Linux 4.8-rc8
[linux/fpc-iii.git] / net / ipv6 / ipv6_sockglue.c
blob5330262ab673c022fbf700d22782a74ccd1494fe
1 /*
2 * IPv6 BSD socket options interface
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on linux/net/ipv4/ip_sockglue.c
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
15 * FIXME: Make the setsockopt code POSIX compliant: That is
17 * o Truncate getsockopt returns
18 * o Return an optlen of the truncated length if need be
20 * Changes:
21 * David L Stevens <dlstevens@us.ibm.com>:
22 * - added multicast source filtering API for MLDv2
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39 #include <linux/slab.h>
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/transp_v6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/inet_common.h>
50 #include <net/tcp.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/xfrm.h>
54 #include <net/compat.h>
56 #include <asm/uaccess.h>
58 struct ip6_ra_chain *ip6_ra_chain;
59 DEFINE_RWLOCK(ip6_ra_lock);
61 int ip6_ra_control(struct sock *sk, int sel)
63 struct ip6_ra_chain *ra, *new_ra, **rap;
65 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
66 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
67 return -ENOPROTOOPT;
69 new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
71 write_lock_bh(&ip6_ra_lock);
72 for (rap = &ip6_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
73 if (ra->sk == sk) {
74 if (sel >= 0) {
75 write_unlock_bh(&ip6_ra_lock);
76 kfree(new_ra);
77 return -EADDRINUSE;
80 *rap = ra->next;
81 write_unlock_bh(&ip6_ra_lock);
83 sock_put(sk);
84 kfree(ra);
85 return 0;
88 if (!new_ra) {
89 write_unlock_bh(&ip6_ra_lock);
90 return -ENOBUFS;
92 new_ra->sk = sk;
93 new_ra->sel = sel;
94 new_ra->next = ra;
95 *rap = new_ra;
96 sock_hold(sk);
97 write_unlock_bh(&ip6_ra_lock);
98 return 0;
101 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
102 struct ipv6_txoptions *opt)
104 if (inet_sk(sk)->is_icsk) {
105 if (opt &&
106 !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
107 inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
108 struct inet_connection_sock *icsk = inet_csk(sk);
109 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
110 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
113 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
114 opt);
115 sk_dst_reset(sk);
117 return opt;
120 static bool setsockopt_needs_rtnl(int optname)
122 switch (optname) {
123 case IPV6_ADD_MEMBERSHIP:
124 case IPV6_DROP_MEMBERSHIP:
125 case IPV6_JOIN_ANYCAST:
126 case IPV6_LEAVE_ANYCAST:
127 case MCAST_JOIN_GROUP:
128 case MCAST_LEAVE_GROUP:
129 case MCAST_JOIN_SOURCE_GROUP:
130 case MCAST_LEAVE_SOURCE_GROUP:
131 case MCAST_BLOCK_SOURCE:
132 case MCAST_UNBLOCK_SOURCE:
133 case MCAST_MSFILTER:
134 return true;
136 return false;
139 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
140 char __user *optval, unsigned int optlen)
142 struct ipv6_pinfo *np = inet6_sk(sk);
143 struct net *net = sock_net(sk);
144 int val, valbool;
145 int retv = -ENOPROTOOPT;
146 bool needs_rtnl = setsockopt_needs_rtnl(optname);
148 if (!optval)
149 val = 0;
150 else {
151 if (optlen >= sizeof(int)) {
152 if (get_user(val, (int __user *) optval))
153 return -EFAULT;
154 } else
155 val = 0;
158 valbool = (val != 0);
160 if (ip6_mroute_opt(optname))
161 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
163 if (needs_rtnl)
164 rtnl_lock();
165 lock_sock(sk);
167 switch (optname) {
169 case IPV6_ADDRFORM:
170 if (optlen < sizeof(int))
171 goto e_inval;
172 if (val == PF_INET) {
173 struct ipv6_txoptions *opt;
174 struct sk_buff *pktopt;
176 if (sk->sk_type == SOCK_RAW)
177 break;
179 if (sk->sk_protocol == IPPROTO_UDP ||
180 sk->sk_protocol == IPPROTO_UDPLITE) {
181 struct udp_sock *up = udp_sk(sk);
182 if (up->pending == AF_INET6) {
183 retv = -EBUSY;
184 break;
186 } else if (sk->sk_protocol != IPPROTO_TCP)
187 break;
189 if (sk->sk_state != TCP_ESTABLISHED) {
190 retv = -ENOTCONN;
191 break;
194 if (ipv6_only_sock(sk) ||
195 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
196 retv = -EADDRNOTAVAIL;
197 break;
200 fl6_free_socklist(sk);
201 ipv6_sock_mc_close(sk);
204 * Sock is moving from IPv6 to IPv4 (sk_prot), so
205 * remove it from the refcnt debug socks count in the
206 * original family...
