mm: introduce vmf_insert_pfn_prot()
[linux/fpc-iii.git] / net / ipv6 / ipv6_sockglue.c
blob381ce38940aee2deb7e9725ff7f2aa35bc3ba12b
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>
55 #include <net/seg6.h>
57 #include <linux/uaccess.h>
59 struct ip6_ra_chain *ip6_ra_chain;
60 DEFINE_RWLOCK(ip6_ra_lock);
62 int ip6_ra_control(struct sock *sk, int sel)
64 struct ip6_ra_chain *ra, *new_ra, **rap;
66 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
67 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
68 return -ENOPROTOOPT;
70 new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
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;
81 *rap = ra->next;
82 write_unlock_bh(&ip6_ra_lock);
84 sock_put(sk);
85 kfree(ra);
86 return 0;
89 if (!new_ra) {
90 write_unlock_bh(&ip6_ra_lock);
91 return -ENOBUFS;
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;
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 struct ipv6_txoptions *opt)
105 if (inet_sk(sk)->is_icsk) {
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);
114 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
115 opt);
116 sk_dst_reset(sk);
118 return opt;
121 static bool setsockopt_needs_rtnl(int optname)
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;
138 return false;
141 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
142 char __user *optval, unsigned int optlen)
144 struct ipv6_pinfo *np = inet6_sk(sk);
145 struct net *net = sock_net(sk);
146 int val, valbool;
147 int retv = -ENOPROTOOPT;
148 bool needs_rtnl = setsockopt_needs_rtnl(optname);
150 if (!optval)
151 val = 0;
152 else {
153 if (optlen >= sizeof(int)) {
154 if (get_user(val, (int __user *) optval))
155 return -EFAULT;
156 } else
157 val = 0;
160 valbool = (val != 0);
162 if (ip6_mroute_opt(optname))
163 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
165 if (needs_rtnl)
166 rtnl_lock();
167 lock_sock(sk);
169 switch (optname) {
171 case IPV6_ADDRFORM:
172 if (optlen < sizeof(int))
173 goto e_inval;
174 if (val == PF_INET) {
175 struct ipv6_txoptions *opt;
176 struct sk_buff *pktopt;
178 if (sk->sk_type == SOCK_RAW)
179 break;
181 if (sk->sk_protocol == IPPROTO_UDP ||
182 sk->sk_protocol == IPPROTO_UDPLITE) {
183 struct udp_sock *up = udp_sk(sk);
184 if (up->pending == AF_INET6) {
185 retv = -EBUSY;
186 break;
188 } else if (sk->sk_protocol != IPPROTO_TCP)
189 break;
191 if (sk->sk_state != TCP_ESTABLISHED) {
192 retv = -ENOTCONN;
193 break;
196 if (ipv6_only_sock(sk) ||
197 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
198 retv = -EADDRNOTAVAIL;
199 break;
202 fl6_free_socklist(sk);
203 __ipv6_sock_mc_close(sk);
206 * Sock is moving from IPv6 to IPv4 (sk_prot), so
207 * remove it from the refcnt debug socks count in the
208 * original family...
210 sk_refcnt_debug_dec(sk);
212 if (sk->sk_protocol == IPPROTO_TCP) {
213 struct inet_connection_sock *icsk = inet_csk(sk);
214 local_bh_disable();
215 sock_prot_inuse_add(net, sk->sk_prot, -1);
216 sock_prot_inuse_add(net, &tcp_prot, 1);
217 local_bh_enable();
218 sk->sk_prot = &tcp_prot;
219 icsk->icsk_af_ops = &ipv4_specific;
220 sk->sk_socket->ops = &inet_stream_ops;
221 sk->sk_family = PF_INET;
222 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
223 } else {
224 struct proto *prot = &udp_prot;
226 if (sk->sk_protocol == IPPROTO_UDPLITE)
227 prot = &udplite_prot;
228 local_bh_disable();
229 sock_prot_inuse_add(net, sk->sk_prot, -1);
230 sock_prot_inuse_add(net, prot, 1);
231 local_bh_enable();
232 sk->sk_prot = prot;
233 sk->sk_socket->ops = &inet_dgram_ops;
234 sk->sk_family = PF_INET;
236 opt = xchg((__force struct ipv6_txoptions **)&np->opt,
