• 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 #include <net/ndisc.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 #include <net/calipso.h>
64 #include <net/seg6.h>
65 
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
68 
69 #include "ip6_offload.h"
70 
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
74 
75 /* The inetsw6 table contains everything that inet6_create needs to
76  * build a new socket.
77  */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80 
81 struct ipv6_params ipv6_defaults = {
82 	.disable_ipv6 = 0,
83 	.autoconf = 1,
84 };
85 
86 static int disable_ipv6_mod;
87 
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90 
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93 
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96 
ipv6_mod_enabled(void)97 bool ipv6_mod_enabled(void)
98 {
99 	return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102 
inet6_sk_generic(struct sock * sk)103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104 {
105 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106 
107 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108 }
109 
inet6_create(struct net * net,struct socket * sock,int protocol,int kern)110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 			int kern)
112 {
113 	struct inet_sock *inet;
114 	struct ipv6_pinfo *np;
115 	struct sock *sk;
116 	struct inet_protosw *answer;
117 	struct proto *answer_prot;
118 	unsigned char answer_flags;
119 	int try_loading_module = 0;
120 	int err;
121 
122 	if (protocol < 0 || protocol >= IPPROTO_MAX)
123 		return -EINVAL;
124 
125 	/* Look for the requested type/protocol pair. */
126 lookup_protocol:
127 	err = -ESOCKTNOSUPPORT;
128 	rcu_read_lock();
129 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
130 
131 		err = 0;
132 		/* Check the non-wild match. */
133 		if (protocol == answer->protocol) {
134 			if (protocol != IPPROTO_IP)
135 				break;
136 		} else {
137 			/* Check for the two wild cases. */
138 			if (IPPROTO_IP == protocol) {
139 				protocol = answer->protocol;
140 				break;
141 			}
142 			if (IPPROTO_IP == answer->protocol)
143 				break;
144 		}
145 		err = -EPROTONOSUPPORT;
146 	}
147 
148 	if (err) {
149 		if (try_loading_module < 2) {
150 			rcu_read_unlock();
151 			/*
152 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
154 			 */
155 			if (++try_loading_module == 1)
156 				request_module("net-pf-%d-proto-%d-type-%d",
157 						PF_INET6, protocol, sock->type);
158 			/*
159 			 * Fall back to generic, e.g. net-pf-10-proto-132
160 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
161 			 */
162 			else
163 				request_module("net-pf-%d-proto-%d",
164 						PF_INET6, protocol);
165 			goto lookup_protocol;
166 		} else
167 			goto out_rcu_unlock;
168 	}
169 
170 	err = -EPERM;
171 	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
172 		goto out_rcu_unlock;
173 
174 	sock->ops = answer->ops;
175 	answer_prot = answer->prot;
176 	answer_flags = answer->flags;
177 	rcu_read_unlock();
178 
179 	WARN_ON(!answer_prot->slab);
180 
181 	err = -ENOBUFS;
182 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
183 	if (!sk)
184 		goto out;
185 
186 	sock_init_data(sock, sk);
187 
188 	err = 0;
189 	if (INET_PROTOSW_REUSE & answer_flags)
190 		sk->sk_reuse = SK_CAN_REUSE;
191 
192 	inet = inet_sk(sk);
193 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
194 
195 	if (SOCK_RAW == sock->type) {
196 		inet->inet_num = protocol;
197 		if (IPPROTO_RAW == protocol)
198 			inet->hdrincl = 1;
199 	}
200 
201 	sk->sk_destruct		= inet_sock_destruct;
202 	sk->sk_family		= PF_INET6;
203 	sk->sk_protocol		= protocol;
204 
205 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
206 
207 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
208 	np->hop_limit	= -1;
209 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
210 	np->mc_loop	= 1;
211 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
212 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect;
213 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
214 
215 	/* Init the ipv4 part of the socket since we can have sockets
216 	 * using v6 API for ipv4.
