xref: /openbmc/linux/net/ipv6/af_inet6.c (revision d4295e12)
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 
97 bool ipv6_mod_enabled(void)
98 {
99 	return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102 
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 
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 &&
172 	    !ns_capable(net->user_ns, CAP_NET_RAW))
173 		goto out_rcu_unlock;
174 
175 	sock->ops = answer->ops;
176 	answer_prot = answer->prot;
177 	answer_flags = answer->flags;
178 	rcu_read_unlock();
179 
180 	WARN_ON(!answer_prot->slab);
181 
182 	err = -ENOBUFS;
183 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 	if (!sk)
185 		goto out;
186 
187 	sock_init_data(sock, sk);
188 
189 	err = 0;
190 	if (INET_PROTOSW_REUSE & answer_flags)
191 		sk->sk_reuse = SK_CAN_REUSE;
192 
193 	inet = inet_sk(sk);
194 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
195 
196 	if (SOCK_RAW == sock->type) {
197 		inet->inet_num = protocol;
198 		if (IPPROTO_RAW == protocol)
199 			inet->hdrincl = 1;
200 	}
201 
202 	sk->sk_destruct		= inet_sock_destruct;
203 	sk->sk_family		= PF_INET6;
204 	sk->sk_protocol		= protocol;
205 
206 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
207 
208 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
209 	np->hop_limit	= -1;
210 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
211 	np->mc_loop	= 1;
212 	np->mc_all	= 1;
213 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
214 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect;
215 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
216 
217 	/* Init the ipv4 part of the socket since we can have sockets
218 	 * using v6 API for ipv4.
219 	 */
220 	inet->uc_ttl	= -1;
221 
222 	inet->mc_loop	= 1;
223 	inet->mc_ttl	= 1;
224 	inet->mc_index	= 0;
225 	inet->mc_list	= NULL;
226 	inet->rcv_tos	= 0;
227 
228 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 		inet->pmtudisc = IP_PMTUDISC_DONT;
230 	else
231 		inet->pmtudisc = IP_PMTUDISC_WANT;
232 	/*
233 	 * Increment only the relevant sk_prot->socks debug field, this changes
234 	 * the previous behaviour of incrementing both the equivalent to
235 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
236 	 *
237 	 * This allows better debug granularity as we'll know exactly how many
238 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239 	 * transport protocol socks. -acme
240 	 */
241 	sk_refcnt_debug_inc(sk);
242 
243 	if (inet->inet_num) {
244 		/* It assumes that any protocol which allows
245 		 * the user to assign a number at socket
246 		 * creation time automatically shares.
247 		 */
248 		inet->inet_sport = htons(inet->inet_num);
249 		err = sk->sk_prot->hash(sk);
250 		if (err) {
251 			sk_common_release(sk);
252 			goto out;
253 		}
254 	}
255 	if (sk->sk_prot->init) {
256 		err = sk->sk_prot->init(sk);
257 		if (err) {
258 			sk_common_release(sk);
259 			goto out;
260 		}
261 	}
262 
263 	if (!kern) {
264 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265 		if (err) {
266 			sk_common_release(sk);
267 			goto out;
268 		}
269 	}
270 out:
271 	return err;
272 out_rcu_unlock:
273 	rcu_read_unlock();
274 	goto out;
275 }
276 
277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278 			bool force_bind_address_no_port, bool with_lock)
279 {
280 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
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 (addr->sin6_family != AF_INET6)
291 		return -EAFNOSUPPORT;
292 
293 	addr_type = ipv6_addr_type(&addr->sin6_addr);
294 	if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295 		return -EINVAL;
296 
297 	snum = ntohs(addr->sin6_port);
298 	if (snum && snum < inet_prot_sock(net) &&
299 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
300 		return -EACCES;
301 
302 	if (with_lock)
303 		lock_sock(sk);
304 
