1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 BSD socket options interface
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Based on linux/net/ipv4/ip_sockglue.c
11 * FIXME: Make the setsockopt code POSIX compliant: That is
13 * o Truncate getsockopt returns
14 * o Return an optlen of the truncated length if need be
17 * David L Stevens <dlstevens@us.ibm.com>:
18 * - added multicast source filtering API for MLDv2
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/mroute6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/init.h>
33 #include <linux/sysctl.h>
34 #include <linux/netfilter.h>
35 #include <linux/slab.h>
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_route.h>
44 #include <net/addrconf.h>
45 #include <net/inet_common.h>
48 #include <net/udplite.h>
50 #include <net/compat.h>
53 #include <linux/uaccess.h>
55 struct ip6_ra_chain *ip6_ra_chain;
56 DEFINE_RWLOCK(ip6_ra_lock);
58 DEFINE_STATIC_KEY_FALSE(ip6_min_hopcount);
60 int ip6_ra_control(struct sock *sk, int sel)
62 struct ip6_ra_chain *ra, *new_ra, **rap;
64 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
65 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
68 new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
69 if (sel >= 0 && !new_ra)
72 write_lock_bh(&ip6_ra_lock);
73 for (rap = &ip6_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
76 write_unlock_bh(&ip6_ra_lock);
82 write_unlock_bh(&ip6_ra_lock);
90 write_unlock_bh(&ip6_ra_lock);
98 write_unlock_bh(&ip6_ra_lock);
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 struct ipv6_txoptions *opt)
105 if (inet_test_bit(IS_ICSK, sk)) {
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,
121 static bool setsockopt_needs_rtnl(int optname)
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:
141 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
142 sockptr_t optval, int optlen)
144 if (in_compat_syscall()) {
145 struct compat_group_source_req gr32;
147 if (optlen < sizeof(gr32))
149 if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
151 greqs->gsr_interface = gr32.gsr_interface;
152 greqs->gsr_group = gr32.gsr_group;
153 greqs->gsr_source = gr32.gsr_source;
155 if (optlen < sizeof(*greqs))
157 if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
164 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
165 sockptr_t optval, int optlen)
167 struct group_source_req greqs;
171 ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
175 if (greqs.gsr_group.ss_family != AF_INET6 ||
176 greqs.gsr_source.ss_family != AF_INET6)
177 return -EADDRNOTAVAIL;
179 if (optname == MCAST_BLOCK_SOURCE) {
180 omode = MCAST_EXCLUDE;
182 } else if (optname == MCAST_UNBLOCK_SOURCE) {
183 omode = MCAST_EXCLUDE;
185 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
186 struct sockaddr_in6 *psin6;
189 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
190 retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
193 /* prior join w/ different source is ok */
194 if (retv && retv != -EADDRINUSE)
196 omode = MCAST_INCLUDE;
198 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
199 omode = MCAST_INCLUDE;
202 return ip6_mc_source(add, omode, sk, &greqs);
205 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
208 struct group_filter *gsf;
211 if (optlen < GROUP_FILTER_SIZE(0))
213 if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max))
216 gsf = memdup_sockptr(optval, optlen);
220 /* numsrc >= (4G-140)/128 overflow in 32 bits */
222 if (gsf->gf_numsrc >= 0x1ffffffU ||
223 gsf->gf_numsrc > sysctl_mld_max_msf)
227 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
230 ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
236 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
239 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
240 struct compat_group_filter *gf32;
247 if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max) - 4)
250 p = kmalloc(optlen + 4, GFP_KERNEL);
254 gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
256 if (copy_from_sockptr(gf32, optval, optlen))
259 /* numsrc >= (4G-140)/128 overflow in 32 bits */
262 if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
266 if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
269 ret = ip6_mc_msfilter(sk, &(struct group_filter){
270 .gf_interface = gf32->gf_interface,
271 .gf_group = gf32->gf_group,
272 .gf_fmode = gf32->gf_fmode,
273 .gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
280 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
281 sockptr_t optval, int optlen)
283 struct sockaddr_in6 *psin6;
284 struct group_req greq;
286 if (optlen < sizeof(greq))
288 if (copy_from_sockptr(&greq, optval, sizeof(greq)))
291 if (greq.gr_group.ss_family != AF_INET6)
292 return -EADDRNOTAVAIL;
293 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
294 if (optname == MCAST_JOIN_GROUP)
295 return ipv6_sock_mc_join(sk, greq.gr_interface,
297 return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
300 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
301 sockptr_t optval, int optlen)
303 struct compat_group_req gr32;
304 struct sockaddr_in6 *psin6;
306 if (optlen < sizeof(gr32))
308 if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
311 if (gr32.gr_group.ss_family != AF_INET6)
312 return -EADDRNOTAVAIL;
313 psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
314 if (optname == MCAST_JOIN_GROUP)
315 return ipv6_sock_mc_join(sk, gr32.gr_interface,
317 return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
320 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
323 struct ipv6_pinfo *np = inet6_sk(sk);
324 struct ipv6_opt_hdr *new = NULL;
325 struct net *net = sock_net(sk);
326 struct ipv6_txoptions *opt;
329 /* hop-by-hop / destination options are privileged option */
330 if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
333 /* remove any sticky options header with a zero option
334 * length, per RFC3542.
