xref: /linux/net/ipv6/ipv6_sockglue.c (revision 5121516b0c4736b7977d977b239e36d23ec64401)
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 	}
482 	if (needs_rtnl)
483 		rtnl_lock();
484 	sockopt_lock_sock(sk);
485 
486 	/* Another thread has converted the socket into IPv4 with
487 	 * IPV6_ADDRFORM concurrently.
488 	 */
489 	if (unlikely(sk->sk_family != AF_INET6))
490 		goto unlock;
491 
492 	switch (optname) {
493 
494 	case IPV6_ADDRFORM:
495 		if (optlen < sizeof(int))
496 			goto e_inval;
497 		if (val == PF_INET) {
498 			if (sk->sk_type == SOCK_RAW)
499 				break;
500 
501 			if (sk->sk_protocol == IPPROTO_UDP ||
502 			    sk->sk_protocol == IPPROTO_UDPLITE) {
503 				struct udp_sock *up = udp_sk(sk);
504 				if (up->pending == AF_INET6) {
505 					retv = -EBUSY;
506 					break;
507 				}
508 			} else if (sk->sk_protocol == IPPROTO_TCP) {
509 				if (sk->sk_prot != &tcpv6_prot) {
510 					retv = -EBUSY;
511 					break;
512 				}
513 			} else {
514 				break;
515 			}
516 
517 			if (sk->sk_state != TCP_ESTABLISHED) {
518 				retv = -ENOTCONN;
519 				break;
520 			}
521 
522 			if (ipv6_only_sock(sk) ||
523 			    !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
524 				retv = -EADDRNOTAVAIL;
525 				break;
526 			}
527 
528 			__ipv6_sock_mc_close(sk);
529 			__ipv6_sock_ac_close(sk);
530 
531 			if (sk->sk_protocol == IPPROTO_TCP) {
532 				struct inet_connection_sock *icsk = inet_csk(sk);
533 
534 				sock_prot_inuse_add(net, sk->sk_prot, -1);
535 				sock_prot_inuse_add(net, &tcp_prot, 1);
536 
537 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
538 				WRITE_ONCE(sk->sk_prot, &tcp_prot);
539 				/* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
540 				WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
541 				WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
542 				WRITE_ONCE(sk->sk_family, PF_INET);
543 				tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
544 			} else {
545 				struct proto *prot = &udp_prot;
546 
547 				if (sk->sk_protocol == IPPROTO_UDPLITE)
548 					prot = &udplite_prot;
549 
550 				sock_prot_inuse_add(net, sk->sk_prot, -1);
551 				sock_prot_inuse_add(net, prot, 1);
552 
553 				/* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
554 				WRITE_ONCE(sk->sk_prot, prot);
555 				WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
556 				WRITE_ONCE(sk->sk_family, PF_INET);
557 			}
558 
559 			/* Disable all options not to allocate memory anymore,
560 			 * but there is still a race.  See the lockless path
561 			 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
562 			 */
563 			np->rxopt.all = 0;
564 
565 			inet6_cleanup_sock(sk);
566 
567 			module_put(THIS_MODULE);
568 			retv = 0;
569 			break;
570 		}
571 		goto e_inval;
572 
573 	case IPV6_V6ONLY:
574 		if (optlen < sizeof(int) ||
575 		    inet_sk(sk)->inet_num)
576 			goto e_inval;
577 		sk->sk_ipv6only = valbool;
578 		retv = 0;
579 		break;
580 
581 	case IPV6_RECVPKTINFO:
582 		if (optlen < sizeof(int))
583 			goto e_inval;
584 		np->rxopt.bits.rxinfo = valbool;
585 		retv = 0;
586 		break;
587 
588 	case IPV6_2292PKTINFO:
589 		if (optlen < sizeof(int))
590 			goto e_inval;
591 		np->rxopt.bits.rxoinfo = valbool;
592 		retv = 0;
593 		break;
594 
595 	case IPV6_RECVHOPLIMIT:
596 		if (optlen < sizeof(int))
597 			goto e_inval;
598 		np->rxopt.bits.rxhlim = valbool;
599 		retv = 0;
600 		break;
601 
602 	case IPV6_2292HOPLIMIT:
603 		if (optlen < sizeof(int))
604 			goto e_inval;
605 		np->rxopt.bits.rxohlim = valbool;
606 		retv = 0;
607 		break;
608 
609 	case IPV6_RECVRTHDR:
610 		if (optlen < sizeof(int))
611 			goto e_inval;
612 		np->rxopt.bits.srcrt = valbool;
613 		retv = 0;
