xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision 63dc02bd)
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  *	This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  *
15  *	FIXME: Make the setsockopt code POSIX compliant: That is
16  *
17  *	o	Truncate getsockopt returns
18  *	o	Return an optlen of the truncated length if need be
19  *
20  *	Changes:
21  *	David L Stevens <dlstevens@us.ibm.com>:
22  *		- added multicast source filtering API for MLDv2
23  */
24 
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39 #include <linux/slab.h>
40 
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/transp_v6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/inet_common.h>
50 #include <net/tcp.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/xfrm.h>
54 #include <net/compat.h>
55 
56 #include <asm/uaccess.h>
57 
58 struct ip6_ra_chain *ip6_ra_chain;
59 DEFINE_RWLOCK(ip6_ra_lock);
60 
61 int ip6_ra_control(struct sock *sk, int sel)
62 {
63 	struct ip6_ra_chain *ra, *new_ra, **rap;
64 
65 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
66 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
67 		return -ENOPROTOOPT;
68 
69 	new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
70 
71 	write_lock_bh(&ip6_ra_lock);
72 	for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
73 		if (ra->sk == sk) {
74 			if (sel>=0) {
75 				write_unlock_bh(&ip6_ra_lock);
76 				kfree(new_ra);
77 				return -EADDRINUSE;
78 			}
79 
80 			*rap = ra->next;
81 			write_unlock_bh(&ip6_ra_lock);
82 
83 			sock_put(sk);
84 			kfree(ra);
85 			return 0;
86 		}
87 	}
88 	if (new_ra == NULL) {
89 		write_unlock_bh(&ip6_ra_lock);
90 		return -ENOBUFS;
91 	}
92 	new_ra->sk = sk;
93 	new_ra->sel = sel;
94 	new_ra->next = ra;
95 	*rap = new_ra;
96 	sock_hold(sk);
97 	write_unlock_bh(&ip6_ra_lock);
98 	return 0;
99 }
100 
101 static
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 					   struct ipv6_txoptions *opt)
104 {
105 	if (inet_sk(sk)->is_icsk) {
106 		if (opt &&
107 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108 		    inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109 			struct inet_connection_sock *icsk = inet_csk(sk);
110 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
112 		}
113 		opt = xchg(&inet6_sk(sk)->opt, opt);
114 	} else {
115 		spin_lock(&sk->sk_dst_lock);
116 		opt = xchg(&inet6_sk(sk)->opt, opt);
117 		spin_unlock(&sk->sk_dst_lock);
118 	}
119 	sk_dst_reset(sk);
120 
121 	return opt;
122 }
123 
124 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
125 		    char __user *optval, unsigned int optlen)
126 {
127 	struct ipv6_pinfo *np = inet6_sk(sk);
128 	struct net *net = sock_net(sk);
129 	int val, valbool;
130 	int retv = -ENOPROTOOPT;
131 
132 	if (optval == NULL)
133 		val=0;
134 	else {
135 		if (optlen >= sizeof(int)) {
136 			if (get_user(val, (int __user *) optval))
137 				return -EFAULT;
138 		} else
139 			val = 0;
140 	}
141 
142 	valbool = (val!=0);
143 
144 	if (ip6_mroute_opt(optname))
145 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
146 
147 	lock_sock(sk);
148 
149 	switch (optname) {
150 
151 	case IPV6_ADDRFORM:
152 		if (optlen < sizeof(int))
153 			goto e_inval;
154 		if (val == PF_INET) {
155 			struct ipv6_txoptions *opt;
156 			struct sk_buff *pktopt;
157 
158 			if (sk->sk_type == SOCK_RAW)
159 				break;
160 
161 			if (sk->sk_protocol == IPPROTO_UDP ||
162 			    sk->sk_protocol == IPPROTO_UDPLITE) {
163 				struct udp_sock *up = udp_sk(sk);
164 				if (up->pending == AF_INET6) {
165 					retv = -EBUSY;
166 					break;
167 				}
168 			} else if (sk->sk_protocol != IPPROTO_TCP)
169 				break;
170 
171 			if (sk->sk_state != TCP_ESTABLISHED) {
172 				retv = -ENOTCONN;
173 				break;
174 			}
175 
176 			if (ipv6_only_sock(sk) ||
177 			    !ipv6_addr_v4mapped(&np->daddr)) {
178 				retv = -EADDRNOTAVAIL;
179 				break;
180 			}
181 
182 			fl6_free_socklist(sk);
183 			ipv6_sock_mc_close(sk);
184 
185 			/*
186 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
187 			 * remove it from the refcnt debug socks count in the
188 			 * original family...
