300					err = -EINVAL;
301					goto out;
302				}
303				dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
304				if (!dev) {
305					err = -ENODEV;
306					goto out;
307				}
308			}
309
310			/* ipv4 addr of the socket is invalid.  Only the
311			 * unspecified and mapped address have a v4 equivalent.
312			 */
313			v4addr = LOOPBACK4_IPV6;
314			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
315				if (!ipv6_chk_addr(&init_net, &addr->sin6_addr,
316						   dev, 0)) {
317					if (dev)
318						dev_put(dev);
319					err = -EADDRNOTAVAIL;
320					goto out;
321				}
322			}
323			if (dev)
324				dev_put(dev);
325		}
326	}
327
328	inet->rcv_saddr = v4addr;
329	inet->saddr = v4addr;
330
331	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
332
333	if (!(addr_type & IPV6_ADDR_MULTICAST))
334		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
335
336	/* Make sure we are allowed to bind here. */
337	if (sk->sk_prot->get_port(sk, snum)) {
338		inet_reset_saddr(sk);
339		err = -EADDRINUSE;
340		goto out;
341	}
342
343	if (addr_type != IPV6_ADDR_ANY)
344		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
345	if (snum)
346		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
347	inet->sport = htons(inet->num);
348	inet->dport = 0;
349	inet->daddr = 0;
350out:
351	release_sock(sk);
352	return err;
353}
354
355EXPORT_SYMBOL(inet6_bind);
356
357int inet6_release(struct socket *sock)
358{
359	struct sock *sk = sock->sk;
360
361	if (sk == NULL)
362		return -EINVAL;
363
364	/* Free mc lists */
365	ipv6_sock_mc_close(sk);
366
367	/* Free ac lists */
368	ipv6_sock_ac_close(sk);
369
370	return inet_release(sock);
371}
372
373EXPORT_SYMBOL(inet6_release);
374
375int inet6_destroy_sock(struct sock *sk)
376{
377	struct ipv6_pinfo *np = inet6_sk(sk);
378	struct sk_buff *skb;
379	struct ipv6_txoptions *opt;
380
381	/* Release rx options */
382
383	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
384		kfree_skb(skb);
385
386	/* Free flowlabels */
387	fl6_free_socklist(sk);
388
389	/* Free tx options */
390
391	if ((opt = xchg(&np->opt, NULL)) != NULL)
392		sock_kfree_s(sk, opt, opt->tot_len);
393
394	return 0;
395}
396
397EXPORT_SYMBOL_GPL(inet6_destroy_sock);
398
399/*
400 *	This does both peername and sockname.
401 */
402
403int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
404		 int *uaddr_len, int peer)
405{
406	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
407	struct sock *sk = sock->sk;
408	struct inet_sock *inet = inet_sk(sk);
409	struct ipv6_pinfo *np = inet6_sk(sk);
410
411	sin->sin6_family = AF_INET6;
412	sin->sin6_flowinfo = 0;
413	sin->sin6_scope_id = 0;
414	if (peer) {
415		if (!inet->dport)
416			return -ENOTCONN;
417		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
418		    peer == 1)
419			return -ENOTCONN;
420		sin->sin6_port = inet->dport;
421		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
422		if (np->sndflow)
423			sin->sin6_flowinfo = np->flow_label;
424	} else {
425		if (ipv6_addr_any(&np->rcv_saddr))
426			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
427		else
428			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
429
430		sin->sin6_port = inet->sport;
431	}
432	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
433		sin->sin6_scope_id = sk->sk_bound_dev_if;
434	*uaddr_len = sizeof(*sin);
435	return(0);
436}
437
438EXPORT_SYMBOL(inet6_getname);
439
440int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
441{
442	struct sock *sk = sock->sk;
443	struct net *net = sk->sk_net;
444
445	switch(cmd)
446	{
447	case SIOCGSTAMP:
448		return sock_get_timestamp(sk, (struct timeval __user *)arg);
449
450	case SIOCGSTAMPNS:
451		return sock_get_timestampns(sk, (struct timespec __user *)arg);
452
453	case SIOCADDRT:
454	case SIOCDELRT:
455