305 	/* Check these errors (active socket, double bind). */
306 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
307 		err = -EINVAL;
308 		goto out;
309 	}
310 
311 	/* Check if the address belongs to the host. */
312 	if (addr_type == IPV6_ADDR_MAPPED) {
313 		int chk_addr_ret;
314 
315 		/* Binding to v4-mapped address on a v6-only socket
316 		 * makes no sense
317 		 */
318 		if (sk->sk_ipv6only) {
319 			err = -EINVAL;
320 			goto out;
321 		}
322 
323 		/* Reproduce AF_INET checks to make the bindings consistent */
324 		v4addr = addr->sin6_addr.s6_addr32[3];
325 		chk_addr_ret = inet_addr_type(net, v4addr);
326 		if (!inet_can_nonlocal_bind(net, inet) &&
327 		    v4addr != htonl(INADDR_ANY) &&
328 		    chk_addr_ret != RTN_LOCAL &&
329 		    chk_addr_ret != RTN_MULTICAST &&
330 		    chk_addr_ret != RTN_BROADCAST) {
331 			err = -EADDRNOTAVAIL;
332 			goto out;
333 		}
334 	} else {
335 		if (addr_type != IPV6_ADDR_ANY) {
336 			struct net_device *dev = NULL;
337 
338 			rcu_read_lock();
339 			if (__ipv6_addr_needs_scope_id(addr_type)) {
340 				if (addr_len >= sizeof(struct sockaddr_in6) &&
341 				    addr->sin6_scope_id) {
342 					/* Override any existing binding, if another one
343 					 * is supplied by user.
344 					 */
345 					sk->sk_bound_dev_if = addr->sin6_scope_id;
346 				}
347 
348 				/* Binding to link-local address requires an interface */
349 				if (!sk->sk_bound_dev_if) {
350 					err = -EINVAL;
351 					goto out_unlock;
352 				}
353 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
354 				if (!dev) {
355 					err = -ENODEV;
356 					goto out_unlock;
357 				}
358 			}
359 
360 			/* ipv4 addr of the socket is invalid.  Only the
361 			 * unspecified and mapped address have a v4 equivalent.
362 			 */
363 			v4addr = LOOPBACK4_IPV6;
364 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
365 				if (!ipv6_can_nonlocal_bind(net, inet) &&
366 				    !ipv6_chk_addr(net, &addr->sin6_addr,
367 						   dev, 0)) {
368 					err = -EADDRNOTAVAIL;
369 					goto out_unlock;
370 				}
371 			}
372 			rcu_read_unlock();
373 		}
374 	}
375 
376 	inet->inet_rcv_saddr = v4addr;
377 	inet->inet_saddr = v4addr;
378 
379 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
380 
381 	if (!(addr_type & IPV6_ADDR_MULTICAST))
382 		np->saddr = addr->sin6_addr;
383 
384 	saved_ipv6only = sk->sk_ipv6only;
385 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
386 		sk->sk_ipv6only = 1;
387 
388 	/* Make sure we are allowed to bind here. */
389 	if (snum || !(inet->bind_address_no_port ||
390 		      force_bind_address_no_port)) {
391 		if (sk->sk_prot->get_port(sk, snum)) {
392 			sk->sk_ipv6only = saved_ipv6only;
393 			inet_reset_saddr(sk);
394 			err = -EADDRINUSE;
395 			goto out;
396 		}
397 		err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
398 		if (err) {
399 			sk->sk_ipv6only = saved_ipv6only;
400 			inet_reset_saddr(sk);
401 			goto out;
402 		}
403 	}
404 
405 	if (addr_type != IPV6_ADDR_ANY)
406 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
407 	if (snum)
408 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
409 	inet->inet_sport = htons(inet->inet_num);
410 	inet->inet_dport = 0;
411 	inet->inet_daddr = 0;
412 out:
413 	if (with_lock)
414 		release_sock(sk);
415 	return err;
416 out_unlock:
417 	rcu_read_unlock();
418 	goto out;
419 }
420 
421 /* bind for INET6 API */
422 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
423 {
424 	struct sock *sk = sock->sk;
425 	int err = 0;
426 
427 	/* If the socket has its own bind function then use it. */
428 	if (sk->sk_prot->bind)
429 		return sk->sk_prot->bind(sk, uaddr, addr_len);
430 
431 	if (addr_len < SIN6_LEN_RFC2133)
432 		return -EINVAL;
433 
434 	/* BPF prog is run before any checks are done so that if the prog
435 	 * changes context in a wrong way it will be caught.
