xref: /linux/net/ipv6/ipv6_sockglue.c (revision f7fdf13bf103bbe8885f722d63fd9377d034d35f)
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 = unrcu_pointer(xchg(&inet6_sk(sk)->opt, RCU_INITIALIZER(opt)));
115 	sk_dst_reset(sk);
116 
117 	return opt;
118 }
119 
120 static bool setsockopt_needs_rtnl(int optname)
121 {
122 	switch (optname) {
123 	case IPV6_JOIN_ANYCAST:
124 		return true;
125 	}
126 	return false;
127 }
128 
129 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
130 		sockptr_t optval, int optlen)
131 {
132 	if (in_compat_syscall()) {
133 		struct compat_group_source_req gr32;
134 
135 		if (optlen < sizeof(gr32))
136 			return -EINVAL;
137 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
138 			return -EFAULT;
139 		greqs->gsr_interface = gr32.gsr_interface;
140 		greqs->gsr_group = gr32.gsr_group;
141 		greqs->gsr_source = gr32.gsr_source;
142 	} else {
143 		if (optlen < sizeof(*greqs))
144 			return -EINVAL;
145 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
146 			return -EFAULT;
147 	}
148 
149 	return 0;
150 }
151 
152 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
153 		sockptr_t optval, int optlen)
154 {
155 	struct group_source_req greqs;
156 	int omode, add;
157 	int ret;
158 
159 	ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
160 	if (ret)
161 		return ret;
162 
163 	if (greqs.gsr_group.ss_family != AF_INET6 ||
164 	    greqs.gsr_source.ss_family != AF_INET6)
165 		return -EADDRNOTAVAIL;
166 
167 	if (optname == MCAST_BLOCK_SOURCE) {
168 		omode = MCAST_EXCLUDE;
169 		add = 1;
170 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
171 		omode = MCAST_EXCLUDE;
172 		add = 0;
173 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
174 		struct sockaddr_in6 *psin6;
175 		int retv;
176 
177 		psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
178 		retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
179 					     &psin6->sin6_addr,
180 					     MCAST_INCLUDE);
181 		/* prior join w/ different source is ok */
182 		if (retv && retv != -EADDRINUSE)
183 			return retv;
184 		omode = MCAST_INCLUDE;
185 		add = 1;
186 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
187 		omode = MCAST_INCLUDE;
188 		add = 0;
189 	}
190 	return ip6_mc_source(add, omode, sk, &greqs);
191 }
192 
193 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
194 		int optlen)
195 {
196 	struct group_filter *gsf;
197 	int ret;
198 
199 	if (optlen < GROUP_FILTER_SIZE(0))
200 		return -EINVAL;
201 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max))
202 		return -ENOBUFS;
203 
204 	gsf = memdup_sockptr(optval, optlen);
205 	if (IS_ERR(gsf))
206 		return PTR_ERR(gsf);
207 
208 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
209 	ret = -ENOBUFS;
210 	if (gsf->gf_numsrc >= 0x1ffffffU ||
211 	    gsf->gf_numsrc > sysctl_mld_max_msf)
212 		goto out_free_gsf;
213 
214 	ret = -EINVAL;
215 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
216 		goto out_free_gsf;
217 
218 	ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
219 out_free_gsf:
220 	kfree(gsf);
221 	return ret;
222 }
223 
224 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
225 		int optlen)
226 {
227 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
228 	struct compat_group_filter *gf32;
229 	void *p;
230 	int ret;
231 	int n;
232 
233 	if (optlen < size0)
234 		return -EINVAL;
235 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max) - 4)
236 		return -ENOBUFS;
237 
238 	p = kmalloc(optlen + 4, GFP_KERNEL);
239 	if (!p)
240 		return -ENOMEM;
241 
242 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
243 	ret = -EFAULT;
244 	if (copy_from_sockptr(gf32, optval, optlen))
245 		goto out_free_p;
246 
247 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
248 	ret = -ENOBUFS;
249 	n = gf32->gf_numsrc;
250 	if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
251 		goto out_free_p;
252 
253 	ret = -EINVAL;
254 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
255 		goto out_free_p;
256 
257 	ret = ip6_mc_msfilter(sk, &(struct group_filter){
258 			.gf_interface = gf32->gf_interface,
259 			.gf_group = gf32->gf_group,
260 			.gf_fmode = gf32->gf_fmode,
261 			.gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
262 
263 out_free_p:
264 	kfree(p);
265 	return ret;
266 }
267 
268 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
269 		sockptr_t optval, int optlen)
270 {
271 	struct sockaddr_in6 *psin6;
272 	struct group_req greq;
273 
274 	if (optlen < sizeof(greq))
275 		return -EINVAL;
276 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
277 		return -EFAULT;
278 
279 	if (greq.gr_group.ss_family != AF_INET6)
280 		return -EADDRNOTAVAIL;
281 	psin6 = (struct sockaddr_in6 *)&greq.gr_group;
282 	if (optname == MCAST_JOIN_GROUP)
283 		return ipv6_sock_mc_join(sk, greq.gr_interface,
284 					 &psin6->sin6_addr);
285 	return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
286 }
287 
288 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
289 		sockptr_t optval, int optlen)
290 {
291 	struct compat_group_req gr32;
292 	struct sockaddr_in6 *psin6;
293 
294 	if (optlen < sizeof(gr32))
295 		return -EINVAL;
296 	if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
297 		return -EFAULT;
298 
299 	if (gr32.gr_group.ss_family != AF_INET6)
300 		return -EADDRNOTAVAIL;
301 	psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
302 	if (optname == MCAST_JOIN_GROUP)
303 		return ipv6_sock_mc_join(sk, gr32.gr_interface,
304 					&psin6->sin6_addr);
305 	return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
306 }
307 
308 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
309 		int optlen)
310 {
311 	struct ipv6_pinfo *np = inet6_sk(sk);
312 	struct ipv6_opt_hdr *new = NULL;
313 	struct net *net = sock_net(sk);
314 	struct ipv6_txoptions *opt;
315 	int err;
316 
317 	/* hop-by-hop / destination options are privileged option */
318 	if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
319 		return -EPERM;
320 
321 	/* remove any sticky options header with a zero option
322 	 * length, per RFC3542.
