xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision b04b4f78)
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 
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/transp_v6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/inet_common.h>
49 #include <net/tcp.h>
50 #include <net/udp.h>
51 #include <net/udplite.h>
52 #include <net/xfrm.h>
53 #include <net/compat.h>
54 
55 #include <asm/uaccess.h>
56 
57 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
58 
59 struct ip6_ra_chain *ip6_ra_chain;
60 DEFINE_RWLOCK(ip6_ra_lock);
61 
62 int ip6_ra_control(struct sock *sk, int sel)
63 {
64 	struct ip6_ra_chain *ra, *new_ra, **rap;
65 
66 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
67 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
68 		return -ENOPROTOOPT;
69 
70 	new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
71 
72 	write_lock_bh(&ip6_ra_lock);
73 	for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
74 		if (ra->sk == sk) {
75 			if (sel>=0) {
76 				write_unlock_bh(&ip6_ra_lock);
77 				kfree(new_ra);
78 				return -EADDRINUSE;
79 			}
80 
81 			*rap = ra->next;
82 			write_unlock_bh(&ip6_ra_lock);
83 
84 			sock_put(sk);
85 			kfree(ra);
86 			return 0;
87 		}
88 	}
89 	if (new_ra == NULL) {
90 		write_unlock_bh(&ip6_ra_lock);
91 		return -ENOBUFS;
92 	}
93 	new_ra->sk = sk;
94 	new_ra->sel = sel;
95 	new_ra->next = ra;
96 	*rap = new_ra;
97 	sock_hold(sk);
98 	write_unlock_bh(&ip6_ra_lock);
99 	return 0;
100 }
101 
102 static
103 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
104 					   struct ipv6_txoptions *opt)
105 {
106 	if (inet_sk(sk)->is_icsk) {
107 		if (opt &&
108 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
109 		    inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
110 			struct inet_connection_sock *icsk = inet_csk(sk);
111 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
112 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
113 		}
114 		opt = xchg(&inet6_sk(sk)->opt, opt);
115 	} else {
116 		write_lock(&sk->sk_dst_lock);
117 		opt = xchg(&inet6_sk(sk)->opt, opt);
118 		write_unlock(&sk->sk_dst_lock);
119 	}
120 	sk_dst_reset(sk);
121 
122 	return opt;
123 }
124 
125 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
126 		    char __user *optval, int optlen)
127 {
128 	struct ipv6_pinfo *np = inet6_sk(sk);
129 	struct net *net = sock_net(sk);
130 	int val, valbool;
131 	int retv = -ENOPROTOOPT;
132 
133 	if (optval == NULL)
134 		val=0;
135 	else {
136 		if (optlen >= sizeof(int)) {
137 			if (get_user(val, (int __user *) optval))
138 				return -EFAULT;
139 		} else
140 			val = 0;
141 	}
142 
143 	valbool = (val!=0);
144 
145 	if (ip6_mroute_opt(optname))
146 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
147 
148 	lock_sock(sk);
149 
150 	switch (optname) {
151 
152 	case IPV6_ADDRFORM:
153 		if (optlen < sizeof(int))
154 			goto e_inval;
155 		if (val == PF_INET) {
156 			struct ipv6_txoptions *opt;
157 			struct sk_buff *pktopt;
158 
159 			if (sk->sk_type == SOCK_RAW)
160 				break;
161 
162 			if (sk->sk_protocol == IPPROTO_UDP ||
163 			    sk->sk_protocol == IPPROTO_UDPLITE) {
164 				struct udp_sock *up = udp_sk(sk);
165 				if (up->pending == AF_INET6) {
166 					retv = -EBUSY;
167 					break;
168 				}
169 			} else if (sk->sk_protocol != IPPROTO_TCP)
170 				break;
171 
172 			if (sk->sk_state != TCP_ESTABLISHED) {
173 				retv = -ENOTCONN;
174 				break;
175 			}
176 
177 			if (ipv6_only_sock(sk) ||
178 			    !ipv6_addr_v4mapped(&np->daddr)) {
179 				retv = -EADDRNOTAVAIL;
180 				break;
181 			}
182 
183 			fl6_free_socklist(sk);
184 			ipv6_sock_mc_close(sk);
185 
186 			/*
187 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
188 			 * remove it from the refcnt debug socks count in the
189 			 * original family...
