xref: /linux/net/ipv6/ipv6_sockglue.c (revision 2fe3c78a2c26dd5ee811024a1b7d6cfb4d654319)
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_ADDRFORM:
124 	case IPV6_ADD_MEMBERSHIP:
125 	case IPV6_DROP_MEMBERSHIP:
126 	case IPV6_JOIN_ANYCAST:
127 	case IPV6_LEAVE_ANYCAST:
128 	case MCAST_JOIN_GROUP:
129 	case MCAST_LEAVE_GROUP:
130 	case MCAST_JOIN_SOURCE_GROUP:
131 	case MCAST_LEAVE_SOURCE_GROUP:
132 	case MCAST_BLOCK_SOURCE:
133 	case MCAST_UNBLOCK_SOURCE:
134 	case MCAST_MSFILTER:
135 		return true;
136 	}
137 	return false;
138 }
139 
140 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
141 		sockptr_t optval, int optlen)
142 {
143 	if (in_compat_syscall()) {
144 		struct compat_group_source_req gr32;
145 
146 		if (optlen < sizeof(gr32))
147 			return -EINVAL;
148 		if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
149 			return -EFAULT;
150 		greqs->gsr_interface = gr32.gsr_interface;
151 		greqs->gsr_group = gr32.gsr_group;
152 		greqs->gsr_source = gr32.gsr_source;
153 	} else {
154 		if (optlen < sizeof(*greqs))
155 			return -EINVAL;
156 		if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
157 			return -EFAULT;
158 	}
159 
160 	return 0;
161 }
162 
163 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
164 		sockptr_t optval, int optlen)
165 {
166 	struct group_source_req greqs;
167 	int omode, add;
168 	int ret;
169 
170 	ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
171 	if (ret)
172 		return ret;
173 
174 	if (greqs.gsr_group.ss_family != AF_INET6 ||
175 	    greqs.gsr_source.ss_family != AF_INET6)
176 		return -EADDRNOTAVAIL;
177 
178 	if (optname == MCAST_BLOCK_SOURCE) {
179 		omode = MCAST_EXCLUDE;
180 		add = 1;
181 	} else if (optname == MCAST_UNBLOCK_SOURCE) {
182 		omode = MCAST_EXCLUDE;
183 		add = 0;
184 	} else if (optname == MCAST_JOIN_SOURCE_GROUP) {
185 		struct sockaddr_in6 *psin6;
186 		int retv;
187 
188 		psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
189 		retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
190 					     &psin6->sin6_addr,
191 					     MCAST_INCLUDE);
192 		/* prior join w/ different source is ok */
193 		if (retv && retv != -EADDRINUSE)
194 			return retv;
195 		omode = MCAST_INCLUDE;
196 		add = 1;
197 	} else /* MCAST_LEAVE_SOURCE_GROUP */ {
198 		omode = MCAST_INCLUDE;
199 		add = 0;
200 	}
201 	return ip6_mc_source(add, omode, sk, &greqs);
202 }
203 
204 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
205 		int optlen)
206 {
207 	struct group_filter *gsf;
208 	int ret;
209 
210 	if (optlen < GROUP_FILTER_SIZE(0))
211 		return -EINVAL;
212 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max))
213 		return -ENOBUFS;
214 
215 	gsf = memdup_sockptr(optval, optlen);
216 	if (IS_ERR(gsf))
217 		return PTR_ERR(gsf);
218 
219 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
220 	ret = -ENOBUFS;
221 	if (gsf->gf_numsrc >= 0x1ffffffU ||
222 	    gsf->gf_numsrc > sysctl_mld_max_msf)
223 		goto out_free_gsf;
224 
225 	ret = -EINVAL;
226 	if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
227 		goto out_free_gsf;
228 
229 	ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
230 out_free_gsf:
231 	kfree(gsf);
232 	return ret;
233 }
234 
235 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
236 		int optlen)
237 {
238 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
239 	struct compat_group_filter *gf32;
240 	void *p;
241 	int ret;
242 	int n;
243 
244 	if (optlen < size0)
245 		return -EINVAL;
246 	if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max) - 4)
247 		return -ENOBUFS;
248 
249 	p = kmalloc(optlen + 4, GFP_KERNEL);
250 	if (!p)
251 		return -ENOMEM;
252 
253 	gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
254 	ret = -EFAULT;
255 	if (copy_from_sockptr(gf32, optval, optlen))
256 		goto out_free_p;
257 
258 	/* numsrc >= (4G-140)/128 overflow in 32 bits */
259 	ret = -ENOBUFS;
260 	n = gf32->gf_numsrc;
261 	if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
262 		goto out_free_p;
263 
264 	ret = -EINVAL;
265 	if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
266 		goto out_free_p;
267 
268 	ret = ip6_mc_msfilter(sk, &(struct group_filter){
269 			.gf_interface = gf32->gf_interface,
270 			.gf_group = gf32->gf_group,
271 			.gf_fmode = gf32->gf_fmode,
272 			.gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
273 
274 out_free_p:
275 	kfree(p);
276 	return ret;
277 }
278 
279 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
280 		sockptr_t optval, int optlen)
281 {
282 	struct sockaddr_in6 *psin6;
283 	struct group_req greq;
284 
285 	if (optlen < sizeof(greq))
286 		return -EINVAL;
287 	if (copy_from_sockptr(&greq, optval, sizeof(greq)))
288 		return -EFAULT;
289 
290 	if (greq.gr_group.ss_family != AF_INET6)
291 		return -EADDRNOTAVAIL;
292 	psin6 = (struct sockaddr_in6 *)&greq.gr_group;
293 	if (optname == MCAST_JOIN_GROUP)
294 		return ipv6_sock_mc_join(sk, greq.gr_interface,
295 					 &psin6->sin6_addr);
296 	return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
297 }
298 
299 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
300 		sockptr_t optval, int optlen)
301 {
302 	struct compat_group_req gr32;
303 	struct sockaddr_in6 *psin6;
304 
305 	if (optlen < sizeof(gr32))
306 		return -EINVAL;
307 	if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
308 		return -EFAULT;
309 
310 	if (gr32.gr_group.ss_family != AF_INET6)
311 		return -EADDRNOTAVAIL;
312 	psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
313 	if (optname == MCAST_JOIN_GROUP)
314 		return ipv6_sock_mc_join(sk, gr32.gr_interface,
315 					&psin6->sin6_addr);
316 	return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
317 }
318 
319 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
320 		int optlen)
321 {
322 	struct ipv6_pinfo *np = inet6_sk(sk);
323 	struct ipv6_opt_hdr *new = NULL;
324 	struct net *net = sock_net(sk);
325 	struct ipv6_txoptions *opt;
326 	int err;
327 
328 	/* hop-by-hop / destination options are privileged option */
329 	if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
330 		return -EPERM;
331 
332 	/* remove any sticky options header with a zero option
333 	 * length, per RFC3542.
