xref: /linux/net/ipv6/ipv6_sockglue.c (revision 1086ca7cce292bb498d7f8f85f4593c9ef4902b7)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 BSD socket options interface
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on linux/net/ipv4/ip_sockglue.c
10  *
11  *	FIXME: Make the setsockopt code POSIX compliant: That is
12  *
13  *	o	Truncate getsockopt returns
14  *	o	Return an optlen of the truncated length if need be
15  *
16  *	Changes:
17  *	David L Stevens <dlstevens@us.ibm.com>:
18  *		- added multicast source filtering API for MLDv2
19  */
20 
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/mroute6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/init.h>
33 #include <linux/sysctl.h>
34 #include <linux/netfilter.h>
35 #include <linux/slab.h>
36 
37 #include <net/sock.h>
38 #include <net/snmp.h>
39 #include <net/ipv6.h>
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_route.h>
44 #include <net/addrconf.h>
45 #include <net/inet_common.h>
46 #include <net/tcp.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/xfrm.h>
50 #include <net/compat.h>
51 #include <net/seg6.h>
52 
53 #include <linux/uaccess.h>
54 
55 struct ip6_ra_chain *ip6_ra_chain;
56 DEFINE_RWLOCK(ip6_ra_lock);
57 
58 DEFINE_STATIC_KEY_FALSE(ip6_min_hopcount);
59 
60 int ip6_ra_control(struct sock *sk, int sel)
61 {
62 	struct ip6_ra_chain *ra, *new_ra, **rap;
63 
64 	/* RA packet may be delivered ONLY to IPPROTO_RAW socket */
65 	if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
66 		return -ENOPROTOOPT;
67 
68 	new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
69 	if (sel >= 0 && !new_ra)
70 		return -ENOMEM;
71 
72 	write_lock_bh(&ip6_ra_lock);
73 	for (rap = &ip6_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
74 		if (ra->sk == sk) {
75 			if (sel >= 0) {
76 				write_unlock_bh(&ip6_ra_lock);
77 				kfree(new_ra);
78 				return -EADDRINUSE;
79 			}
80 
81 			*rap = ra->next;
82 			write_unlock_bh(&ip6_ra_lock);
83 
84 			sock_put(sk);
85 			kfree(ra);
86 			return 0;
87 		}
88 	}
89 	if (!new_ra) {
90 		write_unlock_bh(&ip6_ra_lock);
91 		return -ENOBUFS;
92 	}
93 	new_ra->sk = sk;
94 	new_ra->sel = sel;
95 	new_ra->next = ra;
96 	*rap = new_ra;
97 	sock_hold(sk);
98 	write_unlock_bh(&ip6_ra_lock);
99 	return 0;
100 }
101 
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 					   struct ipv6_txoptions *opt)
104 {
105 	if (inet_test_bit(IS_ICSK, sk)) {
106 		if (opt &&
107 		    !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108 		    inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109 			struct inet_connection_sock *icsk = inet_csk(sk);
110 			icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111 			icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
112 		}
113 	}
114 	opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
115 		   opt);
116 	sk_dst_reset(sk);
117 
118 	return opt;
119 }
120 
121 static bool setsockopt_needs_rtnl(int optname)
122 {
123 	switch (optname) {
124 	case IPV6_ADDRFORM:
125 	case IPV6_ADD_MEMBERSHIP:
126 	case IPV6_DROP_MEMBERSHIP:
127 	case IPV6_JOIN_ANYCAST:
128 	case IPV6_LEAVE_ANYCAST:
129 	case MCAST_JOIN_GROUP:
130 	case MCAST_LEAVE_GROUP:
131 	case MCAST_JOIN_SOURCE_GROUP:
132 	case MCAST_LEAVE_SOURCE_GROUP:
133 	case MCAST_BLOCK_SOURCE:
134 	case MCAST_UNBLOCK_SOURCE:
135 	case MCAST_MSFILTER:
136 		return true;
137 	}
138 	return false;
139 }
140 
141 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
142 		sockptr_t optval, int optlen)
143 {
144 	if (in_compat_syscall()) {
145 		struct compat_group_source_req gr32;
146 
147 		if (optlen < sizeof(gr32))
148 			return -EINVAL;
149 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
150 			return -EFAULT;
151 		greqs->gsr_interface = gr32.gsr_interface;
152 		greqs->gsr_group = gr32.gsr_group;
153 		greqs->gsr_source = gr32.gsr_source;
154 	} else {
155 		if (optlen < sizeof(*greqs))
156 			return -EINVAL;
157 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
158 			return -EFAULT;
159 	}
160 
161 	return 0;
162 }
163 
164 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
165 		sockptr_t optval, int optlen)
166 {
167 	struct group_source_req greqs;
168 	int omode, add;
169 	int ret;
170 
171 	ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
172 	if (ret)
173 		return ret;
174 
175 	if (greqs.gsr_group.ss_family != AF_INET6 ||
176 	    greqs.gsr_source.ss_family != AF_INET6)
177 		return -EADDRNOTAVAIL;
178 
179 	if (optname == MCAST_BLOCK_SOURCE) {
180 		omode = MCAST_EXCLUDE;
181 		add = 1;
182 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
183 		omode = MCAST_EXCLUDE;
184 		add = 0;
185 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
186 		struct sockaddr_in6 *psin6;
187 		int retv;
188 
189 		psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
190 		retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
191 					     &psin6->sin6_addr,
192 					     MCAST_INCLUDE);
193 		/* prior join w/ different source is ok */
194 		if (retv && retv != -EADDRINUSE)
195 			return retv;
196 		omode = MCAST_INCLUDE;
197 		add = 1;
