xref: /openbmc/linux/net/ipv6/af_inet6.c (revision b9ccfda2)
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/ipip.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 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
69 
70 /* The inetsw6 table contains everything that inet6_create needs to
71  * build a new socket.
72  */
73 static struct list_head inetsw6[SOCK_MAX];
74 static DEFINE_SPINLOCK(inetsw6_lock);
75 
76 struct ipv6_params ipv6_defaults = {
77 	.disable_ipv6 = 0,
78 	.autoconf = 1,
79 };
80 
81 static int disable_ipv6_mod;
82 
83 module_param_named(disable, disable_ipv6_mod, int, 0444);
84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
85 
86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
88 
89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
91 
92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
93 {
94 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
95 
96 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
97 }
98 
99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
100 			int kern)
101 {
102 	struct inet_sock *inet;
103 	struct ipv6_pinfo *np;
104 	struct sock *sk;
105 	struct inet_protosw *answer;
106 	struct proto *answer_prot;
107 	unsigned char answer_flags;
108 	char answer_no_check;
109 	int try_loading_module = 0;
110 	int err;
111 
112 	if (sock->type != SOCK_RAW &&
113 	    sock->type != SOCK_DGRAM &&
114 	    !inet_ehash_secret)
115 		build_ehash_secret();
116 
117 	/* Look for the requested type/protocol pair. */
118 lookup_protocol:
119 	err = -ESOCKTNOSUPPORT;
120 	rcu_read_lock();
121 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122 
123 		err = 0;
124 		/* Check the non-wild match. */
125 		if (protocol == answer->protocol) {
126 			if (protocol != IPPROTO_IP)
127 				break;
128 		} else {
129 			/* Check for the two wild cases. */
130 			if (IPPROTO_IP == protocol) {
131 				protocol = answer->protocol;
132 				break;
133 			}
134 			if (IPPROTO_IP == answer->protocol)
135 				break;
136 		}
137 		err = -EPROTONOSUPPORT;
138 	}
139 
140 	if (err) {
141 		if (try_loading_module < 2) {
142 			rcu_read_unlock();
143 			/*
144 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146 			 */
147 			if (++try_loading_module == 1)
148 				request_module("net-pf-%d-proto-%d-type-%d",
149 						PF_INET6, protocol, sock->type);
150 			/*
151 			 * Fall back to generic, e.g. net-pf-10-proto-132
152 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 			 */
154 			else
155 				request_module("net-pf-%d-proto-%d",
156 						PF_INET6, protocol);
157 			goto lookup_protocol;
158 		} else
159 			goto out_rcu_unlock;
160 	}
161 
162 	err = -EPERM;
163 	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
164 		goto out_rcu_unlock;
165 
166 	sock->ops = answer->ops;
167 	answer_prot = answer->prot;
168 	answer_no_check = answer->no_check;
169 	answer_flags = answer->flags;
170 	rcu_read_unlock();
171 
172 	WARN_ON(answer_prot->slab == NULL);
173 
174 	err = -ENOBUFS;
175 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
176 	if (sk == NULL)
177 		goto out;
178 
179 	sock_init_data(sock, sk);
180 
181 	err = 0;
182 	sk->sk_no_check = answer_no_check;
183 	if (INET_PROTOSW_REUSE & answer_flags)
184 		sk->sk_reuse = SK_CAN_REUSE;
185 
186 	inet = inet_sk(sk);
187 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
188 
189 	if (SOCK_RAW == sock->type) {
190 		inet->inet_num = protocol;
191 		if (IPPROTO_RAW == protocol)
192 			inet->hdrincl = 1;
193 	}
194 
195 	sk->sk_destruct		= inet_sock_destruct;
196 	sk->sk_family		= PF_INET6;
197 	sk->sk_protocol		= protocol;
198 
199 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
200 
201 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
202 	np->hop_limit	= -1;
203 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
204 	np->mc_loop	= 1;
205 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
206 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
207 
208 	/* Init the ipv4 part of the socket since we can have sockets
209 	 * using v6 API for ipv4.
210 	 */
211 	inet->uc_ttl	= -1;
212 
213 	inet->mc_loop	= 1;
214 	inet->mc_ttl	= 1;
215 	inet->mc_index	= 0;
216 	inet->mc_list	= NULL;
217 	inet->rcv_tos	= 0;
218 
219 	if (ipv4_config.no_pmtu_disc)
220 		inet->pmtudisc = IP_PMTUDISC_DONT;
221 	else
222 		inet->pmtudisc = IP_PMTUDISC_WANT;
223 	/*
224 	 * Increment only the relevant sk_prot->socks debug field, this changes
225 	 * the previous behaviour of incrementing both the equivalent to
226 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
227 	 *
228 	 * This allows better debug granularity as we'll know exactly how many
229 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
230 	 * transport protocol socks. -acme
231 	 */
232 	sk_refcnt_debug_inc(sk);
233 
234 	if (inet->inet_num) {
235 		/* It assumes that any protocol which allows
236 		 * the user to assign a number at socket
237 		 * creation time automatically shares.