323 	 */
324 	if (optlen > 0) {
325 		if (sockptr_is_null(optval))
326 			return -EINVAL;
327 		if (optlen < sizeof(struct ipv6_opt_hdr) ||
328 		    optlen & 0x7 ||
329 		    optlen > 8 * 255)
330 			return -EINVAL;
331 
332 		new = memdup_sockptr(optval, optlen);
333 		if (IS_ERR(new))
334 			return PTR_ERR(new);
335 		if (unlikely(ipv6_optlen(new) > optlen)) {
336 			kfree(new);
337 			return -EINVAL;
338 		}
339 	}
340 
341 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
342 	opt = ipv6_renew_options(sk, opt, optname, new);
343 	kfree(new);
344 	if (IS_ERR(opt))
345 		return PTR_ERR(opt);
346 
347 	/* routing header option needs extra check */
348 	err = -EINVAL;
349 	if (optname == IPV6_RTHDR && opt && opt->srcrt) {
350 		struct ipv6_rt_hdr *rthdr = opt->srcrt;
351 		switch (rthdr->type) {
352 #if IS_ENABLED(CONFIG_IPV6_MIP6)
353 		case IPV6_SRCRT_TYPE_2:
354 			if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
355 				goto sticky_done;
356 			break;
357 #endif
358 		case IPV6_SRCRT_TYPE_4:
359 		{
360 			struct ipv6_sr_hdr *srh =
361 				(struct ipv6_sr_hdr *)opt->srcrt;
362 
363 			if (!seg6_validate_srh(srh, optlen, false))
364 				goto sticky_done;
365 			break;
366 		}
367 		default:
368 			goto sticky_done;
369 		}
370 	}
371 
372 	err = 0;
373 	opt = ipv6_update_options(sk, opt);
374 sticky_done:
375 	if (opt) {
376 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
377 		txopt_put(opt);
378 	}
379 	return err;
380 }
381 
382 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
383 		       sockptr_t optval, unsigned int optlen)
384 {
385 	struct ipv6_pinfo *np = inet6_sk(sk);
386 	struct net *net = sock_net(sk);
387 	int val, valbool;
388 	int retv = -ENOPROTOOPT;
389 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
390 
391 	if (sockptr_is_null(optval))
392 		val = 0;
393 	else {
394 		if (optlen >= sizeof(int)) {
395 			if (copy_from_sockptr(&val, optval, sizeof(val)))
396 				return -EFAULT;
397 		} else
398 			val = 0;
399 	}
400 
401 	valbool = (val != 0);
402 
403 	if (ip6_mroute_opt(optname))
404 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
405 
406 	/* Handle options that can be set without locking the socket. */
407 	switch (optname) {
408 	case IPV6_UNICAST_HOPS:
409 		if (optlen < sizeof(int))
410 			return -EINVAL;
411 		if (val > 255 || val < -1)
412 			return -EINVAL;
413 		WRITE_ONCE(np->hop_limit, val);
414 		return 0;
415 	case IPV6_MULTICAST_LOOP:
416 		if (optlen < sizeof(int))
417 			return -EINVAL;
418 		if (val != valbool)
419 			return -EINVAL;
420 		inet6_assign_bit(MC6_LOOP, sk, valbool);
421 		return 0;
422 	case IPV6_MULTICAST_HOPS:
423 		if (sk->sk_type == SOCK_STREAM)
424 			return retv;
425 		if (optlen < sizeof(int))
426 			return -EINVAL;
427 		if (val > 255 || val < -1)
428 			return -EINVAL;
429 		WRITE_ONCE(np->mcast_hops,
430 			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
431 		return 0;
432 	case IPV6_MTU:
433 		if (optlen < sizeof(int))
434 			return -EINVAL;
435 		if (val && val < IPV6_MIN_MTU)
436 			return -EINVAL;
437 		WRITE_ONCE(np->frag_size, val);
438 		return 0;
439 	case IPV6_MINHOPCOUNT:
440 		if (optlen < sizeof(int))
441 			return -EINVAL;
442 		if (val < 0 || val > 255)
443 			return -EINVAL;
444 
445 		if (val)
446 			static_branch_enable(&ip6_min_hopcount);
447 
448 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
449 		 * while we are changing it.