334 	 */
335 	if (optlen > 0) {
336 		if (sockptr_is_null(optval))
337 			return -EINVAL;
338 		if (optlen < sizeof(struct ipv6_opt_hdr) ||
339 		    optlen & 0x7 ||
340 		    optlen > 8 * 255)
341 			return -EINVAL;
342 
343 		new = memdup_sockptr(optval, optlen);
344 		if (IS_ERR(new))
345 			return PTR_ERR(new);
346 		if (unlikely(ipv6_optlen(new) > optlen)) {
347 			kfree(new);
348 			return -EINVAL;
349 		}
350 	}
351 
352 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
353 	opt = ipv6_renew_options(sk, opt, optname, new);
354 	kfree(new);
355 	if (IS_ERR(opt))
356 		return PTR_ERR(opt);
357 
358 	/* routing header option needs extra check */
359 	err = -EINVAL;
360 	if (optname == IPV6_RTHDR && opt && opt->srcrt) {
361 		struct ipv6_rt_hdr *rthdr = opt->srcrt;
362 		switch (rthdr->type) {
363 #if IS_ENABLED(CONFIG_IPV6_MIP6)
364 		case IPV6_SRCRT_TYPE_2:
365 			if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
366 				goto sticky_done;
367 			break;
368 #endif
369 		case IPV6_SRCRT_TYPE_4:
370 		{
371 			struct ipv6_sr_hdr *srh =
372 				(struct ipv6_sr_hdr *)opt->srcrt;
373 
374 			if (!seg6_validate_srh(srh, optlen, false))
375 				goto sticky_done;
376 			break;
377 		}
378 		default:
379 			goto sticky_done;
380 		}
381 	}
382 
383 	err = 0;
384 	opt = ipv6_update_options(sk, opt);
385 sticky_done:
386 	if (opt) {
387 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
388 		txopt_put(opt);
389 	}
390 	return err;
391 }
392 
393 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
394 		       sockptr_t optval, unsigned int optlen)
395 {
396 	struct ipv6_pinfo *np = inet6_sk(sk);
397 	struct net *net = sock_net(sk);
398 	int val, valbool;
399 	int retv = -ENOPROTOOPT;
400 	bool needs_rtnl = setsockopt_needs_rtnl(optname);
401 
402 	if (sockptr_is_null(optval))
403 		val = 0;
404 	else {
405 		if (optlen >= sizeof(int)) {
406 			if (copy_from_sockptr(&val, optval, sizeof(val)))
407 				return -EFAULT;
408 		} else
409 			val = 0;
410 	}
411 
412 	valbool = (val != 0);
413 
414 	if (ip6_mroute_opt(optname))
415 		return ip6_mroute_setsockopt(sk, optname, optval, optlen);
416 
417 	/* Handle options that can be set without locking the socket. */
418 	switch (optname) {
419 	case IPV6_UNICAST_HOPS:
420 		if (optlen < sizeof(int))
421 			return -EINVAL;
422 		if (val > 255 || val < -1)
423 			return -EINVAL;
424 		WRITE_ONCE(np->hop_limit, val);
425 		return 0;
426 	case IPV6_MULTICAST_LOOP:
427 		if (optlen < sizeof(int))
428 			return -EINVAL;
429 		if (val != valbool)
430 			return -EINVAL;
431 		inet6_assign_bit(MC6_LOOP, sk, valbool);
432 		return 0;
433 	case IPV6_MULTICAST_HOPS:
434 		if (sk->sk_type == SOCK_STREAM)
435 			return retv;
436 		if (optlen < sizeof(int))
437 			return -EINVAL;
438 		if (val > 255 || val < -1)
439 			return -EINVAL;
440 		WRITE_ONCE(np->mcast_hops,
441 			   val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
442 		return 0;
443 	case IPV6_MTU:
444 		if (optlen < sizeof(int))
445 			return -EINVAL;
446 		if (val && val < IPV6_MIN_MTU)
447 			return -EINVAL;
448 		WRITE_ONCE(np->frag_size, val);
449 		return 0;
450 	case IPV6_MINHOPCOUNT:
451 		if (optlen < sizeof(int))
452 			return -EINVAL;
453 		if (val < 0 || val > 255)
454 			return -EINVAL;
455 
456 		if (val)
457 			static_branch_enable(&ip6_min_hopcount);
458 
459 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
460 		 * while we are changing it.
