xref: /openbmc/linux/net/ipv6/ipv6_sockglue.c (revision 1b36955c)
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 	if (needs_rtnl)
419 		rtnl_lock();
420 	sockopt_lock_sock(sk);
421 
422 	/* Another thread has converted the socket into IPv4 with
423 	 * IPV6_ADDRFORM concurrently.
424 	 */
425 	if (unlikely(sk->sk_family != AF_INET6))
426 		goto unlock;
427 
428 	switch (optname) {
429 
430 	case IPV6_ADDRFORM:
431 		if (optlen < sizeof(int))
432 			goto e_inval;
433 		if (val == PF_INET) {
434 			if (sk->sk_type == SOCK_RAW)
435 				break;
436 
437 			if (sk->sk_protocol == IPPROTO_UDP ||
438 			    sk->sk_protocol == IPPROTO_UDPLITE) {
439 				struct udp_sock *up = udp_sk(sk);
440 				if (up->pending == AF_INET6) {
441 					retv = -EBUSY;
442 					break;
443 				}
444 			} else if (sk->sk_protocol == IPPROTO_TCP) {
445 				if (sk->sk_prot != &tcpv6_prot) {
446 					retv = -EBUSY;
447 					break;
448 				}
449 			} else {
450 				break;
451 			}
452 
453 			if (sk->sk_state != TCP_ESTABLISHED) {
454 				retv = -ENOTCONN;
455 				break;
456 			}
457 
458 			if (ipv6_only_sock(sk) ||
459 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
460 				retv = -EADDRNOTAVAIL;
461 				break;
462 			}
463 
464 			__ipv6_sock_mc_close(sk);
465 			__ipv6_sock_ac_close(sk);
466 
467 			if (sk->sk_protocol == IPPROTO_TCP) {
468 				struct inet_connection_sock *icsk = inet_csk(sk);
469 
470 				sock_prot_inuse_add(net, sk->sk_prot, -1);
471 				sock_prot_inuse_add(net, &tcp_prot, 1);
472 
473 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
474 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
475 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
476 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
477 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
478 				WRITE_ONCE(sk->sk_family, PF_INET);
479 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
480 			} else {
481 				struct proto *prot = &udp_prot;
482 
483 				if (sk->sk_protocol == IPPROTO_UDPLITE)
484 					prot = &udplite_prot;
485 
486 				sock_prot_inuse_add(net, sk->sk_prot, -1);
487 				sock_prot_inuse_add(net, prot, 1);
488 
489 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
490 				WRITE_ONCE(sk->sk_prot, prot);
491 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
492 				WRITE_ONCE(sk->sk_family, PF_INET);
493 			}
494 
495 			/* Disable all options not to allocate memory anymore,
496 			 * but there is still a race.  See the lockless path
497 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
498 			 */
499 			np->rxopt.all = 0;
500 
501 			inet6_cleanup_sock(sk);
502 
503 			module_put(THIS_MODULE);
504 			retv = 0;
505 			break;
506 		}
507 		goto e_inval;
508 
509 	case IPV6_V6ONLY:
510 		if (optlen < sizeof(int) ||
511 		    inet_sk(sk)->inet_num)
512 			goto e_inval;
513 		sk->sk_ipv6only = valbool;
514 		retv = 0;
515 		break;
516 
517 	case IPV6_RECVPKTINFO:
518 		if (optlen < sizeof(int))
519 			goto e_inval;
520 		np->rxopt.bits.rxinfo = valbool;
521 		retv = 0;
522 		break;
523 
524 	case IPV6_2292PKTINFO:
525 		if (optlen < sizeof(int))
526 			goto e_inval;
527 		np->rxopt.bits.rxoinfo = valbool;
528 		retv = 0;
529 		break;
530 
531 	case IPV6_RECVHOPLIMIT:
532 		if (optlen < sizeof(int))
533 			goto e_inval;
534 		np->rxopt.bits.rxhlim = valbool;
535 		retv = 0;
536 		break;
537 
538 	case IPV6_2292HOPLIMIT:
539 		if (optlen < sizeof(int))
540 			goto e_inval;
541 		np->rxopt.bits.rxohlim = valbool;
542 		retv = 0;
543 		break;
544 
545 	case IPV6_RECVRTHDR:
546 		if (optlen < sizeof(int))
547 			goto e_inval;
548 		np->rxopt.bits.srcrt = valbool;
549 		retv = 0;
550 		break;
