xref: /linux/net/ipv6/ipv6_sockglue.c (revision ebc733e54a1a79ea2dde2ba5121ae73a188e20d4)
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_test_bit(IS_ICSK, sk)) {
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 	/* Handle options that can be set without locking the socket. */
419 	switch (optname) {
420 	case IPV6_UNICAST_HOPS:
421 		if (optlen < sizeof(int))
422 			return -EINVAL;
423 		if (val > 255 || val < -1)
424 			return -EINVAL;
425 		WRITE_ONCE(np->hop_limit, val);
426 		return 0;
427 	case IPV6_MULTICAST_LOOP:
428 		if (optlen < sizeof(int))
429 			return -EINVAL;
430 		if (val != valbool)
431 			return -EINVAL;
432 		inet6_assign_bit(MC6_LOOP, sk, valbool);
433 		return 0;
434 	case IPV6_MULTICAST_HOPS:
435 		if (sk->sk_type == SOCK_STREAM)
436 			return retv;
437 		if (optlen < sizeof(int))
438 			return -EINVAL;
439 		if (val > 255 || val < -1)
440 			return -EINVAL;
441 		WRITE_ONCE(np->mcast_hops,
442 			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
443 		return 0;
444 	case IPV6_MTU:
445 		if (optlen < sizeof(int))
446 			return -EINVAL;
447 		if (val && val < IPV6_MIN_MTU)
448 			return -EINVAL;
449 		WRITE_ONCE(np->frag_size, val);
450 		return 0;
451 	case IPV6_MINHOPCOUNT:
452 		if (optlen < sizeof(int))
453 			return -EINVAL;
454 		if (val < 0 || val > 255)
455 			return -EINVAL;
456 
457 		if (val)
458 			static_branch_enable(&ip6_min_hopcount);
459 
460 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
461 		 * while we are changing it.
462 		 */
463 		WRITE_ONCE(np->min_hopcount, val);
464 		return 0;
465 	case IPV6_RECVERR_RFC4884:
466 		if (optlen < sizeof(int))
467 			return -EINVAL;
468 		if (val < 0 || val > 1)
469 			return -EINVAL;
470 		inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
471 		return 0;
472 	case IPV6_MULTICAST_ALL:
473 		if (optlen < sizeof(int))
474 			return -EINVAL;
475 		inet6_assign_bit(MC6_ALL, sk, valbool);
476 		return 0;
477 	case IPV6_AUTOFLOWLABEL:
478 		inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
479 		inet6_set_bit(AUTOFLOWLABEL_SET, sk);
480 		return 0;
481 	case IPV6_DONTFRAG:
482 		inet6_assign_bit(DONTFRAG, sk, valbool);
483 		return 0;
484 	case IPV6_RECVERR:
485 		if (optlen < sizeof(int))
486 			return -EINVAL;
487 		inet6_assign_bit(RECVERR6, sk, valbool);
488 		if (!val)
489 			skb_errqueue_purge(&sk->sk_error_queue);
490 		return 0;
491 	case IPV6_ROUTER_ALERT_ISOLATE:
492 		if (optlen < sizeof(int))
493 			return -EINVAL;
494 		inet6_assign_bit(RTALERT_ISOLATE, sk, valbool);
495 		return 0;
496 	case IPV6_MTU_DISCOVER:
497 		if (optlen < sizeof(int))
498 			return -EINVAL;
499 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
500 			return -EINVAL;
501 		WRITE_ONCE(np->pmtudisc, val);
502 		return 0;
503 	case IPV6_FLOWINFO_SEND:
504 		if (optlen < sizeof(int))
505 			return -EINVAL;
506 		inet6_assign_bit(SNDFLOW, sk, valbool);
507 		return 0;
508 	case IPV6_ADDR_PREFERENCES:
509 		if (optlen < sizeof(int))
510 			return -EINVAL;
511 		return ip6_sock_set_addr_preferences(sk, val);
512 	}
513 	if (needs_rtnl)
514 		rtnl_lock();
515 	sockopt_lock_sock(sk);
516 
517 	/* Another thread has converted the socket into IPv4 with
518 	 * IPV6_ADDRFORM concurrently.