190 			 */
191 			sk_refcnt_debug_dec(sk);
192 
193 			if (sk->sk_protocol == IPPROTO_TCP) {
194 				struct inet_connection_sock *icsk = inet_csk(sk);
195 				local_bh_disable();
196 				sock_prot_inuse_add(net, sk->sk_prot, -1);
197 				sock_prot_inuse_add(net, &tcp_prot, 1);
198 				local_bh_enable();
199 				sk->sk_prot = &tcp_prot;
200 				icsk->icsk_af_ops = &ipv4_specific;
201 				sk->sk_socket->ops = &inet_stream_ops;
202 				sk->sk_family = PF_INET;
203 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
204 			} else {
205 				struct proto *prot = &udp_prot;
206 
207 				if (sk->sk_protocol == IPPROTO_UDPLITE)
208 					prot = &udplite_prot;
209 				local_bh_disable();
210 				sock_prot_inuse_add(net, sk->sk_prot, -1);
211 				sock_prot_inuse_add(net, prot, 1);
212 				local_bh_enable();
213 				sk->sk_prot = prot;
214 				sk->sk_socket->ops = &inet_dgram_ops;
215 				sk->sk_family = PF_INET;
216 			}
217 			opt = xchg(&np->opt, NULL);
218 			if (opt)
219 				sock_kfree_s(sk, opt, opt->tot_len);
220 			pktopt = xchg(&np->pktoptions, NULL);
221 			kfree_skb(pktopt);
222 
223 			sk->sk_destruct = inet_sock_destruct;
224 			/*
225 			 * ... and add it to the refcnt debug socks count
226 			 * in the new family. -acme
227 			 */
228 			sk_refcnt_debug_inc(sk);
229 			module_put(THIS_MODULE);
230 			retv = 0;
231 			break;
232 		}
233 		goto e_inval;
234 
235 	case IPV6_V6ONLY:
236 		if (optlen < sizeof(int) ||
237 		    inet_sk(sk)->num)
238 			goto e_inval;
239 		np->ipv6only = valbool;
240 		retv = 0;
241 		break;
242 
243 	case IPV6_RECVPKTINFO:
244 		if (optlen < sizeof(int))
245 			goto e_inval;
246 		np->rxopt.bits.rxinfo = valbool;
247 		retv = 0;
248 		break;
249 
250 	case IPV6_2292PKTINFO:
251 		if (optlen < sizeof(int))
252 			goto e_inval;
253 		np->rxopt.bits.rxoinfo = valbool;
254 		retv = 0;
255 		break;
256 
257 	case IPV6_RECVHOPLIMIT:
258 		if (optlen < sizeof(int))
259 			goto e_inval;
260 		np->rxopt.bits.rxhlim = valbool;
261 		retv = 0;
262 		break;
263 
264 	case IPV6_2292HOPLIMIT:
265 		if (optlen < sizeof(int))
266 			goto e_inval;
267 		np->rxopt.bits.rxohlim = valbool;
268 		retv = 0;
269 		break;
270 
271 	case IPV6_RECVRTHDR:
272 		if (optlen < sizeof(int))
273 			goto e_inval;
274 		np->rxopt.bits.srcrt = valbool;
275 		retv = 0;
276 		break;
277 
278 	case IPV6_2292RTHDR:
279 		if (optlen < sizeof(int))
280 			goto e_inval;
281 		np->rxopt.bits.osrcrt = valbool;
282 		retv = 0;
283 		break;
284 
285 	case IPV6_RECVHOPOPTS:
286 		if (optlen < sizeof(int))
287 			goto e_inval;
288 		np->rxopt.bits.hopopts = valbool;
289 		retv = 0;
290 		break;
291 
292 	case IPV6_2292HOPOPTS:
293 		if (optlen < sizeof(int))
294 			goto e_inval;
295 		np->rxopt.bits.ohopopts = valbool;
296 		retv = 0;
297 		break;
298 
299 	case IPV6_RECVDSTOPTS:
300 		if (optlen < sizeof(int))
301 			goto e_inval;
302 		np->rxopt.bits.dstopts = valbool;
303 		retv = 0;
304 		break;
305 
306 	case IPV6_2292DSTOPTS:
307 		if (optlen < sizeof(int))
308 			goto e_inval;
309 		np->rxopt.bits.odstopts = valbool;
310 		retv = 0;
311 		break;
312 
313 	case IPV6_TCLASS:
314 		if (optlen < sizeof(int))
315 			goto e_inval;
316 		if (val < -1 || val > 0xff)
317 			goto e_inval;
318 		np->tclass = val;
319 		retv = 0;
320 		break;
321 
322 	case IPV6_RECVTCLASS:
323 		if (optlen < sizeof(int))
324 			goto e_inval;
325 		np->rxopt.bits.rxtclass = valbool;
326 		retv = 0;
327 		break;
328 
329 	case IPV6_FLOWINFO:
330 		if (optlen < sizeof(int))
331 			goto e_inval;
332 		np->rxopt.bits.rxflow = valbool;
333 		retv = 0;
334 		break;
335 
336 	case IPV6_HOPOPTS:
337 	case IPV6_RTHDRDSTOPTS:
338 	case IPV6_RTHDR:
339 	case IPV6_DSTOPTS:
340 	{
341 		struct ipv6_txoptions *opt;
342 
343 		/* remove any sticky options header with a zero option
344 		 * length, per RFC3542.
