xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision bab55037)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 BSD socket options interface
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on linux/net/ipv4/ip_sockglue.c
10  *
11  *	FIXME: Make the setsockopt code POSIX compliant: That is
12  *
13  *	o	Truncate getsockopt returns
14  *	o	Return an optlen of the truncated length if need be
15  *
16  *	Changes:
17  *	David L Stevens <dlstevens@us.ibm.com>:
18  *		- added multicast source filtering API for MLDv2
19  */
20 
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/mroute6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/init.h>
33 #include <linux/sysctl.h>
34 #include <linux/netfilter.h>
35 #include <linux/slab.h>
36 
37 #include <net/sock.h>
38 #include <net/snmp.h>
39 #include <net/ipv6.h>
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_route.h>
44 #include <net/addrconf.h>
45 #include <net/inet_common.h>
46 #include <net/tcp.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/xfrm.h>
50 #include <net/compat.h>
51 #include <net/seg6.h>
52 
53 #include <linux/uaccess.h>
54 
55 struct ip6_ra_chain *ip6_ra_chain;
56 DEFINE_RWLOCK(ip6_ra_lock);
57 
58 DEFINE_STATIC_KEY_FALSE(ip6_min_hopcount);
59 
60 int ip6_ra_control(struct sock *sk, int sel)
61 {
62 	struct ip6_ra_chain *ra, *new_ra, **rap;
63 
64 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
65 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
66 		return -ENOPROTOOPT;
67 
68 	new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
69 	if (sel >= 0 && !new_ra)
70 		return -ENOMEM;
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) {
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 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 					   struct ipv6_txoptions *opt)
104 {
105 	if (inet_sk(sk)->is_icsk) {
106 		if (opt &&
107 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108 		    inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109 			struct inet_connection_sock *icsk = inet_csk(sk);
110 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
112 		}
113 	}
114 	opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
115 		   opt);
116 	sk_dst_reset(sk);
117 
118 	return opt;
119 }
120 
121 static bool setsockopt_needs_rtnl(int optname)
122 {
123 	switch (optname) {
124 	case IPV6_ADDRFORM:
125 	case IPV6_ADD_MEMBERSHIP:
126 	case IPV6_DROP_MEMBERSHIP:
127 	case IPV6_JOIN_ANYCAST:
128 	case IPV6_LEAVE_ANYCAST:
129 	case MCAST_JOIN_GROUP:
130 	case MCAST_LEAVE_GROUP:
131 	case MCAST_JOIN_SOURCE_GROUP:
132 	case MCAST_LEAVE_SOURCE_GROUP:
133 	case MCAST_BLOCK_SOURCE:
134 	case MCAST_UNBLOCK_SOURCE:
135 	case MCAST_MSFILTER:
136 		return true;
137 	}
138 	return false;
139 }
140 
141 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
142 		sockptr_t optval, int optlen)
143 {
144 	if (in_compat_syscall()) {
145 		struct compat_group_source_req gr32;
146 
147 		if (optlen < sizeof(gr32))
148 			return -EINVAL;
149 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
150 			return -EFAULT;
151 		greqs->gsr_interface = gr32.gsr_interface;
152 		greqs->gsr_group = gr32.gsr_group;
153 		greqs->gsr_source = gr32.gsr_source;
154 	} else {
155 		if (optlen < sizeof(*greqs))
156 			return -EINVAL;
157 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
158 			return -EFAULT;
159 	}
160 
161 	return 0;
162 }
163 
164 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
165 		sockptr_t optval, int optlen)
166 {
167 	struct group_source_req greqs;
168 	int omode, add;
169 	int ret;
170 
171 	ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
172 	if (ret)
173 		return ret;
174 
175 	if (greqs.gsr_group.ss_family != AF_INET6 ||
176 	    greqs.gsr_source.ss_family != AF_INET6)
177 		return -EADDRNOTAVAIL;
178 
179 	if (optname == MCAST_BLOCK_SOURCE) {
180 		omode = MCAST_EXCLUDE;
181 		add = 1;
182 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
183 		omode = MCAST_EXCLUDE;
184 		add = 0;
185 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
186 		struct sockaddr_in6 *psin6;
187 		int retv;
188 
189 		psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
190 		retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
191 					     &psin6->sin6_addr,
192 					     MCAST_INCLUDE);
193 		/* prior join w/ different source is ok */
194 		if (retv && retv != -EADDRINUSE)
195 			return retv;
196 		omode = MCAST_INCLUDE;
197 		add = 1;
198 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
199 		omode = MCAST_INCLUDE;
200 		add = 0;
201 	}
202 	return ip6_mc_source(add, omode, sk, &greqs);
203 }
204 
205 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
206 		int optlen)
207 {
208 	struct group_filter *gsf;
209 	int ret;
210 
211 	if (optlen < GROUP_FILTER_SIZE(0))
212 		return -EINVAL;
213 	if (optlen > READ_ONCE(sysctl_optmem_max))
214 		return -ENOBUFS;
215 
216 	gsf = memdup_sockptr(optval, optlen);
217 	if (IS_ERR(gsf))
218 		return PTR_ERR(gsf);
219 
220 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
221 	ret = -ENOBUFS;
222 	if (gsf->gf_numsrc >= 0x1ffffffU ||
223 	    gsf->gf_numsrc > sysctl_mld_max_msf)
224 		goto out_free_gsf;
225 
226 	ret = -EINVAL;
227 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
228 		goto out_free_gsf;
229 
230 	ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
