xref: /linux/net/ipv6/ipv6_sockglue.c (revision c537b994505099b7197e7d3125b942ecbcc51eb6)
1 /*
2  *	IPv6 BSD socket options interface
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Based on linux/net/ipv4/ip_sockglue.c
9  *
10  *	$Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
11  *
12  *	This program is free software; you can redistribute it and/or
13  *      modify it under the terms of the GNU General Public License
14  *      as published by the Free Software Foundation; either version
15  *      2 of the License, or (at your option) any later version.
16  *
17  *	FIXME: Make the setsockopt code POSIX compliant: That is
18  *
19  *	o	Return -EINVAL for setsockopt of short lengths
20  *	o	Truncate getsockopt returns
21  *	o	Return an optlen of the truncated length if need be
22  *
23  *	Changes:
24  *	David L Stevens <dlstevens@us.ibm.com>:
25  *		- added multicast source filtering API for MLDv2
26  */
27 
28 #include <linux/module.h>
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/netdevice.h>
37 #include <linux/if_arp.h>
38 #include <linux/init.h>
39 #include <linux/sysctl.h>
40 #include <linux/netfilter.h>
41 
42 #include <net/sock.h>
43 #include <net/snmp.h>
44 #include <net/ipv6.h>
45 #include <net/ndisc.h>
46 #include <net/protocol.h>
47 #include <net/transp_v6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/inet_common.h>
51 #include <net/tcp.h>
52 #include <net/udp.h>
53 #include <net/udplite.h>
54 #include <net/xfrm.h>
55 
56 #include <asm/uaccess.h>
57 
58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
59 
60 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
61 						    int proto)
62 {
63 	struct inet6_protocol *ops = NULL;
64 
65 	for (;;) {
66 		struct ipv6_opt_hdr *opth;
67 		int len;
68 
69 		if (proto != NEXTHDR_HOP) {
70 			ops = rcu_dereference(inet6_protos[proto]);
71 
72 			if (unlikely(!ops))
73 				break;
74 
75 			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
76 				break;
77 		}
78 
79 		if (unlikely(!pskb_may_pull(skb, 8)))
80 			break;
81 
82 		opth = (void *)skb->data;
83 		len = opth->hdrlen * 8 + 8;
84 
85 		if (unlikely(!pskb_may_pull(skb, len)))
86 			break;
87 
88 		proto = opth->nexthdr;
89 		__skb_pull(skb, len);
90 	}
91 
92 	return ops;
93 }
94 
95 static int ipv6_gso_send_check(struct sk_buff *skb)
96 {
97 	struct ipv6hdr *ipv6h;
98 	struct inet6_protocol *ops;
99 	int err = -EINVAL;
100 
101 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
102 		goto out;
103 
104 	ipv6h = skb->nh.ipv6h;
105 	__skb_pull(skb, sizeof(*ipv6h));
106 	err = -EPROTONOSUPPORT;
107 
108 	rcu_read_lock();
109 	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
110 	if (likely(ops && ops->gso_send_check)) {
111 		skb->h.raw = skb->data;
112 		err = ops->gso_send_check(skb);
113 	}
114 	rcu_read_unlock();
115 
116 out:
117 	return err;
118 }
119 
120 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
121 {
122 	struct sk_buff *segs = ERR_PTR(-EINVAL);
123 	struct ipv6hdr *ipv6h;
124 	struct inet6_protocol *ops;
125 
126 	if (!(features & NETIF_F_HW_CSUM))
127 		features &= ~NETIF_F_SG;
128 
129 	if (unlikely(skb_shinfo(skb)->gso_type &
130 		     ~(SKB_GSO_UDP |
131 		       SKB_GSO_DODGY |
132 		       SKB_GSO_TCP_ECN |
133 		       SKB_GSO_TCPV6 |