217 	 */
218 	inet->uc_ttl	= -1;
219 
220 	inet->mc_loop	= 1;
221 	inet->mc_ttl	= 1;
222 	inet->mc_index	= 0;
223 	inet->mc_list	= NULL;
224 	inet->rcv_tos	= 0;
225 
226 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
227 		inet->pmtudisc = IP_PMTUDISC_DONT;
228 	else
229 		inet->pmtudisc = IP_PMTUDISC_WANT;
230 	/*
231 	 * Increment only the relevant sk_prot->socks debug field, this changes
232 	 * the previous behaviour of incrementing both the equivalent to
233 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
234 	 *
235 	 * This allows better debug granularity as we'll know exactly how many
236 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
237 	 * transport protocol socks. -acme
238 	 */
239 	sk_refcnt_debug_inc(sk);
240 
241 	if (inet->inet_num) {
242 		/* It assumes that any protocol which allows
243 		 * the user to assign a number at socket
244 		 * creation time automatically shares.
245 		 */
246 		inet->inet_sport = htons(inet->inet_num);
247 		err = sk->sk_prot->hash(sk);
248 		if (err) {
249 			sk_common_release(sk);
250 			goto out;
251 		}
252 	}
253 	if (sk->sk_prot->init) {
254 		err = sk->sk_prot->init(sk);
255 		if (err) {
256 			sk_common_release(sk);
257 			goto out;
258 		}
259 	}
260 
261 	if (!kern) {
262 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
263 		if (err) {
264 			sk_common_release(sk);
265 			goto out;
266 		}
267 	}
268 out:
269 	return err;
270 out_rcu_unlock:
271 	rcu_read_unlock();
272 	goto out;
273 }
274 
275 
276 /* bind for INET6 API */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)277 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
278 {
279 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
280 	struct sock *sk = sock->sk;
281 	struct inet_sock *inet = inet_sk(sk);
282 	struct ipv6_pinfo *np = inet6_sk(sk);
283 	struct net *net = sock_net(sk);
284 	__be32 v4addr = 0;
285 	unsigned short snum;
286 	bool saved_ipv6only;
287 	int addr_type = 0;
288 	int err = 0;
289 
290 	/* If the socket has its own bind function then use it. */
291 	if (sk->sk_prot->bind)
292 		return sk->sk_prot->bind(sk, uaddr, addr_len);
293 
294 	if (addr_len < SIN6_LEN_RFC2133)
295 		return -EINVAL;
296 
297 	if (addr->sin6_family != AF_INET6)
298 		return -EAFNOSUPPORT;
299 
300 	addr_type = ipv6_addr_type(&addr->sin6_addr);
301 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
302 		return -EINVAL;
303 
304 	snum = ntohs(addr->sin6_port);
305 	if (snum && snum < inet_prot_sock(net) &&
306 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
307 		return -EACCES;
308 
309 	lock_sock(sk);
310 
311 	/* Check these errors (active socket, double bind). */
312 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
313 		err = -EINVAL;
314 		goto out;
315 	}
316 
317 	/* Check if the address belongs to the host. */
318 	if (addr_type == IPV6_ADDR_MAPPED) {
319 		struct net_device *dev = NULL;
320 		int chk_addr_ret;
321 
322 		/* Binding to v4-mapped address on a v6-only socket
323 		 * makes no sense
324 		 */
325 		if (sk->sk_ipv6only) {
326 			err = -EINVAL;
327 			goto out;
328 		}
329 
330 		rcu_read_lock();
331 		if (sk->sk_bound_dev_if) {
332 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
333 			if (!dev) {
334 				err = -ENODEV;
335 				goto out_unlock;
336 			}
337 		}
338 
339 		/* Reproduce AF_INET checks to make the bindings consistent */
340 		v4addr = addr->sin6_addr.s6_addr32[3];
341 		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
342 		rcu_read_unlock();
343 
344 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
345 		    !(inet->freebind || inet->transparent) &&
346 		    v4addr != htonl(INADDR_ANY) &&
347 		    chk_addr_ret != RTN_LOCAL &&
348 		    chk_addr_ret != RTN_MULTICAST &&
349 		    chk_addr_ret != RTN_BROADCAST) {
350 			err = -EADDRNOTAVAIL;
351 			goto out;
352 		}
353 	} else {
354 		if (addr_type != IPV6_ADDR_ANY) {
355 			struct net_device *dev = NULL;
356 
357 			rcu_read_lock();
358 			if (__ipv6_addr_needs_scope_id(addr_type)) {
359 				if (addr_len >= sizeof(struct sockaddr_in6) &&
360 				    addr->sin6_scope_id) {
361 					/* Override any existing binding, if another one
362 					 * is supplied by user.
