xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision 7dd65feb)
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)->inet_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)->inet_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)->inet_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 		fl.mark = sk->sk_mark;
428 
429 		if (optlen == 0)
430 			goto update;
431 
432 		/* 1K is probably excessive
433 		 * 1K is surely not enough, 2K per standard header is 16K.
434 		 */
435 		retv = -EINVAL;
436 		if (optlen > 64*1024)
437 			break;
438 
439 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
440 		retv = -ENOBUFS;
441 		if (opt == NULL)
442 			break;
443 
444 		memset(opt, 0, sizeof(*opt));
445 		opt->tot_len = sizeof(*opt) + optlen;
446 		retv = -EFAULT;
447 		if (copy_from_user(opt+1, optval, optlen))
448 			goto done;
449 
450 		msg.msg_controllen = optlen;
451 		msg.msg_control = (void*)(opt+1);
452 
453 		retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
454 		if (retv)
455 			goto done;
456 update:
457 		retv = 0;
458 		opt = ipv6_update_options(sk, opt);
459 done:
460 		if (opt)
461 			sock_kfree_s(sk, opt, opt->tot_len);
462 		break;
463 	}
464 	case IPV6_UNICAST_HOPS:
465 		if (optlen < sizeof(int))
466 			goto e_inval;
467 		if (val > 255 || val < -1)
468 			goto e_inval;
469 		np->hop_limit = val;
470 		retv = 0;
471 		break;
472 
473 	case IPV6_MULTICAST_HOPS:
474 		if (sk->sk_type == SOCK_STREAM)
475 			break;
476 		if (optlen < sizeof(int))
477 			goto e_inval;
478 		if (val > 255 || val < -1)
479 			goto e_inval;
480 		np->mcast_hops = val;
481 		retv = 0;
482 		break;
483 
484 	case IPV6_MULTICAST_LOOP:
485 		if (optlen < sizeof(int))
486 			goto e_inval;
487 		if (val != valbool)
488 			goto e_inval;
489 		np->mc_loop = valbool;
490 		retv = 0;
491 		break;
492 
493 	case IPV6_MULTICAST_IF:
494 		if (sk->sk_type == SOCK_STREAM)
495 			break;
496 		if (optlen < sizeof(int))
497 			goto e_inval;
498 
499 		if (val) {
500 			struct net_device *dev;
501 
502 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
503 				goto e_inval;
504 
505 			dev = dev_get_by_index(net, val);
506 			if (!dev) {
507 				retv = -ENODEV;
508 				break;
509 			}
510 			dev_put(dev);
511 		}
512 		np->mcast_oif = val;
513 		retv = 0;
514 		break;
515 	case IPV6_ADD_MEMBERSHIP:
516 	case IPV6_DROP_MEMBERSHIP:
517 	{
518 		struct ipv6_mreq mreq;
519 
520 		if (optlen < sizeof(struct ipv6_mreq))
521 			goto e_inval;
522 
523 		retv = -EPROTO;
524 		if (inet_sk(sk)->is_icsk)
525 			break;
526 
527 		retv = -EFAULT;
528 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
529 			break;
530 
531 		if (optname == IPV6_ADD_MEMBERSHIP)
532 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
533 		else
534 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
535 		break;
536 	}
537 	case IPV6_JOIN_ANYCAST:
538 	case IPV6_LEAVE_ANYCAST:
539 	{
540 		struct ipv6_mreq mreq;
541 
542 		if (optlen < sizeof(struct ipv6_mreq))
543 			goto e_inval;
544 
545 		retv = -EFAULT;
546 		if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
547 			break;
548 
549 		if (optname == IPV6_JOIN_ANYCAST)
550 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
551 		else
552 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
553 		break;
554 	}
555 	case MCAST_JOIN_GROUP:
556 	case MCAST_LEAVE_GROUP:
557 	{
558 		struct group_req greq;
559 		struct sockaddr_in6 *psin6;
560 
561 		if (optlen < sizeof(struct group_req))
562 			goto e_inval;
563 
564 		retv = -EFAULT;
565 		if (copy_from_user(&greq, optval, sizeof(struct group_req)))
566 			break;
567 		if (greq.gr_group.ss_family != AF_INET6) {
568 			retv = -EADDRNOTAVAIL;
569 			break;
570 		}
571 		psin6 = (struct sockaddr_in6 *)&greq.gr_group;
572 		if (optname == MCAST_JOIN_GROUP)
573 			retv = ipv6_sock_mc_join(sk, greq.gr_interface,
574 				&psin6->sin6_addr);
575 		else
576 			retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
577 				&psin6->sin6_addr);
578 		break;
579 	}
580 	case MCAST_JOIN_SOURCE_GROUP:
581 	case MCAST_LEAVE_SOURCE_GROUP:
582 	case MCAST_BLOCK_SOURCE:
583 	case MCAST_UNBLOCK_SOURCE:
584 	{
585 		struct group_source_req greqs;
586 		int omode, add;
587 
588 		if (optlen < sizeof(struct group_source_req))
589 			goto e_inval;
590 		if (copy_from_user(&greqs, optval, sizeof(greqs))) {
591 			retv = -EFAULT;
592 			break;
593 		}
594 		if (greqs.gsr_group.ss_family != AF_INET6 ||
595 		    greqs.gsr_source.ss_family != AF_INET6) {
596 			retv = -EADDRNOTAVAIL;
597 			break;
598 		}
599 		if (optname == MCAST_BLOCK_SOURCE) {
600 			omode = MCAST_EXCLUDE;
601 			add = 1;
602 		} else if (optname == MCAST_UNBLOCK_SOURCE) {
603 			omode = MCAST_EXCLUDE;
604 			add = 0;
605 		} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
606 			struct sockaddr_in6 *psin6;
607 
608 			psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
609 			retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
610 				&psin6->sin6_addr);
611 			/* prior join w/ different source is ok */
612 			if (retv && retv != -EADDRINUSE)
613 				break;
614 			omode = MCAST_INCLUDE;
615 			add = 1;
616 		} else /* MCAST_LEAVE_SOURCE_GROUP */ {
617 			omode = MCAST_INCLUDE;
618 			add = 0;
619 		}
620 		retv = ip6_mc_source(add, omode, sk, &greqs);
621 		break;
622 	}
623 	case MCAST_MSFILTER:
624 	{
625 		extern int sysctl_mld_max_msf;
626 		struct group_filter *gsf;
627 
628 		if (optlen < GROUP_FILTER_SIZE(0))
629 			goto e_inval;
630 		if (optlen > sysctl_optmem_max) {
631 			retv = -ENOBUFS;
632 			break;
633 		}
634 		gsf = kmalloc(optlen,GFP_KERNEL);
635 		if (!gsf) {
636 			retv = -ENOBUFS;
637 			break;
638 		}
639 		retv = -EFAULT;
640 		if (copy_from_user(gsf, optval, optlen)) {
641 			kfree(gsf);
642 			break;
643 		}
644 		/* numsrc >= (4G-140)/128 overflow in 32 bits */
645 		if (gsf->gf_numsrc >= 0x1ffffffU ||
646 		    gsf->gf_numsrc > sysctl_mld_max_msf) {
647 			kfree(gsf);
648 			retv = -ENOBUFS;
649 			break;
650 		}
651 		if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
652 			kfree(gsf);
653 			retv = -EINVAL;
654 			break;
655 		}
656 		retv = ip6_mc_msfilter(sk, gsf);
657 		kfree(gsf);
658 
659 		break;
660 	}
661 	case IPV6_ROUTER_ALERT:
662 		if (optlen < sizeof(int))
663 			goto e_inval;
664 		retv = ip6_ra_control(sk, val);
665 		break;
666 	case IPV6_MTU_DISCOVER:
667 		if (optlen < sizeof(int))
668 			goto e_inval;
669 		if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
670 			goto e_inval;
671 		np->pmtudisc = val;
672 		retv = 0;
673 		break;
674 	case IPV6_MTU:
675 		if (optlen < sizeof(int))
676 			goto e_inval;
677 		if (val && val < IPV6_MIN_MTU)
678 			goto e_inval;
679 		np->frag_size = val;