436 	 */
437 	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
438 	if (err)
439 		return err;
440 
441 	return __inet6_bind(sk, uaddr, addr_len, false, true);
442 }
443 EXPORT_SYMBOL(inet6_bind);
444 
445 int inet6_release(struct socket *sock)
446 {
447 	struct sock *sk = sock->sk;
448 
449 	if (!sk)
450 		return -EINVAL;
451 
452 	/* Free mc lists */
453 	ipv6_sock_mc_close(sk);
454 
455 	/* Free ac lists */
456 	ipv6_sock_ac_close(sk);
457 
458 	return inet_release(sock);
459 }
460 EXPORT_SYMBOL(inet6_release);
461 
462 void inet6_destroy_sock(struct sock *sk)
463 {
464 	struct ipv6_pinfo *np = inet6_sk(sk);
465 	struct sk_buff *skb;
466 	struct ipv6_txoptions *opt;
467 
468 	/* Release rx options */
469 
470 	skb = xchg(&np->pktoptions, NULL);
471 	kfree_skb(skb);
472 
473 	skb = xchg(&np->rxpmtu, NULL);
474 	kfree_skb(skb);
475 
476 	/* Free flowlabels */
477 	fl6_free_socklist(sk);
478 
479 	/* Free tx options */
480 
481 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
482 	if (opt) {
483 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
484 		txopt_put(opt);
485 	}
486 }
487 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
488 
489 /*
490  *	This does both peername and sockname.
491  */
492 
493 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
494 		 int peer)
495 {
496 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
497 	struct sock *sk = sock->sk;
498 	struct inet_sock *inet = inet_sk(sk);
499 	struct ipv6_pinfo *np = inet6_sk(sk);
500 
501 	sin->sin6_family = AF_INET6;
502 	sin->sin6_flowinfo = 0;
503 	sin->sin6_scope_id = 0;
504 	if (peer) {
505 		if (!inet->inet_dport)
506 			return -ENOTCONN;
507 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
508 		    peer == 1)
509 			return -ENOTCONN;
510 		sin->sin6_port = inet->inet_dport;
511 		sin->sin6_addr = sk->sk_v6_daddr;
512 		if (np->sndflow)
513 			sin->sin6_flowinfo = np->flow_label;
514 	} else {
515 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
516 			sin->sin6_addr = np->saddr;
517 		else
518 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
519 
520 		sin->sin6_port = inet->inet_sport;
521 	}
522 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
523 						 sk->sk_bound_dev_if);
524 	return sizeof(*sin);
525 }
526 EXPORT_SYMBOL(inet6_getname);
527 
528 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
529 {
530 	struct sock *sk = sock->sk;
531 	struct net *net = sock_net(sk);
532 
533 	switch (cmd) {
534 	case SIOCGSTAMP:
535 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
536 
537 	case SIOCGSTAMPNS:
538 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
539 
540 	case SIOCADDRT:
541 	case SIOCDELRT:
542 
543 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
544 
545 	case SIOCSIFADDR:
546 		return addrconf_add_ifaddr(net, (void __user *) arg);
547 	case SIOCDIFADDR:
548 		return addrconf_del_ifaddr(net, (void __user *) arg);
549 	case SIOCSIFDSTADDR:
550 		return addrconf_set_dstaddr(net, (void __user *) arg);
551 	default:
552 		if (!sk->sk_prot->ioctl)
553 			return -ENOIOCTLCMD;
554 		return sk->sk_prot->ioctl(sk, cmd, arg);
555 	}
556 	/*NOTREACHED*/
557 	return 0;
558 }
559 EXPORT_SYMBOL(inet6_ioctl);
560 
561 const struct proto_ops inet6_stream_ops = {
562 	.family		   = PF_INET6,
563 	.owner		   = THIS_MODULE,
564 	.release	   = inet6_release,
565 	.bind		   = inet6_bind,
566 	.connect	   = inet_stream_connect,	/* ok		*/
