xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision e8e0929d)
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, unsigned 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 		/* RFC 3542, 6.5: default traffic class of 0x0 */
319 		if (val == -1)
320 			val = 0;
321 		np->tclass = val;
322 		retv = 0;
323 		break;
324 
325 	case IPV6_RECVTCLASS:
326 		if (optlen < sizeof(int))
327 			goto e_inval;
328 		np->rxopt.bits.rxtclass = valbool;
329 		retv = 0;
330 		break;
331 
332 	case IPV6_FLOWINFO:
333 		if (optlen < sizeof(int))
334 			goto e_inval;
335 		np->rxopt.bits.rxflow = valbool;
336 		retv = 0;
337 		break;
338 
339 	case IPV6_HOPOPTS:
340 	case IPV6_RTHDRDSTOPTS:
341 	case IPV6_RTHDR:
342 	case IPV6_DSTOPTS:
343 	{
344 		struct ipv6_txoptions *opt;
345 
346 		/* remove any sticky options header with a zero option
347 		 * length, per RFC3542.
348 		 */
349 		if (optlen == 0)
350 			optval = NULL;
351 		else if (optval == NULL)
352 			goto e_inval;
353 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
354 			 optlen & 0x7 || optlen > 8 * 255)
355 			goto e_inval;
356 
357 		/* hop-by-hop / destination options are privileged option */
358 		retv = -EPERM;
359 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
360 			break;
361 
362 		opt = ipv6_renew_options(sk, np->opt, optname,
363 					 (struct ipv6_opt_hdr __user *)optval,
364 					 optlen);
365 		if (IS_ERR(opt)) {
366 			retv = PTR_ERR(opt);
367 			break;
368 		}
369 
370 		/* routing header option needs extra check */
371 		retv = -EINVAL;
372 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
373 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
374 			switch (rthdr->type) {
375 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
376 			case IPV6_SRCRT_TYPE_2:
377 				if (rthdr->hdrlen != 2 ||
378 				    rthdr->segments_left != 1)
379 					goto sticky_done;
380 
381 				break;
382 #endif
383 			default:
384 				goto sticky_done;
385 			}
386 		}
387 
388 		retv = 0;
389 		opt = ipv6_update_options(sk, opt);
390 sticky_done:
391 		if (opt)
392 			sock_kfree_s(sk, opt, opt->tot_len);
393 		break;
394 	}
395 
396 	case IPV6_PKTINFO:
397 	{
398 		struct in6_pktinfo pkt;
399 
400 		if (optlen == 0)
401 			goto e_inval;
402 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
403 			goto e_inval;
404 
405 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
406 				retv = -EFAULT;
407 				break;
408 		}
409 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
410 			goto e_inval;
411 
412 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
413 		ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
414 		retv = 0;
415 		break;
416 	}
417 
418 	case IPV6_2292PKTOPTIONS:
419 	{
420 		struct ipv6_txoptions *opt = NULL;
421 		struct msghdr msg;
422 		struct flowi fl;
423 		int junk;
424 
425 		fl.fl6_flowlabel = 0;
426 		fl.oif = sk->sk_bound_dev_if;
427 
428 		if (optlen == 0)
429 			goto update;
430 
431 		/* 1K is probably excessive
432 		 * 1K is surely not enough, 2K per standard header is 16K.
