• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 #define pr_fmt(fmt) "IPv6: " fmt
22 
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/socket.h>
28 #include <linux/in.h>
29 #include <linux/kernel.h>
30 #include <linux/timer.h>
31 #include <linux/string.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/fcntl.h>
35 #include <linux/mm.h>
36 #include <linux/interrupt.h>
37 #include <linux/proc_fs.h>
38 #include <linux/stat.h>
39 #include <linux/init.h>
40 #include <linux/slab.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/ping.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/route.h>
56 #include <net/transp_v6.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #ifdef CONFIG_IPV6_TUNNEL
60 #include <net/ip6_tunnel.h>
61 #endif
62 
63 #include <asm/uaccess.h>
64 #include <linux/mroute6.h>
65 
66 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
67 #include <linux/android_aid.h>
68 
current_has_network(void)69 static inline int current_has_network(void)
70 {
71 	return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
72 }
73 #else
current_has_network(void)74 static inline int current_has_network(void)
75 {
76 	return 1;
77 }
78 #endif
79 
80 MODULE_AUTHOR("Cast of dozens");
81 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
82 MODULE_LICENSE("GPL");
83 
84 /* The inetsw6 table contains everything that inet6_create needs to
85  * build a new socket.
86  */
87 static struct list_head inetsw6[SOCK_MAX];
88 static DEFINE_SPINLOCK(inetsw6_lock);
89 
90 struct ipv6_params ipv6_defaults = {
91 	.disable_ipv6 = 0,
92 	.autoconf = 1,
93 };
94 
95 static int disable_ipv6_mod;
96 
97 module_param_named(disable, disable_ipv6_mod, int, 0444);
98 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
99 
100 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
101 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
102 
103 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
104 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
105 
inet6_sk_generic(struct sock * sk)106 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
107 {
108 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
109 
110 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
111 }
112 
inet6_create(struct net * net,struct socket * sock,int protocol,int kern)113 static int inet6_create(struct net *net, struct socket *sock, int protocol,
114 			int kern)
115 {
116 	struct inet_sock *inet;
117 	struct ipv6_pinfo *np;
118 	struct sock *sk;
119 	struct inet_protosw *answer;
120 	struct proto *answer_prot;
121 	unsigned char answer_flags;
122 	char answer_no_check;
123 	int try_loading_module = 0;
124 	int err;
125 
126 	if (!current_has_network())
127 		return -EACCES;
128 
129 	if (sock->type != SOCK_RAW &&
130 	    sock->type != SOCK_DGRAM &&
131 	    !inet_ehash_secret)
132 		build_ehash_secret();
133 
134 	/* Look for the requested type/protocol pair. */
135 lookup_protocol:
136 	err = -ESOCKTNOSUPPORT;
137 	rcu_read_lock();
138 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
139 
140 		err = 0;
141 		/* Check the non-wild match. */
142 		if (protocol == answer->protocol) {
143 			if (protocol != IPPROTO_IP)
144 				break;
145 		} else {
146 			/* Check for the two wild cases. */
147 			if (IPPROTO_IP == protocol) {
148 				protocol = answer->protocol;
149 				break;
150 			}
151 			if (IPPROTO_IP == answer->protocol)
152 				break;
153 		}
154 		err = -EPROTONOSUPPORT;
155 	}
156 
157 	if (err) {
158 		if (try_loading_module < 2) {
159 			rcu_read_unlock();
160 			/*
161 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
162 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
163 			 */
164 			if (++try_loading_module == 1)
165 				request_module("net-pf-%d-proto-%d-type-%d",
166 						PF_INET6, protocol, sock->type);
167 			/*
168 			 * Fall back to generic, e.g. net-pf-10-proto-132
169 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
170 			 */
171 			else
172 				request_module("net-pf-%d-proto-%d",
173 						PF_INET6, protocol);
174 			goto lookup_protocol;
175 		} else
176 			goto out_rcu_unlock;
177 	}
178 
179 	err = -EPERM;
180 	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
181 		goto out_rcu_unlock;
182 
183 	sock->ops = answer->ops;
184 	answer_prot = answer->prot;
185 	answer_no_check = answer->no_check;
186 	answer_flags = answer->flags;
187 	rcu_read_unlock();
188 
189 	WARN_ON(answer_prot->slab == NULL);
190 
191 	err = -ENOBUFS;
192 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
193 	if (sk == NULL)
194 		goto out;
195 
196 	sock_init_data(sock, sk);
197 
198 	err = 0;
199 	sk->sk_no_check = answer_no_check;
200 	if (INET_PROTOSW_REUSE & answer_flags)
201 		sk->sk_reuse = SK_CAN_REUSE;
202 
203 	inet = inet_sk(sk);
204 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
205 
206 	if (SOCK_RAW == sock->type) {
207 		inet->inet_num = protocol;
208 		if (IPPROTO_RAW == protocol)
209 			inet->hdrincl = 1;
210 	}
211 
212 	sk->sk_destruct		= inet_sock_destruct;
213 	sk->sk_family		= PF_INET6;
214 	sk->sk_protocol		= protocol;
215 
216 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
217 
218 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
219 	np->hop_limit	= -1;
220 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
221 	np->mc_loop	= 1;
222 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
223 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
224 
225 	/* Init the ipv4 part of the socket since we can have sockets
226 	 * using v6 API for ipv4.