461 		 */
462 		WRITE_ONCE(np->min_hopcount, val);
463 		return 0;
464 	case IPV6_RECVERR_RFC4884:
465 		if (optlen < sizeof(int))
466 			return -EINVAL;
467 		if (val < 0 || val > 1)
468 			return -EINVAL;
469 		inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
470 		return 0;
471 	case IPV6_MULTICAST_ALL:
472 		if (optlen < sizeof(int))
473 			return -EINVAL;
474 		inet6_assign_bit(MC6_ALL, sk, valbool);
475 		return 0;
476 	case IPV6_AUTOFLOWLABEL:
477 		inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
478 		inet6_set_bit(AUTOFLOWLABEL_SET, sk);
479 		return 0;
480 	case IPV6_DONTFRAG:
481 		inet6_assign_bit(DONTFRAG, sk, valbool);
482 		return 0;
483 	case IPV6_RECVERR:
484 		if (optlen < sizeof(int))
485 			return -EINVAL;
486 		inet6_assign_bit(RECVERR6, sk, valbool);
487 		if (!val)
488 			skb_errqueue_purge(&sk->sk_error_queue);
489 		return 0;
490 	case IPV6_ROUTER_ALERT_ISOLATE:
491 		if (optlen < sizeof(int))
492 			return -EINVAL;
493 		inet6_assign_bit(RTALERT_ISOLATE, sk, valbool);
494 		return 0;
495 	case IPV6_MTU_DISCOVER:
496 		if (optlen < sizeof(int))
497 			return -EINVAL;
498 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
499 			return -EINVAL;
500 		WRITE_ONCE(np->pmtudisc, val);
501 		return 0;
502 	case IPV6_FLOWINFO_SEND:
503 		if (optlen < sizeof(int))
504 			return -EINVAL;
505 		inet6_assign_bit(SNDFLOW, sk, valbool);
506 		return 0;
507 	case IPV6_ADDR_PREFERENCES:
508 		if (optlen < sizeof(int))
509 			return -EINVAL;
510 		return ip6_sock_set_addr_preferences(sk, val);
511 	case IPV6_MULTICAST_IF:
512 		if (sk->sk_type == SOCK_STREAM)
513 			return -ENOPROTOOPT;
514 		if (optlen < sizeof(int))
515 			return -EINVAL;
516 		if (val) {
517 			struct net_device *dev;
518 			int bound_dev_if, midx;
519 
520 			rcu_read_lock();
521 
522 			dev = dev_get_by_index_rcu(net, val);
523 			if (!dev) {
524 				rcu_read_unlock();
525 				return -ENODEV;
526 			}
527 			midx = l3mdev_master_ifindex_rcu(dev);
528 
529 			rcu_read_unlock();
530 
531 			bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
532 			if (bound_dev_if &&
533 			    bound_dev_if != val &&
534 			    (!midx || midx != bound_dev_if))
535 				return -EINVAL;
536 		}
537 		WRITE_ONCE(np->mcast_oif, val);
538 		return 0;
539 	case IPV6_UNICAST_IF:
540 	{
541 		struct net_device *dev;
542 		int ifindex;
543 
544 		if (optlen != sizeof(int))
545 			return -EINVAL;
546 
547 		ifindex = (__force int)ntohl((__force __be32)val);
548 		if (!ifindex) {
549 			WRITE_ONCE(np->ucast_oif, 0);
550 			return 0;
551 		}
552 
553 		dev = dev_get_by_index(net, ifindex);
554 		if (!dev)
555 			return -EADDRNOTAVAIL;
556 		dev_put(dev);
557 
558 		if (READ_ONCE(sk->sk_bound_dev_if))
559 			return -EINVAL;
560 
561 		WRITE_ONCE(np->ucast_oif, ifindex);
562 		return 0;
563 	}
564 	}
565 	if (needs_rtnl)
566 		rtnl_lock();
567 	sockopt_lock_sock(sk);
568 
569 	/* Another thread has converted the socket into IPv4 with
570 	 * IPV6_ADDRFORM concurrently.