450 		 */
451 		WRITE_ONCE(np->min_hopcount, val);
452 		return 0;
453 	case IPV6_RECVERR_RFC4884:
454 		if (optlen < sizeof(int))
455 			return -EINVAL;
456 		if (val < 0 || val > 1)
457 			return -EINVAL;
458 		inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
459 		return 0;
460 	case IPV6_MULTICAST_ALL:
461 		if (optlen < sizeof(int))
462 			return -EINVAL;
463 		inet6_assign_bit(MC6_ALL, sk, valbool);
464 		return 0;
465 	case IPV6_AUTOFLOWLABEL:
466 		inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
467 		inet6_set_bit(AUTOFLOWLABEL_SET, sk);
468 		return 0;
469 	case IPV6_DONTFRAG:
470 		inet6_assign_bit(DONTFRAG, sk, valbool);
471 		return 0;
472 	case IPV6_RECVERR:
473 		if (optlen < sizeof(int))
474 			return -EINVAL;
475 		inet6_assign_bit(RECVERR6, sk, valbool);
476 		if (!val)
477 			skb_errqueue_purge(&sk->sk_error_queue);
478 		return 0;
479 	case IPV6_ROUTER_ALERT_ISOLATE:
480 		if (optlen < sizeof(int))
481 			return -EINVAL;
482 		inet6_assign_bit(RTALERT_ISOLATE, sk, valbool);
483 		return 0;
484 	case IPV6_MTU_DISCOVER:
485 		if (optlen < sizeof(int))
486 			return -EINVAL;
487 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
488 			return -EINVAL;
489 		WRITE_ONCE(np->pmtudisc, val);
490 		return 0;
491 	case IPV6_FLOWINFO_SEND:
492 		if (optlen < sizeof(int))
493 			return -EINVAL;
494 		inet6_assign_bit(SNDFLOW, sk, valbool);
495 		return 0;
496 	case IPV6_ADDR_PREFERENCES:
497 		if (optlen < sizeof(int))
498 			return -EINVAL;
499 		return ip6_sock_set_addr_preferences(sk, val);
500 	case IPV6_MULTICAST_IF:
501 		if (sk->sk_type == SOCK_STREAM)
502 			return -ENOPROTOOPT;
503 		if (optlen < sizeof(int))
504 			return -EINVAL;
505 		if (val) {
506 			struct net_device *dev;
507 			int bound_dev_if, midx;
508 
509 			rcu_read_lock();
510 
511 			dev = dev_get_by_index_rcu(net, val);
512 			if (!dev) {
513 				rcu_read_unlock();
514 				return -ENODEV;
515 			}
516 			midx = l3mdev_master_ifindex_rcu(dev);
517 
518 			rcu_read_unlock();
519 
520 			bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
521 			if (bound_dev_if &&
522 			    bound_dev_if != val &&
523 			    (!midx || midx != bound_dev_if))
524 				return -EINVAL;
525 		}
526 		WRITE_ONCE(np->mcast_oif, val);
527 		return 0;
528 	case IPV6_UNICAST_IF:
529 	{
530 		struct net_device *dev;
531 		int ifindex;
532 
533 		if (optlen != sizeof(int))
534 			return -EINVAL;
535 
536 		ifindex = (__force int)ntohl((__force __be32)val);
537 		if (!ifindex) {
538 			WRITE_ONCE(np->ucast_oif, 0);
539 			return 0;
540 		}
541 
542 		dev = dev_get_by_index(net, ifindex);
543 		if (!dev)
544 			return -EADDRNOTAVAIL;
545 		dev_put(dev);
546 
547 		if (READ_ONCE(sk->sk_bound_dev_if))
548 			return -EINVAL;
549 
550 		WRITE_ONCE(np->ucast_oif, ifindex);
551 		return 0;
552 	}
553 	}
554 	if (needs_rtnl)
555 		rtnl_lock();
556 	sockopt_lock_sock(sk);
557 
558 	/* Another thread has converted the socket into IPv4 with
559 	 * IPV6_ADDRFORM concurrently.
560 	 */
561 	if (unlikely(sk->sk_family != AF_INET6))
562 		goto unlock;
563 
564 	switch (optname) {
565 
566 	case IPV6_ADDRFORM:
567 		if (optlen < sizeof(int))
568 			goto e_inval;
569 		if (val == PF_INET) {
570 			if (sk->sk_type == SOCK_RAW)
571 				break;
572 
573 			if (sk->sk_protocol == IPPROTO_UDP ||
574 			    sk->sk_protocol == IPPROTO_UDPLITE) {
575 				struct udp_sock *up = udp_sk(sk);
576 				if (up->pending == AF_INET6) {
577 					retv = -EBUSY;
578 					break;
579 				}
580 			} else if (sk->sk_protocol == IPPROTO_TCP) {
581 				if (sk->sk_prot != &tcpv6_prot) {
582 					retv = -EBUSY;
583 					break;
584 				}
585 			} else {
586 				break;
587 			}
588 
589 			if (sk->sk_state != TCP_ESTABLISHED) {
590 				retv = -ENOTCONN;
591 				break;
592 			}
593 
594 			if (ipv6_only_sock(sk) ||
595 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
596 				retv = -EADDRNOTAVAIL;
597 				break;
598 			}
599 
600 			__ipv6_sock_mc_close(sk);
601 			__ipv6_sock_ac_close(sk);
602 
603 			if (sk->sk_protocol == IPPROTO_TCP) {
604 				struct inet_connection_sock *icsk = inet_csk(sk);
605 
606 				sock_prot_inuse_add(net, sk->sk_prot, -1);
607 				sock_prot_inuse_add(net, &tcp_prot, 1);
608 
609 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
610 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
611 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
612 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
613 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
614 				WRITE_ONCE(sk->sk_family, PF_INET);
615 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
616 			} else {
617 				struct proto *prot = &udp_prot;
618 
619 				if (sk->sk_protocol == IPPROTO_UDPLITE)
620 					prot = &udplite_prot;
621 
622 				sock_prot_inuse_add(net, sk->sk_prot, -1);
623 				sock_prot_inuse_add(net, prot, 1);
624 
625 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
626 				WRITE_ONCE(sk->sk_prot, prot);
627 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