231 out_free_gsf:
232 	kfree(gsf);
233 	return ret;
234 }
235 
236 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
237 		int optlen)
238 {
239 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
240 	struct compat_group_filter *gf32;
241 	void *p;
242 	int ret;
243 	int n;
244 
245 	if (optlen < size0)
246 		return -EINVAL;
247 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
248 		return -ENOBUFS;
249 
250 	p = kmalloc(optlen + 4, GFP_KERNEL);
251 	if (!p)
252 		return -ENOMEM;
253 
254 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
255 	ret = -EFAULT;
256 	if (copy_from_sockptr(gf32, optval, optlen))
257 		goto out_free_p;
258 
259 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
260 	ret = -ENOBUFS;
261 	n = gf32->gf_numsrc;
262 	if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
263 		goto out_free_p;
264 
265 	ret = -EINVAL;
266 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
267 		goto out_free_p;
268 
269 	ret = ip6_mc_msfilter(sk, &(struct group_filter){
270 			.gf_interface = gf32->gf_interface,
271 			.gf_group = gf32->gf_group,
272 			.gf_fmode = gf32->gf_fmode,
273 			.gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
274 
275 out_free_p:
276 	kfree(p);
277 	return ret;
278 }
279 
280 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
281 		sockptr_t optval, int optlen)
282 {
283 	struct sockaddr_in6 *psin6;
284 	struct group_req greq;
285 
286 	if (optlen < sizeof(greq))
287 		return -EINVAL;
288 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
289 		return -EFAULT;
290 
291 	if (greq.gr_group.ss_family != AF_INET6)
292 		return -EADDRNOTAVAIL;
293 	psin6 = (struct sockaddr_in6 *)&greq.gr_group;
294 	if (optname == MCAST_JOIN_GROUP)
295 		return ipv6_sock_mc_join(sk, greq.gr_interface,
296 					 &psin6->sin6_addr);
297 	return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
298 }
299 
300 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
301 		sockptr_t optval, int optlen)
302 {
303 	struct compat_group_req gr32;
304 	struct sockaddr_in6 *psin6;
305 
306 	if (optlen < sizeof(gr32))
307 		return -EINVAL;
308 	if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
309 		return -EFAULT;
310 
311 	if (gr32.gr_group.ss_family != AF_INET6)
312 		return -EADDRNOTAVAIL;
313 	psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
314 	if (optname == MCAST_JOIN_GROUP)
315 		return ipv6_sock_mc_join(sk, gr32.gr_interface,
316 					&psin6->sin6_addr);
317 	return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
318 }
319 
320 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
321 		int optlen)
322 {
323 	struct ipv6_pinfo *np = inet6_sk(sk);
324 	struct ipv6_opt_hdr *new = NULL;
325 	struct net *net = sock_net(sk);
326 	struct ipv6_txoptions *opt;
327 	int err;
328 
329 	/* hop-by-hop / destination options are privileged option */
330 	if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
331 		return -EPERM;
332 
333 	/* remove any sticky options header with a zero option
334 	 * length, per RFC3542.
335 	 */
336 	if (optlen > 0) {
337 		if (sockptr_is_null(optval))
338 			return -EINVAL;
339 		if (optlen < sizeof(struct ipv6_opt_hdr) ||
340 		    optlen & 0x7 ||
341 		    optlen > 8 * 255)
342 			return -EINVAL;
343 
344 		new = memdup_sockptr(optval, optlen);
345 		if (IS_ERR(new))
346 			return PTR_ERR(new);
347 		if (unlikely(ipv6_optlen(new) > optlen)) {
348 			kfree(new);
349 			return -EINVAL;
350 		}
351 	}
352 
353 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
354 	opt = ipv6_renew_options(sk, opt, optname, new);
355 	kfree(new);
356 	if (IS_ERR(opt))
357 		return PTR_ERR(opt);
358 
359 	/* routing header option needs extra check */
360 	err = -EINVAL;
361 	if (optname == IPV6_RTHDR && opt && opt->srcrt) {
362 		struct ipv6_rt_hdr *rthdr = opt->srcrt;
363 		switch (rthdr->type) {
364 #if IS_ENABLED(CONFIG_IPV6_MIP6)
365 		case IPV6_SRCRT_TYPE_2:
366 			if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
367 				goto sticky_done;
368 			break;
369 #endif
370 		case IPV6_SRCRT_TYPE_4:
371 		{
372 			struct ipv6_sr_hdr *srh =
373 				(struct ipv6_sr_hdr *)opt->srcrt;
374 
375 			if (!seg6_validate_srh(srh, optlen, false))
376 				goto sticky_done;
377 			break;
378 		}
379 		default:
380 			goto sticky_done;
381 		}
382 	}
383 
384 	err = 0;
385 	opt = ipv6_update_options(sk, opt);
386 sticky_done:
387 	if (opt) {
388 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
389 		txopt_put(opt);
390 	}
391 	return err;
392 }
393 
394 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
395 		       sockptr_t optval, unsigned int optlen)
396 {
397 	struct ipv6_pinfo *np = inet6_sk(sk);
398 	struct net *net = sock_net(sk);
399 	int val, valbool;
400 	int retv = -ENOPROTOOPT;
401 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
402 
403 	if (sockptr_is_null(optval))
404 		val = 0;
405 	else {
406 		if (optlen >= sizeof(int)) {
407 			if (copy_from_sockptr(&val, optval, sizeof(val)))
408 				return -EFAULT;
409 		} else
410 			val = 0;
411 	}
412 
413 	valbool = (val != 0);
414 
415 	if (ip6_mroute_opt(optname))
416 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
417 
418 	if (needs_rtnl)
419 		rtnl_lock();
420 	sockopt_lock_sock(sk);
421 
422 	/* Another thread has converted the socket into IPv4 with
423 	 * IPV6_ADDRFORM concurrently.
