xref: /linux/net/ipv6/ipv6_sockglue.c (revision e73d5fb75d67538b13fe04d14a55f32ea4efe227)
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 	}
509 	if (needs_rtnl)
510 		rtnl_lock();
511 	sockopt_lock_sock(sk);
512 
513 	/* Another thread has converted the socket into IPv4 with
514 	 * IPV6_ADDRFORM concurrently.
515 	 */
516 	if (unlikely(sk->sk_family != AF_INET6))
517 		goto unlock;
518 
519 	switch (optname) {
520 
521 	case IPV6_ADDRFORM:
522 		if (optlen < sizeof(int))
523 			goto e_inval;
524 		if (val == PF_INET) {
525 			if (sk->sk_type == SOCK_RAW)
526 				break;
527 
528 			if (sk->sk_protocol == IPPROTO_UDP ||
529 			    sk->sk_protocol == IPPROTO_UDPLITE) {
530 				struct udp_sock *up = udp_sk(sk);
531 				if (up->pending == AF_INET6) {
532 					retv = -EBUSY;
533 					break;
534 				}
535 			} else if (sk->sk_protocol == IPPROTO_TCP) {
536 				if (sk->sk_prot != &tcpv6_prot) {
537 					retv = -EBUSY;
538 					break;
539 				}
540 			} else {
541 				break;
542 			}
543 
544 			if (sk->sk_state != TCP_ESTABLISHED) {
545 				retv = -ENOTCONN;
546 				break;
547 			}
548 
549 			if (ipv6_only_sock(sk) ||
550 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
551 				retv = -EADDRNOTAVAIL;
552 				break;
553 			}
554 
555 			__ipv6_sock_mc_close(sk);
556 			__ipv6_sock_ac_close(sk);
557 
558 			if (sk->sk_protocol == IPPROTO_TCP) {
559 				struct inet_connection_sock *icsk = inet_csk(sk);
560 
561 				sock_prot_inuse_add(net, sk->sk_prot, -1);
562 				sock_prot_inuse_add(net, &tcp_prot, 1);
563 
564 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
565 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
566 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
567 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
568 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
569 				WRITE_ONCE(sk->sk_family, PF_INET);
570 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
571 			} else {
572 				struct proto *prot = &udp_prot;
573 
574 				if (sk->sk_protocol == IPPROTO_UDPLITE)
575 					prot = &udplite_prot;
576 
577 				sock_prot_inuse_add(net, sk->sk_prot, -1);
578 				sock_prot_inuse_add(net, prot, 1);
579 
580 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
581 				WRITE_ONCE(sk->sk_prot, prot);
582 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
583 				WRITE_ONCE(sk->sk_family, PF_INET);
584 			}
585 
586 			/* Disable all options not to allocate memory anymore,
587 			 * but there is still a race.  See the lockless path
588 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
589 			 */
590 			np->rxopt.all = 0;
591 
592 			inet6_cleanup_sock(sk);
593 
594 			module_put(THIS_MODULE);
595 			retv = 0;
596 			break;
597 		}
598 		goto e_inval;
599 
600 	case IPV6_V6ONLY:
601 		if (optlen < sizeof(int) ||
602 		    inet_sk(sk)->inet_num)
603 			goto e_inval;
604 		sk->sk_ipv6only = valbool;
605 		retv = 0;
606 		break;
607 
608 	case IPV6_RECVPKTINFO:
609 		if (optlen < sizeof(int))
610 			goto e_inval;
611 		np->rxopt.bits.rxinfo = valbool;
612 		retv = 0;
613 		break;
614 
615 	case IPV6_2292PKTINFO:
616 		if (optlen < sizeof(int))
617 			goto e_inval;
618 		np->rxopt.bits.rxoinfo = valbool;
619 		retv = 0;
620 		break;