433 		 */
434 		retv = -EINVAL;
435 		if (optlen > 64*1024)
436 			break;
437 
438 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
439 		retv = -ENOBUFS;
440 		if (opt == NULL)
441 			break;
442 
443 		memset(opt, 0, sizeof(*opt));
444 		opt->tot_len = sizeof(*opt) + optlen;
445 		retv = -EFAULT;
446 		if (copy_from_user(opt+1, optval, optlen))
447 			goto done;
448 
449 		msg.msg_controllen = optlen;
450 		msg.msg_control = (void*)(opt+1);
451 
452 		retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
453 		if (retv)
454 			goto done;
455 update:
456 		retv = 0;
457 		opt = ipv6_update_options(sk, opt);
458 done:
459 		if (opt)
460 			sock_kfree_s(sk, opt, opt->tot_len);
461 		break;
462 	}
463 	case IPV6_UNICAST_HOPS:
464 		if (optlen < sizeof(int))
465 			goto e_inval;
466 		if (val > 255 || val < -1)
467 			goto e_inval;
468 		np->hop_limit = val;
469 		retv = 0;
470 		break;
471 
472 	case IPV6_MULTICAST_HOPS:
473 		if (sk->sk_type == SOCK_STREAM)
474 			break;
475 		if (optlen < sizeof(int))
476 			goto e_inval;
477 		if (val > 255 || val < -1)
478 			goto e_inval;
479 		np->mcast_hops = val;
480 		retv = 0;
481 		break;
482 
483 	case IPV6_MULTICAST_LOOP:
484 		if (optlen < sizeof(int))
485 			goto e_inval;
486 		if (val != valbool)
487 			goto e_inval;
488 		np->mc_loop = valbool;
489 		retv = 0;
490 		break;
491 
492 	case IPV6_MULTICAST_IF:
493 		if (sk->sk_type == SOCK_STREAM)
494 			break;
495 		if (optlen < sizeof(int))
496 			goto e_inval;
497 
498 		if (val) {
499 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
500 				goto e_inval;
501 
502 			if (__dev_get_by_index(net, val) == NULL) {
503 				retv = -ENODEV;
504 				break;
505 			}
506 		}
507 		np->mcast_oif = val;
508 		retv = 0;
509 		break;
510 	case IPV6_ADD_MEMBERSHIP:
511 	case IPV6_DROP_MEMBERSHIP:
512 	{
513 		struct ipv6_mreq mreq;
514 
515 		if (optlen < sizeof(struct ipv6_mreq))
516 			goto e_inval;
517 
518 		retv = -EPROTO;
519 		if (inet_sk(sk)->is_icsk)
520 			break;
521 
522 		retv = -EFAULT;
523 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
524 			break;
525 
526 		if (optname == IPV6_ADD_MEMBERSHIP)
527 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
528 		else
529 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
530 		break;
531 	}
532 	case IPV6_JOIN_ANYCAST:
533 	case IPV6_LEAVE_ANYCAST:
534 	{
535 		struct ipv6_mreq mreq;
536 
537 		if (optlen < sizeof(struct ipv6_mreq))
538 			goto e_inval;
539 
540 		retv = -EFAULT;
541 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
542 			break;
543 
544 		if (optname == IPV6_JOIN_ANYCAST)
545 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
546 		else
547 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
548 		break;
549 	}
550 	case MCAST_JOIN_GROUP:
551 	case MCAST_LEAVE_GROUP:
552 	{
553 		struct group_req greq;
554 		struct sockaddr_in6 *psin6;
555 
556 		if (optlen < sizeof(struct group_req))
557 			goto e_inval;
558 
559 		retv = -EFAULT;
560 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
561 			break;
562 		if (greq.gr_group.ss_family != AF_INET6) {
563 			retv = -EADDRNOTAVAIL;
564 			break;
565 		}
566 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
567 		if (optname == MCAST_JOIN_GROUP)
568 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
569 				&psin6->sin6_addr);
570 		else
571 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
572 				&psin6->sin6_addr);
573 		break;
574 	}
575 	case MCAST_JOIN_SOURCE_GROUP:
576 	case MCAST_LEAVE_SOURCE_GROUP:
577 	case MCAST_BLOCK_SOURCE:
578 	case MCAST_UNBLOCK_SOURCE:
579 	{
580 		struct group_source_req greqs;
581 		int omode, add;
582 
583 		if (optlen < sizeof(struct group_source_req))
584 			goto e_inval;
585 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
586 			retv = -EFAULT;
587 			break;
588 		}
589 		if (greqs.gsr_group.ss_family != AF_INET6 ||