237 NULL);
238 if (opt) {
239 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
240 txopt_put(opt);
242 pktopt = xchg(&np->pktoptions, NULL);
243 kfree_skb(pktopt);
246 * ... and add it to the refcnt debug socks count
247 * in the new family. -acme
249 sk_refcnt_debug_inc(sk);
250 module_put(THIS_MODULE);
251 retv = 0;
252 break;
254 goto e_inval;
256 case IPV6_V6ONLY:
257 if (optlen < sizeof(int) ||
258 inet_sk(sk)->inet_num)
259 goto e_inval;
260 sk->sk_ipv6only = valbool;
261 retv = 0;
262 break;
264 case IPV6_RECVPKTINFO:
265 if (optlen < sizeof(int))
266 goto e_inval;
267 np->rxopt.bits.rxinfo = valbool;
268 retv = 0;
269 break;
271 case IPV6_2292PKTINFO:
272 if (optlen < sizeof(int))
273 goto e_inval;
274 np->rxopt.bits.rxoinfo = valbool;
275 retv = 0;
276 break;
278 case IPV6_RECVHOPLIMIT:
279 if (optlen < sizeof(int))
280 goto e_inval;
281 np->rxopt.bits.rxhlim = valbool;
282 retv = 0;
283 break;
285 case IPV6_2292HOPLIMIT:
286 if (optlen < sizeof(int))
287 goto e_inval;
288 np->rxopt.bits.rxohlim = valbool;
289 retv = 0;
290 break;
292 case IPV6_RECVRTHDR:
293 if (optlen < sizeof(int))
294 goto e_inval;
295 np->rxopt.bits.srcrt = valbool;
296 retv = 0;
297 break;
299 case IPV6_2292RTHDR:
300 if (optlen < sizeof(int))
301 goto e_inval;
302 np->rxopt.bits.osrcrt = valbool;
303 retv = 0;
304 break;
306 case IPV6_RECVHOPOPTS:
307 if (optlen < sizeof(int))
308 goto e_inval;
309 np->rxopt.bits.hopopts = valbool;
310 retv = 0;
311 break;
313 case IPV6_2292HOPOPTS:
314 if (optlen < sizeof(int))
315 goto e_inval;
316 np->rxopt.bits.ohopopts = valbool;
317 retv = 0;
318 break;
320 case IPV6_RECVDSTOPTS:
321 if (optlen < sizeof(int))
322 goto e_inval;
323 np->rxopt.bits.dstopts = valbool;
324 retv = 0;
325 break;
327 case IPV6_2292DSTOPTS:
328 if (optlen < sizeof(int))
329 goto e_inval;
330 np->rxopt.bits.odstopts = valbool;
331 retv = 0;
332 break;
334 case IPV6_TCLASS:
335 if (optlen < sizeof(int))
336 goto e_inval;
337 if (val < -1 || val > 0xff)
338 goto e_inval;
339 /* RFC 3542, 6.5: default traffic class of 0x0 */
340 if (val == -1)
341 val = 0;
342 np->tclass = val;
343 retv = 0;
344 break;
346 case IPV6_RECVTCLASS:
347 if (optlen < sizeof(int))
348 goto e_inval;
349 np->rxopt.bits.rxtclass = valbool;
350 retv = 0;
351 break;
353 case IPV6_FLOWINFO:
354 if (optlen < sizeof(int))
355 goto e_inval;
356 np->rxopt.bits.rxflow = valbool;
357 retv = 0;
358 break;
360 case IPV6_RECVPATHMTU:
361 if (optlen < sizeof(int))
362 goto e_inval;
363 np->rxopt.bits.rxpmtu = valbool;
364 retv = 0;
365 break;
367 case IPV6_TRANSPARENT:
368 if (valbool && !ns_capable(net->user_ns, CAP_NET_ADMIN) &&
369 !ns_capable(net->user_ns, CAP_NET_RAW)) {
370 retv = -EPERM;
371 break;
373 if (optlen < sizeof(int))
374 goto e_inval;
375 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
376 inet_sk(sk)->transparent = valbool;
377 retv = 0;
378 break;
380 case IPV6_FREEBIND:
381 if (optlen < sizeof(int))
382 goto e_inval;
383 /* we also don't have a separate freebind bit for IPV6 */
384 inet_sk(sk)->freebind = valbool;
385 retv = 0;
386 break;
388 case IPV6_RECVORIGDSTADDR:
389 if (optlen < sizeof(int))
390 goto e_inval;
391 np->rxopt.bits.rxorigdstaddr = valbool;
392 retv = 0;
393 break;
395 case IPV6_HOPOPTS:
396 case IPV6_RTHDRDSTOPTS:
397 case IPV6_RTHDR:
398 case IPV6_DSTOPTS:
400 struct ipv6_txoptions *opt;
401 struct ipv6_opt_hdr *new = NULL;
403 /* hop-by-hop / destination options are privileged option */
404 retv = -EPERM;
405 if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
406 break;
408 /* remove any sticky options header with a zero option
409 * length, per RFC3542.
