xref: /linux/net/ipv6/ipv6_sockglue.c (revision 5d4a2e29fba5b2bef95b96a46b338ec4d76fa4fd)
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 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
59 
60 struct ip6_ra_chain *ip6_ra_chain;
61 DEFINE_RWLOCK(ip6_ra_lock);
62 
63 int ip6_ra_control(struct sock *sk, int sel)
64 {
65 	struct ip6_ra_chain *ra, *new_ra, **rap;
66 
67 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
68 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
69 		return -ENOPROTOOPT;
70 
71 	new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
72 
73 	write_lock_bh(&ip6_ra_lock);
74 	for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
75 		if (ra->sk == sk) {
76 			if (sel>=0) {
77 				write_unlock_bh(&ip6_ra_lock);
78 				kfree(new_ra);
79 				return -EADDRINUSE;
80 			}
81 
82 			*rap = ra->next;
83 			write_unlock_bh(&ip6_ra_lock);
84 
85 			sock_put(sk);
86 			kfree(ra);
87 			return 0;
88 		}
89 	}
90 	if (new_ra == NULL) {
91 		write_unlock_bh(&ip6_ra_lock);
92 		return -ENOBUFS;
93 	}
94 	new_ra->sk = sk;
95 	new_ra->sel = sel;
96 	new_ra->next = ra;
97 	*rap = new_ra;
98 	sock_hold(sk);
99 	write_unlock_bh(&ip6_ra_lock);
100 	return 0;
101 }
102 
103 static
104 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
105 					   struct ipv6_txoptions *opt)
106 {
107 	if (inet_sk(sk)->is_icsk) {
108 		if (opt &&
109 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
110 		    inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
111 			struct inet_connection_sock *icsk = inet_csk(sk);
112 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
113 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
114 		}
115 		opt = xchg(&inet6_sk(sk)->opt, opt);
116 	} else {
117 		spin_lock(&sk->sk_dst_lock);
118 		opt = xchg(&inet6_sk(sk)->opt, opt);
119 		spin_unlock(&sk->sk_dst_lock);
120 	}
121 	sk_dst_reset(sk);
122 
123 	return opt;
124 }
125 
126 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
127 		    char __user *optval, unsigned int optlen)
128 {
129 	struct ipv6_pinfo *np = inet6_sk(sk);
130 	struct net *net = sock_net(sk);
131 	int val, valbool;
132 	int retv = -ENOPROTOOPT;
133 
134 	if (optval == NULL)
135 		val=0;
136 	else {
137 		if (optlen >= sizeof(int)) {
138 			if (get_user(val, (int __user *) optval))
139 				return -EFAULT;
140 		} else
141 			val = 0;
142 	}
143 
144 	valbool = (val!=0);
145 
146 	if (ip6_mroute_opt(optname))
147 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
148 
149 	lock_sock(sk);
150 
151 	switch (optname) {
152 
153 	case IPV6_ADDRFORM:
154 		if (optlen < sizeof(int))
155 			goto e_inval;
156 		if (val == PF_INET) {
157 			struct ipv6_txoptions *opt;
158 			struct sk_buff *pktopt;
159 
160 			if (sk->sk_type == SOCK_RAW)
161 				break;
162 
163 			if (sk->sk_protocol == IPPROTO_UDP ||
164 			    sk->sk_protocol == IPPROTO_UDPLITE) {
165 				struct udp_sock *up = udp_sk(sk);
166 				if (up->pending == AF_INET6) {
167 					retv = -EBUSY;
168 					break;
169 				}
170 			} else if (sk->sk_protocol != IPPROTO_TCP)
171 				break;
172 
173 			if (sk->sk_state != TCP_ESTABLISHED) {
174 				retv = -ENOTCONN;
175 				break;
176 			}
177 
178 			if (ipv6_only_sock(sk) ||
179 			    !ipv6_addr_v4mapped(&np->daddr)) {
180 				retv = -EADDRNOTAVAIL;
181 				break;
182 			}
183 
184 			fl6_free_socklist(sk);
185 			ipv6_sock_mc_close(sk);
186 
187 			/*
188 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
189 			 * remove it from the refcnt debug socks count in the
190 			 * original family...