227 	 */
228 	inet->uc_ttl	= -1;
229 
230 	inet->mc_loop	= 1;
231 	inet->mc_ttl	= 1;
232 	inet->mc_index	= 0;
233 	inet->mc_list	= NULL;
234 	inet->rcv_tos	= 0;
235 
236 	if (ipv4_config.no_pmtu_disc)
237 		inet->pmtudisc = IP_PMTUDISC_DONT;
238 	else
239 		inet->pmtudisc = IP_PMTUDISC_WANT;
240 	/*
241 	 * Increment only the relevant sk_prot->socks debug field, this changes
242 	 * the previous behaviour of incrementing both the equivalent to
243 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
244 	 *
245 	 * This allows better debug granularity as we'll know exactly how many
246 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
247 	 * transport protocol socks. -acme
248 	 */
249 	sk_refcnt_debug_inc(sk);
250 
251 	if (inet->inet_num) {
252 		/* It assumes that any protocol which allows
253 		 * the user to assign a number at socket
254 		 * creation time automatically shares.
255 		 */
256 		inet->inet_sport = htons(inet->inet_num);
257 		sk->sk_prot->hash(sk);
258 	}
259 	if (sk->sk_prot->init) {
260 		err = sk->sk_prot->init(sk);
261 		if (err) {
262 			sk_common_release(sk);
263 			goto out;
264 		}
265 	}
266 out:
267 	return err;
268 out_rcu_unlock:
269 	rcu_read_unlock();
270 	goto out;
271 }
272 
273 
274 /* bind for INET6 API */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)275 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
276 {
277 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
278 	struct sock *sk = sock->sk;
279 	struct inet_sock *inet = inet_sk(sk);
280 	struct ipv6_pinfo *np = inet6_sk(sk);
281 	struct net *net = sock_net(sk);
282 	__be32 v4addr = 0;
283 	unsigned short snum;
284 	int addr_type = 0;
285 	int err = 0;
286 
287 	/* If the socket has its own bind function then use it. */
288 	if (sk->sk_prot->bind)
289 		return sk->sk_prot->bind(sk, uaddr, addr_len);
290 
291 	if (addr_len < SIN6_LEN_RFC2133)
292 		return -EINVAL;
293 
294 	if (addr->sin6_family != AF_INET6)
295 		return -EAFNOSUPPORT;
296 
297 	addr_type = ipv6_addr_type(&addr->sin6_addr);
298 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
299 		return -EINVAL;
300 
301 	snum = ntohs(addr->sin6_port);
302 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
303 		return -EACCES;
304 
305 	lock_sock(sk);
306 
307 	/* Check these errors (active socket, double bind). */
308 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
309 		err = -EINVAL;
310 		goto out;
311 	}
312 
313 	/* Check if the address belongs to the host. */
314 	if (addr_type == IPV6_ADDR_MAPPED) {
315 		int chk_addr_ret;
316 
317 		/* Binding to v4-mapped address on a v6-only socket
318 		 * makes no sense
319 		 */
320 		if (np->ipv6only) {
321 			err = -EINVAL;
322 			goto out;
323 		}
324 
325 		/* Reproduce AF_INET checks to make the bindings consistent */
326 		v4addr = addr->sin6_addr.s6_addr32[3];
327 		chk_addr_ret = inet_addr_type(net, v4addr);
328 		if (!sysctl_ip_nonlocal_bind &&
329 		    !(inet->freebind || inet->transparent) &&
330 		    v4addr != htonl(INADDR_ANY) &&
331 		    chk_addr_ret != RTN_LOCAL &&
332 		    chk_addr_ret != RTN_MULTICAST &&
333 		    chk_addr_ret != RTN_BROADCAST) {
334 			err = -EADDRNOTAVAIL;
335 			goto out;
336 		}
337 	} else {
338 		if (addr_type != IPV6_ADDR_ANY) {
339 			struct net_device *dev = NULL;
340 
341 			rcu_read_lock();
342 			if (__ipv6_addr_needs_scope_id(addr_type)) {
343 				if (addr_len >= sizeof(struct sockaddr_in6) &&
344 				    addr->sin6_scope_id) {
345 					/* Override any existing binding, if another one
346 					 * is supplied by user.