519 	 */
520 	if (unlikely(sk->sk_family != AF_INET6))
521 		goto unlock;
522 
523 	switch (optname) {
524 
525 	case IPV6_ADDRFORM:
526 		if (optlen < sizeof(int))
527 			goto e_inval;
528 		if (val == PF_INET) {
529 			if (sk->sk_type == SOCK_RAW)
530 				break;
531 
532 			if (sk->sk_protocol == IPPROTO_UDP ||
533 			    sk->sk_protocol == IPPROTO_UDPLITE) {
534 				struct udp_sock *up = udp_sk(sk);
535 				if (up->pending == AF_INET6) {
536 					retv = -EBUSY;
537 					break;
538 				}
539 			} else if (sk->sk_protocol == IPPROTO_TCP) {
540 				if (sk->sk_prot != &tcpv6_prot) {
541 					retv = -EBUSY;
542 					break;
543 				}
544 			} else {
545 				break;
546 			}
547 
548 			if (sk->sk_state != TCP_ESTABLISHED) {
549 				retv = -ENOTCONN;
550 				break;
551 			}
552 
553 			if (ipv6_only_sock(sk) ||
554 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
555 				retv = -EADDRNOTAVAIL;
556 				break;
557 			}
558 
559 			__ipv6_sock_mc_close(sk);
560 			__ipv6_sock_ac_close(sk);
561 
562 			if (sk->sk_protocol == IPPROTO_TCP) {
563 				struct inet_connection_sock *icsk = inet_csk(sk);
564 
565 				sock_prot_inuse_add(net, sk->sk_prot, -1);
566 				sock_prot_inuse_add(net, &tcp_prot, 1);
567 
568 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
569 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
570 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
571 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
572 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
573 				WRITE_ONCE(sk->sk_family, PF_INET);
574 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
575 			} else {
576 				struct proto *prot = &udp_prot;
577 
578 				if (sk->sk_protocol == IPPROTO_UDPLITE)
579 					prot = &udplite_prot;
580 
581 				sock_prot_inuse_add(net, sk->sk_prot, -1);
582 				sock_prot_inuse_add(net, prot, 1);
583 
584 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
585 				WRITE_ONCE(sk->sk_prot, prot);
586 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
587 				WRITE_ONCE(sk->sk_family, PF_INET);
588 			}
589 
590 			/* Disable all options not to allocate memory anymore,
591 			 * but there is still a race.  See the lockless path
592 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
593 			 */
594 			np->rxopt.all = 0;
595 
596 			inet6_cleanup_sock(sk);
597 
598 			module_put(THIS_MODULE);
599 			retv = 0;
600 			break;
601 		}
602 		goto e_inval;
603 
604 	case IPV6_V6ONLY:
605 		if (optlen < sizeof(int) ||
606 		    inet_sk(sk)->inet_num)
607 			goto e_inval;
608 		sk->sk_ipv6only = valbool;
609 		retv = 0;
610 		break;
611 
612 	case IPV6_RECVPKTINFO:
613 		if (optlen < sizeof(int))
614 			goto e_inval;
615 		np->rxopt.bits.rxinfo = valbool;
616 		retv = 0;
617 		break;
618 
619 	case IPV6_2292PKTINFO:
620 		if (optlen < sizeof(int))
621 			goto e_inval;
622 		np->rxopt.bits.rxoinfo = valbool;
623 		retv = 0;
624 		break;
625 
626 	case IPV6_RECVHOPLIMIT:
627 		if (optlen < sizeof(int))
628 			goto e_inval;
629 		np->rxopt.bits.rxhlim = valbool;
630 		retv = 0;
631 		break;
632 
633 	case IPV6_2292HOPLIMIT:
634 		if (optlen < sizeof(int))
635 			goto e_inval;
636 		np->rxopt.bits.rxohlim = valbool;
637 		retv = 0;
638 		break;
639 
640 	case IPV6_RECVRTHDR:
641 		if (optlen < sizeof(int))
642 			goto e_inval;
643 		np->rxopt.bits.srcrt = valbool;
644 		retv = 0;
645 		break;
646 
647 	case IPV6_2292RTHDR:
648 		if (optlen < sizeof(int))
649 			goto e_inval;
650 		np->rxopt.bits.osrcrt = valbool;
651 		retv = 0;