621 
622 	case IPV6_RECVHOPLIMIT:
623 		if (optlen < sizeof(int))
624 			goto e_inval;
625 		np->rxopt.bits.rxhlim = valbool;
626 		retv = 0;
627 		break;
628 
629 	case IPV6_2292HOPLIMIT:
630 		if (optlen < sizeof(int))
631 			goto e_inval;
632 		np->rxopt.bits.rxohlim = valbool;
633 		retv = 0;
634 		break;
635 
636 	case IPV6_RECVRTHDR:
637 		if (optlen < sizeof(int))
638 			goto e_inval;
639 		np->rxopt.bits.srcrt = valbool;
640 		retv = 0;
641 		break;
642 
643 	case IPV6_2292RTHDR:
644 		if (optlen < sizeof(int))
645 			goto e_inval;
646 		np->rxopt.bits.osrcrt = valbool;
647 		retv = 0;
648 		break;
649 
650 	case IPV6_RECVHOPOPTS:
651 		if (optlen < sizeof(int))
652 			goto e_inval;
653 		np->rxopt.bits.hopopts = valbool;
654 		retv = 0;
655 		break;
656 
657 	case IPV6_2292HOPOPTS:
658 		if (optlen < sizeof(int))
659 			goto e_inval;
660 		np->rxopt.bits.ohopopts = valbool;
661 		retv = 0;
662 		break;
663 
664 	case IPV6_RECVDSTOPTS:
665 		if (optlen < sizeof(int))
666 			goto e_inval;
667 		np->rxopt.bits.dstopts = valbool;
668 		retv = 0;
669 		break;
670 
671 	case IPV6_2292DSTOPTS:
672 		if (optlen < sizeof(int))
673 			goto e_inval;
674 		np->rxopt.bits.odstopts = valbool;
675 		retv = 0;
676 		break;
677 
678 	case IPV6_TCLASS:
679 		if (optlen < sizeof(int))
680 			goto e_inval;
681 		if (val < -1 || val > 0xff)
682 			goto e_inval;
683 		/* RFC 3542, 6.5: default traffic class of 0x0 */
684 		if (val == -1)
685 			val = 0;
686 		if (sk->sk_type == SOCK_STREAM) {
687 			val &= ~INET_ECN_MASK;
688 			val |= np->tclass & INET_ECN_MASK;
689 		}
690 		if (np->tclass != val) {
691 			np->tclass = val;
692 			sk_dst_reset(sk);
693 		}
694 		retv = 0;
695 		break;
696 
697 	case IPV6_RECVTCLASS:
698 		if (optlen < sizeof(int))
699 			goto e_inval;
700 		np->rxopt.bits.rxtclass = valbool;
701 		retv = 0;
702 		break;
703 
704 	case IPV6_FLOWINFO:
705 		if (optlen < sizeof(int))
706 			goto e_inval;
707 		np->rxopt.bits.rxflow = valbool;
708 		retv = 0;
709 		break;
710 
711 	case IPV6_RECVPATHMTU:
712 		if (optlen < sizeof(int))
713 			goto e_inval;
714 		np->rxopt.bits.rxpmtu = valbool;
715 		retv = 0;
716 		break;
717 
718 	case IPV6_TRANSPARENT:
719 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
720 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
721 			retv = -EPERM;
722 			break;
723 		}
724 		if (optlen < sizeof(int))
725 			goto e_inval;
726 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
727 		inet_assign_bit(TRANSPARENT, sk, valbool);
728 		retv = 0;
729 		break;
730 
731 	case IPV6_FREEBIND:
732 		if (optlen < sizeof(int))
733 			goto e_inval;
734 		/* we also don't have a separate freebind bit for IPV6 */
735 		inet_assign_bit(FREEBIND, sk, valbool);
736 		retv = 0;
737 		break;
738 
739 	case IPV6_RECVORIGDSTADDR:
740 		if (optlen < sizeof(int))
741 			goto e_inval;
742 		np->rxopt.bits.rxorigdstaddr = valbool;
743 		retv = 0;
744 		break;
745 
746 	case IPV6_HOPOPTS:
747 	case IPV6_RTHDRDSTOPTS:
748 	case IPV6_RTHDR:
749 	case IPV6_DSTOPTS:
750 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
751 		break;
752 
753 	case IPV6_PKTINFO:
754 	{
755 		struct in6_pktinfo pkt;
756 
757 		if (optlen == 0)
758 			goto e_inval;