571 	 */
572 	if (unlikely(sk->sk_family != AF_INET6))
573 		goto unlock;
574 
575 	switch (optname) {
576 
577 	case IPV6_ADDRFORM:
578 		if (optlen < sizeof(int))
579 			goto e_inval;
580 		if (val == PF_INET) {
581 			if (sk->sk_type == SOCK_RAW)
582 				break;
583 
584 			if (sk->sk_protocol == IPPROTO_UDP ||
585 			    sk->sk_protocol == IPPROTO_UDPLITE) {
586 				struct udp_sock *up = udp_sk(sk);
587 				if (up->pending == AF_INET6) {
588 					retv = -EBUSY;
589 					break;
590 				}
591 			} else if (sk->sk_protocol == IPPROTO_TCP) {
592 				if (sk->sk_prot != &tcpv6_prot) {
593 					retv = -EBUSY;
594 					break;
595 				}
596 			} else {
597 				break;
598 			}
599 
600 			if (sk->sk_state != TCP_ESTABLISHED) {
601 				retv = -ENOTCONN;
602 				break;
603 			}
604 
605 			if (ipv6_only_sock(sk) ||
606 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
607 				retv = -EADDRNOTAVAIL;
608 				break;
609 			}
610 
611 			__ipv6_sock_mc_close(sk);
612 			__ipv6_sock_ac_close(sk);
613 
614 			if (sk->sk_protocol == IPPROTO_TCP) {
615 				struct inet_connection_sock *icsk = inet_csk(sk);
616 
617 				sock_prot_inuse_add(net, sk->sk_prot, -1);
618 				sock_prot_inuse_add(net, &tcp_prot, 1);
619 
620 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
621 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
622 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
623 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
624 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
625 				WRITE_ONCE(sk->sk_family, PF_INET);
626 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
627 			} else {
628 				struct proto *prot = &udp_prot;
629 
630 				if (sk->sk_protocol == IPPROTO_UDPLITE)
631 					prot = &udplite_prot;
632 
633 				sock_prot_inuse_add(net, sk->sk_prot, -1);
634 				sock_prot_inuse_add(net, prot, 1);
635 
636 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
637 				WRITE_ONCE(sk->sk_prot, prot);
638 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
639 				WRITE_ONCE(sk->sk_family, PF_INET);
640 			}
641 
642 			/* Disable all options not to allocate memory anymore,
643 			 * but there is still a race.  See the lockless path
644 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
645 			 */
646 			np->rxopt.all = 0;
647 
648 			inet6_cleanup_sock(sk);
649 
650 			module_put(THIS_MODULE);
651 			retv = 0;
652 			break;
653 		}
654 		goto e_inval;
655 
656 	case IPV6_V6ONLY:
657 		if (optlen < sizeof(int) ||
658 		    inet_sk(sk)->inet_num)
659 			goto e_inval;
660 		sk->sk_ipv6only = valbool;
661 		retv = 0;
662 		break;
663 
664 	case IPV6_RECVPKTINFO:
665 		if (optlen < sizeof(int))
666 			goto e_inval;
667 		np->rxopt.bits.rxinfo = valbool;
668 		retv = 0;
669 		break;
670 
671 	case IPV6_2292PKTINFO:
672 		if (optlen < sizeof(int))
673 			goto e_inval;
674 		np->rxopt.bits.rxoinfo = valbool;
675 		retv = 0;
676 		break;
677 
678 	case IPV6_RECVHOPLIMIT:
679 		if (optlen < sizeof(int))
680 			goto e_inval;
681 		np->rxopt.bits.rxhlim = valbool;
682 		retv = 0;
683 		break;
684 
685 	case IPV6_2292HOPLIMIT:
686 		if (optlen < sizeof(int))
687 			goto e_inval;
688 		np->rxopt.bits.rxohlim = valbool;
689 		retv = 0;
690 		break;
691 
692 	case IPV6_RECVRTHDR:
693 		if (optlen < sizeof(int))
694 			goto e_inval;
695 		np->rxopt.bits.srcrt = valbool;
696 		retv = 0;
697 		break;
698 
699 	case IPV6_2292RTHDR:
700 		if (optlen < sizeof(int))
701 			goto e_inval;
702 		np->rxopt.bits.osrcrt = valbool;
703 		retv = 0;
704 		break;
705 
706 	case IPV6_RECVHOPOPTS:
707 		if (optlen < sizeof(int))
708 			goto e_inval;
709 		np->rxopt.bits.hopopts = valbool;
710 		retv = 0;
711 		break;
712 
713 	case IPV6_2292HOPOPTS:
714 		if (optlen < sizeof(int))
715 			goto e_inval;
716 		np->rxopt.bits.ohopopts = valbool;
717 		retv = 0;
718 		break;
719 
720 	case IPV6_RECVDSTOPTS:
721 		if (optlen < sizeof(int))
722 			goto e_inval;
723 		np->rxopt.bits.dstopts = valbool;
724 		retv = 0;
725 		break;
726 
727 	case IPV6_2292DSTOPTS:
728 		if (optlen < sizeof(int))
729 			goto e_inval;
730 		np->rxopt.bits.odstopts = valbool;
731 		retv = 0;
732 		break;
733 
734 	case IPV6_TCLASS:
735 		if (optlen < sizeof(int))
736 			goto e_inval;
737 		if (val < -1 || val > 0xff)
738 			goto e_inval;
739 		/* RFC 3542, 6.5: default traffic class of 0x0 */
740 		if (val == -1)
741 			val = 0;
742 		if (sk->sk_type == SOCK_STREAM) {
743 			val &= ~INET_ECN_MASK;
744 			val |= np->tclass & INET_ECN_MASK;
745 		}
746 		if (np->tclass != val) {
747 			np->tclass = val;
748 			sk_dst_reset(sk);
749 		}
750 		retv = 0;
751 		break;
752 
753 	case IPV6_RECVTCLASS:
754 		if (optlen < sizeof(int))
755 			goto e_inval;
756 		np->rxopt.bits.rxtclass = valbool;
757 		retv = 0;
758 		break;
759 
760 	case IPV6_FLOWINFO:
761 		if (optlen < sizeof(int))
762 			goto e_inval;
763 		np->rxopt.bits.rxflow = valbool;
764 		retv = 0;
765 		break;
766 
767 	case IPV6_RECVPATHMTU:
768 		if (optlen < sizeof(int))
769 			goto e_inval;
770 		np->rxopt.bits.rxpmtu = valbool;
771 		retv = 0;
772 		break;
773 
774 	case IPV6_TRANSPARENT:
775 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
776 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
777 			retv = -EPERM;
778 			break;
779 		}
780 		if (optlen < sizeof(int))
781 			goto e_inval;
782 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
783 		inet_assign_bit(TRANSPARENT, sk, valbool);
784 		retv = 0;
785 		break;
786 
787 	case IPV6_FREEBIND:
788 		if (optlen < sizeof(int))
789 			goto e_inval;
790 		/* we also don't have a separate freebind bit for IPV6 */
791 		inet_assign_bit(FREEBIND, sk, valbool);
792 		retv = 0;
793 		break;
794 
795 	case IPV6_RECVORIGDSTADDR:
796 		if (optlen < sizeof(int))
797 			goto e_inval;
798 		np->rxopt.bits.rxorigdstaddr = valbool;
799 		retv = 0;
800 		break;
801 
802 	case IPV6_HOPOPTS:
803 	case IPV6_RTHDRDSTOPTS:
804 	case IPV6_RTHDR:
805 	case IPV6_DSTOPTS:
806 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
807 		break;
808 
809 	case IPV6_PKTINFO:
810 	{
811 		struct in6_pktinfo pkt;
812 
813 		if (optlen == 0)
814 			goto e_inval;
815 		else if (optlen < sizeof(struct in6_pktinfo) ||
816 			 sockptr_is_null(optval))
817 			goto e_inval;
818 
819 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
820 			retv = -EFAULT;
821 			break;
822 		}
823 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
824 			goto e_inval;
825 
826 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
827 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
828 		retv = 0;
829 		break;
830 	}
831 
832 	case IPV6_2292PKTOPTIONS:
833 	{
834 		struct ipv6_txoptions *opt = NULL;
835 		struct msghdr msg;
836 		struct flowi6 fl6;
837 		struct ipcm6_cookie ipc6;
838 
839 		memset(&fl6, 0, sizeof(fl6));
840 		fl6.flowi6_oif = sk->sk_bound_dev_if;
841 		fl6.flowi6_mark = sk->sk_mark;
842 
843 		if (optlen == 0)
844 			goto update;
845 
846 		/* 1K is probably excessive
847 		 * 1K is surely not enough, 2K per standard header is 16K.