456		return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
457
458	case SIOCSIFADDR:
459		return addrconf_add_ifaddr((void __user *) arg);
460	case SIOCDIFADDR:
461		return addrconf_del_ifaddr((void __user *) arg);
462	case SIOCSIFDSTADDR:
463		return addrconf_set_dstaddr((void __user *) arg);
464	default:
465		if (!sk->sk_prot->ioctl)
466			return -ENOIOCTLCMD;
467		return sk->sk_prot->ioctl(sk, cmd, arg);
468	}
469	/*NOTREACHED*/
470	return(0);
471}
472
473EXPORT_SYMBOL(inet6_ioctl);
474
475const struct proto_ops inet6_stream_ops = {
476	.family		   = PF_INET6,
477	.owner		   = THIS_MODULE,
478	.release	   = inet6_release,
479	.bind		   = inet6_bind,
480	.connect	   = inet_stream_connect,	/* ok		*/
481	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
482	.accept		   = inet_accept,		/* ok		*/
483	.getname	   = inet6_getname,
484	.poll		   = tcp_poll,			/* ok		*/
485	.ioctl		   = inet6_ioctl,		/* must change  */
486	.listen		   = inet_listen,		/* ok		*/
487	.shutdown	   = inet_shutdown,		/* ok		*/
488	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
489	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
490	.sendmsg	   = tcp_sendmsg,		/* ok		*/
491	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
492	.mmap		   = sock_no_mmap,
493	.sendpage	   = tcp_sendpage,
494	.splice_read	   = tcp_splice_read,
495#ifdef CONFIG_COMPAT
496	.compat_setsockopt = compat_sock_common_setsockopt,
497	.compat_getsockopt = compat_sock_common_getsockopt,
498#endif
499};
500
501const struct proto_ops inet6_dgram_ops = {
502	.family		   = PF_INET6,
503	.owner		   = THIS_MODULE,
504	.release	   = inet6_release,
505	.bind		   = inet6_bind,
506	.connect	   = inet_dgram_connect,	/* ok		*/
507	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
508	.accept		   = sock_no_accept,		/* a do nothing	*/
509	.getname	   = inet6_getname,
510	.poll		   = udp_poll,			/* ok		*/
511	.ioctl		   = inet6_ioctl,		/* must change  */
512	.listen		   = sock_no_listen,		/* ok		*/
513	.shutdown	   = inet_shutdown,		/* ok		*/
514	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
515	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
516	.sendmsg	   = inet_sendmsg,		/* ok		*/
517	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
518	.mmap		   = sock_no_mmap,
519	.sendpage	   = sock_no_sendpage,
520#ifdef CONFIG_COMPAT
521	.compat_setsockopt = compat_sock_common_setsockopt,
522	.compat_getsockopt = compat_sock_common_getsockopt,
523#endif
524};
525
526static struct net_proto_family inet6_family_ops = {
527	.family = PF_INET6,
528	.create = inet6_create,
529	.owner	= THIS_MODULE,
530};
531
532int inet6_register_protosw(struct inet_protosw *p)
533{
534	struct list_head *lh;
535	struct inet_protosw *answer;
536	struct list_head *last_perm;
537	int protocol = p->protocol;
538	int ret;
539
540	spin_lock_bh(&inetsw6_lock);
541
542	ret = -EINVAL;
543	if (p->type >= SOCK_MAX)
544		goto out_illegal;
545
546	/* If we are trying to override a permanent protocol, bail. */
547	answer = NULL;
548	ret = -EPERM;
549	last_perm = &inetsw6[p->type];
550	list_for_each(lh, &inetsw6[p->type]) {
551		answer = list_entry(lh, struct inet_protosw, list);
552
553		/* Check only the non-wild match. */
554		if (INET_PROTOSW_PERMANENT & answer->flags) {
555			if (protocol == answer->protocol)
556				break;
557			last_perm = lh;
558		}
559
560		answer = NULL;
561	}
562	if (answer)
563		goto out_permanent;
564
565	/* Add the new entry after the last permanent entry if any, so that
566	 * the new entry does not override a permanent entry when matched with
567	 * a wild-card protocol. But it is allowed to override any existing
568	 * non-permanent entry.  This means that when we remove this entry, the
569	 * system automatically returns to the old behavior.