238 		 */
239 		inet->inet_sport = htons(inet->inet_num);
240 		sk->sk_prot->hash(sk);
241 	}
242 	if (sk->sk_prot->init) {
243 		err = sk->sk_prot->init(sk);
244 		if (err) {
245 			sk_common_release(sk);
246 			goto out;
247 		}
248 	}
249 out:
250 	return err;
251 out_rcu_unlock:
252 	rcu_read_unlock();
253 	goto out;
254 }
255 
256 
257 /* bind for INET6 API */
258 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
259 {
260 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
261 	struct sock *sk = sock->sk;
262 	struct inet_sock *inet = inet_sk(sk);
263 	struct ipv6_pinfo *np = inet6_sk(sk);
264 	struct net *net = sock_net(sk);
265 	__be32 v4addr = 0;
266 	unsigned short snum;
267 	int addr_type = 0;
268 	int err = 0;
269 
270 	/* If the socket has its own bind function then use it. */
271 	if (sk->sk_prot->bind)
272 		return sk->sk_prot->bind(sk, uaddr, addr_len);
273 
274 	if (addr_len < SIN6_LEN_RFC2133)
275 		return -EINVAL;
276 
277 	if (addr->sin6_family != AF_INET6)
278 		return -EAFNOSUPPORT;
279 
280 	addr_type = ipv6_addr_type(&addr->sin6_addr);
281 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
282 		return -EINVAL;
283 
284 	snum = ntohs(addr->sin6_port);
285 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
286 		return -EACCES;
287 
288 	lock_sock(sk);
289 
290 	/* Check these errors (active socket, double bind). */
291 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
292 		err = -EINVAL;
293 		goto out;
294 	}
295 
296 	/* Check if the address belongs to the host. */
297 	if (addr_type == IPV6_ADDR_MAPPED) {
298 		int chk_addr_ret;
299 
300 		/* Binding to v4-mapped address on a v6-only socket
301 		 * makes no sense
302 		 */
303 		if (np->ipv6only) {
304 			err = -EINVAL;
305 			goto out;
306 		}
307 
308 		/* Reproduce AF_INET checks to make the bindings consistent */
309 		v4addr = addr->sin6_addr.s6_addr32[3];
310 		chk_addr_ret = inet_addr_type(net, v4addr);
311 		if (!sysctl_ip_nonlocal_bind &&
312 		    !(inet->freebind || inet->transparent) &&
313 		    v4addr != htonl(INADDR_ANY) &&
314 		    chk_addr_ret != RTN_LOCAL &&
315 		    chk_addr_ret != RTN_MULTICAST &&
316 		    chk_addr_ret != RTN_BROADCAST) {
317 			err = -EADDRNOTAVAIL;
318 			goto out;
319 		}
320 	} else {
321 		if (addr_type != IPV6_ADDR_ANY) {
322 			struct net_device *dev = NULL;
323 
324 			rcu_read_lock();
325 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
326 				if (addr_len >= sizeof(struct sockaddr_in6) &&
327 				    addr->sin6_scope_id) {
328 					/* Override any existing binding, if another one
329 					 * is supplied by user.
330 					 */
331 					sk->sk_bound_dev_if = addr->sin6_scope_id;
332 				}
333 
334 				/* Binding to link-local address requires an interface */
335 				if (!sk->sk_bound_dev_if) {
336 					err = -EINVAL;
337 					goto out_unlock;
338 				}
339 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
340 				if (!dev) {
341 					err = -ENODEV;
342 					goto out_unlock;
343 				}
344 			}
345 
346 			/* ipv4 addr of the socket is invalid.  Only the
347 			 * unspecified and mapped address have a v4 equivalent.