551 
552 	case IPV6_2292RTHDR:
553 		if (optlen < sizeof(int))
554 			goto e_inval;
555 		np->rxopt.bits.osrcrt = valbool;
556 		retv = 0;
557 		break;
558 
559 	case IPV6_RECVHOPOPTS:
560 		if (optlen < sizeof(int))
561 			goto e_inval;
562 		np->rxopt.bits.hopopts = valbool;
563 		retv = 0;
564 		break;
565 
566 	case IPV6_2292HOPOPTS:
567 		if (optlen < sizeof(int))
568 			goto e_inval;
569 		np->rxopt.bits.ohopopts = valbool;
570 		retv = 0;
571 		break;
572 
573 	case IPV6_RECVDSTOPTS:
574 		if (optlen < sizeof(int))
575 			goto e_inval;
576 		np->rxopt.bits.dstopts = valbool;
577 		retv = 0;
578 		break;
579 
580 	case IPV6_2292DSTOPTS:
581 		if (optlen < sizeof(int))
582 			goto e_inval;
583 		np->rxopt.bits.odstopts = valbool;
584 		retv = 0;
585 		break;
586 
587 	case IPV6_TCLASS:
588 		if (optlen < sizeof(int))
589 			goto e_inval;
590 		if (val < -1 || val > 0xff)
591 			goto e_inval;
592 		/* RFC 3542, 6.5: default traffic class of 0x0 */
593 		if (val == -1)
594 			val = 0;
595 		if (sk->sk_type == SOCK_STREAM) {
596 			val &= ~INET_ECN_MASK;
597 			val |= np->tclass & INET_ECN_MASK;
598 		}
599 		if (np->tclass != val) {
600 			np->tclass = val;
601 			sk_dst_reset(sk);
602 		}
603 		retv = 0;
604 		break;
605 
606 	case IPV6_RECVTCLASS:
607 		if (optlen < sizeof(int))
608 			goto e_inval;
609 		np->rxopt.bits.rxtclass = valbool;
610 		retv = 0;
611 		break;
612 
613 	case IPV6_FLOWINFO:
614 		if (optlen < sizeof(int))
615 			goto e_inval;
616 		np->rxopt.bits.rxflow = valbool;
617 		retv = 0;
618 		break;
619 
620 	case IPV6_RECVPATHMTU:
621 		if (optlen < sizeof(int))
622 			goto e_inval;
623 		np->rxopt.bits.rxpmtu = valbool;
624 		retv = 0;
625 		break;
626 
627 	case IPV6_TRANSPARENT:
628 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
629 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
630 			retv = -EPERM;
631 			break;
632 		}
633 		if (optlen < sizeof(int))
634 			goto e_inval;
635 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
636 		inet_assign_bit(TRANSPARENT, sk, valbool);
637 		retv = 0;
638 		break;
639 
640 	case IPV6_FREEBIND:
641 		if (optlen < sizeof(int))
642 			goto e_inval;
643 		/* we also don't have a separate freebind bit for IPV6 */
644 		inet_assign_bit(FREEBIND, sk, valbool);
645 		retv = 0;
646 		break;
647 
648 	case IPV6_RECVORIGDSTADDR:
649 		if (optlen < sizeof(int))
650 			goto e_inval;
651 		np->rxopt.bits.rxorigdstaddr = valbool;
652 		retv = 0;
653 		break;
654 
655 	case IPV6_HOPOPTS:
656 	case IPV6_RTHDRDSTOPTS:
657 	case IPV6_RTHDR:
658 	case IPV6_DSTOPTS:
659 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
660 		break;
661 
662 	case IPV6_PKTINFO:
663 	{
664 		struct in6_pktinfo pkt;
665 
666 		if (optlen == 0)
667 			goto e_inval;
668 		else if (optlen < sizeof(struct in6_pktinfo) ||
669 			 sockptr_is_null(optval))
670 			goto e_inval;
671 
672 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
673 			retv = -EFAULT;
674 			break;
675 		}
676 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
677 			goto e_inval;
678 
679 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
680 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
681 		retv = 0;
682 		break;
683 	}
684 
685 	case IPV6_2292PKTOPTIONS:
686 	{
687 		struct ipv6_txoptions *opt = NULL;
688 		struct msghdr msg;
689 		struct flowi6 fl6;
690 		struct ipcm6_cookie ipc6;
691 
692 		memset(&fl6, 0, sizeof(fl6));
693 		fl6.flowi6_oif = sk->sk_bound_dev_if;
694 		fl6.flowi6_mark = sk->sk_mark;
695 
696 		if (optlen == 0)
697 			goto update;
698 
699 		/* 1K is probably excessive
700 		 * 1K is surely not enough, 2K per standard header is 16K.
