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/sched.h>
35 #include <linux/net.h>
36 #include <linux/in6.h>
37 #include <linux/netdevice.h>
38 #include <linux/if_arp.h>
39 #include <linux/init.h>
40 #include <linux/sysctl.h>
41 #include <linux/netfilter.h>
46 #include <net/ndisc.h>
47 #include <net/protocol.h>
48 #include <net/transp_v6.h>
49 #include <net/ip6_route.h>
50 #include <net/addrconf.h>
51 #include <net/inet_common.h>
56 #include <asm/uaccess.h>
58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
60 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
62 struct sk_buff *segs = ERR_PTR(-EINVAL);
63 struct ipv6hdr *ipv6h;
64 struct inet6_protocol *ops;
67 if (unlikely(skb_shinfo(skb)->gso_type &
75 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
78 ipv6h = skb->nh.ipv6h;
79 proto = ipv6h->nexthdr;
80 __skb_pull(skb, sizeof(*ipv6h));
84 struct ipv6_opt_hdr *opth;
87 if (proto != NEXTHDR_HOP) {
88 ops = rcu_dereference(inet6_protos[proto]);
93 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
97 if (unlikely(!pskb_may_pull(skb, 8)))
100 opth = (void *)skb->data;
101 len = opth->hdrlen * 8 + 8;
103 if (unlikely(!pskb_may_pull(skb, len)))
106 proto = opth->nexthdr;
107 __skb_pull(skb, len);
110 skb->h.raw = skb->data;
111 if (likely(ops->gso_segment))
112 segs = ops->gso_segment(skb, features);
117 if (unlikely(IS_ERR(segs)))
120 for (skb = segs; skb; skb = skb->next) {
121 ipv6h = skb->nh.ipv6h;
122 ipv6h->payload_len = htons(skb->len - skb->mac_len -
130 static struct packet_type ipv6_packet_type = {
131 .type = __constant_htons(ETH_P_IPV6),
133 .gso_segment = ipv6_gso_segment,
136 struct ip6_ra_chain *ip6_ra_chain;
137 DEFINE_RWLOCK(ip6_ra_lock);
139 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
141 struct ip6_ra_chain *ra, *new_ra, **rap;
143 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
144 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
147 new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
149 write_lock_bh(&ip6_ra_lock);
150 for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
153 write_unlock_bh(&ip6_ra_lock);
159 write_unlock_bh(&ip6_ra_lock);
168 if (new_ra == NULL) {
169 write_unlock_bh(&ip6_ra_lock);
174 new_ra->destructor = destructor;
178 write_unlock_bh(&ip6_ra_lock);
182 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
183 char __user *optval, int optlen)
185 struct ipv6_pinfo *np = inet6_sk(sk);
187 int retv = -ENOPROTOOPT;
191 else if (get_user(val, (int __user *) optval))
201 if (val == PF_INET) {
202 struct ipv6_txoptions *opt;
203 struct sk_buff *pktopt;
205 if (sk->sk_protocol != IPPROTO_UDP &&
206 sk->sk_protocol != IPPROTO_TCP)
209 if (sk->sk_state != TCP_ESTABLISHED) {
214 if (ipv6_only_sock(sk) ||
215 !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
216 retv = -EADDRNOTAVAIL;
220 fl6_free_socklist(sk);
221 ipv6_sock_mc_close(sk);
224 * Sock is moving from IPv6 to IPv4 (sk_prot), so
225 * remove it from the refcnt debug socks count in the
228 sk_refcnt_debug_dec(sk);
230 if (sk->sk_protocol == IPPROTO_TCP) {
231 struct inet_connection_sock *icsk = inet_csk(sk);
234 sock_prot_dec_use(sk->sk_prot);
235 sock_prot_inc_use(&tcp_prot);
237 sk->sk_prot = &tcp_prot;
238 icsk->icsk_af_ops = &ipv4_specific;
239 sk->sk_socket->ops = &inet_stream_ops;
240 sk->sk_family = PF_INET;
241 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
244 sock_prot_dec_use(sk->sk_prot);
245 sock_prot_inc_use(&udp_prot);
247 sk->sk_prot = &udp_prot;
248 sk->sk_socket->ops = &inet_dgram_ops;
249 sk->sk_family = PF_INET;
251 opt = xchg(&np->opt, NULL);
253 sock_kfree_s(sk, opt, opt->tot_len);
254 pktopt = xchg(&np->pktoptions, NULL);
258 sk->sk_destruct = inet_sock_destruct;
260 * ... and add it to the refcnt debug socks count
261 * in the new family. -acme
263 sk_refcnt_debug_inc(sk);
264 module_put(THIS_MODULE);
271 if (inet_sk(sk)->num)
273 np->ipv6only = valbool;
277 case IPV6_RECVPKTINFO:
278 np->rxopt.bits.rxinfo = valbool;