363 					 */
364 					sk->sk_bound_dev_if = addr->sin6_scope_id;
365 				}
366 
367 				/* Binding to link-local address requires an interface */
368 				if (!sk->sk_bound_dev_if) {
369 					err = -EINVAL;
370 					goto out_unlock;
371 				}
372 			}
373 
374 			if (sk->sk_bound_dev_if) {
375 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
376 				if (!dev) {
377 					err = -ENODEV;
378 					goto out_unlock;
379 				}
380 			}
381 
382 			/* ipv4 addr of the socket is invalid.  Only the
383 			 * unspecified and mapped address have a v4 equivalent.
384 			 */
385 			v4addr = LOOPBACK4_IPV6;
386 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
387 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
388 				    !(inet->freebind || inet->transparent) &&
389 				    !ipv6_chk_addr(net, &addr->sin6_addr,
390 						   dev, 0)) {
391 					err = -EADDRNOTAVAIL;
392 					goto out_unlock;
393 				}
394 			}
395 			rcu_read_unlock();
396 		}
397 	}
398 
399 	inet->inet_rcv_saddr = v4addr;
400 	inet->inet_saddr = v4addr;
401 
402 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
403 
404 	if (!(addr_type & IPV6_ADDR_MULTICAST))
405 		np->saddr = addr->sin6_addr;
406 
407 	saved_ipv6only = sk->sk_ipv6only;
408 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
409 		sk->sk_ipv6only = 1;
410 
411 	/* Make sure we are allowed to bind here. */
412 	if ((snum || !inet->bind_address_no_port) &&
413 	    sk->sk_prot->get_port(sk, snum)) {
414 		sk->sk_ipv6only = saved_ipv6only;
415 		inet_reset_saddr(sk);
416 		err = -EADDRINUSE;
417 		goto out;
418 	}
419 
420 	if (addr_type != IPV6_ADDR_ANY)
421 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
422 	if (snum)
423 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
424 	inet->inet_sport = htons(inet->inet_num);
425 	inet->inet_dport = 0;
426 	inet->inet_daddr = 0;
427 out:
428 	release_sock(sk);
429 	return err;
430 out_unlock:
431 	rcu_read_unlock();
432 	goto out;
433 }
434 EXPORT_SYMBOL(inet6_bind);
435 
inet6_release(struct socket * sock)436 int inet6_release(struct socket *sock)
437 {
438 	struct sock *sk = sock->sk;
439 
440 	if (!sk)
441 		return -EINVAL;
442 
443 	/* Free mc lists */
444 	ipv6_sock_mc_close(sk);
445 
446 	/* Free ac lists */
447 	ipv6_sock_ac_close(sk);
448 
449 	return inet_release(sock);
450 }
451 EXPORT_SYMBOL(inet6_release);
452 
inet6_destroy_sock(struct sock * sk)453 void inet6_destroy_sock(struct sock *sk)
454 {
455 	struct ipv6_pinfo *np = inet6_sk(sk);
456 	struct sk_buff *skb;
457 	struct ipv6_txoptions *opt;
458 
459 	/* Release rx options */
460 
461 	skb = xchg(&np->pktoptions, NULL);
462 	if (skb)
463 		kfree_skb(skb);
464 
465 	skb = xchg(&np->rxpmtu, NULL);
466 	if (skb)
467 		kfree_skb(skb);
468 
469 	/* Free flowlabels */
470 	fl6_free_socklist(sk);
471 
472 	/* Free tx options */
473 
474 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
475 	if (opt) {
476 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
477 		txopt_put(opt);
478 	}
479 }
480 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
481 
482 /*
483  *	This does both peername and sockname.
