xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision 12eb4683)
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(&sk->sk_v6_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 && !ns_capable(net->user_ns, CAP_NET_ADMIN) &&
347 		    !ns_capable(net->user_ns, CAP_NET_RAW)) {
348 			retv = -EPERM;
349 			break;
350 		}
351 		if (optlen < sizeof(int))
352 			goto e_inval;
353 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
354 		inet_sk(sk)->transparent = valbool;
355 		retv = 0;
356 		break;
357 
358 	case IPV6_RECVORIGDSTADDR:
359 		if (optlen < sizeof(int))
360 			goto e_inval;
361 		np->rxopt.bits.rxorigdstaddr = valbool;
362 		retv = 0;
363 		break;
364 
365 	case IPV6_HOPOPTS:
366 	case IPV6_RTHDRDSTOPTS:
367 	case IPV6_RTHDR:
368 	case IPV6_DSTOPTS:
369 	{
370 		struct ipv6_txoptions *opt;
371 
372 		/* remove any sticky options header with a zero option
373 		 * length, per RFC3542.
374 		 */
375 		if (optlen == 0)
376 			optval = NULL;
377 		else if (optval == NULL)
378 			goto e_inval;
379 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
380 			 optlen & 0x7 || optlen > 8 * 255)
381 			goto e_inval;
382 
383 		/* hop-by-hop / destination options are privileged option */
384 		retv = -EPERM;
385 		if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
386 			break;
387 
388 		opt = ipv6_renew_options(sk, np->opt, optname,
389 					 (struct ipv6_opt_hdr __user *)optval,
390 					 optlen);
391 		if (IS_ERR(opt)) {
392 			retv = PTR_ERR(opt);
393 			break;
394 		}
395 
396 		/* routing header option needs extra check */
397 		retv = -EINVAL;
398 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
399 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
400 			switch (rthdr->type) {
401 #if IS_ENABLED(CONFIG_IPV6_MIP6)
402 			case IPV6_SRCRT_TYPE_2:
403 				if (rthdr->hdrlen != 2 ||
404 				    rthdr->segments_left != 1)
405 					goto sticky_done;
406 
407 				break;
408 #endif
409 			default:
410 				goto sticky_done;
411 			}
412 		}
413 
414 		retv = 0;
415 		opt = ipv6_update_options(sk, opt);
416 sticky_done:
417 		if (opt)
418 			sock_kfree_s(sk, opt, opt->tot_len);
419 		break;
420 	}
421 
422 	case IPV6_PKTINFO:
423 	{
424 		struct in6_pktinfo pkt;
425 
426 		if (optlen == 0)
427 			goto e_inval;
428 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
429 			goto e_inval;
430 
431 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
432 				retv = -EFAULT;
433 				break;
434 		}
435 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
436 			goto e_inval;
437 
438 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
439 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
440 		retv = 0;
441 		break;
442 	}
443 
444 	case IPV6_2292PKTOPTIONS:
445 	{
446 		struct ipv6_txoptions *opt = NULL;
447 		struct msghdr msg;
448 		struct flowi6 fl6;
449 		int junk;
450 
451 		memset(&fl6, 0, sizeof(fl6));
452 		fl6.flowi6_oif = sk->sk_bound_dev_if;
453 		fl6.flowi6_mark = sk->sk_mark;
454 
455 		if (optlen == 0)
456 			goto update;
457 
458 		/* 1K is probably excessive
459 		 * 1K is surely not enough, 2K per standard header is 16K.