590 		    greqs.gsr_source.ss_family != AF_INET6) {
591 			retv = -EADDRNOTAVAIL;
592 			break;
593 		}
594 		if (optname == MCAST_BLOCK_SOURCE) {
595 			omode = MCAST_EXCLUDE;
596 			add = 1;
597 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
598 			omode = MCAST_EXCLUDE;
599 			add = 0;
600 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
601 			struct sockaddr_in6 *psin6;
602 
603 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
604 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
605 				&psin6->sin6_addr);
606 			/* prior join w/ different source is ok */
607 			if (retv && retv != -EADDRINUSE)
608 				break;
609 			omode = MCAST_INCLUDE;
610 			add = 1;
611 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
612 			omode = MCAST_INCLUDE;
613 			add = 0;
614 		}
615 		retv = ip6_mc_source(add, omode, sk, &greqs);
616 		break;
617 	}
618 	case MCAST_MSFILTER:
619 	{
620 		extern int sysctl_mld_max_msf;
621 		struct group_filter *gsf;
622 
623 		if (optlen < GROUP_FILTER_SIZE(0))
624 			goto e_inval;
625 		if (optlen > sysctl_optmem_max) {
626 			retv = -ENOBUFS;
627 			break;
628 		}
629 		gsf = kmalloc(optlen,GFP_KERNEL);
630 		if (!gsf) {
631 			retv = -ENOBUFS;
632 			break;
633 		}
634 		retv = -EFAULT;
635 		if (copy_from_user(gsf, optval, optlen)) {
636 			kfree(gsf);
637 			break;
638 		}
639 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
640 		if (gsf->gf_numsrc >= 0x1ffffffU ||
641 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
642 			kfree(gsf);
643 			retv = -ENOBUFS;
644 			break;
645 		}
646 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
647 			kfree(gsf);
648 			retv = -EINVAL;
649 			break;
650 		}
651 		retv = ip6_mc_msfilter(sk, gsf);
652 		kfree(gsf);
653 
654 		break;
655 	}
656 	case IPV6_ROUTER_ALERT:
657 		if (optlen < sizeof(int))
658 			goto e_inval;
659 		retv = ip6_ra_control(sk, val);
660 		break;
661 	case IPV6_MTU_DISCOVER:
662 		if (optlen < sizeof(int))
663 			goto e_inval;
664 		if (val<0 || val>3)
665 			goto e_inval;
666 		np->pmtudisc = val;
667 		retv = 0;
668 		break;
669 	case IPV6_MTU:
670 		if (optlen < sizeof(int))
671 			goto e_inval;
672 		if (val && val < IPV6_MIN_MTU)
673 			goto e_inval;
674 		np->frag_size = val;
675 		retv = 0;
676 		break;
677 	case IPV6_RECVERR:
678 		if (optlen < sizeof(int))
679 			goto e_inval;
680 		np->recverr = valbool;
681 		if (!val)
682 			skb_queue_purge(&sk->sk_error_queue);
683 		retv = 0;
684 		break;
685 	case IPV6_FLOWINFO_SEND:
686 		if (optlen < sizeof(int))
687 			goto e_inval;
688 		np->sndflow = valbool;
689 		retv = 0;
690 		break;
691 	case IPV6_FLOWLABEL_MGR:
692 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
693 		break;
694 	case IPV6_IPSEC_POLICY:
695 	case IPV6_XFRM_POLICY:
696 		retv = -EPERM;
697 		if (!capable(CAP_NET_ADMIN))
698 			break;
699 		retv = xfrm_user_policy(sk, optname, optval, optlen);
700 		break;
701 
702 	case IPV6_ADDR_PREFERENCES:
703 	    {
704 		unsigned int pref = 0;
705 		unsigned int prefmask = ~0;
706 
707 		if (optlen < sizeof(int))
708 			goto e_inval;
709 
710 		retv = -EINVAL;
711 
712 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
713 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
714 			       IPV6_PREFER_SRC_TMP|
715 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
716 		case IPV6_PREFER_SRC_PUBLIC:
717 			pref |= IPV6_PREFER_SRC_PUBLIC;
718 			break;
719 		case IPV6_PREFER_SRC_TMP:
720 			pref |= IPV6_PREFER_SRC_TMP;
721 			break;
722 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
723 			break;
724 		case 0:
725 			goto pref_skip_pubtmp;
726 		default:
727 			goto e_inval;
728 		}
729 
730 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
731 			      IPV6_PREFER_SRC_TMP);
732 pref_skip_pubtmp:
733 
734 		/* check HOME/COA conflicts */
735 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
736 		case IPV6_PREFER_SRC_HOME:
737 			break;