484  */
485 
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int * uaddr_len,int peer)486 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
487 		 int *uaddr_len, int peer)
488 {
489 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
490 	struct sock *sk = sock->sk;
491 	struct inet_sock *inet = inet_sk(sk);
492 	struct ipv6_pinfo *np = inet6_sk(sk);
493 
494 	sin->sin6_family = AF_INET6;
495 	sin->sin6_flowinfo = 0;
496 	sin->sin6_scope_id = 0;
497 	if (peer) {
498 		if (!inet->inet_dport)
499 			return -ENOTCONN;
500 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
501 		    peer == 1)
502 			return -ENOTCONN;
503 		sin->sin6_port = inet->inet_dport;
504 		sin->sin6_addr = sk->sk_v6_daddr;
505 		if (np->sndflow)
506 			sin->sin6_flowinfo = np->flow_label;
507 	} else {
508 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
509 			sin->sin6_addr = np->saddr;
510 		else
511 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
512 
513 		sin->sin6_port = inet->inet_sport;
514 	}
515 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
516 						 sk->sk_bound_dev_if);
517 	*uaddr_len = sizeof(*sin);
518 	return 0;
519 }
520 EXPORT_SYMBOL(inet6_getname);
521 
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)522 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
523 {
524 	struct sock *sk = sock->sk;
525 	struct net *net = sock_net(sk);
526 
527 	switch (cmd) {
528 	case SIOCGSTAMP:
529 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
530 
531 	case SIOCGSTAMPNS:
532 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
533 
534 	case SIOCADDRT:
535 	case SIOCDELRT:
536 
537 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
538 
539 	case SIOCSIFADDR:
540 		return addrconf_add_ifaddr(net, (void __user *) arg);
541 	case SIOCDIFADDR:
542 		return addrconf_del_ifaddr(net, (void __user *) arg);
543 	case SIOCSIFDSTADDR:
544 		return addrconf_set_dstaddr(net, (void __user *) arg);
545 	default:
546 		if (!sk->sk_prot->ioctl)
547 			return -ENOIOCTLCMD;
548 		return sk->sk_prot->ioctl(sk, cmd, arg);
549 	}
550 	/*NOTREACHED*/
551 	return 0;
552 }
553 EXPORT_SYMBOL(inet6_ioctl);
554 
555 const struct proto_ops inet6_stream_ops = {
556 	.family		   = PF_INET6,
557 	.owner		   = THIS_MODULE,
558 	.release	   = inet6_release,
559 	.bind		   = inet6_bind,
560 	.connect	   = inet_stream_connect,	/* ok		*/
561 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
562 	.accept		   = inet_accept,		/* ok		*/
563 	.getname	   = inet6_getname,
564 	.poll		   = tcp_poll,			/* ok		*/
565 	.ioctl		   = inet6_ioctl,		/* must change  */
566 	.listen		   = inet_listen,		/* ok		*/
567 	.shutdown	   = inet_shutdown,		/* ok		*/
568 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
569 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
570 	.sendmsg	   = inet_sendmsg,		/* ok		*/
571 	.recvmsg	   = inet_recvmsg,		/* ok		*/
572 	.mmap		   = sock_no_mmap,
573 	.sendpage	   = inet_sendpage,
574 	.sendmsg_locked    = tcp_sendmsg_locked,
575 	.sendpage_locked   = tcp_sendpage_locked,
576 	.splice_read	   = tcp_splice_read,
577 	.read_sock	   = tcp_read_sock,
578 	.peek_len	   = tcp_peek_len,
579 #ifdef CONFIG_COMPAT
580 	.compat_setsockopt = compat_sock_common_setsockopt,
581 	.compat_getsockopt = compat_sock_common_getsockopt,
582 #endif
583 };
584 
585 const struct proto_ops inet6_dgram_ops = {
586 	.family		   = PF_INET6,
587 	.owner		   = THIS_MODULE,
588 	.release	   = inet6_release,
589 	.bind		   = inet6_bind,
590 	.connect	   = inet_dgram_connect,	/* ok		*/
591 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