460 		 */
461 		retv = -EINVAL;
462 		if (optlen > 64*1024)
463 			break;
464 
465 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
466 		retv = -ENOBUFS;
467 		if (opt == NULL)
468 			break;
469 
470 		memset(opt, 0, sizeof(*opt));
471 		opt->tot_len = sizeof(*opt) + optlen;
472 		retv = -EFAULT;
473 		if (copy_from_user(opt+1, optval, optlen))
474 			goto done;
475 
476 		msg.msg_controllen = optlen;
477 		msg.msg_control = (void*)(opt+1);
478 
479 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, opt, &junk,
480 					     &junk, &junk);
481 		if (retv)
482 			goto done;
483 update:
484 		retv = 0;
485 		opt = ipv6_update_options(sk, opt);
486 done:
487 		if (opt)
488 			sock_kfree_s(sk, opt, opt->tot_len);
489 		break;
490 	}
491 	case IPV6_UNICAST_HOPS:
492 		if (optlen < sizeof(int))
493 			goto e_inval;
494 		if (val > 255 || val < -1)
495 			goto e_inval;
496 		np->hop_limit = val;
497 		retv = 0;
498 		break;
499 
500 	case IPV6_MULTICAST_HOPS:
501 		if (sk->sk_type == SOCK_STREAM)
502 			break;
503 		if (optlen < sizeof(int))
504 			goto e_inval;
505 		if (val > 255 || val < -1)
506 			goto e_inval;
507 		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
508 		retv = 0;
509 		break;
510 
511 	case IPV6_MULTICAST_LOOP:
512 		if (optlen < sizeof(int))
513 			goto e_inval;
514 		if (val != valbool)
515 			goto e_inval;
516 		np->mc_loop = valbool;
517 		retv = 0;
518 		break;
519 
520 	case IPV6_UNICAST_IF:
521 	{
522 		struct net_device *dev = NULL;
523 		int ifindex;
524 
525 		if (optlen != sizeof(int))
526 			goto e_inval;
527 
528 		ifindex = (__force int)ntohl((__force __be32)val);
529 		if (ifindex == 0) {
530 			np->ucast_oif = 0;
531 			retv = 0;
532 			break;
533 		}
534 
535 		dev = dev_get_by_index(net, ifindex);
536 		retv = -EADDRNOTAVAIL;
537 		if (!dev)
538 			break;
539 		dev_put(dev);
540 
541 		retv = -EINVAL;
542 		if (sk->sk_bound_dev_if)
543 			break;
544 
545 		np->ucast_oif = ifindex;
546 		retv = 0;
547 		break;
548 	}
549 
550 	case IPV6_MULTICAST_IF:
551 		if (sk->sk_type == SOCK_STREAM)
552 			break;
553 		if (optlen < sizeof(int))
554 			goto e_inval;
555 
556 		if (val) {
557 			struct net_device *dev;
558 
559 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
560 				goto e_inval;
561 
562 			dev = dev_get_by_index(net, val);
563 			if (!dev) {
564 				retv = -ENODEV;
565 				break;
566 			}
567 			dev_put(dev);
568 		}
569 		np->mcast_oif = val;
570 		retv = 0;
571 		break;
572 	case IPV6_ADD_MEMBERSHIP:
573 	case IPV6_DROP_MEMBERSHIP:
574 	{
575 		struct ipv6_mreq mreq;
576 
577 		if (optlen < sizeof(struct ipv6_mreq))
578 			goto e_inval;
579 
580 		retv = -EPROTO;
581 		if (inet_sk(sk)->is_icsk)
582 			break;
583 
584 		retv = -EFAULT;
585 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
586 			break;
587 
588 		if (optname == IPV6_ADD_MEMBERSHIP)
589 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
590 		else
591 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
592 		break;
593 	}
594 	case IPV6_JOIN_ANYCAST:
595 	case IPV6_LEAVE_ANYCAST:
596 	{
597 		struct ipv6_mreq mreq;
598 
599 		if (optlen < sizeof(struct ipv6_mreq))
600 			goto e_inval;
601 
602 		retv = -EFAULT;
603 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
604 			break;
605 
606 		if (optname == IPV6_JOIN_ANYCAST)
607 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
608 		else
609 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
610 		break;
611 	}
612 	case MCAST_JOIN_GROUP:
613 	case MCAST_LEAVE_GROUP:
614 	{
615 		struct group_req greq;
616 		struct sockaddr_in6 *psin6;
617 
618 		if (optlen < sizeof(struct group_req))
619 			goto e_inval;
620 
621 		retv = -EFAULT;
622 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
623 			break;
