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/config.h>
31 #include <linux/errno.h>
32 #include <linux/types.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/sched.h>
36 #include <linux/net.h>
37 #include <linux/in6.h>
38 #include <linux/netdevice.h>
39 #include <linux/if_arp.h>
40 #include <linux/init.h>
41 #include <linux/sysctl.h>
42 #include <linux/netfilter.h>
47 #include <net/ndisc.h>
48 #include <net/protocol.h>
49 #include <net/transp_v6.h>
50 #include <net/ip6_route.h>
51 #include <net/addrconf.h>
52 #include <net/inet_common.h>
57 #include <asm/uaccess.h>
59 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
61 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
63 struct sk_buff *segs = ERR_PTR(-EINVAL);
64 struct ipv6hdr *ipv6h;
65 struct inet6_protocol *ops;
68 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
71 ipv6h = skb->nh.ipv6h;
72 proto = ipv6h->nexthdr;
73 __skb_pull(skb, sizeof(*ipv6h));
77 struct ipv6_opt_hdr *opth;
80 if (proto != NEXTHDR_HOP) {
81 ops = rcu_dereference(inet6_protos[proto]);
86 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
90 if (unlikely(!pskb_may_pull(skb, 8)))
93 opth = (void *)skb->data;
94 len = opth->hdrlen * 8 + 8;
96 if (unlikely(!pskb_may_pull(skb, len)))
99 proto = opth->nexthdr;
100 __skb_pull(skb, len);
103 skb->h.raw = skb->data;
104 if (likely(ops->gso_segment))
105 segs = ops->gso_segment(skb, features);
110 if (unlikely(IS_ERR(segs)))
113 for (skb = segs; skb; skb = skb->next) {
114 ipv6h = skb->nh.ipv6h;
115 ipv6h->payload_len = htons(skb->len - skb->mac_len);
122 static struct packet_type ipv6_packet_type = {
123 .type = __constant_htons(ETH_P_IPV6),
125 .gso_segment = ipv6_gso_segment,
128 struct ip6_ra_chain *ip6_ra_chain;
129 DEFINE_RWLOCK(ip6_ra_lock);
131 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
133 struct ip6_ra_chain *ra, *new_ra, **rap;
135 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
136 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
139 new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
141 write_lock_bh(&ip6_ra_lock);
142 for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
145 write_unlock_bh(&ip6_ra_lock);
151 write_unlock_bh(&ip6_ra_lock);
160 if (new_ra == NULL) {
161 write_unlock_bh(&ip6_ra_lock);
166 new_ra->destructor = destructor;
170 write_unlock_bh(&ip6_ra_lock);
174 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
175 char __user *optval, int optlen)
177 struct ipv6_pinfo *np = inet6_sk(sk);
179 int retv = -ENOPROTOOPT;
183 else if (get_user(val, (int __user *) optval))
193 if (val == PF_INET) {
194 struct ipv6_txoptions *opt;
195 struct sk_buff *pktopt;
197 if (sk->sk_protocol != IPPROTO_UDP &&
198 sk->sk_protocol != IPPROTO_TCP)
201 if (sk->sk_state != TCP_ESTABLISHED) {
206 if (ipv6_only_sock(sk) ||
207 !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
208 retv = -EADDRNOTAVAIL;
212 fl6_free_socklist(sk);
213 ipv6_sock_mc_close(sk);
216 * Sock is moving from IPv6 to IPv4 (sk_prot), so
217 * remove it from the refcnt debug socks count in the
220 sk_refcnt_debug_dec(sk);
222 if (sk->sk_protocol == IPPROTO_TCP) {
223 struct inet_connection_sock *icsk = inet_csk(sk);
226 sock_prot_dec_use(sk->sk_prot);
227 sock_prot_inc_use(&tcp_prot);
229 sk->sk_prot = &tcp_prot;
230 icsk->icsk_af_ops = &ipv4_specific;
231 sk->sk_socket->ops = &inet_stream_ops;
232 sk->sk_family = PF_INET;
233 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
236 sock_prot_dec_use(sk->sk_prot);
237 sock_prot_inc_use(&udp_prot);
239 sk->sk_prot = &udp_prot;
240 sk->sk_socket->ops = &inet_dgram_ops;
241 sk->sk_family = PF_INET;
243 opt = xchg(&np->opt, NULL);
245 sock_kfree_s(sk, opt, opt->tot_len);
246 pktopt = xchg(&np->pktoptions, NULL);
250 sk->sk_destruct = inet_sock_destruct;
252 * ... and add it to the refcnt debug socks count
253 * in the new family. -acme
255 sk_refcnt_debug_inc(sk);
256 module_put(THIS_MODULE);
263 if (inet_sk(sk)->num)
265 np->ipv6only = valbool;
269 case IPV6_RECVPKTINFO:
270 np->rxopt.bits.rxinfo = valbool;
274 case IPV6_2292PKTINFO:
275 np->rxopt.bits.rxoinfo = valbool;
279 case IPV6_RECVHOPLIMIT:
280 np->rxopt.bits.rxhlim = valbool;
284 case IPV6_2292HOPLIMIT:
285 np->rxopt.bits.rxohlim = valbool;
290 if (val < 0 || val > 2)
292 np->rxopt.bits.srcrt = val;
297 if (val < 0 || val > 2)
299 np->rxopt.bits.osrcrt = val;
303 case IPV6_RECVHOPOPTS:
304 np->rxopt.bits.hopopts = valbool;
308 case IPV6_2292HOPOPTS:
309 np->rxopt.bits.ohopopts = valbool;
313 case IPV6_RECVDSTOPTS:
314 np->rxopt.bits.dstopts = valbool;
318 case IPV6_2292DSTOPTS:
319 np->rxopt.bits.odstopts = valbool;
324 if (val < 0 || val > 0xff)
330 case IPV6_RECVTCLASS:
331 np->rxopt.bits.rxtclass = valbool;
336 np->rxopt.bits.rxflow = valbool;
341 case IPV6_RTHDRDSTOPTS:
345 struct ipv6_txoptions *opt;
349 /* hop-by-hop / destination options are privileged option */
351 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
355 if (optlen & 0x7 || optlen > 8 * 255)
358 opt = ipv6_renew_options(sk, np->opt, optname,
359 (struct ipv6_opt_hdr __user *)optval,
366 /* routing header option needs extra check */
367 if (optname == IPV6_RTHDR && opt->srcrt) {
368 struct ipv6_rt_hdr *rthdr = opt->srcrt;
371 if ((rthdr->hdrlen & 1) ||
372 (rthdr->hdrlen >> 1) != rthdr->segments_left)
377 if (inet_sk(sk)->is_icsk) {
379 struct inet_connection_sock *icsk = inet_csk(sk);
380 if (!((1 << sk->sk_state) &
381 (TCPF_LISTEN | TCPF_CLOSE))
382 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
383 icsk->icsk_ext_hdr_len =
384 opt->opt_flen + opt->opt_nflen;
385 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
388 opt = xchg(&np->opt, opt);
391 write_lock(&sk->sk_dst_lock);
392 opt = xchg(&np->opt, opt);
393 write_unlock(&sk->sk_dst_lock);
398 sock_kfree_s(sk, opt, opt->tot_len);
402 case IPV6_2292PKTOPTIONS:
404 struct ipv6_txoptions *opt = NULL;
409 fl.fl6_flowlabel = 0;
410 fl.oif = sk->sk_bound_dev_if;
415 /* 1K is probably excessive
416 * 1K is surely not enough, 2K per standard header is 16K.
419 if (optlen > 64*1024)
422 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
427 memset(opt, 0, sizeof(*opt));
428 opt->tot_len = sizeof(*opt) + optlen;
430 if (copy_from_user(opt+1, optval, optlen))
433 msg.msg_controllen = optlen;
434 msg.msg_control = (void*)(opt+1);
436 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
441 if (inet_sk(sk)->is_icsk) {
443 struct inet_connection_sock *icsk = inet_csk(sk);
444 if (!((1 << sk->sk_state) &
445 (TCPF_LISTEN | TCPF_CLOSE))
446 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
447 icsk->icsk_ext_hdr_len =
448 opt->opt_flen + opt->opt_nflen;
449 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
452 opt = xchg(&np->opt, opt);
455 write_lock(&sk->sk_dst_lock);
456 opt = xchg(&np->opt, opt);
457 write_unlock(&sk->sk_dst_lock);
463 sock_kfree_s(sk, opt, opt->tot_len);
466 case IPV6_UNICAST_HOPS:
467 if (val > 255 || val < -1)
473 case IPV6_MULTICAST_HOPS:
474 if (sk->sk_type == SOCK_STREAM)
476 if (val > 255 || val < -1)
478 np->mcast_hops = val;
482 case IPV6_MULTICAST_LOOP:
483 np->mc_loop = valbool;
487 case IPV6_MULTICAST_IF:
488 if (sk->sk_type == SOCK_STREAM)
490 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
493 if (__dev_get_by_index(val) == NULL) {
500 case IPV6_ADD_MEMBERSHIP:
501 case IPV6_DROP_MEMBERSHIP:
503 struct ipv6_mreq mreq;
506 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
509 if (optname == IPV6_ADD_MEMBERSHIP)
510 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
512 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
515 case IPV6_JOIN_ANYCAST:
516 case IPV6_LEAVE_ANYCAST:
518 struct ipv6_mreq mreq;
520 if (optlen != sizeof(struct ipv6_mreq))
524 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
527 if (optname == IPV6_JOIN_ANYCAST)
528 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
530 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
533 case MCAST_JOIN_GROUP:
534 case MCAST_LEAVE_GROUP:
536 struct group_req greq;