345 		 */
346 		if (optlen == 0)
347 			optval = NULL;
348 		else if (optval == NULL)
349 			goto e_inval;
350 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
351 			 optlen & 0x7 || optlen > 8 * 255)
352 			goto e_inval;
353 
354 		/* hop-by-hop / destination options are privileged option */
355 		retv = -EPERM;
356 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
357 			break;
358 
359 		opt = ipv6_renew_options(sk, np->opt, optname,
360 					 (struct ipv6_opt_hdr __user *)optval,
361 					 optlen);
362 		if (IS_ERR(opt)) {
363 			retv = PTR_ERR(opt);
364 			break;
365 		}
366 
367 		/* routing header option needs extra check */
368 		retv = -EINVAL;
369 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
370 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
371 			switch (rthdr->type) {
372 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
373 			case IPV6_SRCRT_TYPE_2:
374 				if (rthdr->hdrlen != 2 ||
375 				    rthdr->segments_left != 1)
376 					goto sticky_done;
377 
378 				break;
379 #endif
380 			default:
381 				goto sticky_done;
382 			}
383 		}
384 
385 		retv = 0;
386 		opt = ipv6_update_options(sk, opt);
387 sticky_done:
388 		if (opt)
389 			sock_kfree_s(sk, opt, opt->tot_len);
390 		break;
391 	}
392 
393 	case IPV6_PKTINFO:
394 	{
395 		struct in6_pktinfo pkt;
396 
397 		if (optlen == 0)
398 			goto e_inval;
399 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
400 			goto e_inval;
401 
402 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
403 				retv = -EFAULT;
404 				break;
405 		}
406 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
407 			goto e_inval;
408 
409 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
410 		ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
411 		retv = 0;
412 		break;
413 	}
414 
415 	case IPV6_2292PKTOPTIONS:
416 	{
417 		struct ipv6_txoptions *opt = NULL;
418 		struct msghdr msg;
419 		struct flowi fl;
420 		int junk;
421 
422 		fl.fl6_flowlabel = 0;
423 		fl.oif = sk->sk_bound_dev_if;
424 
425 		if (optlen == 0)
426 			goto update;
427 
428 		/* 1K is probably excessive
429 		 * 1K is surely not enough, 2K per standard header is 16K.