191 			 */
192 			sk_refcnt_debug_dec(sk);
193 
194 			if (sk->sk_protocol == IPPROTO_TCP) {
195 				struct inet_connection_sock *icsk = inet_csk(sk);
196 				local_bh_disable();
197 				sock_prot_inuse_add(net, sk->sk_prot, -1);
198 				sock_prot_inuse_add(net, &tcp_prot, 1);
199 				local_bh_enable();
200 				sk->sk_prot = &tcp_prot;
201 				icsk->icsk_af_ops = &ipv4_specific;
202 				sk->sk_socket->ops = &inet_stream_ops;
203 				sk->sk_family = PF_INET;
204 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
205 			} else {
206 				struct proto *prot = &udp_prot;
207 
208 				if (sk->sk_protocol == IPPROTO_UDPLITE)
209 					prot = &udplite_prot;
210 				local_bh_disable();
211 				sock_prot_inuse_add(net, sk->sk_prot, -1);
212 				sock_prot_inuse_add(net, prot, 1);
213 				local_bh_enable();
214 				sk->sk_prot = prot;
215 				sk->sk_socket->ops = &inet_dgram_ops;
216 				sk->sk_family = PF_INET;
217 			}
218 			opt = xchg(&np->opt, NULL);
219 			if (opt)
220 				sock_kfree_s(sk, opt, opt->tot_len);
221 			pktopt = xchg(&np->pktoptions, NULL);
222 			kfree_skb(pktopt);
223 
224 			sk->sk_destruct = inet_sock_destruct;
225 			/*
226 			 * ... and add it to the refcnt debug socks count
227 			 * in the new family. -acme
228 			 */
229 			sk_refcnt_debug_inc(sk);
230 			module_put(THIS_MODULE);
231 			retv = 0;
232 			break;
233 		}
234 		goto e_inval;
235 
236 	case IPV6_V6ONLY:
237 		if (optlen < sizeof(int) ||
238 		    inet_sk(sk)->inet_num)
239 			goto e_inval;
240 		np->ipv6only = valbool;
241 		retv = 0;
242 		break;
243 
244 	case IPV6_RECVPKTINFO:
245 		if (optlen < sizeof(int))
246 			goto e_inval;
247 		np->rxopt.bits.rxinfo = valbool;
248 		retv = 0;
249 		break;
250 
251 	case IPV6_2292PKTINFO:
252 		if (optlen < sizeof(int))
253 			goto e_inval;
254 		np->rxopt.bits.rxoinfo = valbool;
255 		retv = 0;
256 		break;
257 
258 	case IPV6_RECVHOPLIMIT:
259 		if (optlen < sizeof(int))
260 			goto e_inval;
261 		np->rxopt.bits.rxhlim = valbool;
262 		retv = 0;
263 		break;
264 
265 	case IPV6_2292HOPLIMIT:
266 		if (optlen < sizeof(int))
267 			goto e_inval;
268 		np->rxopt.bits.rxohlim = valbool;
269 		retv = 0;
270 		break;
271 
272 	case IPV6_RECVRTHDR:
273 		if (optlen < sizeof(int))
274 			goto e_inval;
275 		np->rxopt.bits.srcrt = valbool;
276 		retv = 0;
277 		break;
278 
279 	case IPV6_2292RTHDR:
280 		if (optlen < sizeof(int))
281 			goto e_inval;
282 		np->rxopt.bits.osrcrt = valbool;
283 		retv = 0;
284 		break;
285 
286 	case IPV6_RECVHOPOPTS:
287 		if (optlen < sizeof(int))
288 			goto e_inval;
289 		np->rxopt.bits.hopopts = valbool;
290 		retv = 0;
291 		break;
292 
293 	case IPV6_2292HOPOPTS:
294 		if (optlen < sizeof(int))
295 			goto e_inval;
296 		np->rxopt.bits.ohopopts = valbool;
297 		retv = 0;
298 		break;
299 
300 	case IPV6_RECVDSTOPTS:
301 		if (optlen < sizeof(int))
302 			goto e_inval;
303 		np->rxopt.bits.dstopts = valbool;
304 		retv = 0;
305 		break;
306 
307 	case IPV6_2292DSTOPTS:
308 		if (optlen < sizeof(int))
309 			goto e_inval;
310 		np->rxopt.bits.odstopts = valbool;
311 		retv = 0;
312 		break;
313 
314 	case IPV6_TCLASS:
315 		if (optlen < sizeof(int))
316 			goto e_inval;
317 		if (val < -1 || val > 0xff)
318 			goto e_inval;
319 		/* RFC 3542, 6.5: default traffic class of 0x0 */
320 		if (val == -1)
321 			val = 0;
322 		np->tclass = val;
323 		retv = 0;
324 		break;
325 
326 	case IPV6_RECVTCLASS:
327 		if (optlen < sizeof(int))
328 			goto e_inval;
329 		np->rxopt.bits.rxtclass = valbool;
330 		retv = 0;
331 		break;
332 
333 	case IPV6_FLOWINFO:
334 		if (optlen < sizeof(int))
335 			goto e_inval;
336 		np->rxopt.bits.rxflow = valbool;
337 		retv = 0;
338 		break;
339 
340 	case IPV6_RECVPATHMTU:
341 		if (optlen < sizeof(int))
342 			goto e_inval;
343 		np->rxopt.bits.rxpmtu = valbool;
344 		retv = 0;
345 		break;
346 
347 	case IPV6_HOPOPTS:
348 	case IPV6_RTHDRDSTOPTS:
349 	case IPV6_RTHDR:
350 	case IPV6_DSTOPTS:
351 	{
352 		struct ipv6_txoptions *opt;
353 
354 		/* remove any sticky options header with a zero option
355 		 * length, per RFC3542.
