xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision 643d1f7f)
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 = ipv6_hdr(skb);
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_reset_transport_header(skb);
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_V6_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 = ipv6_hdr(skb);
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_reset_transport_header(skb);
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 = ipv6_hdr(skb);
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_v4mapped(&np->daddr)) {
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_inuse_add(sk->sk_prot, -1);
272 				sock_prot_inuse_add(&tcp_prot, 1);
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_inuse_add(sk->sk_prot, -1);
286 				sock_prot_inuse_add(prot, 1);
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 		np->rxopt.bits.srcrt = valbool;
340 		retv = 0;
341 		break;
342 
343 	case IPV6_2292RTHDR:
344 		np->rxopt.bits.osrcrt = valbool;
345 		retv = 0;
346 		break;
347 
348 	case IPV6_RECVHOPOPTS:
349 		np->rxopt.bits.hopopts = valbool;
350 		retv = 0;
351 		break;
352 
353 	case IPV6_2292HOPOPTS:
354 		np->rxopt.bits.ohopopts = valbool;
355 		retv = 0;
356 		break;
357 
358 	case IPV6_RECVDSTOPTS:
359 		np->rxopt.bits.dstopts = valbool;
360 		retv = 0;
361 		break;
362 
363 	case IPV6_2292DSTOPTS:
364 		np->rxopt.bits.odstopts = valbool;
365 		retv = 0;
366 		break;
367 
368 	case IPV6_TCLASS:
369 		if (val < -1 || val > 0xff)
370 			goto e_inval;
371 		np->tclass = val;
372 		retv = 0;
373 		break;
374 
375 	case IPV6_RECVTCLASS:
376 		np->rxopt.bits.rxtclass = valbool;
377 		retv = 0;
378 		break;
379 
380 	case IPV6_FLOWINFO:
381 		np->rxopt.bits.rxflow = valbool;
382 		retv = 0;
383 		break;
384 
385 	case IPV6_HOPOPTS:
386 	case IPV6_RTHDRDSTOPTS:
387 	case IPV6_RTHDR:
388 	case IPV6_DSTOPTS:
389 	{
390 		struct ipv6_txoptions *opt;
391 		if (optlen == 0)
392 			optval = NULL;
393 
394 		/* hop-by-hop / destination options are privileged option */
395 		retv = -EPERM;
396 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
397 			break;
398 
399 		retv = -EINVAL;
400 		if (optlen & 0x7 || optlen > 8 * 255)
401 			break;
402 
403 		opt = ipv6_renew_options(sk, np->opt, optname,
404 					 (struct ipv6_opt_hdr __user *)optval,
405 					 optlen);
406 		if (IS_ERR(opt)) {
407 			retv = PTR_ERR(opt);
408 			break;
409 		}
410 
411 		/* routing header option needs extra check */
412 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
413 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
414 			switch (rthdr->type) {
415 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
416 			case IPV6_SRCRT_TYPE_2:
417 				break;
418 #endif
419 			default:
420 				goto sticky_done;
421 			}
422 
423 			if ((rthdr->hdrlen & 1) ||
424 			    (rthdr->hdrlen >> 1) != rthdr->segments_left)
425 				goto sticky_done;
426 		}
427 
428 		retv = 0;
429 		if (inet_sk(sk)->is_icsk) {
430 			if (opt) {
431 				struct inet_connection_sock *icsk = inet_csk(sk);
432 				if (!((1 << sk->sk_state) &
433 				      (TCPF_LISTEN | TCPF_CLOSE))
434 				    && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
435 					icsk->icsk_ext_hdr_len =
436 						opt->opt_flen + opt->opt_nflen;
437 					icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
438 				}
439 			}
440 			opt = xchg(&np->opt, opt);
441 			sk_dst_reset(sk);
442 		} else {
443 			write_lock(&sk->sk_dst_lock);
444 			opt = xchg(&np->opt, opt);
445 			write_unlock(&sk->sk_dst_lock);
446 			sk_dst_reset(sk);
447 		}
448 sticky_done:
449 		if (opt)
450 			sock_kfree_s(sk, opt, opt->tot_len);
451 		break;
452 	}
453 
454 	case IPV6_2292PKTOPTIONS:
455 	{
456 		struct ipv6_txoptions *opt = NULL;
457 		struct msghdr msg;
458 		struct flowi fl;
459 		int junk;
460 
461 		fl.fl6_flowlabel = 0;
462 		fl.oif = sk->sk_bound_dev_if;
463 
464 		if (optlen == 0)
465 			goto update;
466 
467 		/* 1K is probably excessive
468 		 * 1K is surely not enough, 2K per standard header is 16K.