652 		break;
653 
654 	case IPV6_RECVHOPOPTS:
655 		if (optlen < sizeof(int))
656 			goto e_inval;
657 		np->rxopt.bits.hopopts = valbool;
658 		retv = 0;
659 		break;
660 
661 	case IPV6_2292HOPOPTS:
662 		if (optlen < sizeof(int))
663 			goto e_inval;
664 		np->rxopt.bits.ohopopts = valbool;
665 		retv = 0;
666 		break;
667 
668 	case IPV6_RECVDSTOPTS:
669 		if (optlen < sizeof(int))
670 			goto e_inval;
671 		np->rxopt.bits.dstopts = valbool;
672 		retv = 0;
673 		break;
674 
675 	case IPV6_2292DSTOPTS:
676 		if (optlen < sizeof(int))
677 			goto e_inval;
678 		np->rxopt.bits.odstopts = valbool;
679 		retv = 0;
680 		break;
681 
682 	case IPV6_TCLASS:
683 		if (optlen < sizeof(int))
684 			goto e_inval;
685 		if (val < -1 || val > 0xff)
686 			goto e_inval;
687 		/* RFC 3542, 6.5: default traffic class of 0x0 */
688 		if (val == -1)
689 			val = 0;
690 		if (sk->sk_type == SOCK_STREAM) {
691 			val &= ~INET_ECN_MASK;
692 			val |= np->tclass & INET_ECN_MASK;
693 		}
694 		if (np->tclass != val) {
695 			np->tclass = val;
696 			sk_dst_reset(sk);
697 		}
698 		retv = 0;
699 		break;
700 
701 	case IPV6_RECVTCLASS:
702 		if (optlen < sizeof(int))
703 			goto e_inval;
704 		np->rxopt.bits.rxtclass = valbool;
705 		retv = 0;
706 		break;
707 
708 	case IPV6_FLOWINFO:
709 		if (optlen < sizeof(int))
710 			goto e_inval;
711 		np->rxopt.bits.rxflow = valbool;
712 		retv = 0;
713 		break;
714 
715 	case IPV6_RECVPATHMTU:
716 		if (optlen < sizeof(int))
717 			goto e_inval;
718 		np->rxopt.bits.rxpmtu = valbool;
719 		retv = 0;
720 		break;
721 
722 	case IPV6_TRANSPARENT:
723 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
724 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
725 			retv = -EPERM;
726 			break;
727 		}
728 		if (optlen < sizeof(int))
729 			goto e_inval;
730 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
731 		inet_assign_bit(TRANSPARENT, sk, valbool);
732 		retv = 0;
733 		break;
734 
735 	case IPV6_FREEBIND:
736 		if (optlen < sizeof(int))
737 			goto e_inval;
738 		/* we also don't have a separate freebind bit for IPV6 */
739 		inet_assign_bit(FREEBIND, sk, valbool);
740 		retv = 0;
741 		break;
742 
743 	case IPV6_RECVORIGDSTADDR:
744 		if (optlen < sizeof(int))
745 			goto e_inval;
746 		np->rxopt.bits.rxorigdstaddr = valbool;
747 		retv = 0;
748 		break;
749 
750 	case IPV6_HOPOPTS:
751 	case IPV6_RTHDRDSTOPTS:
752 	case IPV6_RTHDR:
753 	case IPV6_DSTOPTS:
754 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
755 		break;
756 
757 	case IPV6_PKTINFO:
758 	{
759 		struct in6_pktinfo pkt;
760 
761 		if (optlen == 0)
762 			goto e_inval;
763 		else if (optlen < sizeof(struct in6_pktinfo) ||
764 			 sockptr_is_null(optval))
765 			goto e_inval;
766 
767 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
768 			retv = -EFAULT;
769 			break;
770 		}
771 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
772 			goto e_inval;
773 
774 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
775 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
776 		retv = 0;
777 		break;
778 	}
779 
780 	case IPV6_2292PKTOPTIONS:
781 	{
782 		struct ipv6_txoptions *opt = NULL;
783 		struct msghdr msg;
784 		struct flowi6 fl6;
785 		struct ipcm6_cookie ipc6;
786 
787 		memset(&fl6, 0, sizeof(fl6));
788 		fl6.flowi6_oif = sk->sk_bound_dev_if;
789 		fl6.flowi6_mark = sk->sk_mark;
790 
791 		if (optlen == 0)
792 			goto update;
793 
794 		/* 1K is probably excessive
795 		 * 1K is surely not enough, 2K per standard header is 16K.