759 		else if (optlen < sizeof(struct in6_pktinfo) ||
760 			 sockptr_is_null(optval))
761 			goto e_inval;
762 
763 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
764 			retv = -EFAULT;
765 			break;
766 		}
767 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
768 			goto e_inval;
769 
770 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
771 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
772 		retv = 0;
773 		break;
774 	}
775 
776 	case IPV6_2292PKTOPTIONS:
777 	{
778 		struct ipv6_txoptions *opt = NULL;
779 		struct msghdr msg;
780 		struct flowi6 fl6;
781 		struct ipcm6_cookie ipc6;
782 
783 		memset(&fl6, 0, sizeof(fl6));
784 		fl6.flowi6_oif = sk->sk_bound_dev_if;
785 		fl6.flowi6_mark = sk->sk_mark;
786 
787 		if (optlen == 0)
788 			goto update;
789 
790 		/* 1K is probably excessive
791 		 * 1K is surely not enough, 2K per standard header is 16K.
792 		 */
793 		retv = -EINVAL;
794 		if (optlen > 64*1024)
795 			break;
796 
797 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
798 		retv = -ENOBUFS;
799 		if (!opt)
800 			break;
801 
802 		memset(opt, 0, sizeof(*opt));
803 		refcount_set(&opt->refcnt, 1);
804 		opt->tot_len = sizeof(*opt) + optlen;
805 		retv = -EFAULT;
806 		if (copy_from_sockptr(opt + 1, optval, optlen))
807 			goto done;
808 
809 		msg.msg_controllen = optlen;
810 		msg.msg_control_is_user = false;
811 		msg.msg_control = (void *)(opt+1);
812 		ipc6.opt = opt;
813 
814 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
815 		if (retv)
816 			goto done;
817 update:
818 		retv = 0;
819 		opt = ipv6_update_options(sk, opt);
820 done:
821 		if (opt) {
822 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
823 			txopt_put(opt);
824 		}
825 		break;
826 	}
827 
828 
829 	case IPV6_UNICAST_IF:
830 	{
831 		struct net_device *dev = NULL;
832 		int ifindex;
833 
834 		if (optlen != sizeof(int))
835 			goto e_inval;
836 
837 		ifindex = (__force int)ntohl((__force __be32)val);
838 		if (ifindex == 0) {
839 			np->ucast_oif = 0;
840 			retv = 0;
841 			break;
842 		}
843 
844 		dev = dev_get_by_index(net, ifindex);
845 		retv = -EADDRNOTAVAIL;
846 		if (!dev)
847 			break;
848 		dev_put(dev);
849 
850 		retv = -EINVAL;
851 		if (sk->sk_bound_dev_if)
852 			break;
853 
854 		np->ucast_oif = ifindex;
855 		retv = 0;
856 		break;
857 	}
858 
859 	case IPV6_MULTICAST_IF:
860 		if (sk->sk_type == SOCK_STREAM)
861 			break;
862 		if (optlen < sizeof(int))
863 			goto e_inval;
864 
865 		if (val) {
866 			struct net_device *dev;
867 			int midx;
868 
869 			rcu_read_lock();
870 
871 			dev = dev_get_by_index_rcu(net, val);
872 			if (!dev) {
873 				rcu_read_unlock();
874 				retv = -ENODEV;
875 				break;
876 			}
877 			midx = l3mdev_master_ifindex_rcu(dev);
878 
879 			rcu_read_unlock();
880 
881 			if (sk->sk_bound_dev_if &&
882 			    sk->sk_bound_dev_if != val &&
883 			    (!midx || midx != sk->sk_bound_dev_if))
884 				goto e_inval;
885 		}
886 		np->mcast_oif = val;
887 		retv = 0;
888 		break;
889 	case IPV6_ADD_MEMBERSHIP:
890 	case IPV6_DROP_MEMBERSHIP:
891 	{
892 		struct ipv6_mreq mreq;
893 
894 		if (optlen < sizeof(struct ipv6_mreq))
895 			goto e_inval;
896 
897 		retv = -EPROTO;