282 case IPV6_2292PKTINFO:
283 np->rxopt.bits.rxoinfo = valbool;
287 case IPV6_RECVHOPLIMIT:
288 np->rxopt.bits.rxhlim = valbool;
292 case IPV6_2292HOPLIMIT:
293 np->rxopt.bits.rxohlim = valbool;
298 if (val < 0 || val > 2)
300 np->rxopt.bits.srcrt = val;
305 if (val < 0 || val > 2)
307 np->rxopt.bits.osrcrt = val;
311 case IPV6_RECVHOPOPTS:
312 np->rxopt.bits.hopopts = valbool;
316 case IPV6_2292HOPOPTS:
317 np->rxopt.bits.ohopopts = valbool;
321 case IPV6_RECVDSTOPTS:
322 np->rxopt.bits.dstopts = valbool;
326 case IPV6_2292DSTOPTS:
327 np->rxopt.bits.odstopts = valbool;
332 if (val < 0 || val > 0xff)
338 case IPV6_RECVTCLASS:
339 np->rxopt.bits.rxtclass = valbool;
344 np->rxopt.bits.rxflow = valbool;
349 case IPV6_RTHDRDSTOPTS:
353 struct ipv6_txoptions *opt;
357 /* hop-by-hop / destination options are privileged option */
359 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
363 if (optlen & 0x7 || optlen > 8 * 255)
366 opt = ipv6_renew_options(sk, np->opt, optname,
367 (struct ipv6_opt_hdr __user *)optval,
374 /* routing header option needs extra check */
375 if (optname == IPV6_RTHDR && opt->srcrt) {
376 struct ipv6_rt_hdr *rthdr = opt->srcrt;
379 if ((rthdr->hdrlen & 1) ||
380 (rthdr->hdrlen >> 1) != rthdr->segments_left)
385 if (inet_sk(sk)->is_icsk) {
387 struct inet_connection_sock *icsk = inet_csk(sk);
388 if (!((1 << sk->sk_state) &
389 (TCPF_LISTEN | TCPF_CLOSE))
390 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
391 icsk->icsk_ext_hdr_len =
392 opt->opt_flen + opt->opt_nflen;
393 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
396 opt = xchg(&np->opt, opt);
399 write_lock(&sk->sk_dst_lock);
400 opt = xchg(&np->opt, opt);
401 write_unlock(&sk->sk_dst_lock);
406 sock_kfree_s(sk, opt, opt->tot_len);
410 case IPV6_2292PKTOPTIONS:
412 struct ipv6_txoptions *opt = NULL;
417 fl.fl6_flowlabel = 0;
418 fl.oif = sk->sk_bound_dev_if;
423 /* 1K is probably excessive
424 * 1K is surely not enough, 2K per standard header is 16K.
427 if (optlen > 64*1024)
430 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
435 memset(opt, 0, sizeof(*opt));
436 opt->tot_len = sizeof(*opt) + optlen;
438 if (copy_from_user(opt+1, optval, optlen))
441 msg.msg_controllen = optlen;
442 msg.msg_control = (void*)(opt+1);
444 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
449 if (inet_sk(sk)->is_icsk) {
451 struct inet_connection_sock *icsk = inet_csk(sk);
452 if (!((1 << sk->sk_state) &
453 (TCPF_LISTEN | TCPF_CLOSE))
454 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
455 icsk->icsk_ext_hdr_len =
456 opt->opt_flen + opt->opt_nflen;
457 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
460 opt = xchg(&np->opt, opt);
463 write_lock(&sk->sk_dst_lock);
464 opt = xchg(&np->opt, opt);
465 write_unlock(&sk->sk_dst_lock);
471 sock_kfree_s(sk, opt, opt->tot_len);
474 case IPV6_UNICAST_HOPS:
475 if (val > 255 || val < -1)
481 case IPV6_MULTICAST_HOPS:
482 if (sk->sk_type == SOCK_STREAM)
484 if (val > 255 || val < -1)
486 np->mcast_hops = val;
490 case IPV6_MULTICAST_LOOP:
491 np->mc_loop = valbool;
495 case IPV6_MULTICAST_IF:
496 if (sk->sk_type == SOCK_STREAM)
498 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
501 if (__dev_get_by_index(val) == NULL) {
508 case IPV6_ADD_MEMBERSHIP:
509 case IPV6_DROP_MEMBERSHIP:
511 struct ipv6_mreq mreq;
514 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
517 if (optname == IPV6_ADD_MEMBERSHIP)
518 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
520 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
523 case IPV6_JOIN_ANYCAST:
524 case IPV6_LEAVE_ANYCAST:
526 struct ipv6_mreq mreq;
528 if (optlen != sizeof(struct ipv6_mreq))
532 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
535 if (optname == IPV6_JOIN_ANYCAST)
536 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
538 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
541 case MCAST_JOIN_GROUP:
542 case MCAST_LEAVE_GROUP:
544 struct group_req greq;
545 struct sockaddr_in6 *psin6;