614 		break;
615 
616 	case IPV6_2292RTHDR:
617 		if (optlen < sizeof(int))
618 			goto e_inval;
619 		np->rxopt.bits.osrcrt = valbool;
620 		retv = 0;
621 		break;
622 
623 	case IPV6_RECVHOPOPTS:
624 		if (optlen < sizeof(int))
625 			goto e_inval;
626 		np->rxopt.bits.hopopts = valbool;
627 		retv = 0;
628 		break;
629 
630 	case IPV6_2292HOPOPTS:
631 		if (optlen < sizeof(int))
632 			goto e_inval;
633 		np->rxopt.bits.ohopopts = valbool;
634 		retv = 0;
635 		break;
636 
637 	case IPV6_RECVDSTOPTS:
638 		if (optlen < sizeof(int))
639 			goto e_inval;
640 		np->rxopt.bits.dstopts = valbool;
641 		retv = 0;
642 		break;
643 
644 	case IPV6_2292DSTOPTS:
645 		if (optlen < sizeof(int))
646 			goto e_inval;
647 		np->rxopt.bits.odstopts = valbool;
648 		retv = 0;
649 		break;
650 
651 	case IPV6_TCLASS:
652 		if (optlen < sizeof(int))
653 			goto e_inval;
654 		if (val < -1 || val > 0xff)
655 			goto e_inval;
656 		/* RFC 3542, 6.5: default traffic class of 0x0 */
657 		if (val == -1)
658 			val = 0;
659 		if (sk->sk_type == SOCK_STREAM) {
660 			val &= ~INET_ECN_MASK;
661 			val |= np->tclass & INET_ECN_MASK;
662 		}
663 		if (np->tclass != val) {
664 			np->tclass = val;
665 			sk_dst_reset(sk);
666 		}
667 		retv = 0;
668 		break;
669 
670 	case IPV6_RECVTCLASS:
671 		if (optlen < sizeof(int))
672 			goto e_inval;
673 		np->rxopt.bits.rxtclass = valbool;
674 		retv = 0;
675 		break;
676 
677 	case IPV6_FLOWINFO:
678 		if (optlen < sizeof(int))
679 			goto e_inval;
680 		np->rxopt.bits.rxflow = valbool;
681 		retv = 0;
682 		break;
683 
684 	case IPV6_RECVPATHMTU:
685 		if (optlen < sizeof(int))
686 			goto e_inval;
687 		np->rxopt.bits.rxpmtu = valbool;
688 		retv = 0;
689 		break;
690 
691 	case IPV6_TRANSPARENT:
692 		if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
693 		    !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
694 			retv = -EPERM;
695 			break;
696 		}
697 		if (optlen < sizeof(int))
698 			goto e_inval;
699 		/* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
700 		inet_assign_bit(TRANSPARENT, sk, valbool);
701 		retv = 0;
702 		break;
703 
704 	case IPV6_FREEBIND:
705 		if (optlen < sizeof(int))
706 			goto e_inval;
707 		/* we also don't have a separate freebind bit for IPV6 */
708 		inet_assign_bit(FREEBIND, sk, valbool);
709 		retv = 0;
710 		break;
711 
712 	case IPV6_RECVORIGDSTADDR:
713 		if (optlen < sizeof(int))
714 			goto e_inval;
715 		np->rxopt.bits.rxorigdstaddr = valbool;
716 		retv = 0;
717 		break;
718 
719 	case IPV6_HOPOPTS:
720 	case IPV6_RTHDRDSTOPTS:
721 	case IPV6_RTHDR:
722 	case IPV6_DSTOPTS:
723 		retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
724 		break;
725 
726 	case IPV6_PKTINFO:
727 	{
728 		struct in6_pktinfo pkt;
729 
730 		if (optlen == 0)
731 			goto e_inval;
732 		else if (optlen < sizeof(struct in6_pktinfo) ||
733 			 sockptr_is_null(optval))
734 			goto e_inval;
735 
736 		if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
737 			retv = -EFAULT;
738 			break;
739 		}
740 		if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
741 			goto e_inval;
742 
743 		np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
744 		np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
745 		retv = 0;
746 		break;
747 	}
748 
749 	case IPV6_2292PKTOPTIONS:
750 	{
751 		struct ipv6_txoptions *opt = NULL;
752 		struct msghdr msg;
753 		struct flowi6 fl6;
754 		struct ipcm6_cookie ipc6;
755 
756 		memset(&fl6, 0, sizeof(fl6));
757 		fl6.flowi6_oif = sk->sk_bound_dev_if;
758 		fl6.flowi6_mark = sk->sk_mark;
759 
760 		if (optlen == 0)
761 			goto update;
762 
763 		/* 1K is probably excessive
764 		 * 1K is surely not enough, 2K per standard header is 16K.
