xref: /linux/net/ipv6/ipv6_sockglue.c (revision dcae74622c051b219ee628669a31716473efda2c)
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 	}
473 	if (needs_rtnl)
474 		rtnl_lock();
475 	sockopt_lock_sock(sk);
476 
477 	/* Another thread has converted the socket into IPv4 with
478 	 * IPV6_ADDRFORM concurrently.
479 	 */
480 	if (unlikely(sk->sk_family != AF_INET6))
481 		goto unlock;
482 
483 	switch (optname) {
484 
485 	case IPV6_ADDRFORM:
486 		if (optlen < sizeof(int))
487 			goto e_inval;
488 		if (val == PF_INET) {
489 			if (sk->sk_type == SOCK_RAW)
490 				break;
491 
492 			if (sk->sk_protocol == IPPROTO_UDP ||
493 			    sk->sk_protocol == IPPROTO_UDPLITE) {
494 				struct udp_sock *up = udp_sk(sk);
495 				if (up->pending == AF_INET6) {
496 					retv = -EBUSY;
497 					break;
498 				}
499 			} else if (sk->sk_protocol == IPPROTO_TCP) {
500 				if (sk->sk_prot != &tcpv6_prot) {
501 					retv = -EBUSY;
502 					break;
503 				}
504 			} else {
505 				break;
506 			}
507 
508 			if (sk->sk_state != TCP_ESTABLISHED) {
509 				retv = -ENOTCONN;
510 				break;
511 			}
512 
513 			if (ipv6_only_sock(sk) ||
514 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
515 				retv = -EADDRNOTAVAIL;
516 				break;
517 			}
518 
519 			__ipv6_sock_mc_close(sk);
520 			__ipv6_sock_ac_close(sk);
521 
522 			if (sk->sk_protocol == IPPROTO_TCP) {
523 				struct inet_connection_sock *icsk = inet_csk(sk);
524 
525 				sock_prot_inuse_add(net, sk->sk_prot, -1);
526 				sock_prot_inuse_add(net, &tcp_prot, 1);
527 
528 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
529 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
530 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
531 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
532 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
533 				WRITE_ONCE(sk->sk_family, PF_INET);
534 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
535 			} else {
536 				struct proto *prot = &udp_prot;
537 
538 				if (sk->sk_protocol == IPPROTO_UDPLITE)
539 					prot = &udplite_prot;
540 
541 				sock_prot_inuse_add(net, sk->sk_prot, -1);
542 				sock_prot_inuse_add(net, prot, 1);
543 
544 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
545 				WRITE_ONCE(sk->sk_prot, prot);
546 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
547 				WRITE_ONCE(sk->sk_family, PF_INET);
548 			}
549 
550 			/* Disable all options not to allocate memory anymore,
551 			 * but there is still a race.  See the lockless path
552 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
553 			 */
554 			np->rxopt.all = 0;
555 
556 			inet6_cleanup_sock(sk);
557 
558 			module_put(THIS_MODULE);
559 			retv = 0;
560 			break;
561 		}
562 		goto e_inval;
563 
564 	case IPV6_V6ONLY:
565 		if (optlen < sizeof(int) ||
566 		    inet_sk(sk)->inet_num)
567 			goto e_inval;
568 		sk->sk_ipv6only = valbool;
569 		retv = 0;
570 		break;
571 
572 	case IPV6_RECVPKTINFO:
573 		if (optlen < sizeof(int))
574 			goto e_inval;
575 		np->rxopt.bits.rxinfo = valbool;
576 		retv = 0;
577 		break;
578 
579 	case IPV6_2292PKTINFO:
580 		if (optlen < sizeof(int))
581 			goto e_inval;
582 		np->rxopt.bits.rxoinfo = valbool;
583 		retv = 0;
584 		break;
585 
586 	case IPV6_RECVHOPLIMIT:
587 		if (optlen < sizeof(int))
588 			goto e_inval;
589 		np->rxopt.bits.rxhlim = valbool;
590 		retv = 0;
591 		break;
592 
593 	case IPV6_2292HOPLIMIT:
594 		if (optlen < sizeof(int))
595 			goto e_inval;
596 		np->rxopt.bits.rxohlim = valbool;
597 		retv = 0;
598 		break;
599 
600 	case IPV6_RECVRTHDR:
601 		if (optlen < sizeof(int))
602 			goto e_inval;
603 		np->rxopt.bits.srcrt = valbool;
604 		retv = 0;