347 					 */
348 					sk->sk_bound_dev_if = addr->sin6_scope_id;
349 				}
350 
351 				/* Binding to link-local address requires an interface */
352 				if (!sk->sk_bound_dev_if) {
353 					err = -EINVAL;
354 					goto out_unlock;
355 				}
356 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
357 				if (!dev) {
358 					err = -ENODEV;
359 					goto out_unlock;
360 				}
361 			}
362 
363 			/* ipv4 addr of the socket is invalid.  Only the
364 			 * unspecified and mapped address have a v4 equivalent.
365 			 */
366 			v4addr = LOOPBACK4_IPV6;
367 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
368 				if (!(inet->freebind || inet->transparent) &&
369 				    !ipv6_chk_addr(net, &addr->sin6_addr,
370 						   dev, 0)) {
371 					err = -EADDRNOTAVAIL;
372 					goto out_unlock;
373 				}
374 			}
375 			rcu_read_unlock();
376 		}
377 	}
378 
379 	inet->inet_rcv_saddr = v4addr;
380 	inet->inet_saddr = v4addr;
381 
382 	np->rcv_saddr = addr->sin6_addr;
383 
384 	if (!(addr_type & IPV6_ADDR_MULTICAST))
385 		np->saddr = addr->sin6_addr;
386 
387 	/* Make sure we are allowed to bind here. */
388 	if (sk->sk_prot->get_port(sk, snum)) {
389 		inet_reset_saddr(sk);
390 		err = -EADDRINUSE;
391 		goto out;
392 	}
393 
394 	if (addr_type != IPV6_ADDR_ANY) {
395 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
396 		if (addr_type != IPV6_ADDR_MAPPED)
397 			np->ipv6only = 1;
398 	}
399 	if (snum)
400 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
401 	inet->inet_sport = htons(inet->inet_num);
402 	inet->inet_dport = 0;
403 	inet->inet_daddr = 0;
404 out:
405 	release_sock(sk);
406 	return err;
407 out_unlock:
408 	rcu_read_unlock();
409 	goto out;
410 }
411 EXPORT_SYMBOL(inet6_bind);
412 
inet6_release(struct socket * sock)413 int inet6_release(struct socket *sock)
414 {
415 	struct sock *sk = sock->sk;
416 
417 	if (sk == NULL)
418 		return -EINVAL;
419 
420 	/* Free mc lists */
421 	ipv6_sock_mc_close(sk);
422 
423 	/* Free ac lists */
424 	ipv6_sock_ac_close(sk);
425 
426 	return inet_release(sock);
427 }
428 EXPORT_SYMBOL(inet6_release);
429 
inet6_destroy_sock(struct sock * sk)430 void inet6_destroy_sock(struct sock *sk)
431 {
432 	struct ipv6_pinfo *np = inet6_sk(sk);
433 	struct sk_buff *skb;
434 	struct ipv6_txoptions *opt;
435 
436 	/* Release rx options */
437 
438 	skb = xchg(&np->pktoptions, NULL);
439 	if (skb != NULL)
440 		kfree_skb(skb);
441 
442 	skb = xchg(&np->rxpmtu, NULL);
443 	if (skb != NULL)
444 		kfree_skb(skb);
445 
446 	/* Free flowlabels */
447 	fl6_free_socklist(sk);
448 
449 	/* Free tx options */
450 
451 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
452 	if (opt) {
453 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
454 		txopt_put(opt);
455 	}
456 }
457 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
458 
459 /*
460  *	This does both peername and sockname.