592 	.accept		   = sock_no_accept,		/* a do nothing	*/
593 	.getname	   = inet6_getname,
594 	.poll		   = udp_poll,			/* ok		*/
595 	.ioctl		   = inet6_ioctl,		/* must change  */
596 	.listen		   = sock_no_listen,		/* ok		*/
597 	.shutdown	   = inet_shutdown,		/* ok		*/
598 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
599 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
600 	.sendmsg	   = inet_sendmsg,		/* ok		*/
601 	.recvmsg	   = inet_recvmsg,		/* ok		*/
602 	.mmap		   = sock_no_mmap,
603 	.sendpage	   = sock_no_sendpage,
604 	.set_peek_off	   = sk_set_peek_off,
605 #ifdef CONFIG_COMPAT
606 	.compat_setsockopt = compat_sock_common_setsockopt,
607 	.compat_getsockopt = compat_sock_common_getsockopt,
608 #endif
609 };
610 
611 static const struct net_proto_family inet6_family_ops = {
612 	.family = PF_INET6,
613 	.create = inet6_create,
614 	.owner	= THIS_MODULE,
615 };
616 
inet6_register_protosw(struct inet_protosw * p)617 int inet6_register_protosw(struct inet_protosw *p)
618 {
619 	struct list_head *lh;
620 	struct inet_protosw *answer;
621 	struct list_head *last_perm;
622 	int protocol = p->protocol;
623 	int ret;
624 
625 	spin_lock_bh(&inetsw6_lock);
626 
627 	ret = -EINVAL;
628 	if (p->type >= SOCK_MAX)
629 		goto out_illegal;
630 
631 	/* If we are trying to override a permanent protocol, bail. */
632 	answer = NULL;
633 	ret = -EPERM;
634 	last_perm = &inetsw6[p->type];
635 	list_for_each(lh, &inetsw6[p->type]) {
636 		answer = list_entry(lh, struct inet_protosw, list);
637 
638 		/* Check only the non-wild match. */
639 		if (INET_PROTOSW_PERMANENT & answer->flags) {
640 			if (protocol == answer->protocol)
641 				break;
642 			last_perm = lh;
643 		}
644 
645 		answer = NULL;
646 	}
647 	if (answer)
648 		goto out_permanent;
649 
650 	/* Add the new entry after the last permanent entry if any, so that
651 	 * the new entry does not override a permanent entry when matched with
652 	 * a wild-card protocol. But it is allowed to override any existing
653 	 * non-permanent entry.  This means that when we remove this entry, the
654 	 * system automatically returns to the old behavior.
655 	 */
656 	list_add_rcu(&p->list, last_perm);
657 	ret = 0;
658 out:
659 	spin_unlock_bh(&inetsw6_lock);
660 	return ret;
661 
662 out_permanent:
663 	pr_err("Attempt to override permanent protocol %d\n", protocol);
664 	goto out;
665 
666 out_illegal:
667 	pr_err("Ignoring attempt to register invalid socket type %d\n",
668 	       p->type);
669 	goto out;
670 }
671 EXPORT_SYMBOL(inet6_register_protosw);
672 
673 void
inet6_unregister_protosw(struct inet_protosw * p)674 inet6_unregister_protosw(struct inet_protosw *p)
675 {
676 	if (INET_PROTOSW_PERMANENT & p->flags) {
677 		pr_err("Attempt to unregister permanent protocol %d\n",
678 		       p->protocol);
679 	} else {
680 		spin_lock_bh(&inetsw6_lock);
681 		list_del_rcu(&p->list);
682 		spin_unlock_bh(&inetsw6_lock);
683 
684 		synchronize_net();
685 	}
686 }
687 EXPORT_SYMBOL(inet6_unregister_protosw);
688 
inet6_sk_rebuild_header(struct sock * sk)689 int inet6_sk_rebuild_header(struct sock *sk)
690 {
691 	struct ipv6_pinfo *np = inet6_sk(sk);
692 	struct dst_entry *dst;
693 
694 	dst = __sk_dst_check(sk, np->dst_cookie);
695 
696 	if (!dst) {
697 		struct inet_sock *inet = inet_sk(sk);
698 		struct in6_addr *final_p, final;
699 		struct flowi6 fl6;
700 
701 		memset(&fl6, 0, sizeof(fl6));
702 		fl6.flowi6_proto = sk->sk_protocol;
703 		fl6.daddr = sk->sk_v6_daddr;
704 		fl6.saddr = np->saddr;
705 		fl6.flowlabel = np->flow_label;
706 		fl6.flowi6_oif = sk->sk_bound_dev_if;
707 		fl6.flowi6_mark = sk->sk_mark;