134 		       0)))
135 		goto out;
136 
137 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
138 		goto out;
139 
140 	ipv6h = skb->nh.ipv6h;
141 	__skb_pull(skb, sizeof(*ipv6h));
142 	segs = ERR_PTR(-EPROTONOSUPPORT);
143 
144 	rcu_read_lock();
145 	ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
146 	if (likely(ops && ops->gso_segment)) {
147 		skb->h.raw = skb->data;
148 		segs = ops->gso_segment(skb, features);
149 	}
150 	rcu_read_unlock();
151 
152 	if (unlikely(IS_ERR(segs)))
153 		goto out;
154 
155 	for (skb = segs; skb; skb = skb->next) {
156 		ipv6h = skb->nh.ipv6h;
157 		ipv6h->payload_len = htons(skb->len - skb->mac_len -
158 					   sizeof(*ipv6h));
159 	}
160 
161 out:
162 	return segs;
163 }
164 
165 static struct packet_type ipv6_packet_type = {
166 	.type = __constant_htons(ETH_P_IPV6),
167 	.func = ipv6_rcv,
168 	.gso_send_check = ipv6_gso_send_check,
169 	.gso_segment = ipv6_gso_segment,
170 };
171 
172 struct ip6_ra_chain *ip6_ra_chain;
173 DEFINE_RWLOCK(ip6_ra_lock);
174 
175 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
176 {
177 	struct ip6_ra_chain *ra, *new_ra, **rap;
178 
179 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
180 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
181 		return -EINVAL;
182 
183 	new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
184 
185 	write_lock_bh(&ip6_ra_lock);
186 	for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
187 		if (ra->sk == sk) {
188 			if (sel>=0) {
189 				write_unlock_bh(&ip6_ra_lock);
190 				kfree(new_ra);
191 				return -EADDRINUSE;
192 			}
193 
194 			*rap = ra->next;
195 			write_unlock_bh(&ip6_ra_lock);
196 
197 			if (ra->destructor)
198 				ra->destructor(sk);
199 			sock_put(sk);
200 			kfree(ra);
201 			return 0;
202 		}
203 	}
204 	if (new_ra == NULL) {
205 		write_unlock_bh(&ip6_ra_lock);
206 		return -ENOBUFS;
207 	}
208 	new_ra->sk = sk;
209 	new_ra->sel = sel;
210 	new_ra->destructor = destructor;
211 	new_ra->next = ra;
212 	*rap = new_ra;
213 	sock_hold(sk);
214 	write_unlock_bh(&ip6_ra_lock);
215 	return 0;
216 }
217 
218 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
219 		    char __user *optval, int optlen)
220 {
221 	struct ipv6_pinfo *np = inet6_sk(sk);
222 	int val, valbool;
223 	int retv = -ENOPROTOOPT;
224 
225 	if (optval == NULL)
226 		val=0;
227 	else if (get_user(val, (int __user *) optval))
228 		return -EFAULT;
229 
230 	valbool = (val!=0);
231 
232 	lock_sock(sk);
233 
234 	switch (optname) {
235 
236 	case IPV6_ADDRFORM:
237 		if (val == PF_INET) {
238 			struct ipv6_txoptions *opt;
239 			struct sk_buff *pktopt;
240 
241 			if (sk->sk_protocol != IPPROTO_UDP &&
242 			    sk->sk_protocol != IPPROTO_UDPLITE &&
243 			    sk->sk_protocol != IPPROTO_TCP)
244 				break;
245 
246 			if (sk->sk_state != TCP_ESTABLISHED) {
247 				retv = -ENOTCONN;
248 				break;
249 			}
250 
251 			if (ipv6_only_sock(sk) ||
252 			    !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
253 				retv = -EADDRNOTAVAIL;
254 				break;
255 			}
256 
257 			fl6_free_socklist(sk);
258 			ipv6_sock_mc_close(sk);
259 
260 			/*
261 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
262 			 * remove it from the refcnt debug socks count in the
263 			 * original family...