198 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
199 		omode = MCAST_INCLUDE;
200 		add = 0;
201 	}
202 	return ip6_mc_source(add, omode, sk, &greqs);
203 }
204 
205 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
206 		int optlen)
207 {
208 	struct group_filter *gsf;
209 	int ret;
210 
211 	if (optlen < GROUP_FILTER_SIZE(0))
212 		return -EINVAL;
213 	if (optlen > READ_ONCE(sysctl_optmem_max))
214 		return -ENOBUFS;
215 
216 	gsf = memdup_sockptr(optval, optlen);
217 	if (IS_ERR(gsf))
218 		return PTR_ERR(gsf);
219 
220 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
221 	ret = -ENOBUFS;
222 	if (gsf->gf_numsrc >= 0x1ffffffU ||
223 	    gsf->gf_numsrc > sysctl_mld_max_msf)
224 		goto out_free_gsf;
225 
226 	ret = -EINVAL;
227 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
228 		goto out_free_gsf;
229 
230 	ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
231 out_free_gsf:
232 	kfree(gsf);
233 	return ret;
234 }
235 
236 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
237 		int optlen)
238 {
239 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
240 	struct compat_group_filter *gf32;
241 	void *p;
242 	int ret;
243 	int n;
244 
245 	if (optlen < size0)
246 		return -EINVAL;
247 	if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
248 		return -ENOBUFS;
249 
250 	p = kmalloc(optlen + 4, GFP_KERNEL);
251 	if (!p)
252 		return -ENOMEM;
253 
254 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
255 	ret = -EFAULT;
256 	if (copy_from_sockptr(gf32, optval, optlen))
257 		goto out_free_p;
258 
259 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
260 	ret = -ENOBUFS;
261 	n = gf32->gf_numsrc;
262 	if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
263 		goto out_free_p;
264 
265 	ret = -EINVAL;
266 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
267 		goto out_free_p;
268 
269 	ret = ip6_mc_msfilter(sk, &(struct group_filter){
270 			.gf_interface = gf32->gf_interface,
271 			.gf_group = gf32->gf_group,
272 			.gf_fmode = gf32->gf_fmode,
273 			.gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
274 
275 out_free_p:
276 	kfree(p);
277 	return ret;
278 }
279 
280 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
281 		sockptr_t optval, int optlen)
282 {
283 	struct sockaddr_in6 *psin6;
284 	struct group_req greq;
285 
286 	if (optlen < sizeof(greq))
287 		return -EINVAL;
288 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
289 		return -EFAULT;
290 
291 	if (greq.gr_group.ss_family != AF_INET6)
292 		return -EADDRNOTAVAIL;
293 	psin6 = (struct sockaddr_in6 *)&greq.gr_group;
294 	if (optname == MCAST_JOIN_GROUP)
295 		return ipv6_sock_mc_join(sk, greq.gr_interface,
296 					 &psin6->sin6_addr);
297 	return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
298 }
299 
300 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
301 		sockptr_t optval, int optlen)
302 {
303 	struct compat_group_req gr32;
304 	struct sockaddr_in6 *psin6;
305 
306 	if (optlen < sizeof(gr32))
307 		return -EINVAL;
308 	if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
309 		return -EFAULT;
310 
311 	if (gr32.gr_group.ss_family != AF_INET6)
312 		return -EADDRNOTAVAIL;
313 	psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
314 	if (optname == MCAST_JOIN_GROUP)
315 		return ipv6_sock_mc_join(sk, gr32.gr_interface,
316 					&psin6->sin6_addr);
317 	return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
318 }
319 
320 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
321 		int optlen)
322 {
323 	struct ipv6_pinfo *np = inet6_sk(sk);
324 	struct ipv6_opt_hdr *new = NULL;
325 	struct net *net = sock_net(sk);
326 	struct ipv6_txoptions *opt;
327 	int err;
328 
329 	/* hop-by-hop / destination options are privileged option */
330 	if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
331 		return -EPERM;
332 
333 	/* remove any sticky options header with a zero option
334 	 * length, per RFC3542.
335 	 */
336 	if (optlen > 0) {
337 		if (sockptr_is_null(optval))
338 			return -EINVAL;
339 		if (optlen < sizeof(struct ipv6_opt_hdr) ||
340 		    optlen & 0x7 ||
341 		    optlen > 8 * 255)
342 			return -EINVAL;
343 
344 		new = memdup_sockptr(optval, optlen);
345 		if (IS_ERR(new))
346 			return PTR_ERR(new);
347 		if (unlikely(ipv6_optlen(new) > optlen)) {
348 			kfree(new);
349 			return -EINVAL;
350 		}
351 	}
352 
353 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
354 	opt = ipv6_renew_options(sk, opt, optname, new);
355 	kfree(new);
356 	if (IS_ERR(opt))
357 		return PTR_ERR(opt);
358 
359 	/* routing header option needs extra check */
360 	err = -EINVAL;
361 	if (optname == IPV6_RTHDR && opt && opt->srcrt) {
362 		struct ipv6_rt_hdr *rthdr = opt->srcrt;
363 		switch (rthdr->type) {
364 #if IS_ENABLED(CONFIG_IPV6_MIP6)
365 		case IPV6_SRCRT_TYPE_2:
366 			if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
367 				goto sticky_done;
368 			break;
369 #endif
370 		case IPV6_SRCRT_TYPE_4:
371 		{
372 			struct ipv6_sr_hdr *srh =
373 				(struct ipv6_sr_hdr *)opt->srcrt;
374 
375 			if (!seg6_validate_srh(srh, optlen, false))