348 			 */
349 			v4addr = LOOPBACK4_IPV6;
350 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
351 				if (!(inet->freebind || inet->transparent) &&
352 				    !ipv6_chk_addr(net, &addr->sin6_addr,
353 						   dev, 0)) {
354 					err = -EADDRNOTAVAIL;
355 					goto out_unlock;
356 				}
357 			}
358 			rcu_read_unlock();
359 		}
360 	}
361 
362 	inet->inet_rcv_saddr = v4addr;
363 	inet->inet_saddr = v4addr;
364 
365 	np->rcv_saddr = addr->sin6_addr;
366 
367 	if (!(addr_type & IPV6_ADDR_MULTICAST))
368 		np->saddr = addr->sin6_addr;
369 
370 	/* Make sure we are allowed to bind here. */
371 	if (sk->sk_prot->get_port(sk, snum)) {
372 		inet_reset_saddr(sk);
373 		err = -EADDRINUSE;
374 		goto out;
375 	}
376 
377 	if (addr_type != IPV6_ADDR_ANY) {
378 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
379 		if (addr_type != IPV6_ADDR_MAPPED)
380 			np->ipv6only = 1;
381 	}
382 	if (snum)
383 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
384 	inet->inet_sport = htons(inet->inet_num);
385 	inet->inet_dport = 0;
386 	inet->inet_daddr = 0;
387 out:
388 	release_sock(sk);
389 	return err;
390 out_unlock:
391 	rcu_read_unlock();
392 	goto out;
393 }
394 EXPORT_SYMBOL(inet6_bind);
395 
396 int inet6_release(struct socket *sock)
397 {
398 	struct sock *sk = sock->sk;
399 
400 	if (sk == NULL)
401 		return -EINVAL;
402 
403 	/* Free mc lists */
404 	ipv6_sock_mc_close(sk);
405 
406 	/* Free ac lists */
407 	ipv6_sock_ac_close(sk);
408 
409 	return inet_release(sock);
410 }
411 EXPORT_SYMBOL(inet6_release);
412 
413 void inet6_destroy_sock(struct sock *sk)
414 {
415 	struct ipv6_pinfo *np = inet6_sk(sk);
416 	struct sk_buff *skb;
417 	struct ipv6_txoptions *opt;
418 
419 	/* Release rx options */
420 
421 	skb = xchg(&np->pktoptions, NULL);
422 	if (skb != NULL)
423 		kfree_skb(skb);
424 
425 	skb = xchg(&np->rxpmtu, NULL);
426 	if (skb != NULL)
427 		kfree_skb(skb);
428 
429 	/* Free flowlabels */
430 	fl6_free_socklist(sk);
431 
432 	/* Free tx options */
433 
434 	opt = xchg(&np->opt, NULL);
435 	if (opt != NULL)
436 		sock_kfree_s(sk, opt, opt->tot_len);
437 }
438 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
439 
440 /*
441  *	This does both peername and sockname.
442  */
443 
444 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
445 		 int *uaddr_len, int peer)
446 {
447 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
448 	struct sock *sk = sock->sk;
449 	struct inet_sock *inet = inet_sk(sk);
450 	struct ipv6_pinfo *np = inet6_sk(sk);
451 
452 	sin->sin6_family = AF_INET6;
453 	sin->sin6_flowinfo = 0;
454 	sin->sin6_scope_id = 0;
455 	if (peer) {
456 		if (!inet->inet_dport)
457 			return -ENOTCONN;
458 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
459 		    peer == 1)
460 			return -ENOTCONN;
461 		sin->sin6_port = inet->inet_dport;
462 		sin->sin6_addr = np->daddr;
463 		if (np->sndflow)
464 			sin->sin6_flowinfo = np->flow_label;
465 	} else {
466 		if (ipv6_addr_any(&np->rcv_saddr))
467 			sin->sin6_addr = np->saddr;
468 		else
469 			sin->sin6_addr = np->rcv_saddr;
470 
471 		sin->sin6_port = inet->inet_sport;
472 	}
473 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
474 		sin->sin6_scope_id = sk->sk_bound_dev_if;
475 	*uaddr_len = sizeof(*sin);
476 	return 0;
477 }
478 EXPORT_SYMBOL(inet6_getname);
479 
480 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
481 {
482 	struct sock *sk = sock->sk;
483 	struct net *net = sock_net(sk);
484 
485 	switch (cmd) {
486 	case SIOCGSTAMP:
487 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
488 
489 	case SIOCGSTAMPNS:
490 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
491 
492 	case SIOCADDRT:
493 	case SIOCDELRT:
494 
495 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
496 
497 	case SIOCSIFADDR:
498 		return addrconf_add_ifaddr(net, (void __user *) arg);
499 	case SIOCDIFADDR:
500 		return addrconf_del_ifaddr(net, (void __user *) arg);
501 	case SIOCSIFDSTADDR:
502 		return addrconf_set_dstaddr(net, (void __user *) arg);
503 	default:
504 		if (!sk->sk_prot->ioctl)
505 			return -ENOIOCTLCMD;
506 		return sk->sk_prot->ioctl(sk, cmd, arg);
507 	}
508 	/*NOTREACHED*/
509 	return 0;
510 }
511 EXPORT_SYMBOL(inet6_ioctl);
512 
