xref: /linux/net/ipv6/ipv6_sockglue.c (revision e3cdf6cf5fc6db0643723083e2c70fffe098e249)
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/xfrm.h>
49 #include <net/compat.h>
50 #include <net/seg6.h>
51 #include <net/psp.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_obj(*new_ra) : 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 
111 			icsk->icsk_ext_hdr_len =
112 				psp_sk_overhead(sk) +
113 				opt->opt_flen + opt->opt_nflen;
114 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
115 		}
116 	}
117 	opt = unrcu_pointer(xchg(&inet6_sk(sk)->opt, RCU_INITIALIZER(opt)));
118 	sk_dst_reset(sk);
119 
120 	return opt;
121 }
122 
123 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
124 		sockptr_t optval, int optlen)
125 {
126 	if (in_compat_syscall()) {
127 		struct compat_group_source_req gr32;
128 
129 		if (optlen < sizeof(gr32))
130 			return -EINVAL;
131 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
132 			return -EFAULT;
133 		greqs->gsr_interface = gr32.gsr_interface;
134 		greqs->gsr_group = gr32.gsr_group;
135 		greqs->gsr_source = gr32.gsr_source;
136 	} else {
137 		if (optlen < sizeof(*greqs))
138 			return -EINVAL;
139 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
140 			return -EFAULT;
141 	}
142 
143 	return 0;
144 }
145 
146 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
147 		sockptr_t optval, int optlen)
148 {
149 	struct group_source_req greqs;
150 	int omode, add;
151 	int ret;
152 
153 	ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
154 	if (ret)
155 		return ret;
156 
157 	if (greqs.gsr_group.ss_family != AF_INET6 ||
158 	    greqs.gsr_source.ss_family != AF_INET6)
159 		return -EADDRNOTAVAIL;
160 
161 	if (optname == MCAST_BLOCK_SOURCE) {
162 		omode = MCAST_EXCLUDE;
163 		add = 1;
164 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
165 		omode = MCAST_EXCLUDE;
166 		add = 0;
167 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
168 		struct sockaddr_in6 *psin6;
169 		int retv;
170 
171 		psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
172 		retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
173 					     &psin6->sin6_addr,
174 					     MCAST_INCLUDE);
175 		/* prior join w/ different source is ok */
176 		if (retv && retv != -EADDRINUSE)
177 			return retv;
178 		omode = MCAST_INCLUDE;
179 		add = 1;
180 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
181 		omode = MCAST_INCLUDE;
182 		add = 0;
183 	}
184 	return ip6_mc_source(add, omode, sk, &greqs);
185 }
186 
187 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
188 		int optlen)
189 {
190 	struct group_filter *gsf;
191 	int ret;
192 
193 	if (optlen < GROUP_FILTER_SIZE(0))
194 		return -EINVAL;
195 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max))
196 		return -ENOBUFS;
197 
198 	gsf = memdup_sockptr(optval, optlen);
199 	if (IS_ERR(gsf))
200 		return PTR_ERR(gsf);
201 
202 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
203 	ret = -ENOBUFS;
204 	if (gsf->gf_numsrc >= 0x1ffffffU ||
205 	    gsf->gf_numsrc > sysctl_mld_max_msf)
206 		goto out_free_gsf;
207 
208 	ret = -EINVAL;
209 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
210 		goto out_free_gsf;
211 
212 	ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
213 out_free_gsf:
214 	kfree(gsf);
215 	return ret;
216 }
217 
218 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
219 		int optlen)
220 {
221 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
222 	struct compat_group_filter *gf32;
223 	void *p;
224 	int ret;
225 	int n;
226 
227 	if (optlen < size0)
228 		return -EINVAL;
229 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max) - 4)
230 		return -ENOBUFS;
231 
232 	p = kmalloc(optlen + 4, GFP_KERNEL);
233 	if (!p)
234 		return -ENOMEM;
235 
236 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
237 	ret = -EFAULT;
238 	if (copy_from_sockptr(gf32, optval, optlen))
239 		goto out_free_p;
240 
241 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
242 	ret = -ENOBUFS;
243 	n = gf32->gf_numsrc;
244 	if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
245 		goto out_free_p;
246 
247 	ret = -EINVAL;
248 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
249 		goto out_free_p;
250 
251 	ret = ip6_mc_msfilter(sk, &(struct group_filter){
252 			.gf_interface = gf32->gf_interface,
253 			.gf_group = gf32->gf_group,
254 			.gf_fmode = gf32->gf_fmode,
255 			.gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
256 
257 out_free_p:
258 	kfree(p);
259 	return ret;
260 }
261 
262 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
263 		sockptr_t optval, int optlen)
264 {
265 	struct sockaddr_in6 *psin6;
266 	struct group_req greq;
267 
268 	if (optlen < sizeof(greq))
269 		return -EINVAL;
270 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
271 		return -EFAULT;
272 
273 	if (greq.gr_group.ss_family != AF_INET6)
274 		return -EADDRNOTAVAIL;
275 	psin6 = (struct sockaddr_in6 *)&greq.gr_group;
276 	if (optname == MCAST_JOIN_GROUP)
277 		return ipv6_sock_mc_join(sk, greq.gr_interface,
278 					 &psin6->sin6_addr);
279 	return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
280 }
281 
282 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
283 		sockptr_t optval, int optlen)
284 {
285 	struct compat_group_req gr32;
286 	struct sockaddr_in6 *psin6;
287 
288 	if (optlen < sizeof(gr32))
289 		return -EINVAL;
290 	if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
291 		return -EFAULT;
292 
293 	if (gr32.gr_group.ss_family != AF_INET6)
294 		return -EADDRNOTAVAIL;
295 	psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
296 	if (optname == MCAST_JOIN_GROUP)
297 		return ipv6_sock_mc_join(sk, gr32.gr_interface,
298 					&psin6->sin6_addr);
299 	return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
300 }
301 
302 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
303 		int optlen)
304 {
305 	struct ipv6_pinfo *np = inet6_sk(sk);
306 	struct ipv6_opt_hdr *new = NULL;
307 	struct net *net = sock_net(sk);
308 	struct ipv6_txoptions *opt;
309 	int err;
310 
311 	/* hop-by-hop / destination options are privileged option */
312 	if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
313 		return -EPERM;
314 
315 	/* remove any sticky options header with a zero option
316 	 * length, per RFC3542.