765 		 */
766 		retv = -EINVAL;
767 		if (optlen > 64*1024)
768 			break;
769 
770 		opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
771 		retv = -ENOBUFS;
772 		if (!opt)
773 			break;
774 
775 		memset(opt, 0, sizeof(*opt));
776 		refcount_set(&opt->refcnt, 1);
777 		opt->tot_len = sizeof(*opt) + optlen;
778 		retv = -EFAULT;
779 		if (copy_from_sockptr(opt + 1, optval, optlen))
780 			goto done;
781 
782 		msg.msg_controllen = optlen;
783 		msg.msg_control_is_user = false;
784 		msg.msg_control = (void *)(opt+1);
785 		ipc6.opt = opt;
786 
787 		retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
788 		if (retv)
789 			goto done;
790 update:
791 		retv = 0;
792 		opt = ipv6_update_options(sk, opt);
793 done:
794 		if (opt) {
795 			atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
796 			txopt_put(opt);
797 		}
798 		break;
799 	}
800 
801 
802 	case IPV6_UNICAST_IF:
803 	{
804 		struct net_device *dev = NULL;
805 		int ifindex;
806 
807 		if (optlen != sizeof(int))
808 			goto e_inval;
809 
810 		ifindex = (__force int)ntohl((__force __be32)val);
811 		if (ifindex == 0) {
812 			np->ucast_oif = 0;
813 			retv = 0;
814 			break;
815 		}
816 
817 		dev = dev_get_by_index(net, ifindex);
818 		retv = -EADDRNOTAVAIL;
819 		if (!dev)
820 			break;
821 		dev_put(dev);
822 
823 		retv = -EINVAL;
824 		if (sk->sk_bound_dev_if)
825 			break;
826 
827 		np->ucast_oif = ifindex;
828 		retv = 0;
829 		break;
830 	}
831 
832 	case IPV6_MULTICAST_IF:
833 		if (sk->sk_type == SOCK_STREAM)
834 			break;
835 		if (optlen < sizeof(int))
836 			goto e_inval;
837 
838 		if (val) {
839 			struct net_device *dev;
840 			int midx;
841 
842 			rcu_read_lock();
843 
844 			dev = dev_get_by_index_rcu(net, val);
845 			if (!dev) {
846 				rcu_read_unlock();
847 				retv = -ENODEV;
848 				break;
849 			}
850 			midx = l3mdev_master_ifindex_rcu(dev);
851 
852 			rcu_read_unlock();
853 
854 			if (sk->sk_bound_dev_if &&
855 			    sk->sk_bound_dev_if != val &&
856 			    (!midx || midx != sk->sk_bound_dev_if))
857 				goto e_inval;
858 		}
859 		np->mcast_oif = val;
860 		retv = 0;
861 		break;
862 	case IPV6_ADD_MEMBERSHIP:
863 	case IPV6_DROP_MEMBERSHIP:
864 	{
865 		struct ipv6_mreq mreq;
866 
867 		if (optlen < sizeof(struct ipv6_mreq))
868 			goto e_inval;
869 
870 		retv = -EPROTO;
871 		if (inet_test_bit(IS_ICSK, sk))
872 			break;
873 
874 		retv = -EFAULT;
875 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
876 			break;
877 
878 		if (optname == IPV6_ADD_MEMBERSHIP)
879 			retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
880 		else
881 			retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
882 		break;
883 	}
884 	case IPV6_JOIN_ANYCAST:
885 	case IPV6_LEAVE_ANYCAST:
886 	{
887 		struct ipv6_mreq mreq;
888 
889 		if (optlen < sizeof(struct ipv6_mreq))
890 			goto e_inval;
891 
892 		retv = -EFAULT;
893 		if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
894 			break;
895 
896 		if (optname == IPV6_JOIN_ANYCAST)
897 			retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
898 		else
899 			retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
900 		break;
901 	}
902 	case MCAST_JOIN_GROUP:
903 	case MCAST_LEAVE_GROUP:
904 		if (in_compat_syscall())
905 			retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
906 							    optlen);
907 		else
908 			retv = ipv6_mcast_join_leave(sk, optname, optval,
909 						     optlen);
910 		break;
911 	case MCAST_JOIN_SOURCE_GROUP:
912 	case MCAST_LEAVE_SOURCE_GROUP:
913 	case MCAST_BLOCK_SOURCE:
914 	case MCAST_UNBLOCK_SOURCE:
915 		retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
916 		break;
917 	case MCAST_MSFILTER:
918 		if (in_compat_syscall())
919 			retv = compat_ipv6_set_mcast_msfilter(sk, optval,
920 							      optlen);
921 		else
922 			retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
923 		break;
924 	case IPV6_ROUTER_ALERT:
925 		if (optlen < sizeof(int))
926 			goto e_inval;
927 		retv = ip6_ra_control(sk, val);
928 		break;
929 	case IPV6_ROUTER_ALERT_ISOLATE:
930 		if (optlen < sizeof(int))
931 			goto e_inval;
932 		np->rtalert_isolate = valbool;
933 		retv = 0;
934 		break;
935 	case IPV6_MTU_DISCOVER:
936 		if (optlen < sizeof(int))
937 			goto e_inval;
938 		if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
939 			goto e_inval;
940 		np->pmtudisc = val;
941 		retv = 0;
942 		break;
943 	case IPV6_RECVERR:
944 		if (optlen < sizeof(int))
945 			goto e_inval;
946 		np->recverr = valbool;
947 		if (!val)
948 			skb_errqueue_purge(&sk->sk_error_queue);
949 		retv = 0;
950 		break;
951 	case IPV6_FLOWINFO_SEND:
952 		if (optlen < sizeof(int))
953 			goto e_inval;
954 		np->sndflow = valbool;
955 		retv = 0;
956 		break;
957 	case IPV6_FLOWLABEL_MGR:
958 		retv = ipv6_flowlabel_opt(sk, optval, optlen);
959 		break;
960 	case IPV6_IPSEC_POLICY:
961 	case IPV6_XFRM_POLICY:
962 		retv = -EPERM;
963 		if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
964 			break;
965 		retv = xfrm_user_policy(sk, optname, optval, optlen);
966 		break;
967 
968 	case IPV6_ADDR_PREFERENCES:
969 		if (optlen < sizeof(int))
970 			goto e_inval;
971 		retv = __ip6_sock_set_addr_preferences(sk, val);
972 		break;
973 	case IPV6_DONTFRAG:
974 		np->dontfrag = valbool;
975 		retv = 0;
976 		break;
977 	case IPV6_RECVFRAGSIZE:
978 		np->rxopt.bits.recvfragsize = valbool;
979 		retv = 0;
980 		break;
981 	}
982 
983 unlock:
984 	sockopt_release_sock(sk);
985 	if (needs_rtnl)
986 		rtnl_unlock();
987 
988 	return retv;
989 
990 e_inval:
991 	retv = -EINVAL;
992 	goto unlock;
993 }
994 
995 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
996 		    unsigned int optlen)
997 {
998 	int err;
999 
1000 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1001 		return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1002 
1003 	if (level != SOL_IPV6)
1004 		return -ENOPROTOOPT;
1005 
1006 	err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1007 #ifdef CONFIG_NETFILTER
1008 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1009 	if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1010 			optname != IPV6_XFRM_POLICY)
1011 		err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1012 #endif
1013 	return err;
1014 }
1015 EXPORT_SYMBOL(ipv6_setsockopt);
1016 
1017 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1018 				  int optname, sockptr_t optval, int len)
1019 {
1020 	struct ipv6_opt_hdr *hdr;
1021 
1022 	if (!opt)