701 		 */
702 		retv = -EINVAL;
703 		if (optlen > 64*1024)
704 			break;
705 
706 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
707 		retv = -ENOBUFS;
708 		if (!opt)
709 			break;
710 
711 		memset(opt, 0, sizeof(*opt));
712 		refcount_set(&opt->refcnt, 1);
713 		opt->tot_len = sizeof(*opt) + optlen;
714 		retv = -EFAULT;
715 		if (copy_from_sockptr(opt + 1, optval, optlen))
716 			goto done;
717 
718 		msg.msg_controllen = optlen;
719 		msg.msg_control_is_user = false;
720 		msg.msg_control = (void *)(opt+1);
721 		ipc6.opt = opt;
722 
723 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
724 		if (retv)
725 			goto done;
726 update:
727 		retv = 0;
728 		opt = ipv6_update_options(sk, opt);
729 done:
730 		if (opt) {
731 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
732 			txopt_put(opt);
733 		}
734 		break;
735 	}
736 	case IPV6_UNICAST_HOPS:
737 		if (optlen < sizeof(int))
738 			goto e_inval;
739 		if (val > 255 || val < -1)
740 			goto e_inval;
741 		np->hop_limit = val;
742 		retv = 0;
743 		break;
744 
745 	case IPV6_MULTICAST_HOPS:
746 		if (sk->sk_type == SOCK_STREAM)
747 			break;
748 		if (optlen < sizeof(int))
749 			goto e_inval;
750 		if (val > 255 || val < -1)
751 			goto e_inval;
752 		np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
753 		retv = 0;
754 		break;
755 
756 	case IPV6_MULTICAST_LOOP:
757 		if (optlen < sizeof(int))
758 			goto e_inval;
759 		if (val != valbool)
760 			goto e_inval;
761 		np->mc_loop = valbool;
762 		retv = 0;
763 		break;
764 
765 	case IPV6_UNICAST_IF:
766 	{
767 		struct net_device *dev = NULL;
768 		int ifindex;
769 
770 		if (optlen != sizeof(int))
771 			goto e_inval;
772 
773 		ifindex = (__force int)ntohl((__force __be32)val);
774 		if (ifindex == 0) {
775 			np->ucast_oif = 0;
776 			retv = 0;
777 			break;
778 		}
779 
780 		dev = dev_get_by_index(net, ifindex);
781 		retv = -EADDRNOTAVAIL;
782 		if (!dev)
783 			break;
784 		dev_put(dev);
785 
786 		retv = -EINVAL;
787 		if (sk->sk_bound_dev_if)
788 			break;
789 
790 		np->ucast_oif = ifindex;
791 		retv = 0;
792 		break;
793 	}
794 
795 	case IPV6_MULTICAST_IF:
796 		if (sk->sk_type == SOCK_STREAM)
797 			break;
798 		if (optlen < sizeof(int))
799 			goto e_inval;
800 
801 		if (val) {
802 			struct net_device *dev;
803 			int midx;
804 
805 			rcu_read_lock();
806 
807 			dev = dev_get_by_index_rcu(net, val);
808 			if (!dev) {
809 				rcu_read_unlock();
810 				retv = -ENODEV;
811 				break;
812 			}
813 			midx = l3mdev_master_ifindex_rcu(dev);
814 
815 			rcu_read_unlock();
816 
817 			if (sk->sk_bound_dev_if &&
818 			    sk->sk_bound_dev_if != val &&
819 			    (!midx || midx != sk->sk_bound_dev_if))
820 				goto e_inval;
821 		}
822 		np->mcast_oif = val;
823 		retv = 0;
824 		break;
825 	case IPV6_ADD_MEMBERSHIP:
826 	case IPV6_DROP_MEMBERSHIP:
827 	{
828 		struct ipv6_mreq mreq;
829 
830 		if (optlen < sizeof(struct ipv6_mreq))
831 			goto e_inval;
832 
833 		retv = -EPROTO;
834 		if (inet_test_bit(IS_ICSK, sk))
835 			break;
836 
837 		retv = -EFAULT;
838 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
839 			break;
840 
841 		if (optname == IPV6_ADD_MEMBERSHIP)
842 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