317 	 */
318 	if (optlen > 0) {
319 		if (sockptr_is_null(optval))
320 			return -EINVAL;
321 		if (optlen < sizeof(struct ipv6_opt_hdr) ||
322 		    optlen & 0x7 ||
323 		    optlen > 8 * 255)
324 			return -EINVAL;
325 
326 		new = memdup_sockptr(optval, optlen);
327 		if (IS_ERR(new))
328 			return PTR_ERR(new);
329 		if (unlikely(ipv6_optlen(new) > optlen)) {
330 			kfree(new);
331 			return -EINVAL;
332 		}
333 	}
334 
335 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
336 	opt = ipv6_renew_options(sk, opt, optname, new);
337 	kfree(new);
338 	if (IS_ERR(opt))
339 		return PTR_ERR(opt);
340 
341 	/* routing header option needs extra check */
342 	err = -EINVAL;
343 	if (optname == IPV6_RTHDR && opt && opt->srcrt) {
344 		struct ipv6_rt_hdr *rthdr = opt->srcrt;
345 		switch (rthdr->type) {
346 #if IS_ENABLED(CONFIG_IPV6_MIP6)
347 		case IPV6_SRCRT_TYPE_2:
348 			if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
349 				goto sticky_done;
350 			break;
351 #endif
352 		case IPV6_SRCRT_TYPE_4:
353 		{
354 			struct ipv6_sr_hdr *srh =
355 				(struct ipv6_sr_hdr *)opt->srcrt;
356 
357 			if (!seg6_validate_srh(srh, optlen, false))
358 				goto sticky_done;
359 			break;
360 		}
361 		default:
362 			goto sticky_done;
363 		}
364 	}
365 
366 	err = 0;
367 	opt = ipv6_update_options(sk, opt);
368 sticky_done:
369 	if (opt) {
370 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
371 		txopt_put(opt);
372 	}
373 	return err;
374 }
375 
376 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
377 		       sockptr_t optval, unsigned int optlen)
378 {
379 	struct ipv6_pinfo *np = inet6_sk(sk);
380 	struct net *net = sock_net(sk);
381 	int retv = -ENOPROTOOPT;
382 	int val, valbool;
383 
384 	if (sockptr_is_null(optval))
385 		val = 0;
386 	else {
387 		if (optlen >= sizeof(int)) {
388 			if (copy_from_sockptr(&val, optval, sizeof(val)))
389 				return -EFAULT;
390 		} else
391 			val = 0;
392 	}
393 
394 	valbool = (val != 0);
395 
396 	if (ip6_mroute_opt(optname))
397 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
398 
399 	/* Handle options that can be set without locking the socket. */
400 	switch (optname) {
401 	case IPV6_UNICAST_HOPS:
402 		if (optlen < sizeof(int))
403 			return -EINVAL;
404 		if (val > 255 || val < -1)
405 			return -EINVAL;
406 		WRITE_ONCE(np->hop_limit, val);
407 		return 0;
408 	case IPV6_MULTICAST_LOOP:
409 		if (optlen < sizeof(int))
410 			return -EINVAL;
411 		if (val != valbool)
412 			return -EINVAL;
413 		inet6_assign_bit(MC6_LOOP, sk, valbool);
414 		return 0;
415 	case IPV6_MULTICAST_HOPS:
416 		if (sk->sk_type == SOCK_STREAM)
417 			return retv;
418 		if (optlen < sizeof(int))
419 			return -EINVAL;
420 		if (val > 255 || val < -1)
421 			return -EINVAL;
422 		WRITE_ONCE(np->mcast_hops,
423 			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
424 		return 0;
425 	case IPV6_MTU:
426 		if (optlen < sizeof(int))
427 			return -EINVAL;
428 		if (val && val < IPV6_MIN_MTU)
429 			return -EINVAL;
430 		WRITE_ONCE(np->frag_size, val);
431 		return 0;
432 	case IPV6_MINHOPCOUNT:
433 		if (optlen < sizeof(int))
434 			return -EINVAL;
435 		if (val < 0 || val > 255)
436 			return -EINVAL;
437 
438 		if (val)
439 			static_branch_enable(&ip6_min_hopcount);
440 
441 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
442 		 * while we are changing it.