1023 		return 0;
1024 
1025 	switch (optname) {
1026 	case IPV6_HOPOPTS:
1027 		hdr = opt->hopopt;
1028 		break;
1029 	case IPV6_RTHDRDSTOPTS:
1030 		hdr = opt->dst0opt;
1031 		break;
1032 	case IPV6_RTHDR:
1033 		hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1034 		break;
1035 	case IPV6_DSTOPTS:
1036 		hdr = opt->dst1opt;
1037 		break;
1038 	default:
1039 		return -EINVAL;	/* should not happen */
1040 	}
1041 
1042 	if (!hdr)
1043 		return 0;
1044 
1045 	len = min_t(unsigned int, len, ipv6_optlen(hdr));
1046 	if (copy_to_sockptr(optval, hdr, len))
1047 		return -EFAULT;
1048 	return len;
1049 }
1050 
1051 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1052 			     sockptr_t optlen, int len)
1053 {
1054 	const int size0 = offsetof(struct group_filter, gf_slist_flex);
1055 	struct group_filter gsf;
1056 	int num;
1057 	int err;
1058 
1059 	if (len < size0)
1060 		return -EINVAL;
1061 	if (copy_from_sockptr(&gsf, optval, size0))
1062 		return -EFAULT;
1063 	if (gsf.gf_group.ss_family != AF_INET6)
1064 		return -EADDRNOTAVAIL;
1065 	num = gsf.gf_numsrc;
1066 	sockopt_lock_sock(sk);
1067 	err = ip6_mc_msfget(sk, &gsf, optval, size0);
1068 	if (!err) {
1069 		if (num > gsf.gf_numsrc)
1070 			num = gsf.gf_numsrc;
1071 		len = GROUP_FILTER_SIZE(num);
1072 		if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1073 		    copy_to_sockptr(optval, &gsf, size0))
1074 			err = -EFAULT;
1075 	}
1076 	sockopt_release_sock(sk);
1077 	return err;
1078 }
1079 
1080 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1081 				    sockptr_t optlen, int len)
1082 {
1083 	const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1084 	struct compat_group_filter gf32;
1085 	struct group_filter gf;
1086 	int err;
1087 	int num;
1088 
1089 	if (len < size0)
1090 		return -EINVAL;
1091 
1092 	if (copy_from_sockptr(&gf32, optval, size0))
1093 		return -EFAULT;
1094 	gf.gf_interface = gf32.gf_interface;
1095 	gf.gf_fmode = gf32.gf_fmode;
1096 	num = gf.gf_numsrc = gf32.gf_numsrc;
1097 	gf.gf_group = gf32.gf_group;
1098 
1099 	if (gf.gf_group.ss_family != AF_INET6)
1100 		return -EADDRNOTAVAIL;
1101 
1102 	sockopt_lock_sock(sk);
1103 	err = ip6_mc_msfget(sk, &gf, optval, size0);
1104 	sockopt_release_sock(sk);
1105 	if (err)
1106 		return err;
1107 	if (num > gf.gf_numsrc)
1108 		num = gf.gf_numsrc;
1109 	len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1110 	if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1111 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1112 				   &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1113 	    copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1114 				   &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1115 		return -EFAULT;
1116 	return 0;
1117 }
1118 
1119 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1120 		       sockptr_t optval, sockptr_t optlen)
1121 {
1122 	struct ipv6_pinfo *np = inet6_sk(sk);
1123 	int len;
1124 	int val;
1125 
1126 	if (ip6_mroute_opt(optname))
1127 		return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1128 
1129 	if (copy_from_sockptr(&len, optlen, sizeof(int)))
1130 		return -EFAULT;
1131 	switch (optname) {
1132 	case IPV6_ADDRFORM:
1133 		if (sk->sk_protocol != IPPROTO_UDP &&