513 const struct proto_ops inet6_stream_ops = {
514 	.family		   = PF_INET6,
515 	.owner		   = THIS_MODULE,
516 	.release	   = inet6_release,
517 	.bind		   = inet6_bind,
518 	.connect	   = inet_stream_connect,	/* ok		*/
519 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
520 	.accept		   = inet_accept,		/* ok		*/
521 	.getname	   = inet6_getname,
522 	.poll		   = tcp_poll,			/* ok		*/
523 	.ioctl		   = inet6_ioctl,		/* must change  */
524 	.listen		   = inet_listen,		/* ok		*/
525 	.shutdown	   = inet_shutdown,		/* ok		*/
526 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
527 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
528 	.sendmsg	   = inet_sendmsg,		/* ok		*/
529 	.recvmsg	   = inet_recvmsg,		/* ok		*/
530 	.mmap		   = sock_no_mmap,
531 	.sendpage	   = inet_sendpage,
532 	.splice_read	   = tcp_splice_read,
533 #ifdef CONFIG_COMPAT
534 	.compat_setsockopt = compat_sock_common_setsockopt,
535 	.compat_getsockopt = compat_sock_common_getsockopt,
536 #endif
537 };
538 
539 const struct proto_ops inet6_dgram_ops = {
540 	.family		   = PF_INET6,
541 	.owner		   = THIS_MODULE,
542 	.release	   = inet6_release,
543 	.bind		   = inet6_bind,
544 	.connect	   = inet_dgram_connect,	/* ok		*/
545 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
546 	.accept		   = sock_no_accept,		/* a do nothing	*/
547 	.getname	   = inet6_getname,
548 	.poll		   = udp_poll,			/* ok		*/
549 	.ioctl		   = inet6_ioctl,		/* must change  */
550 	.listen		   = sock_no_listen,		/* ok		*/
551 	.shutdown	   = inet_shutdown,		/* ok		*/
552 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
553 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
554 	.sendmsg	   = inet_sendmsg,		/* ok		*/
555 	.recvmsg	   = inet_recvmsg,		/* ok		*/
556 	.mmap		   = sock_no_mmap,
557 	.sendpage	   = sock_no_sendpage,
558 #ifdef CONFIG_COMPAT
559 	.compat_setsockopt = compat_sock_common_setsockopt,
560 	.compat_getsockopt = compat_sock_common_getsockopt,
561 #endif
562 };
563 
564 static const struct net_proto_family inet6_family_ops = {
565 	.family = PF_INET6,
566 	.create = inet6_create,
567 	.owner	= THIS_MODULE,
568 };
569 
570 int inet6_register_protosw(struct inet_protosw *p)
571 {
572 	struct list_head *lh;
573 	struct inet_protosw *answer;
574 	struct list_head *last_perm;
575 	int protocol = p->protocol;
576 	int ret;
577 
578 	spin_lock_bh(&inetsw6_lock);
579 
580 	ret = -EINVAL;
581 	if (p->type >= SOCK_MAX)
582 		goto out_illegal;
583 
584 	/* If we are trying to override a permanent protocol, bail. */
585 	answer = NULL;
586 	ret = -EPERM;
587 	last_perm = &inetsw6[p->type];
588 	list_for_each(lh, &inetsw6[p->type]) {
589 		answer = list_entry(lh, struct inet_protosw, list);
590 
591 		/* Check only the non-wild match. */
592 		if (INET_PROTOSW_PERMANENT & answer->flags) {
593 			if (protocol == answer->protocol)
594 				break;
595 			last_perm = lh;
596 		}
597 
598 		answer = NULL;
599 	}
600 	if (answer)
601 		goto out_permanent;
602 
603 	/* Add the new entry after the last permanent entry if any, so that
604 	 * the new entry does not override a permanent entry when matched with
605 	 * a wild-card protocol. But it is allowed to override any existing
606 	 * non-permanent entry.  This means that when we remove this entry, the
607 	 * system automatically returns to the old behavior.
608 	 */
609 	list_add_rcu(&p->list, last_perm);
610 	ret = 0;
611 out:
612 	spin_unlock_bh(&inetsw6_lock);
613 	return ret;
614 
615 out_permanent:
616 	pr_err("Attempt to override permanent protocol %d\n", protocol);
617 	goto out;
618 
619 out_illegal:
620 	pr_err("Ignoring attempt to register invalid socket type %d\n",
621 	       p->type);
622 	goto out;
623 }
624 EXPORT_SYMBOL(inet6_register_protosw);
625 
626 void
627 inet6_unregister_protosw(struct inet_protosw *p)
628 {
629 	if (INET_PROTOSW_PERMANENT & p->flags) {
630 		pr_err("Attempt to unregister permanent protocol %d\n",
631 		       p->protocol);
632 	} else {
633 		spin_lock_bh(&inetsw6_lock);
634 		list_del_rcu(&p->list);
635 		spin_unlock_bh(&inetsw6_lock);
636 
637 		synchronize_net();
638 	}
639 }
640 EXPORT_SYMBOL(inet6_unregister_protosw);
641 
642 int inet6_sk_rebuild_header(struct sock *sk)