411 if (optlen == 0)
412 optval = NULL;
413 else if (!optval)
414 goto e_inval;
415 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
416 optlen & 0x7 || optlen > 8 * 255)
417 goto e_inval;
418 else {
419 new = memdup_user(optval, optlen);
420 if (IS_ERR(new)) {
421 retv = PTR_ERR(new);
422 break;
424 if (unlikely(ipv6_optlen(new) > optlen)) {
425 kfree(new);
426 goto e_inval;
430 opt = rcu_dereference_protected(np->opt,
431 lockdep_sock_is_held(sk));
432 opt = ipv6_renew_options(sk, opt, optname, new);
433 kfree(new);
434 if (IS_ERR(opt)) {
435 retv = PTR_ERR(opt);
436 break;
439 /* routing header option needs extra check */
440 retv = -EINVAL;
441 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
442 struct ipv6_rt_hdr *rthdr = opt->srcrt;
443 switch (rthdr->type) {
444 #if IS_ENABLED(CONFIG_IPV6_MIP6)
445 case IPV6_SRCRT_TYPE_2:
446 if (rthdr->hdrlen != 2 ||
447 rthdr->segments_left != 1)
448 goto sticky_done;
450 break;
451 #endif
452 case IPV6_SRCRT_TYPE_4:
454 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)
455 opt->srcrt;
457 if (!seg6_validate_srh(srh, optlen))
458 goto sticky_done;
459 break;
461 default:
462 goto sticky_done;
466 retv = 0;
467 opt = ipv6_update_options(sk, opt);
468 sticky_done:
469 if (opt) {
470 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
471 txopt_put(opt);
473 break;
476 case IPV6_PKTINFO:
478 struct in6_pktinfo pkt;
480 if (optlen == 0)
481 goto e_inval;
482 else if (optlen < sizeof(struct in6_pktinfo) || !optval)
483 goto e_inval;
485 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
486 retv = -EFAULT;
487 break;
489 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
490 goto e_inval;
492 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
493 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
494 retv = 0;
495 break;
498 case IPV6_2292PKTOPTIONS:
500 struct ipv6_txoptions *opt = NULL;
501 struct msghdr msg;
502 struct flowi6 fl6;
503 struct ipcm6_cookie ipc6;
505 memset(&fl6, 0, sizeof(fl6));
506 fl6.flowi6_oif = sk->sk_bound_dev_if;
507 fl6.flowi6_mark = sk->sk_mark;
509 if (optlen == 0)
510 goto update;
512 /* 1K is probably excessive
513 * 1K is surely not enough, 2K per standard header is 16K.
515 retv = -EINVAL;
516 if (optlen > 64*1024)
517 break;
519 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
520 retv = -ENOBUFS;
521 if (!opt)
522 break;
524 memset(opt, 0, sizeof(*opt));
525 refcount_set(&opt->refcnt, 1);
526 opt->tot_len = sizeof(*opt) + optlen;
527 retv = -EFAULT;
528 if (copy_from_user(opt+1, optval, optlen))
529 goto done;
531 msg.msg_controllen = optlen;
532 msg.msg_control = (void *)(opt+1);
533 ipc6.opt = opt;
535 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
536 if (retv)
537 goto done;
538 update:
539 retv = 0;
540 opt = ipv6_update_options(sk, opt);
541 done:
542 if (opt) {
543 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
544 txopt_put(opt);
546 break;
548 case IPV6_UNICAST_HOPS:
549 if (optlen < sizeof(int))
550 goto e_inval;
551 if (val > 255 || val < -1)
552 goto e_inval;
553 np->hop_limit = val;
554 retv = 0;
555 break;
557 case IPV6_MULTICAST_HOPS:
558 if (sk->sk_type == SOCK_STREAM)
559 break;
560 if (optlen < sizeof(int))
561 goto e_inval;
562 if (val > 255 || val < -1)
563 goto e_inval;
564 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
565 retv = 0;
566 break;
568 case IPV6_MULTICAST_LOOP:
569 if (optlen < sizeof(int))
570 goto e_inval;
571 if (val != valbool)
572 goto e_inval;
573 np->mc_loop = valbool;
574 retv = 0;
575 break;
577 case IPV6_UNICAST_IF:
579 struct net_device *dev = NULL;
580 int ifindex;
582 if (optlen != sizeof(int))
583 goto e_inval;
585 ifindex = (__force int)ntohl((__force __be32)val);
586 if (ifindex == 0) {
587 np->ucast_oif = 0;
588 retv = 0;
589 break;
592 dev = dev_get_by_index(net, ifindex);