898 		if (inet_test_bit(IS_ICSK, sk))
899 			break;
900 
901 		retv = -EFAULT;
902 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
903 			break;
904 
905 		if (optname == IPV6_ADD_MEMBERSHIP)
906 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
907 		else
908 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
909 		break;
910 	}
911 	case IPV6_JOIN_ANYCAST:
912 	case IPV6_LEAVE_ANYCAST:
913 	{
914 		struct ipv6_mreq mreq;
915 
916 		if (optlen < sizeof(struct ipv6_mreq))
917 			goto e_inval;
918 
919 		retv = -EFAULT;
920 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
921 			break;
922 
923 		if (optname == IPV6_JOIN_ANYCAST)
924 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
925 		else
926 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
927 		break;
928 	}
929 	case MCAST_JOIN_GROUP:
930 	case MCAST_LEAVE_GROUP:
931 		if (in_compat_syscall())
932 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
933 							    optlen);
934 		else
935 			retv = ipv6_mcast_join_leave(sk, optname, optval,
936 						     optlen);
937 		break;
938 	case MCAST_JOIN_SOURCE_GROUP:
939 	case MCAST_LEAVE_SOURCE_GROUP:
940 	case MCAST_BLOCK_SOURCE:
941 	case MCAST_UNBLOCK_SOURCE:
942 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
943 		break;
944 	case MCAST_MSFILTER:
945 		if (in_compat_syscall())
946 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
947 							      optlen);
948 		else
949 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
950 		break;
951 	case IPV6_ROUTER_ALERT:
952 		if (optlen < sizeof(int))
953 			goto e_inval;
954 		retv = ip6_ra_control(sk, val);
955 		break;
956 	case IPV6_FLOWLABEL_MGR:
957 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
958 		break;
959 	case IPV6_IPSEC_POLICY:
960 	case IPV6_XFRM_POLICY:
961 		retv = -EPERM;
962 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
963 			break;
964 		retv = xfrm_user_policy(sk, optname, optval, optlen);
965 		break;
966 
967 	case IPV6_ADDR_PREFERENCES:
968 		if (optlen < sizeof(int))
969 			goto e_inval;
970 		retv = __ip6_sock_set_addr_preferences(sk, val);
971 		break;
972 	case IPV6_RECVFRAGSIZE:
973 		np->rxopt.bits.recvfragsize = valbool;
974 		retv = 0;
975 		break;
976 	}
977 
978 unlock:
979 	sockopt_release_sock(sk);
980 	if (needs_rtnl)
981 		rtnl_unlock();
982 
983 	return retv;
984 
985 e_inval:
986 	retv = -EINVAL;
987 	goto unlock;
988 }
989 
990 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
991 		    unsigned int optlen)
992 {
993 	int err;
994 
995 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
996 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
997 
998 	if (level != SOL_IPV6)
999 		return -ENOPROTOOPT;
1000 
1001 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1002 #ifdef CONFIG_NETFILTER
1003 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1004 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1005 			optname != IPV6_XFRM_POLICY)
1006 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1007 #endif
1008 	return err;
1009 }
1010 EXPORT_SYMBOL(ipv6_setsockopt);
1011 
1012 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1013 				  int optname, sockptr_t optval, int len)
1014 {
1015 	struct ipv6_opt_hdr *hdr;