443 		 */
444 		WRITE_ONCE(np->min_hopcount, val);
445 		return 0;
446 	case IPV6_RECVERR_RFC4884:
447 		if (optlen < sizeof(int))
448 			return -EINVAL;
449 		if (val < 0 || val > 1)
450 			return -EINVAL;
451 		inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
452 		return 0;
453 	case IPV6_MULTICAST_ALL:
454 		if (optlen < sizeof(int))
455 			return -EINVAL;
456 		inet6_assign_bit(MC6_ALL, sk, valbool);
457 		return 0;
458 	case IPV6_AUTOFLOWLABEL:
459 		inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
460 		inet6_set_bit(AUTOFLOWLABEL_SET, sk);
461 		return 0;
462 	case IPV6_DONTFRAG:
463 		inet6_assign_bit(DONTFRAG, sk, valbool);
464 		return 0;
465 	case IPV6_RECVERR:
466 		if (optlen < sizeof(int))
467 			return -EINVAL;
468 		inet6_assign_bit(RECVERR6, sk, valbool);
469 		if (!val)
470 			skb_errqueue_purge(&sk->sk_error_queue);
471 		return 0;
472 	case IPV6_ROUTER_ALERT_ISOLATE:
473 		if (optlen < sizeof(int))
474 			return -EINVAL;
475 		inet6_assign_bit(RTALERT_ISOLATE, sk, valbool);
476 		return 0;
477 	case IPV6_MTU_DISCOVER:
478 		if (optlen < sizeof(int))
479 			return -EINVAL;
480 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
481 			return -EINVAL;
482 		WRITE_ONCE(np->pmtudisc, val);
483 		return 0;
484 	case IPV6_FLOWINFO_SEND:
485 		if (optlen < sizeof(int))
486 			return -EINVAL;
487 		inet6_assign_bit(SNDFLOW, sk, valbool);
488 		return 0;
489 	case IPV6_ADDR_PREFERENCES:
490 		if (optlen < sizeof(int))
491 			return -EINVAL;
492 		return ip6_sock_set_addr_preferences(sk, val);
493 	case IPV6_MULTICAST_IF:
494 		if (sk->sk_type == SOCK_STREAM)
495 			return -ENOPROTOOPT;
496 		if (optlen < sizeof(int))
497 			return -EINVAL;
498 		if (val) {
499 			struct net_device *dev;
500 			int bound_dev_if, midx;
501 
502 			rcu_read_lock();
503 
504 			dev = dev_get_by_index_rcu(net, val);
505 			if (!dev) {
506 				rcu_read_unlock();
507 				return -ENODEV;
508 			}
509 			midx = l3mdev_master_ifindex_rcu(dev);
510 
511 			rcu_read_unlock();
512 
513 			bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
514 			if (bound_dev_if &&
515 			    bound_dev_if != val &&
516 			    (!midx || midx != bound_dev_if))
517 				return -EINVAL;
518 		}
519 		WRITE_ONCE(np->mcast_oif, val);
520 		return 0;
521 	case IPV6_UNICAST_IF:
522 	{
523 		struct net_device *dev;
524 		int ifindex;
525 
526 		if (optlen != sizeof(int))
527 			return -EINVAL;
528 
529 		ifindex = (__force int)ntohl((__force __be32)val);
530 		if (!ifindex) {
531 			WRITE_ONCE(np->ucast_oif, 0);
532 			return 0;
533 		}
534 
535 		dev = dev_get_by_index(net, ifindex);
536 		if (!dev)
537 			return -EADDRNOTAVAIL;
538 		dev_put(dev);
539 
540 		if (READ_ONCE(sk->sk_bound_dev_if))
541 			return -EINVAL;
542 
543 		WRITE_ONCE(np->ucast_oif, ifindex);
544 		return 0;
545 	}
546 	}
547 
548 	sockopt_lock_sock(sk);
549 
550 	/* Another thread has converted the socket into IPv4 with
551 	 * IPV6_ADDRFORM concurrently.
552 	 */
553 	if (unlikely(sk->sk_family != AF_INET6))
554 		goto unlock;
555 
556 	switch (optname) {
557 
558 	case IPV6_ADDRFORM:
559 		if (optlen < sizeof(int))
560 			goto e_inval;
561 		if (val == PF_INET) {
562 			if (sk->sk_type == SOCK_RAW)
563 				break;
564 
565 			if (sk->sk_protocol == IPPROTO_UDP) {
566 				if (udp_sk(sk)->pending == AF_INET6) {
567 					retv = -EBUSY;
568 					break;
569 				}
570 			} else if (sk->sk_protocol == IPPROTO_TCP) {
571 				if (sk->sk_prot != &tcpv6_prot) {
572 					retv = -EBUSY;
573 					break;
574 				}
575 			} else {
576 				break;
577 			}
578 
579 			if (sk->sk_state != TCP_ESTABLISHED) {
580 				retv = -ENOTCONN;
581 				break;
582 			}
583 
584 			if (ipv6_only_sock(sk) ||
585 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
586 				retv = -EADDRNOTAVAIL;
587 				break;
588 			}
589 
590 			__ipv6_sock_mc_close(sk);
591 			__ipv6_sock_ac_close(sk);
592 
593 			if (sk->sk_protocol == IPPROTO_TCP) {
594 				struct inet_connection_sock *icsk = inet_csk(sk);
595 
596 				sock_prot_inuse_add(net, sk->sk_prot, -1);
597 				sock_prot_inuse_add(net, &tcp_prot, 1);
598 
599 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
600 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
601 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
602 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
603 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
604 				WRITE_ONCE(sk->sk_family, PF_INET);
605 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
606 			} else {
607 				sock_prot_inuse_add(net, sk->sk_prot, -1);
608 				sock_prot_inuse_add(net, &udp_prot, 1);
609 
610 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
611 				WRITE_ONCE(sk->sk_prot, &udp_prot);
612 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
613 				WRITE_ONCE(sk->sk_family, PF_INET);
614 			}
615 
616 			/* Disable all options not to allocate memory anymore,