189 			 */
190 			sk_refcnt_debug_dec(sk);
191 
192 			if (sk->sk_protocol == IPPROTO_TCP) {
193 				struct inet_connection_sock *icsk = inet_csk(sk);
194 				local_bh_disable();
195 				sock_prot_inuse_add(net, sk->sk_prot, -1);
196 				sock_prot_inuse_add(net, &tcp_prot, 1);
197 				local_bh_enable();
198 				sk->sk_prot = &tcp_prot;
199 				icsk->icsk_af_ops = &ipv4_specific;
200 				sk->sk_socket->ops = &inet_stream_ops;
201 				sk->sk_family = PF_INET;
202 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
203 			} else {
204 				struct proto *prot = &udp_prot;
205 
206 				if (sk->sk_protocol == IPPROTO_UDPLITE)
207 					prot = &udplite_prot;
208 				local_bh_disable();
209 				sock_prot_inuse_add(net, sk->sk_prot, -1);
210 				sock_prot_inuse_add(net, prot, 1);
211 				local_bh_enable();
212 				sk->sk_prot = prot;
213 				sk->sk_socket->ops = &inet_dgram_ops;
214 				sk->sk_family = PF_INET;
215 			}
216 			opt = xchg(&np->opt, NULL);
217 			if (opt)
218 				sock_kfree_s(sk, opt, opt->tot_len);
219 			pktopt = xchg(&np->pktoptions, NULL);
220 			kfree_skb(pktopt);
221 
222 			sk->sk_destruct = inet_sock_destruct;
223 			/*
224 			 * ... and add it to the refcnt debug socks count
225 			 * in the new family. -acme
226 			 */
227 			sk_refcnt_debug_inc(sk);
228 			module_put(THIS_MODULE);
229 			retv = 0;
230 			break;
231 		}
232 		goto e_inval;
233 
234 	case IPV6_V6ONLY:
235 		if (optlen < sizeof(int) ||
236 		    inet_sk(sk)->inet_num)
237 			goto e_inval;
238 		np->ipv6only = valbool;
239 		retv = 0;
240 		break;
241 
242 	case IPV6_RECVPKTINFO:
243 		if (optlen < sizeof(int))
244 			goto e_inval;
245 		np->rxopt.bits.rxinfo = valbool;
246 		retv = 0;
247 		break;
248 
249 	case IPV6_2292PKTINFO:
250 		if (optlen < sizeof(int))
251 			goto e_inval;
252 		np->rxopt.bits.rxoinfo = valbool;
253 		retv = 0;
254 		break;
255 
256 	case IPV6_RECVHOPLIMIT:
257 		if (optlen < sizeof(int))
258 			goto e_inval;
259 		np->rxopt.bits.rxhlim = valbool;
260 		retv = 0;
261 		break;
262 
263 	case IPV6_2292HOPLIMIT:
264 		if (optlen < sizeof(int))
265 			goto e_inval;
266 		np->rxopt.bits.rxohlim = valbool;
267 		retv = 0;
268 		break;
269 
270 	case IPV6_RECVRTHDR:
271 		if (optlen < sizeof(int))
272 			goto e_inval;
273 		np->rxopt.bits.srcrt = valbool;
274 		retv = 0;
275 		break;
276 
277 	case IPV6_2292RTHDR:
278 		if (optlen < sizeof(int))
279 			goto e_inval;
280 		np->rxopt.bits.osrcrt = valbool;
281 		retv = 0;
282 		break;
283 
284 	case IPV6_RECVHOPOPTS:
285 		if (optlen < sizeof(int))
286 			goto e_inval;
287 		np->rxopt.bits.hopopts = valbool;
288 		retv = 0;
289 		break;
290 
291 	case IPV6_2292HOPOPTS:
292 		if (optlen < sizeof(int))
293 			goto e_inval;
294 		np->rxopt.bits.ohopopts = valbool;
295 		retv = 0;
296 		break;
297 
298 	case IPV6_RECVDSTOPTS:
299 		if (optlen < sizeof(int))
300 			goto e_inval;
301 		np->rxopt.bits.dstopts = valbool;
302 		retv = 0;
303 		break;
304 
305 	case IPV6_2292DSTOPTS:
306 		if (optlen < sizeof(int))
307 			goto e_inval;
308 		np->rxopt.bits.odstopts = valbool;
309 		retv = 0;
310 		break;
311 
312 	case IPV6_TCLASS:
313 		if (optlen < sizeof(int))
314 			goto e_inval;
315 		if (val < -1 || val > 0xff)
316 			goto e_inval;
317 		/* RFC 3542, 6.5: default traffic class of 0x0 */
318 		if (val == -1)
319 			val = 0;
320 		np->tclass = val;
321 		retv = 0;
322 		break;
323 
324 	case IPV6_RECVTCLASS:
325 		if (optlen < sizeof(int))
326 			goto e_inval;
327 		np->rxopt.bits.rxtclass = valbool;
328 		retv = 0;
329 		break;
330 
331 	case IPV6_FLOWINFO:
332 		if (optlen < sizeof(int))
333 			goto e_inval;
334 		np->rxopt.bits.rxflow = valbool;
335 		retv = 0;
336 		break;
337 
338 	case IPV6_RECVPATHMTU:
339 		if (optlen < sizeof(int))
340 			goto e_inval;
341 		np->rxopt.bits.rxpmtu = valbool;
342 		retv = 0;
343 		break;
344 
345 	case IPV6_TRANSPARENT:
346 		if (valbool && !capable(CAP_NET_ADMIN) && !capable(CAP_NET_RAW)) {
347 			retv = -EPERM;
348 			break;
349 		}
350 		if (optlen < sizeof(int))
351 			goto e_inval;
352 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
353 		inet_sk(sk)->transparent = valbool;
354 		retv = 0;
355 		break;
356 
357 	case IPV6_RECVORIGDSTADDR:
358 		if (optlen < sizeof(int))
359 			goto e_inval;
360 		np->rxopt.bits.rxorigdstaddr = valbool;
361 		retv = 0;
362 		break;
363 
364 	case IPV6_HOPOPTS:
365 	case IPV6_RTHDRDSTOPTS:
366 	case IPV6_RTHDR:
367 	case IPV6_DSTOPTS:
368 	{
369 		struct ipv6_txoptions *opt;
370 
371 		/* remove any sticky options header with a zero option
372 		 * length, per RFC3542.