570	 */
571	list_add_rcu(&p->list, last_perm);
572	ret = 0;
573out:
574	spin_unlock_bh(&inetsw6_lock);
575	return ret;
576
577out_permanent:
578	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
579	       protocol);
580	goto out;
581
582out_illegal:
583	printk(KERN_ERR
584	       "Ignoring attempt to register invalid socket type %d.\n",
585	       p->type);
586	goto out;
587}
588
589EXPORT_SYMBOL(inet6_register_protosw);
590
591void
592inet6_unregister_protosw(struct inet_protosw *p)
593{
594	if (INET_PROTOSW_PERMANENT & p->flags) {
595		printk(KERN_ERR
596		       "Attempt to unregister permanent protocol %d.\n",
597		       p->protocol);
598	} else {
599		spin_lock_bh(&inetsw6_lock);
600		list_del_rcu(&p->list);
601		spin_unlock_bh(&inetsw6_lock);
602
603		synchronize_net();
604	}
605}
606
607EXPORT_SYMBOL(inet6_unregister_protosw);
608
609int inet6_sk_rebuild_header(struct sock *sk)
610{
611	int err;
612	struct dst_entry *dst;
613	struct ipv6_pinfo *np = inet6_sk(sk);
614
615	dst = __sk_dst_check(sk, np->dst_cookie);
616
617	if (dst == NULL) {
618		struct inet_sock *inet = inet_sk(sk);
619		struct in6_addr *final_p = NULL, final;
620		struct flowi fl;
621
622		memset(&fl, 0, sizeof(fl));
623		fl.proto = sk->sk_protocol;
624		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
625		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
626		fl.fl6_flowlabel = np->flow_label;
627		fl.oif = sk->sk_bound_dev_if;
628		fl.fl_ip_dport = inet->dport;
629		fl.fl_ip_sport = inet->sport;
630		security_sk_classify_flow(sk, &fl);
631
632		if (np->opt && np->opt->srcrt) {
633			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
634			ipv6_addr_copy(&final, &fl.fl6_dst);
635			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
636			final_p = &final;
637		}
638
639		err = ip6_dst_lookup(sk, &dst, &fl);
640		if (err) {
641			sk->sk_route_caps = 0;
642			return err;
643		}
644		if (final_p)
645			ipv6_addr_copy(&fl.fl6_dst, final_p);
646
647		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
648			sk->sk_err_soft = -err;
649			return err;
650		}
651
652		__ip6_dst_store(sk, dst, NULL, NULL);
653	}
654
655	return 0;
656}
657
658EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
659
660int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
661{
662	struct ipv6_pinfo *np = inet6_sk(sk);
663	struct inet6_skb_parm *opt = IP6CB(skb);
664
665	if (np->rxopt.all) {
666		if ((opt->hop && (np->rxopt.bits.hopopts ||
667				  np->rxopt.bits.ohopopts)) ||
668		    ((IPV6_FLOWINFO_MASK &
669		      *(__be32 *)skb_network_header(skb)) &&
670		     np->rxopt.bits.rxflow) ||
671		    (opt->srcrt && (np->rxopt.bits.srcrt ||
672		     np->rxopt.bits.osrcrt)) ||
673		    ((opt->dst1 || opt->dst0) &&
674		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
675			return 1;
676	}
677	return 0;
678}
679
680EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
681
682static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
683						    int proto)
684{
685	struct inet6_protocol *ops = NULL;
686
687	for (;;) {
688		struct ipv6_opt_hdr *opth;
689		int len;
690
691		if (proto != NEXTHDR_HOP) {
692			ops = rcu_dereference(inet6_protos[proto]);
693
694			if (unlikely(!ops))
695				break;
696
697			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
698				break;
699		}
700
701		if (unlikely(!pskb_may_pull(skb, 8)))
702			break;
703
704		opth = (void *)skb->data;
705		len = ipv6_optlen(opth);
706
707		if (unlikely(!pskb_may_pull(skb, len)))
708			break;
709
710		proto = opth->nexthdr;
711		__skb_pull(skb, len);
712	}
713
714	return ops;
715}
716
717static int ipv6_gso_send_check(struct sk_buff *skb)
718{
719	struct ipv6hdr *ipv6h;
720	struct inet6_protocol *ops;
721	int err = -EINVAL;
722
723	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
724		goto out;
725
726	ipv6h = ipv6_hdr(skb);
727	__skb_pull(skb, sizeof(*ipv6h));
728	err = -EPROTONOSUPPORT;
729
730	rcu_read_lock();
731	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
732	if (likely(ops && ops->gso_send_check)) {
733		skb_reset_transport_header(skb);
734		err = ops->gso_send_check(skb);
735	}
736	rcu_read_unlock();
737
738out:
739	return err;
740}
741
742static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
743{
744	struct sk_buff *segs = ERR_PTR(-EINVAL);
745	struct ipv6hdr *ipv6h;