567 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
568 	.accept		   = inet_accept,		/* ok		*/
569 	.getname	   = inet6_getname,
570 	.poll		   = tcp_poll,			/* ok		*/
571 	.ioctl		   = inet6_ioctl,		/* must change  */
572 	.listen		   = inet_listen,		/* ok		*/
573 	.shutdown	   = inet_shutdown,		/* ok		*/
574 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
575 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
576 	.sendmsg	   = inet_sendmsg,		/* ok		*/
577 	.recvmsg	   = inet_recvmsg,		/* ok		*/
578 #ifdef CONFIG_MMU
579 	.mmap		   = tcp_mmap,
580 #endif
581 	.sendpage	   = inet_sendpage,
582 	.sendmsg_locked    = tcp_sendmsg_locked,
583 	.sendpage_locked   = tcp_sendpage_locked,
584 	.splice_read	   = tcp_splice_read,
585 	.read_sock	   = tcp_read_sock,
586 	.peek_len	   = tcp_peek_len,
587 #ifdef CONFIG_COMPAT
588 	.compat_setsockopt = compat_sock_common_setsockopt,
589 	.compat_getsockopt = compat_sock_common_getsockopt,
590 #endif
591 	.set_rcvlowat	   = tcp_set_rcvlowat,
592 };
593 
594 const struct proto_ops inet6_dgram_ops = {
595 	.family		   = PF_INET6,
596 	.owner		   = THIS_MODULE,
597 	.release	   = inet6_release,
598 	.bind		   = inet6_bind,
599 	.connect	   = inet_dgram_connect,	/* ok		*/
600 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
601 	.accept		   = sock_no_accept,		/* a do nothing	*/
602 	.getname	   = inet6_getname,
603 	.poll		   = udp_poll,			/* ok		*/
604 	.ioctl		   = inet6_ioctl,		/* must change  */
605 	.listen		   = sock_no_listen,		/* ok		*/
606 	.shutdown	   = inet_shutdown,		/* ok		*/
607 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
608 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
609 	.sendmsg	   = inet_sendmsg,		/* ok		*/
610 	.recvmsg	   = inet_recvmsg,		/* ok		*/
611 	.mmap		   = sock_no_mmap,
612 	.sendpage	   = sock_no_sendpage,
613 	.set_peek_off	   = sk_set_peek_off,
614 #ifdef CONFIG_COMPAT
615 	.compat_setsockopt = compat_sock_common_setsockopt,
616 	.compat_getsockopt = compat_sock_common_getsockopt,
617 #endif
618 };
619 
620 static const struct net_proto_family inet6_family_ops = {
621 	.family = PF_INET6,
622 	.create = inet6_create,
623 	.owner	= THIS_MODULE,
624 };
625 
626 int inet6_register_protosw(struct inet_protosw *p)
627 {
628 	struct list_head *lh;
629 	struct inet_protosw *answer;
630 	struct list_head *last_perm;
631 	int protocol = p->protocol;
632 	int ret;
633 
634 	spin_lock_bh(&inetsw6_lock);
635 
636 	ret = -EINVAL;
637 	if (p->type >= SOCK_MAX)
638 		goto out_illegal;
639 
640 	/* If we are trying to override a permanent protocol, bail. */
641 	answer = NULL;
642 	ret = -EPERM;
643 	last_perm = &inetsw6[p->type];
644 	list_for_each(lh, &inetsw6[p->type]) {
645 		answer = list_entry(lh, struct inet_protosw, list);
646 
647 		/* Check only the non-wild match. */
648 		if (INET_PROTOSW_PERMANENT & answer->flags) {
649 			if (protocol == answer->protocol)
650 				break;
651 			last_perm = lh;
652 		}
653 
654 		answer = NULL;
655 	}
656 	if (answer)
657 		goto out_permanent;
658 
659 	/* Add the new entry after the last permanent entry if any, so that
660 	 * the new entry does not override a permanent entry when matched with
661 	 * a wild-card protocol. But it is allowed to override any existing
662 	 * non-permanent entry.  This means that when we remove this entry, the
663 	 * system automatically returns to the old behavior.