461  */
462 
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int * uaddr_len,int peer)463 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
464 		 int *uaddr_len, int peer)
465 {
466 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
467 	struct sock *sk = sock->sk;
468 	struct inet_sock *inet = inet_sk(sk);
469 	struct ipv6_pinfo *np = inet6_sk(sk);
470 
471 	sin->sin6_family = AF_INET6;
472 	sin->sin6_flowinfo = 0;
473 	sin->sin6_scope_id = 0;
474 	if (peer) {
475 		if (!inet->inet_dport)
476 			return -ENOTCONN;
477 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
478 		    peer == 1)
479 			return -ENOTCONN;
480 		sin->sin6_port = inet->inet_dport;
481 		sin->sin6_addr = np->daddr;
482 		if (np->sndflow)
483 			sin->sin6_flowinfo = np->flow_label;
484 	} else {
485 		if (ipv6_addr_any(&np->rcv_saddr))
486 			sin->sin6_addr = np->saddr;
487 		else
488 			sin->sin6_addr = np->rcv_saddr;
489 
490 		sin->sin6_port = inet->inet_sport;
491 	}
492 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
493 						 sk->sk_bound_dev_if);
494 	*uaddr_len = sizeof(*sin);
495 	return 0;
496 }
497 EXPORT_SYMBOL(inet6_getname);
498 
inet6_killaddr_ioctl(struct net * net,void __user * arg)499 int inet6_killaddr_ioctl(struct net *net, void __user *arg) {
500 	struct in6_ifreq ireq;
501 	struct sockaddr_in6 sin6;
502 
503 	if (!capable(CAP_NET_ADMIN))
504 		return -EACCES;
505 
506 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
507 		return -EFAULT;
508 
509 	sin6.sin6_family = AF_INET6;
510 	sin6.sin6_addr = ireq.ifr6_addr;
511 	return tcp_nuke_addr(net, (struct sockaddr *) &sin6);
512 }
513 
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)514 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
515 {
516 	struct sock *sk = sock->sk;
517 	struct net *net = sock_net(sk);
518 
519 	switch (cmd) {
520 	case SIOCGSTAMP:
521 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
522 
523 	case SIOCGSTAMPNS:
524 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
525 
526 	case SIOCADDRT:
527 	case SIOCDELRT:
528 
529 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
530 
531 	case SIOCSIFADDR:
532 		return addrconf_add_ifaddr(net, (void __user *) arg);
533 	case SIOCDIFADDR:
534 		return addrconf_del_ifaddr(net, (void __user *) arg);
535 	case SIOCSIFDSTADDR:
536 		return addrconf_set_dstaddr(net, (void __user *) arg);
537 	case SIOCKILLADDR:
538 		return inet6_killaddr_ioctl(net, (void __user *) arg);
539 	default:
540 		if (!sk->sk_prot->ioctl)
541 			return -ENOIOCTLCMD;
542 		return sk->sk_prot->ioctl(sk, cmd, arg);
543 	}
544 	/*NOTREACHED*/
545 	return 0;
546 }
547 EXPORT_SYMBOL(inet6_ioctl);
548 
549 const struct proto_ops inet6_stream_ops = {
550 	.family		   = PF_INET6,
551 	.owner		   = THIS_MODULE,
552 	.release	   = inet6_release,
553 	.bind		   = inet6_bind,
554 	.connect	   = inet_stream_connect,	/* ok		*/
555 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
556 	.accept		   = inet_accept,		/* ok		*/
557 	.getname	   = inet6_getname,
558 	.poll		   = tcp_poll,			/* ok		*/
559 	.ioctl		   = inet6_ioctl,		/* must change  */
560 	.listen		   = inet_listen,		/* ok		*/
561 	.shutdown	   = inet_shutdown,		/* ok		*/
562 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
563 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
564 	.sendmsg	   = inet_sendmsg,		/* ok		*/
565 	.recvmsg	   = inet_recvmsg,		/* ok		*/
566 	.mmap		   = sock_no_mmap,
567 	.sendpage	   = inet_sendpage,
568 	.splice_read	   = tcp_splice_read,
569 #ifdef CONFIG_COMPAT
570 	.compat_setsockopt = compat_sock_common_setsockopt,
571 	.compat_getsockopt = compat_sock_common_getsockopt,