356 		 */
357 		if (optlen == 0)
358 			optval = NULL;
359 		else if (optval == NULL)
360 			goto e_inval;
361 		else if (optlen < sizeof(struct ipv6_opt_hdr) ||
362 			 optlen & 0x7 || optlen > 8 * 255)
363 			goto e_inval;
364 
365 		/* hop-by-hop / destination options are privileged option */
366 		retv = -EPERM;
367 		if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
368 			break;
369 
370 		opt = ipv6_renew_options(sk, np->opt, optname,
371 					 (struct ipv6_opt_hdr __user *)optval,
372 					 optlen);
373 		if (IS_ERR(opt)) {
374 			retv = PTR_ERR(opt);
375 			break;
376 		}
377 
378 		/* routing header option needs extra check */
379 		retv = -EINVAL;
380 		if (optname == IPV6_RTHDR && opt && opt->srcrt) {
381 			struct ipv6_rt_hdr *rthdr = opt->srcrt;
382 			switch (rthdr->type) {
383 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
384 			case IPV6_SRCRT_TYPE_2:
385 				if (rthdr->hdrlen != 2 ||
386 				    rthdr->segments_left != 1)
387 					goto sticky_done;
388 
389 				break;
390 #endif
391 			default:
392 				goto sticky_done;
393 			}
394 		}
395 
396 		retv = 0;
397 		opt = ipv6_update_options(sk, opt);
398 sticky_done:
399 		if (opt)
400 			sock_kfree_s(sk, opt, opt->tot_len);
401 		break;
402 	}
403 
404 	case IPV6_PKTINFO:
405 	{
406 		struct in6_pktinfo pkt;
407 
408 		if (optlen == 0)
409 			goto e_inval;
410 		else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
411 			goto e_inval;
412 
413 		if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
414 				retv = -EFAULT;
415 				break;
416 		}
417 		if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
418 			goto e_inval;
419 
420 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
421 		ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
422 		retv = 0;
423 		break;
424 	}
425 
426 	case IPV6_2292PKTOPTIONS:
427 	{
428 		struct ipv6_txoptions *opt = NULL;
429 		struct msghdr msg;
430 		struct flowi fl;
431 		int junk;
432 
433 		fl.fl6_flowlabel = 0;
434 		fl.oif = sk->sk_bound_dev_if;
435 		fl.mark = sk->sk_mark;
436 
437 		if (optlen == 0)
438 			goto update;
439 
440 		/* 1K is probably excessive
441 		 * 1K is surely not enough, 2K per standard header is 16K.