843 		else
844 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
845 		break;
846 	}
847 	case IPV6_JOIN_ANYCAST:
848 	case IPV6_LEAVE_ANYCAST:
849 	{
850 		struct ipv6_mreq mreq;
851 
852 		if (optlen < sizeof(struct ipv6_mreq))
853 			goto e_inval;
854 
855 		retv = -EFAULT;
856 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
857 			break;
858 
859 		if (optname == IPV6_JOIN_ANYCAST)
860 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
861 		else
862 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
863 		break;
864 	}
865 	case IPV6_MULTICAST_ALL:
866 		if (optlen < sizeof(int))
867 			goto e_inval;
868 		np->mc_all = valbool;
869 		retv = 0;
870 		break;
871 
872 	case MCAST_JOIN_GROUP:
873 	case MCAST_LEAVE_GROUP:
874 		if (in_compat_syscall())
875 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
876 							    optlen);
877 		else
878 			retv = ipv6_mcast_join_leave(sk, optname, optval,
879 						     optlen);
880 		break;
881 	case MCAST_JOIN_SOURCE_GROUP:
882 	case MCAST_LEAVE_SOURCE_GROUP:
883 	case MCAST_BLOCK_SOURCE:
884 	case MCAST_UNBLOCK_SOURCE:
885 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
886 		break;
887 	case MCAST_MSFILTER:
888 		if (in_compat_syscall())
889 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
890 							      optlen);
891 		else
892 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
893 		break;
894 	case IPV6_ROUTER_ALERT:
895 		if (optlen < sizeof(int))
896 			goto e_inval;
897 		retv = ip6_ra_control(sk, val);
898 		break;
899 	case IPV6_ROUTER_ALERT_ISOLATE:
900 		if (optlen < sizeof(int))
901 			goto e_inval;
902 		np->rtalert_isolate = valbool;
903 		retv = 0;
904 		break;
905 	case IPV6_MTU_DISCOVER:
906 		if (optlen < sizeof(int))
907 			goto e_inval;
908 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
909 			goto e_inval;
910 		np->pmtudisc = val;
911 		retv = 0;
912 		break;
913 	case IPV6_MTU:
914 		if (optlen < sizeof(int))
915 			goto e_inval;
916 		if (val && val < IPV6_MIN_MTU)
917 			goto e_inval;
918 		np->frag_size = val;
919 		retv = 0;
920 		break;
921 	case IPV6_RECVERR:
922 		if (optlen < sizeof(int))
923 			goto e_inval;
924 		np->recverr = valbool;
925 		if (!val)
926 			skb_errqueue_purge(&sk->sk_error_queue);
927 		retv = 0;
928 		break;
929 	case IPV6_FLOWINFO_SEND:
930 		if (optlen < sizeof(int))
931 			goto e_inval;
932 		np->sndflow = valbool;
933 		retv = 0;
934 		break;
935 	case IPV6_FLOWLABEL_MGR:
936 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
937 		break;
938 	case IPV6_IPSEC_POLICY:
939 	case IPV6_XFRM_POLICY:
940 		retv = -EPERM;
941 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
942 			break;
943 		retv = xfrm_user_policy(sk, optname, optval, optlen);
944 		break;
945 
946 	case IPV6_ADDR_PREFERENCES:
947 		if (optlen < sizeof(int))
948 			goto e_inval;
949 		retv = __ip6_sock_set_addr_preferences(sk, val);
950 		break;
951 	case IPV6_MINHOPCOUNT:
952 		if (optlen < sizeof(int))
953 			goto e_inval;
954 		if (val < 0 || val > 255)
955 			goto e_inval;
956 
957 		if (val)
958 			static_branch_enable(&ip6_min_hopcount);
959 
960 		/* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
961 		 * while we are changing it.