424 	 */
425 	if (unlikely(sk->sk_family != AF_INET6))
426 		goto unlock;
427 
428 	switch (optname) {
429 
430 	case IPV6_ADDRFORM:
431 		if (optlen < sizeof(int))
432 			goto e_inval;
433 		if (val == PF_INET) {
434 			if (sk->sk_type == SOCK_RAW)
435 				break;
436 
437 			if (sk->sk_protocol == IPPROTO_UDP ||
438 			    sk->sk_protocol == IPPROTO_UDPLITE) {
439 				struct udp_sock *up = udp_sk(sk);
440 				if (up->pending == AF_INET6) {
441 					retv = -EBUSY;
442 					break;
443 				}
444 			} else if (sk->sk_protocol == IPPROTO_TCP) {
445 				if (sk->sk_prot != &tcpv6_prot) {
446 					retv = -EBUSY;
447 					break;
448 				}
449 			} else {
450 				break;
451 			}
452 
453 			if (sk->sk_state != TCP_ESTABLISHED) {
454 				retv = -ENOTCONN;
455 				break;
456 			}
457 
458 			if (ipv6_only_sock(sk) ||
459 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
460 				retv = -EADDRNOTAVAIL;
461 				break;
462 			}
463 
464 			__ipv6_sock_mc_close(sk);
465 			__ipv6_sock_ac_close(sk);
466 
467 			/*
468 			 * Sock is moving from IPv6 to IPv4 (sk_prot), so
469 			 * remove it from the refcnt debug socks count in the
470 			 * original family...
471 			 */
472 			sk_refcnt_debug_dec(sk);
473 
474 			if (sk->sk_protocol == IPPROTO_TCP) {
475 				struct inet_connection_sock *icsk = inet_csk(sk);
476 
477 				sock_prot_inuse_add(net, sk->sk_prot, -1);
478 				sock_prot_inuse_add(net, &tcp_prot, 1);
479 
480 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
481 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
482 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
483 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
484 				sk->sk_socket->ops = &inet_stream_ops;
485 				sk->sk_family = PF_INET;
486 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
487 			} else {
488 				struct proto *prot = &udp_prot;
489 
490 				if (sk->sk_protocol == IPPROTO_UDPLITE)
491 					prot = &udplite_prot;
492 
493 				sock_prot_inuse_add(net, sk->sk_prot, -1);
494 				sock_prot_inuse_add(net, prot, 1);
495 
496 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
497 				WRITE_ONCE(sk->sk_prot, prot);
498 				sk->sk_socket->ops = &inet_dgram_ops;
499 				sk->sk_family = PF_INET;
500 			}
501 
502 			/* Disable all options not to allocate memory anymore,
503 			 * but there is still a race.  See the lockless path
504 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
505 			 */
506 			np->rxopt.all = 0;
507 
508 			inet6_cleanup_sock(sk);
509 
510 			/*
511 			 * ... and add it to the refcnt debug socks count
512 			 * in the new family. -acme
513 			 */
514 			sk_refcnt_debug_inc(sk);
515 			module_put(THIS_MODULE);
516 			retv = 0;
517 			break;
518 		}
519 		goto e_inval;
520 
521 	case IPV6_V6ONLY:
522 		if (optlen < sizeof(int) ||
523 		    inet_sk(sk)->inet_num)
524 			goto e_inval;
525 		sk->sk_ipv6only = valbool;
526 		retv = 0;
527 		break;
528 
529 	case IPV6_RECVPKTINFO:
530 		if (optlen < sizeof(int))
531 			goto e_inval;
532 		np->rxopt.bits.rxinfo = valbool;
533 		retv = 0;
534 		break;
535 
536 	case IPV6_2292PKTINFO:
537 		if (optlen < sizeof(int))
538 			goto e_inval;
539 		np->rxopt.bits.rxoinfo = valbool;
540 		retv = 0;
541 		break;
542 
543 	case IPV6_RECVHOPLIMIT:
544 		if (optlen < sizeof(int))
545 			goto e_inval;
546 		np->rxopt.bits.rxhlim = valbool;
547 		retv = 0;
548 		break;
549 
550 	case IPV6_2292HOPLIMIT:
551 		if (optlen < sizeof(int))
552 			goto e_inval;
553 		np->rxopt.bits.rxohlim = valbool;
554 		retv = 0;
555 		break;
556 
557 	case IPV6_RECVRTHDR:
558 		if (optlen < sizeof(int))
559 			goto e_inval;
560 		np->rxopt.bits.srcrt = valbool;
561 		retv = 0;
562 		break;
563 
564 	case IPV6_2292RTHDR:
565 		if (optlen < sizeof(int))
566 			goto e_inval;
567 		np->rxopt.bits.osrcrt = valbool;
568 		retv = 0;
569 		break;
570 
571 	case IPV6_RECVHOPOPTS:
572 		if (optlen < sizeof(int))
573 			goto e_inval;
574 		np->rxopt.bits.hopopts = valbool;
575 		retv = 0;
576 		break;
577 
578 	case IPV6_2292HOPOPTS:
579 		if (optlen < sizeof(int))
580 			goto e_inval;
581 		np->rxopt.bits.ohopopts = valbool;
582 		retv = 0;
583 		break;
584 
585 	case IPV6_RECVDSTOPTS:
586 		if (optlen < sizeof(int))
587 			goto e_inval;
588 		np->rxopt.bits.dstopts = valbool;
589 		retv = 0;
590 		break;
591 
592 	case IPV6_2292DSTOPTS:
593 		if (optlen < sizeof(int))
594 			goto e_inval;
595 		np->rxopt.bits.odstopts = valbool;
596 		retv = 0;
597 		break;
598 
599 	case IPV6_TCLASS:
600 		if (optlen < sizeof(int))
601 			goto e_inval;
602 		if (val < -1 || val > 0xff)