605 		break;
606 
607 	case IPV6_2292RTHDR:
608 		if (optlen < sizeof(int))
609 			goto e_inval;
610 		np->rxopt.bits.osrcrt = valbool;
611 		retv = 0;
612 		break;
613 
614 	case IPV6_RECVHOPOPTS:
615 		if (optlen < sizeof(int))
616 			goto e_inval;
617 		np->rxopt.bits.hopopts = valbool;
618 		retv = 0;
619 		break;
620 
621 	case IPV6_2292HOPOPTS:
622 		if (optlen < sizeof(int))
623 			goto e_inval;
624 		np->rxopt.bits.ohopopts = valbool;
625 		retv = 0;
626 		break;
627 
628 	case IPV6_RECVDSTOPTS:
629 		if (optlen < sizeof(int))
630 			goto e_inval;
631 		np->rxopt.bits.dstopts = valbool;
632 		retv = 0;
633 		break;
634 
635 	case IPV6_2292DSTOPTS:
636 		if (optlen < sizeof(int))
637 			goto e_inval;
638 		np->rxopt.bits.odstopts = valbool;
639 		retv = 0;
640 		break;
641 
642 	case IPV6_TCLASS:
643 		if (optlen < sizeof(int))
644 			goto e_inval;
645 		if (val < -1 || val > 0xff)
646 			goto e_inval;
647 		/* RFC 3542, 6.5: default traffic class of 0x0 */
648 		if (val == -1)
649 			val = 0;
650 		if (sk->sk_type == SOCK_STREAM) {
651 			val &= ~INET_ECN_MASK;
652 			val |= np->tclass & INET_ECN_MASK;
653 		}
654 		if (np->tclass != val) {
655 			np->tclass = val;
656 			sk_dst_reset(sk);
657 		}
658 		retv = 0;
659 		break;
660 
661 	case IPV6_RECVTCLASS:
662 		if (optlen < sizeof(int))
663 			goto e_inval;
664 		np->rxopt.bits.rxtclass = valbool;
665 		retv = 0;
666 		break;
667 
668 	case IPV6_FLOWINFO:
669 		if (optlen < sizeof(int))
670 			goto e_inval;
671 		np->rxopt.bits.rxflow = valbool;
672 		retv = 0;
673 		break;
674 
675 	case IPV6_RECVPATHMTU:
676 		if (optlen < sizeof(int))
677 			goto e_inval;
678 		np->rxopt.bits.rxpmtu = valbool;
679 		retv = 0;
680 		break;
681 
682 	case IPV6_TRANSPARENT:
683 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
684 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
685 			retv = -EPERM;
686 			break;
687 		}
688 		if (optlen < sizeof(int))
689 			goto e_inval;
690 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
691 		inet_assign_bit(TRANSPARENT, sk, valbool);
692 		retv = 0;
693 		break;
694 
695 	case IPV6_FREEBIND:
696 		if (optlen < sizeof(int))
697 			goto e_inval;
698 		/* we also don't have a separate freebind bit for IPV6 */
699 		inet_assign_bit(FREEBIND, sk, valbool);
700 		retv = 0;
701 		break;
702 
703 	case IPV6_RECVORIGDSTADDR:
704 		if (optlen < sizeof(int))
705 			goto e_inval;
706 		np->rxopt.bits.rxorigdstaddr = valbool;
707 		retv = 0;
708 		break;
709 
710 	case IPV6_HOPOPTS:
711 	case IPV6_RTHDRDSTOPTS:
712 	case IPV6_RTHDR:
713 	case IPV6_DSTOPTS:
714 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
715 		break;
716 
717 	case IPV6_PKTINFO:
718 	{
719 		struct in6_pktinfo pkt;
720 
721 		if (optlen == 0)
722 			goto e_inval;
723 		else if (optlen < sizeof(struct in6_pktinfo) ||
724 			 sockptr_is_null(optval))
725 			goto e_inval;
726 
727 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
728 			retv = -EFAULT;
729 			break;
730 		}
731 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
732 			goto e_inval;
733 
734 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
735 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
736 		retv = 0;
737 		break;
738 	}
739 
740 	case IPV6_2292PKTOPTIONS:
741 	{
742 		struct ipv6_txoptions *opt = NULL;
743 		struct msghdr msg;
744 		struct flowi6 fl6;
745 		struct ipcm6_cookie ipc6;
746 
747 		memset(&fl6, 0, sizeof(fl6));
748 		fl6.flowi6_oif = sk->sk_bound_dev_if;
749 		fl6.flowi6_mark = sk->sk_mark;
750 
751 		if (optlen == 0)
752 			goto update;
753 
754 		/* 1K is probably excessive
755 		 * 1K is surely not enough, 2K per standard header is 16K.