208 sk_refcnt_debug_dec(sk);
210 if (sk->sk_protocol == IPPROTO_TCP) {
211 struct inet_connection_sock *icsk = inet_csk(sk);
212 local_bh_disable();
213 sock_prot_inuse_add(net, sk->sk_prot, -1);
214 sock_prot_inuse_add(net, &tcp_prot, 1);
215 local_bh_enable();
216 sk->sk_prot = &tcp_prot;
217 icsk->icsk_af_ops = &ipv4_specific;
218 sk->sk_socket->ops = &inet_stream_ops;
219 sk->sk_family = PF_INET;
220 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
221 } else {
222 struct proto *prot = &udp_prot;
224 if (sk->sk_protocol == IPPROTO_UDPLITE)
225 prot = &udplite_prot;
226 local_bh_disable();
227 sock_prot_inuse_add(net, sk->sk_prot, -1);
228 sock_prot_inuse_add(net, prot, 1);
229 local_bh_enable();
230 sk->sk_prot = prot;
231 sk->sk_socket->ops = &inet_dgram_ops;
232 sk->sk_family = PF_INET;
234 opt = xchg((__force struct ipv6_txoptions **)&np->opt,
235 NULL);
236 if (opt) {
237 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
238 txopt_put(opt);
240 pktopt = xchg(&np->pktoptions, NULL);
241 kfree_skb(pktopt);
243 sk->sk_destruct = inet_sock_destruct;
245 * ... and add it to the refcnt debug socks count
246 * in the new family. -acme
248 sk_refcnt_debug_inc(sk);
249 module_put(THIS_MODULE);
250 retv = 0;
251 break;
253 goto e_inval;
255 case IPV6_V6ONLY:
256 if (optlen < sizeof(int) ||
257 inet_sk(sk)->inet_num)
258 goto e_inval;
259 sk->sk_ipv6only = valbool;
260 retv = 0;
261 break;
263 case IPV6_RECVPKTINFO:
264 if (optlen < sizeof(int))
265 goto e_inval;
266 np->rxopt.bits.rxinfo = valbool;
267 retv = 0;
268 break;
270 case IPV6_2292PKTINFO:
271 if (optlen < sizeof(int))
272 goto e_inval;
273 np->rxopt.bits.rxoinfo = valbool;
274 retv = 0;
275 break;
277 case IPV6_RECVHOPLIMIT:
278 if (optlen < sizeof(int))
279 goto e_inval;
280 np->rxopt.bits.rxhlim = valbool;
281 retv = 0;
282 break;
284 case IPV6_2292HOPLIMIT:
285 if (optlen < sizeof(int))
286 goto e_inval;
287 np->rxopt.bits.rxohlim = valbool;
288 retv = 0;
289 break;
291 case IPV6_RECVRTHDR:
292 if (optlen < sizeof(int))
293 goto e_inval;
294 np->rxopt.bits.srcrt = valbool;
295 retv = 0;
296 break;
298 case IPV6_2292RTHDR:
299 if (optlen < sizeof(int))
300 goto e_inval;
301 np->rxopt.bits.osrcrt = valbool;
302 retv = 0;
303 break;
305 case IPV6_RECVHOPOPTS:
306 if (optlen < sizeof(int))
307 goto e_inval;
308 np->rxopt.bits.hopopts = valbool;
309 retv = 0;
310 break;
312 case IPV6_2292HOPOPTS:
313 if (optlen < sizeof(int))
314 goto e_inval;
315 np->rxopt.bits.ohopopts = valbool;
316 retv = 0;
317 break;
319 case IPV6_RECVDSTOPTS:
320 if (optlen < sizeof(int))
321 goto e_inval;
322 np->rxopt.bits.dstopts = valbool;
323 retv = 0;
324 break;
326 case IPV6_2292DSTOPTS:
327 if (optlen < sizeof(int))
328 goto e_inval;
329 np->rxopt.bits.odstopts = valbool;
330 retv = 0;
331 break;
333 case IPV6_TCLASS:
334 if (optlen < sizeof(int))
335 goto e_inval;
336 if (val < -1 || val > 0xff)
337 goto e_inval;
338 /* RFC 3542, 6.5: default traffic class of 0x0 */
339 if (val == -1)
340 val = 0;
341 np->tclass = val;
342 retv = 0;
343 break;
345 case IPV6_RECVTCLASS:
346 if (optlen < sizeof(int))
347 goto e_inval;
348 np->rxopt.bits.rxtclass = valbool;
349 retv = 0;
350 break;
352 case IPV6_FLOWINFO:
353 if (optlen < sizeof(int))
354 goto e_inval;
355 np->rxopt.bits.rxflow = valbool;
356 retv = 0;
357 break;
359 case IPV6_RECVPATHMTU:
360 if (optlen < sizeof(int))
361 goto e_inval;
362 np->rxopt.bits.rxpmtu = valbool;
363 retv = 0;
364 break;
366 case IPV6_TRANSPARENT:
367 if (valbool && !ns_capable(net->user_ns, CAP_NET_ADMIN) &&
368 !ns_capable(net->user_ns, CAP_NET_RAW)) {
369 retv = -EPERM;
370 break;
372 if (optlen < sizeof(int))
373 goto e_inval;
374 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
375 inet_sk(sk)->transparent = valbool;
376 retv = 0;
377 break;
379 case IPV6_RECVORIGDSTADDR:
380 if (optlen < sizeof(int))
381 goto e_inval;
382 np->rxopt.bits.rxorigdstaddr = valbool;
383 retv = 0;
384 break;
386 case IPV6_HOPOPTS:
387 case IPV6_RTHDRDSTOPTS:
388 case IPV6_RTHDR:
389 case IPV6_DSTOPTS:
391 struct ipv6_txoptions *opt;
393 /* remove any sticky options header with a zero option
394 * length, per RFC3542.