617 			 * but there is still a race.  See the lockless path
618 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
619 			 */
620 			np->rxopt.all = 0;
621 
622 			inet6_cleanup_sock(sk);
623 
624 			module_put(THIS_MODULE);
625 			retv = 0;
626 			break;
627 		}
628 		goto e_inval;
629 
630 	case IPV6_V6ONLY:
631 		if (optlen < sizeof(int) ||
632 		    inet_sk(sk)->inet_num)
633 			goto e_inval;
634 		sk->sk_ipv6only = valbool;
635 		retv = 0;
636 		break;
637 
638 	case IPV6_RECVPKTINFO:
639 		if (optlen < sizeof(int))
640 			goto e_inval;
641 		np->rxopt.bits.rxinfo = valbool;
642 		retv = 0;
643 		break;
644 
645 	case IPV6_2292PKTINFO:
646 		if (optlen < sizeof(int))
647 			goto e_inval;
648 		np->rxopt.bits.rxoinfo = valbool;
649 		retv = 0;
650 		break;
651 
652 	case IPV6_RECVHOPLIMIT:
653 		if (optlen < sizeof(int))
654 			goto e_inval;
655 		np->rxopt.bits.rxhlim = valbool;
656 		retv = 0;
657 		break;
658 
659 	case IPV6_2292HOPLIMIT:
660 		if (optlen < sizeof(int))
661 			goto e_inval;
662 		np->rxopt.bits.rxohlim = valbool;
663 		retv = 0;
664 		break;
665 
666 	case IPV6_RECVRTHDR:
667 		if (optlen < sizeof(int))
668 			goto e_inval;
669 		np->rxopt.bits.srcrt = valbool;
670 		retv = 0;
671 		break;
672 
673 	case IPV6_2292RTHDR:
674 		if (optlen < sizeof(int))
675 			goto e_inval;
676 		np->rxopt.bits.osrcrt = valbool;
677 		retv = 0;
678 		break;
679 
680 	case IPV6_RECVHOPOPTS:
681 		if (optlen < sizeof(int))
682 			goto e_inval;
683 		np->rxopt.bits.hopopts = valbool;
684 		retv = 0;
685 		break;
686 
687 	case IPV6_2292HOPOPTS:
688 		if (optlen < sizeof(int))
689 			goto e_inval;
690 		np->rxopt.bits.ohopopts = valbool;
691 		retv = 0;
692 		break;
693 
694 	case IPV6_RECVDSTOPTS:
695 		if (optlen < sizeof(int))
696 			goto e_inval;
697 		np->rxopt.bits.dstopts = valbool;
698 		retv = 0;
699 		break;
700 
701 	case IPV6_2292DSTOPTS:
702 		if (optlen < sizeof(int))
703 			goto e_inval;
704 		np->rxopt.bits.odstopts = valbool;
705 		retv = 0;
706 		break;
707 
708 	case IPV6_TCLASS:
709 		if (optlen < sizeof(int))
710 			goto e_inval;
711 		if (val < -1 || val > 0xff)
712 			goto e_inval;
713 		/* RFC 3542, 6.5: default traffic class of 0x0 */
714 		if (val == -1)
715 			val = 0;
716 		if (sk->sk_type == SOCK_STREAM) {
717 			val &= ~INET_ECN_MASK;
718 			val |= np->tclass & INET_ECN_MASK;
719 		}
720 		if (np->tclass != val) {
721 			np->tclass = val;
722 			sk_dst_reset(sk);
723 		}
724 		retv = 0;
725 		break;
726 
727 	case IPV6_RECVTCLASS:
728 		if (optlen < sizeof(int))
729 			goto e_inval;
730 		np->rxopt.bits.rxtclass = valbool;
731 		retv = 0;
732 		break;
733 
734 	case IPV6_FLOWINFO:
735 		if (optlen < sizeof(int))
736 			goto e_inval;
737 		np->rxopt.bits.rxflow = valbool;
738 		retv = 0;
739 		break;
740 
741 	case IPV6_RECVPATHMTU:
742 		if (optlen < sizeof(int))
743 			goto e_inval;
744 		np->rxopt.bits.rxpmtu = valbool;
745 		retv = 0;
746 		break;
747 
748 	case IPV6_TRANSPARENT:
749 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
750 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
751 			retv = -EPERM;
752 			break;
753 		}
754 		if (optlen < sizeof(int))
755 			goto e_inval;
756 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
757 		inet_assign_bit(TRANSPARENT, sk, valbool);
758 		retv = 0;
759 		break;
760 
761 	case IPV6_FREEBIND:
762 		if (optlen < sizeof(int))
763 			goto e_inval;
764 		/* we also don't have a separate freebind bit for IPV6 */
765 		inet_assign_bit(FREEBIND, sk, valbool);
766 		retv = 0;
767 		break;
768 
769 	case IPV6_RECVORIGDSTADDR:
770 		if (optlen < sizeof(int))
771 			goto e_inval;
772 		np->rxopt.bits.rxorigdstaddr = valbool;
773 		retv = 0;
774 		break;
775 
776 	case IPV6_HOPOPTS:
777 	case IPV6_RTHDRDSTOPTS:
778 	case IPV6_RTHDR:
779 	case IPV6_DSTOPTS:
780 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
781 		break;
782 
783 	case IPV6_PKTINFO:
784 	{
785 		struct in6_pktinfo pkt;
786 
787 		if (optlen == 0)
788 			goto e_inval;
789 		else if (optlen < sizeof(struct in6_pktinfo) ||
790 			 sockptr_is_null(optval))
791 			goto e_inval;
792 
793 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
794 			retv = -EFAULT;
795 			break;
796 		}
797 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
798 			goto e_inval;
799 
800 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
801 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
802 		retv = 0;
803 		break;
804 	}
805 
806 	case IPV6_2292PKTOPTIONS:
807 	{
808 		struct ipv6_txoptions *opt = NULL;
809 		struct msghdr msg;
810 		struct flowi6 fl6;
811 		struct ipcm6_cookie ipc6;
812 
813 		memset(&fl6, 0, sizeof(fl6));
814 		fl6.flowi6_oif = sk->sk_bound_dev_if;
815 		fl6.flowi6_mark = sk->sk_mark;
816 
817 		if (optlen == 0)
818 			goto update;
819 
820 		/* 1K is probably excessive
821 		 * 1K is surely not enough, 2K per standard header is 16K.