643 {
644 	struct ipv6_pinfo *np = inet6_sk(sk);
645 	struct dst_entry *dst;
646 
647 	dst = __sk_dst_check(sk, np->dst_cookie);
648 
649 	if (dst == NULL) {
650 		struct inet_sock *inet = inet_sk(sk);
651 		struct in6_addr *final_p, final;
652 		struct flowi6 fl6;
653 
654 		memset(&fl6, 0, sizeof(fl6));
655 		fl6.flowi6_proto = sk->sk_protocol;
656 		fl6.daddr = np->daddr;
657 		fl6.saddr = np->saddr;
658 		fl6.flowlabel = np->flow_label;
659 		fl6.flowi6_oif = sk->sk_bound_dev_if;
660 		fl6.flowi6_mark = sk->sk_mark;
661 		fl6.fl6_dport = inet->inet_dport;
662 		fl6.fl6_sport = inet->inet_sport;
663 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
664 
665 		final_p = fl6_update_dst(&fl6, np->opt, &final);
666 
667 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
668 		if (IS_ERR(dst)) {
669 			sk->sk_route_caps = 0;
670 			sk->sk_err_soft = -PTR_ERR(dst);
671 			return PTR_ERR(dst);
672 		}
673 
674 		__ip6_dst_store(sk, dst, NULL, NULL);
675 	}
676 
677 	return 0;
678 }
679 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
680 
681 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
682 {
683 	const struct ipv6_pinfo *np = inet6_sk(sk);
684 	const struct inet6_skb_parm *opt = IP6CB(skb);
685 
686 	if (np->rxopt.all) {
687 		if ((opt->hop && (np->rxopt.bits.hopopts ||
688 				  np->rxopt.bits.ohopopts)) ||
689 		    ((IPV6_FLOWINFO_MASK &
690 		      *(__be32 *)skb_network_header(skb)) &&
691 		     np->rxopt.bits.rxflow) ||
692 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
693 		     np->rxopt.bits.osrcrt)) ||
694 		    ((opt->dst1 || opt->dst0) &&
695 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
696 			return true;
697 	}
698 	return false;
699 }
700 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
701 
702 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
703 {
704 	const struct inet6_protocol *ops = NULL;
705 
706 	for (;;) {
707 		struct ipv6_opt_hdr *opth;
708 		int len;
709 
710 		if (proto != NEXTHDR_HOP) {
711 			ops = rcu_dereference(inet6_protos[proto]);
712 
713 			if (unlikely(!ops))
714 				break;
715 
716 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
717 				break;
718 		}
719 
720 		if (unlikely(!pskb_may_pull(skb, 8)))
721 			break;
722 
723 		opth = (void *)skb->data;
724 		len = ipv6_optlen(opth);
725 
726 		if (unlikely(!pskb_may_pull(skb, len)))
727 			break;
728 
729 		proto = opth->nexthdr;
730 		__skb_pull(skb, len);
731 	}
732 
733 	return proto;
734 }
735 
736 static int ipv6_gso_send_check(struct sk_buff *skb)
737 {
738 	const struct ipv6hdr *ipv6h;
739 	const struct inet6_protocol *ops;
740 	int err = -EINVAL;
741 
742 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
743 		goto out;
744 
745 	ipv6h = ipv6_hdr(skb);
746 	__skb_pull(skb, sizeof(*ipv6h));
747 	err = -EPROTONOSUPPORT;
748 
749 	rcu_read_lock();
750 	ops = rcu_dereference(inet6_protos[
751 		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
752 
753 	if (likely(ops && ops->gso_send_check)) {
754 		skb_reset_transport_header(skb);
755 		err = ops->gso_send_check(skb);
756 	}
757 	rcu_read_unlock();
758 
759 out:
760 	return err;
761 }
762 
763 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
764 	netdev_features_t features)
765 {
766 	struct sk_buff *segs = ERR_PTR(-EINVAL);
767 	struct ipv6hdr *ipv6h;
768 	const struct inet6_protocol *ops;
769 	int proto;
770 	struct frag_hdr *fptr;
771 	unsigned int unfrag_ip6hlen;
772 	u8 *prevhdr;
773 	int offset = 0;
774 
775 	if (!(features & NETIF_F_V6_CSUM))
776 		features &= ~NETIF_F_SG;
777 
778 	if (unlikely(skb_shinfo(skb)->gso_type &
779 		     ~(SKB_GSO_UDP |
780 		       SKB_GSO_DODGY |
781 		       SKB_GSO_TCP_ECN |
782 		       SKB_GSO_TCPV6 |
783 		       0)))
784 		goto out;
785 
786 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
787 		goto out;
788 
789 	ipv6h = ipv6_hdr(skb);
790 	__skb_pull(skb, sizeof(*ipv6h));
791 	segs = ERR_PTR(-EPROTONOSUPPORT);
792 
793 	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
794 	rcu_read_lock();
795 	ops = rcu_dereference(inet6_protos[proto]);
796 	if (likely(ops && ops->gso_segment)) {
797 		skb_reset_transport_header(skb);
798 		segs = ops->gso_segment(skb, features);
799 	}