962 		 */
963 		WRITE_ONCE(np->min_hopcount, val);
964 		retv = 0;
965 		break;
966 	case IPV6_DONTFRAG:
967 		np->dontfrag = valbool;
968 		retv = 0;
969 		break;
970 	case IPV6_AUTOFLOWLABEL:
971 		np->autoflowlabel = valbool;
972 		np->autoflowlabel_set = 1;
973 		retv = 0;
974 		break;
975 	case IPV6_RECVFRAGSIZE:
976 		np->rxopt.bits.recvfragsize = valbool;
977 		retv = 0;
978 		break;
979 	case IPV6_RECVERR_RFC4884:
980 		if (optlen < sizeof(int))
981 			goto e_inval;
982 		if (val < 0 || val > 1)
983 			goto e_inval;
984 		np->recverr_rfc4884 = valbool;
985 		retv = 0;
986 		break;
987 	}
988 
989 unlock:
990 	sockopt_release_sock(sk);
991 	if (needs_rtnl)
992 		rtnl_unlock();
993 
994 	return retv;
995 
996 e_inval:
997 	retv = -EINVAL;
998 	goto unlock;
999 }
1000 
1001 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1002 		    unsigned int optlen)
1003 {
1004 	int err;
1005 
1006 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1007 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1008 
1009 	if (level != SOL_IPV6)
1010 		return -ENOPROTOOPT;
1011 
1012 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1013 #ifdef CONFIG_NETFILTER
1014 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1015 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1016 			optname != IPV6_XFRM_POLICY)
1017 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1018 #endif
1019 	return err;
1020 }
1021 EXPORT_SYMBOL(ipv6_setsockopt);
1022 
1023 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1024 				  int optname, sockptr_t optval, int len)
1025 {
1026 	struct ipv6_opt_hdr *hdr;
1027 
1028 	if (!opt)
1029 		return 0;
1030 
1031 	switch (optname) {
1032 	case IPV6_HOPOPTS:
1033 		hdr = opt->hopopt;
1034 		break;
1035 	case IPV6_RTHDRDSTOPTS:
1036 		hdr = opt->dst0opt;
1037 		break;
1038 	case IPV6_RTHDR:
1039 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1040 		break;
1041 	case IPV6_DSTOPTS:
1042 		hdr = opt->dst1opt;
1043 		break;
1044 	default:
1045 		return -EINVAL;	/* should not happen */
1046 	}
1047 
1048 	if (!hdr)
1049 		return 0;
1050 
1051 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1052 	if (copy_to_sockptr(optval, hdr, len))
1053 		return -EFAULT;
1054 	return len;
1055 }
1056 
1057 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1058 			     sockptr_t optlen, int len)
1059 {
1060 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1061 	struct group_filter gsf;
1062 	int num;
1063 	int err;
1064 
1065 	if (len < size0)
1066 		return -EINVAL;
1067 	if (copy_from_sockptr(&gsf, optval, size0))
1068 		return -EFAULT;
1069 	if (gsf.gf_group.ss_family != AF_INET6)
1070 		return -EADDRNOTAVAIL;
1071 	num = gsf.gf_numsrc;
1072 	sockopt_lock_sock(sk);
1073 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1074 	if (!err) {
1075 		if (num > gsf.gf_numsrc)
1076 			num = gsf.gf_numsrc;
1077 		len = GROUP_FILTER_SIZE(num);
1078 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1079 		    copy_to_sockptr(optval, &gsf, size0))
1080 			err = -EFAULT;
1081 	}
1082 	sockopt_release_sock(sk);
1083 	return err;
1084 }
1085 
1086 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1087 				    sockptr_t optlen, int len)
1088 {
1089 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1090 	struct compat_group_filter gf32;
1091 	struct group_filter gf;
1092 	int err;
1093 	int num;
1094 
1095 	if (len < size0)
1096 		return -EINVAL;
1097 
1098 	if (copy_from_sockptr(&gf32, optval, size0))
1099 		return -EFAULT;
1100 	gf.gf_interface = gf32.gf_interface;
1101 	gf.gf_fmode = gf32.gf_fmode;