264 			 */
265 			sk_refcnt_debug_dec(sk);
266 
267 			if (sk->sk_protocol == IPPROTO_TCP) {
268 				struct inet_connection_sock *icsk = inet_csk(sk);
269 
270 				local_bh_disable();
271 				sock_prot_dec_use(sk->sk_prot);
272 				sock_prot_inc_use(&tcp_prot);
273 				local_bh_enable();
274 				sk->sk_prot = &tcp_prot;
275 				icsk->icsk_af_ops = &ipv4_specific;
276 				sk->sk_socket->ops = &inet_stream_ops;
277 				sk->sk_family = PF_INET;
278 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
279 			} else {
280 				struct proto *prot = &udp_prot;
281 
282 				if (sk->sk_protocol == IPPROTO_UDPLITE)
283 					prot = &udplite_prot;
284 				local_bh_disable();
285 				sock_prot_dec_use(sk->sk_prot);
286 				sock_prot_inc_use(prot);
287 				local_bh_enable();
288 				sk->sk_prot = prot;
289 				sk->sk_socket->ops = &inet_dgram_ops;
290 				sk->sk_family = PF_INET;
291 			}
292 			opt = xchg(&np->opt, NULL);
293 			if (opt)
294 				sock_kfree_s(sk, opt, opt->tot_len);
295 			pktopt = xchg(&np->pktoptions, NULL);
296 			if (pktopt)
297 				kfree_skb(pktopt);
298 
299 			sk->sk_destruct = inet_sock_destruct;
300 			/*
301 			 * ... and add it to the refcnt debug socks count
302 			 * in the new family. -acme
303 			 */
304 			sk_refcnt_debug_inc(sk);
305 			module_put(THIS_MODULE);
306 			retv = 0;
307 			break;
308 		}
309 		goto e_inval;
310 
311 	case IPV6_V6ONLY:
312 		if (inet_sk(sk)->num)
313 			goto e_inval;
314 		np->ipv6only = valbool;
315 		retv = 0;
316 		break;
317 
318 	case IPV6_RECVPKTINFO:
319 		np->rxopt.bits.rxinfo = valbool;
320 		retv = 0;
321 		break;
322 
323 	case IPV6_2292PKTINFO:
324 		np->rxopt.bits.rxoinfo = valbool;
325 		retv = 0;
326 		break;
327 
328 	case IPV6_RECVHOPLIMIT:
329 		np->rxopt.bits.rxhlim = valbool;
330 		retv = 0;
331 		break;
332 
333 	case IPV6_2292HOPLIMIT:
334 		np->rxopt.bits.rxohlim = valbool;
335 		retv = 0;
336 		break;
337 
338 	case IPV6_RECVRTHDR:
339 		if (val < 0 || val > 2)
340 			goto e_inval;
341 		np->rxopt.bits.srcrt = val;
342 		retv = 0;
343 		break;
344 
345 	case IPV6_2292RTHDR:
346 		if (val < 0 || val > 2)
347 			goto e_inval;
348 		np->rxopt.bits.osrcrt = val;
349 		retv = 0;
350 		break;
351 
352 	case IPV6_RECVHOPOPTS:
353 		np->rxopt.bits.hopopts = valbool;
354 		retv = 0;
355 		break;
356 
357 	case IPV6_2292HOPOPTS:
358 		np->rxopt.bits.ohopopts = valbool;
359 		retv = 0;
360 		break;
361 
362 	case IPV6_RECVDSTOPTS:
363 		np->rxopt.bits.dstopts = valbool;
364 		retv = 0;
365 		break;
366 
367 	case IPV6_2292DSTOPTS:
368 		np->rxopt.bits.odstopts = valbool;
369 		retv = 0;
370 		break;
371 
372 	case IPV6_TCLASS:
373 		if (val < -1 || val > 0xff)
374 			goto e_inval;
375 		np->tclass = val;
376 		retv = 0;
377 		break;
378 
379 	case IPV6_RECVTCLASS:
380 		np->rxopt.bits.rxtclass = valbool;
381 		retv = 0;
382 		break;
383 
384 	case IPV6_FLOWINFO:
385 		np->rxopt.bits.rxflow = valbool;
386 		retv = 0;
387 		break;
388 
389 	case IPV6_HOPOPTS:
390 	case IPV6_RTHDRDSTOPTS:
391 	case IPV6_RTHDR:
392 	case IPV6_DSTOPTS:
393 	{
394 		struct ipv6_txoptions *opt;
395 		if (optlen == 0)
396 			optval = NULL;
397 
398 		/* hop-by-hop / destination options are privileged option */
399 		retv = -EPERM;
400 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
401 			break;
402 
403 		retv = -EINVAL;
404 		if (optlen & 0x7 || optlen > 8 * 255)
405 			break;
406 
407 		opt = ipv6_renew_options(sk, np->opt, optname,
408 					 (struct ipv6_opt_hdr __user *)optval,
409 					 optlen);
410 		if (IS_ERR(opt)) {
411 			retv = PTR_ERR(opt);
412 			break;
413 		}