800 	rcu_read_unlock();
801 
802 	if (IS_ERR(segs))
803 		goto out;
804 
805 	for (skb = segs; skb; skb = skb->next) {
806 		ipv6h = ipv6_hdr(skb);
807 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
808 					   sizeof(*ipv6h));
809 		if (proto == IPPROTO_UDP) {
810 			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
811 			fptr = (struct frag_hdr *)(skb_network_header(skb) +
812 				unfrag_ip6hlen);
813 			fptr->frag_off = htons(offset);
814 			if (skb->next != NULL)
815 				fptr->frag_off |= htons(IP6_MF);
816 			offset += (ntohs(ipv6h->payload_len) -
817 				   sizeof(struct frag_hdr));
818 		}
819 	}
820 
821 out:
822 	return segs;
823 }
824 
825 struct ipv6_gro_cb {
826 	struct napi_gro_cb napi;
827 	int proto;
828 };
829 
830 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
831 
832 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
833 					 struct sk_buff *skb)
834 {
835 	const struct inet6_protocol *ops;
836 	struct sk_buff **pp = NULL;
837 	struct sk_buff *p;
838 	struct ipv6hdr *iph;
839 	unsigned int nlen;
840 	unsigned int hlen;
841 	unsigned int off;
842 	int flush = 1;
843 	int proto;
844 	__wsum csum;
845 
846 	off = skb_gro_offset(skb);
847 	hlen = off + sizeof(*iph);
848 	iph = skb_gro_header_fast(skb, off);
849 	if (skb_gro_header_hard(skb, hlen)) {
850 		iph = skb_gro_header_slow(skb, hlen, off);
851 		if (unlikely(!iph))
852 			goto out;
853 	}
854 
855 	skb_gro_pull(skb, sizeof(*iph));
856 	skb_set_transport_header(skb, skb_gro_offset(skb));
857 
858 	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
859 
860 	rcu_read_lock();
861 	proto = iph->nexthdr;
862 	ops = rcu_dereference(inet6_protos[proto]);
863 	if (!ops || !ops->gro_receive) {
864 		__pskb_pull(skb, skb_gro_offset(skb));
865 		proto = ipv6_gso_pull_exthdrs(skb, proto);
866 		skb_gro_pull(skb, -skb_transport_offset(skb));
867 		skb_reset_transport_header(skb);
868 		__skb_push(skb, skb_gro_offset(skb));
869 
870 		ops = rcu_dereference(inet6_protos[proto]);
871 		if (!ops || !ops->gro_receive)
872 			goto out_unlock;
873 
874 		iph = ipv6_hdr(skb);
875 	}
876 
877 	IPV6_GRO_CB(skb)->proto = proto;
878 
879 	flush--;
880 	nlen = skb_network_header_len(skb);
881 
882 	for (p = *head; p; p = p->next) {
883 		struct ipv6hdr *iph2;
884 
885 		if (!NAPI_GRO_CB(p)->same_flow)
886 			continue;
887 
888 		iph2 = ipv6_hdr(p);
889 
890 		/* All fields must match except length. */
891 		if (nlen != skb_network_header_len(p) ||
892 		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
893 		    memcmp(&iph->nexthdr, &iph2->nexthdr,
894 			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
895 			NAPI_GRO_CB(p)->same_flow = 0;
896 			continue;
897 		}
898 
899 		NAPI_GRO_CB(p)->flush |= flush;
900 	}
901 
902 	NAPI_GRO_CB(skb)->flush |= flush;
903 
904 	csum = skb->csum;
905 	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
906 
907 	pp = ops->gro_receive(head, skb);
908 
909 	skb->csum = csum;
910 
911 out_unlock:
912 	rcu_read_unlock();
913 
914 out:
915 	NAPI_GRO_CB(skb)->flush |= flush;
916 
917 	return pp;
918 }
919 
920 static int ipv6_gro_complete(struct sk_buff *skb)
921 {
922 	const struct inet6_protocol *ops;
923 	struct ipv6hdr *iph = ipv6_hdr(skb);
924 	int err = -ENOSYS;
925 
926 	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
927 				 sizeof(*iph));
928 
929 	rcu_read_lock();
930 	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
931 	if (WARN_ON(!ops || !ops->gro_complete))
932 		goto out_unlock;
933 
934 	err = ops->gro_complete(skb);
935 
936 out_unlock:
937 	rcu_read_unlock();
938 
939 	return err;
940 }
941 
942 static struct packet_type ipv6_packet_type __read_mostly = {
943 	.type = cpu_to_be16(ETH_P_IPV6),
944 	.func = ipv6_rcv,
945 	.gso_send_check = ipv6_gso_send_check,
946 	.gso_segment = ipv6_gso_segment,
947 	.gro_receive = ipv6_gro_receive,
948 	.gro_complete = ipv6_gro_complete,
949 };
950 
951 static int __init ipv6_packet_init(void)
952 {
953 	dev_add_pack(&ipv6_packet_type);
954 	return 0;
955 }
956 
957 static void ipv6_packet_cleanup(void)
958 {
959 	dev_remove_pack(&ipv6_packet_type);
960 }
961 
962 static int __net_init ipv6_init_mibs(struct net *net)
963 {
964 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