664 	 */
665 	list_add_rcu(&p->list, last_perm);
666 	ret = 0;
667 out:
668 	spin_unlock_bh(&inetsw6_lock);
669 	return ret;
670 
671 out_permanent:
672 	pr_err("Attempt to override permanent protocol %d\n", protocol);
673 	goto out;
674 
675 out_illegal:
676 	pr_err("Ignoring attempt to register invalid socket type %d\n",
677 	       p->type);
678 	goto out;
679 }
680 EXPORT_SYMBOL(inet6_register_protosw);
681 
682 void
683 inet6_unregister_protosw(struct inet_protosw *p)
684 {
685 	if (INET_PROTOSW_PERMANENT & p->flags) {
686 		pr_err("Attempt to unregister permanent protocol %d\n",
687 		       p->protocol);
688 	} else {
689 		spin_lock_bh(&inetsw6_lock);
690 		list_del_rcu(&p->list);
691 		spin_unlock_bh(&inetsw6_lock);
692 
693 		synchronize_net();
694 	}
695 }
696 EXPORT_SYMBOL(inet6_unregister_protosw);
697 
698 int inet6_sk_rebuild_header(struct sock *sk)
699 {
700 	struct ipv6_pinfo *np = inet6_sk(sk);
701 	struct dst_entry *dst;
702 
703 	dst = __sk_dst_check(sk, np->dst_cookie);
704 
705 	if (!dst) {
706 		struct inet_sock *inet = inet_sk(sk);
707 		struct in6_addr *final_p, final;
708 		struct flowi6 fl6;
709 
710 		memset(&fl6, 0, sizeof(fl6));
711 		fl6.flowi6_proto = sk->sk_protocol;
712 		fl6.daddr = sk->sk_v6_daddr;
713 		fl6.saddr = np->saddr;
714 		fl6.flowlabel = np->flow_label;
715 		fl6.flowi6_oif = sk->sk_bound_dev_if;
716 		fl6.flowi6_mark = sk->sk_mark;
717 		fl6.fl6_dport = inet->inet_dport;
718 		fl6.fl6_sport = inet->inet_sport;
719 		fl6.flowi6_uid = sk->sk_uid;
720 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
721 
722 		rcu_read_lock();
723 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
724 					 &final);
725 		rcu_read_unlock();
726 
727 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
728 		if (IS_ERR(dst)) {
729 			sk->sk_route_caps = 0;
730 			sk->sk_err_soft = -PTR_ERR(dst);
731 			return PTR_ERR(dst);
732 		}
733 
734 		ip6_dst_store(sk, dst, NULL, NULL);
735 	}
736 
737 	return 0;
738 }
739 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
740 
741 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
742 		       const struct inet6_skb_parm *opt)
743 {
744 	const struct ipv6_pinfo *np = inet6_sk(sk);
745 
746 	if (np->rxopt.all) {
747 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
748 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
749 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
750 		     np->rxopt.bits.rxflow) ||
751 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
752 		     np->rxopt.bits.osrcrt)) ||
753 		    ((opt->dst1 || opt->dst0) &&
754 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
755 			return true;
756 	}
757 	return false;
758 }
759 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
760 
761 static struct packet_type ipv6_packet_type __read_mostly = {
762 	.type = cpu_to_be16(ETH_P_IPV6),
763 	.func = ipv6_rcv,
764 	.list_func = ipv6_list_rcv,
765 };
766 
767 static int __init ipv6_packet_init(void)
768 {
769 	dev_add_pack(&ipv6_packet_type);
770 	return 0;
771 }
772 
773 static void ipv6_packet_cleanup(void)
774 {
775 	dev_remove_pack(&ipv6_packet_type);
776 }
777 
778 static int __net_init ipv6_init_mibs(struct net *net)
779 {
780 	int i;
781 