624 		if (greq.gr_group.ss_family != AF_INET6) {
625 			retv = -EADDRNOTAVAIL;
626 			break;
627 		}
628 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
629 		if (optname == MCAST_JOIN_GROUP)
630 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
631 				&psin6->sin6_addr);
632 		else
633 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
634 				&psin6->sin6_addr);
635 		break;
636 	}
637 	case MCAST_JOIN_SOURCE_GROUP:
638 	case MCAST_LEAVE_SOURCE_GROUP:
639 	case MCAST_BLOCK_SOURCE:
640 	case MCAST_UNBLOCK_SOURCE:
641 	{
642 		struct group_source_req greqs;
643 		int omode, add;
644 
645 		if (optlen < sizeof(struct group_source_req))
646 			goto e_inval;
647 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
648 			retv = -EFAULT;
649 			break;
650 		}
651 		if (greqs.gsr_group.ss_family != AF_INET6 ||
652 		    greqs.gsr_source.ss_family != AF_INET6) {
653 			retv = -EADDRNOTAVAIL;
654 			break;
655 		}
656 		if (optname == MCAST_BLOCK_SOURCE) {
657 			omode = MCAST_EXCLUDE;
658 			add = 1;
659 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
660 			omode = MCAST_EXCLUDE;
661 			add = 0;
662 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
663 			struct sockaddr_in6 *psin6;
664 
665 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
666 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
667 				&psin6->sin6_addr);
668 			/* prior join w/ different source is ok */
669 			if (retv && retv != -EADDRINUSE)
670 				break;
671 			omode = MCAST_INCLUDE;
672 			add = 1;
673 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
674 			omode = MCAST_INCLUDE;
675 			add = 0;
676 		}
677 		retv = ip6_mc_source(add, omode, sk, &greqs);
678 		break;
679 	}
680 	case MCAST_MSFILTER:
681 	{
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 (!ns_capable(net->user_ns, 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 		retv = 0;
832 		break;
833 	case IPV6_DONTFRAG:
834 		np->dontfrag = valbool;
835 		retv = 0;
836 		break;
837 	}
838 
839 	release_sock(sk);
840 
841 	return retv;
842 
843 e_inval:
844 	release_sock(sk);
845 	return -EINVAL;
846 }
847 
848 int ipv6_setsockopt(struct sock *sk, int level, int optname,
849 		    char __user *optval, unsigned int optlen)
850 {
851 	int err;
852 
853 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
854 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
855 
856 	if (level != SOL_IPV6)
857 		return -ENOPROTOOPT;
858 
859 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
860 #ifdef CONFIG_NETFILTER
861 	/* we need to exclude all possible ENOPROTOOPTs except default case */
862 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
863 			optname != IPV6_XFRM_POLICY) {
864 		lock_sock(sk);
865 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
866 				optlen);
867 		release_sock(sk);
868 	}
869 #endif
870 	return err;
871 }
872 
873 EXPORT_SYMBOL(ipv6_setsockopt);
874 
875 #ifdef CONFIG_COMPAT
876 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
877 			   char __user *optval, unsigned int optlen)
878 {
879 	int err;
880 
881 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
882 		if (udp_prot.compat_setsockopt != NULL)
883 			return udp_prot.compat_setsockopt(sk, level, optname,
884 							  optval, optlen);
885 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
886 	}
887 
888 	if (level != SOL_IPV6)
889 		return -ENOPROTOOPT;
890 
891 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
892 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
893 			ipv6_setsockopt);
894 
895 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
896 #ifdef CONFIG_NETFILTER
897 	/* we need to exclude all possible ENOPROTOOPTs except default case */
898 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
899 	    optname != IPV6_XFRM_POLICY) {