469 		 */
470 		retv = -EINVAL;
471 		if (optlen > 64*1024)
472 			break;
473 
474 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
475 		retv = -ENOBUFS;
476 		if (opt == NULL)
477 			break;
478 
479 		memset(opt, 0, sizeof(*opt));
480 		opt->tot_len = sizeof(*opt) + optlen;
481 		retv = -EFAULT;
482 		if (copy_from_user(opt+1, optval, optlen))
483 			goto done;
484 
485 		msg.msg_controllen = optlen;
486 		msg.msg_control = (void*)(opt+1);
487 
488 		retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
489 		if (retv)
490 			goto done;
491 update:
492 		retv = 0;
493 		if (inet_sk(sk)->is_icsk) {
494 			if (opt) {
495 				struct inet_connection_sock *icsk = inet_csk(sk);
496 				if (!((1 << sk->sk_state) &
497 				      (TCPF_LISTEN | TCPF_CLOSE))
498 				    && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
499 					icsk->icsk_ext_hdr_len =
500 						opt->opt_flen + opt->opt_nflen;
501 					icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
502 				}
503 			}
504 			opt = xchg(&np->opt, opt);
505 			sk_dst_reset(sk);
506 		} else {
507 			write_lock(&sk->sk_dst_lock);
508 			opt = xchg(&np->opt, opt);
509 			write_unlock(&sk->sk_dst_lock);
510 			sk_dst_reset(sk);
511 		}
512 
513 done:
514 		if (opt)
515 			sock_kfree_s(sk, opt, opt->tot_len);
516 		break;
517 	}
518 	case IPV6_UNICAST_HOPS:
519 		if (val > 255 || val < -1)
520 			goto e_inval;
521 		np->hop_limit = val;
522 		retv = 0;
523 		break;
524 
525 	case IPV6_MULTICAST_HOPS:
526 		if (sk->sk_type == SOCK_STREAM)
527 			goto e_inval;
528 		if (val > 255 || val < -1)
529 			goto e_inval;
530 		np->mcast_hops = val;
531 		retv = 0;
532 		break;
533 
534 	case IPV6_MULTICAST_LOOP:
535 		np->mc_loop = valbool;
536 		retv = 0;
537 		break;
538 
539 	case IPV6_MULTICAST_IF:
540 		if (sk->sk_type == SOCK_STREAM)
541 			goto e_inval;
542 
543 		if (val) {
544 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
545 				goto e_inval;
546 
547 			if (__dev_get_by_index(&init_net, val) == NULL) {
548 				retv = -ENODEV;
549 				break;
550 			}
551 		}
552 		np->mcast_oif = val;
553 		retv = 0;
554 		break;
555 	case IPV6_ADD_MEMBERSHIP:
556 	case IPV6_DROP_MEMBERSHIP:
557 	{
558 		struct ipv6_mreq mreq;
559 
560 		retv = -EPROTO;
561 		if (inet_sk(sk)->is_icsk)
562 			break;
563 
564 		retv = -EFAULT;
565 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
566 			break;
567 
568 		if (optname == IPV6_ADD_MEMBERSHIP)
569 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
570 		else
571 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
572 		break;
573 	}
574 	case IPV6_JOIN_ANYCAST:
575 	case IPV6_LEAVE_ANYCAST:
576 	{
577 		struct ipv6_mreq mreq;
578 
579 		if (optlen != sizeof(struct ipv6_mreq))
580 			goto e_inval;
581 
582 		retv = -EFAULT;
583 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
584 			break;
585 
586 		if (optname == IPV6_JOIN_ANYCAST)
587 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
588 		else
589 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
590 		break;
591 	}
592 	case MCAST_JOIN_GROUP:
593 	case MCAST_LEAVE_GROUP:
594 	{
595 		struct group_req greq;
596 		struct sockaddr_in6 *psin6;
597 
598 		retv = -EFAULT;
599 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
600 			break;
601 		if (greq.gr_group.ss_family != AF_INET6) {
602 			retv = -EADDRNOTAVAIL;
603 			break;
604 		}