796 		 */
797 		retv = -EINVAL;
798 		if (optlen > 64*1024)
799 			break;
800 
801 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
802 		retv = -ENOBUFS;
803 		if (!opt)
804 			break;
805 
806 		memset(opt, 0, sizeof(*opt));
807 		refcount_set(&opt->refcnt, 1);
808 		opt->tot_len = sizeof(*opt) + optlen;
809 		retv = -EFAULT;
810 		if (copy_from_sockptr(opt + 1, optval, optlen))
811 			goto done;
812 
813 		msg.msg_controllen = optlen;
814 		msg.msg_control_is_user = false;
815 		msg.msg_control = (void *)(opt+1);
816 		ipc6.opt = opt;
817 
818 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
819 		if (retv)
820 			goto done;
821 update:
822 		retv = 0;
823 		opt = ipv6_update_options(sk, opt);
824 done:
825 		if (opt) {
826 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
827 			txopt_put(opt);
828 		}
829 		break;
830 	}
831 
832 
833 	case IPV6_UNICAST_IF:
834 	{
835 		struct net_device *dev = NULL;
836 		int ifindex;
837 
838 		if (optlen != sizeof(int))
839 			goto e_inval;
840 
841 		ifindex = (__force int)ntohl((__force __be32)val);
842 		if (ifindex == 0) {
843 			np->ucast_oif = 0;
844 			retv = 0;
845 			break;
846 		}
847 
848 		dev = dev_get_by_index(net, ifindex);
849 		retv = -EADDRNOTAVAIL;
850 		if (!dev)
851 			break;
852 		dev_put(dev);
853 
854 		retv = -EINVAL;
855 		if (sk->sk_bound_dev_if)
856 			break;
857 
858 		np->ucast_oif = ifindex;
859 		retv = 0;
860 		break;
861 	}
862 
863 	case IPV6_MULTICAST_IF:
864 		if (sk->sk_type == SOCK_STREAM)
865 			break;
866 		if (optlen < sizeof(int))
867 			goto e_inval;
868 
869 		if (val) {
870 			struct net_device *dev;
871 			int midx;
872 
873 			rcu_read_lock();
874 
875 			dev = dev_get_by_index_rcu(net, val);
876 			if (!dev) {
877 				rcu_read_unlock();
878 				retv = -ENODEV;
879 				break;
880 			}
881 			midx = l3mdev_master_ifindex_rcu(dev);
882 
883 			rcu_read_unlock();
884 
885 			if (sk->sk_bound_dev_if &&
886 			    sk->sk_bound_dev_if != val &&
887 			    (!midx || midx != sk->sk_bound_dev_if))
888 				goto e_inval;
889 		}
890 		np->mcast_oif = val;
891 		retv = 0;
892 		break;
893 	case IPV6_ADD_MEMBERSHIP:
894 	case IPV6_DROP_MEMBERSHIP:
895 	{
896 		struct ipv6_mreq mreq;
897 
898 		if (optlen < sizeof(struct ipv6_mreq))
899 			goto e_inval;
900 
901 		retv = -EPROTO;
902 		if (inet_test_bit(IS_ICSK, sk))
903 			break;
904 
905 		retv = -EFAULT;
906 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
907 			break;
908 
909 		if (optname == IPV6_ADD_MEMBERSHIP)
910 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
911 		else
912 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
913 		break;
914 	}
915 	case IPV6_JOIN_ANYCAST:
916 	case IPV6_LEAVE_ANYCAST:
917 	{
918 		struct ipv6_mreq mreq;
919 
920 		if (optlen < sizeof(struct ipv6_mreq))
921 			goto e_inval;
922 
923 		retv = -EFAULT;
924 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
925 			break;
926 
927 		if (optname == IPV6_JOIN_ANYCAST)
928 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
929 		else
930 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
931 		break;
932 	}
933 	case MCAST_JOIN_GROUP:
934 	case MCAST_LEAVE_GROUP:
935 		if (in_compat_syscall())
936 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
937 							    optlen);
938 		else
939 			retv = ipv6_mcast_join_leave(sk, optname, optval,
940 						     optlen);
941 		break;
942 	case MCAST_JOIN_SOURCE_GROUP:
943 	case MCAST_LEAVE_SOURCE_GROUP:
944 	case MCAST_BLOCK_SOURCE:
945 	case MCAST_UNBLOCK_SOURCE:
946 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
947 		break;
948 	case MCAST_MSFILTER:
949 		if (in_compat_syscall())