430 		 */
431 		retv = -EINVAL;
432 		if (optlen > 64*1024)
433 			break;
434 
435 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
436 		retv = -ENOBUFS;
437 		if (opt == NULL)
438 			break;
439 
440 		memset(opt, 0, sizeof(*opt));
441 		opt->tot_len = sizeof(*opt) + optlen;
442 		retv = -EFAULT;
443 		if (copy_from_user(opt+1, optval, optlen))
444 			goto done;
445 
446 		msg.msg_controllen = optlen;
447 		msg.msg_control = (void*)(opt+1);
448 
449 		retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
450 		if (retv)
451 			goto done;
452 update:
453 		retv = 0;
454 		opt = ipv6_update_options(sk, opt);
455 done:
456 		if (opt)
457 			sock_kfree_s(sk, opt, opt->tot_len);
458 		break;
459 	}
460 	case IPV6_UNICAST_HOPS:
461 		if (optlen < sizeof(int))
462 			goto e_inval;
463 		if (val > 255 || val < -1)
464 			goto e_inval;
465 		np->hop_limit = val;
466 		retv = 0;
467 		break;
468 
469 	case IPV6_MULTICAST_HOPS:
470 		if (sk->sk_type == SOCK_STREAM)
471 			break;
472 		if (optlen < sizeof(int))
473 			goto e_inval;
474 		if (val > 255 || val < -1)
475 			goto e_inval;
476 		np->mcast_hops = val;
477 		retv = 0;
478 		break;
479 
480 	case IPV6_MULTICAST_LOOP:
481 		if (optlen < sizeof(int))
482 			goto e_inval;
483 		if (val != valbool)
484 			goto e_inval;
485 		np->mc_loop = valbool;
486 		retv = 0;
487 		break;
488 
489 	case IPV6_MULTICAST_IF:
490 		if (sk->sk_type == SOCK_STREAM)
491 			break;
492 		if (optlen < sizeof(int))
493 			goto e_inval;
494 
495 		if (val) {
496 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
497 				goto e_inval;
498 
499 			if (__dev_get_by_index(net, val) == NULL) {
500 				retv = -ENODEV;
501 				break;
502 			}
503 		}
504 		np->mcast_oif = val;
505 		retv = 0;
506 		break;
507 	case IPV6_ADD_MEMBERSHIP:
508 	case IPV6_DROP_MEMBERSHIP:
509 	{
510 		struct ipv6_mreq mreq;
511 
512 		if (optlen < sizeof(struct ipv6_mreq))
513 			goto e_inval;
514 
515 		retv = -EPROTO;
516 		if (inet_sk(sk)->is_icsk)
517 			break;
518 
519 		retv = -EFAULT;
520 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
521 			break;
522 
523 		if (optname == IPV6_ADD_MEMBERSHIP)
524 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
525 		else
526 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
527 		break;
528 	}
529 	case IPV6_JOIN_ANYCAST:
530 	case IPV6_LEAVE_ANYCAST:
531 	{
532 		struct ipv6_mreq mreq;
533 
534 		if (optlen < sizeof(struct ipv6_mreq))
535 			goto e_inval;
536 
537 		retv = -EFAULT;
538 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
539 			break;
540 
541 		if (optname == IPV6_JOIN_ANYCAST)
542 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
543 		else
544 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
545 		break;
546 	}
547 	case MCAST_JOIN_GROUP:
548 	case MCAST_LEAVE_GROUP:
549 	{
550 		struct group_req greq;
551 		struct sockaddr_in6 *psin6;
552 
553 		if (optlen < sizeof(struct group_req))
554 			goto e_inval;
555 
556 		retv = -EFAULT;
557 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
558 			break;
559 		if (greq.gr_group.ss_family != AF_INET6) {
560 			retv = -EADDRNOTAVAIL;
561 			break;
562 		}
563 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
564 		if (optname == MCAST_JOIN_GROUP)
565 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
566 				&psin6->sin6_addr);
567 		else
568 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
569 				&psin6->sin6_addr);
570 		break;
571 	}
572 	case MCAST_JOIN_SOURCE_GROUP:
573 	case MCAST_LEAVE_SOURCE_GROUP:
574 	case MCAST_BLOCK_SOURCE:
575 	case MCAST_UNBLOCK_SOURCE:
576 	{
577 		struct group_source_req greqs;
578 		int omode, add;
579 
580 		if (optlen < sizeof(struct group_source_req))
581 			goto e_inval;
582 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
583 			retv = -EFAULT;
584 			break;
585 		}
586 		if (greqs.gsr_group.ss_family != AF_INET6 ||