414 
415 		/* routing header option needs extra check */
416 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
417 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
418 			switch (rthdr->type) {
419 			case IPV6_SRCRT_TYPE_0:
420 #ifdef CONFIG_IPV6_MIP6
421 			case IPV6_SRCRT_TYPE_2:
422 #endif
423 				break;
424 			default:
425 				goto sticky_done;
426 			}
427 
428 			if ((rthdr->hdrlen & 1) ||
429 			    (rthdr->hdrlen >> 1) != rthdr->segments_left)
430 				goto sticky_done;
431 		}
432 
433 		retv = 0;
434 		if (inet_sk(sk)->is_icsk) {
435 			if (opt) {
436 				struct inet_connection_sock *icsk = inet_csk(sk);
437 				if (!((1 << sk->sk_state) &
438 				      (TCPF_LISTEN | TCPF_CLOSE))
439 				    && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
440 					icsk->icsk_ext_hdr_len =
441 						opt->opt_flen + opt->opt_nflen;
442 					icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
443 				}
444 			}
445 			opt = xchg(&np->opt, opt);
446 			sk_dst_reset(sk);
447 		} else {
448 			write_lock(&sk->sk_dst_lock);
449 			opt = xchg(&np->opt, opt);
450 			write_unlock(&sk->sk_dst_lock);
451 			sk_dst_reset(sk);
452 		}
453 sticky_done:
454 		if (opt)
455 			sock_kfree_s(sk, opt, opt->tot_len);
456 		break;
457 	}
458 
459 	case IPV6_2292PKTOPTIONS:
460 	{
461 		struct ipv6_txoptions *opt = NULL;
462 		struct msghdr msg;
463 		struct flowi fl;
464 		int junk;
465 
466 		fl.fl6_flowlabel = 0;
467 		fl.oif = sk->sk_bound_dev_if;
468 
469 		if (optlen == 0)
470 			goto update;
471 
472 		/* 1K is probably excessive
473 		 * 1K is surely not enough, 2K per standard header is 16K.
474 		 */
475 		retv = -EINVAL;
476 		if (optlen > 64*1024)
477 			break;
478 
479 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
480 		retv = -ENOBUFS;
481 		if (opt == NULL)
482 			break;
483 
484 		memset(opt, 0, sizeof(*opt));
485 		opt->tot_len = sizeof(*opt) + optlen;
486 		retv = -EFAULT;
487 		if (copy_from_user(opt+1, optval, optlen))
488 			goto done;
489 
490 		msg.msg_controllen = optlen;
491 		msg.msg_control = (void*)(opt+1);
492 
493 		retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
494 		if (retv)
495 			goto done;
496 update:
497 		retv = 0;
498 		if (inet_sk(sk)->is_icsk) {
499 			if (opt) {
500 				struct inet_connection_sock *icsk = inet_csk(sk);
501 				if (!((1 << sk->sk_state) &
502 				      (TCPF_LISTEN | TCPF_CLOSE))
503 				    && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
504 					icsk->icsk_ext_hdr_len =
505 						opt->opt_flen + opt->opt_nflen;
506 					icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
507 				}
508 			}
509 			opt = xchg(&np->opt, opt);
510 			sk_dst_reset(sk);
511 		} else {
512 			write_lock(&sk->sk_dst_lock);
513 			opt = xchg(&np->opt, opt);
514 			write_unlock(&sk->sk_dst_lock);
515 			sk_dst_reset(sk);
516 		}
517 
518 done:
519 		if (opt)
520 			sock_kfree_s(sk, opt, opt->tot_len);
521 		break;
522 	}
523 	case IPV6_UNICAST_HOPS:
524 		if (val > 255 || val < -1)
525 			goto e_inval;
526 		np->hop_limit = val;
527 		retv = 0;
528 		break;
529 
530 	case IPV6_MULTICAST_HOPS:
531 		if (sk->sk_type == SOCK_STREAM)
532 			goto e_inval;
533 		if (val > 255 || val < -1)
534 			goto e_inval;
535 		np->mcast_hops = val;
536 		retv = 0;
537 		break;
538 
539 	case IPV6_MULTICAST_LOOP:
540 		np->mc_loop = valbool;
541 		retv = 0;
542 		break;
543 
544 	case IPV6_MULTICAST_IF:
545 		if (sk->sk_type == SOCK_STREAM)
546 			goto e_inval;
547 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
548 			goto e_inval;
549 
550 		if (__dev_get_by_index(val) == NULL) {