1016 
1017 	if (!opt)
1018 		return 0;
1019 
1020 	switch (optname) {
1021 	case IPV6_HOPOPTS:
1022 		hdr = opt->hopopt;
1023 		break;
1024 	case IPV6_RTHDRDSTOPTS:
1025 		hdr = opt->dst0opt;
1026 		break;
1027 	case IPV6_RTHDR:
1028 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1029 		break;
1030 	case IPV6_DSTOPTS:
1031 		hdr = opt->dst1opt;
1032 		break;
1033 	default:
1034 		return -EINVAL;	/* should not happen */
1035 	}
1036 
1037 	if (!hdr)
1038 		return 0;
1039 
1040 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1041 	if (copy_to_sockptr(optval, hdr, len))
1042 		return -EFAULT;
1043 	return len;
1044 }
1045 
1046 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1047 			     sockptr_t optlen, int len)
1048 {
1049 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1050 	struct group_filter gsf;
1051 	int num;
1052 	int err;
1053 
1054 	if (len < size0)
1055 		return -EINVAL;
1056 	if (copy_from_sockptr(&gsf, optval, size0))
1057 		return -EFAULT;
1058 	if (gsf.gf_group.ss_family != AF_INET6)
1059 		return -EADDRNOTAVAIL;
1060 	num = gsf.gf_numsrc;
1061 	sockopt_lock_sock(sk);
1062 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1063 	if (!err) {
1064 		if (num > gsf.gf_numsrc)
1065 			num = gsf.gf_numsrc;
1066 		len = GROUP_FILTER_SIZE(num);
1067 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1068 		    copy_to_sockptr(optval, &gsf, size0))
1069 			err = -EFAULT;
1070 	}
1071 	sockopt_release_sock(sk);
1072 	return err;
1073 }
1074 
1075 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1076 				    sockptr_t optlen, int len)
1077 {
1078 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1079 	struct compat_group_filter gf32;
1080 	struct group_filter gf;
1081 	int err;
1082 	int num;
1083 
1084 	if (len < size0)
1085 		return -EINVAL;
1086 
1087 	if (copy_from_sockptr(&gf32, optval, size0))
1088 		return -EFAULT;
1089 	gf.gf_interface = gf32.gf_interface;
1090 	gf.gf_fmode = gf32.gf_fmode;
1091 	num = gf.gf_numsrc = gf32.gf_numsrc;
1092 	gf.gf_group = gf32.gf_group;
1093 
1094 	if (gf.gf_group.ss_family != AF_INET6)
1095 		return -EADDRNOTAVAIL;
1096 
1097 	sockopt_lock_sock(sk);
1098 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1099 	sockopt_release_sock(sk);
1100 	if (err)
1101 		return err;
1102 	if (num > gf.gf_numsrc)
1103 		num = gf.gf_numsrc;
1104 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1105 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1106 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1107 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1108 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1109 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1110 		return -EFAULT;
1111 	return 0;
1112 }
1113 
1114 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1115 		       sockptr_t optval, sockptr_t optlen)
1116 {
1117 	struct ipv6_pinfo *np = inet6_sk(sk);
1118 	int len;
1119 	int val;
1120 
1121 	if (ip6_mroute_opt(optname))
1122 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1123 
1124 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1125 		return -EFAULT;
1126 	switch (optname) {
1127 	case IPV6_ADDRFORM:
1128 		if (sk->sk_protocol != IPPROTO_UDP &&