376 				goto sticky_done;
377 			break;
378 		}
379 		default:
380 			goto sticky_done;
381 		}
382 	}
383 
384 	err = 0;
385 	opt = ipv6_update_options(sk, opt);
386 sticky_done:
387 	if (opt) {
388 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
389 		txopt_put(opt);
390 	}
391 	return err;
392 }
393 
394 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
395 		       sockptr_t optval, unsigned int optlen)
396 {
397 	struct ipv6_pinfo *np = inet6_sk(sk);
398 	struct net *net = sock_net(sk);
399 	int val, valbool;
400 	int retv = -ENOPROTOOPT;
401 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
402 
403 	if (sockptr_is_null(optval))
404 		val = 0;
405 	else {
406 		if (optlen >= sizeof(int)) {
407 			if (copy_from_sockptr(&val, optval, sizeof(val)))
408 				return -EFAULT;
409 		} else
410 			val = 0;
411 	}
412 
413 	valbool = (val != 0);
414 
415 	if (ip6_mroute_opt(optname))
416 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
417 
418 	/* Handle options that can be set without locking the socket. */
419 	switch (optname) {
420 	case IPV6_UNICAST_HOPS:
421 		if (optlen < sizeof(int))
422 			return -EINVAL;
423 		if (val > 255 || val < -1)
424 			return -EINVAL;
425 		WRITE_ONCE(np->hop_limit, val);
426 		return 0;
427 	case IPV6_MULTICAST_LOOP:
428 		if (optlen < sizeof(int))
429 			return -EINVAL;
430 		if (val != valbool)
431 			return -EINVAL;
432 		inet6_assign_bit(MC6_LOOP, sk, valbool);
433 		return 0;
434 	case IPV6_MULTICAST_HOPS:
435 		if (sk->sk_type == SOCK_STREAM)
436 			return retv;
437 		if (optlen < sizeof(int))
438 			return -EINVAL;
439 		if (val > 255 || val < -1)
440 			return -EINVAL;
441 		WRITE_ONCE(np->mcast_hops,
442 			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
443 		return 0;
444 	case IPV6_MTU:
445 		if (optlen < sizeof(int))
446 			return -EINVAL;
447 		if (val && val < IPV6_MIN_MTU)
448 			return -EINVAL;
449 		WRITE_ONCE(np->frag_size, val);
450 		return 0;
451 	case IPV6_MINHOPCOUNT:
452 		if (optlen < sizeof(int))
453 			return -EINVAL;
454 		if (val < 0 || val > 255)
455 			return -EINVAL;
456 
457 		if (val)
458 			static_branch_enable(&ip6_min_hopcount);
459 
460 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
461 		 * while we are changing it.
462 		 */
463 		WRITE_ONCE(np->min_hopcount, val);
464 		return 0;
465 	case IPV6_RECVERR_RFC4884:
466 		if (optlen < sizeof(int))
467 			return -EINVAL;
468 		if (val < 0 || val > 1)
469 			return -EINVAL;
470 		inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
471 		return 0;
472 	case IPV6_MULTICAST_ALL:
473 		if (optlen < sizeof(int))
474 			return -EINVAL;
475 		inet6_assign_bit(MC6_ALL, sk, valbool);
476 		return 0;
477 	case IPV6_AUTOFLOWLABEL:
478 		inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
479 		inet6_set_bit(AUTOFLOWLABEL_SET, sk);
480 		return 0;
481 	case IPV6_DONTFRAG:
482 		inet6_assign_bit(DONTFRAG, sk, valbool);
483 		return 0;
484 	}
485 	if (needs_rtnl)
486 		rtnl_lock();
487 	sockopt_lock_sock(sk);
488 
489 	/* Another thread has converted the socket into IPv4 with
490 	 * IPV6_ADDRFORM concurrently.
491 	 */
492 	if (unlikely(sk->sk_family != AF_INET6))
493 		goto unlock;
494 
495 	switch (optname) {
496 
497 	case IPV6_ADDRFORM:
498 		if (optlen < sizeof(int))
499 			goto e_inval;
500 		if (val == PF_INET) {
501 			if (sk->sk_type == SOCK_RAW)
502 				break;
503 
504 			if (sk->sk_protocol == IPPROTO_UDP ||
505 			    sk->sk_protocol == IPPROTO_UDPLITE) {
506 				struct udp_sock *up = udp_sk(sk);
507 				if (up->pending == AF_INET6) {
508 					retv = -EBUSY;
509 					break;
510 				}
511 			} else if (sk->sk_protocol == IPPROTO_TCP) {
512 				if (sk->sk_prot != &tcpv6_prot) {
513 					retv = -EBUSY;
514 					break;
515 				}
516 			} else {
517 				break;
518 			}
519 
520 			if (sk->sk_state != TCP_ESTABLISHED) {
521 				retv = -ENOTCONN;
522 				break;
523 			}
524 
525 			if (ipv6_only_sock(sk) ||
526 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
527 				retv = -EADDRNOTAVAIL;
528 				break;
529 			}
530 
531 			__ipv6_sock_mc_close(sk);
532 			__ipv6_sock_ac_close(sk);
533 
534 			if (sk->sk_protocol == IPPROTO_TCP) {
535 				struct inet_connection_sock *icsk = inet_csk(sk);
536 
537 				sock_prot_inuse_add(net, sk->sk_prot, -1);
538 				sock_prot_inuse_add(net, &tcp_prot, 1);
539 
540 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
541 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
542 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
543 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
544 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
545 				WRITE_ONCE(sk->sk_family, PF_INET);
546 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
547 			} else {
548 				struct proto *prot = &udp_prot;
549 
550 				if (sk->sk_protocol == IPPROTO_UDPLITE)
551 					prot = &udplite_prot;
552 
553 				sock_prot_inuse_add(net, sk->sk_prot, -1);
554 				sock_prot_inuse_add(net, prot, 1);
555 
556 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
557 				WRITE_ONCE(sk->sk_prot, prot);