396 if (optlen == 0)
397 optval = NULL;
398 else if (!optval)
399 goto e_inval;
400 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
401 optlen & 0x7 || optlen > 8 * 255)
402 goto e_inval;
404 /* hop-by-hop / destination options are privileged option */
405 retv = -EPERM;
406 if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
407 break;
409 opt = rcu_dereference_protected(np->opt,
410 lockdep_sock_is_held(sk));
411 opt = ipv6_renew_options(sk, opt, optname,
412 (struct ipv6_opt_hdr __user *)optval,
413 optlen);
414 if (IS_ERR(opt)) {
415 retv = PTR_ERR(opt);
416 break;
419 /* routing header option needs extra check */
420 retv = -EINVAL;
421 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
422 struct ipv6_rt_hdr *rthdr = opt->srcrt;
423 switch (rthdr->type) {
424 #if IS_ENABLED(CONFIG_IPV6_MIP6)
425 case IPV6_SRCRT_TYPE_2:
426 if (rthdr->hdrlen != 2 ||
427 rthdr->segments_left != 1)
428 goto sticky_done;
430 break;
431 #endif
432 default:
433 goto sticky_done;
437 retv = 0;
438 opt = ipv6_update_options(sk, opt);
439 sticky_done:
440 if (opt) {
441 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
442 txopt_put(opt);
444 break;
447 case IPV6_PKTINFO:
449 struct in6_pktinfo pkt;
451 if (optlen == 0)
452 goto e_inval;
453 else if (optlen < sizeof(struct in6_pktinfo) || !optval)
454 goto e_inval;
456 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
457 retv = -EFAULT;
458 break;
460 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
461 goto e_inval;
463 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
464 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
465 retv = 0;
466 break;
469 case IPV6_2292PKTOPTIONS:
471 struct ipv6_txoptions *opt = NULL;
472 struct msghdr msg;
473 struct flowi6 fl6;
474 struct sockcm_cookie sockc_junk;
475 struct ipcm6_cookie ipc6;
477 memset(&fl6, 0, sizeof(fl6));
478 fl6.flowi6_oif = sk->sk_bound_dev_if;
479 fl6.flowi6_mark = sk->sk_mark;
481 if (optlen == 0)
482 goto update;
484 /* 1K is probably excessive
485 * 1K is surely not enough, 2K per standard header is 16K.
