2 * IPv6 BSD socket options interface
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on linux/net/ipv4/ip_sockglue.c
10 * $Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
17 * FIXME: Make the setsockopt code POSIX compliant: That is
19 * o Return -EINVAL for setsockopt of short lengths
20 * o Truncate getsockopt returns
21 * o Return an optlen of the truncated length if need be
24 * David L Stevens <dlstevens@us.ibm.com>:
25 * - added multicast source filtering API for MLDv2
28 #include <linux/module.h>
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/in6.h>
36 #include <linux/netdevice.h>
37 #include <linux/if_arp.h>
38 #include <linux/init.h>
39 #include <linux/sysctl.h>
40 #include <linux/netfilter.h>
45 #include <net/ndisc.h>
46 #include <net/protocol.h>
47 #include <net/transp_v6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/inet_common.h>
53 #include <net/udplite.h>
56 #include <asm/uaccess.h>
58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
60 struct ip6_ra_chain *ip6_ra_chain;
61 DEFINE_RWLOCK(ip6_ra_lock);
63 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
65 struct ip6_ra_chain *ra, *new_ra, **rap;
67 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
68 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
71 new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
73 write_lock_bh(&ip6_ra_lock);
74 for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
77 write_unlock_bh(&ip6_ra_lock);
83 write_unlock_bh(&ip6_ra_lock);
93 write_unlock_bh(&ip6_ra_lock);
98 new_ra->destructor = destructor;
102 write_unlock_bh(&ip6_ra_lock);
106 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
107 char __user *optval, int optlen)
109 struct ipv6_pinfo *np = inet6_sk(sk);
111 int retv = -ENOPROTOOPT;
115 else if (get_user(val, (int __user *) optval))
125 if (val == PF_INET) {
126 struct ipv6_txoptions *opt;
127 struct sk_buff *pktopt;
129 if (sk->sk_protocol != IPPROTO_UDP &&
130 #ifdef CONFIG_IP_UDPLITE
131 sk->sk_protocol != IPPROTO_UDPLITE &&
133 sk->sk_protocol != IPPROTO_TCP)
136 if (sk->sk_state != TCP_ESTABLISHED) {
141 if (ipv6_only_sock(sk) ||
142 !ipv6_addr_v4mapped(&np->daddr)) {
143 retv = -EADDRNOTAVAIL;
147 fl6_free_socklist(sk);
148 ipv6_sock_mc_close(sk);
151 * Sock is moving from IPv6 to IPv4 (sk_prot), so
152 * remove it from the refcnt debug socks count in the
155 sk_refcnt_debug_dec(sk);
157 if (sk->sk_protocol == IPPROTO_TCP) {
158 struct inet_connection_sock *icsk = inet_csk(sk);
161 sock_prot_inuse_add(sk->sk_prot, -1);
162 sock_prot_inuse_add(&tcp_prot, 1);
164 sk->sk_prot = &tcp_prot;
165 icsk->icsk_af_ops = &ipv4_specific;
166 sk->sk_socket->ops = &inet_stream_ops;
167 sk->sk_family = PF_INET;
168 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
170 struct proto *prot = &udp_prot;
172 if (IS_PROTO_UDPLITE(sk->sk_protocol))
173 prot = &udplite_prot;
175 sock_prot_inuse_add(sk->sk_prot, -1);
176 sock_prot_inuse_add(prot, 1);
179 sk->sk_socket->ops = &inet_dgram_ops;
180 sk->sk_family = PF_INET;
182 opt = xchg(&np->opt, NULL);
184 sock_kfree_s(sk, opt, opt->tot_len);
185 pktopt = xchg(&np->pktoptions, NULL);
189 sk->sk_destruct = inet_sock_destruct;
191 * ... and add it to the refcnt debug socks count
192 * in the new family. -acme
194 sk_refcnt_debug_inc(sk);
195 module_put(THIS_MODULE);
202 if (inet_sk(sk)->num)
204 np->ipv6only = valbool;
208 case IPV6_RECVPKTINFO:
209 np->rxopt.bits.rxinfo = valbool;
213 case IPV6_2292PKTINFO:
214 np->rxopt.bits.rxoinfo = valbool;
218 case IPV6_RECVHOPLIMIT:
219 np->rxopt.bits.rxhlim = valbool;
223 case IPV6_2292HOPLIMIT:
224 np->rxopt.bits.rxohlim = valbool;
229 np->rxopt.bits.srcrt = valbool;
234 np->rxopt.bits.osrcrt = valbool;
238 case IPV6_RECVHOPOPTS:
239 np->rxopt.bits.hopopts = valbool;
243 case IPV6_2292HOPOPTS:
244 np->rxopt.bits.ohopopts = valbool;
248 case IPV6_RECVDSTOPTS:
249 np->rxopt.bits.dstopts = valbool;
253 case IPV6_2292DSTOPTS:
254 np->rxopt.bits.odstopts = valbool;