822 		 */
823 		retv = -EINVAL;
824 		if (optlen > 64*1024)
825 			break;
826 
827 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
828 		retv = -ENOBUFS;
829 		if (!opt)
830 			break;
831 
832 		memset(opt, 0, sizeof(*opt));
833 		refcount_set(&opt->refcnt, 1);
834 		opt->tot_len = sizeof(*opt) + optlen;
835 		retv = -EFAULT;
836 		if (copy_from_sockptr(opt + 1, optval, optlen))
837 			goto done;
838 
839 		msg.msg_controllen = optlen;
840 		msg.msg_control_is_user = false;
841 		msg.msg_control = (void *)(opt+1);
842 		ipc6.opt = opt;
843 
844 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
845 		if (retv)
846 			goto done;
847 update:
848 		retv = 0;
849 		opt = ipv6_update_options(sk, opt);
850 done:
851 		if (opt) {
852 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
853 			txopt_put(opt);
854 		}
855 		break;
856 	}
857 
858 	case IPV6_ADD_MEMBERSHIP:
859 	case IPV6_DROP_MEMBERSHIP:
860 	{
861 		struct ipv6_mreq mreq;
862 
863 		if (optlen < sizeof(struct ipv6_mreq))
864 			goto e_inval;
865 
866 		retv = -EPROTO;
867 		if (inet_test_bit(IS_ICSK, sk))
868 			break;
869 
870 		retv = -EFAULT;
871 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
872 			break;
873 
874 		if (optname == IPV6_ADD_MEMBERSHIP)
875 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
876 		else
877 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
878 		break;
879 	}
880 	case IPV6_JOIN_ANYCAST:
881 	case IPV6_LEAVE_ANYCAST:
882 	{
883 		struct ipv6_mreq mreq;
884 
885 		if (optlen < sizeof(struct ipv6_mreq))
886 			goto e_inval;
887 
888 		retv = -EFAULT;
889 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
890 			break;
891 
892 		if (optname == IPV6_JOIN_ANYCAST)
893 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
894 		else
895 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
896 		break;
897 	}
898 	case MCAST_JOIN_GROUP:
899 	case MCAST_LEAVE_GROUP:
900 		if (in_compat_syscall())
901 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
902 							    optlen);
903 		else
904 			retv = ipv6_mcast_join_leave(sk, optname, optval,
905 						     optlen);
906 		break;
907 	case MCAST_JOIN_SOURCE_GROUP:
908 	case MCAST_LEAVE_SOURCE_GROUP:
909 	case MCAST_BLOCK_SOURCE:
910 	case MCAST_UNBLOCK_SOURCE:
911 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
912 		break;
913 	case MCAST_MSFILTER:
914 		if (in_compat_syscall())
915 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
916 							      optlen);
917 		else
918 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
919 		break;
920 	case IPV6_ROUTER_ALERT:
921 		if (optlen < sizeof(int))
922 			goto e_inval;
923 		retv = ip6_ra_control(sk, val);
924 		if (retv == 0)
925 			inet6_assign_bit(RTALERT, sk, valbool);
926 		break;
927 	case IPV6_FLOWLABEL_MGR:
928 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
929 		break;
930 	case IPV6_IPSEC_POLICY:
931 	case IPV6_XFRM_POLICY:
932 		retv = -EPERM;
933 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
934 			break;
935 		retv = xfrm_user_policy(sk, optname, optval, optlen);
936 		break;
937 
938 	case IPV6_RECVFRAGSIZE:
939 		np->rxopt.bits.recvfragsize = valbool;
940 		retv = 0;
941 		break;
942 	}
943 
944 unlock:
945 	sockopt_release_sock(sk);
946 
947 	return retv;
948 
949 e_inval:
950 	retv = -EINVAL;
951 	goto unlock;
952 }
953 
954 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
955 		    unsigned int optlen)
956 {
957 	int err;
958 
959 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
960 		return ip_setsockopt(sk, level, optname, optval, optlen);
961 
962 	if (level != SOL_IPV6)
963 		return -ENOPROTOOPT;
964 
965 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
966 #ifdef CONFIG_NETFILTER
967 	/* we need to exclude all possible ENOPROTOOPTs except default case */
968 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
969 			optname != IPV6_XFRM_POLICY)
970 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
971 #endif
972 	return err;
973 }
974 EXPORT_SYMBOL(ipv6_setsockopt);
975 
976 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
977 				  int optname, sockptr_t optval, int len)
978 {
979 	struct ipv6_opt_hdr *hdr;
980 
981 	if (!opt)
982 		return 0;
983 
984 	switch (optname) {
985 	case IPV6_HOPOPTS:
986 		hdr = opt->hopopt;
987 		break;