373 		 */
374 		if (optlen == 0)
375 			optval = NULL;
376 		else if (optval == NULL)
377 			goto e_inval;
378 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
379 			 optlen & 0x7 || optlen > 8 * 255)
380 			goto e_inval;
381 
382 		/* hop-by-hop / destination options are privileged option */
383 		retv = -EPERM;
384 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
385 			break;
386 
387 		opt = ipv6_renew_options(sk, np->opt, optname,
388 					 (struct ipv6_opt_hdr __user *)optval,
389 					 optlen);
390 		if (IS_ERR(opt)) {
391 			retv = PTR_ERR(opt);
392 			break;
393 		}
394 
395 		/* routing header option needs extra check */
396 		retv = -EINVAL;
397 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
398 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
399 			switch (rthdr->type) {
400 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
401 			case IPV6_SRCRT_TYPE_2:
402 				if (rthdr->hdrlen != 2 ||
403 				    rthdr->segments_left != 1)
404 					goto sticky_done;
405 
406 				break;
407 #endif
408 			default:
409 				goto sticky_done;
410 			}
411 		}
412 
413 		retv = 0;
414 		opt = ipv6_update_options(sk, opt);
415 sticky_done:
416 		if (opt)
417 			sock_kfree_s(sk, opt, opt->tot_len);
418 		break;
419 	}
420 
421 	case IPV6_PKTINFO:
422 	{
423 		struct in6_pktinfo pkt;
424 
425 		if (optlen == 0)
426 			goto e_inval;
427 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
428 			goto e_inval;
429 
430 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
431 				retv = -EFAULT;
432 				break;
433 		}
434 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
435 			goto e_inval;
436 
437 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
438 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
439 		retv = 0;
440 		break;
441 	}
442 
443 	case IPV6_2292PKTOPTIONS:
444 	{
445 		struct ipv6_txoptions *opt = NULL;
446 		struct msghdr msg;
447 		struct flowi6 fl6;
448 		int junk;
449 
450 		memset(&fl6, 0, sizeof(fl6));
451 		fl6.flowi6_oif = sk->sk_bound_dev_if;
452 		fl6.flowi6_mark = sk->sk_mark;
453 
454 		if (optlen == 0)
455 			goto update;
456 
457 		/* 1K is probably excessive
458 		 * 1K is surely not enough, 2K per standard header is 16K.