950 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
951 							      optlen);
952 		else
953 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
954 		break;
955 	case IPV6_ROUTER_ALERT:
956 		if (optlen < sizeof(int))
957 			goto e_inval;
958 		retv = ip6_ra_control(sk, val);
959 		break;
960 	case IPV6_FLOWLABEL_MGR:
961 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
962 		break;
963 	case IPV6_IPSEC_POLICY:
964 	case IPV6_XFRM_POLICY:
965 		retv = -EPERM;
966 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
967 			break;
968 		retv = xfrm_user_policy(sk, optname, optval, optlen);
969 		break;
970 
971 	case IPV6_RECVFRAGSIZE:
972 		np->rxopt.bits.recvfragsize = valbool;
973 		retv = 0;
974 		break;
975 	}
976 
977 unlock:
978 	sockopt_release_sock(sk);
979 	if (needs_rtnl)
980 		rtnl_unlock();
981 
982 	return retv;
983 
984 e_inval:
985 	retv = -EINVAL;
986 	goto unlock;
987 }
988 
989 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
990 		    unsigned int optlen)
991 {
992 	int err;
993 
994 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
995 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
996 
997 	if (level != SOL_IPV6)
998 		return -ENOPROTOOPT;
999 
1000 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1001 #ifdef CONFIG_NETFILTER
1002 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1003 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1004 			optname != IPV6_XFRM_POLICY)
1005 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1006 #endif
1007 	return err;
1008 }
1009 EXPORT_SYMBOL(ipv6_setsockopt);
1010 
1011 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1012 				  int optname, sockptr_t optval, int len)
1013 {
1014 	struct ipv6_opt_hdr *hdr;
1015 
1016 	if (!opt)
1017 		return 0;
1018 
1019 	switch (optname) {
1020 	case IPV6_HOPOPTS:
1021 		hdr = opt->hopopt;
1022 		break;
1023 	case IPV6_RTHDRDSTOPTS:
1024 		hdr = opt->dst0opt;
1025 		break;
1026 	case IPV6_RTHDR:
1027 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1028 		break;
1029 	case IPV6_DSTOPTS:
1030 		hdr = opt->dst1opt;
1031 		break;
1032 	default:
1033 		return -EINVAL;	/* should not happen */
1034 	}
1035 
1036 	if (!hdr)
1037 		return 0;
1038 
1039 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1040 	if (copy_to_sockptr(optval, hdr, len))
1041 		return -EFAULT;
1042 	return len;
1043 }
1044 
1045 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1046 			     sockptr_t optlen, int len)
1047 {
1048 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1049 	struct group_filter gsf;
1050 	int num;
1051 	int err;
1052 
1053 	if (len < size0)
1054 		return -EINVAL;
1055 	if (copy_from_sockptr(&gsf, optval, size0))
1056 		return -EFAULT;
1057 	if (gsf.gf_group.ss_family != AF_INET6)
1058 		return -EADDRNOTAVAIL;
1059 	num = gsf.gf_numsrc;
1060 	sockopt_lock_sock(sk);
1061 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1062 	if (!err) {
1063 		if (num > gsf.gf_numsrc)
1064 			num = gsf.gf_numsrc;
1065 		len = GROUP_FILTER_SIZE(num);
1066 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1067 		    copy_to_sockptr(optval, &gsf, size0))
1068 			err = -EFAULT;
1069 	}
1070 	sockopt_release_sock(sk);
1071 	return err;
1072 }
1073 
1074 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1075 				    sockptr_t optlen, int len)
1076 {
1077 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1078 	struct compat_group_filter gf32;
1079 	struct group_filter gf;
1080 	int err;
1081 	int num;
1082 
1083 	if (len < size0)
1084 		return -EINVAL;
1085 
1086 	if (copy_from_sockptr(&gf32, optval, size0))
1087 		return -EFAULT;
1088 	gf.gf_interface = gf32.gf_interface;