558 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
559 				WRITE_ONCE(sk->sk_family, PF_INET);
560 			}
561 
562 			/* Disable all options not to allocate memory anymore,
563 			 * but there is still a race.  See the lockless path
564 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
565 			 */
566 			np->rxopt.all = 0;
567 
568 			inet6_cleanup_sock(sk);
569 
570 			module_put(THIS_MODULE);
571 			retv = 0;
572 			break;
573 		}
574 		goto e_inval;
575 
576 	case IPV6_V6ONLY:
577 		if (optlen < sizeof(int) ||
578 		    inet_sk(sk)->inet_num)
579 			goto e_inval;
580 		sk->sk_ipv6only = valbool;
581 		retv = 0;
582 		break;
583 
584 	case IPV6_RECVPKTINFO:
585 		if (optlen < sizeof(int))
586 			goto e_inval;
587 		np->rxopt.bits.rxinfo = valbool;
588 		retv = 0;
589 		break;
590 
591 	case IPV6_2292PKTINFO:
592 		if (optlen < sizeof(int))
593 			goto e_inval;
594 		np->rxopt.bits.rxoinfo = valbool;
595 		retv = 0;
596 		break;
597 
598 	case IPV6_RECVHOPLIMIT:
599 		if (optlen < sizeof(int))
600 			goto e_inval;
601 		np->rxopt.bits.rxhlim = valbool;
602 		retv = 0;
603 		break;
604 
605 	case IPV6_2292HOPLIMIT:
606 		if (optlen < sizeof(int))
607 			goto e_inval;
608 		np->rxopt.bits.rxohlim = valbool;
609 		retv = 0;
610 		break;
611 
612 	case IPV6_RECVRTHDR:
613 		if (optlen < sizeof(int))
614 			goto e_inval;
615 		np->rxopt.bits.srcrt = valbool;
616 		retv = 0;
617 		break;
618 
619 	case IPV6_2292RTHDR:
620 		if (optlen < sizeof(int))
621 			goto e_inval;
622 		np->rxopt.bits.osrcrt = valbool;
623 		retv = 0;
624 		break;
625 
626 	case IPV6_RECVHOPOPTS:
627 		if (optlen < sizeof(int))
628 			goto e_inval;
629 		np->rxopt.bits.hopopts = valbool;
630 		retv = 0;
631 		break;
632 
633 	case IPV6_2292HOPOPTS:
634 		if (optlen < sizeof(int))
635 			goto e_inval;
636 		np->rxopt.bits.ohopopts = valbool;
637 		retv = 0;
638 		break;
639 
640 	case IPV6_RECVDSTOPTS:
641 		if (optlen < sizeof(int))
642 			goto e_inval;
643 		np->rxopt.bits.dstopts = valbool;
644 		retv = 0;
645 		break;
646 
647 	case IPV6_2292DSTOPTS:
648 		if (optlen < sizeof(int))
649 			goto e_inval;
650 		np->rxopt.bits.odstopts = valbool;
651 		retv = 0;
652 		break;
653 
654 	case IPV6_TCLASS:
655 		if (optlen < sizeof(int))
656 			goto e_inval;
657 		if (val < -1 || val > 0xff)
658 			goto e_inval;
659 		/* RFC 3542, 6.5: default traffic class of 0x0 */
660 		if (val == -1)
661 			val = 0;
662 		if (sk->sk_type == SOCK_STREAM) {
663 			val &= ~INET_ECN_MASK;
664 			val |= np->tclass & INET_ECN_MASK;
665 		}
666 		if (np->tclass != val) {
667 			np->tclass = val;
668 			sk_dst_reset(sk);
669 		}
670 		retv = 0;
671 		break;
672 
673 	case IPV6_RECVTCLASS:
674 		if (optlen < sizeof(int))
675 			goto e_inval;
676 		np->rxopt.bits.rxtclass = valbool;
677 		retv = 0;
678 		break;
679 
680 	case IPV6_FLOWINFO:
681 		if (optlen < sizeof(int))
682 			goto e_inval;
683 		np->rxopt.bits.rxflow = valbool;
684 		retv = 0;
685 		break;
686 
687 	case IPV6_RECVPATHMTU:
688 		if (optlen < sizeof(int))
689 			goto e_inval;
690 		np->rxopt.bits.rxpmtu = valbool;
691 		retv = 0;
692 		break;
693 
694 	case IPV6_TRANSPARENT:
695 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
696 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
697 			retv = -EPERM;
698 			break;
699 		}
700 		if (optlen < sizeof(int))
701 			goto e_inval;
702 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
703 		inet_assign_bit(TRANSPARENT, sk, valbool);
704 		retv = 0;
705 		break;
706 
707 	case IPV6_FREEBIND:
708 		if (optlen < sizeof(int))
709 			goto e_inval;
710 		/* we also don't have a separate freebind bit for IPV6 */
711 		inet_assign_bit(FREEBIND, sk, valbool);
712 		retv = 0;
713 		break;
714 
715 	case IPV6_RECVORIGDSTADDR:
716 		if (optlen < sizeof(int))
717 			goto e_inval;
718 		np->rxopt.bits.rxorigdstaddr = valbool;
719 		retv = 0;
720 		break;
721 
722 	case IPV6_HOPOPTS:
723 	case IPV6_RTHDRDSTOPTS:
724 	case IPV6_RTHDR:
725 	case IPV6_DSTOPTS:
726 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
727 		break;
728 
729 	case IPV6_PKTINFO:
730 	{
731 		struct in6_pktinfo pkt;
732 
733 		if (optlen == 0)
734 			goto e_inval;
735 		else if (optlen < sizeof(struct in6_pktinfo) ||
736 			 sockptr_is_null(optval))
737 			goto e_inval;
738 
739 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
740 			retv = -EFAULT;
741 			break;
742 		}
743 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
744 			goto e_inval;
745 
746 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
747 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
748 		retv = 0;
749 		break;
750 	}
751 
752 	case IPV6_2292PKTOPTIONS:
753 	{
754 		struct ipv6_txoptions *opt = NULL;
755 		struct msghdr msg;
756 		struct flowi6 fl6;
757 		struct ipcm6_cookie ipc6;
758 
759 		memset(&fl6, 0, sizeof(fl6));
760 		fl6.flowi6_oif = sk->sk_bound_dev_if;
761 		fl6.flowi6_mark = sk->sk_mark;
762 
763 		if (optlen == 0)
764 			goto update;
765 
766 		/* 1K is probably excessive
767 		 * 1K is surely not enough, 2K per standard header is 16K.