593 retv = -EADDRNOTAVAIL;
594 if (!dev)
595 break;
596 dev_put(dev);
598 retv = -EINVAL;
599 if (sk->sk_bound_dev_if)
600 break;
602 np->ucast_oif = ifindex;
603 retv = 0;
604 break;
607 case IPV6_MULTICAST_IF:
608 if (sk->sk_type == SOCK_STREAM)
609 break;
610 if (optlen < sizeof(int))
611 goto e_inval;
613 if (val) {
614 struct net_device *dev;
615 int midx;
617 rcu_read_lock();
619 dev = dev_get_by_index_rcu(net, val);
620 if (!dev) {
621 rcu_read_unlock();
622 retv = -ENODEV;
623 break;
625 midx = l3mdev_master_ifindex_rcu(dev);
627 rcu_read_unlock();
629 if (sk->sk_bound_dev_if &&
630 sk->sk_bound_dev_if != val &&
631 (!midx || midx != sk->sk_bound_dev_if))
632 goto e_inval;
634 np->mcast_oif = val;
635 retv = 0;
636 break;
637 case IPV6_ADD_MEMBERSHIP:
638 case IPV6_DROP_MEMBERSHIP:
640 struct ipv6_mreq mreq;
642 if (optlen < sizeof(struct ipv6_mreq))
643 goto e_inval;
645 retv = -EPROTO;
646 if (inet_sk(sk)->is_icsk)
647 break;
649 retv = -EFAULT;
650 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
651 break;
653 if (optname == IPV6_ADD_MEMBERSHIP)
654 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
655 else
656 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
657 break;
659 case IPV6_JOIN_ANYCAST:
660 case IPV6_LEAVE_ANYCAST:
662 struct ipv6_mreq mreq;
664 if (optlen < sizeof(struct ipv6_mreq))
665 goto e_inval;
667 retv = -EFAULT;
668 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
669 break;
671 if (optname == IPV6_JOIN_ANYCAST)
672 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
673 else
674 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
675 break;
677 case IPV6_MULTICAST_ALL:
678 if (optlen < sizeof(int))
679 goto e_inval;
680 np->mc_all = valbool;
681 retv = 0;
682 break;
684 case MCAST_JOIN_GROUP:
685 case MCAST_LEAVE_GROUP:
687 struct group_req greq;
688 struct sockaddr_in6 *psin6;
690 if (optlen < sizeof(struct group_req))
691 goto e_inval;
693 retv = -EFAULT;
694 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
695 break;
696 if (greq.gr_group.ss_family != AF_INET6) {
697 retv = -EADDRNOTAVAIL;
698 break;
700 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
701 if (optname == MCAST_JOIN_GROUP)
702 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
703 &psin6->sin6_addr);
704 else
705 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
706 &psin6->sin6_addr);
707 break;
709 case MCAST_JOIN_SOURCE_GROUP:
710 case MCAST_LEAVE_SOURCE_GROUP:
711 case MCAST_BLOCK_SOURCE:
712 case MCAST_UNBLOCK_SOURCE:
714 struct group_source_req greqs;
715 int omode, add;
717 if (optlen < sizeof(struct group_source_req))
718 goto e_inval;
719 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
720 retv = -EFAULT;
721 break;
723 if (greqs.gsr_group.ss_family != AF_INET6 ||
724 greqs.gsr_source.ss_family != AF_INET6) {
725 retv = -EADDRNOTAVAIL;
726 break;
728 if (optname == MCAST_BLOCK_SOURCE) {
729 omode = MCAST_EXCLUDE;
730 add = 1;
731 } else if (optname == MCAST_UNBLOCK_SOURCE) {
732 omode = MCAST_EXCLUDE;
733 add = 0;
734 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
735 struct sockaddr_in6 *psin6;
737 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
738 retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
739 &psin6->sin6_addr,
740 MCAST_INCLUDE);
741 /* prior join w/ different source is ok */
742 if (retv && retv != -EADDRINUSE)
743 break;
744 omode = MCAST_INCLUDE;
745 add = 1;
746 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
747 omode = MCAST_INCLUDE;
748 add = 0;
750 retv = ip6_mc_source(add, omode, sk, &greqs);
751 break;
753 case MCAST_MSFILTER:
755 struct group_filter *gsf;
757 if (optlen < GROUP_FILTER_SIZE(0))
758 goto e_inval;
759 if (optlen > sysctl_optmem_max) {
760 retv = -ENOBUFS;
761 break;
763 gsf = memdup_user(optval, optlen);
764 if (IS_ERR(gsf)) {
765 retv = PTR_ERR(gsf);