628 				WRITE_ONCE(sk->sk_family, PF_INET);
629 			}
630 
631 			/* Disable all options not to allocate memory anymore,
632 			 * but there is still a race.  See the lockless path
633 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
634 			 */
635 			np->rxopt.all = 0;
636 
637 			inet6_cleanup_sock(sk);
638 
639 			module_put(THIS_MODULE);
640 			retv = 0;
641 			break;
642 		}
643 		goto e_inval;
644 
645 	case IPV6_V6ONLY:
646 		if (optlen < sizeof(int) ||
647 		    inet_sk(sk)->inet_num)
648 			goto e_inval;
649 		sk->sk_ipv6only = valbool;
650 		retv = 0;
651 		break;
652 
653 	case IPV6_RECVPKTINFO:
654 		if (optlen < sizeof(int))
655 			goto e_inval;
656 		np->rxopt.bits.rxinfo = valbool;
657 		retv = 0;
658 		break;
659 
660 	case IPV6_2292PKTINFO:
661 		if (optlen < sizeof(int))
662 			goto e_inval;
663 		np->rxopt.bits.rxoinfo = valbool;
664 		retv = 0;
665 		break;
666 
667 	case IPV6_RECVHOPLIMIT:
668 		if (optlen < sizeof(int))
669 			goto e_inval;
670 		np->rxopt.bits.rxhlim = valbool;
671 		retv = 0;
672 		break;
673 
674 	case IPV6_2292HOPLIMIT:
675 		if (optlen < sizeof(int))
676 			goto e_inval;
677 		np->rxopt.bits.rxohlim = valbool;
678 		retv = 0;
679 		break;
680 
681 	case IPV6_RECVRTHDR:
682 		if (optlen < sizeof(int))
683 			goto e_inval;
684 		np->rxopt.bits.srcrt = valbool;
685 		retv = 0;
686 		break;
687 
688 	case IPV6_2292RTHDR:
689 		if (optlen < sizeof(int))
690 			goto e_inval;
691 		np->rxopt.bits.osrcrt = valbool;
692 		retv = 0;
693 		break;
694 
695 	case IPV6_RECVHOPOPTS:
696 		if (optlen < sizeof(int))
697 			goto e_inval;
698 		np->rxopt.bits.hopopts = valbool;
699 		retv = 0;
700 		break;
701 
702 	case IPV6_2292HOPOPTS:
703 		if (optlen < sizeof(int))
704 			goto e_inval;
705 		np->rxopt.bits.ohopopts = valbool;
706 		retv = 0;
707 		break;
708 
709 	case IPV6_RECVDSTOPTS:
710 		if (optlen < sizeof(int))
711 			goto e_inval;
712 		np->rxopt.bits.dstopts = valbool;
713 		retv = 0;
714 		break;
715 
716 	case IPV6_2292DSTOPTS:
717 		if (optlen < sizeof(int))
718 			goto e_inval;
719 		np->rxopt.bits.odstopts = valbool;
720 		retv = 0;
721 		break;
722 
723 	case IPV6_TCLASS:
724 		if (optlen < sizeof(int))
725 			goto e_inval;
726 		if (val < -1 || val > 0xff)
727 			goto e_inval;
728 		/* RFC 3542, 6.5: default traffic class of 0x0 */
729 		if (val == -1)
730 			val = 0;
731 		if (sk->sk_type == SOCK_STREAM) {
732 			val &= ~INET_ECN_MASK;
733 			val |= np->tclass & INET_ECN_MASK;
734 		}
735 		if (np->tclass != val) {
736 			np->tclass = val;
737 			sk_dst_reset(sk);
738 		}
739 		retv = 0;
740 		break;
741 
742 	case IPV6_RECVTCLASS:
743 		if (optlen < sizeof(int))
744 			goto e_inval;
745 		np->rxopt.bits.rxtclass = valbool;
746 		retv = 0;
747 		break;
748 
749 	case IPV6_FLOWINFO:
750 		if (optlen < sizeof(int))
751 			goto e_inval;
752 		np->rxopt.bits.rxflow = valbool;
753 		retv = 0;
754 		break;
755 
756 	case IPV6_RECVPATHMTU:
757 		if (optlen < sizeof(int))
758 			goto e_inval;
759 		np->rxopt.bits.rxpmtu = valbool;
760 		retv = 0;
761 		break;
762 
763 	case IPV6_TRANSPARENT:
764 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
765 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
766 			retv = -EPERM;
767 			break;
768 		}
769 		if (optlen < sizeof(int))
770 			goto e_inval;
771 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
772 		inet_assign_bit(TRANSPARENT, sk, valbool);
773 		retv = 0;
774 		break;
775 
776 	case IPV6_FREEBIND:
777 		if (optlen < sizeof(int))
778 			goto e_inval;
779 		/* we also don't have a separate freebind bit for IPV6 */
780 		inet_assign_bit(FREEBIND, sk, valbool);
781 		retv = 0;
782 		break;
783 
784 	case IPV6_RECVORIGDSTADDR:
785 		if (optlen < sizeof(int))
786 			goto e_inval;
787 		np->rxopt.bits.rxorigdstaddr = valbool;
788 		retv = 0;
789 		break;
790 
791 	case IPV6_HOPOPTS:
792 	case IPV6_RTHDRDSTOPTS:
793 	case IPV6_RTHDR:
794 	case IPV6_DSTOPTS:
795 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
796 		break;
797 
798 	case IPV6_PKTINFO:
799 	{
800 		struct in6_pktinfo pkt;
801 
802 		if (optlen == 0)
803 			goto e_inval;
804 		else if (optlen < sizeof(struct in6_pktinfo) ||
805 			 sockptr_is_null(optval))
806 			goto e_inval;
807 
808 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
809 			retv = -EFAULT;
810 			break;
811 		}
812 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
813 			goto e_inval;
814 
815 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
816 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
817 		retv = 0;
818 		break;
819 	}
820 
821 	case IPV6_2292PKTOPTIONS:
822 	{
823 		struct ipv6_txoptions *opt = NULL;
824 		struct msghdr msg;
825 		struct flowi6 fl6;
826 		struct ipcm6_cookie ipc6;
827 
828 		memset(&fl6, 0, sizeof(fl6));
829 		fl6.flowi6_oif = sk->sk_bound_dev_if;
830 		fl6.flowi6_mark = sk->sk_mark;
831 
832 		if (optlen == 0)
833 			goto update;
834 
835 		/* 1K is probably excessive
836 		 * 1K is surely not enough, 2K per standard header is 16K.