603 			goto e_inval;
604 		/* RFC 3542, 6.5: default traffic class of 0x0 */
605 		if (val == -1)
606 			val = 0;
607 		if (sk->sk_type == SOCK_STREAM) {
608 			val &= ~INET_ECN_MASK;
609 			val |= np->tclass & INET_ECN_MASK;
610 		}
611 		if (np->tclass != val) {
612 			np->tclass = val;
613 			sk_dst_reset(sk);
614 		}
615 		retv = 0;
616 		break;
617 
618 	case IPV6_RECVTCLASS:
619 		if (optlen < sizeof(int))
620 			goto e_inval;
621 		np->rxopt.bits.rxtclass = valbool;
622 		retv = 0;
623 		break;
624 
625 	case IPV6_FLOWINFO:
626 		if (optlen < sizeof(int))
627 			goto e_inval;
628 		np->rxopt.bits.rxflow = valbool;
629 		retv = 0;
630 		break;
631 
632 	case IPV6_RECVPATHMTU:
633 		if (optlen < sizeof(int))
634 			goto e_inval;
635 		np->rxopt.bits.rxpmtu = valbool;
636 		retv = 0;
637 		break;
638 
639 	case IPV6_TRANSPARENT:
640 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
641 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
642 			retv = -EPERM;
643 			break;
644 		}
645 		if (optlen < sizeof(int))
646 			goto e_inval;
647 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
648 		inet_sk(sk)->transparent = valbool;
649 		retv = 0;
650 		break;
651 
652 	case IPV6_FREEBIND:
653 		if (optlen < sizeof(int))
654 			goto e_inval;
655 		/* we also don't have a separate freebind bit for IPV6 */
656 		inet_sk(sk)->freebind = valbool;
657 		retv = 0;
658 		break;
659 
660 	case IPV6_RECVORIGDSTADDR:
661 		if (optlen < sizeof(int))
662 			goto e_inval;
663 		np->rxopt.bits.rxorigdstaddr = valbool;
664 		retv = 0;
665 		break;
666 
667 	case IPV6_HOPOPTS:
668 	case IPV6_RTHDRDSTOPTS:
669 	case IPV6_RTHDR:
670 	case IPV6_DSTOPTS:
671 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
672 		break;
673 
674 	case IPV6_PKTINFO:
675 	{
676 		struct in6_pktinfo pkt;
677 
678 		if (optlen == 0)
679 			goto e_inval;
680 		else if (optlen < sizeof(struct in6_pktinfo) ||
681 			 sockptr_is_null(optval))
682 			goto e_inval;
683 
684 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
685 			retv = -EFAULT;
686 			break;
687 		}
688 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
689 			goto e_inval;
690 
691 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
692 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
693 		retv = 0;
694 		break;
695 	}
696 
697 	case IPV6_2292PKTOPTIONS:
698 	{
699 		struct ipv6_txoptions *opt = NULL;
700 		struct msghdr msg;
701 		struct flowi6 fl6;
702 		struct ipcm6_cookie ipc6;
703 
704 		memset(&fl6, 0, sizeof(fl6));
705 		fl6.flowi6_oif = sk->sk_bound_dev_if;
706 		fl6.flowi6_mark = sk->sk_mark;
707 
708 		if (optlen == 0)
709 			goto update;
710 
711 		/* 1K is probably excessive
712 		 * 1K is surely not enough, 2K per standard header is 16K.
713 		 */
714 		retv = -EINVAL;
715 		if (optlen > 64*1024)
716 			break;
717 
718 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
719 		retv = -ENOBUFS;
720 		if (!opt)
721 			break;
722 
723 		memset(opt, 0, sizeof(*opt));
724 		refcount_set(&opt->refcnt, 1);
725 		opt->tot_len = sizeof(*opt) + optlen;
726 		retv = -EFAULT;
727 		if (copy_from_sockptr(opt + 1, optval, optlen))
728 			goto done;
729 
730 		msg.msg_controllen = optlen;
731 		msg.msg_control = (void *)(opt+1);
732 		ipc6.opt = opt;
733 
734 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
735 		if (retv)
736 			goto done;
737 update:
738 		retv = 0;
739 		opt = ipv6_update_options(sk, opt);
740 done:
741 		if (opt) {
742 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
743 			txopt_put(opt);
744 		}
745 		break;
746 	}
747 	case IPV6_UNICAST_HOPS:
748 		if (optlen < sizeof(int))
749 			goto e_inval;
750 		if (val > 255 || val < -1)
751 			goto e_inval;
752 		np->hop_limit = val;
753 		retv = 0;
754 		break;
755 
756 	case IPV6_MULTICAST_HOPS:
757 		if (sk->sk_type == SOCK_STREAM)
758 			break;
759 		if (optlen < sizeof(int))
760 			goto e_inval;
761 		if (val > 255 || val < -1)
762 			goto e_inval;
763 		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
764 		retv = 0;
765 		break;
766 
767 	case IPV6_MULTICAST_LOOP:
768 		if (optlen < sizeof(int))
769 			goto e_inval;
770 		if (val != valbool)
771 			goto e_inval;
772 		np->mc_loop = valbool;
773 		retv = 0;
774 		break;
775 
776 	case IPV6_UNICAST_IF:
777 	{
778 		struct net_device *dev = NULL;
779 		int ifindex;
780 
781 		if (optlen != sizeof(int))
782 			goto e_inval;
783 
784 		ifindex = (__force int)ntohl((__force __be32)val);
785 		if (ifindex == 0) {
786 			np->ucast_oif = 0;
787 			retv = 0;
788 			break;
789 		}
790 
791 		dev = dev_get_by_index(net, ifindex);