766 break;
768 /* numsrc >= (4G-140)/128 overflow in 32 bits */
769 if (gsf->gf_numsrc >= 0x1ffffffU ||
770 gsf->gf_numsrc > sysctl_mld_max_msf) {
771 kfree(gsf);
772 retv = -ENOBUFS;
773 break;
775 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
776 kfree(gsf);
777 retv = -EINVAL;
778 break;
780 retv = ip6_mc_msfilter(sk, gsf);
781 kfree(gsf);
783 break;
785 case IPV6_ROUTER_ALERT:
786 if (optlen < sizeof(int))
787 goto e_inval;
788 retv = ip6_ra_control(sk, val);
789 break;
790 case IPV6_MTU_DISCOVER:
791 if (optlen < sizeof(int))
792 goto e_inval;
793 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
794 goto e_inval;
795 np->pmtudisc = val;
796 retv = 0;
797 break;
798 case IPV6_MTU:
799 if (optlen < sizeof(int))
800 goto e_inval;
801 if (val && val < IPV6_MIN_MTU)
802 goto e_inval;
803 np->frag_size = val;
804 retv = 0;
805 break;
806 case IPV6_RECVERR:
807 if (optlen < sizeof(int))
808 goto e_inval;
809 np->recverr = valbool;
810 if (!val)
811 skb_queue_purge(&sk->sk_error_queue);
812 retv = 0;
813 break;
814 case IPV6_FLOWINFO_SEND:
815 if (optlen < sizeof(int))
816 goto e_inval;
817 np->sndflow = valbool;
818 retv = 0;
819 break;
820 case IPV6_FLOWLABEL_MGR:
821 retv = ipv6_flowlabel_opt(sk, optval, optlen);
822 break;
823 case IPV6_IPSEC_POLICY:
824 case IPV6_XFRM_POLICY:
825 retv = -EPERM;
826 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
827 break;
828 retv = xfrm_user_policy(sk, optname, optval, optlen);
829 break;
831 case IPV6_ADDR_PREFERENCES:
833 unsigned int pref = 0;
834 unsigned int prefmask = ~0;
836 if (optlen < sizeof(int))
837 goto e_inval;
839 retv = -EINVAL;
841 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
842 switch (val & (IPV6_PREFER_SRC_PUBLIC|
843 IPV6_PREFER_SRC_TMP|
844 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
845 case IPV6_PREFER_SRC_PUBLIC:
846 pref |= IPV6_PREFER_SRC_PUBLIC;
847 break;
848 case IPV6_PREFER_SRC_TMP:
849 pref |= IPV6_PREFER_SRC_TMP;
850 break;
851 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
852 break;
853 case 0:
854 goto pref_skip_pubtmp;
855 default:
856 goto e_inval;
859 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
860 IPV6_PREFER_SRC_TMP);
861 pref_skip_pubtmp:
863 /* check HOME/COA conflicts */
864 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
865 case IPV6_PREFER_SRC_HOME:
866 break;
867 case IPV6_PREFER_SRC_COA:
868 pref |= IPV6_PREFER_SRC_COA;
869 case 0:
870 goto pref_skip_coa;
871 default:
872 goto e_inval;
875 prefmask &= ~IPV6_PREFER_SRC_COA;
876 pref_skip_coa:
878 /* check CGA/NONCGA conflicts */
879 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
880 case IPV6_PREFER_SRC_CGA:
881 case IPV6_PREFER_SRC_NONCGA:
882 case 0:
883 break;
884 default:
885 goto e_inval;
888 np->srcprefs = (np->srcprefs & prefmask) | pref;
889 retv = 0;
891 break;
893 case IPV6_MINHOPCOUNT:
894 if (optlen < sizeof(int))
895 goto e_inval;
896 if (val < 0 || val > 255)
897 goto e_inval;
898 np->min_hopcount = val;
899 retv = 0;
900 break;
901 case IPV6_DONTFRAG:
902 np->dontfrag = valbool;
903 retv = 0;
904 break;
905 case IPV6_AUTOFLOWLABEL:
906 np->autoflowlabel = valbool;
907 np->autoflowlabel_set = 1;
908 retv = 0;
909 break;
910 case IPV6_RECVFRAGSIZE:
911 np->rxopt.bits.recvfragsize = valbool;
912 retv = 0;
913 break;
916 release_sock(sk);
917 if (needs_rtnl)
918 rtnl_unlock();
920 return retv;
922 e_inval:
923 release_sock(sk);
924 if (needs_rtnl)
925 rtnl_unlock();
926 return -EINVAL;
929 int ipv6_setsockopt(struct sock *sk, int level, int optname,
930 char __user *optval, unsigned int optlen)
932 int err;
934 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
935 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
937 if (level != SOL_IPV6)
938 return -ENOPROTOOPT;
940 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