572 #endif
573 };
574 
575 const struct proto_ops inet6_dgram_ops = {
576 	.family		   = PF_INET6,
577 	.owner		   = THIS_MODULE,
578 	.release	   = inet6_release,
579 	.bind		   = inet6_bind,
580 	.connect	   = inet_dgram_connect,	/* ok		*/
581 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
582 	.accept		   = sock_no_accept,		/* a do nothing	*/
583 	.getname	   = inet6_getname,
584 	.poll		   = udp_poll,			/* ok		*/
585 	.ioctl		   = inet6_ioctl,		/* must change  */
586 	.listen		   = sock_no_listen,		/* ok		*/
587 	.shutdown	   = inet_shutdown,		/* ok		*/
588 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
589 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
590 	.sendmsg	   = inet_sendmsg,		/* ok		*/
591 	.recvmsg	   = inet_recvmsg,		/* ok		*/
592 	.mmap		   = sock_no_mmap,
593 	.sendpage	   = sock_no_sendpage,
594 #ifdef CONFIG_COMPAT
595 	.compat_setsockopt = compat_sock_common_setsockopt,
596 	.compat_getsockopt = compat_sock_common_getsockopt,
597 #endif
598 };
599 
600 static const struct net_proto_family inet6_family_ops = {
601 	.family = PF_INET6,
602 	.create = inet6_create,
603 	.owner	= THIS_MODULE,
604 };
605 
inet6_register_protosw(struct inet_protosw * p)606 int inet6_register_protosw(struct inet_protosw *p)
607 {
608 	struct list_head *lh;
609 	struct inet_protosw *answer;
610 	struct list_head *last_perm;
611 	int protocol = p->protocol;
612 	int ret;
613 
614 	spin_lock_bh(&inetsw6_lock);
615 
616 	ret = -EINVAL;
617 	if (p->type >= SOCK_MAX)
618 		goto out_illegal;
619 
620 	/* If we are trying to override a permanent protocol, bail. */
621 	answer = NULL;
622 	ret = -EPERM;
623 	last_perm = &inetsw6[p->type];
624 	list_for_each(lh, &inetsw6[p->type]) {
625 		answer = list_entry(lh, struct inet_protosw, list);
626 
627 		/* Check only the non-wild match. */
628 		if (INET_PROTOSW_PERMANENT & answer->flags) {
629 			if (protocol == answer->protocol)
630 				break;
631 			last_perm = lh;
632 		}
633 
634 		answer = NULL;
635 	}
636 	if (answer)
637 		goto out_permanent;
638 
639 	/* Add the new entry after the last permanent entry if any, so that
640 	 * the new entry does not override a permanent entry when matched with
641 	 * a wild-card protocol. But it is allowed to override any existing
642 	 * non-permanent entry.  This means that when we remove this entry, the
643 	 * system automatically returns to the old behavior.
644 	 */
645 	list_add_rcu(&p->list, last_perm);
646 	ret = 0;
647 out:
648 	spin_unlock_bh(&inetsw6_lock);
649 	return ret;
650 
651 out_permanent:
652 	pr_err("Attempt to override permanent protocol %d\n", protocol);
653 	goto out;
654 
655 out_illegal:
656 	pr_err("Ignoring attempt to register invalid socket type %d\n",
657 	       p->type);
658 	goto out;
659 }
660 EXPORT_SYMBOL(inet6_register_protosw);
661 
662 void
inet6_unregister_protosw(struct inet_protosw * p)663 inet6_unregister_protosw(struct inet_protosw *p)
664 {
665 	if (INET_PROTOSW_PERMANENT & p->flags) {
666 		pr_err("Attempt to unregister permanent protocol %d\n",
667 		       p->protocol);
668 	} else {
669 		spin_lock_bh(&inetsw6_lock);
670 		list_del_rcu(&p->list);
671 		spin_unlock_bh(&inetsw6_lock);
672 
673 		synchronize_net();
674 	}
675 }
676 EXPORT_SYMBOL(inet6_unregister_protosw);
677 
inet6_sk_rebuild_header(struct sock * sk)678 int inet6_sk_rebuild_header(struct sock *sk)
679 {
680 	struct ipv6_pinfo *np = inet6_sk(sk);
681 	struct dst_entry *dst;
682 
683 	dst = __sk_dst_check(sk, np->dst_cookie);
684 
685 	if (dst == NULL) {
686 		struct inet_sock *inet = inet_sk(sk);
687 		struct in6_addr *final_p, final;
688 		struct flowi6 fl6;
689 