988 	case IPV6_RTHDRDSTOPTS:
989 		hdr = opt->dst0opt;
990 		break;
991 	case IPV6_RTHDR:
992 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
993 		break;
994 	case IPV6_DSTOPTS:
995 		hdr = opt->dst1opt;
996 		break;
997 	default:
998 		return -EINVAL;	/* should not happen */
999 	}
1000 
1001 	if (!hdr)
1002 		return 0;
1003 
1004 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1005 	if (copy_to_sockptr(optval, hdr, len))
1006 		return -EFAULT;
1007 	return len;
1008 }
1009 
1010 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1011 			     sockptr_t optlen, int len)
1012 {
1013 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1014 	struct group_filter gsf;
1015 	int num;
1016 	int err;
1017 
1018 	if (len < size0)
1019 		return -EINVAL;
1020 	if (copy_from_sockptr(&gsf, optval, size0))
1021 		return -EFAULT;
1022 	if (gsf.gf_group.ss_family != AF_INET6)
1023 		return -EADDRNOTAVAIL;
1024 	num = gsf.gf_numsrc;
1025 	sockopt_lock_sock(sk);
1026 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1027 	if (!err) {
1028 		if (num > gsf.gf_numsrc)
1029 			num = gsf.gf_numsrc;
1030 		len = GROUP_FILTER_SIZE(num);
1031 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1032 		    copy_to_sockptr(optval, &gsf, size0))
1033 			err = -EFAULT;
1034 	}
1035 	sockopt_release_sock(sk);
1036 	return err;
1037 }
1038 
1039 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1040 				    sockptr_t optlen, int len)
1041 {
1042 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1043 	struct compat_group_filter gf32;
1044 	struct group_filter gf;
1045 	int err;
1046 	int num;
1047 
1048 	if (len < size0)
1049 		return -EINVAL;
1050 
1051 	if (copy_from_sockptr(&gf32, optval, size0))
1052 		return -EFAULT;
1053 	gf.gf_interface = gf32.gf_interface;
1054 	gf.gf_fmode = gf32.gf_fmode;
1055 	num = gf.gf_numsrc = gf32.gf_numsrc;
1056 	gf.gf_group = gf32.gf_group;
1057 
1058 	if (gf.gf_group.ss_family != AF_INET6)
1059 		return -EADDRNOTAVAIL;
1060 
1061 	sockopt_lock_sock(sk);
1062 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1063 	sockopt_release_sock(sk);
1064 	if (err)
1065 		return err;
1066 	if (num > gf.gf_numsrc)
1067 		num = gf.gf_numsrc;
1068 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1069 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1070 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1071 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1072 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1073 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1074 		return -EFAULT;
1075 	return 0;
1076 }
1077 
1078 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1079 		       sockptr_t optval, sockptr_t optlen)
1080 {
1081 	struct ipv6_pinfo *np = inet6_sk(sk);
1082 	int len;
1083 	int val;
1084 
1085 	if (ip6_mroute_opt(optname))
1086 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1087 
1088 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1089 		return -EFAULT;
1090 	switch (optname) {
1091 	case IPV6_ADDRFORM:
1092 		if (sk->sk_protocol != IPPROTO_UDP &&
1093 		    sk->sk_protocol != IPPROTO_TCP)
1094 			return -ENOPROTOOPT;
1095 		if (sk->sk_state != TCP_ESTABLISHED)
1096 			return -ENOTCONN;
1097 		val = sk->sk_family;
1098 		break;
1099 	case MCAST_MSFILTER:
1100 		if (in_compat_syscall())
1101 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1102 		return ipv6_get_msfilter(sk, optval, optlen, len);
1103 	case IPV6_2292PKTOPTIONS:
1104 	{
1105 		struct msghdr msg;
1106 		struct sk_buff *skb;
1107 
1108 		if (sk->sk_type != SOCK_STREAM)
1109 			return -ENOPROTOOPT;
1110 
1111 		if (optval.is_kernel) {
1112 			msg.msg_control_is_user = false;
1113 			msg.msg_control = optval.kernel;
1114 		} else {
1115 			msg.msg_control_is_user = true;
1116 			msg.msg_control_user = optval.user;
1117 		}
1118 		msg.msg_controllen = len;
1119 		msg.msg_flags = 0;
1120 
1121 		sockopt_lock_sock(sk);
1122 		skb = np->pktoptions;
1123 		if (skb)
1124 			ip6_datagram_recv_ctl(sk, &msg, skb);
1125 		sockopt_release_sock(sk);
1126 		if (!skb) {
1127 			if (np->rxopt.bits.rxinfo) {
1128 				int mcast_oif = READ_ONCE(np->mcast_oif);
1129 				struct in6_pktinfo src_info;
1130 
1131 				src_info.ipi6_ifindex = mcast_oif ? :
1132 					np->sticky_pktinfo.ipi6_ifindex;