1089 	gf.gf_fmode = gf32.gf_fmode;
1090 	num = gf.gf_numsrc = gf32.gf_numsrc;
1091 	gf.gf_group = gf32.gf_group;
1092 
1093 	if (gf.gf_group.ss_family != AF_INET6)
1094 		return -EADDRNOTAVAIL;
1095 
1096 	sockopt_lock_sock(sk);
1097 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1098 	sockopt_release_sock(sk);
1099 	if (err)
1100 		return err;
1101 	if (num > gf.gf_numsrc)
1102 		num = gf.gf_numsrc;
1103 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1104 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1105 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1106 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1107 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1108 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1109 		return -EFAULT;
1110 	return 0;
1111 }
1112 
1113 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1114 		       sockptr_t optval, sockptr_t optlen)
1115 {
1116 	struct ipv6_pinfo *np = inet6_sk(sk);
1117 	int len;
1118 	int val;
1119 
1120 	if (ip6_mroute_opt(optname))
1121 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1122 
1123 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1124 		return -EFAULT;
1125 	switch (optname) {
1126 	case IPV6_ADDRFORM:
1127 		if (sk->sk_protocol != IPPROTO_UDP &&
1128 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1129 		    sk->sk_protocol != IPPROTO_TCP)
1130 			return -ENOPROTOOPT;
1131 		if (sk->sk_state != TCP_ESTABLISHED)
1132 			return -ENOTCONN;
1133 		val = sk->sk_family;
1134 		break;
1135 	case MCAST_MSFILTER:
1136 		if (in_compat_syscall())
1137 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1138 		return ipv6_get_msfilter(sk, optval, optlen, len);
1139 	case IPV6_2292PKTOPTIONS:
1140 	{
1141 		struct msghdr msg;
1142 		struct sk_buff *skb;
1143 
1144 		if (sk->sk_type != SOCK_STREAM)
1145 			return -ENOPROTOOPT;
1146 
1147 		if (optval.is_kernel) {
1148 			msg.msg_control_is_user = false;
1149 			msg.msg_control = optval.kernel;
1150 		} else {
1151 			msg.msg_control_is_user = true;
1152 			msg.msg_control_user = optval.user;
1153 		}
1154 		msg.msg_controllen = len;
1155 		msg.msg_flags = 0;
1156 
1157 		sockopt_lock_sock(sk);
1158 		skb = np->pktoptions;
1159 		if (skb)
1160 			ip6_datagram_recv_ctl(sk, &msg, skb);
1161 		sockopt_release_sock(sk);
1162 		if (!skb) {
1163 			if (np->rxopt.bits.rxinfo) {
1164 				struct in6_pktinfo src_info;
1165 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1166 					np->sticky_pktinfo.ipi6_ifindex;
1167 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1168 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1169 			}
1170 			if (np->rxopt.bits.rxhlim) {
1171 				int hlim = READ_ONCE(np->mcast_hops);
1172 
1173 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1174 			}
1175 			if (np->rxopt.bits.rxtclass) {
1176 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1177 
1178 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1179 			}
1180 			if (np->rxopt.bits.rxoinfo) {
1181 				struct in6_pktinfo src_info;
1182 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1183 					np->sticky_pktinfo.ipi6_ifindex;
1184 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1185 								     np->sticky_pktinfo.ipi6_addr;
1186 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1187 			}
1188 			if (np->rxopt.bits.rxohlim) {
1189 				int hlim = READ_ONCE(np->mcast_hops);
1190 
1191 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1192 			}
1193 			if (np->rxopt.bits.rxflow) {
1194 				__be32 flowinfo = np->rcv_flowinfo;
1195 
1196 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1197 			}
1198 		}