941 #ifdef CONFIG_NETFILTER
942 /* we need to exclude all possible ENOPROTOOPTs except default case */
943 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
944 optname != IPV6_XFRM_POLICY)
945 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
946 #endif
947 return err;
949 EXPORT_SYMBOL(ipv6_setsockopt);
951 #ifdef CONFIG_COMPAT
952 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
953 char __user *optval, unsigned int optlen)
955 int err;
957 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
958 if (udp_prot.compat_setsockopt != NULL)
959 return udp_prot.compat_setsockopt(sk, level, optname,
960 optval, optlen);
961 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
964 if (level != SOL_IPV6)
965 return -ENOPROTOOPT;
967 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
968 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
969 ipv6_setsockopt);
971 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
972 #ifdef CONFIG_NETFILTER
973 /* we need to exclude all possible ENOPROTOOPTs except default case */
974 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
975 optname != IPV6_XFRM_POLICY)
976 err = compat_nf_setsockopt(sk, PF_INET6, optname, optval,
977 optlen);
978 #endif
979 return err;
981 EXPORT_SYMBOL(compat_ipv6_setsockopt);
982 #endif
984 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
985 int optname, char __user *optval, int len)
987 struct ipv6_opt_hdr *hdr;
989 if (!opt)
990 return 0;
992 switch (optname) {
993 case IPV6_HOPOPTS:
994 hdr = opt->hopopt;
995 break;
996 case IPV6_RTHDRDSTOPTS:
997 hdr = opt->dst0opt;
998 break;
999 case IPV6_RTHDR:
1000 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1001 break;
1002 case IPV6_DSTOPTS:
1003 hdr = opt->dst1opt;
1004 break;
1005 default:
1006 return -EINVAL; /* should not happen */
1009 if (!hdr)
1010 return 0;
1012 len = min_t(unsigned int, len, ipv6_optlen(hdr));
1013 if (copy_to_user(optval, hdr, len))
1014 return -EFAULT;
1015 return len;
1018 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1019 char __user *optval, int __user *optlen, unsigned int flags)
1021 struct ipv6_pinfo *np = inet6_sk(sk);
1022 int len;
1023 int val;
1025 if (ip6_mroute_opt(optname))
1026 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1028 if (get_user(len, optlen))
1029 return -EFAULT;
1030 switch (optname) {
1031 case IPV6_ADDRFORM:
1032 if (sk->sk_protocol != IPPROTO_UDP &&
1033 sk->sk_protocol != IPPROTO_UDPLITE &&
1034 sk->sk_protocol != IPPROTO_TCP)
1035 return -ENOPROTOOPT;
1036 if (sk->sk_state != TCP_ESTABLISHED)
1037 return -ENOTCONN;
1038 val = sk->sk_family;
1039 break;
1040 case MCAST_MSFILTER:
1042 struct group_filter gsf;
1043 int err;
1045 if (len < GROUP_FILTER_SIZE(0))
1046 return -EINVAL;
1047 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
1048 return -EFAULT;
1049 if (gsf.gf_group.ss_family != AF_INET6)
1050 return -EADDRNOTAVAIL;
1051 lock_sock(sk);
1052 err = ip6_mc_msfget(sk, &gsf,
1053 (struct group_filter __user *)optval, optlen);
1054 release_sock(sk);
1055 return err;
1058 case IPV6_2292PKTOPTIONS:
1060 struct msghdr msg;
1061 struct sk_buff *skb;
1063 if (sk->sk_type != SOCK_STREAM)
1064 return -ENOPROTOOPT;
1066 msg.msg_control = optval;
1067 msg.msg_controllen = len;
1068 msg.msg_flags = flags;
1070 lock_sock(sk);
1071 skb = np->pktoptions;
1072 if (skb)
1073 ip6_datagram_recv_ctl(sk, &msg, skb);
1074 release_sock(sk);
1075 if (!skb) {
1076 if (np->rxopt.bits.rxinfo) {
1077 struct in6_pktinfo src_info;
1078 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1079 np->sticky_pktinfo.ipi6_ifindex;
1080 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1081 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1083 if (np->rxopt.bits.rxhlim) {
1084 int hlim = np->mcast_hops;
1085 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1087 if (np->rxopt.bits.rxtclass) {