442 		 */
443 		retv = -EINVAL;
444 		if (optlen > 64*1024)
445 			break;
446 
447 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
448 		retv = -ENOBUFS;
449 		if (opt == NULL)
450 			break;
451 
452 		memset(opt, 0, sizeof(*opt));
453 		opt->tot_len = sizeof(*opt) + optlen;
454 		retv = -EFAULT;
455 		if (copy_from_user(opt+1, optval, optlen))
456 			goto done;
457 
458 		msg.msg_controllen = optlen;
459 		msg.msg_control = (void*)(opt+1);
460 
461 		retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk,
462 					 &junk);
463 		if (retv)
464 			goto done;
465 update:
466 		retv = 0;
467 		opt = ipv6_update_options(sk, opt);
468 done:
469 		if (opt)
470 			sock_kfree_s(sk, opt, opt->tot_len);
471 		break;
472 	}
473 	case IPV6_UNICAST_HOPS:
474 		if (optlen < sizeof(int))
475 			goto e_inval;
476 		if (val > 255 || val < -1)
477 			goto e_inval;
478 		np->hop_limit = val;
479 		retv = 0;
480 		break;
481 
482 	case IPV6_MULTICAST_HOPS:
483 		if (sk->sk_type == SOCK_STREAM)
484 			break;
485 		if (optlen < sizeof(int))
486 			goto e_inval;
487 		if (val > 255 || val < -1)
488 			goto e_inval;
489 		np->mcast_hops = val;
490 		retv = 0;
491 		break;
492 
493 	case IPV6_MULTICAST_LOOP:
494 		if (optlen < sizeof(int))
495 			goto e_inval;
496 		if (val != valbool)
497 			goto e_inval;
498 		np->mc_loop = valbool;
499 		retv = 0;
500 		break;
501 
502 	case IPV6_MULTICAST_IF:
503 		if (sk->sk_type == SOCK_STREAM)
504 			break;
505 		if (optlen < sizeof(int))
506 			goto e_inval;
507 
508 		if (val) {
509 			struct net_device *dev;
510 
511 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
512 				goto e_inval;
513 
514 			dev = dev_get_by_index(net, val);
515 			if (!dev) {
516 				retv = -ENODEV;
517 				break;
518 			}
519 			dev_put(dev);
520 		}
521 		np->mcast_oif = val;
522 		retv = 0;
523 		break;
524 	case IPV6_ADD_MEMBERSHIP:
525 	case IPV6_DROP_MEMBERSHIP:
526 	{
527 		struct ipv6_mreq mreq;
528 
529 		if (optlen < sizeof(struct ipv6_mreq))
530 			goto e_inval;
531 
532 		retv = -EPROTO;
533 		if (inet_sk(sk)->is_icsk)
534 			break;
535 
536 		retv = -EFAULT;
537 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
538 			break;
539 
540 		if (optname == IPV6_ADD_MEMBERSHIP)
541 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
542 		else
543 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
544 		break;
545 	}
546 	case IPV6_JOIN_ANYCAST:
547 	case IPV6_LEAVE_ANYCAST:
548 	{
549 		struct ipv6_mreq mreq;
550 
551 		if (optlen < sizeof(struct ipv6_mreq))
552 			goto e_inval;
553 
554 		retv = -EFAULT;
555 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
556 			break;
557 
558 		if (optname == IPV6_JOIN_ANYCAST)
559 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
560 		else
561 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
562 		break;
563 	}
564 	case MCAST_JOIN_GROUP:
565 	case MCAST_LEAVE_GROUP:
566 	{
567 		struct group_req greq;
568 		struct sockaddr_in6 *psin6;
569 
570 		if (optlen < sizeof(struct group_req))
571 			goto e_inval;
572 
573 		retv = -EFAULT;
574 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
575 			break;
576 		if (greq.gr_group.ss_family != AF_INET6) {
577 			retv = -EADDRNOTAVAIL;
578 			break;
579 		}
580 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
581 		if (optname == MCAST_JOIN_GROUP)
582 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
583 				&psin6->sin6_addr);
584 		else
585 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
586 				&psin6->sin6_addr);
587 		break;
588 	}
589 	case MCAST_JOIN_SOURCE_GROUP:
590 	case MCAST_LEAVE_SOURCE_GROUP:
591 	case MCAST_BLOCK_SOURCE:
592 	case MCAST_UNBLOCK_SOURCE:
593 	{
594 		struct group_source_req greqs;
595 		int omode, add;
596 
597 		if (optlen < sizeof(struct group_source_req))
598 			goto e_inval;
599 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
600 			retv = -EFAULT;
601 			break;
602 		}
603 		if (greqs.gsr_group.ss_family != AF_INET6 ||
604 		    greqs.gsr_source.ss_family != AF_INET6) {
605 			retv = -EADDRNOTAVAIL;
606 			break;
607 		}
608 		if (optname == MCAST_BLOCK_SOURCE) {
609 			omode = MCAST_EXCLUDE;
610 			add = 1;