837 		 */
838 		retv = -EINVAL;
839 		if (optlen > 64*1024)
840 			break;
841 
842 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
843 		retv = -ENOBUFS;
844 		if (!opt)
845 			break;
846 
847 		memset(opt, 0, sizeof(*opt));
848 		refcount_set(&opt->refcnt, 1);
849 		opt->tot_len = sizeof(*opt) + optlen;
850 		retv = -EFAULT;
851 		if (copy_from_sockptr(opt + 1, optval, optlen))
852 			goto done;
853 
854 		msg.msg_controllen = optlen;
855 		msg.msg_control_is_user = false;
856 		msg.msg_control = (void *)(opt+1);
857 		ipc6.opt = opt;
858 
859 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
860 		if (retv)
861 			goto done;
862 update:
863 		retv = 0;
864 		opt = ipv6_update_options(sk, opt);
865 done:
866 		if (opt) {
867 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
868 			txopt_put(opt);
869 		}
870 		break;
871 	}
872 
873 	case IPV6_ADD_MEMBERSHIP:
874 	case IPV6_DROP_MEMBERSHIP:
875 	{
876 		struct ipv6_mreq mreq;
877 
878 		if (optlen < sizeof(struct ipv6_mreq))
879 			goto e_inval;
880 
881 		retv = -EPROTO;
882 		if (inet_test_bit(IS_ICSK, sk))
883 			break;
884 
885 		retv = -EFAULT;
886 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
887 			break;
888 
889 		if (optname == IPV6_ADD_MEMBERSHIP)
890 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
891 		else
892 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
893 		break;
894 	}
895 	case IPV6_JOIN_ANYCAST:
896 	case IPV6_LEAVE_ANYCAST:
897 	{
898 		struct ipv6_mreq mreq;
899 
900 		if (optlen < sizeof(struct ipv6_mreq))
901 			goto e_inval;
902 
903 		retv = -EFAULT;
904 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
905 			break;
906 
907 		if (optname == IPV6_JOIN_ANYCAST)
908 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
909 		else
910 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
911 		break;
912 	}
913 	case MCAST_JOIN_GROUP:
914 	case MCAST_LEAVE_GROUP:
915 		if (in_compat_syscall())
916 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
917 							    optlen);
918 		else
919 			retv = ipv6_mcast_join_leave(sk, optname, optval,
920 						     optlen);
921 		break;
922 	case MCAST_JOIN_SOURCE_GROUP:
923 	case MCAST_LEAVE_SOURCE_GROUP:
924 	case MCAST_BLOCK_SOURCE:
925 	case MCAST_UNBLOCK_SOURCE:
926 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
927 		break;
928 	case MCAST_MSFILTER:
929 		if (in_compat_syscall())
930 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
931 							      optlen);
932 		else
933 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
934 		break;
935 	case IPV6_ROUTER_ALERT:
936 		if (optlen < sizeof(int))
937 			goto e_inval;
938 		retv = ip6_ra_control(sk, val);
939 		if (retv == 0)
940 			inet6_assign_bit(RTALERT, sk, valbool);
941 		break;
942 	case IPV6_FLOWLABEL_MGR:
943 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
944 		break;
945 	case IPV6_IPSEC_POLICY:
946 	case IPV6_XFRM_POLICY:
947 		retv = -EPERM;
948 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
949 			break;
950 		retv = xfrm_user_policy(sk, optname, optval, optlen);
951 		break;
952 
953 	case IPV6_RECVFRAGSIZE:
954 		np->rxopt.bits.recvfragsize = valbool;
955 		retv = 0;
956 		break;
957 	}
958 
959 unlock:
960 	sockopt_release_sock(sk);
961 	if (needs_rtnl)
962 		rtnl_unlock();
963 
964 	return retv;
965 
966 e_inval:
967 	retv = -EINVAL;
968 	goto unlock;
969 }
970 
971 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
972 		    unsigned int optlen)
973 {
974 	int err;
975 
976 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
977 		return ip_setsockopt(sk, level, optname, optval, optlen);
978 
979 	if (level != SOL_IPV6)
980 		return -ENOPROTOOPT;
981 
982 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
983 #ifdef CONFIG_NETFILTER
984 	/* we need to exclude all possible ENOPROTOOPTs except default case */
985 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
986 			optname != IPV6_XFRM_POLICY)
987 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
988 #endif
989 	return err;
990 }
991 EXPORT_SYMBOL(ipv6_setsockopt);
992 
993 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
994 				  int optname, sockptr_t optval, int len)
995 {
996 	struct ipv6_opt_hdr *hdr;
997 
998 	if (!opt)
999 		return 0;
1000 
1001 	switch (optname) {
1002 	case IPV6_HOPOPTS:
1003 		hdr = opt->hopopt;