1102 	num = gf.gf_numsrc = gf32.gf_numsrc;
1103 	gf.gf_group = gf32.gf_group;
1104 
1105 	if (gf.gf_group.ss_family != AF_INET6)
1106 		return -EADDRNOTAVAIL;
1107 
1108 	sockopt_lock_sock(sk);
1109 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1110 	sockopt_release_sock(sk);
1111 	if (err)
1112 		return err;
1113 	if (num > gf.gf_numsrc)
1114 		num = gf.gf_numsrc;
1115 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1116 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1117 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1118 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1119 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1120 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1121 		return -EFAULT;
1122 	return 0;
1123 }
1124 
1125 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1126 		       sockptr_t optval, sockptr_t optlen)
1127 {
1128 	struct ipv6_pinfo *np = inet6_sk(sk);
1129 	int len;
1130 	int val;
1131 
1132 	if (ip6_mroute_opt(optname))
1133 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1134 
1135 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1136 		return -EFAULT;
1137 	switch (optname) {
1138 	case IPV6_ADDRFORM:
1139 		if (sk->sk_protocol != IPPROTO_UDP &&
1140 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1141 		    sk->sk_protocol != IPPROTO_TCP)
1142 			return -ENOPROTOOPT;
1143 		if (sk->sk_state != TCP_ESTABLISHED)
1144 			return -ENOTCONN;
1145 		val = sk->sk_family;
1146 		break;
1147 	case MCAST_MSFILTER:
1148 		if (in_compat_syscall())
1149 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1150 		return ipv6_get_msfilter(sk, optval, optlen, len);
1151 	case IPV6_2292PKTOPTIONS:
1152 	{
1153 		struct msghdr msg;
1154 		struct sk_buff *skb;
1155 
1156 		if (sk->sk_type != SOCK_STREAM)
1157 			return -ENOPROTOOPT;
1158 
1159 		if (optval.is_kernel) {
1160 			msg.msg_control_is_user = false;
1161 			msg.msg_control = optval.kernel;
1162 		} else {
1163 			msg.msg_control_is_user = true;
1164 			msg.msg_control_user = optval.user;
1165 		}
1166 		msg.msg_controllen = len;
1167 		msg.msg_flags = 0;
1168 
1169 		sockopt_lock_sock(sk);
1170 		skb = np->pktoptions;
1171 		if (skb)
1172 			ip6_datagram_recv_ctl(sk, &msg, skb);
1173 		sockopt_release_sock(sk);
1174 		if (!skb) {
1175 			if (np->rxopt.bits.rxinfo) {
1176 				struct in6_pktinfo src_info;
1177 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1178 					np->sticky_pktinfo.ipi6_ifindex;
1179 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1180 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1181 			}
1182 			if (np->rxopt.bits.rxhlim) {
1183 				int hlim = np->mcast_hops;
1184 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1185 			}
1186 			if (np->rxopt.bits.rxtclass) {
1187 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1188 
1189 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1190 			}
1191 			if (np->rxopt.bits.rxoinfo) {
1192 				struct in6_pktinfo src_info;
1193 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1194 					np->sticky_pktinfo.ipi6_ifindex;
1195 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1196 								     np->sticky_pktinfo.ipi6_addr;
1197 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1198 			}
1199 			if (np->rxopt.bits.rxohlim) {
1200 				int hlim = np->mcast_hops;
1201 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1202 			}
1203 			if (np->rxopt.bits.rxflow) {
1204 				__be32 flowinfo = np->rcv_flowinfo;
1205 
1206 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1207 			}