259 if (val < -1 || val > 0xff)
265 case IPV6_RECVTCLASS:
266 np->rxopt.bits.rxtclass = valbool;
271 np->rxopt.bits.rxflow = valbool;
276 case IPV6_RTHDRDSTOPTS:
280 struct ipv6_txoptions *opt;
284 /* hop-by-hop / destination options are privileged option */
286 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
290 if (optlen & 0x7 || optlen > 8 * 255)
293 opt = ipv6_renew_options(sk, np->opt, optname,
294 (struct ipv6_opt_hdr __user *)optval,
301 /* routing header option needs extra check */
302 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
303 struct ipv6_rt_hdr *rthdr = opt->srcrt;
304 switch (rthdr->type) {
305 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
306 case IPV6_SRCRT_TYPE_2:
313 if ((rthdr->hdrlen & 1) ||
314 (rthdr->hdrlen >> 1) != rthdr->segments_left)
319 if (inet_sk(sk)->is_icsk) {
321 struct inet_connection_sock *icsk = inet_csk(sk);
322 if (!((1 << sk->sk_state) &
323 (TCPF_LISTEN | TCPF_CLOSE))
324 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
325 icsk->icsk_ext_hdr_len =
326 opt->opt_flen + opt->opt_nflen;
327 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
330 opt = xchg(&np->opt, opt);
333 write_lock(&sk->sk_dst_lock);
334 opt = xchg(&np->opt, opt);
335 write_unlock(&sk->sk_dst_lock);
340 sock_kfree_s(sk, opt, opt->tot_len);
344 case IPV6_2292PKTOPTIONS:
346 struct ipv6_txoptions *opt = NULL;
351 fl.fl6_flowlabel = 0;
352 fl.oif = sk->sk_bound_dev_if;
357 /* 1K is probably excessive
358 * 1K is surely not enough, 2K per standard header is 16K.
361 if (optlen > 64*1024)
364 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
369 memset(opt, 0, sizeof(*opt));
370 opt->tot_len = sizeof(*opt) + optlen;
372 if (copy_from_user(opt+1, optval, optlen))
375 msg.msg_controllen = optlen;
376 msg.msg_control = (void*)(opt+1);
378 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
383 if (inet_sk(sk)->is_icsk) {
385 struct inet_connection_sock *icsk = inet_csk(sk);
386 if (!((1 << sk->sk_state) &
387 (TCPF_LISTEN | TCPF_CLOSE))
388 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
389 icsk->icsk_ext_hdr_len =
390 opt->opt_flen + opt->opt_nflen;
391 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
394 opt = xchg(&np->opt, opt);
397 write_lock(&sk->sk_dst_lock);
398 opt = xchg(&np->opt, opt);
399 write_unlock(&sk->sk_dst_lock);
405 sock_kfree_s(sk, opt, opt->tot_len);
408 case IPV6_UNICAST_HOPS:
409 if (val > 255 || val < -1)
415 case IPV6_MULTICAST_HOPS:
416 if (sk->sk_type == SOCK_STREAM)
418 if (val > 255 || val < -1)
420 np->mcast_hops = val;
424 case IPV6_MULTICAST_LOOP:
425 np->mc_loop = valbool;
429 case IPV6_MULTICAST_IF:
430 if (sk->sk_type == SOCK_STREAM)
434 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
437 if (__dev_get_by_index(&init_net, val) == NULL) {
445 case IPV6_ADD_MEMBERSHIP:
446 case IPV6_DROP_MEMBERSHIP:
448 struct ipv6_mreq mreq;
451 if (inet_sk(sk)->is_icsk)
455 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
458 if (optname == IPV6_ADD_MEMBERSHIP)
459 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
461 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
464 case IPV6_JOIN_ANYCAST:
465 case IPV6_LEAVE_ANYCAST:
467 struct ipv6_mreq mreq;
469 if (optlen != sizeof(struct ipv6_mreq))
473 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
476 if (optname == IPV6_JOIN_ANYCAST)
477 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
479 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
482 case MCAST_JOIN_GROUP:
483 case MCAST_LEAVE_GROUP:
485 struct group_req greq;
486 struct sockaddr_in6 *psin6;
489 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
491 if (greq.gr_group.ss_family != AF_INET6) {
492 retv = -EADDRNOTAVAIL;
495 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
496 if (optname == MCAST_JOIN_GROUP)
497 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
500 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