756 		 */
757 		retv = -EINVAL;
758 		if (optlen > 64*1024)
759 			break;
760 
761 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
762 		retv = -ENOBUFS;
763 		if (!opt)
764 			break;
765 
766 		memset(opt, 0, sizeof(*opt));
767 		refcount_set(&opt->refcnt, 1);
768 		opt->tot_len = sizeof(*opt) + optlen;
769 		retv = -EFAULT;
770 		if (copy_from_sockptr(opt + 1, optval, optlen))
771 			goto done;
772 
773 		msg.msg_controllen = optlen;
774 		msg.msg_control_is_user = false;
775 		msg.msg_control = (void *)(opt+1);
776 		ipc6.opt = opt;
777 
778 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
779 		if (retv)
780 			goto done;
781 update:
782 		retv = 0;
783 		opt = ipv6_update_options(sk, opt);
784 done:
785 		if (opt) {
786 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
787 			txopt_put(opt);
788 		}
789 		break;
790 	}
791 
792 
793 	case IPV6_UNICAST_IF:
794 	{
795 		struct net_device *dev = NULL;
796 		int ifindex;
797 
798 		if (optlen != sizeof(int))
799 			goto e_inval;
800 
801 		ifindex = (__force int)ntohl((__force __be32)val);
802 		if (ifindex == 0) {
803 			np->ucast_oif = 0;
804 			retv = 0;
805 			break;
806 		}
807 
808 		dev = dev_get_by_index(net, ifindex);
809 		retv = -EADDRNOTAVAIL;
810 		if (!dev)
811 			break;
812 		dev_put(dev);
813 
814 		retv = -EINVAL;
815 		if (sk->sk_bound_dev_if)
816 			break;
817 
818 		np->ucast_oif = ifindex;
819 		retv = 0;
820 		break;
821 	}
822 
823 	case IPV6_MULTICAST_IF:
824 		if (sk->sk_type == SOCK_STREAM)
825 			break;
826 		if (optlen < sizeof(int))
827 			goto e_inval;
828 
829 		if (val) {
830 			struct net_device *dev;
831 			int midx;
832 
833 			rcu_read_lock();
834 
835 			dev = dev_get_by_index_rcu(net, val);
836 			if (!dev) {
837 				rcu_read_unlock();
838 				retv = -ENODEV;
839 				break;
840 			}
841 			midx = l3mdev_master_ifindex_rcu(dev);
842 
843 			rcu_read_unlock();
844 
845 			if (sk->sk_bound_dev_if &&
846 			    sk->sk_bound_dev_if != val &&
847 			    (!midx || midx != sk->sk_bound_dev_if))
848 				goto e_inval;
849 		}
850 		np->mcast_oif = val;
851 		retv = 0;
852 		break;
853 	case IPV6_ADD_MEMBERSHIP:
854 	case IPV6_DROP_MEMBERSHIP:
855 	{
856 		struct ipv6_mreq mreq;
857 
858 		if (optlen < sizeof(struct ipv6_mreq))
859 			goto e_inval;
860 
861 		retv = -EPROTO;
862 		if (inet_test_bit(IS_ICSK, sk))
863 			break;
864 
865 		retv = -EFAULT;
866 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
867 			break;
868 
869 		if (optname == IPV6_ADD_MEMBERSHIP)
870 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
871 		else
872 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
873 		break;
874 	}
875 	case IPV6_JOIN_ANYCAST:
876 	case IPV6_LEAVE_ANYCAST:
877 	{
878 		struct ipv6_mreq mreq;
879 
880 		if (optlen < sizeof(struct ipv6_mreq))
881 			goto e_inval;
882 
883 		retv = -EFAULT;
884 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
885 			break;
886 
887 		if (optname == IPV6_JOIN_ANYCAST)
888 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
889 		else
890 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
891 		break;
892 	}
893 	case IPV6_MULTICAST_ALL:
894 		if (optlen < sizeof(int))
895 			goto e_inval;
896 		np->mc_all = valbool;
897 		retv = 0;