587 		    greqs.gsr_source.ss_family != AF_INET6) {
588 			retv = -EADDRNOTAVAIL;
589 			break;
590 		}
591 		if (optname == MCAST_BLOCK_SOURCE) {
592 			omode = MCAST_EXCLUDE;
593 			add = 1;
594 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
595 			omode = MCAST_EXCLUDE;
596 			add = 0;
597 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
598 			struct sockaddr_in6 *psin6;
599 
600 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
601 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
602 				&psin6->sin6_addr);
603 			/* prior join w/ different source is ok */
604 			if (retv && retv != -EADDRINUSE)
605 				break;
606 			omode = MCAST_INCLUDE;
607 			add = 1;
608 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
609 			omode = MCAST_INCLUDE;
610 			add = 0;
611 		}
612 		retv = ip6_mc_source(add, omode, sk, &greqs);
613 		break;
614 	}
615 	case MCAST_MSFILTER:
616 	{
617 		extern int sysctl_mld_max_msf;
618 		struct group_filter *gsf;
619 
620 		if (optlen < GROUP_FILTER_SIZE(0))
621 			goto e_inval;
622 		if (optlen > sysctl_optmem_max) {
623 			retv = -ENOBUFS;
624 			break;
625 		}
626 		gsf = kmalloc(optlen,GFP_KERNEL);
627 		if (!gsf) {
628 			retv = -ENOBUFS;
629 			break;
630 		}
631 		retv = -EFAULT;
632 		if (copy_from_user(gsf, optval, optlen)) {
633 			kfree(gsf);
634 			break;
635 		}
636 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
637 		if (gsf->gf_numsrc >= 0x1ffffffU ||
638 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
639 			kfree(gsf);
640 			retv = -ENOBUFS;
641 			break;
642 		}
643 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
644 			kfree(gsf);
645 			retv = -EINVAL;
646 			break;
647 		}
648 		retv = ip6_mc_msfilter(sk, gsf);
649 		kfree(gsf);
650 
651 		break;
652 	}
653 	case IPV6_ROUTER_ALERT:
654 		if (optlen < sizeof(int))
655 			goto e_inval;
656 		retv = ip6_ra_control(sk, val);
657 		break;
658 	case IPV6_MTU_DISCOVER:
659 		if (optlen < sizeof(int))
660 			goto e_inval;
661 		if (val<0 || val>3)
662 			goto e_inval;
663 		np->pmtudisc = val;
664 		retv = 0;
665 		break;
666 	case IPV6_MTU:
667 		if (optlen < sizeof(int))
668 			goto e_inval;
669 		if (val && val < IPV6_MIN_MTU)
670 			goto e_inval;
671 		np->frag_size = val;
672 		retv = 0;
673 		break;
674 	case IPV6_RECVERR:
675 		if (optlen < sizeof(int))
676 			goto e_inval;
677 		np->recverr = valbool;
678 		if (!val)
679 			skb_queue_purge(&sk->sk_error_queue);
680 		retv = 0;
681 		break;
682 	case IPV6_FLOWINFO_SEND:
683 		if (optlen < sizeof(int))
684 			goto e_inval;
685 		np->sndflow = valbool;
686 		retv = 0;
687 		break;
688 	case IPV6_FLOWLABEL_MGR:
689 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
690 		break;
691 	case IPV6_IPSEC_POLICY:
692 	case IPV6_XFRM_POLICY:
693 		retv = -EPERM;
694 		if (!capable(CAP_NET_ADMIN))
695 			break;
696 		retv = xfrm_user_policy(sk, optname, optval, optlen);
697 		break;
698 
699 	case IPV6_ADDR_PREFERENCES:
700 	    {
701 		unsigned int pref = 0;
702 		unsigned int prefmask = ~0;
703 
704 		if (optlen < sizeof(int))
705 			goto e_inval;
706 
707 		retv = -EINVAL;
708 
709 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
710 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
711 			       IPV6_PREFER_SRC_TMP|
712 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
713 		case IPV6_PREFER_SRC_PUBLIC:
714 			pref |= IPV6_PREFER_SRC_PUBLIC;
715 			break;
716 		case IPV6_PREFER_SRC_TMP:
717 			pref |= IPV6_PREFER_SRC_TMP;
718 			break;
719 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
720 			break;
721 		case 0:
722 			goto pref_skip_pubtmp;
723 		default:
724 			goto e_inval;
725 		}
726 
727 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
728 			      IPV6_PREFER_SRC_TMP);
729 pref_skip_pubtmp:
730 
731 		/* check HOME/COA conflicts */
732 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
733 		case IPV6_PREFER_SRC_HOME:
734 			break;