1004 		break;
1005 	case IPV6_RTHDRDSTOPTS:
1006 		hdr = opt->dst0opt;
1007 		break;
1008 	case IPV6_RTHDR:
1009 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1010 		break;
1011 	case IPV6_DSTOPTS:
1012 		hdr = opt->dst1opt;
1013 		break;
1014 	default:
1015 		return -EINVAL;	/* should not happen */
1016 	}
1017 
1018 	if (!hdr)
1019 		return 0;
1020 
1021 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1022 	if (copy_to_sockptr(optval, hdr, len))
1023 		return -EFAULT;
1024 	return len;
1025 }
1026 
1027 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1028 			     sockptr_t optlen, int len)
1029 {
1030 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1031 	struct group_filter gsf;
1032 	int num;
1033 	int err;
1034 
1035 	if (len < size0)
1036 		return -EINVAL;
1037 	if (copy_from_sockptr(&gsf, optval, size0))
1038 		return -EFAULT;
1039 	if (gsf.gf_group.ss_family != AF_INET6)
1040 		return -EADDRNOTAVAIL;
1041 	num = gsf.gf_numsrc;
1042 	sockopt_lock_sock(sk);
1043 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1044 	if (!err) {
1045 		if (num > gsf.gf_numsrc)
1046 			num = gsf.gf_numsrc;
1047 		len = GROUP_FILTER_SIZE(num);
1048 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1049 		    copy_to_sockptr(optval, &gsf, size0))
1050 			err = -EFAULT;
1051 	}
1052 	sockopt_release_sock(sk);
1053 	return err;
1054 }
1055 
1056 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1057 				    sockptr_t optlen, int len)
1058 {
1059 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1060 	struct compat_group_filter gf32;
1061 	struct group_filter gf;
1062 	int err;
1063 	int num;
1064 
1065 	if (len < size0)
1066 		return -EINVAL;
1067 
1068 	if (copy_from_sockptr(&gf32, optval, size0))
1069 		return -EFAULT;
1070 	gf.gf_interface = gf32.gf_interface;
1071 	gf.gf_fmode = gf32.gf_fmode;
1072 	num = gf.gf_numsrc = gf32.gf_numsrc;
1073 	gf.gf_group = gf32.gf_group;
1074 
1075 	if (gf.gf_group.ss_family != AF_INET6)
1076 		return -EADDRNOTAVAIL;
1077 
1078 	sockopt_lock_sock(sk);
1079 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1080 	sockopt_release_sock(sk);
1081 	if (err)
1082 		return err;
1083 	if (num > gf.gf_numsrc)
1084 		num = gf.gf_numsrc;
1085 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1086 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1087 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1088 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1089 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1090 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1091 		return -EFAULT;
1092 	return 0;
1093 }
1094 
1095 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1096 		       sockptr_t optval, sockptr_t optlen)
1097 {
1098 	struct ipv6_pinfo *np = inet6_sk(sk);
1099 	int len;
1100 	int val;
1101 
1102 	if (ip6_mroute_opt(optname))
1103 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1104 
1105 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1106 		return -EFAULT;
1107 	switch (optname) {
1108 	case IPV6_ADDRFORM:
1109 		if (sk->sk_protocol != IPPROTO_UDP &&
1110 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1111 		    sk->sk_protocol != IPPROTO_TCP)
1112 			return -ENOPROTOOPT;
1113 		if (sk->sk_state != TCP_ESTABLISHED)
1114 			return -ENOTCONN;
1115 		val = sk->sk_family;
1116 		break;
1117 	case MCAST_MSFILTER:
1118 		if (in_compat_syscall())
1119 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1120 		return ipv6_get_msfilter(sk, optval, optlen, len);
1121 	case IPV6_2292PKTOPTIONS:
1122 	{
1123 		struct msghdr msg;
1124 		struct sk_buff *skb;
1125 
1126 		if (sk->sk_type != SOCK_STREAM)
1127 			return -ENOPROTOOPT;
1128 
1129 		if (optval.is_kernel) {
1130 			msg.msg_control_is_user = false;
1131 			msg.msg_control = optval.kernel;
1132 		} else {
1133 			msg.msg_control_is_user = true;
1134 			msg.msg_control_user = optval.user;
1135 		}
1136 		msg.msg_controllen = len;
1137 		msg.msg_flags = 0;
1138 
1139 		sockopt_lock_sock(sk);
1140 		skb = np->pktoptions;
1141 		if (skb)
1142 			ip6_datagram_recv_ctl(sk, &msg, skb);
1143 		sockopt_release_sock(sk);
1144 		if (!skb) {
1145 			if (np->rxopt.bits.rxinfo) {
1146 				int mcast_oif = READ_ONCE(np->mcast_oif);
1147 				struct in6_pktinfo src_info;
1148 
1149 				src_info.ipi6_ifindex = mcast_oif ? :