848 		 */
849 		retv = -EINVAL;
850 		if (optlen > 64*1024)
851 			break;
852 
853 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
854 		retv = -ENOBUFS;
855 		if (!opt)
856 			break;
857 
858 		memset(opt, 0, sizeof(*opt));
859 		refcount_set(&opt->refcnt, 1);
860 		opt->tot_len = sizeof(*opt) + optlen;
861 		retv = -EFAULT;
862 		if (copy_from_sockptr(opt + 1, optval, optlen))
863 			goto done;
864 
865 		msg.msg_controllen = optlen;
866 		msg.msg_control_is_user = false;
867 		msg.msg_control = (void *)(opt+1);
868 		ipc6.opt = opt;
869 
870 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
871 		if (retv)
872 			goto done;
873 update:
874 		retv = 0;
875 		opt = ipv6_update_options(sk, opt);
876 done:
877 		if (opt) {
878 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
879 			txopt_put(opt);
880 		}
881 		break;
882 	}
883 
884 	case IPV6_ADD_MEMBERSHIP:
885 	case IPV6_DROP_MEMBERSHIP:
886 	{
887 		struct ipv6_mreq mreq;
888 
889 		if (optlen < sizeof(struct ipv6_mreq))
890 			goto e_inval;
891 
892 		retv = -EPROTO;
893 		if (inet_test_bit(IS_ICSK, sk))
894 			break;
895 
896 		retv = -EFAULT;
897 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
898 			break;
899 
900 		if (optname == IPV6_ADD_MEMBERSHIP)
901 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
902 		else
903 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
904 		break;
905 	}
906 	case IPV6_JOIN_ANYCAST:
907 	case IPV6_LEAVE_ANYCAST:
908 	{
909 		struct ipv6_mreq mreq;
910 
911 		if (optlen < sizeof(struct ipv6_mreq))
912 			goto e_inval;
913 
914 		retv = -EFAULT;
915 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
916 			break;
917 
918 		if (optname == IPV6_JOIN_ANYCAST)
919 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
920 		else
921 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
922 		break;
923 	}
924 	case MCAST_JOIN_GROUP:
925 	case MCAST_LEAVE_GROUP:
926 		if (in_compat_syscall())
927 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
928 							    optlen);
929 		else
930 			retv = ipv6_mcast_join_leave(sk, optname, optval,
931 						     optlen);
932 		break;
933 	case MCAST_JOIN_SOURCE_GROUP:
934 	case MCAST_LEAVE_SOURCE_GROUP:
935 	case MCAST_BLOCK_SOURCE:
936 	case MCAST_UNBLOCK_SOURCE:
937 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
938 		break;
939 	case MCAST_MSFILTER:
940 		if (in_compat_syscall())
941 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
942 							      optlen);
943 		else
944 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
945 		break;
946 	case IPV6_ROUTER_ALERT:
947 		if (optlen < sizeof(int))
948 			goto e_inval;
949 		retv = ip6_ra_control(sk, val);
950 		if (retv == 0)
951 			inet6_assign_bit(RTALERT, sk, valbool);
952 		break;
953 	case IPV6_FLOWLABEL_MGR:
954 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
955 		break;
956 	case IPV6_IPSEC_POLICY:
957 	case IPV6_XFRM_POLICY:
958 		retv = -EPERM;
959 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
960 			break;
961 		retv = xfrm_user_policy(sk, optname, optval, optlen);
962 		break;
963 
964 	case IPV6_RECVFRAGSIZE:
965 		np->rxopt.bits.recvfragsize = valbool;
966 		retv = 0;
967 		break;
968 	}
969 
970 unlock:
971 	sockopt_release_sock(sk);
972 	if (needs_rtnl)
973 		rtnl_unlock();
974 
975 	return retv;
976 
977 e_inval:
978 	retv = -EINVAL;
979 	goto unlock;
980 }
981 
982 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
983 		    unsigned int optlen)
984 {
985 	int err;
986 
987 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
988 		return ip_setsockopt(sk, level, optname, optval, optlen);
989 
990 	if (level != SOL_IPV6)
991 		return -ENOPROTOOPT;
992 
993 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
994 #ifdef CONFIG_NETFILTER
995 	/* we need to exclude all possible ENOPROTOOPTs except default case */
996 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
997 			optname != IPV6_XFRM_POLICY)
998 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
999 #endif
1000 	return err;
1001 }
1002 EXPORT_SYMBOL(ipv6_setsockopt);
1003 
1004 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1005 				  int optname, sockptr_t optval, int len)
1006 {
1007 	struct ipv6_opt_hdr *hdr;
1008 
1009 	if (!opt)
1010 		return 0;
1011 
1012 	switch (optname) {
1013 	case IPV6_HOPOPTS:
1014 		hdr = opt->hopopt;