548 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
550 if (greq.gr_group.ss_family != AF_INET6) {
551 retv = -EADDRNOTAVAIL;
554 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
555 if (optname == MCAST_JOIN_GROUP)
556 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
559 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
563 case MCAST_JOIN_SOURCE_GROUP:
564 case MCAST_LEAVE_SOURCE_GROUP:
565 case MCAST_BLOCK_SOURCE:
566 case MCAST_UNBLOCK_SOURCE:
568 struct group_source_req greqs;
571 if (optlen != sizeof(struct group_source_req))
573 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
577 if (greqs.gsr_group.ss_family != AF_INET6 ||
578 greqs.gsr_source.ss_family != AF_INET6) {
579 retv = -EADDRNOTAVAIL;
582 if (optname == MCAST_BLOCK_SOURCE) {
583 omode = MCAST_EXCLUDE;
585 } else if (optname == MCAST_UNBLOCK_SOURCE) {
586 omode = MCAST_EXCLUDE;
588 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
589 struct sockaddr_in6 *psin6;
591 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
592 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
594 /* prior join w/ different source is ok */
595 if (retv && retv != -EADDRINUSE)
597 omode = MCAST_INCLUDE;
599 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
600 omode = MCAST_INCLUDE;
603 retv = ip6_mc_source(add, omode, sk, &greqs);
608 extern int sysctl_mld_max_msf;
609 struct group_filter *gsf;
611 if (optlen < GROUP_FILTER_SIZE(0))
613 if (optlen > sysctl_optmem_max) {
617 gsf = kmalloc(optlen,GFP_KERNEL);
623 if (copy_from_user(gsf, optval, optlen)) {
627 /* numsrc >= (4G-140)/128 overflow in 32 bits */
628 if (gsf->gf_numsrc >= 0x1ffffffU ||
629 gsf->gf_numsrc > sysctl_mld_max_msf) {
634 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
639 retv = ip6_mc_msfilter(sk, gsf);
644 case IPV6_ROUTER_ALERT:
645 retv = ip6_ra_control(sk, val, NULL);
647 case IPV6_MTU_DISCOVER:
654 if (val && val < IPV6_MIN_MTU)
660 np->recverr = valbool;
662 skb_queue_purge(&sk->sk_error_queue);
665 case IPV6_FLOWINFO_SEND:
666 np->sndflow = valbool;
669 case IPV6_FLOWLABEL_MGR:
670 retv = ipv6_flowlabel_opt(sk, optval, optlen);
672 case IPV6_IPSEC_POLICY:
673 case IPV6_XFRM_POLICY:
675 if (!capable(CAP_NET_ADMIN))
677 retv = xfrm_user_policy(sk, optname, optval, optlen);
690 int ipv6_setsockopt(struct sock *sk, int level, int optname,
691 char __user *optval, int optlen)
695 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
696 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
698 if (level != SOL_IPV6)
701 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
702 #ifdef CONFIG_NETFILTER
703 /* we need to exclude all possible ENOPROTOOPTs except default case */
704 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
705 optname != IPV6_XFRM_POLICY) {
707 err = nf_setsockopt(sk, PF_INET6, optname, optval,
717 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
718 char __user *optval, int optlen)
722 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
723 if (udp_prot.compat_setsockopt != NULL)
724 return udp_prot.compat_setsockopt(sk, level, optname,
726 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
729 if (level != SOL_IPV6)
732 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
733 #ifdef CONFIG_NETFILTER
734 /* we need to exclude all possible ENOPROTOOPTs except default case */
735 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
736 optname != IPV6_XFRM_POLICY) {
738 err = compat_nf_setsockopt(sk, PF_INET6, optname,
746 EXPORT_SYMBOL(compat_ipv6_setsockopt);
749 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
750 char __user *optval, int len)
754 len = min_t(int, len, ipv6_optlen(hdr));
755 if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
760 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
761 char __user *optval, int __user *optlen)
763 struct ipv6_pinfo *np = inet6_sk(sk);
767 if (get_user(len, optlen))
771 if (sk->sk_protocol != IPPROTO_UDP &&
772 sk->sk_protocol != IPPROTO_TCP)
