2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
11 * Many : Split from ip.c , see ip.c for history.
12 * Martin Mares : TOS setting fixed.
13 * Alan Cox : Fixed a couple of oopses in Martin's
15 * Mike McLagan : Routing by source
18 #include <linux/module.h>
19 #include <linux/types.h>
21 #include <linux/skbuff.h>
23 #include <linux/icmp.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/slab.h>
30 #include <net/tcp_states.h>
31 #include <linux/udp.h>
32 #include <linux/igmp.h>
33 #include <linux/netfilter.h>
34 #include <linux/route.h>
35 #include <linux/mroute.h>
36 #include <net/inet_ecn.h>
37 #include <net/route.h>
39 #include <net/compat.h>
40 #include <net/checksum.h>
41 #if IS_ENABLED(CONFIG_IPV6)
42 #include <net/transp_v6.h>
44 #include <net/ip_fib.h>
46 #include <linux/errqueue.h>
47 #include <asm/uaccess.h>
50 * SOL_IP control messages.
53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
55 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
57 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
59 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
64 int ttl = ip_hdr(skb)->ttl;
65 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
70 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
75 if (IPCB(skb)->opt.optlen == 0)
78 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
85 unsigned char optbuf[sizeof(struct ip_options) + 40];
86 struct ip_options *opt = (struct ip_options *)optbuf;
88 if (IPCB(skb)->opt.optlen == 0)
91 if (ip_options_echo(opt, skb)) {
92 msg->msg_flags |= MSG_CTRUNC;
97 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
100 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
103 __wsum csum = skb->csum;
105 if (skb->ip_summed != CHECKSUM_COMPLETE)
109 csum = csum_sub(csum, csum_partial(skb_transport_header(skb),
112 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
115 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
121 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
125 err = security_secid_to_secctx(secid, &secdata, &seclen);
129 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
130 security_release_secctx(secdata, seclen);
133 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
135 struct sockaddr_in sin;
136 const struct iphdr *iph = ip_hdr(skb);
137 __be16 *ports = (__be16 *)skb_transport_header(skb);
139 if (skb_transport_offset(skb) + 4 > skb->len)
142 /* All current transport protocols have the port numbers in the
143 * first four bytes of the transport header and this function is
144 * written with this assumption in mind.
147 sin.sin_family = AF_INET;
148 sin.sin_addr.s_addr = iph->daddr;
149 sin.sin_port = ports[1];
150 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
152 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
155 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
158 struct inet_sock *inet = inet_sk(skb->sk);
159 unsigned int flags = inet->cmsg_flags;
161 /* Ordered by supposed usage frequency */
162 if (flags & IP_CMSG_PKTINFO) {
163 ip_cmsg_recv_pktinfo(msg, skb);
165 flags &= ~IP_CMSG_PKTINFO;
170 if (flags & IP_CMSG_TTL) {
171 ip_cmsg_recv_ttl(msg, skb);
173 flags &= ~IP_CMSG_TTL;
178 if (flags & IP_CMSG_TOS) {
179 ip_cmsg_recv_tos(msg, skb);
181 flags &= ~IP_CMSG_TOS;
186 if (flags & IP_CMSG_RECVOPTS) {
187 ip_cmsg_recv_opts(msg, skb);
189 flags &= ~IP_CMSG_RECVOPTS;
194 if (flags & IP_CMSG_RETOPTS) {
195 ip_cmsg_recv_retopts(msg, skb);
197 flags &= ~IP_CMSG_RETOPTS;
202 if (flags & IP_CMSG_PASSSEC) {
203 ip_cmsg_recv_security(msg, skb);
205 flags &= ~IP_CMSG_PASSSEC;
210 if (flags & IP_CMSG_ORIGDSTADDR) {
211 ip_cmsg_recv_dstaddr(msg, skb);
213 flags &= ~IP_CMSG_ORIGDSTADDR;
218 if (flags & IP_CMSG_CHECKSUM)
219 ip_cmsg_recv_checksum(msg, skb, offset);
221 EXPORT_SYMBOL(ip_cmsg_recv_offset);
223 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
227 struct cmsghdr *cmsg;
228 struct net *net = sock_net(sk);
230 for_each_cmsghdr(cmsg, msg) {
231 if (!CMSG_OK(msg, cmsg))
233 #if IS_ENABLED(CONFIG_IPV6)
235 cmsg->cmsg_level == SOL_IPV6 &&
236 cmsg->cmsg_type == IPV6_PKTINFO) {
237 struct in6_pktinfo *src_info;
239 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
241 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
242 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
244 ipc->oif = src_info->ipi6_ifindex;
245 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
249 if (cmsg->cmsg_level == SOL_SOCKET) {
250 if (__sock_cmsg_send(sk, msg, cmsg, &ipc->sockc))
255 if (cmsg->cmsg_level != SOL_IP)
257 switch (cmsg->cmsg_type) {
259 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
261 /* Our caller is responsible for freeing ipc->opt */
262 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
263 err < 40 ? err : 40);
269 struct in_pktinfo *info;
270 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
272 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
273 ipc->oif = info->ipi_ifindex;
274 ipc->addr = info->ipi_spec_dst.s_addr;
278 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
280 val = *(int *)CMSG_DATA(cmsg);
281 if (val < 1 || val > 255)
286 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
288 val = *(int *)CMSG_DATA(cmsg);
289 if (val < 0 || val > 255)
292 ipc->priority = rt_tos2priority(ipc->tos);
303 /* Special input handler for packets caught by router alert option.