551 			retv = -ENODEV;
552 			break;
553 		}
554 		np->mcast_oif = val;
555 		retv = 0;
556 		break;
557 	case IPV6_ADD_MEMBERSHIP:
558 	case IPV6_DROP_MEMBERSHIP:
559 	{
560 		struct ipv6_mreq mreq;
561 
562 		retv = -EFAULT;
563 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
564 			break;
565 
566 		if (optname == IPV6_ADD_MEMBERSHIP)
567 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
568 		else
569 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
570 		break;
571 	}
572 	case IPV6_JOIN_ANYCAST:
573 	case IPV6_LEAVE_ANYCAST:
574 	{
575 		struct ipv6_mreq mreq;
576 
577 		if (optlen != sizeof(struct ipv6_mreq))
578 			goto e_inval;
579 
580 		retv = -EFAULT;
581 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
582 			break;
583 
584 		if (optname == IPV6_JOIN_ANYCAST)
585 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
586 		else
587 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
588 		break;
589 	}
590 	case MCAST_JOIN_GROUP:
591 	case MCAST_LEAVE_GROUP:
592 	{
593 		struct group_req greq;
594 		struct sockaddr_in6 *psin6;
595 
596 		retv = -EFAULT;
597 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
598 			break;
599 		if (greq.gr_group.ss_family != AF_INET6) {
600 			retv = -EADDRNOTAVAIL;
601 			break;
602 		}
603 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
604 		if (optname == MCAST_JOIN_GROUP)
605 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
606 				&psin6->sin6_addr);
607 		else
608 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
609 				&psin6->sin6_addr);
610 		break;
611 	}
612 	case MCAST_JOIN_SOURCE_GROUP:
613 	case MCAST_LEAVE_SOURCE_GROUP:
614 	case MCAST_BLOCK_SOURCE:
615 	case MCAST_UNBLOCK_SOURCE:
616 	{
617 		struct group_source_req greqs;
618 		int omode, add;
619 
620 		if (optlen != sizeof(struct group_source_req))
621 			goto e_inval;
622 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
623 			retv = -EFAULT;
624 			break;
625 		}
626 		if (greqs.gsr_group.ss_family != AF_INET6 ||
627 		    greqs.gsr_source.ss_family != AF_INET6) {
628 			retv = -EADDRNOTAVAIL;
629 			break;
630 		}
631 		if (optname == MCAST_BLOCK_SOURCE) {
632 			omode = MCAST_EXCLUDE;
633 			add = 1;
634 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
635 			omode = MCAST_EXCLUDE;
636 			add = 0;
637 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
638 			struct sockaddr_in6 *psin6;
639 
640 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
641 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
642 				&psin6->sin6_addr);
643 			/* prior join w/ different source is ok */
644 			if (retv && retv != -EADDRINUSE)
645 				break;
646 			omode = MCAST_INCLUDE;
647 			add = 1;
648 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
649 			omode = MCAST_INCLUDE;
650 			add = 0;
651 		}
652 		retv = ip6_mc_source(add, omode, sk, &greqs);
653 		break;
654 	}
655 	case MCAST_MSFILTER:
656 	{
657 		extern int sysctl_mld_max_msf;
658 		struct group_filter *gsf;
659 
660 		if (optlen < GROUP_FILTER_SIZE(0))
661 			goto e_inval;
662 		if (optlen > sysctl_optmem_max) {
663 			retv = -ENOBUFS;
664 			break;
665 		}
666 		gsf = kmalloc(optlen,GFP_KERNEL);
667 		if (gsf == 0) {
668 			retv = -ENOBUFS;
669 			break;
670 		}
671 		retv = -EFAULT;
672 		if (copy_from_user(gsf, optval, optlen)) {
673 			kfree(gsf);
674 			break;
675 		}
676 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
677 		if (gsf->gf_numsrc >= 0x1ffffffU ||
678 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
679 			kfree(gsf);
680 			retv = -ENOBUFS;
681 			break;
682 		}