782 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
783 	if (!net->mib.udp_stats_in6)
784 		return -ENOMEM;
785 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
786 	if (!net->mib.udplite_stats_in6)
787 		goto err_udplite_mib;
788 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
789 	if (!net->mib.ipv6_statistics)
790 		goto err_ip_mib;
791 
792 	for_each_possible_cpu(i) {
793 		struct ipstats_mib *af_inet6_stats;
794 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
795 		u64_stats_init(&af_inet6_stats->syncp);
796 	}
797 
798 
799 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
800 	if (!net->mib.icmpv6_statistics)
801 		goto err_icmp_mib;
802 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
803 						GFP_KERNEL);
804 	if (!net->mib.icmpv6msg_statistics)
805 		goto err_icmpmsg_mib;
806 	return 0;
807 
808 err_icmpmsg_mib:
809 	free_percpu(net->mib.icmpv6_statistics);
810 err_icmp_mib:
811 	free_percpu(net->mib.ipv6_statistics);
812 err_ip_mib:
813 	free_percpu(net->mib.udplite_stats_in6);
814 err_udplite_mib:
815 	free_percpu(net->mib.udp_stats_in6);
816 	return -ENOMEM;
817 }
818 
819 static void ipv6_cleanup_mibs(struct net *net)
820 {
821 	free_percpu(net->mib.udp_stats_in6);
822 	free_percpu(net->mib.udplite_stats_in6);
823 	free_percpu(net->mib.ipv6_statistics);
824 	free_percpu(net->mib.icmpv6_statistics);
825 	kfree(net->mib.icmpv6msg_statistics);
826 }
827 
828 static int __net_init inet6_net_init(struct net *net)
829 {
830 	int err = 0;
831 
832 	net->ipv6.sysctl.bindv6only = 0;
833 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
834 	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
835 	net->ipv6.sysctl.flowlabel_consistency = 1;
836 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
837 	net->ipv6.sysctl.idgen_retries = 3;
838 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
839 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
840 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
841 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
842 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
843 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
844 	atomic_set(&net->ipv6.fib6_sernum, 1);
845 
846 	err = ipv6_init_mibs(net);
847 	if (err)
848 		return err;
849 #ifdef CONFIG_PROC_FS
850 	err = udp6_proc_init(net);
851 	if (err)
852 		goto out;
853 	err = tcp6_proc_init(net);
854 	if (err)
855 		goto proc_tcp6_fail;
856 	err = ac6_proc_init(net);
857 	if (err)
858 		goto proc_ac6_fail;
859 #endif
860 	return err;
861 
862 #ifdef CONFIG_PROC_FS
863 proc_ac6_fail:
864 	tcp6_proc_exit(net);
865 proc_tcp6_fail:
866 	udp6_proc_exit(net);
867 out:
868 	ipv6_cleanup_mibs(net);
869 	return err;
870 #endif
871 }
872 
873 static void __net_exit inet6_net_exit(struct net *net)
874 {
875 #ifdef CONFIG_PROC_FS
876 	udp6_proc_exit(net);
877 	tcp6_proc_exit(net);
878 	ac6_proc_exit(net);
879 #endif
880 	ipv6_cleanup_mibs(net);
881 }
882 
883 static struct pernet_operations inet6_net_ops = {
884 	.init = inet6_net_init,
885 	.exit = inet6_net_exit,
886 };
887 
888 static const struct ipv6_stub ipv6_stub_impl = {
889 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
890 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