708 		fl6.fl6_dport = inet->inet_dport;
709 		fl6.fl6_sport = inet->inet_sport;
710 		fl6.flowi6_uid = sk->sk_uid;
711 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
712 
713 		rcu_read_lock();
714 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
715 					 &final);
716 		rcu_read_unlock();
717 
718 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
719 		if (IS_ERR(dst)) {
720 			sk->sk_route_caps = 0;
721 			sk->sk_err_soft = -PTR_ERR(dst);
722 			return PTR_ERR(dst);
723 		}
724 
725 		ip6_dst_store(sk, dst, NULL, NULL);
726 	}
727 
728 	return 0;
729 }
730 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
731 
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb,const struct inet6_skb_parm * opt)732 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
733 		       const struct inet6_skb_parm *opt)
734 {
735 	const struct ipv6_pinfo *np = inet6_sk(sk);
736 
737 	if (np->rxopt.all) {
738 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
739 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
740 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
741 		     np->rxopt.bits.rxflow) ||
742 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
743 		     np->rxopt.bits.osrcrt)) ||
744 		    ((opt->dst1 || opt->dst0) &&
745 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
746 			return true;
747 	}
748 	return false;
749 }
750 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
751 
752 static struct packet_type ipv6_packet_type __read_mostly = {
753 	.type = cpu_to_be16(ETH_P_IPV6),
754 	.func = ipv6_rcv,
755 };
756 
ipv6_packet_init(void)757 static int __init ipv6_packet_init(void)
758 {
759 	dev_add_pack(&ipv6_packet_type);
760 	return 0;
761 }
762 
ipv6_packet_cleanup(void)763 static void ipv6_packet_cleanup(void)
764 {
765 	dev_remove_pack(&ipv6_packet_type);
766 }
767 
ipv6_init_mibs(struct net * net)768 static int __net_init ipv6_init_mibs(struct net *net)
769 {
770 	int i;
771 
772 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
773 	if (!net->mib.udp_stats_in6)
774 		return -ENOMEM;
775 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
776 	if (!net->mib.udplite_stats_in6)
777 		goto err_udplite_mib;
778 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
779 	if (!net->mib.ipv6_statistics)
780 		goto err_ip_mib;
781 
782 	for_each_possible_cpu(i) {
783 		struct ipstats_mib *af_inet6_stats;
784 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
785 		u64_stats_init(&af_inet6_stats->syncp);
786 	}
787 
788 
789 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
790 	if (!net->mib.icmpv6_statistics)
791 		goto err_icmp_mib;
792 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
793 						GFP_KERNEL);
794 	if (!net->mib.icmpv6msg_statistics)
795 		goto err_icmpmsg_mib;
796 	return 0;
797 
798 err_icmpmsg_mib:
799 	free_percpu(net->mib.icmpv6_statistics);
800 err_icmp_mib:
801 	free_percpu(net->mib.ipv6_statistics);
802 err_ip_mib:
803 	free_percpu(net->mib.udplite_stats_in6);
804 err_udplite_mib:
805 	free_percpu(net->mib.udp_stats_in6);
806 	return -ENOMEM;
807 }
808 
ipv6_cleanup_mibs(struct net * net)809 static void ipv6_cleanup_mibs(struct net *net)
810 {
811 	free_percpu(net->mib.udp_stats_in6);
812 	free_percpu(net->mib.udplite_stats_in6);
813 	free_percpu(net->mib.ipv6_statistics);
814 	free_percpu(net->mib.icmpv6_statistics);
815 	kfree(net->mib.icmpv6msg_statistics);
816 }
817 
inet6_net_init(struct net * net)818 static int __net_init inet6_net_init(struct net *net)
819 {
820 	int err = 0;
821 
822 	net->ipv6.sysctl.bindv6only = 0;
823 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