738 		case IPV6_PREFER_SRC_COA:
739 			pref |= IPV6_PREFER_SRC_COA;
740 		case 0:
741 			goto pref_skip_coa;
742 		default:
743 			goto e_inval;
744 		}
745 
746 		prefmask &= ~IPV6_PREFER_SRC_COA;
747 pref_skip_coa:
748 
749 		/* check CGA/NONCGA conflicts */
750 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
751 		case IPV6_PREFER_SRC_CGA:
752 		case IPV6_PREFER_SRC_NONCGA:
753 		case 0:
754 			break;
755 		default:
756 			goto e_inval;
757 		}
758 
759 		np->srcprefs = (np->srcprefs & prefmask) | pref;
760 		retv = 0;
761 
762 		break;
763 	    }
764 	}
765 
766 	release_sock(sk);
767 
768 	return retv;
769 
770 e_inval:
771 	release_sock(sk);
772 	return -EINVAL;
773 }
774 
775 int ipv6_setsockopt(struct sock *sk, int level, int optname,
776 		    char __user *optval, unsigned int optlen)
777 {
778 	int err;
779 
780 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
781 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
782 
783 	if (level != SOL_IPV6)
784 		return -ENOPROTOOPT;
785 
786 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
787 #ifdef CONFIG_NETFILTER
788 	/* we need to exclude all possible ENOPROTOOPTs except default case */
789 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
790 			optname != IPV6_XFRM_POLICY) {
791 		lock_sock(sk);
792 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
793 				optlen);
794 		release_sock(sk);
795 	}
796 #endif
797 	return err;
798 }
799 
800 EXPORT_SYMBOL(ipv6_setsockopt);
801 
802 #ifdef CONFIG_COMPAT
803 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
804 			   char __user *optval, unsigned int optlen)
805 {
806 	int err;
807 
808 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
809 		if (udp_prot.compat_setsockopt != NULL)
810 			return udp_prot.compat_setsockopt(sk, level, optname,
811 							  optval, optlen);
812 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
813 	}
814 
815 	if (level != SOL_IPV6)
816 		return -ENOPROTOOPT;
817 
818 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
819 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
820 			ipv6_setsockopt);
821 
822 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
823 #ifdef CONFIG_NETFILTER
824 	/* we need to exclude all possible ENOPROTOOPTs except default case */
825 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
826 	    optname != IPV6_XFRM_POLICY) {
827 		lock_sock(sk);
828 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
829 					   optval, optlen);
830 		release_sock(sk);
831 	}
832 #endif
833 	return err;
834 }
835 
836 EXPORT_SYMBOL(compat_ipv6_setsockopt);
837 #endif
838 
839 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
840 				  int optname, char __user *optval, int len)
841 {
842 	struct ipv6_opt_hdr *hdr;
843 
844 	if (!opt)
845 		return 0;
846 
847 	switch(optname) {
848 	case IPV6_HOPOPTS:
849 		hdr = opt->hopopt;
850 		break;
851 	case IPV6_RTHDRDSTOPTS:
852 		hdr = opt->dst0opt;
853 		break;
854 	case IPV6_RTHDR:
855 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
856 		break;
857 	case IPV6_DSTOPTS:
858 		hdr = opt->dst1opt;
859 		break;
860 	default:
861 		return -EINVAL;	/* should not happen */
862 	}
863 
864 	if (!hdr)
865 		return 0;
866 
867 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
868 	if (copy_to_user(optval, hdr, len))
869 		return -EFAULT;
870 	return len;
871 }
872 
873 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
874 		    char __user *optval, int __user *optlen)
875 {
876 	struct ipv6_pinfo *np = inet6_sk(sk);
877 	int len;
878 	int val;
879 
880 	if (ip6_mroute_opt(optname))
881 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
882 
883 	if (get_user(len, optlen))
884 		return -EFAULT;
885 	switch (optname) {
886 	case IPV6_ADDRFORM:
887 		if (sk->sk_protocol != IPPROTO_UDP &&
888 		    sk->sk_protocol != IPPROTO_UDPLITE &&
889 		    sk->sk_protocol != IPPROTO_TCP)