1129 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1130 		    sk->sk_protocol != IPPROTO_TCP)
1131 			return -ENOPROTOOPT;
1132 		if (sk->sk_state != TCP_ESTABLISHED)
1133 			return -ENOTCONN;
1134 		val = sk->sk_family;
1135 		break;
1136 	case MCAST_MSFILTER:
1137 		if (in_compat_syscall())
1138 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1139 		return ipv6_get_msfilter(sk, optval, optlen, len);
1140 	case IPV6_2292PKTOPTIONS:
1141 	{
1142 		struct msghdr msg;
1143 		struct sk_buff *skb;
1144 
1145 		if (sk->sk_type != SOCK_STREAM)
1146 			return -ENOPROTOOPT;
1147 
1148 		if (optval.is_kernel) {
1149 			msg.msg_control_is_user = false;
1150 			msg.msg_control = optval.kernel;
1151 		} else {
1152 			msg.msg_control_is_user = true;
1153 			msg.msg_control_user = optval.user;
1154 		}
1155 		msg.msg_controllen = len;
1156 		msg.msg_flags = 0;
1157 
1158 		sockopt_lock_sock(sk);
1159 		skb = np->pktoptions;
1160 		if (skb)
1161 			ip6_datagram_recv_ctl(sk, &msg, skb);
1162 		sockopt_release_sock(sk);
1163 		if (!skb) {
1164 			if (np->rxopt.bits.rxinfo) {
1165 				struct in6_pktinfo src_info;
1166 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1167 					np->sticky_pktinfo.ipi6_ifindex;
1168 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1169 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1170 			}
1171 			if (np->rxopt.bits.rxhlim) {
1172 				int hlim = READ_ONCE(np->mcast_hops);
1173 
1174 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1175 			}
1176 			if (np->rxopt.bits.rxtclass) {
1177 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1178 
1179 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1180 			}
1181 			if (np->rxopt.bits.rxoinfo) {
1182 				struct in6_pktinfo src_info;
1183 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1184 					np->sticky_pktinfo.ipi6_ifindex;
1185 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1186 								     np->sticky_pktinfo.ipi6_addr;
1187 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1188 			}
1189 			if (np->rxopt.bits.rxohlim) {
1190 				int hlim = READ_ONCE(np->mcast_hops);
1191 
1192 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1193 			}
1194 			if (np->rxopt.bits.rxflow) {
1195 				__be32 flowinfo = np->rcv_flowinfo;
1196 
1197 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1198 			}
1199 		}
1200 		len -= msg.msg_controllen;
1201 		return copy_to_sockptr(optlen, &len, sizeof(int));
1202 	}
1203 	case IPV6_MTU:
1204 	{
1205 		struct dst_entry *dst;
1206 
1207 		val = 0;
1208 		rcu_read_lock();
1209 		dst = __sk_dst_get(sk);
1210 		if (dst)
1211 			val = dst_mtu(dst);
1212 		rcu_read_unlock();
1213 		if (!val)
1214 			return -ENOTCONN;
1215 		break;
1216 	}
1217 
1218 	case IPV6_V6ONLY:
1219 		val = sk->sk_ipv6only;
1220 		break;
1221 
1222 	case IPV6_RECVPKTINFO:
1223 		val = np->rxopt.bits.rxinfo;
1224 		break;
1225 
1226 	case IPV6_2292PKTINFO:
1227 		val = np->rxopt.bits.rxoinfo;
1228 		break;
1229 
1230 	case IPV6_RECVHOPLIMIT:
1231 		val = np->rxopt.bits.rxhlim;
1232 		break;
1233 
1234 	case IPV6_2292HOPLIMIT:
1235 		val = np->rxopt.bits.rxohlim;
1236 		break;
1237 
1238 	case IPV6_RECVRTHDR:
1239 		val = np->rxopt.bits.srcrt;