792 		retv = -EADDRNOTAVAIL;
793 		if (!dev)
794 			break;
795 		dev_put(dev);
796 
797 		retv = -EINVAL;
798 		if (sk->sk_bound_dev_if)
799 			break;
800 
801 		np->ucast_oif = ifindex;
802 		retv = 0;
803 		break;
804 	}
805 
806 	case IPV6_MULTICAST_IF:
807 		if (sk->sk_type == SOCK_STREAM)
808 			break;
809 		if (optlen < sizeof(int))
810 			goto e_inval;
811 
812 		if (val) {
813 			struct net_device *dev;
814 			int midx;
815 
816 			rcu_read_lock();
817 
818 			dev = dev_get_by_index_rcu(net, val);
819 			if (!dev) {
820 				rcu_read_unlock();
821 				retv = -ENODEV;
822 				break;
823 			}
824 			midx = l3mdev_master_ifindex_rcu(dev);
825 
826 			rcu_read_unlock();
827 
828 			if (sk->sk_bound_dev_if &&
829 			    sk->sk_bound_dev_if != val &&
830 			    (!midx || midx != sk->sk_bound_dev_if))
831 				goto e_inval;
832 		}
833 		np->mcast_oif = val;
834 		retv = 0;
835 		break;
836 	case IPV6_ADD_MEMBERSHIP:
837 	case IPV6_DROP_MEMBERSHIP:
838 	{
839 		struct ipv6_mreq mreq;
840 
841 		if (optlen < sizeof(struct ipv6_mreq))
842 			goto e_inval;
843 
844 		retv = -EPROTO;
845 		if (inet_sk(sk)->is_icsk)
846 			break;
847 
848 		retv = -EFAULT;
849 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
850 			break;
851 
852 		if (optname == IPV6_ADD_MEMBERSHIP)
853 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
854 		else
855 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
856 		break;
857 	}
858 	case IPV6_JOIN_ANYCAST:
859 	case IPV6_LEAVE_ANYCAST:
860 	{
861 		struct ipv6_mreq mreq;
862 
863 		if (optlen < sizeof(struct ipv6_mreq))
864 			goto e_inval;
865 
866 		retv = -EFAULT;
867 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
868 			break;
869 
870 		if (optname == IPV6_JOIN_ANYCAST)
871 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
872 		else
873 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
874 		break;
875 	}
876 	case IPV6_MULTICAST_ALL:
877 		if (optlen < sizeof(int))
878 			goto e_inval;
879 		np->mc_all = valbool;
880 		retv = 0;
881 		break;
882 
883 	case MCAST_JOIN_GROUP:
884 	case MCAST_LEAVE_GROUP:
885 		if (in_compat_syscall())
886 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
887 							    optlen);
888 		else
889 			retv = ipv6_mcast_join_leave(sk, optname, optval,
890 						     optlen);
891 		break;
892 	case MCAST_JOIN_SOURCE_GROUP:
893 	case MCAST_LEAVE_SOURCE_GROUP:
894 	case MCAST_BLOCK_SOURCE:
895 	case MCAST_UNBLOCK_SOURCE:
896 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
897 		break;
898 	case MCAST_MSFILTER:
899 		if (in_compat_syscall())
900 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
901 							      optlen);
902 		else
903 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
904 		break;
905 	case IPV6_ROUTER_ALERT:
906 		if (optlen < sizeof(int))
907 			goto e_inval;
908 		retv = ip6_ra_control(sk, val);
909 		break;
910 	case IPV6_ROUTER_ALERT_ISOLATE:
911 		if (optlen < sizeof(int))
912 			goto e_inval;
913 		np->rtalert_isolate = valbool;
914 		retv = 0;
915 		break;
916 	case IPV6_MTU_DISCOVER:
917 		if (optlen < sizeof(int))
918 			goto e_inval;
919 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
920 			goto e_inval;
921 		np->pmtudisc = val;
922 		retv = 0;
923 		break;
924 	case IPV6_MTU:
925 		if (optlen < sizeof(int))
926 			goto e_inval;
927 		if (val && val < IPV6_MIN_MTU)
928 			goto e_inval;
929 		np->frag_size = val;
930 		retv = 0;
931 		break;
932 	case IPV6_RECVERR:
933 		if (optlen < sizeof(int))
934 			goto e_inval;
935 		np->recverr = valbool;
936 		if (!val)
937 			skb_queue_purge(&sk->sk_error_queue);
938 		retv = 0;
939 		break;
940 	case IPV6_FLOWINFO_SEND:
941 		if (optlen < sizeof(int))
942 			goto e_inval;
943 		np->sndflow = valbool;
944 		retv = 0;
945 		break;
946 	case IPV6_FLOWLABEL_MGR:
947 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
948 		break;
949 	case IPV6_IPSEC_POLICY:
950 	case IPV6_XFRM_POLICY:
951 		retv = -EPERM;
952 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
953 			break;
954 		retv = xfrm_user_policy(sk, optname, optval, optlen);
955 		break;
956 
957 	case IPV6_ADDR_PREFERENCES:
958 		if (optlen < sizeof(int))
959 			goto e_inval;
960 		retv = __ip6_sock_set_addr_preferences(sk, val);
961 		break;
962 	case IPV6_MINHOPCOUNT:
963 		if (optlen < sizeof(int))
964 			goto e_inval;
965 		if (val < 0 || val > 255)
966 			goto e_inval;
967 
968 		if (val)
969 			static_branch_enable(&ip6_min_hopcount);
970 
971 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
972 		 * while we are changing it.