605 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
606 		if (optname == MCAST_JOIN_GROUP)
607 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
608 				&psin6->sin6_addr);
609 		else
610 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
611 				&psin6->sin6_addr);
612 		break;
613 	}
614 	case MCAST_JOIN_SOURCE_GROUP:
615 	case MCAST_LEAVE_SOURCE_GROUP:
616 	case MCAST_BLOCK_SOURCE:
617 	case MCAST_UNBLOCK_SOURCE:
618 	{
619 		struct group_source_req greqs;
620 		int omode, add;
621 
622 		if (optlen != sizeof(struct group_source_req))
623 			goto e_inval;
624 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
625 			retv = -EFAULT;
626 			break;
627 		}
628 		if (greqs.gsr_group.ss_family != AF_INET6 ||
629 		    greqs.gsr_source.ss_family != AF_INET6) {
630 			retv = -EADDRNOTAVAIL;
631 			break;
632 		}
633 		if (optname == MCAST_BLOCK_SOURCE) {
634 			omode = MCAST_EXCLUDE;
635 			add = 1;
636 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
637 			omode = MCAST_EXCLUDE;
638 			add = 0;
639 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
640 			struct sockaddr_in6 *psin6;
641 
642 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
643 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
644 				&psin6->sin6_addr);
645 			/* prior join w/ different source is ok */
646 			if (retv && retv != -EADDRINUSE)
647 				break;
648 			omode = MCAST_INCLUDE;
649 			add = 1;
650 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
651 			omode = MCAST_INCLUDE;
652 			add = 0;
653 		}
654 		retv = ip6_mc_source(add, omode, sk, &greqs);
655 		break;
656 	}
657 	case MCAST_MSFILTER:
658 	{
659 		extern int sysctl_mld_max_msf;
660 		struct group_filter *gsf;
661 
662 		if (optlen < GROUP_FILTER_SIZE(0))
663 			goto e_inval;
664 		if (optlen > sysctl_optmem_max) {
665 			retv = -ENOBUFS;
666 			break;
667 		}
668 		gsf = kmalloc(optlen,GFP_KERNEL);
669 		if (!gsf) {
670 			retv = -ENOBUFS;
671 			break;
672 		}
673 		retv = -EFAULT;
674 		if (copy_from_user(gsf, optval, optlen)) {
675 			kfree(gsf);
676 			break;
677 		}
678 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
679 		if (gsf->gf_numsrc >= 0x1ffffffU ||
680 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
681 			kfree(gsf);
682 			retv = -ENOBUFS;
683 			break;
684 		}
685 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
686 			kfree(gsf);
687 			retv = -EINVAL;
688 			break;
689 		}
690 		retv = ip6_mc_msfilter(sk, gsf);
691 		kfree(gsf);
692 
693 		break;
694 	}
695 	case IPV6_ROUTER_ALERT:
696 		retv = ip6_ra_control(sk, val, NULL);
697 		break;
698 	case IPV6_MTU_DISCOVER:
699 		if (val<0 || val>3)
700 			goto e_inval;
701 		np->pmtudisc = val;
702 		retv = 0;
703 		break;
704 	case IPV6_MTU:
705 		if (val && val < IPV6_MIN_MTU)
706 			goto e_inval;
707 		np->frag_size = val;
708 		retv = 0;
709 		break;
710 	case IPV6_RECVERR:
711 		np->recverr = valbool;
712 		if (!val)
713 			skb_queue_purge(&sk->sk_error_queue);
714 		retv = 0;
715 		break;
716 	case IPV6_FLOWINFO_SEND:
717 		np->sndflow = valbool;
718 		retv = 0;
719 		break;
720 	case IPV6_FLOWLABEL_MGR:
721 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
722 		break;
723 	case IPV6_IPSEC_POLICY:
724 	case IPV6_XFRM_POLICY:
725 		retv = -EPERM;
726 		if (!capable(CAP_NET_ADMIN))
727 			break;
728 		retv = xfrm_user_policy(sk, optname, optval, optlen);