1240 		break;
1241 
1242 	case IPV6_2292RTHDR:
1243 		val = np->rxopt.bits.osrcrt;
1244 		break;
1245 
1246 	case IPV6_HOPOPTS:
1247 	case IPV6_RTHDRDSTOPTS:
1248 	case IPV6_RTHDR:
1249 	case IPV6_DSTOPTS:
1250 	{
1251 		struct ipv6_txoptions *opt;
1252 
1253 		sockopt_lock_sock(sk);
1254 		opt = rcu_dereference_protected(np->opt,
1255 						lockdep_sock_is_held(sk));
1256 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1257 		sockopt_release_sock(sk);
1258 		/* check if ipv6_getsockopt_sticky() returns err code */
1259 		if (len < 0)
1260 			return len;
1261 		return copy_to_sockptr(optlen, &len, sizeof(int));
1262 	}
1263 
1264 	case IPV6_RECVHOPOPTS:
1265 		val = np->rxopt.bits.hopopts;
1266 		break;
1267 
1268 	case IPV6_2292HOPOPTS:
1269 		val = np->rxopt.bits.ohopopts;
1270 		break;
1271 
1272 	case IPV6_RECVDSTOPTS:
1273 		val = np->rxopt.bits.dstopts;
1274 		break;
1275 
1276 	case IPV6_2292DSTOPTS:
1277 		val = np->rxopt.bits.odstopts;
1278 		break;
1279 
1280 	case IPV6_TCLASS:
1281 		val = np->tclass;
1282 		break;
1283 
1284 	case IPV6_RECVTCLASS:
1285 		val = np->rxopt.bits.rxtclass;
1286 		break;
1287 
1288 	case IPV6_FLOWINFO:
1289 		val = np->rxopt.bits.rxflow;
1290 		break;
1291 
1292 	case IPV6_RECVPATHMTU:
1293 		val = np->rxopt.bits.rxpmtu;
1294 		break;
1295 
1296 	case IPV6_PATHMTU:
1297 	{
1298 		struct dst_entry *dst;
1299 		struct ip6_mtuinfo mtuinfo;
1300 
1301 		if (len < sizeof(mtuinfo))
1302 			return -EINVAL;
1303 
1304 		len = sizeof(mtuinfo);
1305 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1306 
1307 		rcu_read_lock();
1308 		dst = __sk_dst_get(sk);
1309 		if (dst)
1310 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1311 		rcu_read_unlock();
1312 		if (!mtuinfo.ip6m_mtu)
1313 			return -ENOTCONN;
1314 
1315 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1316 			return -EFAULT;
1317 		if (copy_to_sockptr(optval, &mtuinfo, len))
1318 			return -EFAULT;
1319 
1320 		return 0;
1321 	}
1322 
1323 	case IPV6_TRANSPARENT:
1324 		val = inet_test_bit(TRANSPARENT, sk);
1325 		break;
1326 
1327 	case IPV6_FREEBIND:
1328 		val = inet_test_bit(FREEBIND, sk);
1329 		break;
1330 
1331 	case IPV6_RECVORIGDSTADDR:
1332 		val = np->rxopt.bits.rxorigdstaddr;
1333 		break;
1334 
1335 	case IPV6_UNICAST_HOPS:
1336 	case IPV6_MULTICAST_HOPS:
1337 	{
1338 		struct dst_entry *dst;
1339 
1340 		if (optname == IPV6_UNICAST_HOPS)
1341 			val = READ_ONCE(np->hop_limit);
1342 		else
1343 			val = READ_ONCE(np->mcast_hops);
1344 
1345 		if (val < 0) {
1346 			rcu_read_lock();
1347 			dst = __sk_dst_get(sk);
1348 			if (dst)
1349 				val = ip6_dst_hoplimit(dst);
1350 			rcu_read_unlock();
1351 		}
1352 
1353 		if (val < 0)
1354 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1355 		break;
1356 	}
1357 
1358 	case IPV6_MULTICAST_LOOP:
1359 		val = inet6_test_bit(MC6_LOOP, sk);
1360 		break;
1361 
1362 	case IPV6_MULTICAST_IF:
1363 		val = np->mcast_oif;
1364 		break;
1365 
1366 	case IPV6_MULTICAST_ALL:
1367 		val = inet6_test_bit(MC6_ALL, sk);
1368 		break;
1369 
1370 	case IPV6_UNICAST_IF:
1371 		val = (__force int)htonl((__u32) np->ucast_oif);
1372 		break;
1373 
1374 	case IPV6_MTU_DISCOVER:
1375 		val = READ_ONCE(np->pmtudisc);