537 struct sockaddr_in6 *psin6;
540 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
542 if (greq.gr_group.ss_family != AF_INET6) {
543 retv = -EADDRNOTAVAIL;
546 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
547 if (optname == MCAST_JOIN_GROUP)
548 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
551 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
555 case MCAST_JOIN_SOURCE_GROUP:
556 case MCAST_LEAVE_SOURCE_GROUP:
557 case MCAST_BLOCK_SOURCE:
558 case MCAST_UNBLOCK_SOURCE:
560 struct group_source_req greqs;
563 if (optlen != sizeof(struct group_source_req))
565 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
569 if (greqs.gsr_group.ss_family != AF_INET6 ||
570 greqs.gsr_source.ss_family != AF_INET6) {
571 retv = -EADDRNOTAVAIL;
574 if (optname == MCAST_BLOCK_SOURCE) {
575 omode = MCAST_EXCLUDE;
577 } else if (optname == MCAST_UNBLOCK_SOURCE) {
578 omode = MCAST_EXCLUDE;
580 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
581 struct sockaddr_in6 *psin6;
583 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
584 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
586 /* prior join w/ different source is ok */
587 if (retv && retv != -EADDRINUSE)
589 omode = MCAST_INCLUDE;
591 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
592 omode = MCAST_INCLUDE;
595 retv = ip6_mc_source(add, omode, sk, &greqs);
600 extern int sysctl_mld_max_msf;
601 struct group_filter *gsf;
603 if (optlen < GROUP_FILTER_SIZE(0))
605 if (optlen > sysctl_optmem_max) {
609 gsf = kmalloc(optlen,GFP_KERNEL);
615 if (copy_from_user(gsf, optval, optlen)) {
619 /* numsrc >= (4G-140)/128 overflow in 32 bits */
620 if (gsf->gf_numsrc >= 0x1ffffffU ||
621 gsf->gf_numsrc > sysctl_mld_max_msf) {
626 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
631 retv = ip6_mc_msfilter(sk, gsf);
636 case IPV6_ROUTER_ALERT:
637 retv = ip6_ra_control(sk, val, NULL);
639 case IPV6_MTU_DISCOVER:
646 if (val && val < IPV6_MIN_MTU)
652 np->recverr = valbool;
654 skb_queue_purge(&sk->sk_error_queue);
657 case IPV6_FLOWINFO_SEND:
658 np->sndflow = valbool;
661 case IPV6_FLOWLABEL_MGR:
662 retv = ipv6_flowlabel_opt(sk, optval, optlen);
664 case IPV6_IPSEC_POLICY:
665 case IPV6_XFRM_POLICY:
667 if (!capable(CAP_NET_ADMIN))
669 retv = xfrm_user_policy(sk, optname, optval, optlen);
682 int ipv6_setsockopt(struct sock *sk, int level, int optname,
683 char __user *optval, int optlen)
687 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
688 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
690 if (level != SOL_IPV6)
693 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
694 #ifdef CONFIG_NETFILTER
695 /* we need to exclude all possible ENOPROTOOPTs except default case */
696 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
697 optname != IPV6_XFRM_POLICY) {
699 err = nf_setsockopt(sk, PF_INET6, optname, optval,
709 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
710 char __user *optval, int optlen)
714 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
715 if (udp_prot.compat_setsockopt != NULL)
716 return udp_prot.compat_setsockopt(sk, level, optname,
718 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
721 if (level != SOL_IPV6)
724 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
725 #ifdef CONFIG_NETFILTER
726 /* we need to exclude all possible ENOPROTOOPTs except default case */
727 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
728 optname != IPV6_XFRM_POLICY) {
730 err = compat_nf_setsockopt(sk, PF_INET6, optname,
738 EXPORT_SYMBOL(compat_ipv6_setsockopt);
741 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
742 char __user *optval, int len)
746 len = min_t(int, len, ipv6_optlen(hdr));
747 if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
752 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
753 char __user *optval, int __user *optlen)
755 struct ipv6_pinfo *np = inet6_sk(sk);
759 if (get_user(len, optlen))
763 if (sk->sk_protocol != IPPROTO_UDP &&