611 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
612 			omode = MCAST_EXCLUDE;
613 			add = 0;
614 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
615 			struct sockaddr_in6 *psin6;
616 
617 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
618 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
619 				&psin6->sin6_addr);
620 			/* prior join w/ different source is ok */
621 			if (retv && retv != -EADDRINUSE)
622 				break;
623 			omode = MCAST_INCLUDE;
624 			add = 1;
625 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
626 			omode = MCAST_INCLUDE;
627 			add = 0;
628 		}
629 		retv = ip6_mc_source(add, omode, sk, &greqs);
630 		break;
631 	}
632 	case MCAST_MSFILTER:
633 	{
634 		extern int sysctl_mld_max_msf;
635 		struct group_filter *gsf;
636 
637 		if (optlen < GROUP_FILTER_SIZE(0))
638 			goto e_inval;
639 		if (optlen > sysctl_optmem_max) {
640 			retv = -ENOBUFS;
641 			break;
642 		}
643 		gsf = kmalloc(optlen,GFP_KERNEL);
644 		if (!gsf) {
645 			retv = -ENOBUFS;
646 			break;
647 		}
648 		retv = -EFAULT;
649 		if (copy_from_user(gsf, optval, optlen)) {
650 			kfree(gsf);
651 			break;
652 		}
653 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
654 		if (gsf->gf_numsrc >= 0x1ffffffU ||
655 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
656 			kfree(gsf);
657 			retv = -ENOBUFS;
658 			break;
659 		}
660 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
661 			kfree(gsf);
662 			retv = -EINVAL;
663 			break;
664 		}
665 		retv = ip6_mc_msfilter(sk, gsf);
666 		kfree(gsf);
667 
668 		break;
669 	}
670 	case IPV6_ROUTER_ALERT:
671 		if (optlen < sizeof(int))
672 			goto e_inval;
673 		retv = ip6_ra_control(sk, val);
674 		break;
675 	case IPV6_MTU_DISCOVER:
676 		if (optlen < sizeof(int))
677 			goto e_inval;
678 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
679 			goto e_inval;
680 		np->pmtudisc = val;
681 		retv = 0;
682 		break;
683 	case IPV6_MTU:
684 		if (optlen < sizeof(int))
685 			goto e_inval;
686 		if (val && val < IPV6_MIN_MTU)
687 			goto e_inval;
688 		np->frag_size = val;
689 		retv = 0;
690 		break;
691 	case IPV6_RECVERR:
692 		if (optlen < sizeof(int))
693 			goto e_inval;
694 		np->recverr = valbool;
695 		if (!val)
696 			skb_queue_purge(&sk->sk_error_queue);
697 		retv = 0;
698 		break;
699 	case IPV6_FLOWINFO_SEND:
700 		if (optlen < sizeof(int))
701 			goto e_inval;
702 		np->sndflow = valbool;
703 		retv = 0;
704 		break;
705 	case IPV6_FLOWLABEL_MGR:
706 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
707 		break;
708 	case IPV6_IPSEC_POLICY:
709 	case IPV6_XFRM_POLICY:
710 		retv = -EPERM;
711 		if (!capable(CAP_NET_ADMIN))
712 			break;
713 		retv = xfrm_user_policy(sk, optname, optval, optlen);
714 		break;
715 
716 	case IPV6_ADDR_PREFERENCES:
717 	    {
718 		unsigned int pref = 0;
719 		unsigned int prefmask = ~0;
720 
721 		if (optlen < sizeof(int))
722 			goto e_inval;
723 
724 		retv = -EINVAL;
725 
726 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
727 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
728 			       IPV6_PREFER_SRC_TMP|
729 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
730 		case IPV6_PREFER_SRC_PUBLIC:
731 			pref |= IPV6_PREFER_SRC_PUBLIC;
732 			break;
733 		case IPV6_PREFER_SRC_TMP:
734 			pref |= IPV6_PREFER_SRC_TMP;
735 			break;
736 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
737 			break;
738 		case 0:
739 			goto pref_skip_pubtmp;
740 		default:
741 			goto e_inval;
742 		}
743 
744 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
745 			      IPV6_PREFER_SRC_TMP);
746 pref_skip_pubtmp:
747 
748 		/* check HOME/COA conflicts */
749 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
750 		case IPV6_PREFER_SRC_HOME:
751 			break;
752 		case IPV6_PREFER_SRC_COA:
753 			pref |= IPV6_PREFER_SRC_COA;
754 		case 0:
755 			goto pref_skip_coa;
756 		default:
757 			goto e_inval;
758 		}
759 
760 		prefmask &= ~IPV6_PREFER_SRC_COA;
761 pref_skip_coa:
762 
763 		/* check CGA/NONCGA conflicts */
764 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
765 		case IPV6_PREFER_SRC_CGA:
766 		case IPV6_PREFER_SRC_NONCGA:
767 		case 0:
768 			break;
769 		default:
770 			goto e_inval;
771 		}
772 
773 		np->srcprefs = (np->srcprefs & prefmask) | pref;
774 		retv = 0;
775 
776 		break;
777 	    }
778 	case IPV6_MINHOPCOUNT:
779 		if (optlen < sizeof(int))
780 			goto e_inval;
781 		if (val < 0 || val > 255)
782 			goto e_inval;
783 		np->min_hopcount = val;
784 		break;
785 	case IPV6_DONTFRAG:
786 		np->dontfrag = valbool;
787 		retv = 0;
788 		break;
789 	}
790 
791 	release_sock(sk);
792 
793 	return retv;
794 
795 e_inval:
796 	release_sock(sk);
797 	return -EINVAL;
798 }
799 
800 int ipv6_setsockopt(struct sock *sk, int level, int optname,
801 		    char __user *optval, unsigned int optlen)
802 {
803 	int err;
804 
805 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
806 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
807 
808 	if (level != SOL_IPV6)
809 		return -ENOPROTOOPT;
810 
811 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
812 #ifdef CONFIG_NETFILTER
813 	/* we need to exclude all possible ENOPROTOOPTs except default case */
814 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
815 			optname != IPV6_XFRM_POLICY) {
816 		lock_sock(sk);
817 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
818 				optlen);
819 		release_sock(sk);
820 	}
821 #endif
822 	return err;
823 }
824 
825 EXPORT_SYMBOL(ipv6_setsockopt);
826 
827 #ifdef CONFIG_COMPAT
828 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
829 			   char __user *optval, unsigned int optlen)
830 {
831 	int err;
832 
833 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
834 		if (udp_prot.compat_setsockopt != NULL)
835 			return udp_prot.compat_setsockopt(sk, level, optname,
836 							  optval, optlen);
837 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
838 	}
839 
840 	if (level != SOL_IPV6)
841 		return -ENOPROTOOPT;
842 
843 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
844 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
845 			ipv6_setsockopt);
846 
847 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
848 #ifdef CONFIG_NETFILTER
849 	/* we need to exclude all possible ENOPROTOOPTs except default case */
850 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
851 	    optname != IPV6_XFRM_POLICY) {
852 		lock_sock(sk);
853 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
854 					   optval, optlen);
855 		release_sock(sk);
856 	}
857 #endif
858 	return err;
859 }
860 
861 EXPORT_SYMBOL(compat_ipv6_setsockopt);
862 #endif
863 
864 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
865 				  int optname, char __user *optval, int len)
866 {
867 	struct ipv6_opt_hdr *hdr;
868 
869 	if (!opt)
870 		return 0;
871 
872 	switch(optname) {
873 	case IPV6_HOPOPTS:
874 		hdr = opt->hopopt;
875 		break;
876 	case IPV6_RTHDRDSTOPTS:
877 		hdr = opt->dst0opt;
878 		break;
879 	case IPV6_RTHDR:
880 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
881 		break;
882 	case IPV6_DSTOPTS:
883 		hdr = opt->dst1opt;
884 		break;
885 	default:
886 		return -EINVAL;	/* should not happen */
887 	}
888 
889 	if (!hdr)
890 		return 0;
891 
892 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
893 	if (copy_to_user(optval, hdr, len))
894 		return -EFAULT;
895 	return len;
896 }
897 
898 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
899 		    char __user *optval, int __user *optlen)
900 {
901 	struct ipv6_pinfo *np = inet6_sk(sk);
902 	int len;
903 	int val;
904 
905 	if (ip6_mroute_opt(optname))
906 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
907 
908 	if (get_user(len, optlen))
909 		return -EFAULT;
910 	switch (optname) {
911 	case IPV6_ADDRFORM:
912 		if (sk->sk_protocol != IPPROTO_UDP &&
913 		    sk->sk_protocol != IPPROTO_UDPLITE &&
914 		    sk->sk_protocol != IPPROTO_TCP)
915 			return -ENOPROTOOPT;
916 		if (sk->sk_state != TCP_ESTABLISHED)
917 			return -ENOTCONN;
918 		val = sk->sk_family;
919 		break;
920 	case MCAST_MSFILTER:
921 	{
922 		struct group_filter gsf;
923 		int err;
924 
925 		if (len < GROUP_FILTER_SIZE(0))
926 			return -EINVAL;
927 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
928 			return -EFAULT;
929 		if (gsf.gf_group.ss_family != AF_INET6)