683 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
684 			kfree(gsf);
685 			retv = -EINVAL;
686 			break;
687 		}
688 		retv = ip6_mc_msfilter(sk, gsf);
689 		kfree(gsf);
690 
691 		break;
692 	}
693 	case IPV6_ROUTER_ALERT:
694 		retv = ip6_ra_control(sk, val, NULL);
695 		break;
696 	case IPV6_MTU_DISCOVER:
697 		if (val<0 || val>2)
698 			goto e_inval;
699 		np->pmtudisc = val;
700 		retv = 0;
701 		break;
702 	case IPV6_MTU:
703 		if (val && val < IPV6_MIN_MTU)
704 			goto e_inval;
705 		np->frag_size = val;
706 		retv = 0;
707 		break;
708 	case IPV6_RECVERR:
709 		np->recverr = valbool;
710 		if (!val)
711 			skb_queue_purge(&sk->sk_error_queue);
712 		retv = 0;
713 		break;
714 	case IPV6_FLOWINFO_SEND:
715 		np->sndflow = valbool;
716 		retv = 0;
717 		break;
718 	case IPV6_FLOWLABEL_MGR:
719 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
720 		break;
721 	case IPV6_IPSEC_POLICY:
722 	case IPV6_XFRM_POLICY:
723 		retv = -EPERM;
724 		if (!capable(CAP_NET_ADMIN))
725 			break;
726 		retv = xfrm_user_policy(sk, optname, optval, optlen);
727 		break;
728 
729 	}
730 	release_sock(sk);
731 
732 	return retv;
733 
734 e_inval:
735 	release_sock(sk);
736 	return -EINVAL;
737 }
738 
739 int ipv6_setsockopt(struct sock *sk, int level, int optname,
740 		    char __user *optval, int optlen)
741 {
742 	int err;
743 
744 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
745 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
746 
747 	if (level != SOL_IPV6)
748 		return -ENOPROTOOPT;
749 
750 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
751 #ifdef CONFIG_NETFILTER
752 	/* we need to exclude all possible ENOPROTOOPTs except default case */
753 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
754 			optname != IPV6_XFRM_POLICY) {
755 		lock_sock(sk);
756 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
757 				optlen);
758 		release_sock(sk);
759 	}
760 #endif
761 	return err;
762 }
763 
764 
765 #ifdef CONFIG_COMPAT
766 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
767 			   char __user *optval, int optlen)
768 {
769 	int err;
770 
771 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
772 		if (udp_prot.compat_setsockopt != NULL)
773 			return udp_prot.compat_setsockopt(sk, level, optname,
774 							  optval, optlen);
775 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
776 	}
777 
778 	if (level != SOL_IPV6)
779 		return -ENOPROTOOPT;
780 
781 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
782 #ifdef CONFIG_NETFILTER
783 	/* we need to exclude all possible ENOPROTOOPTs except default case */
784 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
785 	    optname != IPV6_XFRM_POLICY) {
786 		lock_sock(sk);
787 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
788 					   optval, optlen);
789 		release_sock(sk);
790 	}
791 #endif
792 	return err;
793 }
794 
795 EXPORT_SYMBOL(compat_ipv6_setsockopt);
796 #endif
797 
798 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
799 				  char __user *optval, int len)
800 {
801 	struct ipv6_opt_hdr *hdr;
802 
803 	if (!opt || !opt->hopopt)
804 		return 0;
805 	hdr = opt->hopopt;
806 
807 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
808 	if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
809 		return -EFAULT;
810 	return len;
811 }
812 
813 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
814 		    char __user *optval, int __user *optlen)
815 {
816 	struct ipv6_pinfo *np = inet6_sk(sk);
817 	int len;
818 	int val;
819 
820 	if (get_user(len, optlen))
821 		return -EFAULT;
822 	switch (optname) {