898 		break;
899 
900 	case MCAST_JOIN_GROUP:
901 	case MCAST_LEAVE_GROUP:
902 		if (in_compat_syscall())
903 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
904 							    optlen);
905 		else
906 			retv = ipv6_mcast_join_leave(sk, optname, optval,
907 						     optlen);
908 		break;
909 	case MCAST_JOIN_SOURCE_GROUP:
910 	case MCAST_LEAVE_SOURCE_GROUP:
911 	case MCAST_BLOCK_SOURCE:
912 	case MCAST_UNBLOCK_SOURCE:
913 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
914 		break;
915 	case MCAST_MSFILTER:
916 		if (in_compat_syscall())
917 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
918 							      optlen);
919 		else
920 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
921 		break;
922 	case IPV6_ROUTER_ALERT:
923 		if (optlen < sizeof(int))
924 			goto e_inval;
925 		retv = ip6_ra_control(sk, val);
926 		break;
927 	case IPV6_ROUTER_ALERT_ISOLATE:
928 		if (optlen < sizeof(int))
929 			goto e_inval;
930 		np->rtalert_isolate = valbool;
931 		retv = 0;
932 		break;
933 	case IPV6_MTU_DISCOVER:
934 		if (optlen < sizeof(int))
935 			goto e_inval;
936 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
937 			goto e_inval;
938 		np->pmtudisc = val;
939 		retv = 0;
940 		break;
941 	case IPV6_RECVERR:
942 		if (optlen < sizeof(int))
943 			goto e_inval;
944 		np->recverr = valbool;
945 		if (!val)
946 			skb_errqueue_purge(&sk->sk_error_queue);
947 		retv = 0;
948 		break;
949 	case IPV6_FLOWINFO_SEND:
950 		if (optlen < sizeof(int))
951 			goto e_inval;
952 		np->sndflow = valbool;
953 		retv = 0;
954 		break;
955 	case IPV6_FLOWLABEL_MGR:
956 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
957 		break;
958 	case IPV6_IPSEC_POLICY:
959 	case IPV6_XFRM_POLICY:
960 		retv = -EPERM;
961 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
962 			break;
963 		retv = xfrm_user_policy(sk, optname, optval, optlen);
964 		break;
965 
966 	case IPV6_ADDR_PREFERENCES:
967 		if (optlen < sizeof(int))
968 			goto e_inval;
969 		retv = __ip6_sock_set_addr_preferences(sk, val);
970 		break;
971 	case IPV6_DONTFRAG:
972 		np->dontfrag = valbool;
973 		retv = 0;
974 		break;
975 	case IPV6_AUTOFLOWLABEL:
976 		np->autoflowlabel = valbool;
977 		np->autoflowlabel_set = 1;
978 		retv = 0;
979 		break;
980 	case IPV6_RECVFRAGSIZE:
981 		np->rxopt.bits.recvfragsize = valbool;
982 		retv = 0;
983 		break;
984 	}
985 
986 unlock:
987 	sockopt_release_sock(sk);
988 	if (needs_rtnl)
989 		rtnl_unlock();
990 
991 	return retv;
992 
993 e_inval:
994 	retv = -EINVAL;
995 	goto unlock;
996 }
997 
998 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
999 		    unsigned int optlen)
1000 {
1001 	int err;
1002 
1003 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1004 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1005 
1006 	if (level != SOL_IPV6)
1007 		return -ENOPROTOOPT;
1008 
1009 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1010 #ifdef CONFIG_NETFILTER
1011 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1012 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1013 			optname != IPV6_XFRM_POLICY)
1014 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1015 #endif
1016 	return err;
1017 }
1018 EXPORT_SYMBOL(ipv6_setsockopt);