459 		 */
460 		retv = -EINVAL;
461 		if (optlen > 64*1024)
462 			break;
463 
464 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
465 		retv = -ENOBUFS;
466 		if (opt == NULL)
467 			break;
468 
469 		memset(opt, 0, sizeof(*opt));
470 		opt->tot_len = sizeof(*opt) + optlen;
471 		retv = -EFAULT;
472 		if (copy_from_user(opt+1, optval, optlen))
473 			goto done;
474 
475 		msg.msg_controllen = optlen;
476 		msg.msg_control = (void*)(opt+1);
477 
478 		retv = datagram_send_ctl(net, sk, &msg, &fl6, opt, &junk, &junk,
479 					 &junk);
480 		if (retv)
481 			goto done;
482 update:
483 		retv = 0;
484 		opt = ipv6_update_options(sk, opt);
485 done:
486 		if (opt)
487 			sock_kfree_s(sk, opt, opt->tot_len);
488 		break;
489 	}
490 	case IPV6_UNICAST_HOPS:
491 		if (optlen < sizeof(int))
492 			goto e_inval;
493 		if (val > 255 || val < -1)
494 			goto e_inval;
495 		np->hop_limit = val;
496 		retv = 0;
497 		break;
498 
499 	case IPV6_MULTICAST_HOPS:
500 		if (sk->sk_type == SOCK_STREAM)
501 			break;
502 		if (optlen < sizeof(int))
503 			goto e_inval;
504 		if (val > 255 || val < -1)
505 			goto e_inval;
506 		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
507 		retv = 0;
508 		break;
509 
510 	case IPV6_MULTICAST_LOOP:
511 		if (optlen < sizeof(int))
512 			goto e_inval;
513 		if (val != valbool)
514 			goto e_inval;
515 		np->mc_loop = valbool;
516 		retv = 0;
517 		break;
518 
519 	case IPV6_UNICAST_IF:
520 	{
521 		struct net_device *dev = NULL;
522 		int ifindex;
523 
524 		if (optlen != sizeof(int))
525 			goto e_inval;
526 
527 		ifindex = (__force int)ntohl((__force __be32)val);
528 		if (ifindex == 0) {
529 			np->ucast_oif = 0;
530 			retv = 0;
531 			break;
532 		}
533 
534 		dev = dev_get_by_index(net, ifindex);
535 		retv = -EADDRNOTAVAIL;
536 		if (!dev)
537 			break;
538 		dev_put(dev);
539 
540 		retv = -EINVAL;
541 		if (sk->sk_bound_dev_if)
542 			break;
543 
544 		np->ucast_oif = ifindex;
545 		retv = 0;
546 		break;
547 	}
548 
549 	case IPV6_MULTICAST_IF:
550 		if (sk->sk_type == SOCK_STREAM)
551 			break;
552 		if (optlen < sizeof(int))
553 			goto e_inval;
554 
555 		if (val) {
556 			struct net_device *dev;
557 
558 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
559 				goto e_inval;
560 
561 			dev = dev_get_by_index(net, val);
562 			if (!dev) {
563 				retv = -ENODEV;
564 				break;
565 			}
566 			dev_put(dev);
567 		}
568 		np->mcast_oif = val;
569 		retv = 0;
570 		break;
571 	case IPV6_ADD_MEMBERSHIP:
572 	case IPV6_DROP_MEMBERSHIP:
573 	{
574 		struct ipv6_mreq mreq;
575 
576 		if (optlen < sizeof(struct ipv6_mreq))
577 			goto e_inval;
578 
579 		retv = -EPROTO;
580 		if (inet_sk(sk)->is_icsk)
581 			break;
582 
583 		retv = -EFAULT;
584 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
585 			break;
586 
587 		if (optname == IPV6_ADD_MEMBERSHIP)
588 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
589 		else
590 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
591 		break;
592 	}
593 	case IPV6_JOIN_ANYCAST:
594 	case IPV6_LEAVE_ANYCAST:
595 	{
596 		struct ipv6_mreq mreq;
597 
598 		if (optlen < sizeof(struct ipv6_mreq))
599 			goto e_inval;
600 
601 		retv = -EFAULT;
602 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
603 			break;
604 
605 		if (optname == IPV6_JOIN_ANYCAST)
606 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
607 		else
608 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
609 		break;
610 	}
611 	case MCAST_JOIN_GROUP:
612 	case MCAST_LEAVE_GROUP:
613 	{
614 		struct group_req greq;
615 		struct sockaddr_in6 *psin6;
616 
617 		if (optlen < sizeof(struct group_req))
618 			goto e_inval;
619 
620 		retv = -EFAULT;
621 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
622 			break;
623 		if (greq.gr_group.ss_family != AF_INET6) {
624 			retv = -EADDRNOTAVAIL;
625 			break;
626 		}
627 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
628 		if (optname == MCAST_JOIN_GROUP)
629 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
630 				&psin6->sin6_addr);
631 		else
632 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
633 				&psin6->sin6_addr);
634 		break;
635 	}
636 	case MCAST_JOIN_SOURCE_GROUP:
637 	case MCAST_LEAVE_SOURCE_GROUP:
638 	case MCAST_BLOCK_SOURCE:
639 	case MCAST_UNBLOCK_SOURCE:
640 	{
641 		struct group_source_req greqs;
642 		int omode, add;
643 