304 They are selected only by protocol field, and then processed likely
305 local ones; but only if someone wants them! Otherwise, router
306 not running rsvpd will kill RSVP.
308 It is user level problem, what it will make with them.
309 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
310 but receiver should be enough clever f.e. to forward mtrace requests,
311 sent to multicast group to reach destination designated router.
313 struct ip_ra_chain __rcu *ip_ra_chain;
314 static DEFINE_SPINLOCK(ip_ra_lock);
317 static void ip_ra_destroy_rcu(struct rcu_head *head)
319 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
321 sock_put(ra->saved_sk);
325 int ip_ra_control(struct sock *sk, unsigned char on,
326 void (*destructor)(struct sock *))
328 struct ip_ra_chain *ra, *new_ra;
329 struct ip_ra_chain __rcu **rap;
331 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
334 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
336 spin_lock_bh(&ip_ra_lock);
337 for (rap = &ip_ra_chain;
338 (ra = rcu_dereference_protected(*rap,
339 lockdep_is_held(&ip_ra_lock))) != NULL;
343 spin_unlock_bh(&ip_ra_lock);
347 /* dont let ip_call_ra_chain() use sk again */
349 RCU_INIT_POINTER(*rap, ra->next);
350 spin_unlock_bh(&ip_ra_lock);
355 * Delay sock_put(sk) and kfree(ra) after one rcu grace
356 * period. This guarantee ip_call_ra_chain() dont need
357 * to mess with socket refcounts.
360 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
365 spin_unlock_bh(&ip_ra_lock);
369 new_ra->destructor = destructor;
371 RCU_INIT_POINTER(new_ra->next, ra);
372 rcu_assign_pointer(*rap, new_ra);
374 spin_unlock_bh(&ip_ra_lock);
379 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
380 __be16 port, u32 info, u8 *payload)
382 struct sock_exterr_skb *serr;
384 skb = skb_clone(skb, GFP_ATOMIC);
388 serr = SKB_EXT_ERR(skb);
389 serr->ee.ee_errno = err;
390 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
391 serr->ee.ee_type = icmp_hdr(skb)->type;
392 serr->ee.ee_code = icmp_hdr(skb)->code;
394 serr->ee.ee_info = info;
395 serr->ee.ee_data = 0;
396 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
397 skb_network_header(skb);
400 if (skb_pull(skb, payload - skb->data)) {
401 skb_reset_transport_header(skb);
402 if (sock_queue_err_skb(sk, skb) == 0)
408 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
410 struct inet_sock *inet = inet_sk(sk);
411 struct sock_exterr_skb *serr;
418 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
422 skb_put(skb, sizeof(struct iphdr));
423 skb_reset_network_header(skb);
427 serr = SKB_EXT_ERR(skb);
428 serr->ee.ee_errno = err;
429 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
430 serr->ee.ee_type = 0;
431 serr->ee.ee_code = 0;
433 serr->ee.ee_info = info;
434 serr->ee.ee_data = 0;
435 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
438 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
439 skb_reset_transport_header(skb);
441 if (sock_queue_err_skb(sk, skb))
445 /* For some errors we have valid addr_offset even with zero payload and
446 * zero port. Also, addr_offset should be supported if port is set.
448 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
450 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
451 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
454 /* IPv4 supports cmsg on all imcp errors and some timestamps
456 * Timestamp code paths do not initialize the fields expected by cmsg:
457 * the PKTINFO fields in skb->cb[]. Fill those in here.