768 		 */
769 		retv = -EINVAL;
770 		if (optlen > 64*1024)
771 			break;
772 
773 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
774 		retv = -ENOBUFS;
775 		if (!opt)
776 			break;
777 
778 		memset(opt, 0, sizeof(*opt));
779 		refcount_set(&opt->refcnt, 1);
780 		opt->tot_len = sizeof(*opt) + optlen;
781 		retv = -EFAULT;
782 		if (copy_from_sockptr(opt + 1, optval, optlen))
783 			goto done;
784 
785 		msg.msg_controllen = optlen;
786 		msg.msg_control_is_user = false;
787 		msg.msg_control = (void *)(opt+1);
788 		ipc6.opt = opt;
789 
790 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
791 		if (retv)
792 			goto done;
793 update:
794 		retv = 0;
795 		opt = ipv6_update_options(sk, opt);
796 done:
797 		if (opt) {
798 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
799 			txopt_put(opt);
800 		}
801 		break;
802 	}
803 
804 
805 	case IPV6_UNICAST_IF:
806 	{
807 		struct net_device *dev = NULL;
808 		int ifindex;
809 
810 		if (optlen != sizeof(int))
811 			goto e_inval;
812 
813 		ifindex = (__force int)ntohl((__force __be32)val);
814 		if (ifindex == 0) {
815 			np->ucast_oif = 0;
816 			retv = 0;
817 			break;
818 		}
819 
820 		dev = dev_get_by_index(net, ifindex);
821 		retv = -EADDRNOTAVAIL;
822 		if (!dev)
823 			break;
824 		dev_put(dev);
825 
826 		retv = -EINVAL;
827 		if (sk->sk_bound_dev_if)
828 			break;
829 
830 		np->ucast_oif = ifindex;
831 		retv = 0;
832 		break;
833 	}
834 
835 	case IPV6_MULTICAST_IF:
836 		if (sk->sk_type == SOCK_STREAM)
837 			break;
838 		if (optlen < sizeof(int))
839 			goto e_inval;
840 
841 		if (val) {
842 			struct net_device *dev;
843 			int midx;
844 
845 			rcu_read_lock();
846 
847 			dev = dev_get_by_index_rcu(net, val);
848 			if (!dev) {
849 				rcu_read_unlock();
850 				retv = -ENODEV;
851 				break;
852 			}
853 			midx = l3mdev_master_ifindex_rcu(dev);
854 
855 			rcu_read_unlock();
856 
857 			if (sk->sk_bound_dev_if &&
858 			    sk->sk_bound_dev_if != val &&
859 			    (!midx || midx != sk->sk_bound_dev_if))
860 				goto e_inval;
861 		}
862 		np->mcast_oif = val;
863 		retv = 0;
864 		break;
865 	case IPV6_ADD_MEMBERSHIP:
866 	case IPV6_DROP_MEMBERSHIP:
867 	{
868 		struct ipv6_mreq mreq;
869 
870 		if (optlen < sizeof(struct ipv6_mreq))
871 			goto e_inval;
872 
873 		retv = -EPROTO;
874 		if (inet_test_bit(IS_ICSK, sk))
875 			break;
876 
877 		retv = -EFAULT;
878 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
879 			break;
880 
881 		if (optname == IPV6_ADD_MEMBERSHIP)
882 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
883 		else
884 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
885 		break;
886 	}
887 	case IPV6_JOIN_ANYCAST:
888 	case IPV6_LEAVE_ANYCAST:
889 	{
890 		struct ipv6_mreq mreq;
891 
892 		if (optlen < sizeof(struct ipv6_mreq))
893 			goto e_inval;
894 
895 		retv = -EFAULT;
896 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
897 			break;
898 
899 		if (optname == IPV6_JOIN_ANYCAST)
900 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
901 		else
902 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
903 		break;
904 	}
905 	case MCAST_JOIN_GROUP:
906 	case MCAST_LEAVE_GROUP:
907 		if (in_compat_syscall())
908 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
909 							    optlen);
910 		else
911 			retv = ipv6_mcast_join_leave(sk, optname, optval,
912 						     optlen);
913 		break;
914 	case MCAST_JOIN_SOURCE_GROUP:
915 	case MCAST_LEAVE_SOURCE_GROUP:
916 	case MCAST_BLOCK_SOURCE:
917 	case MCAST_UNBLOCK_SOURCE:
918 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
919 		break;
920 	case MCAST_MSFILTER:
921 		if (in_compat_syscall())
922 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
923 							      optlen);
924 		else
925 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
926 		break;
927 	case IPV6_ROUTER_ALERT:
928 		if (optlen < sizeof(int))
929 			goto e_inval;
930 		retv = ip6_ra_control(sk, val);
931 		break;
932 	case IPV6_ROUTER_ALERT_ISOLATE:
933 		if (optlen < sizeof(int))
934 			goto e_inval;
935 		np->rtalert_isolate = valbool;
936 		retv = 0;
937 		break;
938 	case IPV6_MTU_DISCOVER:
939 		if (optlen < sizeof(int))
940 			goto e_inval;
941 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
942 			goto e_inval;
943 		np->pmtudisc = val;
944 		retv = 0;
945 		break;
946 	case IPV6_RECVERR:
947 		if (optlen < sizeof(int))
948 			goto e_inval;
949 		np->recverr = valbool;
950 		if (!val)
951 			skb_errqueue_purge(&sk->sk_error_queue);
952 		retv = 0;
953 		break;
954 	case IPV6_FLOWINFO_SEND:
955 		if (optlen < sizeof(int))
956 			goto e_inval;
957 		np->sndflow = valbool;
958 		retv = 0;
959 		break;
960 	case IPV6_FLOWLABEL_MGR:
961 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
962 		break;