644 		if (optlen < sizeof(struct group_source_req))
645 			goto e_inval;
646 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
647 			retv = -EFAULT;
648 			break;
649 		}
650 		if (greqs.gsr_group.ss_family != AF_INET6 ||
651 		    greqs.gsr_source.ss_family != AF_INET6) {
652 			retv = -EADDRNOTAVAIL;
653 			break;
654 		}
655 		if (optname == MCAST_BLOCK_SOURCE) {
656 			omode = MCAST_EXCLUDE;
657 			add = 1;
658 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
659 			omode = MCAST_EXCLUDE;
660 			add = 0;
661 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
662 			struct sockaddr_in6 *psin6;
663 
664 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
665 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
666 				&psin6->sin6_addr);
667 			/* prior join w/ different source is ok */
668 			if (retv && retv != -EADDRINUSE)
669 				break;
670 			omode = MCAST_INCLUDE;
671 			add = 1;
672 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
673 			omode = MCAST_INCLUDE;
674 			add = 0;
675 		}
676 		retv = ip6_mc_source(add, omode, sk, &greqs);
677 		break;
678 	}
679 	case MCAST_MSFILTER:
680 	{
681 		extern int sysctl_mld_max_msf;
682 		struct group_filter *gsf;
683 
684 		if (optlen < GROUP_FILTER_SIZE(0))
685 			goto e_inval;
686 		if (optlen > sysctl_optmem_max) {
687 			retv = -ENOBUFS;
688 			break;
689 		}
690 		gsf = kmalloc(optlen,GFP_KERNEL);
691 		if (!gsf) {
692 			retv = -ENOBUFS;
693 			break;
694 		}
695 		retv = -EFAULT;
696 		if (copy_from_user(gsf, optval, optlen)) {
697 			kfree(gsf);
698 			break;
699 		}
700 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
701 		if (gsf->gf_numsrc >= 0x1ffffffU ||
702 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
703 			kfree(gsf);
704 			retv = -ENOBUFS;
705 			break;
706 		}
707 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
708 			kfree(gsf);
709 			retv = -EINVAL;
710 			break;
711 		}
712 		retv = ip6_mc_msfilter(sk, gsf);
713 		kfree(gsf);
714 
715 		break;
716 	}
717 	case IPV6_ROUTER_ALERT:
718 		if (optlen < sizeof(int))
719 			goto e_inval;
720 		retv = ip6_ra_control(sk, val);
721 		break;
722 	case IPV6_MTU_DISCOVER:
723 		if (optlen < sizeof(int))
724 			goto e_inval;
725 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
726 			goto e_inval;
727 		np->pmtudisc = val;
728 		retv = 0;
729 		break;
730 	case IPV6_MTU:
731 		if (optlen < sizeof(int))
732 			goto e_inval;
733 		if (val && val < IPV6_MIN_MTU)
734 			goto e_inval;
735 		np->frag_size = val;
736 		retv = 0;
737 		break;
738 	case IPV6_RECVERR:
739 		if (optlen < sizeof(int))
740 			goto e_inval;
741 		np->recverr = valbool;
742 		if (!val)
743 			skb_queue_purge(&sk->sk_error_queue);
744 		retv = 0;
745 		break;
746 	case IPV6_FLOWINFO_SEND:
747 		if (optlen < sizeof(int))
748 			goto e_inval;
749 		np->sndflow = valbool;
750 		retv = 0;
751 		break;
752 	case IPV6_FLOWLABEL_MGR:
753 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
754 		break;
755 	case IPV6_IPSEC_POLICY:
756 	case IPV6_XFRM_POLICY:
757 		retv = -EPERM;
758 		if (!capable(CAP_NET_ADMIN))
759 			break;
760 		retv = xfrm_user_policy(sk, optname, optval, optlen);
761 		break;
762 
763 	case IPV6_ADDR_PREFERENCES:
764 	    {
765 		unsigned int pref = 0;
766 		unsigned int prefmask = ~0;
767 
768 		if (optlen < sizeof(int))
769 			goto e_inval;
770 
771 		retv = -EINVAL;
772 
773 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
774 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
775 			       IPV6_PREFER_SRC_TMP|
776 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
777 		case IPV6_PREFER_SRC_PUBLIC:
778 			pref |= IPV6_PREFER_SRC_PUBLIC;
779 			break;
780 		case IPV6_PREFER_SRC_TMP:
781 			pref |= IPV6_PREFER_SRC_TMP;
782 			break;
783 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
784 			break;
785 		case 0:
786 			goto pref_skip_pubtmp;
787 		default:
788 			goto e_inval;
789 		}
790 
791 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
792 			      IPV6_PREFER_SRC_TMP);
793 pref_skip_pubtmp:
794 
795 		/* check HOME/COA conflicts */
796 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
797 		case IPV6_PREFER_SRC_HOME:
798 			break;
799 		case IPV6_PREFER_SRC_COA:
800 			pref |= IPV6_PREFER_SRC_COA;
801 		case 0:
802 			goto pref_skip_coa;
803 		default:
804 			goto e_inval;
805 		}
806 
807 		prefmask &= ~IPV6_PREFER_SRC_COA;
808 pref_skip_coa:
809 
810 		/* check CGA/NONCGA conflicts */
811 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
812 		case IPV6_PREFER_SRC_CGA:
813 		case IPV6_PREFER_SRC_NONCGA:
814 		case 0:
815 			break;
816 		default:
817 			goto e_inval;
818 		}
819 
820 		np->srcprefs = (np->srcprefs & prefmask) | pref;
821 		retv = 0;
822 
823 		break;
824 	    }
825 	case IPV6_MINHOPCOUNT:
826 		if (optlen < sizeof(int))
827 			goto e_inval;
828 		if (val < 0 || val > 255)
829 			goto e_inval;
830 		np->min_hopcount = val;
831 		break;
832 	case IPV6_DONTFRAG:
833 		np->dontfrag = valbool;
834 		retv = 0;
835 		break;
836 	}
837 
838 	release_sock(sk);
839 
840 	return retv;
841 
842 e_inval:
843 	release_sock(sk);
844 	return -EINVAL;
845 }
846 
847 int ipv6_setsockopt(struct sock *sk, int level, int optname,
848 		    char __user *optval, unsigned int optlen)
849 {
850 	int err;
851 
852 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
853 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
854 
855 	if (level != SOL_IPV6)
856 		return -ENOPROTOOPT;
857 
858 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
859 #ifdef CONFIG_NETFILTER
860 	/* we need to exclude all possible ENOPROTOOPTs except default case */
861 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
862 			optname != IPV6_XFRM_POLICY) {
863 		lock_sock(sk);
864 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
865 				optlen);
866 		release_sock(sk);
867 	}
868 #endif
869 	return err;
870 }
871 
872 EXPORT_SYMBOL(ipv6_setsockopt);
873 
874 #ifdef CONFIG_COMPAT
875 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
876 			   char __user *optval, unsigned int optlen)
877 {
878 	int err;
879 
880 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
881 		if (udp_prot.compat_setsockopt != NULL)
882 			return udp_prot.compat_setsockopt(sk, level, optname,
883 							  optval, optlen);
884 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
885 	}
886 
887 	if (level != SOL_IPV6)
888 		return -ENOPROTOOPT;
889 
890 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
891 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
892 			ipv6_setsockopt);
893 
894 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
895 #ifdef CONFIG_NETFILTER
896 	/* we need to exclude all possible ENOPROTOOPTs except default case */
897 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
898 	    optname != IPV6_XFRM_POLICY) {
899 		lock_sock(sk);
900 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
901 					   optval, optlen);
902 		release_sock(sk);
903 	}
904 #endif
905 	return err;
906 }
907 
908 EXPORT_SYMBOL(compat_ipv6_setsockopt);
909 #endif
910 
911 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
912 				  int optname, char __user *optval, int len)
913 {
914 	struct ipv6_opt_hdr *hdr;
915 
916 	if (!opt)
917 		return 0;
918 
919 	switch(optname) {
920 	case IPV6_HOPOPTS:
921 		hdr = opt->hopopt;
922 		break;
923 	case IPV6_RTHDRDSTOPTS:
924 		hdr = opt->dst0opt;
925 		break;
926 	case IPV6_RTHDR:
927 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
928 		break;
929 	case IPV6_DSTOPTS:
930 		hdr = opt->dst1opt;
931 		break;
932 	default:
933 		return -EINVAL;	/* should not happen */
934 	}
935 
936 	if (!hdr)
937 		return 0;
938 
939 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
940 	if (copy_to_user(optval, hdr, len))
941 		return -EFAULT;
942 	return len;
943 }
944 
945 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
946 		    char __user *optval, int __user *optlen, unsigned flags)
947 {
948 	struct ipv6_pinfo *np = inet6_sk(sk);
949 	int len;
950 	int val;
951 
952 	if (ip6_mroute_opt(optname))
953 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
954 
955 	if (get_user(len, optlen))
956 		return -EFAULT;
957 	switch (optname) {
958 	case IPV6_ADDRFORM:
959 		if (sk->sk_protocol != IPPROTO_UDP &&
960 		    sk->sk_protocol != IPPROTO_UDPLITE &&
961 		    sk->sk_protocol != IPPROTO_TCP)
962 			return -ENOPROTOOPT;
963 		if (sk->sk_state != TCP_ESTABLISHED)
964 			return -ENOTCONN;
965 		val = sk->sk_family;
966 		break;
967 	case MCAST_MSFILTER:
968 	{
969 		struct group_filter gsf;
970 		int err;
971 
972 		if (len < GROUP_FILTER_SIZE(0))
973 			return -EINVAL;
974 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
975 			return -EFAULT;
976 		if (gsf.gf_group.ss_family != AF_INET6)
977 			return -EADDRNOTAVAIL;