337 if (sockptr_is_null(optval))
339 if (optlen < sizeof(struct ipv6_opt_hdr) ||
344 new = memdup_sockptr(optval, optlen);
347 if (unlikely(ipv6_optlen(new) > optlen)) {
353 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
354 opt = ipv6_renew_options(sk, opt, optname, new);
359 /* routing header option needs extra check */
361 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
362 struct ipv6_rt_hdr *rthdr = opt->srcrt;
363 switch (rthdr->type) {
364 #if IS_ENABLED(CONFIG_IPV6_MIP6)
365 case IPV6_SRCRT_TYPE_2:
366 if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
370 case IPV6_SRCRT_TYPE_4:
372 struct ipv6_sr_hdr *srh =
373 (struct ipv6_sr_hdr *)opt->srcrt;
375 if (!seg6_validate_srh(srh, optlen, false))
385 opt = ipv6_update_options(sk, opt);
388 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
394 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
395 sockptr_t optval, unsigned int optlen)
397 struct ipv6_pinfo *np = inet6_sk(sk);
398 struct net *net = sock_net(sk);
400 int retv = -ENOPROTOOPT;
401 bool needs_rtnl = setsockopt_needs_rtnl(optname);
403 if (sockptr_is_null(optval))
406 if (optlen >= sizeof(int)) {
407 if (copy_from_sockptr(&val, optval, sizeof(val)))
413 valbool = (val != 0);
415 if (ip6_mroute_opt(optname))
416 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
418 /* Handle options that can be set without locking the socket. */
420 case IPV6_UNICAST_HOPS:
421 if (optlen < sizeof(int))
423 if (val > 255 || val < -1)
425 WRITE_ONCE(np->hop_limit, val);
427 case IPV6_MULTICAST_LOOP:
428 if (optlen < sizeof(int))
432 inet6_assign_bit(MC6_LOOP, sk, valbool);
434 case IPV6_MULTICAST_HOPS:
435 if (sk->sk_type == SOCK_STREAM)
437 if (optlen < sizeof(int))
439 if (val > 255 || val < -1)
441 WRITE_ONCE(np->mcast_hops,
442 val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
445 if (optlen < sizeof(int))
447 if (val && val < IPV6_MIN_MTU)
449 WRITE_ONCE(np->frag_size, val);
451 case IPV6_MINHOPCOUNT:
452 if (optlen < sizeof(int))
454 if (val < 0 || val > 255)
458 static_branch_enable(&ip6_min_hopcount);
460 /* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
461 * while we are changing it.