764 sk->sk_protocol != IPPROTO_TCP)
766 if (sk->sk_state != TCP_ESTABLISHED)
772 struct group_filter gsf;
775 if (len < GROUP_FILTER_SIZE(0))
777 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
780 err = ip6_mc_msfget(sk, &gsf,
781 (struct group_filter __user *)optval, optlen);
786 case IPV6_2292PKTOPTIONS:
791 if (sk->sk_type != SOCK_STREAM)
794 msg.msg_control = optval;
795 msg.msg_controllen = len;
799 skb = np->pktoptions;
801 atomic_inc(&skb->users);
805 int err = datagram_recv_ctl(sk, &msg, skb);
810 if (np->rxopt.bits.rxinfo) {
811 struct in6_pktinfo src_info;
812 src_info.ipi6_ifindex = np->mcast_oif;
813 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
814 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
816 if (np->rxopt.bits.rxhlim) {
817 int hlim = np->mcast_hops;
818 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
820 if (np->rxopt.bits.rxoinfo) {
821 struct in6_pktinfo src_info;
822 src_info.ipi6_ifindex = np->mcast_oif;
823 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
824 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
826 if (np->rxopt.bits.rxohlim) {
827 int hlim = np->mcast_hops;
828 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
831 len -= msg.msg_controllen;
832 return put_user(len, optlen);
836 struct dst_entry *dst;
839 dst = sk_dst_get(sk);
854 case IPV6_RECVPKTINFO:
855 val = np->rxopt.bits.rxinfo;
858 case IPV6_2292PKTINFO:
859 val = np->rxopt.bits.rxoinfo;
862 case IPV6_RECVHOPLIMIT:
863 val = np->rxopt.bits.rxhlim;
866 case IPV6_2292HOPLIMIT:
867 val = np->rxopt.bits.rxohlim;
871 val = np->rxopt.bits.srcrt;
875 val = np->rxopt.bits.osrcrt;
879 case IPV6_RTHDRDSTOPTS:
885 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
888 return put_user(len, optlen);
891 case IPV6_RECVHOPOPTS:
892 val = np->rxopt.bits.hopopts;
895 case IPV6_2292HOPOPTS:
896 val = np->rxopt.bits.ohopopts;
899 case IPV6_RECVDSTOPTS:
900 val = np->rxopt.bits.dstopts;
903 case IPV6_2292DSTOPTS:
904 val = np->rxopt.bits.odstopts;
911 case IPV6_RECVTCLASS:
912 val = np->rxopt.bits.rxtclass;
916 val = np->rxopt.bits.rxflow;
919 case IPV6_UNICAST_HOPS:
923 case IPV6_MULTICAST_HOPS:
924 val = np->mcast_hops;
927 case IPV6_MULTICAST_LOOP:
931 case IPV6_MULTICAST_IF:
935 case IPV6_MTU_DISCOVER:
943 case IPV6_FLOWINFO_SEND:
950 len = min_t(unsigned int, sizeof(int), len);
951 if(put_user(len, optlen))
953 if(copy_to_user(optval,&val,len))
958 int ipv6_getsockopt(struct sock *sk, int level, int optname,
959 char __user *optval, int __user *optlen)
963 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
964 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
966 if(level != SOL_IPV6)
969 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
970 #ifdef CONFIG_NETFILTER
971 /* we need to exclude all possible EINVALs except default case */
972 if (err == -EINVAL && optname != IPV6_ADDRFORM &&
973 optname != MCAST_MSFILTER) {
976 if (get_user(len, optlen))
980 err = nf_getsockopt(sk, PF_INET6, optname, optval,
984 err = put_user(len, optlen);
991 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
992 char __user *optval, int __user *optlen)
996 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
997 if (udp_prot.compat_getsockopt != NULL)
998 return udp_prot.compat_getsockopt(sk, level, optname,
1000 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1003 if (level != SOL_IPV6)
1004 return -ENOPROTOOPT;
1006 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1007 #ifdef CONFIG_NETFILTER
1008 /* we need to exclude all possible EINVALs except default case */
1009 if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1010 optname != MCAST_MSFILTER) {
1013 if (get_user(len, optlen))
1017 err = compat_nf_getsockopt(sk, PF_INET6,
1018 optname, optval, &len);
1021 err = put_user(len, optlen);
1027 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1030 void __init ipv6_packet_init(void)
1032 dev_add_pack(&ipv6_packet_type);
1035 void ipv6_packet_cleanup(void)
1037 dev_remove_pack(&ipv6_packet_type);