680 		retv = 0;
681 		break;
682 	case IPV6_RECVERR:
683 		if (optlen < sizeof(int))
684 			goto e_inval;
685 		np->recverr = valbool;
686 		if (!val)
687 			skb_queue_purge(&sk->sk_error_queue);
688 		retv = 0;
689 		break;
690 	case IPV6_FLOWINFO_SEND:
691 		if (optlen < sizeof(int))
692 			goto e_inval;
693 		np->sndflow = valbool;
694 		retv = 0;
695 		break;
696 	case IPV6_FLOWLABEL_MGR:
697 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
698 		break;
699 	case IPV6_IPSEC_POLICY:
700 	case IPV6_XFRM_POLICY:
701 		retv = -EPERM;
702 		if (!capable(CAP_NET_ADMIN))
703 			break;
704 		retv = xfrm_user_policy(sk, optname, optval, optlen);
705 		break;
706 
707 	case IPV6_ADDR_PREFERENCES:
708 	    {
709 		unsigned int pref = 0;
710 		unsigned int prefmask = ~0;
711 
712 		if (optlen < sizeof(int))
713 			goto e_inval;
714 
715 		retv = -EINVAL;
716 
717 		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
718 		switch (val & (IPV6_PREFER_SRC_PUBLIC|
719 			       IPV6_PREFER_SRC_TMP|
720 			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
721 		case IPV6_PREFER_SRC_PUBLIC:
722 			pref |= IPV6_PREFER_SRC_PUBLIC;
723 			break;
724 		case IPV6_PREFER_SRC_TMP:
725 			pref |= IPV6_PREFER_SRC_TMP;
726 			break;
727 		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
728 			break;
729 		case 0:
730 			goto pref_skip_pubtmp;
731 		default:
732 			goto e_inval;
733 		}
734 
735 		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
736 			      IPV6_PREFER_SRC_TMP);
737 pref_skip_pubtmp:
738 
739 		/* check HOME/COA conflicts */
740 		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
741 		case IPV6_PREFER_SRC_HOME:
742 			break;
743 		case IPV6_PREFER_SRC_COA:
744 			pref |= IPV6_PREFER_SRC_COA;
745 		case 0:
746 			goto pref_skip_coa;
747 		default:
748 			goto e_inval;
749 		}
750 
751 		prefmask &= ~IPV6_PREFER_SRC_COA;
752 pref_skip_coa:
753 
754 		/* check CGA/NONCGA conflicts */
755 		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
756 		case IPV6_PREFER_SRC_CGA:
757 		case IPV6_PREFER_SRC_NONCGA:
758 		case 0:
759 			break;
760 		default:
761 			goto e_inval;
762 		}
763 
764 		np->srcprefs = (np->srcprefs & prefmask) | pref;
765 		retv = 0;
766 
767 		break;
768 	    }
769 	}
770 
771 	release_sock(sk);
772 
773 	return retv;
774 
775 e_inval:
776 	release_sock(sk);
777 	return -EINVAL;
778 }
779 
780 int ipv6_setsockopt(struct sock *sk, int level, int optname,
781 		    char __user *optval, unsigned int optlen)
782 {
783 	int err;
784 
785 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
786 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
787 
788 	if (level != SOL_IPV6)
789 		return -ENOPROTOOPT;
790 
791 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
792 #ifdef CONFIG_NETFILTER
793 	/* we need to exclude all possible ENOPROTOOPTs except default case */
794 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
795 			optname != IPV6_XFRM_POLICY) {
796 		lock_sock(sk);
797 		err = nf_setsockopt(sk, PF_INET6, optname, optval,
798 				optlen);
799 		release_sock(sk);
800 	}
801 #endif
802 	return err;
803 }
804 
805 EXPORT_SYMBOL(ipv6_setsockopt);
806 
807 #ifdef CONFIG_COMPAT
808 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
809 			   char __user *optval, unsigned int optlen)
810 {
811 	int err;
812 
813 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