487 retv = -EINVAL;
488 if (optlen > 64*1024)
489 break;
491 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
492 retv = -ENOBUFS;
493 if (!opt)
494 break;
496 memset(opt, 0, sizeof(*opt));
497 atomic_set(&opt->refcnt, 1);
498 opt->tot_len = sizeof(*opt) + optlen;
499 retv = -EFAULT;
500 if (copy_from_user(opt+1, optval, optlen))
501 goto done;
503 msg.msg_controllen = optlen;
504 msg.msg_control = (void *)(opt+1);
505 ipc6.opt = opt;
507 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6, &sockc_junk);
508 if (retv)
509 goto done;
510 update:
511 retv = 0;
512 opt = ipv6_update_options(sk, opt);
513 done:
514 if (opt) {
515 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
516 txopt_put(opt);
518 break;
520 case IPV6_UNICAST_HOPS:
521 if (optlen < sizeof(int))
522 goto e_inval;
523 if (val > 255 || val < -1)
524 goto e_inval;
525 np->hop_limit = val;
526 retv = 0;
527 break;
529 case IPV6_MULTICAST_HOPS:
530 if (sk->sk_type == SOCK_STREAM)
531 break;
532 if (optlen < sizeof(int))
533 goto e_inval;
534 if (val > 255 || val < -1)
535 goto e_inval;
536 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
537 retv = 0;
538 break;
540 case IPV6_MULTICAST_LOOP:
541 if (optlen < sizeof(int))
542 goto e_inval;
543 if (val != valbool)
544 goto e_inval;
545 np->mc_loop = valbool;
546 retv = 0;
547 break;
549 case IPV6_UNICAST_IF:
551 struct net_device *dev = NULL;
552 int ifindex;
554 if (optlen != sizeof(int))
555 goto e_inval;
557 ifindex = (__force int)ntohl((__force __be32)val);
558 if (ifindex == 0) {
559 np->ucast_oif = 0;
560 retv = 0;
561 break;
564 dev = dev_get_by_index(net, ifindex);
565 retv = -EADDRNOTAVAIL;
566 if (!dev)
567 break;
568 dev_put(dev);
570 retv = -EINVAL;
571 if (sk->sk_bound_dev_if)
572 break;
574 np->ucast_oif = ifindex;
575 retv = 0;
576 break;
579 case IPV6_MULTICAST_IF:
580 if (sk->sk_type == SOCK_STREAM)
581 break;
582 if (optlen < sizeof(int))
583 goto e_inval;
585 if (val) {
586 struct net_device *dev;
588 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
589 goto e_inval;
591 dev = dev_get_by_index(net, val);
592 if (!dev) {
593 retv = -ENODEV;
594 break;
596 dev_put(dev);
598 np->mcast_oif = val;
599 retv = 0;
600 break;
601 case IPV6_ADD_MEMBERSHIP:
602 case IPV6_DROP_MEMBERSHIP:
604 struct ipv6_mreq mreq;
606 if (optlen < sizeof(struct ipv6_mreq))
607 goto e_inval;
609 retv = -EPROTO;
610 if (inet_sk(sk)->is_icsk)
611 break;
613 retv = -EFAULT;
614 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
615 break;
617 if (optname == IPV6_ADD_MEMBERSHIP)
618 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
619 else
620 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
621 break;
623 case IPV6_JOIN_ANYCAST:
624 case IPV6_LEAVE_ANYCAST:
626 struct ipv6_mreq mreq;
628 if (optlen < sizeof(struct ipv6_mreq))
629 goto e_inval;
631 retv = -EFAULT;
632 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
633 break;
635 if (optname == IPV6_JOIN_ANYCAST)
636 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
637 else
638 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
639 break;
641 case MCAST_JOIN_GROUP:
642 case MCAST_LEAVE_GROUP:
644 struct group_req greq;
645 struct sockaddr_in6 *psin6;
647 if (optlen < sizeof(struct group_req))
648 goto e_inval;
650 retv = -EFAULT;
651 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
652 break;
653 if (greq.gr_group.ss_family != AF_INET6) {
654 retv = -EADDRNOTAVAIL;
655 break;
657 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
658 if (optname == MCAST_JOIN_GROUP)
659 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
660 &psin6->sin6_addr);
661 else
662 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
663 &psin6->sin6_addr);
664 break;
666 case MCAST_JOIN_SOURCE_GROUP:
667 case MCAST_LEAVE_SOURCE_GROUP:
668 case MCAST_BLOCK_SOURCE:
669 case MCAST_UNBLOCK_SOURCE:
671 struct group_source_req greqs;
672 int omode, add;
674 if (optlen < sizeof(struct group_source_req))
675 goto e_inval;
676 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
677 retv = -EFAULT;
678 break;