1134 		    sk->sk_protocol != IPPROTO_UDPLITE &&
1135 		    sk->sk_protocol != IPPROTO_TCP)
1136 			return -ENOPROTOOPT;
1137 		if (sk->sk_state != TCP_ESTABLISHED)
1138 			return -ENOTCONN;
1139 		val = sk->sk_family;
1140 		break;
1141 	case MCAST_MSFILTER:
1142 		if (in_compat_syscall())
1143 			return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1144 		return ipv6_get_msfilter(sk, optval, optlen, len);
1145 	case IPV6_2292PKTOPTIONS:
1146 	{
1147 		struct msghdr msg;
1148 		struct sk_buff *skb;
1149 
1150 		if (sk->sk_type != SOCK_STREAM)
1151 			return -ENOPROTOOPT;
1152 
1153 		if (optval.is_kernel) {
1154 			msg.msg_control_is_user = false;
1155 			msg.msg_control = optval.kernel;
1156 		} else {
1157 			msg.msg_control_is_user = true;
1158 			msg.msg_control_user = optval.user;
1159 		}
1160 		msg.msg_controllen = len;
1161 		msg.msg_flags = 0;
1162 
1163 		sockopt_lock_sock(sk);
1164 		skb = np->pktoptions;
1165 		if (skb)
1166 			ip6_datagram_recv_ctl(sk, &msg, skb);
1167 		sockopt_release_sock(sk);
1168 		if (!skb) {
1169 			if (np->rxopt.bits.rxinfo) {
1170 				struct in6_pktinfo src_info;
1171 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1172 					np->sticky_pktinfo.ipi6_ifindex;
1173 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1174 				put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1175 			}
1176 			if (np->rxopt.bits.rxhlim) {
1177 				int hlim = READ_ONCE(np->mcast_hops);
1178 
1179 				put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1180 			}
1181 			if (np->rxopt.bits.rxtclass) {
1182 				int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1183 
1184 				put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1185 			}
1186 			if (np->rxopt.bits.rxoinfo) {
1187 				struct in6_pktinfo src_info;
1188 				src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1189 					np->sticky_pktinfo.ipi6_ifindex;
1190 				src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1191 								     np->sticky_pktinfo.ipi6_addr;
1192 				put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1193 			}
1194 			if (np->rxopt.bits.rxohlim) {
1195 				int hlim = READ_ONCE(np->mcast_hops);
1196 
1197 				put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1198 			}
1199 			if (np->rxopt.bits.rxflow) {
1200 				__be32 flowinfo = np->rcv_flowinfo;
1201 
1202 				put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1203 			}
1204 		}
1205 		len -= msg.msg_controllen;
1206 		return copy_to_sockptr(optlen, &len, sizeof(int));
1207 	}
1208 	case IPV6_MTU:
1209 	{
1210 		struct dst_entry *dst;
1211 
1212 		val = 0;
1213 		rcu_read_lock();
1214 		dst = __sk_dst_get(sk);
1215 		if (dst)
1216 			val = dst_mtu(dst);
1217 		rcu_read_unlock();
1218 		if (!val)
1219 			return -ENOTCONN;
1220 		break;
1221 	}
1222 
1223 	case IPV6_V6ONLY:
1224 		val = sk->sk_ipv6only;
1225 		break;
1226 
1227 	case IPV6_RECVPKTINFO:
1228 		val = np->rxopt.bits.rxinfo;
1229 		break;
1230 
1231 	case IPV6_2292PKTINFO:
1232 		val = np->rxopt.bits.rxoinfo;
1233 		break;
1234 
1235 	case IPV6_RECVHOPLIMIT:
1236 		val = np->rxopt.bits.rxhlim;
1237 		break;
1238 
1239 	case IPV6_2292HOPLIMIT:
1240 		val = np->rxopt.bits.rxohlim;
1241 		break;
1242 