900 		lock_sock(sk);
901 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
902 					   optval, optlen);
903 		release_sock(sk);
904 	}
905 #endif
906 	return err;
907 }
908 
909 EXPORT_SYMBOL(compat_ipv6_setsockopt);
910 #endif
911 
912 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
913 				  int optname, char __user *optval, int len)
914 {
915 	struct ipv6_opt_hdr *hdr;
916 
917 	if (!opt)
918 		return 0;
919 
920 	switch(optname) {
921 	case IPV6_HOPOPTS:
922 		hdr = opt->hopopt;
923 		break;
924 	case IPV6_RTHDRDSTOPTS:
925 		hdr = opt->dst0opt;
926 		break;
927 	case IPV6_RTHDR:
928 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
929 		break;
930 	case IPV6_DSTOPTS:
931 		hdr = opt->dst1opt;
932 		break;
933 	default:
934 		return -EINVAL;	/* should not happen */
935 	}
936 
937 	if (!hdr)
938 		return 0;
939 
940 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
941 	if (copy_to_user(optval, hdr, len))
942 		return -EFAULT;
943 	return len;
944 }
945 
946 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
947 		    char __user *optval, int __user *optlen, unsigned int flags)
948 {
949 	struct ipv6_pinfo *np = inet6_sk(sk);
950 	int len;
951 	int val;
952 
953 	if (ip6_mroute_opt(optname))
954 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
955 
956 	if (get_user(len, optlen))
957 		return -EFAULT;
958 	switch (optname) {
959 	case IPV6_ADDRFORM:
960 		if (sk->sk_protocol != IPPROTO_UDP &&
961 		    sk->sk_protocol != IPPROTO_UDPLITE &&
962 		    sk->sk_protocol != IPPROTO_TCP)
963 			return -ENOPROTOOPT;
964 		if (sk->sk_state != TCP_ESTABLISHED)
965 			return -ENOTCONN;
966 		val = sk->sk_family;
967 		break;
968 	case MCAST_MSFILTER:
969 	{
970 		struct group_filter gsf;
971 		int err;
972 
973 		if (len < GROUP_FILTER_SIZE(0))
974 			return -EINVAL;
975 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
976 			return -EFAULT;
977 		if (gsf.gf_group.ss_family != AF_INET6)
978 			return -EADDRNOTAVAIL;
979 		lock_sock(sk);
980 		err = ip6_mc_msfget(sk, &gsf,
981 			(struct group_filter __user *)optval, optlen);
982 		release_sock(sk);
983 		return err;
984 	}
985 
986 	case IPV6_2292PKTOPTIONS:
987 	{
988 		struct msghdr msg;
989 		struct sk_buff *skb;
990 
991 		if (sk->sk_type != SOCK_STREAM)
992 			return -ENOPROTOOPT;
993 
994 		msg.msg_control = optval;
995 		msg.msg_controllen = len;
996 		msg.msg_flags = flags;
997 
998 		lock_sock(sk);
999 		skb = np->pktoptions;
1000 		if (skb)
1001 			atomic_inc(&skb->users);
1002 		release_sock(sk);
1003 
1004 		if (skb) {
1005 			int err = ip6_datagram_recv_ctl(sk, &msg, skb);
1006 			kfree_skb(skb);
1007 			if (err)
1008 				return err;
1009 		} else {
1010 			if (np->rxopt.bits.rxinfo) {
1011 				struct in6_pktinfo src_info;
1012 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1013 					np->sticky_pktinfo.ipi6_ifindex;
1014 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1015 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1016 			}
1017 			if (np->rxopt.bits.rxhlim) {
1018 				int hlim = np->mcast_hops;
1019 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1020 			}
1021 			if (np->rxopt.bits.rxtclass) {
1022 				int tclass = np->rcv_tclass;
1023 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1024 			}
1025 			if (np->rxopt.bits.rxoinfo) {
1026 				struct in6_pktinfo src_info;
1027 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1028 					np->sticky_pktinfo.ipi6_ifindex;
1029 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1030 								     np->sticky_pktinfo.ipi6_addr;
1031 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1032 			}
1033 			if (np->rxopt.bits.rxohlim) {
1034 				int hlim = np->mcast_hops;
1035 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1036 			}
1037 		}