965 			  sizeof(struct udp_mib),
966 			  __alignof__(struct udp_mib)) < 0)
967 		return -ENOMEM;
968 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
969 			  sizeof(struct udp_mib),
970 			  __alignof__(struct udp_mib)) < 0)
971 		goto err_udplite_mib;
972 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
973 			  sizeof(struct ipstats_mib),
974 			  __alignof__(struct ipstats_mib)) < 0)
975 		goto err_ip_mib;
976 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
977 			  sizeof(struct icmpv6_mib),
978 			  __alignof__(struct icmpv6_mib)) < 0)
979 		goto err_icmp_mib;
980 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
981 						GFP_KERNEL);
982 	if (!net->mib.icmpv6msg_statistics)
983 		goto err_icmpmsg_mib;
984 	return 0;
985 
986 err_icmpmsg_mib:
987 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
988 err_icmp_mib:
989 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
990 err_ip_mib:
991 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
992 err_udplite_mib:
993 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
994 	return -ENOMEM;
995 }
996 
997 static void ipv6_cleanup_mibs(struct net *net)
998 {
999 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1000 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1001 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1002 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1003 	kfree(net->mib.icmpv6msg_statistics);
1004 }
1005 
1006 static int __net_init inet6_net_init(struct net *net)
1007 {
1008 	int err = 0;
1009 
1010 	net->ipv6.sysctl.bindv6only = 0;
1011 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
1012 
1013 	err = ipv6_init_mibs(net);
1014 	if (err)
1015 		return err;
1016 #ifdef CONFIG_PROC_FS
1017 	err = udp6_proc_init(net);
1018 	if (err)
1019 		goto out;
1020 	err = tcp6_proc_init(net);
1021 	if (err)
1022 		goto proc_tcp6_fail;
1023 	err = ac6_proc_init(net);
1024 	if (err)
1025 		goto proc_ac6_fail;
1026 #endif
1027 	return err;
1028 
1029 #ifdef CONFIG_PROC_FS
1030 proc_ac6_fail:
1031 	tcp6_proc_exit(net);
1032 proc_tcp6_fail:
1033 	udp6_proc_exit(net);
1034 out:
1035 	ipv6_cleanup_mibs(net);
1036 	return err;
1037 #endif
1038 }
1039 
1040 static void __net_exit inet6_net_exit(struct net *net)
1041 {
1042 #ifdef CONFIG_PROC_FS
1043 	udp6_proc_exit(net);
1044 	tcp6_proc_exit(net);
1045 	ac6_proc_exit(net);
1046 #endif
1047 	ipv6_cleanup_mibs(net);
1048 }
1049 
1050 static struct pernet_operations inet6_net_ops = {
1051 	.init = inet6_net_init,
1052 	.exit = inet6_net_exit,
1053 };
1054 
1055 static int __init inet6_init(void)
1056 {
1057 	struct sk_buff *dummy_skb;
1058 	struct list_head *r;
1059 	int err = 0;
1060 
1061 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1062 
1063 	/* Register the socket-side information for inet6_create.  */
1064 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1065 		INIT_LIST_HEAD(r);
1066 
1067 	if (disable_ipv6_mod) {
1068 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1069 		goto out;
1070 	}
1071 
1072 	err = proto_register(&tcpv6_prot, 1);
1073 	if (err)
1074 		goto out;
1075 
1076 	err = proto_register(&udpv6_prot, 1);
1077 	if (err)
1078 		goto out_unregister_tcp_proto;
1079 
1080 	err = proto_register(&udplitev6_prot, 1);
1081 	if (err)
1082 		goto out_unregister_udp_proto;
1083 
1084 	err = proto_register(&rawv6_prot, 1);
1085 	if (err)
1086 		goto out_unregister_udplite_proto;
1087 
1088 
1089 	/* We MUST register RAW sockets before we create the ICMP6,
1090 	 * IGMP6, or NDISC control sockets.
1091 	 */
1092 	err = rawv6_init();
1093 	if (err)
1094 		goto out_unregister_raw_proto;
1095 
1096 	/* Register the family here so that the init calls below will
1097 	 * be able to create sockets. (?? is this dangerous ??)
1098 	 */
1099 	err = sock_register(&inet6_family_ops);
1100 	if (err)
1101 		goto out_sock_register_fail;
1102 
1103 	tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1104 
1105 	/*
1106 	 *	ipngwg API draft makes clear that the correct semantics
1107 	 *	for TCP and UDP is to consider one TCP and UDP instance
1108 	 *	in a host available by both INET and INET6 APIs and
1109 	 *	able to communicate via both network protocols.