973 		 */
974 		WRITE_ONCE(np->min_hopcount, val);
975 		retv = 0;
976 		break;
977 	case IPV6_DONTFRAG:
978 		np->dontfrag = valbool;
979 		retv = 0;
980 		break;
981 	case IPV6_AUTOFLOWLABEL:
982 		np->autoflowlabel = valbool;
983 		np->autoflowlabel_set = 1;
984 		retv = 0;
985 		break;
986 	case IPV6_RECVFRAGSIZE:
987 		np->rxopt.bits.recvfragsize = valbool;
988 		retv = 0;
989 		break;
990 	case IPV6_RECVERR_RFC4884:
991 		if (optlen < sizeof(int))
992 			goto e_inval;
993 		if (val < 0 || val > 1)
994 			goto e_inval;
995 		np->recverr_rfc4884 = valbool;
996 		retv = 0;
997 		break;
998 	}
999 
1000 unlock:
1001 	sockopt_release_sock(sk);
1002 	if (needs_rtnl)
1003 		rtnl_unlock();
1004 
1005 	return retv;
1006 
1007 e_inval:
1008 	retv = -EINVAL;
1009 	goto unlock;
1010 }
1011 
1012 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1013 		    unsigned int optlen)
1014 {
1015 	int err;
1016 
1017 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1018 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1019 
1020 	if (level != SOL_IPV6)
1021 		return -ENOPROTOOPT;
1022 
1023 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1024 #ifdef CONFIG_NETFILTER
1025 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1026 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1027 			optname != IPV6_XFRM_POLICY)
1028 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1029 #endif
1030 	return err;
1031 }
1032 EXPORT_SYMBOL(ipv6_setsockopt);
1033 
1034 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1035 				  int optname, sockptr_t optval, int len)
1036 {
1037 	struct ipv6_opt_hdr *hdr;
1038 
1039 	if (!opt)
1040 		return 0;
1041 
1042 	switch (optname) {
1043 	case IPV6_HOPOPTS:
1044 		hdr = opt->hopopt;
1045 		break;
1046 	case IPV6_RTHDRDSTOPTS:
1047 		hdr = opt->dst0opt;
1048 		break;
1049 	case IPV6_RTHDR:
1050 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1051 		break;
1052 	case IPV6_DSTOPTS:
1053 		hdr = opt->dst1opt;
1054 		break;
1055 	default:
1056 		return -EINVAL;	/* should not happen */
1057 	}
1058 
1059 	if (!hdr)
1060 		return 0;
1061 
1062 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1063 	if (copy_to_sockptr(optval, hdr, len))
1064 		return -EFAULT;
1065 	return len;
1066 }
1067 
1068 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1069 			     sockptr_t optlen, int len)
1070 {
1071 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1072 	struct group_filter gsf;
1073 	int num;
1074 	int err;
1075 
1076 	if (len < size0)
1077 		return -EINVAL;
1078 	if (copy_from_sockptr(&gsf, optval, size0))
1079 		return -EFAULT;
1080 	if (gsf.gf_group.ss_family != AF_INET6)
1081 		return -EADDRNOTAVAIL;
1082 	num = gsf.gf_numsrc;
1083 	sockopt_lock_sock(sk);
1084 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1085 	if (!err) {
1086 		if (num > gsf.gf_numsrc)
1087 			num = gsf.gf_numsrc;
1088 		len = GROUP_FILTER_SIZE(num);
1089 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1090 		    copy_to_sockptr(optval, &gsf, size0))
1091 			err = -EFAULT;
1092 	}
1093 	sockopt_release_sock(sk);
1094 	return err;
1095 }
1096 
1097 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1098 				    sockptr_t optlen, int len)
1099 {
1100 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1101 	struct compat_group_filter gf32;
1102 	struct group_filter gf;
1103 	int err;
1104 	int num;
1105 
1106 	if (len < size0)
1107 		return -EINVAL;
1108 
1109 	if (copy_from_sockptr(&gf32, optval, size0))
1110 		return -EFAULT;
1111 	gf.gf_interface = gf32.gf_interface;
1112 	gf.gf_fmode = gf32.gf_fmode;
1113 	num = gf.gf_numsrc = gf32.gf_numsrc;
1114 	gf.gf_group = gf32.gf_group;
1115 
1116 	if (gf.gf_group.ss_family != AF_INET6)
1117 		return -EADDRNOTAVAIL;
1118 
1119 	sockopt_lock_sock(sk);
1120 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1121 	sockopt_release_sock(sk);
1122 	if (err)
1123 		return err;
1124 	if (num > gf.gf_numsrc)
1125 		num = gf.gf_numsrc;
1126 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1127 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1128 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1129 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1130 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1131 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1132 		return -EFAULT;
1133 	return 0;
1134 }
1135 
1136 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1137 		       sockptr_t optval, sockptr_t optlen)
1138 {
1139 	struct ipv6_pinfo *np = inet6_sk(sk);
1140 	int len;
1141 	int val;
1142 
1143 	if (ip6_mroute_opt(optname))
1144 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1145 
1146 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1147 		return -EFAULT;
1148 	switch (optname) {