963 	case IPV6_IPSEC_POLICY:
964 	case IPV6_XFRM_POLICY:
965 		retv = -EPERM;
966 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
967 			break;
968 		retv = xfrm_user_policy(sk, optname, optval, optlen);
969 		break;
970 
971 	case IPV6_ADDR_PREFERENCES:
972 		if (optlen < sizeof(int))
973 			goto e_inval;
974 		retv = __ip6_sock_set_addr_preferences(sk, val);
975 		break;
976 	case IPV6_RECVFRAGSIZE:
977 		np->rxopt.bits.recvfragsize = valbool;
978 		retv = 0;
979 		break;
980 	}
981 
982 unlock:
983 	sockopt_release_sock(sk);
984 	if (needs_rtnl)
985 		rtnl_unlock();
986 
987 	return retv;
988 
989 e_inval:
990 	retv = -EINVAL;
991 	goto unlock;
992 }
993 
994 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
995 		    unsigned int optlen)
996 {
997 	int err;
998 
999 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1000 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1001 
1002 	if (level != SOL_IPV6)
1003 		return -ENOPROTOOPT;
1004 
1005 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1006 #ifdef CONFIG_NETFILTER
1007 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1008 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1009 			optname != IPV6_XFRM_POLICY)
1010 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1011 #endif
1012 	return err;
1013 }
1014 EXPORT_SYMBOL(ipv6_setsockopt);
1015 
1016 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1017 				  int optname, sockptr_t optval, int len)
1018 {
1019 	struct ipv6_opt_hdr *hdr;
1020 
1021 	if (!opt)
1022 		return 0;
1023 
1024 	switch (optname) {
1025 	case IPV6_HOPOPTS:
1026 		hdr = opt->hopopt;
1027 		break;
1028 	case IPV6_RTHDRDSTOPTS:
1029 		hdr = opt->dst0opt;
1030 		break;
1031 	case IPV6_RTHDR:
1032 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1033 		break;
1034 	case IPV6_DSTOPTS:
1035 		hdr = opt->dst1opt;
1036 		break;
1037 	default:
1038 		return -EINVAL;	/* should not happen */
1039 	}
1040 
1041 	if (!hdr)
1042 		return 0;
1043 
1044 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1045 	if (copy_to_sockptr(optval, hdr, len))
1046 		return -EFAULT;
1047 	return len;
1048 }
1049 
1050 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1051 			     sockptr_t optlen, int len)
1052 {
1053 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1054 	struct group_filter gsf;
1055 	int num;
1056 	int err;
1057 
1058 	if (len < size0)
1059 		return -EINVAL;
1060 	if (copy_from_sockptr(&gsf, optval, size0))
1061 		return -EFAULT;
1062 	if (gsf.gf_group.ss_family != AF_INET6)
1063 		return -EADDRNOTAVAIL;
1064 	num = gsf.gf_numsrc;
1065 	sockopt_lock_sock(sk);
1066 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1067 	if (!err) {
1068 		if (num > gsf.gf_numsrc)
1069 			num = gsf.gf_numsrc;
1070 		len = GROUP_FILTER_SIZE(num);
1071 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1072 		    copy_to_sockptr(optval, &gsf, size0))
1073 			err = -EFAULT;
1074 	}
1075 	sockopt_release_sock(sk);
1076 	return err;
1077 }
1078 
1079 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1080 				    sockptr_t optlen, int len)
1081 {
1082 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1083 	struct compat_group_filter gf32;
1084 	struct group_filter gf;
1085 	int err;
1086 	int num;
1087 
1088 	if (len < size0)
1089 		return -EINVAL;
1090 
1091 	if (copy_from_sockptr(&gf32, optval, size0))
1092 		return -EFAULT;
1093 	gf.gf_interface = gf32.gf_interface;
1094 	gf.gf_fmode = gf32.gf_fmode;
1095 	num = gf.gf_numsrc = gf32.gf_numsrc;
1096 	gf.gf_group = gf32.gf_group;
1097 
1098 	if (gf.gf_group.ss_family != AF_INET6)
1099 		return -EADDRNOTAVAIL;
1100 
1101 	sockopt_lock_sock(sk);
1102 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1103 	sockopt_release_sock(sk);
1104 	if (err)
1105 		return err;
1106 	if (num > gf.gf_numsrc)
1107 		num = gf.gf_numsrc;
1108 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1109 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1110 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1111 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1112 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1113 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1114 		return -EFAULT;
1115 	return 0;
1116 }
1117 
1118 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1119 		       sockptr_t optval, sockptr_t optlen)
1120 {
1121 	struct ipv6_pinfo *np = inet6_sk(sk);
1122 	int len;
1123 	int val;
1124 
1125 	if (ip6_mroute_opt(optname))
1126 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1127 
1128 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1129 		return -EFAULT;
1130 	switch (optname) {
1131 	case IPV6_ADDRFORM:
1132 		if (sk->sk_protocol != IPPROTO_UDP &&
1133 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1134 		    sk->sk_protocol != IPPROTO_TCP)