1015 		break;
1016 	case IPV6_RTHDRDSTOPTS:
1017 		hdr = opt->dst0opt;
1018 		break;
1019 	case IPV6_RTHDR:
1020 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1021 		break;
1022 	case IPV6_DSTOPTS:
1023 		hdr = opt->dst1opt;
1024 		break;
1025 	default:
1026 		return -EINVAL;	/* should not happen */
1027 	}
1028 
1029 	if (!hdr)
1030 		return 0;
1031 
1032 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1033 	if (copy_to_sockptr(optval, hdr, len))
1034 		return -EFAULT;
1035 	return len;
1036 }
1037 
1038 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1039 			     sockptr_t optlen, int len)
1040 {
1041 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1042 	struct group_filter gsf;
1043 	int num;
1044 	int err;
1045 
1046 	if (len < size0)
1047 		return -EINVAL;
1048 	if (copy_from_sockptr(&gsf, optval, size0))
1049 		return -EFAULT;
1050 	if (gsf.gf_group.ss_family != AF_INET6)
1051 		return -EADDRNOTAVAIL;
1052 	num = gsf.gf_numsrc;
1053 	sockopt_lock_sock(sk);
1054 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1055 	if (!err) {
1056 		if (num > gsf.gf_numsrc)
1057 			num = gsf.gf_numsrc;
1058 		len = GROUP_FILTER_SIZE(num);
1059 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1060 		    copy_to_sockptr(optval, &gsf, size0))
1061 			err = -EFAULT;
1062 	}
1063 	sockopt_release_sock(sk);
1064 	return err;
1065 }
1066 
1067 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1068 				    sockptr_t optlen, int len)
1069 {
1070 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1071 	struct compat_group_filter gf32;
1072 	struct group_filter gf;
1073 	int err;
1074 	int num;
1075 
1076 	if (len < size0)
1077 		return -EINVAL;
1078 
1079 	if (copy_from_sockptr(&gf32, optval, size0))
1080 		return -EFAULT;
1081 	gf.gf_interface = gf32.gf_interface;
1082 	gf.gf_fmode = gf32.gf_fmode;
1083 	num = gf.gf_numsrc = gf32.gf_numsrc;
1084 	gf.gf_group = gf32.gf_group;
1085 
1086 	if (gf.gf_group.ss_family != AF_INET6)
1087 		return -EADDRNOTAVAIL;
1088 
1089 	sockopt_lock_sock(sk);
1090 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1091 	sockopt_release_sock(sk);
1092 	if (err)
1093 		return err;
1094 	if (num > gf.gf_numsrc)
1095 		num = gf.gf_numsrc;
1096 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1097 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1098 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1099 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1100 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1101 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1102 		return -EFAULT;
1103 	return 0;
1104 }
1105 
1106 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1107 		       sockptr_t optval, sockptr_t optlen)
1108 {
1109 	struct ipv6_pinfo *np = inet6_sk(sk);
1110 	int len;
1111 	int val;
1112 
1113 	if (ip6_mroute_opt(optname))
1114 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1115 
1116 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1117 		return -EFAULT;
1118 	switch (optname) {
1119 	case IPV6_ADDRFORM:
1120 		if (sk->sk_protocol != IPPROTO_UDP &&
1121 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1122 		    sk->sk_protocol != IPPROTO_TCP)
1123 			return -ENOPROTOOPT;
1124 		if (sk->sk_state != TCP_ESTABLISHED)
1125 			return -ENOTCONN;
1126 		val = sk->sk_family;
1127 		break;
1128 	case MCAST_MSFILTER:
1129 		if (in_compat_syscall())
1130 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1131 		return ipv6_get_msfilter(sk, optval, optlen, len);
1132 	case IPV6_2292PKTOPTIONS:
1133 	{
1134 		struct msghdr msg;
1135 		struct sk_buff *skb;
1136 
1137 		if (sk->sk_type != SOCK_STREAM)
1138 			return -ENOPROTOOPT;
1139 
1140 		if (optval.is_kernel) {
1141 			msg.msg_control_is_user = false;
1142 			msg.msg_control = optval.kernel;
1143 		} else {
1144 			msg.msg_control_is_user = true;
1145 			msg.msg_control_user = optval.user;
1146 		}
1147 		msg.msg_controllen = len;
1148 		msg.msg_flags = 0;
1149 
1150 		sockopt_lock_sock(sk);
1151 		skb = np->pktoptions;
1152 		if (skb)
1153 			ip6_datagram_recv_ctl(sk, &msg, skb);
1154 		sockopt_release_sock(sk);
1155 		if (!skb) {
1156 			if (np->rxopt.bits.rxinfo) {
1157 				int mcast_oif = READ_ONCE(np->mcast_oif);
1158 				struct in6_pktinfo src_info;
1159 
1160 				src_info.ipi6_ifindex = mcast_oif ? :