978 		lock_sock(sk);
979 		err = ip6_mc_msfget(sk, &gsf,
980 			(struct group_filter __user *)optval, optlen);
981 		release_sock(sk);
982 		return err;
983 	}
984 
985 	case IPV6_2292PKTOPTIONS:
986 	{
987 		struct msghdr msg;
988 		struct sk_buff *skb;
989 
990 		if (sk->sk_type != SOCK_STREAM)
991 			return -ENOPROTOOPT;
992 
993 		msg.msg_control = optval;
994 		msg.msg_controllen = len;
995 		msg.msg_flags = flags;
996 
997 		lock_sock(sk);
998 		skb = np->pktoptions;
999 		if (skb)
1000 			atomic_inc(&skb->users);
1001 		release_sock(sk);
1002 
1003 		if (skb) {
1004 			int err = datagram_recv_ctl(sk, &msg, skb);
1005 			kfree_skb(skb);
1006 			if (err)
1007 				return err;
1008 		} else {
1009 			if (np->rxopt.bits.rxinfo) {
1010 				struct in6_pktinfo src_info;
1011 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1012 					np->sticky_pktinfo.ipi6_ifindex;
1013 				src_info.ipi6_addr = np->mcast_oif ? np->daddr : np->sticky_pktinfo.ipi6_addr;
1014 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1015 			}
1016 			if (np->rxopt.bits.rxhlim) {
1017 				int hlim = np->mcast_hops;
1018 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1019 			}
1020 			if (np->rxopt.bits.rxtclass) {
1021 				int tclass = np->rcv_tclass;
1022 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1023 			}
1024 			if (np->rxopt.bits.rxoinfo) {
1025 				struct in6_pktinfo src_info;
1026 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1027 					np->sticky_pktinfo.ipi6_ifindex;
1028 				src_info.ipi6_addr = np->mcast_oif ? np->daddr : np->sticky_pktinfo.ipi6_addr;
1029 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1030 			}
1031 			if (np->rxopt.bits.rxohlim) {
1032 				int hlim = np->mcast_hops;
1033 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1034 			}
1035 		}
1036 		len -= msg.msg_controllen;
1037 		return put_user(len, optlen);
1038 	}
1039 	case IPV6_MTU:
1040 	{
1041 		struct dst_entry *dst;
1042 
1043 		val = 0;
1044 		rcu_read_lock();
1045 		dst = __sk_dst_get(sk);
1046 		if (dst)
1047 			val = dst_mtu(dst);
1048 		rcu_read_unlock();
1049 		if (!val)
1050 			return -ENOTCONN;
1051 		break;
1052 	}
1053 
1054 	case IPV6_V6ONLY:
1055 		val = np->ipv6only;
1056 		break;
1057 
1058 	case IPV6_RECVPKTINFO:
1059 		val = np->rxopt.bits.rxinfo;
1060 		break;
1061 
1062 	case IPV6_2292PKTINFO:
1063 		val = np->rxopt.bits.rxoinfo;
1064 		break;
1065 
1066 	case IPV6_RECVHOPLIMIT:
1067 		val = np->rxopt.bits.rxhlim;
1068 		break;
1069 
1070 	case IPV6_2292HOPLIMIT:
1071 		val = np->rxopt.bits.rxohlim;
1072 		break;
1073 
1074 	case IPV6_RECVRTHDR:
1075 		val = np->rxopt.bits.srcrt;
1076 		break;
1077 
1078 	case IPV6_2292RTHDR:
1079 		val = np->rxopt.bits.osrcrt;
1080 		break;
1081 
1082 	case IPV6_HOPOPTS:
1083 	case IPV6_RTHDRDSTOPTS:
1084 	case IPV6_RTHDR:
1085 	case IPV6_DSTOPTS:
1086 	{
1087 
1088 		lock_sock(sk);
1089 		len = ipv6_getsockopt_sticky(sk, np->opt,
1090 					     optname, optval, len);
1091 		release_sock(sk);
1092 		/* check if ipv6_getsockopt_sticky() returns err code */
1093 		if (len < 0)
1094 			return len;
1095 		return put_user(len, optlen);
1096 	}
1097 
1098 	case IPV6_RECVHOPOPTS:
1099 		val = np->rxopt.bits.hopopts;
1100 		break;
1101 
1102 	case IPV6_2292HOPOPTS:
1103 		val = np->rxopt.bits.ohopopts;
1104 		break;
1105 
1106 	case IPV6_RECVDSTOPTS:
1107 		val = np->rxopt.bits.dstopts;
1108 		break;
1109 
1110 	case IPV6_2292DSTOPTS:
1111 		val = np->rxopt.bits.odstopts;
1112 		break;
1113 
1114 	case IPV6_TCLASS:
1115 		val = np->tclass;
1116 		break;
1117 
1118 	case IPV6_RECVTCLASS:
1119 		val = np->rxopt.bits.rxtclass;
1120 		break;
1121 
1122 	case IPV6_FLOWINFO:
1123 		val = np->rxopt.bits.rxflow;
1124 		break;
1125 
1126 	case IPV6_RECVPATHMTU:
1127 		val = np->rxopt.bits.rxpmtu;
1128 		break;
1129 
1130 	case IPV6_PATHMTU:
1131 	{
1132 		struct dst_entry *dst;
1133 		struct ip6_mtuinfo mtuinfo;
1134 
1135 		if (len < sizeof(mtuinfo))
1136 			return -EINVAL;
1137 
1138 		len = sizeof(mtuinfo);
1139 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1140 
1141 		rcu_read_lock();
1142 		dst = __sk_dst_get(sk);
1143 		if (dst)
1144 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1145 		rcu_read_unlock();
1146 		if (!mtuinfo.ip6m_mtu)
1147 			return -ENOTCONN;
1148 
1149 		if (put_user(len, optlen))
1150 			return -EFAULT;