1088 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1090 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1092 if (np->rxopt.bits.rxoinfo) {
1093 struct in6_pktinfo src_info;
1094 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1095 np->sticky_pktinfo.ipi6_ifindex;
1096 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1097 np->sticky_pktinfo.ipi6_addr;
1098 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1100 if (np->rxopt.bits.rxohlim) {
1101 int hlim = np->mcast_hops;
1102 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1104 if (np->rxopt.bits.rxflow) {
1105 __be32 flowinfo = np->rcv_flowinfo;
1107 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1110 len -= msg.msg_controllen;
1111 return put_user(len, optlen);
1113 case IPV6_MTU:
1115 struct dst_entry *dst;
1117 val = 0;
1118 rcu_read_lock();
1119 dst = __sk_dst_get(sk);
1120 if (dst)
1121 val = dst_mtu(dst);
1122 rcu_read_unlock();
1123 if (!val)
1124 return -ENOTCONN;
1125 break;
1128 case IPV6_V6ONLY:
1129 val = sk->sk_ipv6only;
1130 break;
1132 case IPV6_RECVPKTINFO:
1133 val = np->rxopt.bits.rxinfo;
1134 break;
1136 case IPV6_2292PKTINFO:
1137 val = np->rxopt.bits.rxoinfo;
1138 break;
1140 case IPV6_RECVHOPLIMIT:
1141 val = np->rxopt.bits.rxhlim;
1142 break;
1144 case IPV6_2292HOPLIMIT:
1145 val = np->rxopt.bits.rxohlim;
1146 break;
1148 case IPV6_RECVRTHDR:
1149 val = np->rxopt.bits.srcrt;
1150 break;
1152 case IPV6_2292RTHDR:
1153 val = np->rxopt.bits.osrcrt;
1154 break;
1156 case IPV6_HOPOPTS:
1157 case IPV6_RTHDRDSTOPTS:
1158 case IPV6_RTHDR:
1159 case IPV6_DSTOPTS:
1161 struct ipv6_txoptions *opt;
1163 lock_sock(sk);
1164 opt = rcu_dereference_protected(np->opt,
1165 lockdep_sock_is_held(sk));
1166 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1167 release_sock(sk);
1168 /* check if ipv6_getsockopt_sticky() returns err code */
1169 if (len < 0)
1170 return len;
1171 return put_user(len, optlen);
1174 case IPV6_RECVHOPOPTS:
1175 val = np->rxopt.bits.hopopts;
1176 break;
1178 case IPV6_2292HOPOPTS:
1179 val = np->rxopt.bits.ohopopts;
1180 break;
1182 case IPV6_RECVDSTOPTS:
1183 val = np->rxopt.bits.dstopts;
1184 break;
1186 case IPV6_2292DSTOPTS:
1187 val = np->rxopt.bits.odstopts;
1188 break;
1190 case IPV6_TCLASS:
1191 val = np->tclass;
1192 break;
1194 case IPV6_RECVTCLASS:
1195 val = np->rxopt.bits.rxtclass;
1196 break;
1198 case IPV6_FLOWINFO:
1199 val = np->rxopt.bits.rxflow;
1200 break;
1202 case IPV6_RECVPATHMTU:
1203 val = np->rxopt.bits.rxpmtu;
1204 break;
1206 case IPV6_PATHMTU:
1208 struct dst_entry *dst;
1209 struct ip6_mtuinfo mtuinfo;
1211 if (len < sizeof(mtuinfo))
1212 return -EINVAL;
1214 len = sizeof(mtuinfo);
1215 memset(&mtuinfo, 0, sizeof(mtuinfo));
1217 rcu_read_lock();
1218 dst = __sk_dst_get(sk);
1219 if (dst)
1220 mtuinfo.ip6m_mtu = dst_mtu(dst);
1221 rcu_read_unlock();
1222 if (!mtuinfo.ip6m_mtu)
1223 return -ENOTCONN;
1225 if (put_user(len, optlen))
1226 return -EFAULT;
1227 if (copy_to_user(optval, &mtuinfo, len))
1228 return -EFAULT;
1230 return 0;
1233 case IPV6_TRANSPARENT:
1234 val = inet_sk(sk)->transparent;
1235 break;
1237 case IPV6_FREEBIND:
1238 val = inet_sk(sk)->freebind;
1239 break;
1241 case IPV6_RECVORIGDSTADDR:
1242 val = np->rxopt.bits.rxorigdstaddr;
1243 break;
1245 case IPV6_UNICAST_HOPS:
1246 case IPV6_MULTICAST_HOPS:
1248 struct dst_entry *dst;
1250 if (optname == IPV6_UNICAST_HOPS)
1251 val = np->hop_limit;
1252 else
1253 val = np->mcast_hops;
1255 if (val < 0) {
1256 rcu_read_lock();
1257 dst = __sk_dst_get(sk);
1258 if (dst)
1259 val = ip6_dst_hoplimit(dst);
1260 rcu_read_unlock();
1263 if (val < 0)
1264 val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1265 break;
1268 case IPV6_MULTICAST_LOOP:
1269 val = np->mc_loop;
1270 break;
1272 case IPV6_MULTICAST_IF:
1273 val = np->mcast_oif;