690 		memset(&fl6, 0, sizeof(fl6));
691 		fl6.flowi6_proto = sk->sk_protocol;
692 		fl6.daddr = np->daddr;
693 		fl6.saddr = np->saddr;
694 		fl6.flowlabel = np->flow_label;
695 		fl6.flowi6_oif = sk->sk_bound_dev_if;
696 		fl6.flowi6_mark = sk->sk_mark;
697 		fl6.fl6_dport = inet->inet_dport;
698 		fl6.fl6_sport = inet->inet_sport;
699 		fl6.flowi6_uid = sk->sk_uid;
700 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
701 
702 		rcu_read_lock();
703 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
704 					 &final);
705 		rcu_read_unlock();
706 
707 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
708 		if (IS_ERR(dst)) {
709 			sk->sk_route_caps = 0;
710 			sk->sk_err_soft = -PTR_ERR(dst);
711 			return PTR_ERR(dst);
712 		}
713 
714 		__ip6_dst_store(sk, dst, NULL, NULL);
715 	}
716 
717 	return 0;
718 }
719 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
720 
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb)721 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
722 {
723 	const struct ipv6_pinfo *np = inet6_sk(sk);
724 	const struct inet6_skb_parm *opt = IP6CB(skb);
725 
726 	if (np->rxopt.all) {
727 		if ((opt->hop && (np->rxopt.bits.hopopts ||
728 				  np->rxopt.bits.ohopopts)) ||
729 		    ((IPV6_FLOWINFO_MASK &
730 		      *(__be32 *)skb_network_header(skb)) &&
731 		     np->rxopt.bits.rxflow) ||
732 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
733 		     np->rxopt.bits.osrcrt)) ||
734 		    ((opt->dst1 || opt->dst0) &&
735 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
736 			return true;
737 	}
738 	return false;
739 }
740 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
741 
742 static struct packet_type ipv6_packet_type __read_mostly = {
743 	.type = cpu_to_be16(ETH_P_IPV6),
744 	.func = ipv6_rcv,
745 };
746 
ipv6_packet_init(void)747 static int __init ipv6_packet_init(void)
748 {
749 	dev_add_pack(&ipv6_packet_type);
750 	return 0;
751 }
752 
ipv6_packet_cleanup(void)753 static void ipv6_packet_cleanup(void)
754 {
755 	dev_remove_pack(&ipv6_packet_type);
756 }
757 
ipv6_init_mibs(struct net * net)758 static int __net_init ipv6_init_mibs(struct net *net)
759 {
760 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
761 			  sizeof(struct udp_mib),
762 			  __alignof__(struct udp_mib)) < 0)
763 		return -ENOMEM;
764 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
765 			  sizeof(struct udp_mib),
766 			  __alignof__(struct udp_mib)) < 0)
767 		goto err_udplite_mib;
768 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
769 			  sizeof(struct ipstats_mib),
770 			  __alignof__(struct ipstats_mib)) < 0)
771 		goto err_ip_mib;
772 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
773 			  sizeof(struct icmpv6_mib),
774 			  __alignof__(struct icmpv6_mib)) < 0)
775 		goto err_icmp_mib;
776 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
777 						GFP_KERNEL);
778 	if (!net->mib.icmpv6msg_statistics)
779 		goto err_icmpmsg_mib;
780 	return 0;
781 
782 err_icmpmsg_mib:
783 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
784 err_icmp_mib:
785 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
786 err_ip_mib:
787 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
788 err_udplite_mib:
789 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
790 	return -ENOMEM;
791 }
792 
ipv6_cleanup_mibs(struct net * net)793 static void ipv6_cleanup_mibs(struct net *net)
794 {
795 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
796 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
797 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
798 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
799 	kfree(net->mib.icmpv6msg_statistics);