680 if (greqs.gsr_group.ss_family != AF_INET6 ||
681 greqs.gsr_source.ss_family != AF_INET6) {
682 retv = -EADDRNOTAVAIL;
683 break;
685 if (optname == MCAST_BLOCK_SOURCE) {
686 omode = MCAST_EXCLUDE;
687 add = 1;
688 } else if (optname == MCAST_UNBLOCK_SOURCE) {
689 omode = MCAST_EXCLUDE;
690 add = 0;
691 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
692 struct sockaddr_in6 *psin6;
694 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
695 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
696 &psin6->sin6_addr);
697 /* prior join w/ different source is ok */
698 if (retv && retv != -EADDRINUSE)
699 break;
700 omode = MCAST_INCLUDE;
701 add = 1;
702 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
703 omode = MCAST_INCLUDE;
704 add = 0;
706 retv = ip6_mc_source(add, omode, sk, &greqs);
707 break;
709 case MCAST_MSFILTER:
711 struct group_filter *gsf;
713 if (optlen < GROUP_FILTER_SIZE(0))
714 goto e_inval;
715 if (optlen > sysctl_optmem_max) {
716 retv = -ENOBUFS;
717 break;
719 gsf = kmalloc(optlen, GFP_KERNEL);
720 if (!gsf) {
721 retv = -ENOBUFS;
722 break;
724 retv = -EFAULT;
725 if (copy_from_user(gsf, optval, optlen)) {
726 kfree(gsf);
727 break;
729 /* numsrc >= (4G-140)/128 overflow in 32 bits */
730 if (gsf->gf_numsrc >= 0x1ffffffU ||
731 gsf->gf_numsrc > sysctl_mld_max_msf) {
732 kfree(gsf);
733 retv = -ENOBUFS;
734 break;
736 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
737 kfree(gsf);
738 retv = -EINVAL;
739 break;
741 retv = ip6_mc_msfilter(sk, gsf);
742 kfree(gsf);
744 break;
746 case IPV6_ROUTER_ALERT:
747 if (optlen < sizeof(int))
748 goto e_inval;
749 retv = ip6_ra_control(sk, val);
750 break;
751 case IPV6_MTU_DISCOVER:
752 if (optlen < sizeof(int))
753 goto e_inval;
754 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
755 goto e_inval;
756 np->pmtudisc = val;
757 retv = 0;
758 break;
759 case IPV6_MTU:
760 if (optlen < sizeof(int))
761 goto e_inval;
762 if (val && val < IPV6_MIN_MTU)
763 goto e_inval;
764 np->frag_size = val;
765 retv = 0;
766 break;
767 case IPV6_RECVERR:
768 if (optlen < sizeof(int))
769 goto e_inval;
770 np->recverr = valbool;
771 if (!val)
772 skb_queue_purge(&sk->sk_error_queue);
773 retv = 0;
774 break;
775 case IPV6_FLOWINFO_SEND:
776 if (optlen < sizeof(int))
777 goto e_inval;
778 np->sndflow = valbool;
779 retv = 0;
780 break;
781 case IPV6_FLOWLABEL_MGR:
782 retv = ipv6_flowlabel_opt(sk, optval, optlen);
783 break;
784 case IPV6_IPSEC_POLICY:
785 case IPV6_XFRM_POLICY:
786 retv = -EPERM;
787 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
788 break;
789 retv = xfrm_user_policy(sk, optname, optval, optlen);
790 break;
792 case IPV6_ADDR_PREFERENCES:
794 unsigned int pref = 0;
795 unsigned int prefmask = ~0;
797 if (optlen < sizeof(int))
798 goto e_inval;
800 retv = -EINVAL;
802 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
803 switch (val & (IPV6_PREFER_SRC_PUBLIC|
804 IPV6_PREFER_SRC_TMP|
805 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
806 case IPV6_PREFER_SRC_PUBLIC:
807 pref |= IPV6_PREFER_SRC_PUBLIC;
808 break;
809 case IPV6_PREFER_SRC_TMP:
810 pref |= IPV6_PREFER_SRC_TMP;
811 break;
812 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
813 break;
814 case 0:
815 goto pref_skip_pubtmp;
816 default:
817 goto e_inval;
820 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
821 IPV6_PREFER_SRC_TMP);
822 pref_skip_pubtmp:
824 /* check HOME/COA conflicts */
825 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
826 case IPV6_PREFER_SRC_HOME:
827 break;
828 case IPV6_PREFER_SRC_COA:
829 pref |= IPV6_PREFER_SRC_COA;
830 case 0:
831 goto pref_skip_coa;
832 default:
833 goto e_inval;
836 prefmask &= ~IPV6_PREFER_SRC_COA;
837 pref_skip_coa:
839 /* check CGA/NONCGA conflicts */
840 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
841 case IPV6_PREFER_SRC_CGA:
842 case IPV6_PREFER_SRC_NONCGA:
843 case 0:
844 break;
845 default:
846 goto e_inval;
849 np->srcprefs = (np->srcprefs & prefmask) | pref;
850 retv = 0;
852 break;
854 case IPV6_MINHOPCOUNT:
855 if (optlen < sizeof(int))
856 goto e_inval;