814 		if (udp_prot.compat_setsockopt != NULL)
815 			return udp_prot.compat_setsockopt(sk, level, optname,
816 							  optval, optlen);
817 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
818 	}
819 
820 	if (level != SOL_IPV6)
821 		return -ENOPROTOOPT;
822 
823 	if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
824 		return compat_mc_setsockopt(sk, level, optname, optval, optlen,
825 			ipv6_setsockopt);
826 
827 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
828 #ifdef CONFIG_NETFILTER
829 	/* we need to exclude all possible ENOPROTOOPTs except default case */
830 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
831 	    optname != IPV6_XFRM_POLICY) {
832 		lock_sock(sk);
833 		err = compat_nf_setsockopt(sk, PF_INET6, optname,
834 					   optval, optlen);
835 		release_sock(sk);
836 	}
837 #endif
838 	return err;
839 }
840 
841 EXPORT_SYMBOL(compat_ipv6_setsockopt);
842 #endif
843 
844 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
845 				  int optname, char __user *optval, int len)
846 {
847 	struct ipv6_opt_hdr *hdr;
848 
849 	if (!opt)
850 		return 0;
851 
852 	switch(optname) {
853 	case IPV6_HOPOPTS:
854 		hdr = opt->hopopt;
855 		break;
856 	case IPV6_RTHDRDSTOPTS:
857 		hdr = opt->dst0opt;
858 		break;
859 	case IPV6_RTHDR:
860 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
861 		break;
862 	case IPV6_DSTOPTS:
863 		hdr = opt->dst1opt;
864 		break;
865 	default:
866 		return -EINVAL;	/* should not happen */
867 	}
868 
869 	if (!hdr)
870 		return 0;
871 
872 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
873 	if (copy_to_user(optval, hdr, len))
874 		return -EFAULT;
875 	return len;
876 }
877 
878 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
879 		    char __user *optval, int __user *optlen)
880 {
881 	struct ipv6_pinfo *np = inet6_sk(sk);
882 	int len;
883 	int val;
884 
885 	if (ip6_mroute_opt(optname))
886 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
887 
888 	if (get_user(len, optlen))
889 		return -EFAULT;
890 	switch (optname) {
891 	case IPV6_ADDRFORM:
892 		if (sk->sk_protocol != IPPROTO_UDP &&
893 		    sk->sk_protocol != IPPROTO_UDPLITE &&
894 		    sk->sk_protocol != IPPROTO_TCP)
895 			return -ENOPROTOOPT;
896 		if (sk->sk_state != TCP_ESTABLISHED)
897 			return -ENOTCONN;
898 		val = sk->sk_family;
899 		break;
900 	case MCAST_MSFILTER:
901 	{
902 		struct group_filter gsf;
903 		int err;
904 
905 		if (len < GROUP_FILTER_SIZE(0))
906 			return -EINVAL;
907 		if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
908 			return -EFAULT;
909 		if (gsf.gf_group.ss_family != AF_INET6)
910 			return -EADDRNOTAVAIL;
911 		lock_sock(sk);
912 		err = ip6_mc_msfget(sk, &gsf,
913 			(struct group_filter __user *)optval, optlen);
914 		release_sock(sk);
915 		return err;
916 	}
917 
918 	case IPV6_2292PKTOPTIONS:
919 	{
920 		struct msghdr msg;
921 		struct sk_buff *skb;
922 
923 		if (sk->sk_type != SOCK_STREAM)
924 			return -ENOPROTOOPT;
925 
926 		msg.msg_control = optval;
927 		msg.msg_controllen = len;
928 		msg.msg_flags = 0;
929 
930 		lock_sock(sk);
931 		skb = np->pktoptions;
932 		if (skb)
933 			atomic_inc(&skb->users);
934 		release_sock(sk);
935 
936 		if (skb) {
937 			int err = datagram_recv_ctl(sk, &msg, skb);
938 			kfree_skb(skb);
939 			if (err)
940 				return err;
941 		} else {