890 			return -ENOPROTOOPT;
891 		if (sk->sk_state != TCP_ESTABLISHED)
892 			return -ENOTCONN;
893 		val = sk->sk_family;
894 		break;
895 	case MCAST_MSFILTER:
896 	{
897 		struct group_filter gsf;
898 		int err;
899 
900 		if (len < GROUP_FILTER_SIZE(0))
901 			return -EINVAL;
902 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
903 			return -EFAULT;
904 		if (gsf.gf_group.ss_family != AF_INET6)
905 			return -EADDRNOTAVAIL;
906 		lock_sock(sk);
907 		err = ip6_mc_msfget(sk, &gsf,
908 			(struct group_filter __user *)optval, optlen);
909 		release_sock(sk);
910 		return err;
911 	}
912 
913 	case IPV6_2292PKTOPTIONS:
914 	{
915 		struct msghdr msg;
916 		struct sk_buff *skb;
917 
918 		if (sk->sk_type != SOCK_STREAM)
919 			return -ENOPROTOOPT;
920 
921 		msg.msg_control = optval;
922 		msg.msg_controllen = len;
923 		msg.msg_flags = 0;
924 
925 		lock_sock(sk);
926 		skb = np->pktoptions;
927 		if (skb)
928 			atomic_inc(&skb->users);
929 		release_sock(sk);
930 
931 		if (skb) {
932 			int err = datagram_recv_ctl(sk, &msg, skb);
933 			kfree_skb(skb);
934 			if (err)
935 				return err;
936 		} else {
937 			if (np->rxopt.bits.rxinfo) {
938 				struct in6_pktinfo src_info;
939 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
940 					np->sticky_pktinfo.ipi6_ifindex;
941 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
942 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
943 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
944 			}
945 			if (np->rxopt.bits.rxhlim) {
946 				int hlim = np->mcast_hops;
947 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
948 			}
949 			if (np->rxopt.bits.rxoinfo) {
950 				struct in6_pktinfo src_info;
951 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
952 					np->sticky_pktinfo.ipi6_ifindex;
953 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
954 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
955 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
956 			}
957 			if (np->rxopt.bits.rxohlim) {
958 				int hlim = np->mcast_hops;
959 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
960 			}
961 		}
962 		len -= msg.msg_controllen;
963 		return put_user(len, optlen);
964 	}
965 	case IPV6_MTU:
966 	{
967 		struct dst_entry *dst;
968 		val = 0;
969 		lock_sock(sk);
970 		dst = sk_dst_get(sk);
971 		if (dst) {
972 			val = dst_mtu(dst);
973 			dst_release(dst);
974 		}
975 		release_sock(sk);
976 		if (!val)
977 			return -ENOTCONN;
978 		break;
979 	}
980 
981 	case IPV6_V6ONLY:
982 		val = np->ipv6only;
983 		break;
984 
985 	case IPV6_RECVPKTINFO:
986 		val = np->rxopt.bits.rxinfo;
987 		break;
988 
989 	case IPV6_2292PKTINFO:
990 		val = np->rxopt.bits.rxoinfo;
991 		break;
992 
993 	case IPV6_RECVHOPLIMIT:
994 		val = np->rxopt.bits.rxhlim;
995 		break;
996 
997 	case IPV6_2292HOPLIMIT:
998 		val = np->rxopt.bits.rxohlim;
999 		break;
1000 
1001 	case IPV6_RECVRTHDR:
1002 		val = np->rxopt.bits.srcrt;
1003 		break;
1004 
1005 	case IPV6_2292RTHDR:
1006 		val = np->rxopt.bits.osrcrt;
1007 		break;
1008 
1009 	case IPV6_HOPOPTS:
1010 	case IPV6_RTHDRDSTOPTS:
1011 	case IPV6_RTHDR:
1012 	case IPV6_DSTOPTS:
1013 	{
1014 
1015 		lock_sock(sk);
1016 		len = ipv6_getsockopt_sticky(sk, np->opt,
1017 					     optname, optval, len);
1018 		release_sock(sk);
1019 		/* check if ipv6_getsockopt_sticky() returns err code */
1020 		if (len < 0)
1021 			return len;
1022 		return put_user(len, optlen);
1023 	}
1024 
1025 	case IPV6_RECVHOPOPTS:
1026 		val = np->rxopt.bits.hopopts;
1027 		break;
1028 
1029 	case IPV6_2292HOPOPTS:
1030 		val = np->rxopt.bits.ohopopts;
1031 		break;
1032 
1033 	case IPV6_RECVDSTOPTS:
1034 		val = np->rxopt.bits.dstopts;
1035 		break;
1036 
1037 	case IPV6_2292DSTOPTS:
1038 		val = np->rxopt.bits.odstopts;
1039 		break;
1040 
1041 	case IPV6_TCLASS:
1042 		val = np->tclass;
1043 		break;
1044 
1045 	case IPV6_RECVTCLASS:
1046 		val = np->rxopt.bits.rxtclass;
1047 		break;
1048 
1049 	case IPV6_FLOWINFO:
1050 		val = np->rxopt.bits.rxflow;
1051 		break;
1052 
1053 	case IPV6_UNICAST_HOPS:
1054 	case IPV6_MULTICAST_HOPS:
1055 	{
1056 		struct dst_entry *dst;
1057 
1058 		if (optname == IPV6_UNICAST_HOPS)
1059 			val = np->hop_limit;
1060 		else
1061 			val = np->mcast_hops;
1062 
1063 		dst = sk_dst_get(sk);
1064 		if (dst) {
1065 			if (val < 0)
1066 				val = ip6_dst_hoplimit(dst);
1067 			dst_release(dst);
1068 		}
1069 		if (val < 0)
1070 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1071 		break;
1072 	}
1073 
1074 	case IPV6_MULTICAST_LOOP:
1075 		val = np->mc_loop;
1076 		break;
1077 
1078 	case IPV6_MULTICAST_IF:
1079 		val = np->mcast_oif;
1080 		break;
1081 
1082 	case IPV6_MTU_DISCOVER:
1083 		val = np->pmtudisc;
1084 		break;
1085 
1086 	case IPV6_RECVERR:
1087 		val = np->recverr;
1088 		break;
1089 
1090 	case IPV6_FLOWINFO_SEND:
1091 		val = np->sndflow;
1092 		break;
1093 
1094 	case IPV6_ADDR_PREFERENCES:
1095 		val = 0;
1096 
1097 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1098 			val |= IPV6_PREFER_SRC_TMP;
1099 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1100 			val |= IPV6_PREFER_SRC_PUBLIC;
1101 		else {
1102 			/* XXX: should we return system default? */
1103 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1104 		}
1105 
1106 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1107 			val |= IPV6_PREFER_SRC_COA;
1108 		else
1109 			val |= IPV6_PREFER_SRC_HOME;
1110 		break;
1111 
1112 	default:
1113 		return -ENOPROTOOPT;
1114 	}
1115 	len = min_t(unsigned int, sizeof(int), len);
1116 	if(put_user(len, optlen))
1117 		return -EFAULT;
1118 	if(copy_to_user(optval,&val,len))
1119 		return -EFAULT;
1120 	return 0;
1121 }
1122 
1123 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1124 		    char __user *optval, int __user *optlen)
1125 {
1126 	int err;
1127 
1128 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1129 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1130 
1131 	if(level != SOL_IPV6)
1132 		return -ENOPROTOOPT;
1133 
1134 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1135 #ifdef CONFIG_NETFILTER
1136 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1137 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1138 		int len;
1139 
1140 		if (get_user(len, optlen))
1141 			return -EFAULT;
1142 
1143 		lock_sock(sk);
1144 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1145 				&len);
1146 		release_sock(sk);
1147 		if (err >= 0)
1148 			err = put_user(len, optlen);
1149 	}
1150 #endif
1151 	return err;
1152 }
1153 
1154 EXPORT_SYMBOL(ipv6_getsockopt);
1155 
1156 #ifdef CONFIG_COMPAT
1157 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1158 			   char __user *optval, int __user *optlen)
1159 {
1160 	int err;
1161 
1162 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1163 		if (udp_prot.compat_getsockopt != NULL)
1164 			return udp_prot.compat_getsockopt(sk, level, optname,
1165 							  optval, optlen);
1166 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1167 	}
1168 
1169 	if (level != SOL_IPV6)
1170 		return -ENOPROTOOPT;
1171 
1172 	if (optname == MCAST_MSFILTER)
1173 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1174 			ipv6_getsockopt);
1175 
1176 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1177 #ifdef CONFIG_NETFILTER
1178 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1179 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1180 		int len;
1181 
1182 		if (get_user(len, optlen))
1183 			return -EFAULT;
1184 
1185 		lock_sock(sk);
1186 		err = compat_nf_getsockopt(sk, PF_INET6,
1187 					   optname, optval, &len);
1188 		release_sock(sk);
1189 		if (err >= 0)
1190 			err = put_user(len, optlen);
1191 	}
1192 #endif
1193 	return err;
1194 }
1195 
1196 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1197 #endif
1198 
1199