463 WRITE_ONCE(np->min_hopcount, val);
465 case IPV6_RECVERR_RFC4884:
466 if (optlen < sizeof(int))
468 if (val < 0 || val > 1)
470 inet6_assign_bit(RECVERR6_RFC4884, sk, valbool);
472 case IPV6_MULTICAST_ALL:
473 if (optlen < sizeof(int))
475 inet6_assign_bit(MC6_ALL, sk, valbool);
477 case IPV6_AUTOFLOWLABEL:
478 inet6_assign_bit(AUTOFLOWLABEL, sk, valbool);
479 inet6_set_bit(AUTOFLOWLABEL_SET, sk);
482 inet6_assign_bit(DONTFRAG, sk, valbool);
485 if (optlen < sizeof(int))
487 inet6_assign_bit(RECVERR6, sk, valbool);
489 skb_errqueue_purge(&sk->sk_error_queue);
491 case IPV6_ROUTER_ALERT_ISOLATE:
492 if (optlen < sizeof(int))
494 inet6_assign_bit(RTALERT_ISOLATE, sk, valbool);
496 case IPV6_MTU_DISCOVER:
497 if (optlen < sizeof(int))
499 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
501 WRITE_ONCE(np->pmtudisc, val);
503 case IPV6_FLOWINFO_SEND:
504 if (optlen < sizeof(int))
506 inet6_assign_bit(SNDFLOW, sk, valbool);
508 case IPV6_ADDR_PREFERENCES:
509 if (optlen < sizeof(int))
511 return ip6_sock_set_addr_preferences(sk, val);
512 case IPV6_MULTICAST_IF:
513 if (sk->sk_type == SOCK_STREAM)
515 if (optlen < sizeof(int))
518 struct net_device *dev;
519 int bound_dev_if, midx;
523 dev = dev_get_by_index_rcu(net, val);
528 midx = l3mdev_master_ifindex_rcu(dev);
532 bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
534 bound_dev_if != val &&
535 (!midx || midx != bound_dev_if))
538 WRITE_ONCE(np->mcast_oif, val);
540 case IPV6_UNICAST_IF:
542 struct net_device *dev;
545 if (optlen != sizeof(int))
548 ifindex = (__force int)ntohl((__force __be32)val);
550 WRITE_ONCE(np->ucast_oif, 0);
554 dev = dev_get_by_index(net, ifindex);
556 return -EADDRNOTAVAIL;
559 if (READ_ONCE(sk->sk_bound_dev_if))
562 WRITE_ONCE(np->ucast_oif, ifindex);
568 sockopt_lock_sock(sk);
570 /* Another thread has converted the socket into IPv4 with
571 * IPV6_ADDRFORM concurrently.
573 if (unlikely(sk->sk_family != AF_INET6))
579 if (optlen < sizeof(int))
581 if (val == PF_INET) {
582 if (sk->sk_type == SOCK_RAW)
585 if (sk->sk_protocol == IPPROTO_UDP ||
586 sk->sk_protocol == IPPROTO_UDPLITE) {
587 struct udp_sock *up = udp_sk(sk);
588 if (up->pending == AF_INET6) {
592 } else if (sk->sk_protocol == IPPROTO_TCP) {
593 if (sk->sk_prot != &tcpv6_prot) {
601 if (sk->sk_state != TCP_ESTABLISHED) {
606 if (ipv6_only_sock(sk) ||
607 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
608 retv = -EADDRNOTAVAIL;
612 __ipv6_sock_mc_close(sk);
613 __ipv6_sock_ac_close(sk);
615 if (sk->sk_protocol == IPPROTO_TCP) {
616 struct inet_connection_sock *icsk = inet_csk(sk);
618 sock_prot_inuse_add(net, sk->sk_prot, -1);
619 sock_prot_inuse_add(net, &tcp_prot, 1);
621 /* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
622 WRITE_ONCE(sk->sk_prot, &tcp_prot);
623 /* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
624 WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
625 WRITE_ONCE(sk->sk_socket->ops, &inet_stream_ops);
626 WRITE_ONCE(sk->sk_family, PF_INET);
627 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
629 struct proto *prot = &udp_prot;
631 if (sk->sk_protocol == IPPROTO_UDPLITE)
632 prot = &udplite_prot;
634 sock_prot_inuse_add(net, sk->sk_prot, -1);
635 sock_prot_inuse_add(net, prot, 1);
637 /* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
638 WRITE_ONCE(sk->sk_prot, prot);
639 WRITE_ONCE(sk->sk_socket->ops, &inet_dgram_ops);
640 WRITE_ONCE(sk->sk_family, PF_INET);
643 /* Disable all options not to allocate memory anymore,
644 * but there is still a race. See the lockless path
645 * in udpv6_sendmsg() and ipv6_local_rxpmtu().