857 if (val < 0 || val > 255)
858 goto e_inval;
859 np->min_hopcount = val;
860 retv = 0;
861 break;
862 case IPV6_DONTFRAG:
863 np->dontfrag = valbool;
864 retv = 0;
865 break;
866 case IPV6_AUTOFLOWLABEL:
867 np->autoflowlabel = valbool;
868 retv = 0;
869 break;
872 release_sock(sk);
873 if (needs_rtnl)
874 rtnl_unlock();
876 return retv;
878 e_inval:
879 release_sock(sk);
880 if (needs_rtnl)
881 rtnl_unlock();
882 return -EINVAL;
885 int ipv6_setsockopt(struct sock *sk, int level, int optname,
886 char __user *optval, unsigned int optlen)
888 int err;
890 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
891 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
893 if (level != SOL_IPV6)
894 return -ENOPROTOOPT;
896 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
897 #ifdef CONFIG_NETFILTER
898 /* we need to exclude all possible ENOPROTOOPTs except default case */
899 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
900 optname != IPV6_XFRM_POLICY) {
901 lock_sock(sk);
902 err = nf_setsockopt(sk, PF_INET6, optname, optval,
903 optlen);
904 release_sock(sk);
906 #endif
907 return err;
909 EXPORT_SYMBOL(ipv6_setsockopt);
911 #ifdef CONFIG_COMPAT
912 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
913 char __user *optval, unsigned int optlen)
915 int err;
917 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
918 if (udp_prot.compat_setsockopt != NULL)
919 return udp_prot.compat_setsockopt(sk, level, optname,
920 optval, optlen);
921 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
924 if (level != SOL_IPV6)
925 return -ENOPROTOOPT;
927 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
928 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
929 ipv6_setsockopt);
931 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
932 #ifdef CONFIG_NETFILTER
933 /* we need to exclude all possible ENOPROTOOPTs except default case */
934 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
935 optname != IPV6_XFRM_POLICY) {
936 lock_sock(sk);
937 err = compat_nf_setsockopt(sk, PF_INET6, optname,
938 optval, optlen);
939 release_sock(sk);
941 #endif
942 return err;
944 EXPORT_SYMBOL(compat_ipv6_setsockopt);
945 #endif
947 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
948 int optname, char __user *optval, int len)
950 struct ipv6_opt_hdr *hdr;
952 if (!opt)
953 return 0;
955 switch (optname) {
956 case IPV6_HOPOPTS:
957 hdr = opt->hopopt;
958 break;
959 case IPV6_RTHDRDSTOPTS:
960 hdr = opt->dst0opt;
961 break;
962 case IPV6_RTHDR:
963 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
964 break;
965 case IPV6_DSTOPTS:
966 hdr = opt->dst1opt;
967 break;
968 default:
969 return -EINVAL; /* should not happen */
972 if (!hdr)
973 return 0;
975 len = min_t(unsigned int, len, ipv6_optlen(hdr));
976 if (copy_to_user(optval, hdr, len))
977 return -EFAULT;
978 return len;
981 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
982 char __user *optval, int __user *optlen, unsigned int flags)
984 struct ipv6_pinfo *np = inet6_sk(sk);
985 int len;
986 int val;
988 if (ip6_mroute_opt(optname))
989 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
991 if (get_user(len, optlen))
992 return -EFAULT;
993 switch (optname) {
994 case IPV6_ADDRFORM:
995 if (sk->sk_protocol != IPPROTO_UDP &&
996 sk->sk_protocol != IPPROTO_UDPLITE &&
997 sk->sk_protocol != IPPROTO_TCP)
998 return -ENOPROTOOPT;
999 if (sk->sk_state != TCP_ESTABLISHED)
1000 return -ENOTCONN;
1001 val = sk->sk_family;
1002 break;
1003 case MCAST_MSFILTER:
1005 struct group_filter gsf;
1006 int err;
1008 if (len < GROUP_FILTER_SIZE(0))
1009 return -EINVAL;
1010 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
1011 return -EFAULT;
1012 if (gsf.gf_group.ss_family != AF_INET6)
1013 return -EADDRNOTAVAIL;
1014 lock_sock(sk);
1015 err = ip6_mc_msfget(sk, &gsf,
1016 (struct group_filter __user *)optval, optlen);
1017 release_sock(sk);
1018 return err;
1021 case IPV6_2292PKTOPTIONS:
1023 struct msghdr msg;
1024 struct sk_buff *skb;
1026 if (sk->sk_type != SOCK_STREAM)
1027 return -ENOPROTOOPT;
1029 msg.msg_control = optval;