1019 
1020 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1021 				  int optname, sockptr_t optval, int len)
1022 {
1023 	struct ipv6_opt_hdr *hdr;
1024 
1025 	if (!opt)
1026 		return 0;
1027 
1028 	switch (optname) {
1029 	case IPV6_HOPOPTS:
1030 		hdr = opt->hopopt;
1031 		break;
1032 	case IPV6_RTHDRDSTOPTS:
1033 		hdr = opt->dst0opt;
1034 		break;
1035 	case IPV6_RTHDR:
1036 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1037 		break;
1038 	case IPV6_DSTOPTS:
1039 		hdr = opt->dst1opt;
1040 		break;
1041 	default:
1042 		return -EINVAL;	/* should not happen */
1043 	}
1044 
1045 	if (!hdr)
1046 		return 0;
1047 
1048 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1049 	if (copy_to_sockptr(optval, hdr, len))
1050 		return -EFAULT;
1051 	return len;
1052 }
1053 
1054 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1055 			     sockptr_t optlen, int len)
1056 {
1057 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1058 	struct group_filter gsf;
1059 	int num;
1060 	int err;
1061 
1062 	if (len < size0)
1063 		return -EINVAL;
1064 	if (copy_from_sockptr(&gsf, optval, size0))
1065 		return -EFAULT;
1066 	if (gsf.gf_group.ss_family != AF_INET6)
1067 		return -EADDRNOTAVAIL;
1068 	num = gsf.gf_numsrc;
1069 	sockopt_lock_sock(sk);
1070 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1071 	if (!err) {
1072 		if (num > gsf.gf_numsrc)
1073 			num = gsf.gf_numsrc;
1074 		len = GROUP_FILTER_SIZE(num);
1075 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1076 		    copy_to_sockptr(optval, &gsf, size0))
1077 			err = -EFAULT;
1078 	}
1079 	sockopt_release_sock(sk);
1080 	return err;
1081 }
1082 
1083 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1084 				    sockptr_t optlen, int len)
1085 {
1086 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1087 	struct compat_group_filter gf32;
1088 	struct group_filter gf;
1089 	int err;
1090 	int num;
1091 
1092 	if (len < size0)
1093 		return -EINVAL;
1094 
1095 	if (copy_from_sockptr(&gf32, optval, size0))
1096 		return -EFAULT;
1097 	gf.gf_interface = gf32.gf_interface;
1098 	gf.gf_fmode = gf32.gf_fmode;
1099 	num = gf.gf_numsrc = gf32.gf_numsrc;
1100 	gf.gf_group = gf32.gf_group;
1101 
1102 	if (gf.gf_group.ss_family != AF_INET6)
1103 		return -EADDRNOTAVAIL;
1104 
1105 	sockopt_lock_sock(sk);
1106 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1107 	sockopt_release_sock(sk);
1108 	if (err)
1109 		return err;
1110 	if (num > gf.gf_numsrc)
1111 		num = gf.gf_numsrc;
1112 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1113 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1114 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1115 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1116 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1117 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1118 		return -EFAULT;
1119 	return 0;
1120 }
1121 
1122 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1123 		       sockptr_t optval, sockptr_t optlen)
1124 {
1125 	struct ipv6_pinfo *np = inet6_sk(sk);
1126 	int len;
1127 	int val;
1128 
1129 	if (ip6_mroute_opt(optname))
1130 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1131 
1132 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1133 		return -EFAULT;