774 if (sk->sk_state != TCP_ESTABLISHED)
780 struct group_filter gsf;
783 if (len < GROUP_FILTER_SIZE(0))
785 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
788 err = ip6_mc_msfget(sk, &gsf,
789 (struct group_filter __user *)optval, optlen);
794 case IPV6_2292PKTOPTIONS:
799 if (sk->sk_type != SOCK_STREAM)
802 msg.msg_control = optval;
803 msg.msg_controllen = len;
807 skb = np->pktoptions;
809 atomic_inc(&skb->users);
813 int err = datagram_recv_ctl(sk, &msg, skb);
818 if (np->rxopt.bits.rxinfo) {
819 struct in6_pktinfo src_info;
820 src_info.ipi6_ifindex = np->mcast_oif;
821 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
822 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
824 if (np->rxopt.bits.rxhlim) {
825 int hlim = np->mcast_hops;
826 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
828 if (np->rxopt.bits.rxoinfo) {
829 struct in6_pktinfo src_info;
830 src_info.ipi6_ifindex = np->mcast_oif;
831 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
832 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
834 if (np->rxopt.bits.rxohlim) {
835 int hlim = np->mcast_hops;
836 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
839 len -= msg.msg_controllen;
840 return put_user(len, optlen);
844 struct dst_entry *dst;
847 dst = sk_dst_get(sk);
862 case IPV6_RECVPKTINFO:
863 val = np->rxopt.bits.rxinfo;
866 case IPV6_2292PKTINFO:
867 val = np->rxopt.bits.rxoinfo;
870 case IPV6_RECVHOPLIMIT:
871 val = np->rxopt.bits.rxhlim;
874 case IPV6_2292HOPLIMIT:
875 val = np->rxopt.bits.rxohlim;
879 val = np->rxopt.bits.srcrt;
883 val = np->rxopt.bits.osrcrt;
887 case IPV6_RTHDRDSTOPTS:
893 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
896 return put_user(len, optlen);
899 case IPV6_RECVHOPOPTS:
900 val = np->rxopt.bits.hopopts;
903 case IPV6_2292HOPOPTS:
904 val = np->rxopt.bits.ohopopts;
907 case IPV6_RECVDSTOPTS:
908 val = np->rxopt.bits.dstopts;
911 case IPV6_2292DSTOPTS:
912 val = np->rxopt.bits.odstopts;
919 case IPV6_RECVTCLASS:
920 val = np->rxopt.bits.rxtclass;
924 val = np->rxopt.bits.rxflow;
927 case IPV6_UNICAST_HOPS:
931 case IPV6_MULTICAST_HOPS:
932 val = np->mcast_hops;
935 case IPV6_MULTICAST_LOOP:
939 case IPV6_MULTICAST_IF:
943 case IPV6_MTU_DISCOVER:
951 case IPV6_FLOWINFO_SEND:
958 len = min_t(unsigned int, sizeof(int), len);
959 if(put_user(len, optlen))
961 if(copy_to_user(optval,&val,len))
966 int ipv6_getsockopt(struct sock *sk, int level, int optname,
967 char __user *optval, int __user *optlen)
971 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
972 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
974 if(level != SOL_IPV6)
977 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
978 #ifdef CONFIG_NETFILTER
979 /* we need to exclude all possible EINVALs except default case */
980 if (err == -EINVAL && optname != IPV6_ADDRFORM &&
981 optname != MCAST_MSFILTER) {
984 if (get_user(len, optlen))
988 err = nf_getsockopt(sk, PF_INET6, optname, optval,
992 err = put_user(len, optlen);
999 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1000 char __user *optval, int __user *optlen)
1004 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1005 if (udp_prot.compat_getsockopt != NULL)
1006 return udp_prot.compat_getsockopt(sk, level, optname,
1008 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1011 if (level != SOL_IPV6)
1012 return -ENOPROTOOPT;
1014 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1015 #ifdef CONFIG_NETFILTER
1016 /* we need to exclude all possible EINVALs except default case */
1017 if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1018 optname != MCAST_MSFILTER) {
1021 if (get_user(len, optlen))
1025 err = compat_nf_getsockopt(sk, PF_INET6,
1026 optname, optval, &len);
1029 err = put_user(len, optlen);
1035 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1038 void __init ipv6_packet_init(void)
1040 dev_add_pack(&ipv6_packet_type);
1043 void ipv6_packet_cleanup(void)
1045 dev_remove_pack(&ipv6_packet_type);