504 case MCAST_JOIN_SOURCE_GROUP:
505 case MCAST_LEAVE_SOURCE_GROUP:
506 case MCAST_BLOCK_SOURCE:
507 case MCAST_UNBLOCK_SOURCE:
509 struct group_source_req greqs;
512 if (optlen != sizeof(struct group_source_req))
514 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
518 if (greqs.gsr_group.ss_family != AF_INET6 ||
519 greqs.gsr_source.ss_family != AF_INET6) {
520 retv = -EADDRNOTAVAIL;
523 if (optname == MCAST_BLOCK_SOURCE) {
524 omode = MCAST_EXCLUDE;
526 } else if (optname == MCAST_UNBLOCK_SOURCE) {
527 omode = MCAST_EXCLUDE;
529 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
530 struct sockaddr_in6 *psin6;
532 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
533 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
535 /* prior join w/ different source is ok */
536 if (retv && retv != -EADDRINUSE)
538 omode = MCAST_INCLUDE;
540 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
541 omode = MCAST_INCLUDE;
544 retv = ip6_mc_source(add, omode, sk, &greqs);
549 extern int sysctl_mld_max_msf;
550 struct group_filter *gsf;
552 if (optlen < GROUP_FILTER_SIZE(0))
554 if (optlen > sysctl_optmem_max) {
558 gsf = kmalloc(optlen,GFP_KERNEL);
564 if (copy_from_user(gsf, optval, optlen)) {
568 /* numsrc >= (4G-140)/128 overflow in 32 bits */
569 if (gsf->gf_numsrc >= 0x1ffffffU ||
570 gsf->gf_numsrc > sysctl_mld_max_msf) {
575 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
580 retv = ip6_mc_msfilter(sk, gsf);
585 case IPV6_ROUTER_ALERT:
586 retv = ip6_ra_control(sk, val, NULL);
588 case IPV6_MTU_DISCOVER:
595 if (val && val < IPV6_MIN_MTU)
601 np->recverr = valbool;
603 skb_queue_purge(&sk->sk_error_queue);
606 case IPV6_FLOWINFO_SEND:
607 np->sndflow = valbool;
610 case IPV6_FLOWLABEL_MGR:
611 retv = ipv6_flowlabel_opt(sk, optval, optlen);
613 case IPV6_IPSEC_POLICY:
614 case IPV6_XFRM_POLICY:
616 if (!capable(CAP_NET_ADMIN))
618 retv = xfrm_user_policy(sk, optname, optval, optlen);
631 int ipv6_setsockopt(struct sock *sk, int level, int optname,
632 char __user *optval, int optlen)
636 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
637 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
639 if (level != SOL_IPV6)
642 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
643 #ifdef CONFIG_NETFILTER
644 /* we need to exclude all possible ENOPROTOOPTs except default case */
645 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
646 optname != IPV6_XFRM_POLICY) {
648 err = nf_setsockopt(sk, PF_INET6, optname, optval,
656 EXPORT_SYMBOL(ipv6_setsockopt);
659 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
660 char __user *optval, int optlen)
664 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
665 if (udp_prot.compat_setsockopt != NULL)
666 return udp_prot.compat_setsockopt(sk, level, optname,
668 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
671 if (level != SOL_IPV6)
674 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
675 #ifdef CONFIG_NETFILTER
676 /* we need to exclude all possible ENOPROTOOPTs except default case */
677 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
678 optname != IPV6_XFRM_POLICY) {
680 err = compat_nf_setsockopt(sk, PF_INET6, optname,
688 EXPORT_SYMBOL(compat_ipv6_setsockopt);
691 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
692 int optname, char __user *optval, int len)
694 struct ipv6_opt_hdr *hdr;
703 case IPV6_RTHDRDSTOPTS:
707 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
713 return -EINVAL; /* should not happen */
719 len = min_t(unsigned int, len, ipv6_optlen(hdr));
720 if (copy_to_user(optval, hdr, len))
722 return ipv6_optlen(hdr);
725 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
726 char __user *optval, int __user *optlen)
728 struct ipv6_pinfo *np = inet6_sk(sk);
732 if (get_user(len, optlen))
736 if (sk->sk_protocol != IPPROTO_UDP &&
737 #ifdef CONFIG_IP_UDPLITE
738 sk->sk_protocol != IPPROTO_UDPLITE &&
740 sk->sk_protocol != IPPROTO_TCP)
742 if (sk->sk_state != TCP_ESTABLISHED)