1149 	case IPV6_ADDRFORM:
1150 		if (sk->sk_protocol != IPPROTO_UDP &&
1151 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1152 		    sk->sk_protocol != IPPROTO_TCP)
1153 			return -ENOPROTOOPT;
1154 		if (sk->sk_state != TCP_ESTABLISHED)
1155 			return -ENOTCONN;
1156 		val = sk->sk_family;
1157 		break;
1158 	case MCAST_MSFILTER:
1159 		if (in_compat_syscall())
1160 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1161 		return ipv6_get_msfilter(sk, optval, optlen, len);
1162 	case IPV6_2292PKTOPTIONS:
1163 	{
1164 		struct msghdr msg;
1165 		struct sk_buff *skb;
1166 
1167 		if (sk->sk_type != SOCK_STREAM)
1168 			return -ENOPROTOOPT;
1169 
1170 		if (optval.is_kernel) {
1171 			msg.msg_control_is_user = false;
1172 			msg.msg_control = optval.kernel;
1173 		} else {
1174 			msg.msg_control_is_user = true;
1175 			msg.msg_control_user = optval.user;
1176 		}
1177 		msg.msg_controllen = len;
1178 		msg.msg_flags = 0;
1179 
1180 		sockopt_lock_sock(sk);
1181 		skb = np->pktoptions;
1182 		if (skb)
1183 			ip6_datagram_recv_ctl(sk, &msg, skb);
1184 		sockopt_release_sock(sk);
1185 		if (!skb) {
1186 			if (np->rxopt.bits.rxinfo) {
1187 				struct in6_pktinfo src_info;
1188 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1189 					np->sticky_pktinfo.ipi6_ifindex;
1190 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1191 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1192 			}
1193 			if (np->rxopt.bits.rxhlim) {
1194 				int hlim = np->mcast_hops;
1195 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1196 			}
1197 			if (np->rxopt.bits.rxtclass) {
1198 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1199 
1200 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1201 			}
1202 			if (np->rxopt.bits.rxoinfo) {
1203 				struct in6_pktinfo src_info;
1204 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1205 					np->sticky_pktinfo.ipi6_ifindex;
1206 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1207 								     np->sticky_pktinfo.ipi6_addr;
1208 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1209 			}
1210 			if (np->rxopt.bits.rxohlim) {
1211 				int hlim = np->mcast_hops;
1212 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1213 			}
1214 			if (np->rxopt.bits.rxflow) {
1215 				__be32 flowinfo = np->rcv_flowinfo;
1216 
1217 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1218 			}
1219 		}
1220 		len -= msg.msg_controllen;
1221 		return copy_to_sockptr(optlen, &len, sizeof(int));
1222 	}
1223 	case IPV6_MTU:
1224 	{
1225 		struct dst_entry *dst;
1226 
1227 		val = 0;
1228 		rcu_read_lock();
1229 		dst = __sk_dst_get(sk);
1230 		if (dst)
1231 			val = dst_mtu(dst);
1232 		rcu_read_unlock();
1233 		if (!val)
1234 			return -ENOTCONN;
1235 		break;
1236 	}
1237 
1238 	case IPV6_V6ONLY:
1239 		val = sk->sk_ipv6only;
1240 		break;
1241 
1242 	case IPV6_RECVPKTINFO:
1243 		val = np->rxopt.bits.rxinfo;
1244 		break;
1245 
1246 	case IPV6_2292PKTINFO:
1247 		val = np->rxopt.bits.rxoinfo;
1248 		break;
1249 
1250 	case IPV6_RECVHOPLIMIT:
1251 		val = np->rxopt.bits.rxhlim;
1252 		break;
1253 
1254 	case IPV6_2292HOPLIMIT:
1255 		val = np->rxopt.bits.rxohlim;
1256 		break;
1257 
1258 	case IPV6_RECVRTHDR:
1259 		val = np->rxopt.bits.srcrt;
1260 		break;
1261 
1262 	case IPV6_2292RTHDR:
1263 		val = np->rxopt.bits.osrcrt;
1264 		break;
1265 
1266 	case IPV6_HOPOPTS:
1267 	case IPV6_RTHDRDSTOPTS:
1268 	case IPV6_RTHDR:
1269 	case IPV6_DSTOPTS:
1270 	{
1271 		struct ipv6_txoptions *opt;
1272 
1273 		sockopt_lock_sock(sk);
1274 		opt = rcu_dereference_protected(np->opt,
1275 						lockdep_sock_is_held(sk));
1276 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1277 		sockopt_release_sock(sk);
1278 		/* check if ipv6_getsockopt_sticky() returns err code */
1279 		if (len < 0)
1280 			return len;
1281 		return copy_to_sockptr(optlen, &len, sizeof(int));
1282 	}
1283 
1284 	case IPV6_RECVHOPOPTS:
1285 		val = np->rxopt.bits.hopopts;
1286 		break;
1287 
1288 	case IPV6_2292HOPOPTS:
1289 		val = np->rxopt.bits.ohopopts;
1290 		break;
1291 
1292 	case IPV6_RECVDSTOPTS:
1293 		val = np->rxopt.bits.dstopts;
1294 		break;
1295 
1296 	case IPV6_2292DSTOPTS:
1297 		val = np->rxopt.bits.odstopts;
1298 		break;
1299 
1300 	case IPV6_TCLASS:
1301 		val = np->tclass;
1302 		break;
1303 
1304 	case IPV6_RECVTCLASS:
1305 		val = np->rxopt.bits.rxtclass;
1306 		break;
1307 
1308 	case IPV6_FLOWINFO:
1309 		val = np->rxopt.bits.rxflow;
1310 		break;
1311 
1312 	case IPV6_RECVPATHMTU:
1313 		val = np->rxopt.bits.rxpmtu;
1314 		break;
1315 
1316 	case IPV6_PATHMTU:
1317 	{
1318 		struct dst_entry *dst;
1319 		struct ip6_mtuinfo mtuinfo;
1320 
1321 		if (len < sizeof(mtuinfo))
1322 			return -EINVAL;
1323 
1324 		len = sizeof(mtuinfo);