735 		case IPV6_PREFER_SRC_COA:
736 			pref |= IPV6_PREFER_SRC_COA;
737 		case 0:
738 			goto pref_skip_coa;
739 		default:
740 			goto e_inval;
741 		}
742 
743 		prefmask &= ~IPV6_PREFER_SRC_COA;
744 pref_skip_coa:
745 
746 		/* check CGA/NONCGA conflicts */
747 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
748 		case IPV6_PREFER_SRC_CGA:
749 		case IPV6_PREFER_SRC_NONCGA:
750 		case 0:
751 			break;
752 		default:
753 			goto e_inval;
754 		}
755 
756 		np->srcprefs = (np->srcprefs & prefmask) | pref;
757 		retv = 0;
758 
759 		break;
760 	    }
761 	}
762 
763 	release_sock(sk);
764 
765 	return retv;
766 
767 e_inval:
768 	release_sock(sk);
769 	return -EINVAL;
770 }
771 
772 int ipv6_setsockopt(struct sock *sk, int level, int optname,
773 		    char __user *optval, int optlen)
774 {
775 	int err;
776 
777 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
778 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
779 
780 	if (level != SOL_IPV6)
781 		return -ENOPROTOOPT;
782 
783 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
784 #ifdef CONFIG_NETFILTER
785 	/* we need to exclude all possible ENOPROTOOPTs except default case */
786 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
787 			optname != IPV6_XFRM_POLICY) {
788 		lock_sock(sk);
789 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
790 				optlen);
791 		release_sock(sk);
792 	}
793 #endif
794 	return err;
795 }
796 
797 EXPORT_SYMBOL(ipv6_setsockopt);
798 
799 #ifdef CONFIG_COMPAT
800 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
801 			   char __user *optval, int optlen)
802 {
803 	int err;
804 
805 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
806 		if (udp_prot.compat_setsockopt != NULL)
807 			return udp_prot.compat_setsockopt(sk, level, optname,
808 							  optval, optlen);
809 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
810 	}
811 
812 	if (level != SOL_IPV6)
813 		return -ENOPROTOOPT;
814 
815 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
816 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
817 			ipv6_setsockopt);
818 
819 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
820 #ifdef CONFIG_NETFILTER
821 	/* we need to exclude all possible ENOPROTOOPTs except default case */
822 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
823 	    optname != IPV6_XFRM_POLICY) {
824 		lock_sock(sk);
825 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
826 					   optval, optlen);
827 		release_sock(sk);
828 	}
829 #endif
830 	return err;
831 }
832 
833 EXPORT_SYMBOL(compat_ipv6_setsockopt);
834 #endif
835 
836 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
837 				  int optname, char __user *optval, int len)
838 {
839 	struct ipv6_opt_hdr *hdr;
840 
841 	if (!opt)
842 		return 0;
843 
844 	switch(optname) {
845 	case IPV6_HOPOPTS:
846 		hdr = opt->hopopt;
847 		break;
848 	case IPV6_RTHDRDSTOPTS:
849 		hdr = opt->dst0opt;
850 		break;
851 	case IPV6_RTHDR:
852 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
853 		break;
854 	case IPV6_DSTOPTS:
855 		hdr = opt->dst1opt;
856 		break;
857 	default:
858 		return -EINVAL;	/* should not happen */
859 	}
860 
861 	if (!hdr)
862 		return 0;
863 
864 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
865 	if (copy_to_user(optval, hdr, len))
866 		return -EFAULT;
867 	return len;
868 }
869 
870 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
871 		    char __user *optval, int __user *optlen)
872 {
873 	struct ipv6_pinfo *np = inet6_sk(sk);
874 	int len;
875 	int val;
876 
877 	if (ip6_mroute_opt(optname))
878 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
879 
880 	if (get_user(len, optlen))
881 		return -EFAULT;
882 	switch (optname) {
883 	case IPV6_ADDRFORM:
884 		if (sk->sk_protocol != IPPROTO_UDP &&
885 		    sk->sk_protocol != IPPROTO_UDPLITE &&
886 		    sk->sk_protocol != IPPROTO_TCP)
887 			return -ENOPROTOOPT;