1133 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1134 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1135 			}
1136 			if (np->rxopt.bits.rxhlim) {
1137 				int hlim = READ_ONCE(np->mcast_hops);
1138 
1139 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1140 			}
1141 			if (np->rxopt.bits.rxtclass) {
1142 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1143 
1144 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1145 			}
1146 			if (np->rxopt.bits.rxoinfo) {
1147 				int mcast_oif = READ_ONCE(np->mcast_oif);
1148 				struct in6_pktinfo src_info;
1149 
1150 				src_info.ipi6_ifindex = mcast_oif ? :
1151 					np->sticky_pktinfo.ipi6_ifindex;
1152 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr :
1153 								 np->sticky_pktinfo.ipi6_addr;
1154 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1155 			}
1156 			if (np->rxopt.bits.rxohlim) {
1157 				int hlim = READ_ONCE(np->mcast_hops);
1158 
1159 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1160 			}
1161 			if (np->rxopt.bits.rxflow) {
1162 				__be32 flowinfo = np->rcv_flowinfo;
1163 
1164 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1165 			}
1166 		}
1167 		len -= msg.msg_controllen;
1168 		return copy_to_sockptr(optlen, &len, sizeof(int));
1169 	}
1170 	case IPV6_MTU:
1171 	{
1172 		struct dst_entry *dst;
1173 
1174 		val = 0;
1175 		rcu_read_lock();
1176 		dst = __sk_dst_get(sk);
1177 		if (dst)
1178 			val = dst6_mtu(dst);
1179 		rcu_read_unlock();
1180 		if (!val)
1181 			return -ENOTCONN;
1182 		break;
1183 	}
1184 
1185 	case IPV6_V6ONLY:
1186 		val = sk->sk_ipv6only;
1187 		break;
1188 
1189 	case IPV6_RECVPKTINFO:
1190 		val = np->rxopt.bits.rxinfo;
1191 		break;
1192 
1193 	case IPV6_2292PKTINFO:
1194 		val = np->rxopt.bits.rxoinfo;
1195 		break;
1196 
1197 	case IPV6_RECVHOPLIMIT:
1198 		val = np->rxopt.bits.rxhlim;
1199 		break;
1200 
1201 	case IPV6_2292HOPLIMIT:
1202 		val = np->rxopt.bits.rxohlim;
1203 		break;
1204 
1205 	case IPV6_RECVRTHDR:
1206 		val = np->rxopt.bits.srcrt;
1207 		break;
1208 
1209 	case IPV6_2292RTHDR:
1210 		val = np->rxopt.bits.osrcrt;
1211 		break;
1212 
1213 	case IPV6_HOPOPTS:
1214 	case IPV6_RTHDRDSTOPTS:
1215 	case IPV6_RTHDR:
1216 	case IPV6_DSTOPTS:
1217 	{
1218 		struct ipv6_txoptions *opt;
1219 
1220 		sockopt_lock_sock(sk);
1221 		opt = rcu_dereference_protected(np->opt,
1222 						lockdep_sock_is_held(sk));
1223 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1224 		sockopt_release_sock(sk);
1225 		/* check if ipv6_getsockopt_sticky() returns err code */
1226 		if (len < 0)
1227 			return len;
1228 		return copy_to_sockptr(optlen, &len, sizeof(int));
1229 	}
1230 
1231 	case IPV6_RECVHOPOPTS:
1232 		val = np->rxopt.bits.hopopts;
1233 		break;
1234 
1235 	case IPV6_2292HOPOPTS:
1236 		val = np->rxopt.bits.ohopopts;
1237 		break;
1238 
1239 	case IPV6_RECVDSTOPTS:
1240 		val = np->rxopt.bits.dstopts;
1241 		break;
1242 
1243 	case IPV6_2292DSTOPTS:
1244 		val = np->rxopt.bits.odstopts;
1245 		break;
1246 
1247 	case IPV6_TCLASS:
1248 		val = np->tclass;
1249 		break;
1250 
1251 	case IPV6_RECVTCLASS:
1252 		val = np->rxopt.bits.rxtclass;
1253 		break;
1254 
1255 	case IPV6_FLOWINFO:
1256 		val = np->rxopt.bits.rxflow;
1257 		break;
1258 
1259 	case IPV6_RECVPATHMTU:
1260 		val = np->rxopt.bits.rxpmtu;
1261 		break;
1262 
1263 	case IPV6_PATHMTU:
1264 	{
1265 		struct dst_entry *dst;
1266 		struct ip6_mtuinfo mtuinfo;
1267 
1268 		if (len < sizeof(mtuinfo))
1269 			return -EINVAL;
1270 
1271 		len = sizeof(mtuinfo);
1272 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1273 
1274 		rcu_read_lock();
1275 		dst = __sk_dst_get(sk);
1276 		if (dst)
1277 			mtuinfo.ip6m_mtu = dst6_mtu(dst);
1278 		rcu_read_unlock();
1279 		if (!mtuinfo.ip6m_mtu)
1280 			return -ENOTCONN;
1281 
1282 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1283 			return -EFAULT;
1284 		if (copy_to_sockptr(optval, &mtuinfo, len))
1285 			return -EFAULT;
1286 
1287 		return 0;
1288 	}
1289 
1290 	case IPV6_TRANSPARENT:
1291 		val = inet_test_bit(TRANSPARENT, sk);
1292 		break;
1293 
1294 	case IPV6_FREEBIND:
1295 		val = inet_test_bit(FREEBIND, sk);
1296 		break;
1297 
1298 	case IPV6_RECVORIGDSTADDR:
1299 		val = np->rxopt.bits.rxorigdstaddr;
1300 		break;
1301 
1302 	case IPV6_UNICAST_HOPS:
1303 	case IPV6_MULTICAST_HOPS:
1304 	{
1305 		struct dst_entry *dst;
1306 
1307 		if (optname == IPV6_UNICAST_HOPS)
1308 			val = READ_ONCE(np->hop_limit);
1309 		else
1310 			val = READ_ONCE(np->mcast_hops);
1311 
1312 		if (val < 0) {
1313 			rcu_read_lock();
1314 			dst = __sk_dst_get(sk);
1315 			if (dst)
1316 				val = ip6_dst_hoplimit(dst);
1317 			rcu_read_unlock();
1318 		}
1319 
1320 		if (val < 0)
1321 			val = READ_ONCE(sock_net(sk)->ipv6.devconf_all->hop_limit);
1322 		break;
1323 	}
1324 
1325 	case IPV6_MULTICAST_LOOP:
1326 		val = inet6_test_bit(MC6_LOOP, sk);
1327 		break;
1328 
1329 	case IPV6_MULTICAST_IF:
1330 		val = READ_ONCE(np->mcast_oif);
1331 		break;
1332 
1333 	case IPV6_MULTICAST_ALL:
1334 		val = inet6_test_bit(MC6_ALL, sk);
1335 		break;
1336 
1337 	case IPV6_UNICAST_IF:
1338 		val = (__force int)htonl((__u32) READ_ONCE(np->ucast_oif));
1339 		break;
1340 
1341 	case IPV6_MTU_DISCOVER:
1342 		val = READ_ONCE(np->pmtudisc);
1343 		break;
1344 
1345 	case IPV6_RECVERR:
1346 		val = inet6_test_bit(RECVERR6, sk);
1347 		break;
1348 
1349 	case IPV6_FLOWINFO_SEND:
1350 		val = inet6_test_bit(SNDFLOW, sk);
1351 		break;
1352 
1353 	case IPV6_FLOWLABEL_MGR:
1354 	{
1355 		struct in6_flowlabel_req freq;
1356 		int flags;
1357 
1358 		if (len < sizeof(freq))
1359 			return -EINVAL;
1360 
1361 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1362 			return -EFAULT;
1363 
1364 		if (freq.flr_action != IPV6_FL_A_GET)
1365 			return -EINVAL;
1366 
1367 		len = sizeof(freq);
1368 		flags = freq.flr_flags;
1369 
1370 		memset(&freq, 0, sizeof(freq));
1371 
1372 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1373 		if (val < 0)
1374 			return val;
1375 
1376 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1377 			return -EFAULT;
1378 		if (copy_to_sockptr(optval, &freq, len))
1379 			return -EFAULT;
1380 
1381 		return 0;
1382 	}
1383 
1384 	case IPV6_ADDR_PREFERENCES:
1385 		{
1386 		u8 srcprefs = READ_ONCE(np->srcprefs);
1387 		val = 0;
1388 
1389 		if (srcprefs & IPV6_PREFER_SRC_TMP)
1390 			val |= IPV6_PREFER_SRC_TMP;
1391 		else if (srcprefs & IPV6_PREFER_SRC_PUBLIC)
1392 			val |= IPV6_PREFER_SRC_PUBLIC;
1393 		else {
1394 			/* XXX: should we return system default? */
1395 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1396 		}
1397 
1398 		if (srcprefs & IPV6_PREFER_SRC_COA)
1399 			val |= IPV6_PREFER_SRC_COA;
1400 		else
1401 			val |= IPV6_PREFER_SRC_HOME;
1402 		break;
1403 		}
1404 	case IPV6_MINHOPCOUNT:
1405 		val = READ_ONCE(np->min_hopcount);
1406 		break;
1407 
1408 	case IPV6_DONTFRAG:
1409 		val = inet6_test_bit(DONTFRAG, sk);
1410 		break;
1411 
1412 	case IPV6_AUTOFLOWLABEL:
1413 		val = ip6_autoflowlabel(sock_net(sk), sk);
1414 		break;
1415 
1416 	case IPV6_RECVFRAGSIZE:
1417 		val = np->rxopt.bits.recvfragsize;
1418 		break;
1419 
1420 	case IPV6_ROUTER_ALERT:
1421 		val = inet6_test_bit(RTALERT, sk);
1422 		break;
1423 
1424 	case IPV6_ROUTER_ALERT_ISOLATE:
1425 		val = inet6_test_bit(RTALERT_ISOLATE, sk);
1426 		break;
1427 
1428 	case IPV6_RECVERR_RFC4884:
1429 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1430 		break;
1431 
1432 	default:
1433 		return -ENOPROTOOPT;
1434 	}
1435 	len = min_t(unsigned int, sizeof(int), len);
1436 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1437 		return -EFAULT;
1438 	if (copy_to_sockptr(optval, &val, len))
1439 		return -EFAULT;
1440 	return 0;
1441 }
1442 
1443 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1444 		    char __user *optval, int __user *optlen)
1445 {
1446 	int err;
1447 
1448 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1449 		return ip_getsockopt(sk, level, optname, optval, optlen);
1450 
1451 	if (level != SOL_IPV6)
1452 		return -ENOPROTOOPT;
1453 
1454 	err = do_ipv6_getsockopt(sk, level, optname,
1455 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1456 #ifdef CONFIG_NETFILTER
1457 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1458 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1459 		int len;
1460 
1461 		if (get_user(len, optlen))
1462 			return -EFAULT;
1463 
1464 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1465 		if (err >= 0)
1466 			err = put_user(len, optlen);
1467 	}
1468 #endif
1469 	return err;
1470 }
1471 EXPORT_SYMBOL(ipv6_getsockopt);
1472