891 	.ipv6_dst_lookup   = ip6_dst_lookup,
892 	.fib6_get_table	   = fib6_get_table,
893 	.fib6_table_lookup = fib6_table_lookup,
894 	.fib6_lookup       = fib6_lookup,
895 	.fib6_multipath_select = fib6_multipath_select,
896 	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
897 	.udpv6_encap_enable = udpv6_encap_enable,
898 	.ndisc_send_na = ndisc_send_na,
899 	.nd_tbl	= &nd_tbl,
900 };
901 
902 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
903 	.inet6_bind = __inet6_bind,
904 	.udp6_lib_lookup = __udp6_lib_lookup,
905 };
906 
907 static int __init inet6_init(void)
908 {
909 	struct list_head *r;
910 	int err = 0;
911 
912 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
913 
914 	/* Register the socket-side information for inet6_create.  */
915 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
916 		INIT_LIST_HEAD(r);
917 
918 	if (disable_ipv6_mod) {
919 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
920 		goto out;
921 	}
922 
923 	err = proto_register(&tcpv6_prot, 1);
924 	if (err)
925 		goto out;
926 
927 	err = proto_register(&udpv6_prot, 1);
928 	if (err)
929 		goto out_unregister_tcp_proto;
930 
931 	err = proto_register(&udplitev6_prot, 1);
932 	if (err)
933 		goto out_unregister_udp_proto;
934 
935 	err = proto_register(&rawv6_prot, 1);
936 	if (err)
937 		goto out_unregister_udplite_proto;
938 
939 	err = proto_register(&pingv6_prot, 1);
940 	if (err)
941 		goto out_unregister_raw_proto;
942 
943 	/* We MUST register RAW sockets before we create the ICMP6,
944 	 * IGMP6, or NDISC control sockets.
945 	 */
946 	err = rawv6_init();
947 	if (err)
948 		goto out_unregister_ping_proto;
949 
950 	/* Register the family here so that the init calls below will
951 	 * be able to create sockets. (?? is this dangerous ??)
952 	 */
953 	err = sock_register(&inet6_family_ops);
954 	if (err)
955 		goto out_sock_register_fail;
956 
957 	/*
958 	 *	ipngwg API draft makes clear that the correct semantics
959 	 *	for TCP and UDP is to consider one TCP and UDP instance
960 	 *	in a host available by both INET and INET6 APIs and
961 	 *	able to communicate via both network protocols.
962 	 */
963 
964 	err = register_pernet_subsys(&inet6_net_ops);
965 	if (err)
966 		goto register_pernet_fail;
967 	err = ip6_mr_init();
968 	if (err)
969 		goto ipmr_fail;
970 	err = icmpv6_init();
971 	if (err)
972 		goto icmp_fail;
973 	err = ndisc_init();
974 	if (err)
975 		goto ndisc_fail;
976 	err = igmp6_init();
977 	if (err)
978 		goto igmp_fail;
979 
980 	err = ipv6_netfilter_init();
981 	if (err)
982 		goto netfilter_fail;
983 	/* Create /proc/foo6 entries. */
984 #ifdef CONFIG_PROC_FS
985 	err = -ENOMEM;
986 	if (raw6_proc_init())
987 		goto proc_raw6_fail;
988 	if (udplite6_proc_init())
989 		goto proc_udplite6_fail;
990 	if (ipv6_misc_proc_init())
991 		goto proc_misc6_fail;
992 	if (if6_proc_init())
993 		goto proc_if6_fail;
994 #endif
995 	err = ip6_route_init();
996 	if (err)
997 		goto ip6_route_fail;
998 	err = ndisc_late_init();
999 	if (err)
1000 		goto ndisc_late_fail;
1001 	err = ip6_flowlabel_init();
1002 	if (err)
1003 		goto ip6_flowlabel_fail;
1004 	err = ipv6_anycast_init();
1005 	if (err)
1006 		goto ipv6_anycast_fail;