649 inet6_cleanup_sock(sk);
651 module_put(THIS_MODULE);
658 if (optlen < sizeof(int) ||
659 inet_sk(sk)->inet_num)
661 sk->sk_ipv6only = valbool;
665 case IPV6_RECVPKTINFO:
666 if (optlen < sizeof(int))
668 np->rxopt.bits.rxinfo = valbool;
672 case IPV6_2292PKTINFO:
673 if (optlen < sizeof(int))
675 np->rxopt.bits.rxoinfo = valbool;
679 case IPV6_RECVHOPLIMIT:
680 if (optlen < sizeof(int))
682 np->rxopt.bits.rxhlim = valbool;
686 case IPV6_2292HOPLIMIT:
687 if (optlen < sizeof(int))
689 np->rxopt.bits.rxohlim = valbool;
694 if (optlen < sizeof(int))
696 np->rxopt.bits.srcrt = valbool;
701 if (optlen < sizeof(int))
703 np->rxopt.bits.osrcrt = valbool;
707 case IPV6_RECVHOPOPTS:
708 if (optlen < sizeof(int))
710 np->rxopt.bits.hopopts = valbool;
714 case IPV6_2292HOPOPTS:
715 if (optlen < sizeof(int))
717 np->rxopt.bits.ohopopts = valbool;
721 case IPV6_RECVDSTOPTS:
722 if (optlen < sizeof(int))
724 np->rxopt.bits.dstopts = valbool;
728 case IPV6_2292DSTOPTS:
729 if (optlen < sizeof(int))
731 np->rxopt.bits.odstopts = valbool;
736 if (optlen < sizeof(int))
738 if (val < -1 || val > 0xff)
740 /* RFC 3542, 6.5: default traffic class of 0x0 */
743 if (sk->sk_type == SOCK_STREAM) {
744 val &= ~INET_ECN_MASK;
745 val |= np->tclass & INET_ECN_MASK;
747 if (np->tclass != val) {
754 case IPV6_RECVTCLASS:
755 if (optlen < sizeof(int))
757 np->rxopt.bits.rxtclass = valbool;
762 if (optlen < sizeof(int))
764 np->rxopt.bits.rxflow = valbool;
768 case IPV6_RECVPATHMTU:
769 if (optlen < sizeof(int))
771 np->rxopt.bits.rxpmtu = valbool;
775 case IPV6_TRANSPARENT:
776 if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
777 !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
781 if (optlen < sizeof(int))
783 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
784 inet_assign_bit(TRANSPARENT, sk, valbool);
789 if (optlen < sizeof(int))
791 /* we also don't have a separate freebind bit for IPV6 */
792 inet_assign_bit(FREEBIND, sk, valbool);
796 case IPV6_RECVORIGDSTADDR:
797 if (optlen < sizeof(int))
799 np->rxopt.bits.rxorigdstaddr = valbool;
804 case IPV6_RTHDRDSTOPTS:
807 retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
812 struct in6_pktinfo pkt;
816 else if (optlen < sizeof(struct in6_pktinfo) ||
817 sockptr_is_null(optval))
820 if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
824 if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
827 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
828 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
833 case IPV6_2292PKTOPTIONS:
835 struct ipv6_txoptions *opt = NULL;
838 struct ipcm6_cookie ipc6;
840 memset(&fl6, 0, sizeof(fl6));
841 fl6.flowi6_oif = sk->sk_bound_dev_if;
842 fl6.flowi6_mark = sk->sk_mark;
847 /* 1K is probably excessive
848 * 1K is surely not enough, 2K per standard header is 16K.