1208 		}
1209 		len -= msg.msg_controllen;
1210 		return copy_to_sockptr(optlen, &len, sizeof(int));
1211 	}
1212 	case IPV6_MTU:
1213 	{
1214 		struct dst_entry *dst;
1215 
1216 		val = 0;
1217 		rcu_read_lock();
1218 		dst = __sk_dst_get(sk);
1219 		if (dst)
1220 			val = dst_mtu(dst);
1221 		rcu_read_unlock();
1222 		if (!val)
1223 			return -ENOTCONN;
1224 		break;
1225 	}
1226 
1227 	case IPV6_V6ONLY:
1228 		val = sk->sk_ipv6only;
1229 		break;
1230 
1231 	case IPV6_RECVPKTINFO:
1232 		val = np->rxopt.bits.rxinfo;
1233 		break;
1234 
1235 	case IPV6_2292PKTINFO:
1236 		val = np->rxopt.bits.rxoinfo;
1237 		break;
1238 
1239 	case IPV6_RECVHOPLIMIT:
1240 		val = np->rxopt.bits.rxhlim;
1241 		break;
1242 
1243 	case IPV6_2292HOPLIMIT:
1244 		val = np->rxopt.bits.rxohlim;
1245 		break;
1246 
1247 	case IPV6_RECVRTHDR:
1248 		val = np->rxopt.bits.srcrt;
1249 		break;
1250 
1251 	case IPV6_2292RTHDR:
1252 		val = np->rxopt.bits.osrcrt;
1253 		break;
1254 
1255 	case IPV6_HOPOPTS:
1256 	case IPV6_RTHDRDSTOPTS:
1257 	case IPV6_RTHDR:
1258 	case IPV6_DSTOPTS:
1259 	{
1260 		struct ipv6_txoptions *opt;
1261 
1262 		sockopt_lock_sock(sk);
1263 		opt = rcu_dereference_protected(np->opt,
1264 						lockdep_sock_is_held(sk));
1265 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1266 		sockopt_release_sock(sk);
1267 		/* check if ipv6_getsockopt_sticky() returns err code */
1268 		if (len < 0)
1269 			return len;
1270 		return copy_to_sockptr(optlen, &len, sizeof(int));
1271 	}
1272 
1273 	case IPV6_RECVHOPOPTS:
1274 		val = np->rxopt.bits.hopopts;
1275 		break;
1276 
1277 	case IPV6_2292HOPOPTS:
1278 		val = np->rxopt.bits.ohopopts;
1279 		break;
1280 
1281 	case IPV6_RECVDSTOPTS:
1282 		val = np->rxopt.bits.dstopts;
1283 		break;
1284 
1285 	case IPV6_2292DSTOPTS:
1286 		val = np->rxopt.bits.odstopts;
1287 		break;
1288 
1289 	case IPV6_TCLASS:
1290 		val = np->tclass;
1291 		break;
1292 
1293 	case IPV6_RECVTCLASS:
1294 		val = np->rxopt.bits.rxtclass;
1295 		break;
1296 
1297 	case IPV6_FLOWINFO:
1298 		val = np->rxopt.bits.rxflow;
1299 		break;
1300 
1301 	case IPV6_RECVPATHMTU:
1302 		val = np->rxopt.bits.rxpmtu;
1303 		break;
1304 
1305 	case IPV6_PATHMTU:
1306 	{
1307 		struct dst_entry *dst;
1308 		struct ip6_mtuinfo mtuinfo;
1309 
1310 		if (len < sizeof(mtuinfo))
1311 			return -EINVAL;
1312 
1313 		len = sizeof(mtuinfo);
1314 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1315 
1316 		rcu_read_lock();
1317 		dst = __sk_dst_get(sk);
1318 		if (dst)
1319 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1320 		rcu_read_unlock();
1321 		if (!mtuinfo.ip6m_mtu)
1322 			return -ENOTCONN;
1323 
1324 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1325 			return -EFAULT;
1326 		if (copy_to_sockptr(optval, &mtuinfo, len))
1327 			return -EFAULT;
1328 
1329 		return 0;
1330 	}
1331 
1332 	case IPV6_TRANSPARENT:
1333 		val = inet_test_bit(TRANSPARENT, sk);
1334 		break;
1335 
1336 	case IPV6_FREEBIND:
1337 		val = inet_test_bit(FREEBIND, sk);
1338 		break;
1339 
1340 	case IPV6_RECVORIGDSTADDR:
1341 		val = np->rxopt.bits.rxorigdstaddr;
1342 		break;
1343 
1344 	case IPV6_UNICAST_HOPS:
1345 	case IPV6_MULTICAST_HOPS:
1346 	{
1347 		struct dst_entry *dst;
1348 
1349 		if (optname == IPV6_UNICAST_HOPS)
1350 			val = np->hop_limit;
1351 		else
1352 			val = np->mcast_hops;
1353 
1354 		if (val < 0) {
1355 			rcu_read_lock();
1356 			dst = __sk_dst_get(sk);
1357 			if (dst)
1358 				val = ip6_dst_hoplimit(dst);
1359 			rcu_read_unlock();
1360 		}
1361 
1362 		if (val < 0)