1161 					np->sticky_pktinfo.ipi6_ifindex;
1162 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1163 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1164 			}
1165 			if (np->rxopt.bits.rxhlim) {
1166 				int hlim = READ_ONCE(np->mcast_hops);
1167 
1168 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1169 			}
1170 			if (np->rxopt.bits.rxtclass) {
1171 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1172 
1173 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1174 			}
1175 			if (np->rxopt.bits.rxoinfo) {
1176 				int mcast_oif = READ_ONCE(np->mcast_oif);
1177 				struct in6_pktinfo src_info;
1178 
1179 				src_info.ipi6_ifindex = mcast_oif ? :
1180 					np->sticky_pktinfo.ipi6_ifindex;
1181 				src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr :
1182 								 np->sticky_pktinfo.ipi6_addr;
1183 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1184 			}
1185 			if (np->rxopt.bits.rxohlim) {
1186 				int hlim = READ_ONCE(np->mcast_hops);
1187 
1188 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1189 			}
1190 			if (np->rxopt.bits.rxflow) {
1191 				__be32 flowinfo = np->rcv_flowinfo;
1192 
1193 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1194 			}
1195 		}
1196 		len -= msg.msg_controllen;
1197 		return copy_to_sockptr(optlen, &len, sizeof(int));
1198 	}
1199 	case IPV6_MTU:
1200 	{
1201 		struct dst_entry *dst;
1202 
1203 		val = 0;
1204 		rcu_read_lock();
1205 		dst = __sk_dst_get(sk);
1206 		if (dst)
1207 			val = dst_mtu(dst);
1208 		rcu_read_unlock();
1209 		if (!val)
1210 			return -ENOTCONN;
1211 		break;
1212 	}
1213 
1214 	case IPV6_V6ONLY:
1215 		val = sk->sk_ipv6only;
1216 		break;
1217 
1218 	case IPV6_RECVPKTINFO:
1219 		val = np->rxopt.bits.rxinfo;
1220 		break;
1221 
1222 	case IPV6_2292PKTINFO:
1223 		val = np->rxopt.bits.rxoinfo;
1224 		break;
1225 
1226 	case IPV6_RECVHOPLIMIT:
1227 		val = np->rxopt.bits.rxhlim;
1228 		break;
1229 
1230 	case IPV6_2292HOPLIMIT:
1231 		val = np->rxopt.bits.rxohlim;
1232 		break;
1233 
1234 	case IPV6_RECVRTHDR:
1235 		val = np->rxopt.bits.srcrt;
1236 		break;
1237 
1238 	case IPV6_2292RTHDR:
1239 		val = np->rxopt.bits.osrcrt;
1240 		break;
1241 
1242 	case IPV6_HOPOPTS:
1243 	case IPV6_RTHDRDSTOPTS:
1244 	case IPV6_RTHDR:
1245 	case IPV6_DSTOPTS:
1246 	{
1247 		struct ipv6_txoptions *opt;
1248 
1249 		sockopt_lock_sock(sk);
1250 		opt = rcu_dereference_protected(np->opt,
1251 						lockdep_sock_is_held(sk));
1252 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1253 		sockopt_release_sock(sk);
1254 		/* check if ipv6_getsockopt_sticky() returns err code */
1255 		if (len < 0)
1256 			return len;
1257 		return copy_to_sockptr(optlen, &len, sizeof(int));
1258 	}
1259 
1260 	case IPV6_RECVHOPOPTS:
1261 		val = np->rxopt.bits.hopopts;
1262 		break;
1263 
1264 	case IPV6_2292HOPOPTS:
1265 		val = np->rxopt.bits.ohopopts;
1266 		break;
1267 
1268 	case IPV6_RECVDSTOPTS:
1269 		val = np->rxopt.bits.dstopts;
1270 		break;
1271 
1272 	case IPV6_2292DSTOPTS:
1273 		val = np->rxopt.bits.odstopts;
1274 		break;
1275 
1276 	case IPV6_TCLASS:
1277 		val = np->tclass;
1278 		break;
1279 
1280 	case IPV6_RECVTCLASS:
1281 		val = np->rxopt.bits.rxtclass;
1282 		break;
1283 
1284 	case IPV6_FLOWINFO:
1285 		val = np->rxopt.bits.rxflow;
1286 		break;
1287 
1288 	case IPV6_RECVPATHMTU:
1289 		val = np->rxopt.bits.rxpmtu;
1290 		break;
1291 
1292 	case IPV6_PATHMTU:
1293 	{
1294 		struct dst_entry *dst;
1295 		struct ip6_mtuinfo mtuinfo;
1296 
1297 		if (len < sizeof(mtuinfo))
1298 			return -EINVAL;
1299 
1300 		len = sizeof(mtuinfo);
1301 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1302 
1303 		rcu_read_lock();
1304 		dst = __sk_dst_get(sk);
1305 		if (dst)
1306 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1307 		rcu_read_unlock();
1308 		if (!mtuinfo.ip6m_mtu)
1309 			return -ENOTCONN;
1310 
1311 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1312 			return -EFAULT;
1313 		if (copy_to_sockptr(optval, &mtuinfo, len))
1314 			return -EFAULT;
1315 
1316 		return 0;
1317 	}
1318 
1319 	case IPV6_TRANSPARENT:
1320 		val = inet_test_bit(TRANSPARENT, sk);
1321 		break;
1322 
1323 	case IPV6_FREEBIND:
1324 		val = inet_test_bit(FREEBIND, sk);
1325 		break;
1326 
1327 	case IPV6_RECVORIGDSTADDR:
1328 		val = np->rxopt.bits.rxorigdstaddr;
1329 		break;
1330 