1030 msg.msg_controllen = len;
1031 msg.msg_flags = flags;
1033 lock_sock(sk);
1034 skb = np->pktoptions;
1035 if (skb)
1036 ip6_datagram_recv_ctl(sk, &msg, skb);
1037 release_sock(sk);
1038 if (!skb) {
1039 if (np->rxopt.bits.rxinfo) {
1040 struct in6_pktinfo src_info;
1041 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1042 np->sticky_pktinfo.ipi6_ifindex;
1043 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1044 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1046 if (np->rxopt.bits.rxhlim) {
1047 int hlim = np->mcast_hops;
1048 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1050 if (np->rxopt.bits.rxtclass) {
1051 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1053 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1055 if (np->rxopt.bits.rxoinfo) {
1056 struct in6_pktinfo src_info;
1057 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1058 np->sticky_pktinfo.ipi6_ifindex;
1059 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1060 np->sticky_pktinfo.ipi6_addr;
1061 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1063 if (np->rxopt.bits.rxohlim) {
1064 int hlim = np->mcast_hops;
1065 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1067 if (np->rxopt.bits.rxflow) {
1068 __be32 flowinfo = np->rcv_flowinfo;
1070 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1073 len -= msg.msg_controllen;
1074 return put_user(len, optlen);
1076 case IPV6_MTU:
1078 struct dst_entry *dst;
1080 val = 0;
1081 rcu_read_lock();
1082 dst = __sk_dst_get(sk);
1083 if (dst)
1084 val = dst_mtu(dst);
1085 rcu_read_unlock();
1086 if (!val)
1087 return -ENOTCONN;
1088 break;
1091 case IPV6_V6ONLY:
1092 val = sk->sk_ipv6only;
1093 break;
1095 case IPV6_RECVPKTINFO:
1096 val = np->rxopt.bits.rxinfo;
1097 break;
1099 case IPV6_2292PKTINFO:
1100 val = np->rxopt.bits.rxoinfo;
1101 break;
1103 case IPV6_RECVHOPLIMIT:
1104 val = np->rxopt.bits.rxhlim;
1105 break;
1107 case IPV6_2292HOPLIMIT:
1108 val = np->rxopt.bits.rxohlim;
1109 break;
1111 case IPV6_RECVRTHDR:
1112 val = np->rxopt.bits.srcrt;
1113 break;
1115 case IPV6_2292RTHDR:
1116 val = np->rxopt.bits.osrcrt;
1117 break;
1119 case IPV6_HOPOPTS:
1120 case IPV6_RTHDRDSTOPTS:
1121 case IPV6_RTHDR:
1122 case IPV6_DSTOPTS:
1124 struct ipv6_txoptions *opt;
1126 lock_sock(sk);
1127 opt = rcu_dereference_protected(np->opt,
1128 lockdep_sock_is_held(sk));
1129 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1130 release_sock(sk);
1131 /* check if ipv6_getsockopt_sticky() returns err code */
1132 if (len < 0)
1133 return len;
1134 return put_user(len, optlen);
1137 case IPV6_RECVHOPOPTS:
1138 val = np->rxopt.bits.hopopts;
1139 break;
1141 case IPV6_2292HOPOPTS:
1142 val = np->rxopt.bits.ohopopts;
1143 break;
1145 case IPV6_RECVDSTOPTS:
1146 val = np->rxopt.bits.dstopts;
1147 break;
1149 case IPV6_2292DSTOPTS:
1150 val = np->rxopt.bits.odstopts;
1151 break;
1153 case IPV6_TCLASS:
1154 val = np->tclass;
1155 break;
1157 case IPV6_RECVTCLASS:
1158 val = np->rxopt.bits.rxtclass;
1159 break;
1161 case IPV6_FLOWINFO:
1162 val = np->rxopt.bits.rxflow;
1163 break;
1165 case IPV6_RECVPATHMTU:
1166 val = np->rxopt.bits.rxpmtu;
1167 break;
1169 case IPV6_PATHMTU:
1171 struct dst_entry *dst;
1172 struct ip6_mtuinfo mtuinfo;
1174 if (len < sizeof(mtuinfo))
1175 return -EINVAL;
1177 len = sizeof(mtuinfo);
1178 memset(&mtuinfo, 0, sizeof(mtuinfo));
1180 rcu_read_lock();
1181 dst = __sk_dst_get(sk);
1182 if (dst)
1183 mtuinfo.ip6m_mtu = dst_mtu(dst);
1184 rcu_read_unlock();
1185 if (!mtuinfo.ip6m_mtu)
1186 return -ENOTCONN;
1188 if (put_user(len, optlen))
1189 return -EFAULT;
1190 if (copy_to_user(optval, &mtuinfo, len))
1191 return -EFAULT;
1193 return 0;
1196 case IPV6_TRANSPARENT:
1197 val = inet_sk(sk)->transparent;
1198 break;
1200 case IPV6_RECVORIGDSTADDR:
1201 val = np->rxopt.bits.rxorigdstaddr;
1202 break;
1204 case IPV6_UNICAST_HOPS:
1205 case IPV6_MULTICAST_HOPS:
1207 struct dst_entry *dst;
1209 if (optname == IPV6_UNICAST_HOPS)
1210 val = np->hop_limit;
1211 else
1212 val = np->mcast_hops;
1214 if (val < 0) {
1215 rcu_read_lock();