746	struct inet6_protocol *ops;
747
748	if (!(features & NETIF_F_V6_CSUM))
749		features &= ~NETIF_F_SG;
750
751	if (unlikely(skb_shinfo(skb)->gso_type &
752		     ~(SKB_GSO_UDP |
753		       SKB_GSO_DODGY |
754		       SKB_GSO_TCP_ECN |
755		       SKB_GSO_TCPV6 |
756		       0)))
757		goto out;
758
759	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
760		goto out;
761
762	ipv6h = ipv6_hdr(skb);
763	__skb_pull(skb, sizeof(*ipv6h));
764	segs = ERR_PTR(-EPROTONOSUPPORT);
765
766	rcu_read_lock();
767	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
768	if (likely(ops && ops->gso_segment)) {
769		skb_reset_transport_header(skb);
770		segs = ops->gso_segment(skb, features);
771	}
772	rcu_read_unlock();
773
774	if (unlikely(IS_ERR(segs)))
775		goto out;
776
777	for (skb = segs; skb; skb = skb->next) {
778		ipv6h = ipv6_hdr(skb);
779		ipv6h->payload_len = htons(skb->len - skb->mac_len -
780					   sizeof(*ipv6h));
781	}
782
783out:
784	return segs;
785}
786
787static struct packet_type ipv6_packet_type = {
788	.type = __constant_htons(ETH_P_IPV6),
789	.func = ipv6_rcv,
790	.gso_send_check = ipv6_gso_send_check,
791	.gso_segment = ipv6_gso_segment,
792};
793
794static int __init ipv6_packet_init(void)
795{
796	dev_add_pack(&ipv6_packet_type);
797	return 0;
798}
799
800static void ipv6_packet_cleanup(void)
801{
802	dev_remove_pack(&ipv6_packet_type);
803}
804
805static int __init init_ipv6_mibs(void)
806{
807	if (snmp_mib_init((void **)ipv6_statistics,
808			  sizeof(struct ipstats_mib)) < 0)
809		goto err_ip_mib;
810	if (snmp_mib_init((void **)icmpv6_statistics,
811			  sizeof(struct icmpv6_mib)) < 0)
812		goto err_icmp_mib;
813	if (snmp_mib_init((void **)icmpv6msg_statistics,
814			  sizeof(struct icmpv6msg_mib)) < 0)
815		goto err_icmpmsg_mib;
816	if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
817		goto err_udp_mib;
818#ifdef CONFIG_IP_UDPLITE
819	if (snmp_mib_init((void **)udplite_stats_in6,
820			  sizeof (struct udp_mib)) < 0)
821		goto err_udplite_mib;
822#endif
823	return 0;
824
825#ifdef CONFIG_IP_UDPLITE
826err_udplite_mib:
827#endif
828	snmp_mib_free((void **)udp_stats_in6);
829err_udp_mib:
830	snmp_mib_free((void **)icmpv6msg_statistics);
831err_icmpmsg_mib:
832	snmp_mib_free((void **)icmpv6_statistics);
833err_icmp_mib:
834	snmp_mib_free((void **)ipv6_statistics);
835err_ip_mib:
836	return -ENOMEM;
837
838}
839
840static void cleanup_ipv6_mibs(void)
841{
842	snmp_mib_free((void **)ipv6_statistics);
843	snmp_mib_free((void **)icmpv6_statistics);
844	snmp_mib_free((void **)icmpv6msg_statistics);
845	snmp_mib_free((void **)udp_stats_in6);
846#ifdef CONFIG_IP_UDPLITE
847	snmp_mib_free((void **)udplite_stats_in6);
848#endif
849}
850
851static int inet6_net_init(struct net *net)
852{
853	net->ipv6.sysctl.bindv6only = 0;
854	net->ipv6.sysctl.flush_delay = 0;
855	net->ipv6.sysctl.ip6_rt_max_size = 4096;
856	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
857	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
858	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
859	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
860	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
861	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
862	net->ipv6.sysctl.icmpv6_time = 1*HZ;
863
864	return 0;
865}
866
867static void inet6_net_exit(struct net *net)
868{
869	return;
870}
871
872static struct pernet_operations inet6_net_ops = {
873	.init = inet6_net_init,
874	.exit = inet6_net_exit,
875};
876
877static int __init inet6_init(void)
878{
879	struct sk_buff *dummy_skb;
880	struct list_head *r;
881	int err;
882
883	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
884
885	err = proto_register(&tcpv6_prot, 1);
886	if (err)
887		goto out;
888
889	err = proto_register(&udpv6_prot, 1);
890	if (err)
891		goto out_unregister_tcp_proto;
892
893#ifdef CONFIG_IP_UDPLITE
894	err = proto_register(&udplitev6_prot, 1);
895	if (err)
896		goto out_unregister_udp_proto;
897#endif
898
899	err = proto_register(&rawv6_prot, 1);
900	if (err)
901		goto out_unregister_udplite_proto;
902
903
904	/* Register the socket-side information for inet6_create.  */
905	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
906		INIT_LIST_HEAD(r);
907
908	/* We MUST register RAW sockets before we create the ICMP6,
909	 * IGMP6, or NDISC control sockets.