1363 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1364 		break;
1365 	}
1366 
1367 	case IPV6_MULTICAST_LOOP:
1368 		val = np->mc_loop;
1369 		break;
1370 
1371 	case IPV6_MULTICAST_IF:
1372 		val = np->mcast_oif;
1373 		break;
1374 
1375 	case IPV6_MULTICAST_ALL:
1376 		val = np->mc_all;
1377 		break;
1378 
1379 	case IPV6_UNICAST_IF:
1380 		val = (__force int)htonl((__u32) np->ucast_oif);
1381 		break;
1382 
1383 	case IPV6_MTU_DISCOVER:
1384 		val = np->pmtudisc;
1385 		break;
1386 
1387 	case IPV6_RECVERR:
1388 		val = np->recverr;
1389 		break;
1390 
1391 	case IPV6_FLOWINFO_SEND:
1392 		val = np->sndflow;
1393 		break;
1394 
1395 	case IPV6_FLOWLABEL_MGR:
1396 	{
1397 		struct in6_flowlabel_req freq;
1398 		int flags;
1399 
1400 		if (len < sizeof(freq))
1401 			return -EINVAL;
1402 
1403 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1404 			return -EFAULT;
1405 
1406 		if (freq.flr_action != IPV6_FL_A_GET)
1407 			return -EINVAL;
1408 
1409 		len = sizeof(freq);
1410 		flags = freq.flr_flags;
1411 
1412 		memset(&freq, 0, sizeof(freq));
1413 
1414 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1415 		if (val < 0)
1416 			return val;
1417 
1418 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1419 			return -EFAULT;
1420 		if (copy_to_sockptr(optval, &freq, len))
1421 			return -EFAULT;
1422 
1423 		return 0;
1424 	}
1425 
1426 	case IPV6_ADDR_PREFERENCES:
1427 		val = 0;
1428 
1429 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1430 			val |= IPV6_PREFER_SRC_TMP;
1431 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1432 			val |= IPV6_PREFER_SRC_PUBLIC;
1433 		else {
1434 			/* XXX: should we return system default? */
1435 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1436 		}
1437 
1438 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1439 			val |= IPV6_PREFER_SRC_COA;
1440 		else
1441 			val |= IPV6_PREFER_SRC_HOME;
1442 		break;
1443 
1444 	case IPV6_MINHOPCOUNT:
1445 		val = np->min_hopcount;
1446 		break;
1447 
1448 	case IPV6_DONTFRAG:
1449 		val = np->dontfrag;
1450 		break;
1451 
1452 	case IPV6_AUTOFLOWLABEL:
1453 		val = ip6_autoflowlabel(sock_net(sk), np);
1454 		break;
1455 
1456 	case IPV6_RECVFRAGSIZE:
1457 		val = np->rxopt.bits.recvfragsize;
1458 		break;
1459 
1460 	case IPV6_ROUTER_ALERT_ISOLATE:
1461 		val = np->rtalert_isolate;
1462 		break;
1463 
1464 	case IPV6_RECVERR_RFC4884:
1465 		val = np->recverr_rfc4884;
1466 		break;
1467 
1468 	default:
1469 		return -ENOPROTOOPT;
1470 	}
1471 	len = min_t(unsigned int, sizeof(int), len);
1472 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1473 		return -EFAULT;
1474 	if (copy_to_sockptr(optval, &val, len))
1475 		return -EFAULT;
1476 	return 0;
1477 }
1478 
1479 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1480 		    char __user *optval, int __user *optlen)
1481 {
1482 	int err;
1483 
1484 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1485 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1486 
1487 	if (level != SOL_IPV6)
1488 		return -ENOPROTOOPT;
1489 
1490 	err = do_ipv6_getsockopt(sk, level, optname,
1491 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1492 #ifdef CONFIG_NETFILTER
1493 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1494 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1495 		int len;
1496 
1497 		if (get_user(len, optlen))
1498 			return -EFAULT;
1499 
1500 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1501 		if (err >= 0)
1502 			err = put_user(len, optlen);
1503 	}
1504 #endif
1505 	return err;
1506 }
1507 EXPORT_SYMBOL(ipv6_getsockopt);
1508