729 		break;
730 
731 	}
732 	release_sock(sk);
733 
734 	return retv;
735 
736 e_inval:
737 	release_sock(sk);
738 	return -EINVAL;
739 }
740 
741 int ipv6_setsockopt(struct sock *sk, int level, int optname,
742 		    char __user *optval, int optlen)
743 {
744 	int err;
745 
746 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
747 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
748 
749 	if (level != SOL_IPV6)
750 		return -ENOPROTOOPT;
751 
752 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
753 #ifdef CONFIG_NETFILTER
754 	/* we need to exclude all possible ENOPROTOOPTs except default case */
755 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
756 			optname != IPV6_XFRM_POLICY) {
757 		lock_sock(sk);
758 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
759 				optlen);
760 		release_sock(sk);
761 	}
762 #endif
763 	return err;
764 }
765 
766 EXPORT_SYMBOL(ipv6_setsockopt);
767 
768 #ifdef CONFIG_COMPAT
769 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
770 			   char __user *optval, int optlen)
771 {
772 	int err;
773 
774 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
775 		if (udp_prot.compat_setsockopt != NULL)
776 			return udp_prot.compat_setsockopt(sk, level, optname,
777 							  optval, optlen);
778 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
779 	}
780 
781 	if (level != SOL_IPV6)
782 		return -ENOPROTOOPT;
783 
784 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
785 #ifdef CONFIG_NETFILTER
786 	/* we need to exclude all possible ENOPROTOOPTs except default case */
787 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
788 	    optname != IPV6_XFRM_POLICY) {
789 		lock_sock(sk);
790 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
791 					   optval, optlen);
792 		release_sock(sk);
793 	}
794 #endif
795 	return err;
796 }
797 
798 EXPORT_SYMBOL(compat_ipv6_setsockopt);
799 #endif
800 
801 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
802 				  int optname, char __user *optval, int len)
803 {
804 	struct ipv6_opt_hdr *hdr;
805 
806 	if (!opt)
807 		return 0;
808 
809 	switch(optname) {
810 	case IPV6_HOPOPTS:
811 		hdr = opt->hopopt;
812 		break;
813 	case IPV6_RTHDRDSTOPTS:
814 		hdr = opt->dst0opt;
815 		break;
816 	case IPV6_RTHDR:
817 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
818 		break;
819 	case IPV6_DSTOPTS:
820 		hdr = opt->dst1opt;
821 		break;
822 	default:
823 		return -EINVAL;	/* should not happen */
824 	}
825 
826 	if (!hdr)
827 		return 0;
828 
829 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
830 	if (copy_to_user(optval, hdr, len))
831 		return -EFAULT;
832 	return ipv6_optlen(hdr);
833 }
834 
835 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
836 		    char __user *optval, int __user *optlen)
837 {
838 	struct ipv6_pinfo *np = inet6_sk(sk);
839 	int len;
840 	int val;
841 
842 	if (get_user(len, optlen))
843 		return -EFAULT;
844 	switch (optname) {
845 	case IPV6_ADDRFORM:
846 		if (sk->sk_protocol != IPPROTO_UDP &&
847 		    sk->sk_protocol != IPPROTO_UDPLITE &&
848 		    sk->sk_protocol != IPPROTO_TCP)
849 			return -EINVAL;
850 		if (sk->sk_state != TCP_ESTABLISHED)
851 			return -ENOTCONN;
852 		val = sk->sk_family;
853 		break;
854 	case MCAST_MSFILTER:
855 	{
856 		struct group_filter gsf;
857 		int err;
858 
859 		if (len < GROUP_FILTER_SIZE(0))