1134 	switch (optname) {
1135 	case IPV6_ADDRFORM:
1136 		if (sk->sk_protocol != IPPROTO_UDP &&
1137 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1138 		    sk->sk_protocol != IPPROTO_TCP)
1139 			return -ENOPROTOOPT;
1140 		if (sk->sk_state != TCP_ESTABLISHED)
1141 			return -ENOTCONN;
1142 		val = sk->sk_family;
1143 		break;
1144 	case MCAST_MSFILTER:
1145 		if (in_compat_syscall())
1146 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1147 		return ipv6_get_msfilter(sk, optval, optlen, len);
1148 	case IPV6_2292PKTOPTIONS:
1149 	{
1150 		struct msghdr msg;
1151 		struct sk_buff *skb;
1152 
1153 		if (sk->sk_type != SOCK_STREAM)
1154 			return -ENOPROTOOPT;
1155 
1156 		if (optval.is_kernel) {
1157 			msg.msg_control_is_user = false;
1158 			msg.msg_control = optval.kernel;
1159 		} else {
1160 			msg.msg_control_is_user = true;
1161 			msg.msg_control_user = optval.user;
1162 		}
1163 		msg.msg_controllen = len;
1164 		msg.msg_flags = 0;
1165 
1166 		sockopt_lock_sock(sk);
1167 		skb = np->pktoptions;
1168 		if (skb)
1169 			ip6_datagram_recv_ctl(sk, &msg, skb);
1170 		sockopt_release_sock(sk);
1171 		if (!skb) {
1172 			if (np->rxopt.bits.rxinfo) {
1173 				struct in6_pktinfo src_info;
1174 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1175 					np->sticky_pktinfo.ipi6_ifindex;
1176 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1177 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1178 			}
1179 			if (np->rxopt.bits.rxhlim) {
1180 				int hlim = READ_ONCE(np->mcast_hops);
1181 
1182 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1183 			}
1184 			if (np->rxopt.bits.rxtclass) {
1185 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1186 
1187 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1188 			}
1189 			if (np->rxopt.bits.rxoinfo) {
1190 				struct in6_pktinfo src_info;
1191 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1192 					np->sticky_pktinfo.ipi6_ifindex;
1193 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1194 								     np->sticky_pktinfo.ipi6_addr;
1195 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1196 			}
1197 			if (np->rxopt.bits.rxohlim) {
1198 				int hlim = READ_ONCE(np->mcast_hops);
1199 
1200 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1201 			}
1202 			if (np->rxopt.bits.rxflow) {
1203 				__be32 flowinfo = np->rcv_flowinfo;
1204 
1205 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1206 			}
1207 		}
1208 		len -= msg.msg_controllen;
1209 		return copy_to_sockptr(optlen, &len, sizeof(int));
1210 	}
1211 	case IPV6_MTU:
1212 	{
1213 		struct dst_entry *dst;
1214 
1215 		val = 0;
1216 		rcu_read_lock();
1217 		dst = __sk_dst_get(sk);
1218 		if (dst)
1219 			val = dst_mtu(dst);
1220 		rcu_read_unlock();
1221 		if (!val)
1222 			return -ENOTCONN;
1223 		break;
1224 	}
1225 
1226 	case IPV6_V6ONLY:
1227 		val = sk->sk_ipv6only;
1228 		break;
1229 
1230 	case IPV6_RECVPKTINFO:
1231 		val = np->rxopt.bits.rxinfo;
1232 		break;
1233 
1234 	case IPV6_2292PKTINFO:
1235 		val = np->rxopt.bits.rxoinfo;
1236 		break;
1237 
1238 	case IPV6_RECVHOPLIMIT:
1239 		val = np->rxopt.bits.rxhlim;
1240 		break;
1241 
1242 	case IPV6_2292HOPLIMIT:
1243 		val = np->rxopt.bits.rxohlim;
1244 		break;
1245 
1246 	case IPV6_RECVRTHDR:
1247 		val = np->rxopt.bits.srcrt;
1248 		break;
1249 
1250 	case IPV6_2292RTHDR:
1251 		val = np->rxopt.bits.osrcrt;
1252 		break;
1253 
1254 	case IPV6_HOPOPTS:
1255 	case IPV6_RTHDRDSTOPTS:
1256 	case IPV6_RTHDR:
1257 	case IPV6_DSTOPTS:
1258 	{
1259 		struct ipv6_txoptions *opt;
1260 
1261 		sockopt_lock_sock(sk);
1262 		opt = rcu_dereference_protected(np->opt,
1263 						lockdep_sock_is_held(sk));
1264 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1265 		sockopt_release_sock(sk);
1266 		/* check if ipv6_getsockopt_sticky() returns err code */
1267 		if (len < 0)
1268 			return len;
1269 		return copy_to_sockptr(optlen, &len, sizeof(int));
1270 	}
1271 
1272 	case IPV6_RECVHOPOPTS:
1273 		val = np->rxopt.bits.hopopts;
1274 		break;
1275 
1276 	case IPV6_2292HOPOPTS:
1277 		val = np->rxopt.bits.ohopopts;
1278 		break;
1279 
1280 	case IPV6_RECVDSTOPTS:
1281 		val = np->rxopt.bits.dstopts;
1282 		break;
1283 
1284 	case IPV6_2292DSTOPTS:
1285 		val = np->rxopt.bits.odstopts;
1286 		break;
1287 
1288 	case IPV6_TCLASS:
1289 		val = np->tclass;
1290 		break;
1291 
1292 	case IPV6_RECVTCLASS:
1293 		val = np->rxopt.bits.rxtclass;
1294 		break;
1295 
1296 	case IPV6_FLOWINFO:
1297 		val = np->rxopt.bits.rxflow;
1298 		break;
1299 
1300 	case IPV6_RECVPATHMTU:
1301 		val = np->rxopt.bits.rxpmtu;
1302 		break;
1303 
1304 	case IPV6_PATHMTU:
1305 	{
1306 		struct dst_entry *dst;
1307 		struct ip6_mtuinfo mtuinfo;
1308 
1309 		if (len < sizeof(mtuinfo))
1310 			return -EINVAL;
1311 
1312 		len = sizeof(mtuinfo);
1313 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1314 
1315 		rcu_read_lock();
1316 		dst = __sk_dst_get(sk);
1317 		if (dst)
1318 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1319 		rcu_read_unlock();
1320 		if (!mtuinfo.ip6m_mtu)
1321 			return -ENOTCONN;
1322 
1323 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1324 			return -EFAULT;
1325 		if (copy_to_sockptr(optval, &mtuinfo, len))
1326 			return -EFAULT;
1327 
1328 		return 0;
1329 	}
1330 
1331 	case IPV6_TRANSPARENT:
1332 		val = inet_test_bit(TRANSPARENT, sk);
1333 		break;
1334 
1335 	case IPV6_FREEBIND:
1336 		val = inet_test_bit(FREEBIND, sk);
1337 		break;
1338 
1339 	case IPV6_RECVORIGDSTADDR:
1340 		val = np->rxopt.bits.rxorigdstaddr;
1341 		break;
1342 
1343 	case IPV6_UNICAST_HOPS:
1344 	case IPV6_MULTICAST_HOPS:
1345 	{
1346 		struct dst_entry *dst;
1347 
1348 		if (optname == IPV6_UNICAST_HOPS)
1349 			val = READ_ONCE(np->hop_limit);
1350 		else
1351 			val = READ_ONCE(np->mcast_hops);
1352 
1353 		if (val < 0) {
1354 			rcu_read_lock();
1355 			dst = __sk_dst_get(sk);
1356 			if (dst)
1357 				val = ip6_dst_hoplimit(dst);
1358 			rcu_read_unlock();
1359 		}
1360 
1361 		if (val < 0)
1362 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1363 		break;
1364 	}
1365 
1366 	case IPV6_MULTICAST_LOOP:
1367 		val = inet6_test_bit(MC6_LOOP, sk);
1368 		break;
1369 
1370 	case IPV6_MULTICAST_IF:
1371 		val = np->mcast_oif;
1372 		break;
1373 
1374 	case IPV6_MULTICAST_ALL:
1375 		val = np->mc_all;
1376 		break;
1377 
1378 	case IPV6_UNICAST_IF:
1379 		val = (__force int)htonl((__u32) np->ucast_oif);