1151 		if (copy_to_user(optval, &mtuinfo, len))
1152 			return -EFAULT;
1153 
1154 		return 0;
1155 		break;
1156 	}
1157 
1158 	case IPV6_TRANSPARENT:
1159 		val = inet_sk(sk)->transparent;
1160 		break;
1161 
1162 	case IPV6_RECVORIGDSTADDR:
1163 		val = np->rxopt.bits.rxorigdstaddr;
1164 		break;
1165 
1166 	case IPV6_UNICAST_HOPS:
1167 	case IPV6_MULTICAST_HOPS:
1168 	{
1169 		struct dst_entry *dst;
1170 
1171 		if (optname == IPV6_UNICAST_HOPS)
1172 			val = np->hop_limit;
1173 		else
1174 			val = np->mcast_hops;
1175 
1176 		if (val < 0) {
1177 			rcu_read_lock();
1178 			dst = __sk_dst_get(sk);
1179 			if (dst)
1180 				val = ip6_dst_hoplimit(dst);
1181 			rcu_read_unlock();
1182 		}
1183 
1184 		if (val < 0)
1185 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1186 		break;
1187 	}
1188 
1189 	case IPV6_MULTICAST_LOOP:
1190 		val = np->mc_loop;
1191 		break;
1192 
1193 	case IPV6_MULTICAST_IF:
1194 		val = np->mcast_oif;
1195 		break;
1196 
1197 	case IPV6_UNICAST_IF:
1198 		val = (__force int)htonl((__u32) np->ucast_oif);
1199 		break;
1200 
1201 	case IPV6_MTU_DISCOVER:
1202 		val = np->pmtudisc;
1203 		break;
1204 
1205 	case IPV6_RECVERR:
1206 		val = np->recverr;
1207 		break;
1208 
1209 	case IPV6_FLOWINFO_SEND:
1210 		val = np->sndflow;
1211 		break;
1212 
1213 	case IPV6_ADDR_PREFERENCES:
1214 		val = 0;
1215 
1216 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1217 			val |= IPV6_PREFER_SRC_TMP;
1218 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1219 			val |= IPV6_PREFER_SRC_PUBLIC;
1220 		else {
1221 			/* XXX: should we return system default? */
1222 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1223 		}
1224 
1225 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1226 			val |= IPV6_PREFER_SRC_COA;
1227 		else
1228 			val |= IPV6_PREFER_SRC_HOME;
1229 		break;
1230 
1231 	case IPV6_MINHOPCOUNT:
1232 		val = np->min_hopcount;
1233 		break;
1234 
1235 	case IPV6_DONTFRAG:
1236 		val = np->dontfrag;
1237 		break;
1238 
1239 	default:
1240 		return -ENOPROTOOPT;
1241 	}
1242 	len = min_t(unsigned int, sizeof(int), len);
1243 	if(put_user(len, optlen))
1244 		return -EFAULT;
1245 	if(copy_to_user(optval,&val,len))
1246 		return -EFAULT;
1247 	return 0;
1248 }
1249 
1250 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1251 		    char __user *optval, int __user *optlen)
1252 {
1253 	int err;
1254 
1255 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1256 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1257 
1258 	if(level != SOL_IPV6)
1259 		return -ENOPROTOOPT;
1260 
1261 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1262 #ifdef CONFIG_NETFILTER
1263 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1264 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1265 		int len;
1266 
1267 		if (get_user(len, optlen))
1268 			return -EFAULT;
1269 
1270 		lock_sock(sk);
1271 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1272 				&len);
1273 		release_sock(sk);
1274 		if (err >= 0)
1275 			err = put_user(len, optlen);
1276 	}
1277 #endif
1278 	return err;
1279 }
1280 
1281 EXPORT_SYMBOL(ipv6_getsockopt);
1282 
1283 #ifdef CONFIG_COMPAT
1284 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1285 			   char __user *optval, int __user *optlen)
1286 {
1287 	int err;
1288 
1289 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1290 		if (udp_prot.compat_getsockopt != NULL)
1291 			return udp_prot.compat_getsockopt(sk, level, optname,
1292 							  optval, optlen);
1293 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1294 	}
1295 
1296 	if (level != SOL_IPV6)
1297 		return -ENOPROTOOPT;
1298 
1299 	if (optname == MCAST_MSFILTER)
1300 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1301 			ipv6_getsockopt);
1302 
1303 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1304 				 MSG_CMSG_COMPAT);
1305 #ifdef CONFIG_NETFILTER
1306 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1307 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1308 		int len;
1309 
1310 		if (get_user(len, optlen))
1311 			return -EFAULT;
1312 
1313 		lock_sock(sk);
1314 		err = compat_nf_getsockopt(sk, PF_INET6,
1315 					   optname, optval, &len);
1316 		release_sock(sk);
1317 		if (err >= 0)
1318 			err = put_user(len, optlen);
1319 	}
1320 #endif
1321 	return err;
1322 }
1323 
1324 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1325 #endif
1326 
1327