851 if (optlen > 64*1024)
854 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
859 memset(opt, 0, sizeof(*opt));
860 refcount_set(&opt->refcnt, 1);
861 opt->tot_len = sizeof(*opt) + optlen;
863 if (copy_from_sockptr(opt + 1, optval, optlen))
866 msg.msg_controllen = optlen;
867 msg.msg_control_is_user = false;
868 msg.msg_control = (void *)(opt+1);
871 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
876 opt = ipv6_update_options(sk, opt);
879 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
885 case IPV6_ADD_MEMBERSHIP:
886 case IPV6_DROP_MEMBERSHIP:
888 struct ipv6_mreq mreq;
890 if (optlen < sizeof(struct ipv6_mreq))
894 if (inet_test_bit(IS_ICSK, sk))
898 if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
901 if (optname == IPV6_ADD_MEMBERSHIP)
902 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
904 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
907 case IPV6_JOIN_ANYCAST:
908 case IPV6_LEAVE_ANYCAST:
910 struct ipv6_mreq mreq;
912 if (optlen < sizeof(struct ipv6_mreq))
916 if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
919 if (optname == IPV6_JOIN_ANYCAST)
920 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
922 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
925 case MCAST_JOIN_GROUP:
926 case MCAST_LEAVE_GROUP:
927 if (in_compat_syscall())
928 retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
931 retv = ipv6_mcast_join_leave(sk, optname, optval,
934 case MCAST_JOIN_SOURCE_GROUP:
935 case MCAST_LEAVE_SOURCE_GROUP:
936 case MCAST_BLOCK_SOURCE:
937 case MCAST_UNBLOCK_SOURCE:
938 retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
941 if (in_compat_syscall())
942 retv = compat_ipv6_set_mcast_msfilter(sk, optval,
945 retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
947 case IPV6_ROUTER_ALERT:
948 if (optlen < sizeof(int))
950 retv = ip6_ra_control(sk, val);
952 case IPV6_FLOWLABEL_MGR:
953 retv = ipv6_flowlabel_opt(sk, optval, optlen);
955 case IPV6_IPSEC_POLICY:
956 case IPV6_XFRM_POLICY:
958 if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
960 retv = xfrm_user_policy(sk, optname, optval, optlen);
963 case IPV6_RECVFRAGSIZE:
964 np->rxopt.bits.recvfragsize = valbool;
970 sockopt_release_sock(sk);
981 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
986 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
987 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
989 if (level != SOL_IPV6)
992 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
993 #ifdef CONFIG_NETFILTER
994 /* we need to exclude all possible ENOPROTOOPTs except default case */
995 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
996 optname != IPV6_XFRM_POLICY)
997 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1001 EXPORT_SYMBOL(ipv6_setsockopt);
1003 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1004 int optname, sockptr_t optval, int len)
1006 struct ipv6_opt_hdr *hdr;
1015 case IPV6_RTHDRDSTOPTS:
1019 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1025 return -EINVAL; /* should not happen */
1031 len = min_t(unsigned int, len, ipv6_optlen(hdr));
1032 if (copy_to_sockptr(optval, hdr, len))
1037 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1038 sockptr_t optlen, int len)
1040 const int size0 = offsetof(struct group_filter, gf_slist_flex);
1041 struct group_filter gsf;
1047 if (copy_from_sockptr(&gsf, optval, size0))
1049 if (gsf.gf_group.ss_family != AF_INET6)
1050 return -EADDRNOTAVAIL;
1051 num = gsf.gf_numsrc;
1052 sockopt_lock_sock(sk);
1053 err = ip6_mc_msfget(sk, &gsf, optval, size0);
1055 if (num > gsf.gf_numsrc)
1056 num = gsf.gf_numsrc;
1057 len = GROUP_FILTER_SIZE(num);
1058 if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1059 copy_to_sockptr(optval, &gsf, size0))
1062 sockopt_release_sock(sk);
1066 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1067 sockptr_t optlen, int len)
1069 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1070 struct compat_group_filter gf32;
1071 struct group_filter gf;