942 			if (np->rxopt.bits.rxinfo) {
943 				struct in6_pktinfo src_info;
944 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
945 					np->sticky_pktinfo.ipi6_ifindex;
946 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
947 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
948 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
949 			}
950 			if (np->rxopt.bits.rxhlim) {
951 				int hlim = np->mcast_hops;
952 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
953 			}
954 			if (np->rxopt.bits.rxoinfo) {
955 				struct in6_pktinfo src_info;
956 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
957 					np->sticky_pktinfo.ipi6_ifindex;
958 				np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
959 					ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
960 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
961 			}
962 			if (np->rxopt.bits.rxohlim) {
963 				int hlim = np->mcast_hops;
964 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
965 			}
966 		}
967 		len -= msg.msg_controllen;
968 		return put_user(len, optlen);
969 	}
970 	case IPV6_MTU:
971 	{
972 		struct dst_entry *dst;
973 		val = 0;
974 		lock_sock(sk);
975 		dst = sk_dst_get(sk);
976 		if (dst) {
977 			val = dst_mtu(dst);
978 			dst_release(dst);
979 		}
980 		release_sock(sk);
981 		if (!val)
982 			return -ENOTCONN;
983 		break;
984 	}
985 
986 	case IPV6_V6ONLY:
987 		val = np->ipv6only;
988 		break;
989 
990 	case IPV6_RECVPKTINFO:
991 		val = np->rxopt.bits.rxinfo;
992 		break;
993 
994 	case IPV6_2292PKTINFO:
995 		val = np->rxopt.bits.rxoinfo;
996 		break;
997 
998 	case IPV6_RECVHOPLIMIT:
999 		val = np->rxopt.bits.rxhlim;
1000 		break;
1001 
1002 	case IPV6_2292HOPLIMIT:
1003 		val = np->rxopt.bits.rxohlim;
1004 		break;
1005 
1006 	case IPV6_RECVRTHDR:
1007 		val = np->rxopt.bits.srcrt;
1008 		break;
1009 
1010 	case IPV6_2292RTHDR:
1011 		val = np->rxopt.bits.osrcrt;
1012 		break;
1013 
1014 	case IPV6_HOPOPTS:
1015 	case IPV6_RTHDRDSTOPTS:
1016 	case IPV6_RTHDR:
1017 	case IPV6_DSTOPTS:
1018 	{
1019 
1020 		lock_sock(sk);
1021 		len = ipv6_getsockopt_sticky(sk, np->opt,
1022 					     optname, optval, len);
1023 		release_sock(sk);
1024 		/* check if ipv6_getsockopt_sticky() returns err code */
1025 		if (len < 0)
1026 			return len;
1027 		return put_user(len, optlen);
1028 	}
1029 
1030 	case IPV6_RECVHOPOPTS:
1031 		val = np->rxopt.bits.hopopts;
1032 		break;
1033 
1034 	case IPV6_2292HOPOPTS:
1035 		val = np->rxopt.bits.ohopopts;
1036 		break;
1037 
1038 	case IPV6_RECVDSTOPTS:
1039 		val = np->rxopt.bits.dstopts;
1040 		break;
1041 
1042 	case IPV6_2292DSTOPTS:
1043 		val = np->rxopt.bits.odstopts;
1044 		break;
1045 
1046 	case IPV6_TCLASS:
1047 		val = np->tclass;
1048 		break;
1049 
1050 	case IPV6_RECVTCLASS:
1051 		val = np->rxopt.bits.rxtclass;
1052 		break;
1053 
1054 	case IPV6_FLOWINFO:
1055 		val = np->rxopt.bits.rxflow;
1056 		break;
1057 
1058 	case IPV6_UNICAST_HOPS:
1059 	case IPV6_MULTICAST_HOPS:
1060 	{
1061 		struct dst_entry *dst;
1062 
1063 		if (optname == IPV6_UNICAST_HOPS)
1064 			val = np->hop_limit;
1065 		else
1066 			val = np->mcast_hops;
1067 
1068 		dst = sk_dst_get(sk);
1069 		if (dst) {
1070 			if (val < 0)
1071 				val = ip6_dst_hoplimit(dst);
1072 			dst_release(dst);
1073 		}
1074 		if (val < 0)
1075 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1076 		break;