888 		if (sk->sk_state != TCP_ESTABLISHED)
889 			return -ENOTCONN;
890 		val = sk->sk_family;
891 		break;
892 	case MCAST_MSFILTER:
893 	{
894 		struct group_filter gsf;
895 		int err;
896 
897 		if (len < GROUP_FILTER_SIZE(0))
898 			return -EINVAL;
899 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
900 			return -EFAULT;
901 		if (gsf.gf_group.ss_family != AF_INET6)
902 			return -EADDRNOTAVAIL;
903 		lock_sock(sk);
904 		err = ip6_mc_msfget(sk, &gsf,
905 			(struct group_filter __user *)optval, optlen);
906 		release_sock(sk);
907 		return err;
908 	}
909 
910 	case IPV6_2292PKTOPTIONS:
911 	{
912 		struct msghdr msg;
913 		struct sk_buff *skb;
914 
915 		if (sk->sk_type != SOCK_STREAM)
916 			return -ENOPROTOOPT;
917 
918 		msg.msg_control = optval;
919 		msg.msg_controllen = len;
920 		msg.msg_flags = 0;
921 
922 		lock_sock(sk);
923 		skb = np->pktoptions;
924 		if (skb)
925 			atomic_inc(&skb->users);
926 		release_sock(sk);
927 
928 		if (skb) {
929 			int err = datagram_recv_ctl(sk, &msg, skb);
930 			kfree_skb(skb);
931 			if (err)
932 				return err;
933 		} else {
934 			if (np->rxopt.bits.rxinfo) {
935 				struct in6_pktinfo src_info;
936 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
937 					np->sticky_pktinfo.ipi6_ifindex;
938 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
939 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
940 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
941 			}
942 			if (np->rxopt.bits.rxhlim) {
943 				int hlim = np->mcast_hops;
944 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
945 			}
946 			if (np->rxopt.bits.rxoinfo) {
947 				struct in6_pktinfo src_info;
948 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
949 					np->sticky_pktinfo.ipi6_ifindex;
950 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
951 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
952 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
953 			}
954 			if (np->rxopt.bits.rxohlim) {
955 				int hlim = np->mcast_hops;
956 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
957 			}
958 		}
959 		len -= msg.msg_controllen;
960 		return put_user(len, optlen);
961 	}
962 	case IPV6_MTU:
963 	{
964 		struct dst_entry *dst;
965 		val = 0;
966 		lock_sock(sk);
967 		dst = sk_dst_get(sk);
968 		if (dst) {
969 			val = dst_mtu(dst);
970 			dst_release(dst);
971 		}
972 		release_sock(sk);
973 		if (!val)
974 			return -ENOTCONN;
975 		break;
976 	}
977 
978 	case IPV6_V6ONLY:
979 		val = np->ipv6only;
980 		break;
981 
982 	case IPV6_RECVPKTINFO:
983 		val = np->rxopt.bits.rxinfo;
984 		break;
985 
986 	case IPV6_2292PKTINFO:
987 		val = np->rxopt.bits.rxoinfo;
988 		break;
989 
990 	case IPV6_RECVHOPLIMIT:
991 		val = np->rxopt.bits.rxhlim;
992 		break;
993 
994 	case IPV6_2292HOPLIMIT:
995 		val = np->rxopt.bits.rxohlim;
996 		break;
997 
998 	case IPV6_RECVRTHDR:
999 		val = np->rxopt.bits.srcrt;
1000 		break;
1001 
1002 	case IPV6_2292RTHDR:
1003 		val = np->rxopt.bits.osrcrt;
1004 		break;
1005 
1006 	case IPV6_HOPOPTS:
1007 	case IPV6_RTHDRDSTOPTS:
1008 	case IPV6_RTHDR:
1009 	case IPV6_DSTOPTS:
1010 	{
1011 
1012 		lock_sock(sk);
1013 		len = ipv6_getsockopt_sticky(sk, np->opt,
1014 					     optname, optval, len);
1015 		release_sock(sk);
1016 		/* check if ipv6_getsockopt_sticky() returns err code */
1017 		if (len < 0)
1018 			return len;
1019 		return put_user(len, optlen);
1020 	}
1021 
1022 	case IPV6_RECVHOPOPTS:
1023 		val = np->rxopt.bits.hopopts;
1024 		break;
1025 
1026 	case IPV6_2292HOPOPTS:
1027 		val = np->rxopt.bits.ohopopts;
1028 		break;
1029 
1030 	case IPV6_RECVDSTOPTS:
1031 		val = np->rxopt.bits.dstopts;
1032 		break;
1033 
1034 	case IPV6_2292DSTOPTS:
1035 		val = np->rxopt.bits.odstopts;
1036 		break;
1037 
1038 	case IPV6_TCLASS:
1039 		val = np->tclass;
1040 		if (val < 0)
1041 			val = 0;
1042 		break;
1043 
1044 	case IPV6_RECVTCLASS:
1045 		val = np->rxopt.bits.rxtclass;
1046 		break;
1047 
1048 	case IPV6_FLOWINFO:
1049 		val = np->rxopt.bits.rxflow;
1050 		break;
1051 
1052 	case IPV6_UNICAST_HOPS:
1053 	case IPV6_MULTICAST_HOPS:
1054 	{
1055 		struct dst_entry *dst;
1056 
1057 		if (optname == IPV6_UNICAST_HOPS)
1058 			val = np->hop_limit;
1059 		else
1060 			val = np->mcast_hops;
1061 
1062 		dst = sk_dst_get(sk);
1063 		if (dst) {
1064 			if (val < 0)
1065 				val = ip6_dst_hoplimit(dst);
1066 			dst_release(dst);
1067 		}
1068 		if (val < 0)
1069 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1070 		break;
1071 	}
1072 
1073 	case IPV6_MULTICAST_LOOP:
1074 		val = np->mc_loop;
1075 		break;
1076 
1077 	case IPV6_MULTICAST_IF:
1078 		val = np->mcast_oif;
1079 		break;
1080 
1081 	case IPV6_MTU_DISCOVER:
1082 		val = np->pmtudisc;
1083 		break;
1084 
1085 	case IPV6_RECVERR:
1086 		val = np->recverr;
1087 		break;
1088 
1089 	case IPV6_FLOWINFO_SEND:
1090 		val = np->sndflow;
1091 		break;
1092 
1093 	case IPV6_ADDR_PREFERENCES:
1094 		val = 0;
1095 
1096 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1097 			val |= IPV6_PREFER_SRC_TMP;
1098 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1099 			val |= IPV6_PREFER_SRC_PUBLIC;
1100 		else {
1101 			/* XXX: should we return system default? */
1102 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1103 		}
1104 
1105 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1106 			val |= IPV6_PREFER_SRC_COA;
1107 		else
1108 			val |= IPV6_PREFER_SRC_HOME;
1109 		break;
1110 
1111 	default:
1112 		return -ENOPROTOOPT;
1113 	}
1114 	len = min_t(unsigned int, sizeof(int), len);
1115 	if(put_user(len, optlen))
1116 		return -EFAULT;
1117 	if(copy_to_user(optval,&val,len))
1118 		return -EFAULT;
1119 	return 0;
1120 }
1121 
1122 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1123 		    char __user *optval, int __user *optlen)
1124 {
1125 	int err;
1126 
1127 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1128 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1129 
1130 	if(level != SOL_IPV6)
1131 		return -ENOPROTOOPT;
1132 
1133 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1134 #ifdef CONFIG_NETFILTER
1135 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1136 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1137 		int len;
1138 
1139 		if (get_user(len, optlen))
1140 			return -EFAULT;
1141 
1142 		lock_sock(sk);
1143 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1144 				&len);
1145 		release_sock(sk);
1146 		if (err >= 0)
1147 			err = put_user(len, optlen);
1148 	}
1149 #endif
1150 	return err;
1151 }
1152 
1153 EXPORT_SYMBOL(ipv6_getsockopt);
1154 
1155 #ifdef CONFIG_COMPAT
1156 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1157 			   char __user *optval, int __user *optlen)
1158 {
1159 	int err;
1160 
1161 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1162 		if (udp_prot.compat_getsockopt != NULL)
1163 			return udp_prot.compat_getsockopt(sk, level, optname,
1164 							  optval, optlen);
1165 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1166 	}
1167 
1168 	if (level != SOL_IPV6)
1169 		return -ENOPROTOOPT;
1170 
1171 	if (optname == MCAST_MSFILTER)
1172 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1173 			ipv6_getsockopt);
1174 
1175 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1176 #ifdef CONFIG_NETFILTER
1177 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1178 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1179 		int len;
1180 
1181 		if (get_user(len, optlen))
1182 			return -EFAULT;
1183 
1184 		lock_sock(sk);
1185 		err = compat_nf_getsockopt(sk, PF_INET6,
1186 					   optname, optval, &len);
1187 		release_sock(sk);
1188 		if (err >= 0)
1189 			err = put_user(len, optlen);
1190 	}
1191 #endif
1192 	return err;
1193 }
1194 
1195 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1196 #endif
1197 
1198