1331 	case IPV6_UNICAST_HOPS:
1332 	case IPV6_MULTICAST_HOPS:
1333 	{
1334 		struct dst_entry *dst;
1335 
1336 		if (optname == IPV6_UNICAST_HOPS)
1337 			val = READ_ONCE(np->hop_limit);
1338 		else
1339 			val = READ_ONCE(np->mcast_hops);
1340 
1341 		if (val < 0) {
1342 			rcu_read_lock();
1343 			dst = __sk_dst_get(sk);
1344 			if (dst)
1345 				val = ip6_dst_hoplimit(dst);
1346 			rcu_read_unlock();
1347 		}
1348 
1349 		if (val < 0)
1350 			val = READ_ONCE(sock_net(sk)->ipv6.devconf_all->hop_limit);
1351 		break;
1352 	}
1353 
1354 	case IPV6_MULTICAST_LOOP:
1355 		val = inet6_test_bit(MC6_LOOP, sk);
1356 		break;
1357 
1358 	case IPV6_MULTICAST_IF:
1359 		val = READ_ONCE(np->mcast_oif);
1360 		break;
1361 
1362 	case IPV6_MULTICAST_ALL:
1363 		val = inet6_test_bit(MC6_ALL, sk);
1364 		break;
1365 
1366 	case IPV6_UNICAST_IF:
1367 		val = (__force int)htonl((__u32) READ_ONCE(np->ucast_oif));
1368 		break;
1369 
1370 	case IPV6_MTU_DISCOVER:
1371 		val = READ_ONCE(np->pmtudisc);
1372 		break;
1373 
1374 	case IPV6_RECVERR:
1375 		val = inet6_test_bit(RECVERR6, sk);
1376 		break;
1377 
1378 	case IPV6_FLOWINFO_SEND:
1379 		val = inet6_test_bit(SNDFLOW, sk);
1380 		break;
1381 
1382 	case IPV6_FLOWLABEL_MGR:
1383 	{
1384 		struct in6_flowlabel_req freq;
1385 		int flags;
1386 
1387 		if (len < sizeof(freq))
1388 			return -EINVAL;
1389 
1390 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1391 			return -EFAULT;
1392 
1393 		if (freq.flr_action != IPV6_FL_A_GET)
1394 			return -EINVAL;
1395 
1396 		len = sizeof(freq);
1397 		flags = freq.flr_flags;
1398 
1399 		memset(&freq, 0, sizeof(freq));
1400 
1401 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1402 		if (val < 0)
1403 			return val;
1404 
1405 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1406 			return -EFAULT;
1407 		if (copy_to_sockptr(optval, &freq, len))
1408 			return -EFAULT;
1409 
1410 		return 0;
1411 	}
1412 
1413 	case IPV6_ADDR_PREFERENCES:
1414 		{
1415 		u8 srcprefs = READ_ONCE(np->srcprefs);
1416 		val = 0;
1417 
1418 		if (srcprefs & IPV6_PREFER_SRC_TMP)
1419 			val |= IPV6_PREFER_SRC_TMP;
1420 		else if (srcprefs & IPV6_PREFER_SRC_PUBLIC)
1421 			val |= IPV6_PREFER_SRC_PUBLIC;
1422 		else {
1423 			/* XXX: should we return system default? */
1424 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1425 		}
1426 
1427 		if (srcprefs & IPV6_PREFER_SRC_COA)
1428 			val |= IPV6_PREFER_SRC_COA;
1429 		else
1430 			val |= IPV6_PREFER_SRC_HOME;
1431 		break;
1432 		}
1433 	case IPV6_MINHOPCOUNT:
1434 		val = READ_ONCE(np->min_hopcount);
1435 		break;
1436 
1437 	case IPV6_DONTFRAG:
1438 		val = inet6_test_bit(DONTFRAG, sk);
1439 		break;
1440 
1441 	case IPV6_AUTOFLOWLABEL:
1442 		val = ip6_autoflowlabel(sock_net(sk), sk);
1443 		break;
1444 
1445 	case IPV6_RECVFRAGSIZE:
1446 		val = np->rxopt.bits.recvfragsize;
1447 		break;
1448 
1449 	case IPV6_ROUTER_ALERT:
1450 		val = inet6_test_bit(RTALERT, sk);
1451 		break;
1452 
1453 	case IPV6_ROUTER_ALERT_ISOLATE:
1454 		val = inet6_test_bit(RTALERT_ISOLATE, sk);
1455 		break;
1456 
1457 	case IPV6_RECVERR_RFC4884:
1458 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1459 		break;
1460 
1461 	default:
1462 		return -ENOPROTOOPT;
1463 	}
1464 	len = min_t(unsigned int, sizeof(int), len);
1465 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1466 		return -EFAULT;
1467 	if (copy_to_sockptr(optval, &val, len))
1468 		return -EFAULT;
1469 	return 0;
1470 }
1471 
1472 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1473 		    char __user *optval, int __user *optlen)
1474 {
1475 	int err;
1476 
1477 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1478 		return ip_getsockopt(sk, level, optname, optval, optlen);
1479 
1480 	if (level != SOL_IPV6)
1481 		return -ENOPROTOOPT;
1482 
1483 	err = do_ipv6_getsockopt(sk, level, optname,
1484 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1485 #ifdef CONFIG_NETFILTER
1486 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1487 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1488 		int len;
1489 
1490 		if (get_user(len, optlen))
1491 			return -EFAULT;
1492 
1493 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1494 		if (err >= 0)
1495 			err = put_user(len, optlen);
1496 	}
1497 #endif
1498 	return err;
1499 }
1500 EXPORT_SYMBOL(ipv6_getsockopt);
1501