1077 	}
1078 
1079 	case IPV6_MULTICAST_LOOP:
1080 		val = np->mc_loop;
1081 		break;
1082 
1083 	case IPV6_MULTICAST_IF:
1084 		val = np->mcast_oif;
1085 		break;
1086 
1087 	case IPV6_MTU_DISCOVER:
1088 		val = np->pmtudisc;
1089 		break;
1090 
1091 	case IPV6_RECVERR:
1092 		val = np->recverr;
1093 		break;
1094 
1095 	case IPV6_FLOWINFO_SEND:
1096 		val = np->sndflow;
1097 		break;
1098 
1099 	case IPV6_ADDR_PREFERENCES:
1100 		val = 0;
1101 
1102 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1103 			val |= IPV6_PREFER_SRC_TMP;
1104 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1105 			val |= IPV6_PREFER_SRC_PUBLIC;
1106 		else {
1107 			/* XXX: should we return system default? */
1108 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1109 		}
1110 
1111 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1112 			val |= IPV6_PREFER_SRC_COA;
1113 		else
1114 			val |= IPV6_PREFER_SRC_HOME;
1115 		break;
1116 
1117 	default:
1118 		return -ENOPROTOOPT;
1119 	}
1120 	len = min_t(unsigned int, sizeof(int), len);
1121 	if(put_user(len, optlen))
1122 		return -EFAULT;
1123 	if(copy_to_user(optval,&val,len))
1124 		return -EFAULT;
1125 	return 0;
1126 }
1127 
1128 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1129 		    char __user *optval, int __user *optlen)
1130 {
1131 	int err;
1132 
1133 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1134 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1135 
1136 	if(level != SOL_IPV6)
1137 		return -ENOPROTOOPT;
1138 
1139 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1140 #ifdef CONFIG_NETFILTER
1141 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1142 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1143 		int len;
1144 
1145 		if (get_user(len, optlen))
1146 			return -EFAULT;
1147 
1148 		lock_sock(sk);
1149 		err = nf_getsockopt(sk, PF_INET6, optname, optval,
1150 				&len);
1151 		release_sock(sk);
1152 		if (err >= 0)
1153 			err = put_user(len, optlen);
1154 	}
1155 #endif
1156 	return err;
1157 }
1158 
1159 EXPORT_SYMBOL(ipv6_getsockopt);
1160 
1161 #ifdef CONFIG_COMPAT
1162 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1163 			   char __user *optval, int __user *optlen)
1164 {
1165 	int err;
1166 
1167 	if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1168 		if (udp_prot.compat_getsockopt != NULL)
1169 			return udp_prot.compat_getsockopt(sk, level, optname,
1170 							  optval, optlen);
1171 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1172 	}
1173 
1174 	if (level != SOL_IPV6)
1175 		return -ENOPROTOOPT;
1176 
1177 	if (optname == MCAST_MSFILTER)
1178 		return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1179 			ipv6_getsockopt);
1180 
1181 	err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1182 #ifdef CONFIG_NETFILTER
1183 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1184 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1185 		int len;
1186 
1187 		if (get_user(len, optlen))
1188 			return -EFAULT;
1189 
1190 		lock_sock(sk);
1191 		err = compat_nf_getsockopt(sk, PF_INET6,
1192 					   optname, optval, &len);
1193 		release_sock(sk);
1194 		if (err >= 0)
1195 			err = put_user(len, optlen);
1196 	}
1197 #endif
1198 	return err;
1199 }
1200 
1201 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1202 #endif
1203 
1204