1274 break;
1276 case IPV6_MULTICAST_ALL:
1277 val = np->mc_all;
1278 break;
1280 case IPV6_UNICAST_IF:
1281 val = (__force int)htonl((__u32) np->ucast_oif);
1282 break;
1284 case IPV6_MTU_DISCOVER:
1285 val = np->pmtudisc;
1286 break;
1288 case IPV6_RECVERR:
1289 val = np->recverr;
1290 break;
1292 case IPV6_FLOWINFO_SEND:
1293 val = np->sndflow;
1294 break;
1296 case IPV6_FLOWLABEL_MGR:
1298 struct in6_flowlabel_req freq;
1299 int flags;
1301 if (len < sizeof(freq))
1302 return -EINVAL;
1304 if (copy_from_user(&freq, optval, sizeof(freq)))
1305 return -EFAULT;
1307 if (freq.flr_action != IPV6_FL_A_GET)
1308 return -EINVAL;
1310 len = sizeof(freq);
1311 flags = freq.flr_flags;
1313 memset(&freq, 0, sizeof(freq));
1315 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1316 if (val < 0)
1317 return val;
1319 if (put_user(len, optlen))
1320 return -EFAULT;
1321 if (copy_to_user(optval, &freq, len))
1322 return -EFAULT;
1324 return 0;
1327 case IPV6_ADDR_PREFERENCES:
1328 val = 0;
1330 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1331 val |= IPV6_PREFER_SRC_TMP;
1332 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1333 val |= IPV6_PREFER_SRC_PUBLIC;
1334 else {
1335 /* XXX: should we return system default? */
1336 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1339 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1340 val |= IPV6_PREFER_SRC_COA;
1341 else
1342 val |= IPV6_PREFER_SRC_HOME;
1343 break;
1345 case IPV6_MINHOPCOUNT:
1346 val = np->min_hopcount;
1347 break;
1349 case IPV6_DONTFRAG:
1350 val = np->dontfrag;
1351 break;
1353 case IPV6_AUTOFLOWLABEL:
1354 val = ip6_autoflowlabel(sock_net(sk), np);
1355 break;
1357 case IPV6_RECVFRAGSIZE:
1358 val = np->rxopt.bits.recvfragsize;
1359 break;
1361 default:
1362 return -ENOPROTOOPT;
1364 len = min_t(unsigned int, sizeof(int), len);
1365 if (put_user(len, optlen))
1366 return -EFAULT;
1367 if (copy_to_user(optval, &val, len))
1368 return -EFAULT;
1369 return 0;
1372 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1373 char __user *optval, int __user *optlen)
1375 int err;
1377 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1378 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1380 if (level != SOL_IPV6)
1381 return -ENOPROTOOPT;
1383 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1384 #ifdef CONFIG_NETFILTER
1385 /* we need to exclude all possible ENOPROTOOPTs except default case */
1386 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1387 int len;
1389 if (get_user(len, optlen))
1390 return -EFAULT;
1392 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1393 if (err >= 0)
1394 err = put_user(len, optlen);
1396 #endif
1397 return err;
1399 EXPORT_SYMBOL(ipv6_getsockopt);
1401 #ifdef CONFIG_COMPAT
1402 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1403 char __user *optval, int __user *optlen)
1405 int err;
1407 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1408 if (udp_prot.compat_getsockopt != NULL)
1409 return udp_prot.compat_getsockopt(sk, level, optname,
1410 optval, optlen);
1411 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1414 if (level != SOL_IPV6)
1415 return -ENOPROTOOPT;
1417 if (optname == MCAST_MSFILTER)
1418 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1419 ipv6_getsockopt);
1421 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1422 MSG_CMSG_COMPAT);
1423 #ifdef CONFIG_NETFILTER
1424 /* we need to exclude all possible ENOPROTOOPTs except default case */
1425 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1426 int len;
1428 if (get_user(len, optlen))
1429 return -EFAULT;
1431 err = compat_nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1432 if (err >= 0)
1433 err = put_user(len, optlen);
1435 #endif
1436 return err;
1438 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1439 #endif