860 			return -EINVAL;
861 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
862 			return -EFAULT;
863 		lock_sock(sk);
864 		err = ip6_mc_msfget(sk, &gsf,
865 			(struct group_filter __user *)optval, optlen);
866 		release_sock(sk);
867 		return err;
868 	}
869 
870 	case IPV6_2292PKTOPTIONS:
871 	{
872 		struct msghdr msg;
873 		struct sk_buff *skb;
874 
875 		if (sk->sk_type != SOCK_STREAM)
876 			return -ENOPROTOOPT;
877 
878 		msg.msg_control = optval;
879 		msg.msg_controllen = len;
880 		msg.msg_flags = 0;
881 
882 		lock_sock(sk);
883 		skb = np->pktoptions;
884 		if (skb)
885 			atomic_inc(&skb->users);
886 		release_sock(sk);
887 
888 		if (skb) {
889 			int err = datagram_recv_ctl(sk, &msg, skb);
890 			kfree_skb(skb);
891 			if (err)
892 				return err;
893 		} else {
894 			if (np->rxopt.bits.rxinfo) {
895 				struct in6_pktinfo src_info;
896 				src_info.ipi6_ifindex = np->mcast_oif;
897 				ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
898 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
899 			}
900 			if (np->rxopt.bits.rxhlim) {
901 				int hlim = np->mcast_hops;
902 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
903 			}
904 			if (np->rxopt.bits.rxoinfo) {
905 				struct in6_pktinfo src_info;
906 				src_info.ipi6_ifindex = np->mcast_oif;
907 				ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
908 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
909 			}
910 			if (np->rxopt.bits.rxohlim) {
911 				int hlim = np->mcast_hops;
912 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
913 			}
914 		}
915 		len -= msg.msg_controllen;
916 		return put_user(len, optlen);
917 	}
918 	case IPV6_MTU:
919 	{
920 		struct dst_entry *dst;
921 		val = 0;
922 		lock_sock(sk);
923 		dst = sk_dst_get(sk);
924 		if (dst) {
925 			val = dst_mtu(dst);
926 			dst_release(dst);
927 		}
928 		release_sock(sk);
929 		if (!val)
930 			return -ENOTCONN;
931 		break;
932 	}
933 
934 	case IPV6_V6ONLY:
935 		val = np->ipv6only;
936 		break;
937 
938 	case IPV6_RECVPKTINFO:
939 		val = np->rxopt.bits.rxinfo;
940 		break;
941 
942 	case IPV6_2292PKTINFO:
943 		val = np->rxopt.bits.rxoinfo;
944 		break;
945 
946 	case IPV6_RECVHOPLIMIT:
947 		val = np->rxopt.bits.rxhlim;
948 		break;
949 
950 	case IPV6_2292HOPLIMIT:
951 		val = np->rxopt.bits.rxohlim;
952 		break;
953 
954 	case IPV6_RECVRTHDR:
955 		val = np->rxopt.bits.srcrt;
956 		break;
957 
958 	case IPV6_2292RTHDR:
959 		val = np->rxopt.bits.osrcrt;
960 		break;
961 
962 	case IPV6_HOPOPTS:
963 	case IPV6_RTHDRDSTOPTS:
964 	case IPV6_RTHDR:
965 	case IPV6_DSTOPTS:
966 	{
967 
968 		lock_sock(sk);
969 		len = ipv6_getsockopt_sticky(sk, np->opt,
970 					     optname, optval, len);
971 		release_sock(sk);
972 		return put_user(len, optlen);
973 	}
974 
975 	case IPV6_RECVHOPOPTS:
976 		val = np->rxopt.bits.hopopts;
977 		break;
978 
979 	case IPV6_2292HOPOPTS:
980 		val = np->rxopt.bits.ohopopts;
981 		break;
982 
983 	case IPV6_RECVDSTOPTS:
984 		val = np->rxopt.bits.dstopts;
985 		break;
986 
987 	case IPV6_2292DSTOPTS:
988 		val = np->rxopt.bits.odstopts;
989 		break;
990 
991 	case IPV6_TCLASS:
992 		val = np->tclass;
993 		if (val < 0)
994 			val = 0;
995 		break;
996 
997 	case IPV6_RECVTCLASS:
998 		val = np->rxopt.bits.rxtclass;
999 		break;
1000 
1001 	case IPV6_FLOWINFO:
1002 		val = np->rxopt.bits.rxflow;
1003 		break;