1135 			return -ENOPROTOOPT;
1136 		if (sk->sk_state != TCP_ESTABLISHED)
1137 			return -ENOTCONN;
1138 		val = sk->sk_family;
1139 		break;
1140 	case MCAST_MSFILTER:
1141 		if (in_compat_syscall())
1142 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1143 		return ipv6_get_msfilter(sk, optval, optlen, len);
1144 	case IPV6_2292PKTOPTIONS:
1145 	{
1146 		struct msghdr msg;
1147 		struct sk_buff *skb;
1148 
1149 		if (sk->sk_type != SOCK_STREAM)
1150 			return -ENOPROTOOPT;
1151 
1152 		if (optval.is_kernel) {
1153 			msg.msg_control_is_user = false;
1154 			msg.msg_control = optval.kernel;
1155 		} else {
1156 			msg.msg_control_is_user = true;
1157 			msg.msg_control_user = optval.user;
1158 		}
1159 		msg.msg_controllen = len;
1160 		msg.msg_flags = 0;
1161 
1162 		sockopt_lock_sock(sk);
1163 		skb = np->pktoptions;
1164 		if (skb)
1165 			ip6_datagram_recv_ctl(sk, &msg, skb);
1166 		sockopt_release_sock(sk);
1167 		if (!skb) {
1168 			if (np->rxopt.bits.rxinfo) {
1169 				struct in6_pktinfo src_info;
1170 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1171 					np->sticky_pktinfo.ipi6_ifindex;
1172 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1173 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1174 			}
1175 			if (np->rxopt.bits.rxhlim) {
1176 				int hlim = READ_ONCE(np->mcast_hops);
1177 
1178 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1179 			}
1180 			if (np->rxopt.bits.rxtclass) {
1181 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1182 
1183 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1184 			}
1185 			if (np->rxopt.bits.rxoinfo) {
1186 				struct in6_pktinfo src_info;
1187 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1188 					np->sticky_pktinfo.ipi6_ifindex;
1189 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1190 								     np->sticky_pktinfo.ipi6_addr;
1191 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1192 			}
1193 			if (np->rxopt.bits.rxohlim) {
1194 				int hlim = READ_ONCE(np->mcast_hops);
1195 
1196 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1197 			}
1198 			if (np->rxopt.bits.rxflow) {
1199 				__be32 flowinfo = np->rcv_flowinfo;
1200 
1201 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1202 			}
1203 		}
1204 		len -= msg.msg_controllen;
1205 		return copy_to_sockptr(optlen, &len, sizeof(int));
1206 	}
1207 	case IPV6_MTU:
1208 	{
1209 		struct dst_entry *dst;
1210 
1211 		val = 0;
1212 		rcu_read_lock();
1213 		dst = __sk_dst_get(sk);
1214 		if (dst)
1215 			val = dst_mtu(dst);
1216 		rcu_read_unlock();
1217 		if (!val)
1218 			return -ENOTCONN;
1219 		break;
1220 	}
1221 
1222 	case IPV6_V6ONLY:
1223 		val = sk->sk_ipv6only;
1224 		break;
1225 
1226 	case IPV6_RECVPKTINFO:
1227 		val = np->rxopt.bits.rxinfo;
1228 		break;
1229 
1230 	case IPV6_2292PKTINFO:
1231 		val = np->rxopt.bits.rxoinfo;
1232 		break;
1233 
1234 	case IPV6_RECVHOPLIMIT:
1235 		val = np->rxopt.bits.rxhlim;
1236 		break;
1237 
1238 	case IPV6_2292HOPLIMIT:
1239 		val = np->rxopt.bits.rxohlim;
1240 		break;
1241 
1242 	case IPV6_RECVRTHDR:
1243 		val = np->rxopt.bits.srcrt;
1244 		break;
1245 
1246 	case IPV6_2292RTHDR:
1247 		val = np->rxopt.bits.osrcrt;
1248 		break;
1249 
1250 	case IPV6_HOPOPTS:
1251 	case IPV6_RTHDRDSTOPTS:
1252 	case IPV6_RTHDR:
1253 	case IPV6_DSTOPTS:
1254 	{
1255 		struct ipv6_txoptions *opt;
1256 
1257 		sockopt_lock_sock(sk);
1258 		opt = rcu_dereference_protected(np->opt,
1259 						lockdep_sock_is_held(sk));
1260 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1261 		sockopt_release_sock(sk);
1262 		/* check if ipv6_getsockopt_sticky() returns err code */
1263 		if (len < 0)
1264 			return len;
1265 		return copy_to_sockptr(optlen, &len, sizeof(int));
1266 	}
1267 
1268 	case IPV6_RECVHOPOPTS:
1269 		val = np->rxopt.bits.hopopts;
1270 		break;
1271 
1272 	case IPV6_2292HOPOPTS:
1273 		val = np->rxopt.bits.ohopopts;
1274 		break;
1275 
1276 	case IPV6_RECVDSTOPTS:
1277 		val = np->rxopt.bits.dstopts;
1278 		break;
1279 
1280 	case IPV6_2292DSTOPTS:
1281 		val = np->rxopt.bits.odstopts;
1282 		break;
1283 
1284 	case IPV6_TCLASS:
1285 		val = np->tclass;
1286 		break;
1287 
1288 	case IPV6_RECVTCLASS:
1289 		val = np->rxopt.bits.rxtclass;
1290 		break;
1291 
1292 	case IPV6_FLOWINFO:
1293 		val = np->rxopt.bits.rxflow;
1294 		break;
1295 
1296 	case IPV6_RECVPATHMTU:
1297 		val = np->rxopt.bits.rxpmtu;
1298 		break;
1299 
1300 	case IPV6_PATHMTU:
1301 	{
1302 		struct dst_entry *dst;
1303 		struct ip6_mtuinfo mtuinfo;
1304 
1305 		if (len < sizeof(mtuinfo))
1306 			return -EINVAL;
1307 
1308 		len = sizeof(mtuinfo);
1309 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1310 
1311 		rcu_read_lock();
1312 		dst = __sk_dst_get(sk);