910	 */
911	err = rawv6_init();
912	if (err)
913		goto out_unregister_raw_proto;
914
915	/* Register the family here so that the init calls below will
916	 * be able to create sockets. (?? is this dangerous ??)
917	 */
918	err = sock_register(&inet6_family_ops);
919	if (err)
920		goto out_sock_register_fail;
921
922	/* Initialise ipv6 mibs */
923	err = init_ipv6_mibs();
924	if (err)
925		goto out_unregister_sock;
926
927	/*
928	 *	ipngwg API draft makes clear that the correct semantics
929	 *	for TCP and UDP is to consider one TCP and UDP instance
930	 *	in a host availiable by both INET and INET6 APIs and
931	 *	able to communicate via both network protocols.
932	 */
933
934	err = register_pernet_subsys(&inet6_net_ops);
935	if (err)
936		goto register_pernet_fail;
937	err = icmpv6_init();
938	if (err)
939		goto icmp_fail;
940	err = ndisc_init();
941	if (err)
942		goto ndisc_fail;
943	err = igmp6_init();
944	if (err)
945		goto igmp_fail;
946	err = ipv6_netfilter_init();
947	if (err)
948		goto netfilter_fail;
949	/* Create /proc/foo6 entries. */
950#ifdef CONFIG_PROC_FS
951	err = -ENOMEM;
952	if (raw6_proc_init())
953		goto proc_raw6_fail;
954	if (tcp6_proc_init())
955		goto proc_tcp6_fail;
956	if (udp6_proc_init())
957		goto proc_udp6_fail;
958	if (udplite6_proc_init())
959		goto proc_udplite6_fail;
960	if (ipv6_misc_proc_init())
961		goto proc_misc6_fail;
962
963	if (ac6_proc_init())
964		goto proc_anycast6_fail;
965	if (if6_proc_init())
966		goto proc_if6_fail;
967#endif
968	err = ip6_route_init();
969	if (err)
970		goto ip6_route_fail;
971	err = ip6_flowlabel_init();
972	if (err)
973		goto ip6_flowlabel_fail;
974	err = addrconf_init();
975	if (err)
976		goto addrconf_fail;
977
978	/* Init v6 extension headers. */
979	err = ipv6_exthdrs_init();
980	if (err)
981		goto ipv6_exthdrs_fail;
982
983	err = ipv6_frag_init();
984	if (err)
985		goto ipv6_frag_fail;
986
987	/* Init v6 transport protocols. */
988	err = udpv6_init();
989	if (err)
990		goto udpv6_fail;
991
992	err = udplitev6_init();
993	if (err)
994		goto udplitev6_fail;
995
996	err = tcpv6_init();
997	if (err)
998		goto tcpv6_fail;
999
1000	err = ipv6_packet_init();
1001	if (err)
1002		goto ipv6_packet_fail;
1003
1004#ifdef CONFIG_SYSCTL
1005	err = ipv6_sysctl_register();
1006	if (err)
1007		goto sysctl_fail;
1008#endif
1009out:
1010	return err;
1011
1012#ifdef CONFIG_SYSCTL
1013sysctl_fail:
1014	ipv6_packet_cleanup();
1015#endif
1016ipv6_packet_fail:
1017	tcpv6_exit();
1018tcpv6_fail:
1019	udplitev6_exit();
1020udplitev6_fail:
1021	udpv6_exit();
1022udpv6_fail:
1023	ipv6_frag_exit();
1024ipv6_frag_fail:
1025	ipv6_exthdrs_exit();
1026ipv6_exthdrs_fail:
1027	addrconf_cleanup();
1028addrconf_fail:
1029	ip6_flowlabel_cleanup();
1030ip6_flowlabel_fail:
1031	ip6_route_cleanup();
1032ip6_route_fail:
1033#ifdef CONFIG_PROC_FS
1034	if6_proc_exit();
1035proc_if6_fail:
1036	ac6_proc_exit();
1037proc_anycast6_fail:
1038	ipv6_misc_proc_exit();
1039proc_misc6_fail:
1040	udplite6_proc_exit();
1041proc_udplite6_fail:
1042	udp6_proc_exit();
1043proc_udp6_fail:
1044	tcp6_proc_exit();
1045proc_tcp6_fail:
1046	raw6_proc_exit();
1047proc_raw6_fail:
1048#endif
1049	ipv6_netfilter_fini();
1050netfilter_fail:
1051	igmp6_cleanup();
1052igmp_fail:
1053	ndisc_cleanup();
1054ndisc_fail:
1055	icmpv6_cleanup();
1056icmp_fail:
1057	unregister_pernet_subsys(&inet6_net_ops);
1058register_pernet_fail:
1059	cleanup_ipv6_mibs();
1060out_unregister_sock:
1061	sock_unregister(PF_INET6);
1062	rtnl_unregister_all(PF_INET6);
1063out_sock_register_fail:
1064	rawv6_exit();
1065out_unregister_raw_proto:
1066	proto_unregister(&rawv6_prot);
1067out_unregister_udplite_proto:
1068#ifdef CONFIG_IP_UDPLITE
1069	proto_unregister(&udplitev6_prot);
1070out_unregister_udp_proto:
1071#endif
1072	proto_unregister(&udpv6_prot);
1073out_unregister_tcp_proto:
1074	proto_unregister(&tcpv6_prot);
1075	goto out;
1076}
1077module_init(inet6_init);
1078
1079static void __exit inet6_exit(void)
1080{
1081	/* First of all disallow new sockets creation. */
1082	sock_unregister(PF_INET6);
1083	/* Disallow any further netlink messages */
1084	rtnl_unregister_all(PF_INET6);
1085
1086#ifdef CONFIG_SYSCTL
1087	ipv6_sysctl_unregister();
1088#endif
1089	udpv6_exit();
1090#ifdef CONFIG_IP_UDPLITE
1091	udplitev6_exit();
1092#endif
1093	tcpv6_exit();
1094
1095	/* Cleanup code parts. */
1096	ipv6_packet_cleanup();
1097	ipv6_frag_exit();
1098	ipv6_exthdrs_exit();
1099	addrconf_cleanup();
1100	ip6_flowlabel_cleanup();
1101	ip6_route_cleanup();
1102#ifdef CONFIG_PROC_FS
1103
1104	/* Cleanup code parts. */
1105	if6_proc_exit();
1106	ac6_proc_exit();
1107	ipv6_misc_proc_exit();
1108	udplite6_proc_exit();
1109	udp6_proc_exit();
1110	tcp6_proc_exit();
1111	raw6_proc_exit();
1112#endif
1113	ipv6_netfilter_fini();
1114	igmp6_cleanup();
1115	ndisc_cleanup();
1116	icmpv6_cleanup();
1117	rawv6_exit();
1118
1119	unregister_pernet_subsys(&inet6_net_ops);
1120	cleanup_ipv6_mibs();
1121	proto_unregister(&rawv6_prot);
1122#ifdef CONFIG_IP_UDPLITE
1123	proto_unregister(&udplitev6_prot);
1124#endif
1125	proto_unregister(&udpv6_prot);
1126	proto_unregister(&tcpv6_prot);
1127}
1128module_exit(inet6_exit);
1129
1130MODULE_ALIAS_NETPROTO(PF_INET6);
1131