1078 if (copy_from_sockptr(&gf32, optval, size0))
1080 gf.gf_interface = gf32.gf_interface;
1081 gf.gf_fmode = gf32.gf_fmode;
1082 num = gf.gf_numsrc = gf32.gf_numsrc;
1083 gf.gf_group = gf32.gf_group;
1085 if (gf.gf_group.ss_family != AF_INET6)
1086 return -EADDRNOTAVAIL;
1088 sockopt_lock_sock(sk);
1089 err = ip6_mc_msfget(sk, &gf, optval, size0);
1090 sockopt_release_sock(sk);
1093 if (num > gf.gf_numsrc)
1095 len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1096 if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1097 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1098 &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1099 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1100 &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1105 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1106 sockptr_t optval, sockptr_t optlen)
1108 struct ipv6_pinfo *np = inet6_sk(sk);
1112 if (ip6_mroute_opt(optname))
1113 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1115 if (copy_from_sockptr(&len, optlen, sizeof(int)))
1119 if (sk->sk_protocol != IPPROTO_UDP &&
1120 sk->sk_protocol != IPPROTO_UDPLITE &&
1121 sk->sk_protocol != IPPROTO_TCP)
1122 return -ENOPROTOOPT;
1123 if (sk->sk_state != TCP_ESTABLISHED)
1125 val = sk->sk_family;
1127 case MCAST_MSFILTER:
1128 if (in_compat_syscall())
1129 return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1130 return ipv6_get_msfilter(sk, optval, optlen, len);
1131 case IPV6_2292PKTOPTIONS:
1134 struct sk_buff *skb;
1136 if (sk->sk_type != SOCK_STREAM)
1137 return -ENOPROTOOPT;
1139 if (optval.is_kernel) {
1140 msg.msg_control_is_user = false;
1141 msg.msg_control = optval.kernel;
1143 msg.msg_control_is_user = true;
1144 msg.msg_control_user = optval.user;
1146 msg.msg_controllen = len;
1149 sockopt_lock_sock(sk);
1150 skb = np->pktoptions;
1152 ip6_datagram_recv_ctl(sk, &msg, skb);
1153 sockopt_release_sock(sk);
1155 if (np->rxopt.bits.rxinfo) {
1156 int mcast_oif = READ_ONCE(np->mcast_oif);
1157 struct in6_pktinfo src_info;
1159 src_info.ipi6_ifindex = mcast_oif ? :
1160 np->sticky_pktinfo.ipi6_ifindex;
1161 src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1162 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1164 if (np->rxopt.bits.rxhlim) {
1165 int hlim = READ_ONCE(np->mcast_hops);
1167 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1169 if (np->rxopt.bits.rxtclass) {
1170 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1172 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1174 if (np->rxopt.bits.rxoinfo) {
1175 int mcast_oif = READ_ONCE(np->mcast_oif);
1176 struct in6_pktinfo src_info;
1178 src_info.ipi6_ifindex = mcast_oif ? :
1179 np->sticky_pktinfo.ipi6_ifindex;
1180 src_info.ipi6_addr = mcast_oif ? sk->sk_v6_daddr :
1181 np->sticky_pktinfo.ipi6_addr;
1182 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1184 if (np->rxopt.bits.rxohlim) {
1185 int hlim = READ_ONCE(np->mcast_hops);
1187 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1189 if (np->rxopt.bits.rxflow) {
1190 __be32 flowinfo = np->rcv_flowinfo;
1192 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1195 len -= msg.msg_controllen;
1196 return copy_to_sockptr(optlen, &len, sizeof(int));
1200 struct dst_entry *dst;
1204 dst = __sk_dst_get(sk);
1214 val = sk->sk_ipv6only;
1217 case IPV6_RECVPKTINFO:
1218 val = np->rxopt.bits.rxinfo;
1221 case IPV6_2292PKTINFO:
1222 val = np->rxopt.bits.rxoinfo;
1225 case IPV6_RECVHOPLIMIT:
1226 val = np->rxopt.bits.rxhlim;
1229 case IPV6_2292HOPLIMIT:
1230 val = np->rxopt.bits.rxohlim;
1233 case IPV6_RECVRTHDR:
1234 val = np->rxopt.bits.srcrt;
1237 case IPV6_2292RTHDR:
1238 val = np->rxopt.bits.osrcrt;
1242 case IPV6_RTHDRDSTOPTS:
1246 struct ipv6_txoptions *opt;
1248 sockopt_lock_sock(sk);
1249 opt = rcu_dereference_protected(np->opt,
1250 lockdep_sock_is_held(sk));
1251 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1252 sockopt_release_sock(sk);
1253 /* check if ipv6_getsockopt_sticky() returns err code */