823 	case IPV6_ADDRFORM:
824 		if (sk->sk_protocol != IPPROTO_UDP &&
825 		    sk->sk_protocol != IPPROTO_UDPLITE &&
826 		    sk->sk_protocol != IPPROTO_TCP)
827 			return -EINVAL;
828 		if (sk->sk_state != TCP_ESTABLISHED)
829 			return -ENOTCONN;
830 		val = sk->sk_family;
831 		break;
832 	case MCAST_MSFILTER:
833 	{
834 		struct group_filter gsf;
835 		int err;
836 
837 		if (len < GROUP_FILTER_SIZE(0))
838 			return -EINVAL;
839 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
840 			return -EFAULT;
841 		lock_sock(sk);
842 		err = ip6_mc_msfget(sk, &gsf,
843 			(struct group_filter __user *)optval, optlen);
844 		release_sock(sk);
845 		return err;
846 	}
847 
848 	case IPV6_2292PKTOPTIONS:
849 	{
850 		struct msghdr msg;
851 		struct sk_buff *skb;
852 
853 		if (sk->sk_type != SOCK_STREAM)
854 			return -ENOPROTOOPT;
855 
856 		msg.msg_control = optval;
857 		msg.msg_controllen = len;
858 		msg.msg_flags = 0;
859 
860 		lock_sock(sk);
861 		skb = np->pktoptions;
862 		if (skb)
863 			atomic_inc(&skb->users);
864 		release_sock(sk);
865 
866 		if (skb) {
867 			int err = datagram_recv_ctl(sk, &msg, skb);
868 			kfree_skb(skb);
869 			if (err)
870 				return err;
871 		} else {
872 			if (np->rxopt.bits.rxinfo) {
873 				struct in6_pktinfo src_info;
874 				src_info.ipi6_ifindex = np->mcast_oif;
875 				ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
876 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
877 			}
878 			if (np->rxopt.bits.rxhlim) {
879 				int hlim = np->mcast_hops;
880 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
881 			}
882 			if (np->rxopt.bits.rxoinfo) {
883 				struct in6_pktinfo src_info;
884 				src_info.ipi6_ifindex = np->mcast_oif;
885 				ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
886 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
887 			}
888 			if (np->rxopt.bits.rxohlim) {
889 				int hlim = np->mcast_hops;
890 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
891 			}
892 		}
893 		len -= msg.msg_controllen;
894 		return put_user(len, optlen);
895 	}
896 	case IPV6_MTU:
897 	{
898 		struct dst_entry *dst;
899 		val = 0;
900 		lock_sock(sk);
901 		dst = sk_dst_get(sk);
902 		if (dst) {
903 			val = dst_mtu(dst);
904 			dst_release(dst);
905 		}
906 		release_sock(sk);
907 		if (!val)
908 			return -ENOTCONN;
909 		break;
910 	}
911 
912 	case IPV6_V6ONLY:
913 		val = np->ipv6only;
914 		break;
915 
916 	case IPV6_RECVPKTINFO:
917 		val = np->rxopt.bits.rxinfo;
918 		break;
919 
920 	case IPV6_2292PKTINFO:
921 		val = np->rxopt.bits.rxoinfo;
922 		break;
923 
924 	case IPV6_RECVHOPLIMIT:
925 		val = np->rxopt.bits.rxhlim;
926 		break;
927 
928 	case IPV6_2292HOPLIMIT:
929 		val = np->rxopt.bits.rxohlim;
930 		break;
931 
932 	case IPV6_RECVRTHDR:
933 		val = np->rxopt.bits.srcrt;
934 		break;
935 
936 	case IPV6_2292RTHDR:
937 		val = np->rxopt.bits.osrcrt;
938 		break;
939 
940 	case IPV6_HOPOPTS:
941 	case IPV6_RTHDRDSTOPTS:
942 	case IPV6_RTHDR:
943 	case IPV6_DSTOPTS:
944 	{
945 
946 		lock_sock(sk);
947 		len = ipv6_getsockopt_sticky(sk, np->opt,
948 					     optval, len);
949 		release_sock(sk);
950 		return put_user(len, optlen);
951 	}
952 
953 	case IPV6_RECVHOPOPTS:
954 		val = np->rxopt.bits.hopopts;
955 		break;
956 
957 	case IPV6_2292HOPOPTS:
958 		val = np->rxopt.bits.ohopopts;
959 		break;
960 
961 	case IPV6_RECVDSTOPTS:
962 		val = np->rxopt.bits.dstopts;
963 		break;
964 
965 	case IPV6_2292DSTOPTS:
966 		val = np->rxopt.bits.odstopts;
967 		break;
968 
969 	case IPV6_TCLASS:
970 		val = np->tclass;
971 		if (val < 0)