1007 	err = addrconf_init();
1008 	if (err)
1009 		goto addrconf_fail;
1010 
1011 	/* Init v6 extension headers. */
1012 	err = ipv6_exthdrs_init();
1013 	if (err)
1014 		goto ipv6_exthdrs_fail;
1015 
1016 	err = ipv6_frag_init();
1017 	if (err)
1018 		goto ipv6_frag_fail;
1019 
1020 	/* Init v6 transport protocols. */
1021 	err = udpv6_init();
1022 	if (err)
1023 		goto udpv6_fail;
1024 
1025 	err = udplitev6_init();
1026 	if (err)
1027 		goto udplitev6_fail;
1028 
1029 	err = udpv6_offload_init();
1030 	if (err)
1031 		goto udpv6_offload_fail;
1032 
1033 	err = tcpv6_init();
1034 	if (err)
1035 		goto tcpv6_fail;
1036 
1037 	err = ipv6_packet_init();
1038 	if (err)
1039 		goto ipv6_packet_fail;
1040 
1041 	err = pingv6_init();
1042 	if (err)
1043 		goto pingv6_fail;
1044 
1045 	err = calipso_init();
1046 	if (err)
1047 		goto calipso_fail;
1048 
1049 	err = seg6_init();
1050 	if (err)
1051 		goto seg6_fail;
1052 
1053 	err = igmp6_late_init();
1054 	if (err)
1055 		goto igmp6_late_err;
1056 
1057 #ifdef CONFIG_SYSCTL
1058 	err = ipv6_sysctl_register();
1059 	if (err)
1060 		goto sysctl_fail;
1061 #endif
1062 
1063 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1064 	wmb();
1065 	ipv6_stub = &ipv6_stub_impl;
1066 	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1067 out:
1068 	return err;
1069 
1070 #ifdef CONFIG_SYSCTL
1071 sysctl_fail:
1072 	igmp6_late_cleanup();
1073 #endif
1074 igmp6_late_err:
1075 	seg6_exit();
1076 seg6_fail:
1077 	calipso_exit();
1078 calipso_fail:
1079 	pingv6_exit();
1080 pingv6_fail:
1081 	ipv6_packet_cleanup();
1082 ipv6_packet_fail:
1083 	tcpv6_exit();
1084 tcpv6_fail:
1085 	udpv6_offload_exit();
1086 udpv6_offload_fail:
1087 	udplitev6_exit();
1088 udplitev6_fail:
1089 	udpv6_exit();
1090 udpv6_fail:
1091 	ipv6_frag_exit();
1092 ipv6_frag_fail:
1093 	ipv6_exthdrs_exit();
1094 ipv6_exthdrs_fail:
1095 	addrconf_cleanup();
1096 addrconf_fail:
1097 	ipv6_anycast_cleanup();
1098 ipv6_anycast_fail:
1099 	ip6_flowlabel_cleanup();
1100 ip6_flowlabel_fail:
1101 	ndisc_late_cleanup();
1102 ndisc_late_fail:
1103 	ip6_route_cleanup();
1104 ip6_route_fail:
1105 #ifdef CONFIG_PROC_FS
1106 	if6_proc_exit();
1107 proc_if6_fail:
1108 	ipv6_misc_proc_exit();
1109 proc_misc6_fail:
1110 	udplite6_proc_exit();
1111 proc_udplite6_fail:
1112 	raw6_proc_exit();
1113 proc_raw6_fail:
1114 #endif
1115 	ipv6_netfilter_fini();
1116 netfilter_fail:
1117 	igmp6_cleanup();
1118 igmp_fail:
1119 	ndisc_cleanup();
1120 ndisc_fail:
1121 	icmpv6_cleanup();
1122 icmp_fail:
1123 	ip6_mr_cleanup();
1124 ipmr_fail:
1125 	unregister_pernet_subsys(&inet6_net_ops);
1126 register_pernet_fail:
1127 	sock_unregister(PF_INET6);
1128 	rtnl_unregister_all(PF_INET6);
1129 out_sock_register_fail:
1130 	rawv6_exit();
1131 out_unregister_ping_proto:
1132 	proto_unregister(&pingv6_prot);
1133 out_unregister_raw_proto:
1134 	proto_unregister(&rawv6_prot);
1135 out_unregister_udplite_proto:
1136 	proto_unregister(&udplitev6_prot);
1137 out_unregister_udp_proto:
1138 	proto_unregister(&udpv6_prot);
1139 out_unregister_tcp_proto:
1140 	proto_unregister(&tcpv6_prot);
1141 	goto out;
1142 }
1143 module_init(inet6_init);
1144 
1145 MODULE_ALIAS_NETPROTO(PF_INET6);
1146