1216 dst = __sk_dst_get(sk);
1217 if (dst)
1218 val = ip6_dst_hoplimit(dst);
1219 rcu_read_unlock();
1222 if (val < 0)
1223 val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1224 break;
1227 case IPV6_MULTICAST_LOOP:
1228 val = np->mc_loop;
1229 break;
1231 case IPV6_MULTICAST_IF:
1232 val = np->mcast_oif;
1233 break;
1235 case IPV6_UNICAST_IF:
1236 val = (__force int)htonl((__u32) np->ucast_oif);
1237 break;
1239 case IPV6_MTU_DISCOVER:
1240 val = np->pmtudisc;
1241 break;
1243 case IPV6_RECVERR:
1244 val = np->recverr;
1245 break;
1247 case IPV6_FLOWINFO_SEND:
1248 val = np->sndflow;
1249 break;
1251 case IPV6_FLOWLABEL_MGR:
1253 struct in6_flowlabel_req freq;
1254 int flags;
1256 if (len < sizeof(freq))
1257 return -EINVAL;
1259 if (copy_from_user(&freq, optval, sizeof(freq)))
1260 return -EFAULT;
1262 if (freq.flr_action != IPV6_FL_A_GET)
1263 return -EINVAL;
1265 len = sizeof(freq);
1266 flags = freq.flr_flags;
1268 memset(&freq, 0, sizeof(freq));
1270 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1271 if (val < 0)
1272 return val;
1274 if (put_user(len, optlen))
1275 return -EFAULT;
1276 if (copy_to_user(optval, &freq, len))
1277 return -EFAULT;
1279 return 0;
1282 case IPV6_ADDR_PREFERENCES:
1283 val = 0;
1285 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1286 val |= IPV6_PREFER_SRC_TMP;
1287 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1288 val |= IPV6_PREFER_SRC_PUBLIC;
1289 else {
1290 /* XXX: should we return system default? */
1291 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1294 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1295 val |= IPV6_PREFER_SRC_COA;
1296 else
1297 val |= IPV6_PREFER_SRC_HOME;
1298 break;
1300 case IPV6_MINHOPCOUNT:
1301 val = np->min_hopcount;
1302 break;
1304 case IPV6_DONTFRAG:
1305 val = np->dontfrag;
1306 break;
1308 case IPV6_AUTOFLOWLABEL:
1309 val = np->autoflowlabel;
1310 break;
1312 default:
1313 return -ENOPROTOOPT;
1315 len = min_t(unsigned int, sizeof(int), len);
1316 if (put_user(len, optlen))
1317 return -EFAULT;
1318 if (copy_to_user(optval, &val, len))
1319 return -EFAULT;
1320 return 0;
1323 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1324 char __user *optval, int __user *optlen)
1326 int err;
1328 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1329 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1331 if (level != SOL_IPV6)
1332 return -ENOPROTOOPT;
1334 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1335 #ifdef CONFIG_NETFILTER
1336 /* we need to exclude all possible ENOPROTOOPTs except default case */
1337 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1338 int len;
1340 if (get_user(len, optlen))
1341 return -EFAULT;
1343 lock_sock(sk);
1344 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1345 &len);
1346 release_sock(sk);
1347 if (err >= 0)
1348 err = put_user(len, optlen);
1350 #endif
1351 return err;
1353 EXPORT_SYMBOL(ipv6_getsockopt);
1355 #ifdef CONFIG_COMPAT
1356 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1357 char __user *optval, int __user *optlen)
1359 int err;
1361 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1362 if (udp_prot.compat_getsockopt != NULL)
1363 return udp_prot.compat_getsockopt(sk, level, optname,
1364 optval, optlen);
1365 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1368 if (level != SOL_IPV6)
1369 return -ENOPROTOOPT;
1371 if (optname == MCAST_MSFILTER)
1372 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1373 ipv6_getsockopt);
1375 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1376 MSG_CMSG_COMPAT);
1377 #ifdef CONFIG_NETFILTER
1378 /* we need to exclude all possible ENOPROTOOPTs except default case */
1379 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1380 int len;
1382 if (get_user(len, optlen))
1383 return -EFAULT;
1385 lock_sock(sk);
1386 err = compat_nf_getsockopt(sk, PF_INET6,
1387 optname, optval, &len);
1388 release_sock(sk);
1389 if (err >= 0)
1390 err = put_user(len, optlen);
1392 #endif
1393 return err;
1395 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1396 #endif