1110 	 */
1111 
1112 	err = register_pernet_subsys(&inet6_net_ops);
1113 	if (err)
1114 		goto register_pernet_fail;
1115 	err = icmpv6_init();
1116 	if (err)
1117 		goto icmp_fail;
1118 	err = ip6_mr_init();
1119 	if (err)
1120 		goto ipmr_fail;
1121 	err = ndisc_init();
1122 	if (err)
1123 		goto ndisc_fail;
1124 	err = igmp6_init();
1125 	if (err)
1126 		goto igmp_fail;
1127 	err = ipv6_netfilter_init();
1128 	if (err)
1129 		goto netfilter_fail;
1130 	/* Create /proc/foo6 entries. */
1131 #ifdef CONFIG_PROC_FS
1132 	err = -ENOMEM;
1133 	if (raw6_proc_init())
1134 		goto proc_raw6_fail;
1135 	if (udplite6_proc_init())
1136 		goto proc_udplite6_fail;
1137 	if (ipv6_misc_proc_init())
1138 		goto proc_misc6_fail;
1139 	if (if6_proc_init())
1140 		goto proc_if6_fail;
1141 #endif
1142 	err = ip6_route_init();
1143 	if (err)
1144 		goto ip6_route_fail;
1145 	err = ip6_flowlabel_init();
1146 	if (err)
1147 		goto ip6_flowlabel_fail;
1148 	err = addrconf_init();
1149 	if (err)
1150 		goto addrconf_fail;
1151 
1152 	/* Init v6 extension headers. */
1153 	err = ipv6_exthdrs_init();
1154 	if (err)
1155 		goto ipv6_exthdrs_fail;
1156 
1157 	err = ipv6_frag_init();
1158 	if (err)
1159 		goto ipv6_frag_fail;
1160 
1161 	/* Init v6 transport protocols. */
1162 	err = udpv6_init();
1163 	if (err)
1164 		goto udpv6_fail;
1165 
1166 	err = udplitev6_init();
1167 	if (err)
1168 		goto udplitev6_fail;
1169 
1170 	err = tcpv6_init();
1171 	if (err)
1172 		goto tcpv6_fail;
1173 
1174 	err = ipv6_packet_init();
1175 	if (err)
1176 		goto ipv6_packet_fail;
1177 
1178 #ifdef CONFIG_SYSCTL
1179 	err = ipv6_sysctl_register();
1180 	if (err)
1181 		goto sysctl_fail;
1182 #endif
1183 out:
1184 	return err;
1185 
1186 #ifdef CONFIG_SYSCTL
1187 sysctl_fail:
1188 	ipv6_packet_cleanup();
1189 #endif
1190 ipv6_packet_fail:
1191 	tcpv6_exit();
1192 tcpv6_fail:
1193 	udplitev6_exit();
1194 udplitev6_fail:
1195 	udpv6_exit();
1196 udpv6_fail:
1197 	ipv6_frag_exit();
1198 ipv6_frag_fail:
1199 	ipv6_exthdrs_exit();
1200 ipv6_exthdrs_fail:
1201 	addrconf_cleanup();
1202 addrconf_fail:
1203 	ip6_flowlabel_cleanup();
1204 ip6_flowlabel_fail:
1205 	ip6_route_cleanup();
1206 ip6_route_fail:
1207 #ifdef CONFIG_PROC_FS
1208 	if6_proc_exit();
1209 proc_if6_fail:
1210 	ipv6_misc_proc_exit();
1211 proc_misc6_fail:
1212 	udplite6_proc_exit();
1213 proc_udplite6_fail:
1214 	raw6_proc_exit();
1215 proc_raw6_fail:
1216 #endif
1217 	ipv6_netfilter_fini();
1218 netfilter_fail:
1219 	igmp6_cleanup();
1220 igmp_fail:
1221 	ndisc_cleanup();
1222 ndisc_fail:
1223 	ip6_mr_cleanup();
1224 ipmr_fail:
1225 	icmpv6_cleanup();
1226 icmp_fail:
1227 	unregister_pernet_subsys(&inet6_net_ops);
1228 register_pernet_fail:
1229 	sock_unregister(PF_INET6);
1230 	rtnl_unregister_all(PF_INET6);
1231 out_sock_register_fail:
1232 	rawv6_exit();
1233 out_unregister_raw_proto:
1234 	proto_unregister(&rawv6_prot);
1235 out_unregister_udplite_proto:
1236 	proto_unregister(&udplitev6_prot);
1237 out_unregister_udp_proto:
1238 	proto_unregister(&udpv6_prot);
1239 out_unregister_tcp_proto:
1240 	proto_unregister(&tcpv6_prot);
1241 	goto out;
1242 }
1243 module_init(inet6_init);
1244 
1245 static void __exit inet6_exit(void)
1246 {
1247 	if (disable_ipv6_mod)
1248 		return;
1249 
1250 	/* First of all disallow new sockets creation. */
1251 	sock_unregister(PF_INET6);
1252 	/* Disallow any further netlink messages */
1253 	rtnl_unregister_all(PF_INET6);
1254 
1255 	udpv6_exit();
1256 	udplitev6_exit();
1257 	tcpv6_exit();
1258 
1259 	/* Cleanup code parts. */
1260 	ipv6_packet_cleanup();
1261 	ipv6_frag_exit();
1262 	ipv6_exthdrs_exit();
1263 	addrconf_cleanup();
1264 	ip6_flowlabel_cleanup();
1265 	ip6_route_cleanup();
1266 #ifdef CONFIG_PROC_FS
1267 
1268 	/* Cleanup code parts. */
1269 	if6_proc_exit();
1270 	ipv6_misc_proc_exit();
1271 	udplite6_proc_exit();
1272 	raw6_proc_exit();
1273 #endif
1274 	ipv6_netfilter_fini();
1275 	igmp6_cleanup();
1276 	ndisc_cleanup();
1277 	ip6_mr_cleanup();
1278 	icmpv6_cleanup();
1279 	rawv6_exit();
1280 
1281 	unregister_pernet_subsys(&inet6_net_ops);
1282 	proto_unregister(&rawv6_prot);
1283 	proto_unregister(&udplitev6_prot);
1284 	proto_unregister(&udpv6_prot);
1285 	proto_unregister(&tcpv6_prot);
1286 
1287 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1288 }
1289 module_exit(inet6_exit);
1290 
1291 MODULE_ALIAS_NETPROTO(PF_INET6);
1292