1004 
1005 	case IPV6_UNICAST_HOPS:
1006 	case IPV6_MULTICAST_HOPS:
1007 	{
1008 		struct dst_entry *dst;
1009 
1010 		if (optname == IPV6_UNICAST_HOPS)
1011 			val = np->hop_limit;
1012 		else
1013 			val = np->mcast_hops;
1014 
1015 		dst = sk_dst_get(sk);
1016 		if (dst) {
1017 			if (val < 0)
1018 				val = dst_metric(dst, RTAX_HOPLIMIT);
1019 			if (val < 0)
1020 				val = ipv6_get_hoplimit(dst->dev);
1021 			dst_release(dst);
1022 		}
1023 		if (val < 0)
1024 			val = ipv6_devconf.hop_limit;
1025 		break;
1026 	}
1027 
1028 	case IPV6_MULTICAST_LOOP:
1029 		val = np->mc_loop;
1030 		break;
1031 
1032 	case IPV6_MULTICAST_IF:
1033 		val = np->mcast_oif;
1034 		break;
1035 
1036 	case IPV6_MTU_DISCOVER:
1037 		val = np->pmtudisc;
1038 		break;
1039 
1040 	case IPV6_RECVERR:
1041 		val = np->recverr;
1042 		break;
1043 
1044 	case IPV6_FLOWINFO_SEND:
1045 		val = np->sndflow;
1046 		break;
1047 
1048 	default:
1049 		return -ENOPROTOOPT;
1050 	}
1051 	len = min_t(unsigned int, sizeof(int), len);
1052 	if(put_user(len, optlen))
1053 		return -EFAULT;
1054 	if(copy_to_user(optval,&val,len))
1055 		return -EFAULT;
1056 	return 0;
1057 }
1058 
1059 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1060 		    char __user *optval, int __user *optlen)
1061 {
1062 	int err;
1063 
1064 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1065 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1066 
1067 	if(level != SOL_IPV6)
1068 		return -ENOPROTOOPT;
1069 
1070 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1071 #ifdef CONFIG_NETFILTER
1072 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1073 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1074 		int len;
1075 
1076 		if (get_user(len, optlen))
1077 			return -EFAULT;
1078 
1079 		lock_sock(sk);
1080 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1081 				&len);
1082 		release_sock(sk);
1083 		if (err >= 0)
1084 			err = put_user(len, optlen);
1085 	}
1086 #endif
1087 	return err;
1088 }
1089 
1090 EXPORT_SYMBOL(ipv6_getsockopt);
1091 
1092 #ifdef CONFIG_COMPAT
1093 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1094 			   char __user *optval, int __user *optlen)
1095 {
1096 	int err;
1097 
1098 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1099 		if (udp_prot.compat_getsockopt != NULL)
1100 			return udp_prot.compat_getsockopt(sk, level, optname,
1101 							  optval, optlen);
1102 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1103 	}
1104 
1105 	if (level != SOL_IPV6)
1106 		return -ENOPROTOOPT;
1107 
1108 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1109 #ifdef CONFIG_NETFILTER
1110 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1111 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1112 		int len;
1113 
1114 		if (get_user(len, optlen))
1115 			return -EFAULT;
1116 
1117 		lock_sock(sk);
1118 		err = compat_nf_getsockopt(sk, PF_INET6,
1119 					   optname, optval, &len);
1120 		release_sock(sk);
1121 		if (err >= 0)
1122 			err = put_user(len, optlen);
1123 	}
1124 #endif
1125 	return err;
1126 }
1127 
1128 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1129 #endif
1130 
1131 int __init ipv6_packet_init(void)
1132 {
1133 	dev_add_pack(&ipv6_packet_type);
1134 	return 0;
1135 }
1136 
1137 void ipv6_packet_cleanup(void)
1138 {
1139 	dev_remove_pack(&ipv6_packet_type);
1140 }
1141