1243 	case IPV6_RECVRTHDR:
1244 		val = np->rxopt.bits.srcrt;
1245 		break;
1246 
1247 	case IPV6_2292RTHDR:
1248 		val = np->rxopt.bits.osrcrt;
1249 		break;
1250 
1251 	case IPV6_HOPOPTS:
1252 	case IPV6_RTHDRDSTOPTS:
1253 	case IPV6_RTHDR:
1254 	case IPV6_DSTOPTS:
1255 	{
1256 		struct ipv6_txoptions *opt;
1257 
1258 		sockopt_lock_sock(sk);
1259 		opt = rcu_dereference_protected(np->opt,
1260 						lockdep_sock_is_held(sk));
1261 		len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1262 		sockopt_release_sock(sk);
1263 		/* check if ipv6_getsockopt_sticky() returns err code */
1264 		if (len < 0)
1265 			return len;
1266 		return copy_to_sockptr(optlen, &len, sizeof(int));
1267 	}
1268 
1269 	case IPV6_RECVHOPOPTS:
1270 		val = np->rxopt.bits.hopopts;
1271 		break;
1272 
1273 	case IPV6_2292HOPOPTS:
1274 		val = np->rxopt.bits.ohopopts;
1275 		break;
1276 
1277 	case IPV6_RECVDSTOPTS:
1278 		val = np->rxopt.bits.dstopts;
1279 		break;
1280 
1281 	case IPV6_2292DSTOPTS:
1282 		val = np->rxopt.bits.odstopts;
1283 		break;
1284 
1285 	case IPV6_TCLASS:
1286 		val = np->tclass;
1287 		break;
1288 
1289 	case IPV6_RECVTCLASS:
1290 		val = np->rxopt.bits.rxtclass;
1291 		break;
1292 
1293 	case IPV6_FLOWINFO:
1294 		val = np->rxopt.bits.rxflow;
1295 		break;
1296 
1297 	case IPV6_RECVPATHMTU:
1298 		val = np->rxopt.bits.rxpmtu;
1299 		break;
1300 
1301 	case IPV6_PATHMTU:
1302 	{
1303 		struct dst_entry *dst;
1304 		struct ip6_mtuinfo mtuinfo;
1305 
1306 		if (len < sizeof(mtuinfo))
1307 			return -EINVAL;
1308 
1309 		len = sizeof(mtuinfo);
1310 		memset(&mtuinfo, 0, sizeof(mtuinfo));
1311 
1312 		rcu_read_lock();
1313 		dst = __sk_dst_get(sk);
1314 		if (dst)
1315 			mtuinfo.ip6m_mtu = dst_mtu(dst);
1316 		rcu_read_unlock();
1317 		if (!mtuinfo.ip6m_mtu)
1318 			return -ENOTCONN;
1319 
1320 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1321 			return -EFAULT;
1322 		if (copy_to_sockptr(optval, &mtuinfo, len))
1323 			return -EFAULT;
1324 
1325 		return 0;
1326 	}
1327 
1328 	case IPV6_TRANSPARENT:
1329 		val = inet_test_bit(TRANSPARENT, sk);
1330 		break;
1331 
1332 	case IPV6_FREEBIND:
1333 		val = inet_test_bit(FREEBIND, sk);
1334 		break;
1335 
1336 	case IPV6_RECVORIGDSTADDR:
1337 		val = np->rxopt.bits.rxorigdstaddr;
1338 		break;
1339 
1340 	case IPV6_UNICAST_HOPS:
1341 	case IPV6_MULTICAST_HOPS:
1342 	{
1343 		struct dst_entry *dst;
1344 
1345 		if (optname == IPV6_UNICAST_HOPS)
1346 			val = READ_ONCE(np->hop_limit);
1347 		else
1348 			val = READ_ONCE(np->mcast_hops);
1349 
1350 		if (val < 0) {
1351 			rcu_read_lock();
1352 			dst = __sk_dst_get(sk);
1353 			if (dst)
1354 				val = ip6_dst_hoplimit(dst);
1355 			rcu_read_unlock();
1356 		}
1357 
1358 		if (val < 0)
1359 			val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1360 		break;
1361 	}
1362 
1363 	case IPV6_MULTICAST_LOOP:
1364 		val = inet6_test_bit(MC6_LOOP, sk);
1365 		break;
1366 
1367 	case IPV6_MULTICAST_IF:
1368 		val = np->mcast_oif;
1369 		break;
1370 
1371 	case IPV6_MULTICAST_ALL:
1372 		val = inet6_test_bit(MC6_ALL, sk);
1373 		break;
1374 
1375 	case IPV6_UNICAST_IF:
1376 		val = (__force int)htonl((__u32) np->ucast_oif);
1377 		break;
1378 
1379 	case IPV6_MTU_DISCOVER:
1380 		val = np->pmtudisc;
1381 		break;
1382 
1383 	case IPV6_RECVERR:
1384 		val = np->recverr;
1385 		break;
1386 
1387 	case IPV6_FLOWINFO_SEND:
1388 		val = np->sndflow;
1389 		break;
1390 
1391 	case IPV6_FLOWLABEL_MGR:
1392 	{
1393 		struct in6_flowlabel_req freq;
1394 		int flags;
1395 
1396 		if (len < sizeof(freq))
1397 			return -EINVAL;
1398 
1399 		if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1400 			return -EFAULT;
1401 
1402 		if (freq.flr_action != IPV6_FL_A_GET)
1403 			return -EINVAL;
1404 
1405 		len = sizeof(freq);
1406 		flags = freq.flr_flags;
1407 
1408 		memset(&freq, 0, sizeof(freq));
1409 
1410 		val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1411 		if (val < 0)
1412 			return val;
1413 
1414 		if (copy_to_sockptr(optlen, &len, sizeof(int)))
1415 			return -EFAULT;
1416 		if (copy_to_sockptr(optval, &freq, len))
1417 			return -EFAULT;
1418 
1419 		return 0;
1420 	}
1421 
1422 	case IPV6_ADDR_PREFERENCES:
1423 		val = 0;
1424 
1425 		if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1426 			val |= IPV6_PREFER_SRC_TMP;
1427 		else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1428 			val |= IPV6_PREFER_SRC_PUBLIC;
1429 		else {
1430 			/* XXX: should we return system default? */
1431 			val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1432 		}
1433 
1434 		if (np->srcprefs & IPV6_PREFER_SRC_COA)
1435 			val |= IPV6_PREFER_SRC_COA;
1436 		else
1437 			val |= IPV6_PREFER_SRC_HOME;
1438 		break;
1439 
1440 	case IPV6_MINHOPCOUNT:
1441 		val = READ_ONCE(np->min_hopcount);
1442 		break;
1443 
1444 	case IPV6_DONTFRAG:
1445 		val = np->dontfrag;
1446 		break;
1447 
1448 	case IPV6_AUTOFLOWLABEL:
1449 		val = ip6_autoflowlabel(sock_net(sk), sk);
1450 		break;
1451 
1452 	case IPV6_RECVFRAGSIZE:
1453 		val = np->rxopt.bits.recvfragsize;
1454 		break;
1455 
1456 	case IPV6_ROUTER_ALERT_ISOLATE:
1457 		val = np->rtalert_isolate;
1458 		break;
1459 
1460 	case IPV6_RECVERR_RFC4884:
1461 		val = inet6_test_bit(RECVERR6_RFC4884, sk);
1462 		break;
1463 
1464 	default:
1465 		return -ENOPROTOOPT;
1466 	}
1467 	len = min_t(unsigned int, sizeof(int), len);
1468 	if (copy_to_sockptr(optlen, &len, sizeof(int)))
1469 		return -EFAULT;
1470 	if (copy_to_sockptr(optval, &val, len))
1471 		return -EFAULT;
1472 	return 0;
1473 }
1474 
1475 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1476 		    char __user *optval, int __user *optlen)
1477 {
1478 	int err;
1479 
1480 	if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1481 		return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1482 
1483 	if (level != SOL_IPV6)
1484 		return -ENOPROTOOPT;
1485 
1486 	err = do_ipv6_getsockopt(sk, level, optname,
1487 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1488 #ifdef CONFIG_NETFILTER
1489 	/* we need to exclude all possible ENOPROTOOPTs except default case */
1490 	if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1491 		int len;
1492 
1493 		if (get_user(len, optlen))
1494 			return -EFAULT;
1495 
1496 		err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1497 		if (err >= 0)
1498 			err = put_user(len, optlen);
1499 	}
1500 #endif
1501 	return err;
1502 }
1503 EXPORT_SYMBOL(ipv6_getsockopt);
1504