1376 		break;
1377 
1378 	case IPV6_RECVERR:
1379 		val = inet6_test_bit(RECVERR6, sk);
1380 		break;
1381 
1382 	case IPV6_FLOWINFO_SEND:
1383 		val = inet6_test_bit(SNDFLOW, sk);
1384 		break;
1385 
1386 	case IPV6_FLOWLABEL_MGR:
1387 	{
1388 		struct in6_flowlabel_req freq;
1389 		int flags;
1390 
1391 		if (len < sizeof(freq))
1392 			return -EINVAL;
1393 
1394 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1395 			return -EFAULT;
1396 
1397 		if (freq.flr_action != IPV6_FL_A_GET)
1398 			return -EINVAL;
1399 
1400 		len = sizeof(freq);
1401 		flags = freq.flr_flags;
1402 
1403 		memset(&freq, 0, sizeof(freq));
1404 
1405 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1406 		if (val < 0)
1407 			return val;
1408 
1409 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1410 			return -EFAULT;
1411 		if (copy_to_sockptr(optval, &freq, len))
1412 			return -EFAULT;
1413 
1414 		return 0;
1415 	}
1416 
1417 	case IPV6_ADDR_PREFERENCES:
1418 		val = 0;
1419 
1420 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1421 			val |= IPV6_PREFER_SRC_TMP;
1422 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1423 			val |= IPV6_PREFER_SRC_PUBLIC;
1424 		else {
1425 			/* XXX: should we return system default? */
1426 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1427 		}
1428 
1429 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1430 			val |= IPV6_PREFER_SRC_COA;
1431 		else
1432 			val |= IPV6_PREFER_SRC_HOME;
1433 		break;
1434 
1435 	case IPV6_MINHOPCOUNT:
1436 		val = READ_ONCE(np->min_hopcount);
1437 		break;
1438 
1439 	case IPV6_DONTFRAG:
1440 		val = inet6_test_bit(DONTFRAG, sk);
1441 		break;
1442 
1443 	case IPV6_AUTOFLOWLABEL:
1444 		val = ip6_autoflowlabel(sock_net(sk), sk);
1445 		break;
1446 
1447 	case IPV6_RECVFRAGSIZE:
1448 		val = np->rxopt.bits.recvfragsize;
1449 		break;
1450 
1451 	case IPV6_ROUTER_ALERT_ISOLATE:
1452 		val = inet6_test_bit(RTALERT_ISOLATE, sk);
1453 		break;
1454 
1455 	case IPV6_RECVERR_RFC4884:
1456 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1457 		break;
1458 
1459 	default:
1460 		return -ENOPROTOOPT;
1461 	}
1462 	len = min_t(unsigned int, sizeof(int), len);
1463 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1464 		return -EFAULT;
1465 	if (copy_to_sockptr(optval, &val, len))
1466 		return -EFAULT;
1467 	return 0;
1468 }
1469 
1470 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1471 		    char __user *optval, int __user *optlen)
1472 {
1473 	int err;
1474 
1475 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1476 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1477 
1478 	if (level != SOL_IPV6)
1479 		return -ENOPROTOOPT;
1480 
1481 	err = do_ipv6_getsockopt(sk, level, optname,
1482 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1483 #ifdef CONFIG_NETFILTER
1484 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1485 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1486 		int len;
1487 
1488 		if (get_user(len, optlen))
1489 			return -EFAULT;
1490 
1491 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1492 		if (err >= 0)
1493 			err = put_user(len, optlen);
1494 	}
1495 #endif
1496 	return err;
1497 }
1498 EXPORT_SYMBOL(ipv6_getsockopt);
1499