748 struct group_filter gsf;
751 if (len < GROUP_FILTER_SIZE(0))
753 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
756 err = ip6_mc_msfget(sk, &gsf,
757 (struct group_filter __user *)optval, optlen);
762 case IPV6_2292PKTOPTIONS:
767 if (sk->sk_type != SOCK_STREAM)
770 msg.msg_control = optval;
771 msg.msg_controllen = len;
775 skb = np->pktoptions;
777 atomic_inc(&skb->users);
781 int err = datagram_recv_ctl(sk, &msg, skb);
786 if (np->rxopt.bits.rxinfo) {
787 struct in6_pktinfo src_info;
788 src_info.ipi6_ifindex = np->mcast_oif;
789 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
790 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
792 if (np->rxopt.bits.rxhlim) {
793 int hlim = np->mcast_hops;
794 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
796 if (np->rxopt.bits.rxoinfo) {
797 struct in6_pktinfo src_info;
798 src_info.ipi6_ifindex = np->mcast_oif;
799 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
800 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
802 if (np->rxopt.bits.rxohlim) {
803 int hlim = np->mcast_hops;
804 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
807 len -= msg.msg_controllen;
808 return put_user(len, optlen);
812 struct dst_entry *dst;
815 dst = sk_dst_get(sk);
830 case IPV6_RECVPKTINFO:
831 val = np->rxopt.bits.rxinfo;
834 case IPV6_2292PKTINFO:
835 val = np->rxopt.bits.rxoinfo;
838 case IPV6_RECVHOPLIMIT:
839 val = np->rxopt.bits.rxhlim;
842 case IPV6_2292HOPLIMIT:
843 val = np->rxopt.bits.rxohlim;
847 val = np->rxopt.bits.srcrt;
851 val = np->rxopt.bits.osrcrt;
855 case IPV6_RTHDRDSTOPTS:
861 len = ipv6_getsockopt_sticky(sk, np->opt,
862 optname, optval, len);
864 return put_user(len, optlen);
867 case IPV6_RECVHOPOPTS:
868 val = np->rxopt.bits.hopopts;
871 case IPV6_2292HOPOPTS:
872 val = np->rxopt.bits.ohopopts;
875 case IPV6_RECVDSTOPTS:
876 val = np->rxopt.bits.dstopts;
879 case IPV6_2292DSTOPTS:
880 val = np->rxopt.bits.odstopts;
889 case IPV6_RECVTCLASS:
890 val = np->rxopt.bits.rxtclass;
894 val = np->rxopt.bits.rxflow;
897 case IPV6_UNICAST_HOPS:
898 case IPV6_MULTICAST_HOPS:
900 struct dst_entry *dst;
902 if (optname == IPV6_UNICAST_HOPS)
905 val = np->mcast_hops;
907 dst = sk_dst_get(sk);
910 val = dst_metric(dst, RTAX_HOPLIMIT);
912 val = ipv6_get_hoplimit(dst->dev);
916 val = ipv6_devconf.hop_limit;
920 case IPV6_MULTICAST_LOOP:
924 case IPV6_MULTICAST_IF:
928 case IPV6_MTU_DISCOVER:
936 case IPV6_FLOWINFO_SEND:
943 len = min_t(unsigned int, sizeof(int), len);
944 if(put_user(len, optlen))
946 if(copy_to_user(optval,&val,len))
951 int ipv6_getsockopt(struct sock *sk, int level, int optname,
952 char __user *optval, int __user *optlen)
956 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
957 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
959 if(level != SOL_IPV6)
962 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
963 #ifdef CONFIG_NETFILTER
964 /* we need to exclude all possible ENOPROTOOPTs except default case */
965 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
968 if (get_user(len, optlen))
972 err = nf_getsockopt(sk, PF_INET6, optname, optval,
976 err = put_user(len, optlen);
982 EXPORT_SYMBOL(ipv6_getsockopt);
985 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
986 char __user *optval, int __user *optlen)
990 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
991 if (udp_prot.compat_getsockopt != NULL)
992 return udp_prot.compat_getsockopt(sk, level, optname,
994 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
997 if (level != SOL_IPV6)
1000 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1001 #ifdef CONFIG_NETFILTER
1002 /* we need to exclude all possible ENOPROTOOPTs except default case */
1003 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1006 if (get_user(len, optlen))
1010 err = compat_nf_getsockopt(sk, PF_INET6,
1011 optname, optval, &len);
1014 err = put_user(len, optlen);
1020 EXPORT_SYMBOL(compat_ipv6_getsockopt);