1313 		if (dst)
1314 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1315 		rcu_read_unlock();
1316 		if (!mtuinfo.ip6m_mtu)
1317 			return -ENOTCONN;
1318 
1319 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1320 			return -EFAULT;
1321 		if (copy_to_sockptr(optval, &mtuinfo, len))
1322 			return -EFAULT;
1323 
1324 		return 0;
1325 	}
1326 
1327 	case IPV6_TRANSPARENT:
1328 		val = inet_test_bit(TRANSPARENT, sk);
1329 		break;
1330 
1331 	case IPV6_FREEBIND:
1332 		val = inet_test_bit(FREEBIND, sk);
1333 		break;
1334 
1335 	case IPV6_RECVORIGDSTADDR:
1336 		val = np->rxopt.bits.rxorigdstaddr;
1337 		break;
1338 
1339 	case IPV6_UNICAST_HOPS:
1340 	case IPV6_MULTICAST_HOPS:
1341 	{
1342 		struct dst_entry *dst;
1343 
1344 		if (optname == IPV6_UNICAST_HOPS)
1345 			val = READ_ONCE(np->hop_limit);
1346 		else
1347 			val = READ_ONCE(np->mcast_hops);
1348 
1349 		if (val < 0) {
1350 			rcu_read_lock();
1351 			dst = __sk_dst_get(sk);
1352 			if (dst)
1353 				val = ip6_dst_hoplimit(dst);
1354 			rcu_read_unlock();
1355 		}
1356 
1357 		if (val < 0)
1358 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1359 		break;
1360 	}
1361 
1362 	case IPV6_MULTICAST_LOOP:
1363 		val = inet6_test_bit(MC6_LOOP, sk);
1364 		break;
1365 
1366 	case IPV6_MULTICAST_IF:
1367 		val = np->mcast_oif;
1368 		break;
1369 
1370 	case IPV6_MULTICAST_ALL:
1371 		val = inet6_test_bit(MC6_ALL, sk);
1372 		break;
1373 
1374 	case IPV6_UNICAST_IF:
1375 		val = (__force int)htonl((__u32) np->ucast_oif);
1376 		break;
1377 
1378 	case IPV6_MTU_DISCOVER:
1379 		val = np->pmtudisc;
1380 		break;
1381 
1382 	case IPV6_RECVERR:
1383 		val = np->recverr;
1384 		break;
1385 
1386 	case IPV6_FLOWINFO_SEND:
1387 		val = np->sndflow;
1388 		break;
1389 
1390 	case IPV6_FLOWLABEL_MGR:
1391 	{
1392 		struct in6_flowlabel_req freq;
1393 		int flags;
1394 
1395 		if (len < sizeof(freq))
1396 			return -EINVAL;
1397 
1398 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1399 			return -EFAULT;
1400 
1401 		if (freq.flr_action != IPV6_FL_A_GET)
1402 			return -EINVAL;
1403 
1404 		len = sizeof(freq);
1405 		flags = freq.flr_flags;
1406 
1407 		memset(&freq, 0, sizeof(freq));
1408 
1409 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1410 		if (val < 0)
1411 			return val;
1412 
1413 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1414 			return -EFAULT;
1415 		if (copy_to_sockptr(optval, &freq, len))
1416 			return -EFAULT;
1417 
1418 		return 0;
1419 	}
1420 
1421 	case IPV6_ADDR_PREFERENCES:
1422 		val = 0;
1423 
1424 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1425 			val |= IPV6_PREFER_SRC_TMP;
1426 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1427 			val |= IPV6_PREFER_SRC_PUBLIC;
1428 		else {
1429 			/* XXX: should we return system default? */
1430 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1431 		}
1432 
1433 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1434 			val |= IPV6_PREFER_SRC_COA;
1435 		else
1436 			val |= IPV6_PREFER_SRC_HOME;
1437 		break;
1438 
1439 	case IPV6_MINHOPCOUNT:
1440 		val = READ_ONCE(np->min_hopcount);
1441 		break;
1442 
1443 	case IPV6_DONTFRAG:
1444 		val = inet6_test_bit(DONTFRAG, sk);
1445 		break;
1446 
1447 	case IPV6_AUTOFLOWLABEL:
1448 		val = ip6_autoflowlabel(sock_net(sk), sk);
1449 		break;
1450 
1451 	case IPV6_RECVFRAGSIZE:
1452 		val = np->rxopt.bits.recvfragsize;
1453 		break;
1454 
1455 	case IPV6_ROUTER_ALERT_ISOLATE:
1456 		val = np->rtalert_isolate;
1457 		break;
1458 
1459 	case IPV6_RECVERR_RFC4884:
1460 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1461 		break;
1462 
1463 	default:
1464 		return -ENOPROTOOPT;
1465 	}
1466 	len = min_t(unsigned int, sizeof(int), len);
1467 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1468 		return -EFAULT;
1469 	if (copy_to_sockptr(optval, &val, len))
1470 		return -EFAULT;
1471 	return 0;
1472 }
1473 
1474 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1475 		    char __user *optval, int __user *optlen)
1476 {
1477 	int err;
1478 
1479 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1480 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1481 
1482 	if (level != SOL_IPV6)
1483 		return -ENOPROTOOPT;
1484 
1485 	err = do_ipv6_getsockopt(sk, level, optname,
1486 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1487 #ifdef CONFIG_NETFILTER
1488 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1489 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1490 		int len;
1491 
1492 		if (get_user(len, optlen))
1493 			return -EFAULT;
1494 
1495 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1496 		if (err >= 0)
1497 			err = put_user(len, optlen);
1498 	}
1499 #endif
1500 	return err;
1501 }
1502 EXPORT_SYMBOL(ipv6_getsockopt);
1503