800 }
801 
inet6_net_init(struct net * net)802 static int __net_init inet6_net_init(struct net *net)
803 {
804 	int err = 0;
805 
806 	net->ipv6.sysctl.bindv6only = 0;
807 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
808 
809 	err = ipv6_init_mibs(net);
810 	if (err)
811 		return err;
812 #ifdef CONFIG_PROC_FS
813 	err = udp6_proc_init(net);
814 	if (err)
815 		goto out;
816 	err = tcp6_proc_init(net);
817 	if (err)
818 		goto proc_tcp6_fail;
819 	err = ac6_proc_init(net);
820 	if (err)
821 		goto proc_ac6_fail;
822 #endif
823 	return err;
824 
825 #ifdef CONFIG_PROC_FS
826 proc_ac6_fail:
827 	tcp6_proc_exit(net);
828 proc_tcp6_fail:
829 	udp6_proc_exit(net);
830 out:
831 	ipv6_cleanup_mibs(net);
832 	return err;
833 #endif
834 }
835 
inet6_net_exit(struct net * net)836 static void __net_exit inet6_net_exit(struct net *net)
837 {
838 #ifdef CONFIG_PROC_FS
839 	udp6_proc_exit(net);
840 	tcp6_proc_exit(net);
841 	ac6_proc_exit(net);
842 #endif
843 	ipv6_cleanup_mibs(net);
844 }
845 
846 static struct pernet_operations inet6_net_ops = {
847 	.init = inet6_net_init,
848 	.exit = inet6_net_exit,
849 };
850 
inet6_init(void)851 static int __init inet6_init(void)
852 {
853 	struct list_head *r;
854 	int err = 0;
855 
856 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
857 
858 	/* Register the socket-side information for inet6_create.  */
859 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
860 		INIT_LIST_HEAD(r);
861 
862 	if (disable_ipv6_mod) {
863 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
864 		goto out;
865 	}
866 
867 	err = proto_register(&tcpv6_prot, 1);
868 	if (err)
869 		goto out;
870 
871 	err = proto_register(&udpv6_prot, 1);
872 	if (err)
873 		goto out_unregister_tcp_proto;
874 
875 	err = proto_register(&udplitev6_prot, 1);
876 	if (err)
877 		goto out_unregister_udp_proto;
878 
879 	err = proto_register(&rawv6_prot, 1);
880 	if (err)
881 		goto out_unregister_udplite_proto;
882 
883 	err = proto_register(&pingv6_prot, 1);
884 	if (err)
885 		goto out_unregister_ping_proto;
886 
887 	/* We MUST register RAW sockets before we create the ICMP6,
888 	 * IGMP6, or NDISC control sockets.
889 	 */
890 	err = rawv6_init();
891 	if (err)
892 		goto out_unregister_raw_proto;
893 
894 	/* Register the family here so that the init calls below will
895 	 * be able to create sockets. (?? is this dangerous ??)
896 	 */
897 	err = sock_register(&inet6_family_ops);
898 	if (err)
899 		goto out_sock_register_fail;
900 
901 	tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
902 
903 	/*
904 	 *	ipngwg API draft makes clear that the correct semantics
905 	 *	for TCP and UDP is to consider one TCP and UDP instance
906 	 *	in a host available by both INET and INET6 APIs and
907 	 *	able to communicate via both network protocols.
908 	 */
909 
910 	err = register_pernet_subsys(&inet6_net_ops);
911 	if (err)
912 		goto register_pernet_fail;
913 	err = icmpv6_init();
914 	if (err)
915 		goto icmp_fail;
916 	err = ip6_mr_init();
917 	if (err)
918 		goto ipmr_fail;
919 	err = ndisc_init();
920 	if (err)
921 		goto ndisc_fail;
922 	err = igmp6_init();
923 	if (err)
924 		goto igmp_fail;
925 	err = ipv6_netfilter_init();
926 	if (err)
927 		goto netfilter_fail;
928 	/* Create /proc/foo6 entries. */
929 #ifdef CONFIG_PROC_FS
930 	err = -ENOMEM;
931 	if (raw6_proc_init())
932 		goto proc_raw6_fail;
933 	if (udplite6_proc_init())
934 		goto proc_udplite6_fail;
935 	if (ipv6_misc_proc_init())
936 		goto proc_misc6_fail;
937 	if (if6_proc_init())
938 		goto proc_if6_fail;
939 #endif
940 	err = ip6_route_init();
941 	if (err)
942 		goto ip6_route_fail;