1199 		len -= msg.msg_controllen;
1200 		return copy_to_sockptr(optlen, &len, sizeof(int));
1201 	}
1202 	case IPV6_MTU:
1203 	{
1204 		struct dst_entry *dst;
1205 
1206 		val = 0;
1207 		rcu_read_lock();
1208 		dst = __sk_dst_get(sk);
1209 		if (dst)
1210 			val = dst_mtu(dst);
1211 		rcu_read_unlock();
1212 		if (!val)
1213 			return -ENOTCONN;
1214 		break;
1215 	}
1216 
1217 	case IPV6_V6ONLY:
1218 		val = sk->sk_ipv6only;
1219 		break;
1220 
1221 	case IPV6_RECVPKTINFO:
1222 		val = np->rxopt.bits.rxinfo;
1223 		break;
1224 
1225 	case IPV6_2292PKTINFO:
1226 		val = np->rxopt.bits.rxoinfo;
1227 		break;
1228 
1229 	case IPV6_RECVHOPLIMIT:
1230 		val = np->rxopt.bits.rxhlim;
1231 		break;
1232 
1233 	case IPV6_2292HOPLIMIT:
1234 		val = np->rxopt.bits.rxohlim;
1235 		break;
1236 
1237 	case IPV6_RECVRTHDR:
1238 		val = np->rxopt.bits.srcrt;
1239 		break;
1240 
1241 	case IPV6_2292RTHDR:
1242 		val = np->rxopt.bits.osrcrt;
1243 		break;
1244 
1245 	case IPV6_HOPOPTS:
1246 	case IPV6_RTHDRDSTOPTS:
1247 	case IPV6_RTHDR:
1248 	case IPV6_DSTOPTS:
1249 	{
1250 		struct ipv6_txoptions *opt;
1251 
1252 		sockopt_lock_sock(sk);
1253 		opt = rcu_dereference_protected(np->opt,
1254 						lockdep_sock_is_held(sk));
1255 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1256 		sockopt_release_sock(sk);
1257 		/* check if ipv6_getsockopt_sticky() returns err code */
1258 		if (len < 0)
1259 			return len;
1260 		return copy_to_sockptr(optlen, &len, sizeof(int));
1261 	}
1262 
1263 	case IPV6_RECVHOPOPTS:
1264 		val = np->rxopt.bits.hopopts;
1265 		break;
1266 
1267 	case IPV6_2292HOPOPTS:
1268 		val = np->rxopt.bits.ohopopts;
1269 		break;
1270 
1271 	case IPV6_RECVDSTOPTS:
1272 		val = np->rxopt.bits.dstopts;
1273 		break;
1274 
1275 	case IPV6_2292DSTOPTS:
1276 		val = np->rxopt.bits.odstopts;
1277 		break;
1278 
1279 	case IPV6_TCLASS:
1280 		val = np->tclass;
1281 		break;
1282 
1283 	case IPV6_RECVTCLASS:
1284 		val = np->rxopt.bits.rxtclass;
1285 		break;
1286 
1287 	case IPV6_FLOWINFO:
1288 		val = np->rxopt.bits.rxflow;
1289 		break;
1290 
1291 	case IPV6_RECVPATHMTU:
1292 		val = np->rxopt.bits.rxpmtu;
1293 		break;
1294 
1295 	case IPV6_PATHMTU:
1296 	{
1297 		struct dst_entry *dst;
1298 		struct ip6_mtuinfo mtuinfo;
1299 
1300 		if (len < sizeof(mtuinfo))
1301 			return -EINVAL;
1302 
1303 		len = sizeof(mtuinfo);
1304 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1305 
1306 		rcu_read_lock();
1307 		dst = __sk_dst_get(sk);
1308 		if (dst)
1309 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1310 		rcu_read_unlock();
1311 		if (!mtuinfo.ip6m_mtu)
1312 			return -ENOTCONN;
1313 
1314 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1315 			return -EFAULT;
1316 		if (copy_to_sockptr(optval, &mtuinfo, len))
1317 			return -EFAULT;
1318 
1319 		return 0;
1320 	}
1321 
1322 	case IPV6_TRANSPARENT:
1323 		val = inet_test_bit(TRANSPARENT, sk);
1324 		break;
1325 
1326 	case IPV6_FREEBIND:
1327 		val = inet_test_bit(FREEBIND, sk);
1328 		break;
1329 
1330 	case IPV6_RECVORIGDSTADDR:
1331 		val = np->rxopt.bits.rxorigdstaddr;
1332 		break;
1333 
1334 	case IPV6_UNICAST_HOPS:
1335 	case IPV6_MULTICAST_HOPS:
1336 	{
1337 		struct dst_entry *dst;
1338 
1339 		if (optname == IPV6_UNICAST_HOPS)
1340 			val = READ_ONCE(np->hop_limit);
1341 		else
1342 			val = READ_ONCE(np->mcast_hops);
1343 
1344 		if (val < 0) {
1345 			rcu_read_lock();
1346 			dst = __sk_dst_get(sk);
1347 			if (dst)
1348 				val = ip6_dst_hoplimit(dst);
1349 			rcu_read_unlock();
1350 		}
1351 
1352 		if (val < 0)
1353 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1354 		break;
1355 	}
1356 
1357 	case IPV6_MULTICAST_LOOP:
1358 		val = inet6_test_bit(MC6_LOOP, sk);
1359 		break;
1360 
1361 	case IPV6_MULTICAST_IF:
1362 		val = np->mcast_oif;
1363 		break;
1364 
1365 	case IPV6_MULTICAST_ALL:
1366 		val = inet6_test_bit(MC6_ALL, sk);
1367 		break;
1368 
1369 	case IPV6_UNICAST_IF:
1370 		val = (__force int)htonl((__u32) np->ucast_oif);
1371 		break;
1372 
1373 	case IPV6_MTU_DISCOVER:
1374 		val = READ_ONCE(np->pmtudisc);
1375 		break;
1376 
1377 	case IPV6_RECVERR:
1378 		val = inet6_test_bit(RECVERR6, sk);
1379 		break;
1380 
1381 	case IPV6_FLOWINFO_SEND:
1382 		val = inet6_test_bit(SNDFLOW, sk);
1383 		break;
1384 
1385 	case IPV6_FLOWLABEL_MGR:
1386 	{
1387 		struct in6_flowlabel_req freq;
1388 		int flags;
1389 
1390 		if (len < sizeof(freq))
1391 			return -EINVAL;
1392 
1393 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1394 			return -EFAULT;
1395 
1396 		if (freq.flr_action != IPV6_FL_A_GET)
1397 			return -EINVAL;
1398 
1399 		len = sizeof(freq);
1400 		flags = freq.flr_flags;
1401 
1402 		memset(&freq, 0, sizeof(freq));
1403 
1404 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1405 		if (val < 0)
1406 			return val;
1407 
1408 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1409 			return -EFAULT;
1410 		if (copy_to_sockptr(optval, &freq, len))
1411 			return -EFAULT;
1412 
1413 		return 0;
1414 	}
1415 
1416 	case IPV6_ADDR_PREFERENCES:
1417 		{
1418 		u8 srcprefs = READ_ONCE(np->srcprefs);
1419 		val = 0;
1420 
1421 		if (srcprefs & IPV6_PREFER_SRC_TMP)
1422 			val |= IPV6_PREFER_SRC_TMP;
1423 		else if (srcprefs & IPV6_PREFER_SRC_PUBLIC)
1424 			val |= IPV6_PREFER_SRC_PUBLIC;
1425 		else {
1426 			/* XXX: should we return system default? */
1427 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1428 		}
1429 
1430 		if (srcprefs & IPV6_PREFER_SRC_COA)
1431 			val |= IPV6_PREFER_SRC_COA;
1432 		else
1433 			val |= IPV6_PREFER_SRC_HOME;
1434 		break;
1435 		}
1436 	case IPV6_MINHOPCOUNT:
1437 		val = READ_ONCE(np->min_hopcount);
1438 		break;
1439 
1440 	case IPV6_DONTFRAG:
1441 		val = inet6_test_bit(DONTFRAG, sk);
1442 		break;
1443 
1444 	case IPV6_AUTOFLOWLABEL:
1445 		val = ip6_autoflowlabel(sock_net(sk), sk);
1446 		break;
1447 
1448 	case IPV6_RECVFRAGSIZE:
1449 		val = np->rxopt.bits.recvfragsize;
1450 		break;
1451 
1452 	case IPV6_ROUTER_ALERT_ISOLATE:
1453 		val = inet6_test_bit(RTALERT_ISOLATE, sk);
1454 		break;
1455 
1456 	case IPV6_RECVERR_RFC4884:
1457 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1458 		break;
1459 
1460 	default:
1461 		return -ENOPROTOOPT;
1462 	}
1463 	len = min_t(unsigned int, sizeof(int), len);
1464 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1465 		return -EFAULT;
1466 	if (copy_to_sockptr(optval, &val, len))
1467 		return -EFAULT;
1468 	return 0;
1469 }
1470 
1471 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1472 		    char __user *optval, int __user *optlen)
1473 {
1474 	int err;
1475 
1476 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1477 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1478 
1479 	if (level != SOL_IPV6)
1480 		return -ENOPROTOOPT;
1481 
1482 	err = do_ipv6_getsockopt(sk, level, optname,
1483 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1484 #ifdef CONFIG_NETFILTER
1485 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1486 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1487 		int len;
1488 
1489 		if (get_user(len, optlen))
1490 			return -EFAULT;
1491 
1492 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1493 		if (err >= 0)
1494 			err = put_user(len, optlen);
1495 	}
1496 #endif
1497 	return err;
1498 }
1499 EXPORT_SYMBOL(ipv6_getsockopt);
1500