1380 		break;
1381 
1382 	case IPV6_MTU_DISCOVER:
1383 		val = np->pmtudisc;
1384 		break;
1385 
1386 	case IPV6_RECVERR:
1387 		val = np->recverr;
1388 		break;
1389 
1390 	case IPV6_FLOWINFO_SEND:
1391 		val = np->sndflow;
1392 		break;
1393 
1394 	case IPV6_FLOWLABEL_MGR:
1395 	{
1396 		struct in6_flowlabel_req freq;
1397 		int flags;
1398 
1399 		if (len < sizeof(freq))
1400 			return -EINVAL;
1401 
1402 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1403 			return -EFAULT;
1404 
1405 		if (freq.flr_action != IPV6_FL_A_GET)
1406 			return -EINVAL;
1407 
1408 		len = sizeof(freq);
1409 		flags = freq.flr_flags;
1410 
1411 		memset(&freq, 0, sizeof(freq));
1412 
1413 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1414 		if (val < 0)
1415 			return val;
1416 
1417 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1418 			return -EFAULT;
1419 		if (copy_to_sockptr(optval, &freq, len))
1420 			return -EFAULT;
1421 
1422 		return 0;
1423 	}
1424 
1425 	case IPV6_ADDR_PREFERENCES:
1426 		val = 0;
1427 
1428 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1429 			val |= IPV6_PREFER_SRC_TMP;
1430 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1431 			val |= IPV6_PREFER_SRC_PUBLIC;
1432 		else {
1433 			/* XXX: should we return system default? */
1434 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1435 		}
1436 
1437 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1438 			val |= IPV6_PREFER_SRC_COA;
1439 		else
1440 			val |= IPV6_PREFER_SRC_HOME;
1441 		break;
1442 
1443 	case IPV6_MINHOPCOUNT:
1444 		val = READ_ONCE(np->min_hopcount);
1445 		break;
1446 
1447 	case IPV6_DONTFRAG:
1448 		val = np->dontfrag;
1449 		break;
1450 
1451 	case IPV6_AUTOFLOWLABEL:
1452 		val = ip6_autoflowlabel(sock_net(sk), np);
1453 		break;
1454 
1455 	case IPV6_RECVFRAGSIZE:
1456 		val = np->rxopt.bits.recvfragsize;
1457 		break;
1458 
1459 	case IPV6_ROUTER_ALERT_ISOLATE:
1460 		val = np->rtalert_isolate;
1461 		break;
1462 
1463 	case IPV6_RECVERR_RFC4884:
1464 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1465 		break;
1466 
1467 	default:
1468 		return -ENOPROTOOPT;
1469 	}
1470 	len = min_t(unsigned int, sizeof(int), len);
1471 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1472 		return -EFAULT;
1473 	if (copy_to_sockptr(optval, &val, len))
1474 		return -EFAULT;
1475 	return 0;
1476 }
1477 
1478 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1479 		    char __user *optval, int __user *optlen)
1480 {
1481 	int err;
1482 
1483 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1484 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1485 
1486 	if (level != SOL_IPV6)
1487 		return -ENOPROTOOPT;
1488 
1489 	err = do_ipv6_getsockopt(sk, level, optname,
1490 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1491 #ifdef CONFIG_NETFILTER
1492 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1493 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1494 		int len;
1495 
1496 		if (get_user(len, optlen))
1497 			return -EFAULT;
1498 
1499 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1500 		if (err >= 0)
1501 			err = put_user(len, optlen);
1502 	}
1503 #endif
1504 	return err;
1505 }
1506 EXPORT_SYMBOL(ipv6_getsockopt);
1507