1150 					np->sticky_pktinfo.ipi6_ifindex;
1151 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1152 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1153 			}
1154 			if (np->rxopt.bits.rxhlim) {
1155 				int hlim = READ_ONCE(np->mcast_hops);
1156 
1157 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1158 			}
1159 			if (np->rxopt.bits.rxtclass) {
1160 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1161 
1162 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1163 			}
1164 			if (np->rxopt.bits.rxoinfo) {
1165 				int mcast_oif = READ_ONCE(np->mcast_oif);
1166 				struct in6_pktinfo src_info;
1167 
1168 				src_info.ipi6_ifindex = mcast_oif ? :
1169 					np->sticky_pktinfo.ipi6_ifindex;
1170 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr :
1171 								 np->sticky_pktinfo.ipi6_addr;
1172 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1173 			}
1174 			if (np->rxopt.bits.rxohlim) {
1175 				int hlim = READ_ONCE(np->mcast_hops);
1176 
1177 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1178 			}
1179 			if (np->rxopt.bits.rxflow) {
1180 				__be32 flowinfo = np->rcv_flowinfo;
1181 
1182 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1183 			}
1184 		}
1185 		len -= msg.msg_controllen;
1186 		return copy_to_sockptr(optlen, &len, sizeof(int));
1187 	}
1188 	case IPV6_MTU:
1189 	{
1190 		struct dst_entry *dst;
1191 
1192 		val = 0;
1193 		rcu_read_lock();
1194 		dst = __sk_dst_get(sk);
1195 		if (dst)
1196 			val = dst_mtu(dst);
1197 		rcu_read_unlock();
1198 		if (!val)
1199 			return -ENOTCONN;
1200 		break;
1201 	}
1202 
1203 	case IPV6_V6ONLY:
1204 		val = sk->sk_ipv6only;
1205 		break;
1206 
1207 	case IPV6_RECVPKTINFO:
1208 		val = np->rxopt.bits.rxinfo;
1209 		break;
1210 
1211 	case IPV6_2292PKTINFO:
1212 		val = np->rxopt.bits.rxoinfo;
1213 		break;
1214 
1215 	case IPV6_RECVHOPLIMIT:
1216 		val = np->rxopt.bits.rxhlim;
1217 		break;
1218 
1219 	case IPV6_2292HOPLIMIT:
1220 		val = np->rxopt.bits.rxohlim;
1221 		break;
1222 
1223 	case IPV6_RECVRTHDR:
1224 		val = np->rxopt.bits.srcrt;
1225 		break;
1226 
1227 	case IPV6_2292RTHDR:
1228 		val = np->rxopt.bits.osrcrt;
1229 		break;
1230 
1231 	case IPV6_HOPOPTS:
1232 	case IPV6_RTHDRDSTOPTS:
1233 	case IPV6_RTHDR:
1234 	case IPV6_DSTOPTS:
1235 	{
1236 		struct ipv6_txoptions *opt;
1237 
1238 		sockopt_lock_sock(sk);
1239 		opt = rcu_dereference_protected(np->opt,
1240 						lockdep_sock_is_held(sk));
1241 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1242 		sockopt_release_sock(sk);
1243 		/* check if ipv6_getsockopt_sticky() returns err code */
1244 		if (len < 0)
1245 			return len;
1246 		return copy_to_sockptr(optlen, &len, sizeof(int));
1247 	}
1248 
1249 	case IPV6_RECVHOPOPTS:
1250 		val = np->rxopt.bits.hopopts;
1251 		break;
1252 
1253 	case IPV6_2292HOPOPTS:
1254 		val = np->rxopt.bits.ohopopts;
1255 		break;
1256 
1257 	case IPV6_RECVDSTOPTS:
1258 		val = np->rxopt.bits.dstopts;
1259 		break;
1260 
1261 	case IPV6_2292DSTOPTS:
1262 		val = np->rxopt.bits.odstopts;
1263 		break;
1264 
1265 	case IPV6_TCLASS:
1266 		val = np->tclass;
1267 		break;
1268 
1269 	case IPV6_RECVTCLASS:
1270 		val = np->rxopt.bits.rxtclass;
1271 		break;
1272 
1273 	case IPV6_FLOWINFO:
1274 		val = np->rxopt.bits.rxflow;
1275 		break;
1276 
1277 	case IPV6_RECVPATHMTU:
1278 		val = np->rxopt.bits.rxpmtu;
1279 		break;
1280 
1281 	case IPV6_PATHMTU:
1282 	{
1283 		struct dst_entry *dst;
1284 		struct ip6_mtuinfo mtuinfo;
1285 
1286 		if (len < sizeof(mtuinfo))
1287 			return -EINVAL;
1288 
1289 		len = sizeof(mtuinfo);
1290 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1291 
1292 		rcu_read_lock();
1293 		dst = __sk_dst_get(sk);
1294 		if (dst)
1295 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1296 		rcu_read_unlock();
1297 		if (!mtuinfo.ip6m_mtu)
1298 			return -ENOTCONN;
1299 
1300 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1301 			return -EFAULT;
1302 		if (copy_to_sockptr(optval, &mtuinfo, len))
1303 			return -EFAULT;
1304 
1305 		return 0;
1306 	}
1307 
1308 	case IPV6_TRANSPARENT:
1309 		val = inet_test_bit(TRANSPARENT, sk);
1310 		break;
1311 
1312 	case IPV6_FREEBIND:
1313 		val = inet_test_bit(FREEBIND, sk);
1314 		break;
1315 
1316 	case IPV6_RECVORIGDSTADDR:
1317 		val = np->rxopt.bits.rxorigdstaddr;
1318 		break;
1319 