824 	net->ipv6.sysctl.flowlabel_consistency = 1;
825 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
826 	net->ipv6.sysctl.idgen_retries = 3;
827 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
828 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
829 	atomic_set(&net->ipv6.fib6_sernum, 1);
830 
831 	err = ipv6_init_mibs(net);
832 	if (err)
833 		return err;
834 #ifdef CONFIG_PROC_FS
835 	err = udp6_proc_init(net);
836 	if (err)
837 		goto out;
838 	err = tcp6_proc_init(net);
839 	if (err)
840 		goto proc_tcp6_fail;
841 	err = ac6_proc_init(net);
842 	if (err)
843 		goto proc_ac6_fail;
844 #endif
845 	return err;
846 
847 #ifdef CONFIG_PROC_FS
848 proc_ac6_fail:
849 	tcp6_proc_exit(net);
850 proc_tcp6_fail:
851 	udp6_proc_exit(net);
852 out:
853 	ipv6_cleanup_mibs(net);
854 	return err;
855 #endif
856 }
857 
inet6_net_exit(struct net * net)858 static void __net_exit inet6_net_exit(struct net *net)
859 {
860 #ifdef CONFIG_PROC_FS
861 	udp6_proc_exit(net);
862 	tcp6_proc_exit(net);
863 	ac6_proc_exit(net);
864 #endif
865 	ipv6_cleanup_mibs(net);
866 }
867 
868 static struct pernet_operations inet6_net_ops = {
869 	.init = inet6_net_init,
870 	.exit = inet6_net_exit,
871 };
872 
873 static const struct ipv6_stub ipv6_stub_impl = {
874 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
875 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
876 	.ipv6_dst_lookup = ip6_dst_lookup,
877 	.udpv6_encap_enable = udpv6_encap_enable,
878 	.ndisc_send_na = ndisc_send_na,
879 	.nd_tbl	= &nd_tbl,
880 };
881 
inet6_init(void)882 static int __init inet6_init(void)
883 {
884 	struct list_head *r;
885 	int err = 0;
886 
887 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
888 
889 	/* Register the socket-side information for inet6_create.  */
890 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
891 		INIT_LIST_HEAD(r);
892 
893 	if (disable_ipv6_mod) {
894 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
895 		goto out;
896 	}
897 
898 	err = proto_register(&tcpv6_prot, 1);
899 	if (err)
900 		goto out;
901 
902 	err = proto_register(&udpv6_prot, 1);
903 	if (err)
904 		goto out_unregister_tcp_proto;
905 
906 	err = proto_register(&udplitev6_prot, 1);
907 	if (err)
908 		goto out_unregister_udp_proto;
909 
910 	err = proto_register(&rawv6_prot, 1);
911 	if (err)
912 		goto out_unregister_udplite_proto;
913 
914 	err = proto_register(&pingv6_prot, 1);
915 	if (err)
916 		goto out_unregister_ping_proto;
917 
918 	/* We MUST register RAW sockets before we create the ICMP6,
919 	 * IGMP6, or NDISC control sockets.
920 	 */
921 	err = rawv6_init();
922 	if (err)
923 		goto out_unregister_raw_proto;
924 
925 	/* Register the family here so that the init calls below will
926 	 * be able to create sockets. (?? is this dangerous ??)
927 	 */
928 	err = sock_register(&inet6_family_ops);
929 	if (err)
930 		goto out_sock_register_fail;
931 
932 	/*
933 	 *	ipngwg API draft makes clear that the correct semantics
934 	 *	for TCP and UDP is to consider one TCP and UDP instance
935 	 *	in a host available by both INET and INET6 APIs and
936 	 *	able to communicate via both network protocols.
937 	 */
938 
939 	err = register_pernet_subsys(&inet6_net_ops);
940 	if (err)
941 		goto register_pernet_fail;
942 	err = ip6_mr_init();
943 	if (err)
944 		goto ipmr_fail;
945 	err = icmpv6_init();
946 	if (err)
947 		goto icmp_fail;
948 	err = ndisc_init();
949 	if (err)
950 		goto ndisc_fail;
951 	err = igmp6_init();
952 	if (err)
953 		goto igmp_fail;
954 
955 	err = ipv6_netfilter_init();
956 	if (err)
957 		goto netfilter_fail;