1325 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1326 
1327 		rcu_read_lock();
1328 		dst = __sk_dst_get(sk);
1329 		if (dst)
1330 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1331 		rcu_read_unlock();
1332 		if (!mtuinfo.ip6m_mtu)
1333 			return -ENOTCONN;
1334 
1335 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1336 			return -EFAULT;
1337 		if (copy_to_sockptr(optval, &mtuinfo, len))
1338 			return -EFAULT;
1339 
1340 		return 0;
1341 	}
1342 
1343 	case IPV6_TRANSPARENT:
1344 		val = inet_sk(sk)->transparent;
1345 		break;
1346 
1347 	case IPV6_FREEBIND:
1348 		val = inet_sk(sk)->freebind;
1349 		break;
1350 
1351 	case IPV6_RECVORIGDSTADDR:
1352 		val = np->rxopt.bits.rxorigdstaddr;
1353 		break;
1354 
1355 	case IPV6_UNICAST_HOPS:
1356 	case IPV6_MULTICAST_HOPS:
1357 	{
1358 		struct dst_entry *dst;
1359 
1360 		if (optname == IPV6_UNICAST_HOPS)
1361 			val = np->hop_limit;
1362 		else
1363 			val = np->mcast_hops;
1364 
1365 		if (val < 0) {
1366 			rcu_read_lock();
1367 			dst = __sk_dst_get(sk);
1368 			if (dst)
1369 				val = ip6_dst_hoplimit(dst);
1370 			rcu_read_unlock();
1371 		}
1372 
1373 		if (val < 0)
1374 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1375 		break;
1376 	}
1377 
1378 	case IPV6_MULTICAST_LOOP:
1379 		val = np->mc_loop;
1380 		break;
1381 
1382 	case IPV6_MULTICAST_IF:
1383 		val = np->mcast_oif;
1384 		break;
1385 
1386 	case IPV6_MULTICAST_ALL:
1387 		val = np->mc_all;
1388 		break;
1389 
1390 	case IPV6_UNICAST_IF:
1391 		val = (__force int)htonl((__u32) np->ucast_oif);
1392 		break;
1393 
1394 	case IPV6_MTU_DISCOVER:
1395 		val = np->pmtudisc;
1396 		break;
1397 
1398 	case IPV6_RECVERR:
1399 		val = np->recverr;
1400 		break;
1401 
1402 	case IPV6_FLOWINFO_SEND:
1403 		val = np->sndflow;
1404 		break;
1405 
1406 	case IPV6_FLOWLABEL_MGR:
1407 	{
1408 		struct in6_flowlabel_req freq;
1409 		int flags;
1410 
1411 		if (len < sizeof(freq))
1412 			return -EINVAL;
1413 
1414 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1415 			return -EFAULT;
1416 
1417 		if (freq.flr_action != IPV6_FL_A_GET)
1418 			return -EINVAL;
1419 
1420 		len = sizeof(freq);
1421 		flags = freq.flr_flags;
1422 
1423 		memset(&freq, 0, sizeof(freq));
1424 
1425 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1426 		if (val < 0)
1427 			return val;
1428 
1429 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1430 			return -EFAULT;
1431 		if (copy_to_sockptr(optval, &freq, len))
1432 			return -EFAULT;
1433 
1434 		return 0;
1435 	}
1436 
1437 	case IPV6_ADDR_PREFERENCES:
1438 		val = 0;
1439 
1440 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1441 			val |= IPV6_PREFER_SRC_TMP;
1442 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1443 			val |= IPV6_PREFER_SRC_PUBLIC;
1444 		else {
1445 			/* XXX: should we return system default? */
1446 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1447 		}
1448 
1449 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1450 			val |= IPV6_PREFER_SRC_COA;
1451 		else
1452 			val |= IPV6_PREFER_SRC_HOME;
1453 		break;
1454 
1455 	case IPV6_MINHOPCOUNT:
1456 		val = np->min_hopcount;
1457 		break;
1458 
1459 	case IPV6_DONTFRAG:
1460 		val = np->dontfrag;
1461 		break;
1462 
1463 	case IPV6_AUTOFLOWLABEL:
1464 		val = ip6_autoflowlabel(sock_net(sk), np);
1465 		break;
1466 
1467 	case IPV6_RECVFRAGSIZE:
1468 		val = np->rxopt.bits.recvfragsize;
1469 		break;
1470 
1471 	case IPV6_ROUTER_ALERT_ISOLATE:
1472 		val = np->rtalert_isolate;
1473 		break;
1474 
1475 	case IPV6_RECVERR_RFC4884:
1476 		val = np->recverr_rfc4884;
1477 		break;
1478 
1479 	default:
1480 		return -ENOPROTOOPT;
1481 	}
1482 	len = min_t(unsigned int, sizeof(int), len);
1483 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1484 		return -EFAULT;
1485 	if (copy_to_sockptr(optval, &val, len))
1486 		return -EFAULT;
1487 	return 0;
1488 }
1489 
1490 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1491 		    char __user *optval, int __user *optlen)
1492 {
1493 	int err;
1494 
1495 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1496 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1497 
1498 	if (level != SOL_IPV6)
1499 		return -ENOPROTOOPT;
1500 
1501 	err = do_ipv6_getsockopt(sk, level, optname,
1502 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1503 #ifdef CONFIG_NETFILTER
1504 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1505 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1506 		int len;
1507 
1508 		if (get_user(len, optlen))
1509 			return -EFAULT;
1510 
1511 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1512 		if (err >= 0)
1513 			err = put_user(len, optlen);
1514 	}
1515 #endif
1516 	return err;
1517 }
1518 EXPORT_SYMBOL(ipv6_getsockopt);
1519