1320 	case IPV6_UNICAST_HOPS:
1321 	case IPV6_MULTICAST_HOPS:
1322 	{
1323 		struct dst_entry *dst;
1324 
1325 		if (optname == IPV6_UNICAST_HOPS)
1326 			val = READ_ONCE(np->hop_limit);
1327 		else
1328 			val = READ_ONCE(np->mcast_hops);
1329 
1330 		if (val < 0) {
1331 			rcu_read_lock();
1332 			dst = __sk_dst_get(sk);
1333 			if (dst)
1334 				val = ip6_dst_hoplimit(dst);
1335 			rcu_read_unlock();
1336 		}
1337 
1338 		if (val < 0)
1339 			val = READ_ONCE(sock_net(sk)->ipv6.devconf_all->hop_limit);
1340 		break;
1341 	}
1342 
1343 	case IPV6_MULTICAST_LOOP:
1344 		val = inet6_test_bit(MC6_LOOP, sk);
1345 		break;
1346 
1347 	case IPV6_MULTICAST_IF:
1348 		val = READ_ONCE(np->mcast_oif);
1349 		break;
1350 
1351 	case IPV6_MULTICAST_ALL:
1352 		val = inet6_test_bit(MC6_ALL, sk);
1353 		break;
1354 
1355 	case IPV6_UNICAST_IF:
1356 		val = (__force int)htonl((__u32) READ_ONCE(np->ucast_oif));
1357 		break;
1358 
1359 	case IPV6_MTU_DISCOVER:
1360 		val = READ_ONCE(np->pmtudisc);
1361 		break;
1362 
1363 	case IPV6_RECVERR:
1364 		val = inet6_test_bit(RECVERR6, sk);
1365 		break;
1366 
1367 	case IPV6_FLOWINFO_SEND:
1368 		val = inet6_test_bit(SNDFLOW, sk);
1369 		break;
1370 
1371 	case IPV6_FLOWLABEL_MGR:
1372 	{
1373 		struct in6_flowlabel_req freq;
1374 		int flags;
1375 
1376 		if (len < sizeof(freq))
1377 			return -EINVAL;
1378 
1379 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1380 			return -EFAULT;
1381 
1382 		if (freq.flr_action != IPV6_FL_A_GET)
1383 			return -EINVAL;
1384 
1385 		len = sizeof(freq);
1386 		flags = freq.flr_flags;
1387 
1388 		memset(&freq, 0, sizeof(freq));
1389 
1390 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1391 		if (val < 0)
1392 			return val;
1393 
1394 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1395 			return -EFAULT;
1396 		if (copy_to_sockptr(optval, &freq, len))
1397 			return -EFAULT;
1398 
1399 		return 0;
1400 	}
1401 
1402 	case IPV6_ADDR_PREFERENCES:
1403 		{
1404 		u8 srcprefs = READ_ONCE(np->srcprefs);
1405 		val = 0;
1406 
1407 		if (srcprefs & IPV6_PREFER_SRC_TMP)
1408 			val |= IPV6_PREFER_SRC_TMP;
1409 		else if (srcprefs & IPV6_PREFER_SRC_PUBLIC)
1410 			val |= IPV6_PREFER_SRC_PUBLIC;
1411 		else {
1412 			/* XXX: should we return system default? */
1413 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1414 		}
1415 
1416 		if (srcprefs & IPV6_PREFER_SRC_COA)
1417 			val |= IPV6_PREFER_SRC_COA;
1418 		else
1419 			val |= IPV6_PREFER_SRC_HOME;
1420 		break;
1421 		}
1422 	case IPV6_MINHOPCOUNT:
1423 		val = READ_ONCE(np->min_hopcount);
1424 		break;
1425 
1426 	case IPV6_DONTFRAG:
1427 		val = inet6_test_bit(DONTFRAG, sk);
1428 		break;
1429 
1430 	case IPV6_AUTOFLOWLABEL:
1431 		val = ip6_autoflowlabel(sock_net(sk), sk);
1432 		break;
1433 
1434 	case IPV6_RECVFRAGSIZE:
1435 		val = np->rxopt.bits.recvfragsize;
1436 		break;
1437 
1438 	case IPV6_ROUTER_ALERT:
1439 		val = inet6_test_bit(RTALERT, sk);
1440 		break;
1441 
1442 	case IPV6_ROUTER_ALERT_ISOLATE:
1443 		val = inet6_test_bit(RTALERT_ISOLATE, sk);
1444 		break;
1445 
1446 	case IPV6_RECVERR_RFC4884:
1447 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1448 		break;
1449 
1450 	default:
1451 		return -ENOPROTOOPT;
1452 	}
1453 	len = min_t(unsigned int, sizeof(int), len);
1454 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1455 		return -EFAULT;
1456 	if (copy_to_sockptr(optval, &val, len))
1457 		return -EFAULT;
1458 	return 0;
1459 }
1460 
1461 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1462 		    char __user *optval, int __user *optlen)
1463 {
1464 	int err;
1465 
1466 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1467 		return ip_getsockopt(sk, level, optname, optval, optlen);
1468 
1469 	if (level != SOL_IPV6)
1470 		return -ENOPROTOOPT;
1471 
1472 	err = do_ipv6_getsockopt(sk, level, optname,
1473 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1474 #ifdef CONFIG_NETFILTER
1475 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1476 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1477 		int len;
1478 
1479 		if (get_user(len, optlen))
1480 			return -EFAULT;
1481 
1482 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1483 		if (err >= 0)
1484 			err = put_user(len, optlen);
1485 	}
1486 #endif
1487 	return err;
1488 }
1489 EXPORT_SYMBOL(ipv6_getsockopt);
1490