958 	/* Create /proc/foo6 entries. */
959 #ifdef CONFIG_PROC_FS
960 	err = -ENOMEM;
961 	if (raw6_proc_init())
962 		goto proc_raw6_fail;
963 	if (udplite6_proc_init())
964 		goto proc_udplite6_fail;
965 	if (ipv6_misc_proc_init())
966 		goto proc_misc6_fail;
967 	if (if6_proc_init())
968 		goto proc_if6_fail;
969 #endif
970 	err = ip6_route_init();
971 	if (err)
972 		goto ip6_route_fail;
973 	err = ndisc_late_init();
974 	if (err)
975 		goto ndisc_late_fail;
976 	err = ip6_flowlabel_init();
977 	if (err)
978 		goto ip6_flowlabel_fail;
979 	err = addrconf_init();
980 	if (err)
981 		goto addrconf_fail;
982 
983 	/* Init v6 extension headers. */
984 	err = ipv6_exthdrs_init();
985 	if (err)
986 		goto ipv6_exthdrs_fail;
987 
988 	err = ipv6_frag_init();
989 	if (err)
990 		goto ipv6_frag_fail;
991 
992 	/* Init v6 transport protocols. */
993 	err = udpv6_init();
994 	if (err)
995 		goto udpv6_fail;
996 
997 	err = udplitev6_init();
998 	if (err)
999 		goto udplitev6_fail;
1000 
1001 	err = udpv6_offload_init();
1002 	if (err)
1003 		goto udpv6_offload_fail;
1004 
1005 	err = tcpv6_init();
1006 	if (err)
1007 		goto tcpv6_fail;
1008 
1009 	err = ipv6_packet_init();
1010 	if (err)
1011 		goto ipv6_packet_fail;
1012 
1013 	err = pingv6_init();
1014 	if (err)
1015 		goto pingv6_fail;
1016 
1017 	err = calipso_init();
1018 	if (err)
1019 		goto calipso_fail;
1020 
1021 	err = seg6_init();
1022 	if (err)
1023 		goto seg6_fail;
1024 
1025 	err = igmp6_late_init();
1026 	if (err)
1027 		goto igmp6_late_err;
1028 
1029 #ifdef CONFIG_SYSCTL
1030 	err = ipv6_sysctl_register();
1031 	if (err)
1032 		goto sysctl_fail;
1033 #endif
1034 
1035 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1036 	wmb();
1037 	ipv6_stub = &ipv6_stub_impl;
1038 out:
1039 	return err;
1040 
1041 #ifdef CONFIG_SYSCTL
1042 sysctl_fail:
1043 	igmp6_late_cleanup();
1044 #endif
1045 igmp6_late_err:
1046 	seg6_exit();
1047 seg6_fail:
1048 	calipso_exit();
1049 calipso_fail:
1050 	pingv6_exit();
1051 pingv6_fail:
1052 	ipv6_packet_cleanup();
1053 ipv6_packet_fail:
1054 	tcpv6_exit();
1055 tcpv6_fail:
1056 	udpv6_offload_exit();
1057 udpv6_offload_fail:
1058 	udplitev6_exit();
1059 udplitev6_fail:
1060 	udpv6_exit();
1061 udpv6_fail:
1062 	ipv6_frag_exit();
1063 ipv6_frag_fail:
1064 	ipv6_exthdrs_exit();
1065 ipv6_exthdrs_fail:
1066 	addrconf_cleanup();
1067 addrconf_fail:
1068 	ip6_flowlabel_cleanup();
1069 ip6_flowlabel_fail:
1070 	ndisc_late_cleanup();
1071 ndisc_late_fail:
1072 	ip6_route_cleanup();
1073 ip6_route_fail:
1074 #ifdef CONFIG_PROC_FS
1075 	if6_proc_exit();
1076 proc_if6_fail:
1077 	ipv6_misc_proc_exit();
1078 proc_misc6_fail:
1079 	udplite6_proc_exit();
1080 proc_udplite6_fail:
1081 	raw6_proc_exit();
1082 proc_raw6_fail:
1083 #endif
1084 	ipv6_netfilter_fini();
1085 netfilter_fail:
1086 	igmp6_cleanup();
1087 igmp_fail:
1088 	ndisc_cleanup();
1089 ndisc_fail:
1090 	ip6_mr_cleanup();
1091 icmp_fail:
1092 	unregister_pernet_subsys(&inet6_net_ops);
1093 ipmr_fail:
1094 	icmpv6_cleanup();
1095 register_pernet_fail:
1096 	sock_unregister(PF_INET6);
1097 	rtnl_unregister_all(PF_INET6);
1098 out_sock_register_fail:
1099 	rawv6_exit();
1100 out_unregister_ping_proto:
1101 	proto_unregister(&pingv6_prot);
1102 out_unregister_raw_proto:
1103 	proto_unregister(&rawv6_prot);
1104 out_unregister_udplite_proto:
1105 	proto_unregister(&udplitev6_prot);
1106 out_unregister_udp_proto:
1107 	proto_unregister(&udpv6_prot);
1108 out_unregister_tcp_proto:
1109 	proto_unregister(&tcpv6_prot);
1110 	goto out;
1111 }
1112 module_init(inet6_init);
1113 
1114 MODULE_ALIAS_NETPROTO(PF_INET6);
1115