1038 		len -= msg.msg_controllen;
1039 		return put_user(len, optlen);
1040 	}
1041 	case IPV6_MTU:
1042 	{
1043 		struct dst_entry *dst;
1044 
1045 		val = 0;
1046 		rcu_read_lock();
1047 		dst = __sk_dst_get(sk);
1048 		if (dst)
1049 			val = dst_mtu(dst);
1050 		rcu_read_unlock();
1051 		if (!val)
1052 			return -ENOTCONN;
1053 		break;
1054 	}
1055 
1056 	case IPV6_V6ONLY:
1057 		val = np->ipv6only;
1058 		break;
1059 
1060 	case IPV6_RECVPKTINFO:
1061 		val = np->rxopt.bits.rxinfo;
1062 		break;
1063 
1064 	case IPV6_2292PKTINFO:
1065 		val = np->rxopt.bits.rxoinfo;
1066 		break;
1067 
1068 	case IPV6_RECVHOPLIMIT:
1069 		val = np->rxopt.bits.rxhlim;
1070 		break;
1071 
1072 	case IPV6_2292HOPLIMIT:
1073 		val = np->rxopt.bits.rxohlim;
1074 		break;
1075 
1076 	case IPV6_RECVRTHDR:
1077 		val = np->rxopt.bits.srcrt;
1078 		break;
1079 
1080 	case IPV6_2292RTHDR:
1081 		val = np->rxopt.bits.osrcrt;
1082 		break;
1083 
1084 	case IPV6_HOPOPTS:
1085 	case IPV6_RTHDRDSTOPTS:
1086 	case IPV6_RTHDR:
1087 	case IPV6_DSTOPTS:
1088 	{
1089 
1090 		lock_sock(sk);
1091 		len = ipv6_getsockopt_sticky(sk, np->opt,
1092 					     optname, optval, len);
1093 		release_sock(sk);
1094 		/* check if ipv6_getsockopt_sticky() returns err code */
1095 		if (len < 0)
1096 			return len;
1097 		return put_user(len, optlen);
1098 	}
1099 
1100 	case IPV6_RECVHOPOPTS:
1101 		val = np->rxopt.bits.hopopts;
1102 		break;
1103 
1104 	case IPV6_2292HOPOPTS:
1105 		val = np->rxopt.bits.ohopopts;
1106 		break;
1107 
1108 	case IPV6_RECVDSTOPTS:
1109 		val = np->rxopt.bits.dstopts;
1110 		break;
1111 
1112 	case IPV6_2292DSTOPTS:
1113 		val = np->rxopt.bits.odstopts;
1114 		break;
1115 
1116 	case IPV6_TCLASS:
1117 		val = np->tclass;
1118 		break;
1119 
1120 	case IPV6_RECVTCLASS:
1121 		val = np->rxopt.bits.rxtclass;
1122 		break;
1123 
1124 	case IPV6_FLOWINFO:
1125 		val = np->rxopt.bits.rxflow;
1126 		break;
1127 
1128 	case IPV6_RECVPATHMTU:
1129 		val = np->rxopt.bits.rxpmtu;
1130 		break;
1131 
1132 	case IPV6_PATHMTU:
1133 	{
1134 		struct dst_entry *dst;
1135 		struct ip6_mtuinfo mtuinfo;
1136 
1137 		if (len < sizeof(mtuinfo))
1138 			return -EINVAL;
1139 
1140 		len = sizeof(mtuinfo);
1141 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1142 
1143 		rcu_read_lock();
1144 		dst = __sk_dst_get(sk);
1145 		if (dst)
1146 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1147 		rcu_read_unlock();
1148 		if (!mtuinfo.ip6m_mtu)
1149 			return -ENOTCONN;
1150 
1151 		if (put_user(len, optlen))
1152 			return -EFAULT;
1153 		if (copy_to_user(optval, &mtuinfo, len))
1154 			return -EFAULT;
1155 
1156 		return 0;
1157 		break;
1158 	}
1159 
1160 	case IPV6_TRANSPARENT:
1161 		val = inet_sk(sk)->transparent;
1162 		break;
1163 
1164 	case IPV6_RECVORIGDSTADDR:
1165 		val = np->rxopt.bits.rxorigdstaddr;
1166 		break;
1167 
1168 	case IPV6_UNICAST_HOPS:
1169 	case IPV6_MULTICAST_HOPS:
1170 	{
1171 		struct dst_entry *dst;
1172 
1173 		if (optname == IPV6_UNICAST_HOPS)
1174 			val = np->hop_limit;
1175 		else
1176 			val = np->mcast_hops;
1177 
1178 		if (val < 0) {
1179 			rcu_read_lock();
1180 			dst = __sk_dst_get(sk);
1181 			if (dst)
1182 				val = ip6_dst_hoplimit(dst);
1183 			rcu_read_unlock();
1184 		}
1185 
1186 		if (val < 0)
1187 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1188 		break;
1189 	}
1190 
1191 	case IPV6_MULTICAST_LOOP:
1192 		val = np->mc_loop;
1193 		break;
1194 
1195 	case IPV6_MULTICAST_IF:
1196 		val = np->mcast_oif;
1197 		break;
1198 
1199 	case IPV6_UNICAST_IF:
1200 		val = (__force int)htonl((__u32) np->ucast_oif);
1201 		break;
1202 
1203 	case IPV6_MTU_DISCOVER:
1204 		val = np->pmtudisc;
1205 		break;
1206 
1207 	case IPV6_RECVERR:
1208 		val = np->recverr;
1209 		break;
1210 
1211 	case IPV6_FLOWINFO_SEND:
1212 		val = np->sndflow;
1213 		break;
1214 
1215 	case IPV6_FLOWLABEL_MGR:
1216 	{
1217 		struct in6_flowlabel_req freq;
1218 
1219 		if (len < sizeof(freq))
1220 			return -EINVAL;
1221 
1222 		if (copy_from_user(&freq, optval, sizeof(freq)))
1223 			return -EFAULT;
1224 
1225 		if (freq.flr_action != IPV6_FL_A_GET)
1226 			return -EINVAL;
1227 
1228 		len = sizeof(freq);
1229 		memset(&freq, 0, sizeof(freq));
1230 
1231 		val = ipv6_flowlabel_opt_get(sk, &freq);
1232 		if (val < 0)
1233 			return val;
1234 
1235 		if (put_user(len, optlen))
1236 			return -EFAULT;
1237 		if (copy_to_user(optval, &freq, len))
1238 			return -EFAULT;
1239 
1240 		return 0;
1241 	}
1242 
1243 	case IPV6_ADDR_PREFERENCES:
1244 		val = 0;
1245 
1246 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1247 			val |= IPV6_PREFER_SRC_TMP;
1248 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1249 			val |= IPV6_PREFER_SRC_PUBLIC;
1250 		else {
1251 			/* XXX: should we return system default? */
1252 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1253 		}
1254 
1255 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1256 			val |= IPV6_PREFER_SRC_COA;
1257 		else
1258 			val |= IPV6_PREFER_SRC_HOME;
1259 		break;
1260 
1261 	case IPV6_MINHOPCOUNT:
1262 		val = np->min_hopcount;
1263 		break;
1264 
1265 	case IPV6_DONTFRAG:
1266 		val = np->dontfrag;
1267 		break;
1268 
1269 	default:
1270 		return -ENOPROTOOPT;
1271 	}
1272 	len = min_t(unsigned int, sizeof(int), len);
1273 	if(put_user(len, optlen))
1274 		return -EFAULT;
1275 	if(copy_to_user(optval,&val,len))
1276 		return -EFAULT;
1277 	return 0;
1278 }
1279 
1280 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1281 		    char __user *optval, int __user *optlen)
1282 {
1283 	int err;
1284 
1285 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1286 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1287 
1288 	if(level != SOL_IPV6)
1289 		return -ENOPROTOOPT;
1290 
1291 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1292 #ifdef CONFIG_NETFILTER
1293 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1294 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1295 		int len;
1296 
1297 		if (get_user(len, optlen))
1298 			return -EFAULT;
1299 
1300 		lock_sock(sk);
1301 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1302 				&len);
1303 		release_sock(sk);
1304 		if (err >= 0)
1305 			err = put_user(len, optlen);
1306 	}
1307 #endif
1308 	return err;
1309 }
1310 
1311 EXPORT_SYMBOL(ipv6_getsockopt);
1312 
1313 #ifdef CONFIG_COMPAT
1314 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1315 			   char __user *optval, int __user *optlen)
1316 {
1317 	int err;
1318 
1319 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1320 		if (udp_prot.compat_getsockopt != NULL)
1321 			return udp_prot.compat_getsockopt(sk, level, optname,
1322 							  optval, optlen);
1323 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1324 	}
1325 
1326 	if (level != SOL_IPV6)
1327 		return -ENOPROTOOPT;
1328 
1329 	if (optname == MCAST_MSFILTER)
1330 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1331 			ipv6_getsockopt);
1332 
1333 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1334 				 MSG_CMSG_COMPAT);
1335 #ifdef CONFIG_NETFILTER
1336 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1337 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1338 		int len;
1339 
1340 		if (get_user(len, optlen))
1341 			return -EFAULT;
1342 
1343 		lock_sock(sk);
1344 		err = compat_nf_getsockopt(sk, PF_INET6,
1345 					   optname, optval, &len);
1346 		release_sock(sk);
1347 		if (err >= 0)
1348 			err = put_user(len, optlen);
1349 	}
1350 #endif
1351 	return err;
1352 }
1353 
1354 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1355 #endif
1356 
1357