972 			val = 0;
973 		break;
974 
975 	case IPV6_RECVTCLASS:
976 		val = np->rxopt.bits.rxtclass;
977 		break;
978 
979 	case IPV6_FLOWINFO:
980 		val = np->rxopt.bits.rxflow;
981 		break;
982 
983 	case IPV6_UNICAST_HOPS:
984 	case IPV6_MULTICAST_HOPS:
985 	{
986 		struct dst_entry *dst;
987 
988 		if (optname == IPV6_UNICAST_HOPS)
989 			val = np->hop_limit;
990 		else
991 			val = np->mcast_hops;
992 
993 		dst = sk_dst_get(sk);
994 		if (dst) {
995 			if (val < 0)
996 				val = dst_metric(dst, RTAX_HOPLIMIT);
997 			if (val < 0)
998 				val = ipv6_get_hoplimit(dst->dev);
999 			dst_release(dst);
1000 		}
1001 		if (val < 0)
1002 			val = ipv6_devconf.hop_limit;
1003 		break;
1004 	}
1005 
1006 	case IPV6_MULTICAST_LOOP:
1007 		val = np->mc_loop;
1008 		break;
1009 
1010 	case IPV6_MULTICAST_IF:
1011 		val = np->mcast_oif;
1012 		break;
1013 
1014 	case IPV6_MTU_DISCOVER:
1015 		val = np->pmtudisc;
1016 		break;
1017 
1018 	case IPV6_RECVERR:
1019 		val = np->recverr;
1020 		break;
1021 
1022 	case IPV6_FLOWINFO_SEND:
1023 		val = np->sndflow;
1024 		break;
1025 
1026 	default:
1027 		return -EINVAL;
1028 	}
1029 	len = min_t(unsigned int, sizeof(int), len);
1030 	if(put_user(len, optlen))
1031 		return -EFAULT;
1032 	if(copy_to_user(optval,&val,len))
1033 		return -EFAULT;
1034 	return 0;
1035 }
1036 
1037 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1038 		    char __user *optval, int __user *optlen)
1039 {
1040 	int err;
1041 
1042 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1043 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1044 
1045 	if(level != SOL_IPV6)
1046 		return -ENOPROTOOPT;
1047 
1048 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1049 #ifdef CONFIG_NETFILTER
1050 	/* we need to exclude all possible EINVALs except default case */
1051 	if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1052 			optname != MCAST_MSFILTER) {
1053 		int len;
1054 
1055 		if (get_user(len, optlen))
1056 			return -EFAULT;
1057 
1058 		lock_sock(sk);
1059 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1060 				&len);
1061 		release_sock(sk);
1062 		if (err >= 0)
1063 			err = put_user(len, optlen);
1064 	}
1065 #endif
1066 	return err;
1067 }
1068 
1069 #ifdef CONFIG_COMPAT
1070 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1071 			   char __user *optval, int __user *optlen)
1072 {
1073 	int err;
1074 
1075 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1076 		if (udp_prot.compat_getsockopt != NULL)
1077 			return udp_prot.compat_getsockopt(sk, level, optname,
1078 							  optval, optlen);
1079 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1080 	}
1081 
1082 	if (level != SOL_IPV6)
1083 		return -ENOPROTOOPT;
1084 
1085 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1086 #ifdef CONFIG_NETFILTER
1087 	/* we need to exclude all possible EINVALs except default case */
1088 	if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1089 			optname != MCAST_MSFILTER) {
1090 		int len;
1091 
1092 		if (get_user(len, optlen))
1093 			return -EFAULT;
1094 
1095 		lock_sock(sk);
1096 		err = compat_nf_getsockopt(sk, PF_INET6,
1097 					   optname, optval, &len);
1098 		release_sock(sk);
1099 		if (err >= 0)
1100 			err = put_user(len, optlen);
1101 	}
1102 #endif
1103 	return err;
1104 }
1105 
1106 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1107 #endif
1108 
1109 void __init ipv6_packet_init(void)
1110 {
1111 	dev_add_pack(&ipv6_packet_type);
1112 }
1113 
1114 void ipv6_packet_cleanup(void)
1115 {
1116 	dev_remove_pack(&ipv6_packet_type);
1117 }
1118