1256 return copy_to_sockptr(optlen, &len, sizeof(int));
1259 case IPV6_RECVHOPOPTS:
1260 val = np->rxopt.bits.hopopts;
1263 case IPV6_2292HOPOPTS:
1264 val = np->rxopt.bits.ohopopts;
1267 case IPV6_RECVDSTOPTS:
1268 val = np->rxopt.bits.dstopts;
1271 case IPV6_2292DSTOPTS:
1272 val = np->rxopt.bits.odstopts;
1279 case IPV6_RECVTCLASS:
1280 val = np->rxopt.bits.rxtclass;
1284 val = np->rxopt.bits.rxflow;
1287 case IPV6_RECVPATHMTU:
1288 val = np->rxopt.bits.rxpmtu;
1293 struct dst_entry *dst;
1294 struct ip6_mtuinfo mtuinfo;
1296 if (len < sizeof(mtuinfo))
1299 len = sizeof(mtuinfo);
1300 memset(&mtuinfo, 0, sizeof(mtuinfo));
1303 dst = __sk_dst_get(sk);
1305 mtuinfo.ip6m_mtu = dst_mtu(dst);
1307 if (!mtuinfo.ip6m_mtu)
1310 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1312 if (copy_to_sockptr(optval, &mtuinfo, len))
1318 case IPV6_TRANSPARENT:
1319 val = inet_test_bit(TRANSPARENT, sk);
1323 val = inet_test_bit(FREEBIND, sk);
1326 case IPV6_RECVORIGDSTADDR:
1327 val = np->rxopt.bits.rxorigdstaddr;
1330 case IPV6_UNICAST_HOPS:
1331 case IPV6_MULTICAST_HOPS:
1333 struct dst_entry *dst;
1335 if (optname == IPV6_UNICAST_HOPS)
1336 val = READ_ONCE(np->hop_limit);
1338 val = READ_ONCE(np->mcast_hops);
1342 dst = __sk_dst_get(sk);
1344 val = ip6_dst_hoplimit(dst);
1349 val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1353 case IPV6_MULTICAST_LOOP:
1354 val = inet6_test_bit(MC6_LOOP, sk);
1357 case IPV6_MULTICAST_IF:
1358 val = READ_ONCE(np->mcast_oif);
1361 case IPV6_MULTICAST_ALL:
1362 val = inet6_test_bit(MC6_ALL, sk);
1365 case IPV6_UNICAST_IF:
1366 val = (__force int)htonl((__u32) READ_ONCE(np->ucast_oif));
1369 case IPV6_MTU_DISCOVER:
1370 val = READ_ONCE(np->pmtudisc);
1374 val = inet6_test_bit(RECVERR6, sk);
1377 case IPV6_FLOWINFO_SEND:
1378 val = inet6_test_bit(SNDFLOW, sk);
1381 case IPV6_FLOWLABEL_MGR:
1383 struct in6_flowlabel_req freq;
1386 if (len < sizeof(freq))
1389 if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1392 if (freq.flr_action != IPV6_FL_A_GET)
1396 flags = freq.flr_flags;
1398 memset(&freq, 0, sizeof(freq));
1400 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1404 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1406 if (copy_to_sockptr(optval, &freq, len))
1412 case IPV6_ADDR_PREFERENCES:
1414 u8 srcprefs = READ_ONCE(np->srcprefs);
1417 if (srcprefs & IPV6_PREFER_SRC_TMP)
1418 val |= IPV6_PREFER_SRC_TMP;
1419 else if (srcprefs & IPV6_PREFER_SRC_PUBLIC)
1420 val |= IPV6_PREFER_SRC_PUBLIC;
1422 /* XXX: should we return system default? */
1423 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1426 if (srcprefs & IPV6_PREFER_SRC_COA)
1427 val |= IPV6_PREFER_SRC_COA;
1429 val |= IPV6_PREFER_SRC_HOME;
1432 case IPV6_MINHOPCOUNT:
1433 val = READ_ONCE(np->min_hopcount);
1437 val = inet6_test_bit(DONTFRAG, sk);
1440 case IPV6_AUTOFLOWLABEL:
1441 val = ip6_autoflowlabel(sock_net(sk), sk);
1444 case IPV6_RECVFRAGSIZE:
1445 val = np->rxopt.bits.recvfragsize;
1448 case IPV6_ROUTER_ALERT_ISOLATE:
1449 val = inet6_test_bit(RTALERT_ISOLATE, sk);
1452 case IPV6_RECVERR_RFC4884:
1453 val = inet6_test_bit(RECVERR6_RFC4884, sk);
1457 return -ENOPROTOOPT;
1459 len = min_t(unsigned int, sizeof(int), len);
1460 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1462 if (copy_to_sockptr(optval, &val, len))
1467 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1468 char __user *optval, int __user *optlen)
1472 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1473 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1475 if (level != SOL_IPV6)
1476 return -ENOPROTOOPT;
1478 err = do_ipv6_getsockopt(sk, level, optname,
1479 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1480 #ifdef CONFIG_NETFILTER
1481 /* we need to exclude all possible ENOPROTOOPTs except default case */
1482 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1485 if (get_user(len, optlen))
1488 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1490 err = put_user(len, optlen);
1495 EXPORT_SYMBOL(ipv6_getsockopt);