930 			return -EADDRNOTAVAIL;
931 		lock_sock(sk);
932 		err = ip6_mc_msfget(sk, &gsf,
933 			(struct group_filter __user *)optval, optlen);
934 		release_sock(sk);
935 		return err;
936 	}
937 
938 	case IPV6_2292PKTOPTIONS:
939 	{
940 		struct msghdr msg;
941 		struct sk_buff *skb;
942 
943 		if (sk->sk_type != SOCK_STREAM)
944 			return -ENOPROTOOPT;
945 
946 		msg.msg_control = optval;
947 		msg.msg_controllen = len;
948 		msg.msg_flags = 0;
949 
950 		lock_sock(sk);
951 		skb = np->pktoptions;
952 		if (skb)
953 			atomic_inc(&skb->users);
954 		release_sock(sk);
955 
956 		if (skb) {
957 			int err = datagram_recv_ctl(sk, &msg, skb);
958 			kfree_skb(skb);
959 			if (err)
960 				return err;
961 		} else {
962 			if (np->rxopt.bits.rxinfo) {
963 				struct in6_pktinfo src_info;
964 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
965 					np->sticky_pktinfo.ipi6_ifindex;
966 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
967 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
968 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
969 			}
970 			if (np->rxopt.bits.rxhlim) {
971 				int hlim = np->mcast_hops;
972 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
973 			}
974 			if (np->rxopt.bits.rxoinfo) {
975 				struct in6_pktinfo src_info;
976 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
977 					np->sticky_pktinfo.ipi6_ifindex;
978 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
979 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
980 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
981 			}
982 			if (np->rxopt.bits.rxohlim) {
983 				int hlim = np->mcast_hops;
984 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
985 			}
986 		}
987 		len -= msg.msg_controllen;
988 		return put_user(len, optlen);
989 	}
990 	case IPV6_MTU:
991 	{
992 		struct dst_entry *dst;
993 
994 		val = 0;
995 		rcu_read_lock();
996 		dst = __sk_dst_get(sk);
997 		if (dst)
998 			val = dst_mtu(dst);
999 		rcu_read_unlock();
1000 		if (!val)
1001 			return -ENOTCONN;
1002 		break;
1003 	}
1004 
1005 	case IPV6_V6ONLY:
1006 		val = np->ipv6only;
1007 		break;
1008 
1009 	case IPV6_RECVPKTINFO:
1010 		val = np->rxopt.bits.rxinfo;
1011 		break;
1012 
1013 	case IPV6_2292PKTINFO:
1014 		val = np->rxopt.bits.rxoinfo;
1015 		break;
1016 
1017 	case IPV6_RECVHOPLIMIT:
1018 		val = np->rxopt.bits.rxhlim;
1019 		break;
1020 
1021 	case IPV6_2292HOPLIMIT:
1022 		val = np->rxopt.bits.rxohlim;
1023 		break;
1024 
1025 	case IPV6_RECVRTHDR:
1026 		val = np->rxopt.bits.srcrt;
1027 		break;
1028 
1029 	case IPV6_2292RTHDR:
1030 		val = np->rxopt.bits.osrcrt;
1031 		break;
1032 
1033 	case IPV6_HOPOPTS:
1034 	case IPV6_RTHDRDSTOPTS:
1035 	case IPV6_RTHDR:
1036 	case IPV6_DSTOPTS:
1037 	{
1038 
1039 		lock_sock(sk);
1040 		len = ipv6_getsockopt_sticky(sk, np->opt,
1041 					     optname, optval, len);
1042 		release_sock(sk);
1043 		/* check if ipv6_getsockopt_sticky() returns err code */
1044 		if (len < 0)
1045 			return len;
1046 		return put_user(len, optlen);
1047 	}
1048 
1049 	case IPV6_RECVHOPOPTS:
1050 		val = np->rxopt.bits.hopopts;
1051 		break;
1052 
1053 	case IPV6_2292HOPOPTS:
1054 		val = np->rxopt.bits.ohopopts;
1055 		break;
1056 
1057 	case IPV6_RECVDSTOPTS:
1058 		val = np->rxopt.bits.dstopts;
1059 		break;
1060 
1061 	case IPV6_2292DSTOPTS:
1062 		val = np->rxopt.bits.odstopts;
1063 		break;
1064 
1065 	case IPV6_TCLASS:
1066 		val = np->tclass;
1067 		break;
1068 
1069 	case IPV6_RECVTCLASS:
1070 		val = np->rxopt.bits.rxtclass;
1071 		break;
1072 
1073 	case IPV6_FLOWINFO:
1074 		val = np->rxopt.bits.rxflow;
1075 		break;
1076 
1077 	case IPV6_RECVPATHMTU:
1078 		val = np->rxopt.bits.rxpmtu;
1079 		break;
1080 
1081 	case IPV6_PATHMTU:
1082 	{
1083 		struct dst_entry *dst;
1084 		struct ip6_mtuinfo mtuinfo;
1085 
1086 		if (len < sizeof(mtuinfo))
1087 			return -EINVAL;
1088 
1089 		len = sizeof(mtuinfo);
1090 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1091 
1092 		rcu_read_lock();
1093 		dst = __sk_dst_get(sk);
1094 		if (dst)
1095 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1096 		rcu_read_unlock();
1097 		if (!mtuinfo.ip6m_mtu)