943 	err = ip6_flowlabel_init();
944 	if (err)
945 		goto ip6_flowlabel_fail;
946 	err = addrconf_init();
947 	if (err)
948 		goto addrconf_fail;
949 
950 	/* Init v6 extension headers. */
951 	err = ipv6_exthdrs_init();
952 	if (err)
953 		goto ipv6_exthdrs_fail;
954 
955 	err = ipv6_frag_init();
956 	if (err)
957 		goto ipv6_frag_fail;
958 
959 	/* Init v6 transport protocols. */
960 	err = udpv6_init();
961 	if (err)
962 		goto udpv6_fail;
963 
964 	err = udplitev6_init();
965 	if (err)
966 		goto udplitev6_fail;
967 
968 	err = tcpv6_init();
969 	if (err)
970 		goto tcpv6_fail;
971 
972 	err = ipv6_packet_init();
973 	if (err)
974 		goto ipv6_packet_fail;
975 
976 	err = pingv6_init();
977 	if (err)
978 		goto pingv6_fail;
979 
980 #ifdef CONFIG_SYSCTL
981 	err = ipv6_sysctl_register();
982 	if (err)
983 		goto sysctl_fail;
984 #endif
985 out:
986 	return err;
987 
988 #ifdef CONFIG_SYSCTL
989 sysctl_fail:
990 	ipv6_packet_cleanup();
991 #endif
992 pingv6_fail:
993 	pingv6_exit();
994 ipv6_packet_fail:
995 	tcpv6_exit();
996 tcpv6_fail:
997 	udplitev6_exit();
998 udplitev6_fail:
999 	udpv6_exit();
1000 udpv6_fail:
1001 	ipv6_frag_exit();
1002 ipv6_frag_fail:
1003 	ipv6_exthdrs_exit();
1004 ipv6_exthdrs_fail:
1005 	addrconf_cleanup();
1006 addrconf_fail:
1007 	ip6_flowlabel_cleanup();
1008 ip6_flowlabel_fail:
1009 	ip6_route_cleanup();
1010 ip6_route_fail:
1011 #ifdef CONFIG_PROC_FS
1012 	if6_proc_exit();
1013 proc_if6_fail:
1014 	ipv6_misc_proc_exit();
1015 proc_misc6_fail:
1016 	udplite6_proc_exit();
1017 proc_udplite6_fail:
1018 	raw6_proc_exit();
1019 proc_raw6_fail:
1020 #endif
1021 	ipv6_netfilter_fini();
1022 netfilter_fail:
1023 	igmp6_cleanup();
1024 igmp_fail:
1025 	ndisc_cleanup();
1026 ndisc_fail:
1027 	ip6_mr_cleanup();
1028 ipmr_fail:
1029 	icmpv6_cleanup();
1030 icmp_fail:
1031 	unregister_pernet_subsys(&inet6_net_ops);
1032 register_pernet_fail:
1033 	sock_unregister(PF_INET6);
1034 	rtnl_unregister_all(PF_INET6);
1035 out_sock_register_fail:
1036 	rawv6_exit();
1037 out_unregister_ping_proto:
1038 	proto_unregister(&pingv6_prot);
1039 out_unregister_raw_proto:
1040 	proto_unregister(&rawv6_prot);
1041 out_unregister_udplite_proto:
1042 	proto_unregister(&udplitev6_prot);
1043 out_unregister_udp_proto:
1044 	proto_unregister(&udpv6_prot);
1045 out_unregister_tcp_proto:
1046 	proto_unregister(&tcpv6_prot);
1047 	goto out;
1048 }
1049 module_init(inet6_init);
1050 
inet6_exit(void)1051 static void __exit inet6_exit(void)
1052 {
1053 	if (disable_ipv6_mod)
1054 		return;
1055 
1056 	/* First of all disallow new sockets creation. */
1057 	sock_unregister(PF_INET6);
1058 	/* Disallow any further netlink messages */
1059 	rtnl_unregister_all(PF_INET6);
1060 
1061 	udpv6_exit();
1062 	udplitev6_exit();
1063 	tcpv6_exit();
1064 
1065 	/* Cleanup code parts. */
1066 	ipv6_packet_cleanup();
1067 	ipv6_frag_exit();
1068 	ipv6_exthdrs_exit();
1069 	addrconf_cleanup();
1070 	ip6_flowlabel_cleanup();
1071 	ip6_route_cleanup();
1072 #ifdef CONFIG_PROC_FS
1073 
1074 	/* Cleanup code parts. */
1075 	if6_proc_exit();
1076 	ipv6_misc_proc_exit();
1077 	udplite6_proc_exit();
1078 	raw6_proc_exit();
1079 #endif
1080 	ipv6_netfilter_fini();
1081 	igmp6_cleanup();
1082 	ndisc_cleanup();
1083 	ip6_mr_cleanup();
1084 	icmpv6_cleanup();
1085 	rawv6_exit();
1086 
1087 	unregister_pernet_subsys(&inet6_net_ops);
1088 	proto_unregister(&rawv6_prot);
1089 	proto_unregister(&udplitev6_prot);
1090 	proto_unregister(&udpv6_prot);
1091 	proto_unregister(&tcpv6_prot);
1092 
1093 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1094 }
1095 module_exit(inet6_exit);
1096 
1097 MODULE_ALIAS_NETPROTO(PF_INET6);
1098