1098 			return -ENOTCONN;
1099 
1100 		if (put_user(len, optlen))
1101 			return -EFAULT;
1102 		if (copy_to_user(optval, &mtuinfo, len))
1103 			return -EFAULT;
1104 
1105 		return 0;
1106 		break;
1107 	}
1108 
1109 	case IPV6_UNICAST_HOPS:
1110 	case IPV6_MULTICAST_HOPS:
1111 	{
1112 		struct dst_entry *dst;
1113 
1114 		if (optname == IPV6_UNICAST_HOPS)
1115 			val = np->hop_limit;
1116 		else
1117 			val = np->mcast_hops;
1118 
1119 		if (val < 0) {
1120 			rcu_read_lock();
1121 			dst = __sk_dst_get(sk);
1122 			if (dst)
1123 				val = ip6_dst_hoplimit(dst);
1124 			rcu_read_unlock();
1125 		}
1126 
1127 		if (val < 0)
1128 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1129 		break;
1130 	}
1131 
1132 	case IPV6_MULTICAST_LOOP:
1133 		val = np->mc_loop;
1134 		break;
1135 
1136 	case IPV6_MULTICAST_IF:
1137 		val = np->mcast_oif;
1138 		break;
1139 
1140 	case IPV6_MTU_DISCOVER:
1141 		val = np->pmtudisc;
1142 		break;
1143 
1144 	case IPV6_RECVERR:
1145 		val = np->recverr;
1146 		break;
1147 
1148 	case IPV6_FLOWINFO_SEND:
1149 		val = np->sndflow;
1150 		break;
1151 
1152 	case IPV6_ADDR_PREFERENCES:
1153 		val = 0;
1154 
1155 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1156 			val |= IPV6_PREFER_SRC_TMP;
1157 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1158 			val |= IPV6_PREFER_SRC_PUBLIC;
1159 		else {
1160 			/* XXX: should we return system default? */
1161 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1162 		}
1163 
1164 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1165 			val |= IPV6_PREFER_SRC_COA;
1166 		else
1167 			val |= IPV6_PREFER_SRC_HOME;
1168 		break;
1169 
1170 	case IPV6_MINHOPCOUNT:
1171 		val = np->min_hopcount;
1172 		break;
1173 
1174 	case IPV6_DONTFRAG:
1175 		val = np->dontfrag;
1176 		break;
1177 
1178 	default:
1179 		return -ENOPROTOOPT;
1180 	}
1181 	len = min_t(unsigned int, sizeof(int), len);
1182 	if(put_user(len, optlen))
1183 		return -EFAULT;
1184 	if(copy_to_user(optval,&val,len))
1185 		return -EFAULT;
1186 	return 0;
1187 }
1188 
1189 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1190 		    char __user *optval, int __user *optlen)
1191 {
1192 	int err;
1193 
1194 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1195 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1196 
1197 	if(level != SOL_IPV6)
1198 		return -ENOPROTOOPT;
1199 
1200 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1201 #ifdef CONFIG_NETFILTER
1202 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1203 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1204 		int len;
1205 
1206 		if (get_user(len, optlen))
1207 			return -EFAULT;
1208 
1209 		lock_sock(sk);
1210 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1211 				&len);
1212 		release_sock(sk);
1213 		if (err >= 0)
1214 			err = put_user(len, optlen);
1215 	}
1216 #endif
1217 	return err;
1218 }
1219 
1220 EXPORT_SYMBOL(ipv6_getsockopt);
1221 
1222 #ifdef CONFIG_COMPAT
1223 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1224 			   char __user *optval, int __user *optlen)
1225 {
1226 	int err;
1227 
1228 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1229 		if (udp_prot.compat_getsockopt != NULL)
1230 			return udp_prot.compat_getsockopt(sk, level, optname,
1231 							  optval, optlen);
1232 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1233 	}
1234 
1235 	if (level != SOL_IPV6)
1236 		return -ENOPROTOOPT;
1237 
1238 	if (optname == MCAST_MSFILTER)
1239 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1240 			ipv6_getsockopt);
1241 
1242 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1243 #ifdef CONFIG_NETFILTER
1244 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1245 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1246 		int len;
1247 
1248 		if (get_user(len, optlen))
1249 			return -EFAULT;
1250 
1251 		lock_sock(sk);
1252 		err = compat_nf_getsockopt(sk, PF_INET6,
1253 					   optname, optval, &len);
1254 		release_sock(sk);
1255 		if (err >= 0)
1256 			err = put_user(len, optlen);
1257 	}
1258 #endif
1259 	return err;
1260 }
1261 
1262 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1263 #endif
1264 
1265