459 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
463 struct in_pktinfo *info;
465 if (ee_origin == SO_EE_ORIGIN_ICMP)
468 if (ee_origin == SO_EE_ORIGIN_LOCAL)
471 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
472 * timestamp with egress dev. Not possible for packets without dev
473 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
475 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
479 info = PKTINFO_SKB_CB(skb);
480 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
481 info->ipi_ifindex = skb->dev->ifindex;
486 * Handle MSG_ERRQUEUE
488 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
490 struct sock_exterr_skb *serr;
492 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
494 struct sock_extended_err ee;
495 struct sockaddr_in offender;
500 WARN_ON_ONCE(sk->sk_family == AF_INET6);
503 skb = sock_dequeue_err_skb(sk);
509 msg->msg_flags |= MSG_TRUNC;
512 err = skb_copy_datagram_msg(skb, 0, msg, copied);
516 sock_recv_timestamp(msg, sk, skb);
518 serr = SKB_EXT_ERR(skb);
520 if (sin && ipv4_datagram_support_addr(serr)) {
521 sin->sin_family = AF_INET;
522 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
524 sin->sin_port = serr->port;
525 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
526 *addr_len = sizeof(*sin);
529 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
530 sin = &errhdr.offender;
531 memset(sin, 0, sizeof(*sin));
533 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
534 sin->sin_family = AF_INET;
535 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
536 if (inet_sk(sk)->cmsg_flags)
537 ip_cmsg_recv(msg, skb);
540 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
542 /* Now we could try to dump offended packet options */
544 msg->msg_flags |= MSG_ERRQUEUE;
555 * Socket option code for IP. This is the end of the line after any
556 * TCP,UDP etc options on an IP socket.
558 static bool setsockopt_needs_rtnl(int optname)
561 case IP_ADD_MEMBERSHIP:
562 case IP_ADD_SOURCE_MEMBERSHIP:
563 case IP_BLOCK_SOURCE:
564 case IP_DROP_MEMBERSHIP:
565 case IP_DROP_SOURCE_MEMBERSHIP:
567 case IP_UNBLOCK_SOURCE:
568 case MCAST_BLOCK_SOURCE:
570 case MCAST_JOIN_GROUP:
571 case MCAST_JOIN_SOURCE_GROUP:
572 case MCAST_LEAVE_GROUP:
573 case MCAST_LEAVE_SOURCE_GROUP:
574 case MCAST_UNBLOCK_SOURCE:
580 static int do_ip_setsockopt(struct sock *sk, int level,
581 int optname, char __user *optval, unsigned int optlen)
583 struct inet_sock *inet = inet_sk(sk);
584 struct net *net = sock_net(sk);
586 bool needs_rtnl = setsockopt_needs_rtnl(optname);
597 case IP_MTU_DISCOVER:
599 case IP_ROUTER_ALERT:
605 case IP_BIND_ADDRESS_NO_PORT:
607 case IP_MULTICAST_TTL:
608 case IP_MULTICAST_ALL:
609 case IP_MULTICAST_LOOP:
610 case IP_RECVORIGDSTADDR:
612 if (optlen >= sizeof(int)) {
613 if (get_user(val, (int __user *) optval))
615 } else if (optlen >= sizeof(char)) {
618 if (get_user(ucval, (unsigned char __user *) optval))
624 /* If optlen==0, it is equivalent to val == 0 */
626 if (ip_mroute_opt(optname))
627 return ip_mroute_setsockopt(sk, optname, optval, optlen);
637 struct ip_options_rcu *old, *opt = NULL;
641 err = ip_options_get_from_user(sock_net(sk), &opt,
645 old = rcu_dereference_protected(inet->inet_opt,
646 lockdep_sock_is_held(sk));
648 struct inet_connection_sock *icsk = inet_csk(sk);
649 #if IS_ENABLED(CONFIG_IPV6)
650 if (sk->sk_family == PF_INET ||
651 (!((1 << sk->sk_state) &
652 (TCPF_LISTEN | TCPF_CLOSE)) &&
653 inet->inet_daddr != LOOPBACK4_IPV6)) {
656 icsk->icsk_ext_hdr_len -= old->opt.optlen;
658 icsk->icsk_ext_hdr_len += opt->opt.optlen;
659 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
660 #if IS_ENABLED(CONFIG_IPV6)
664 rcu_assign_pointer(inet->inet_opt, opt);
671 inet->cmsg_flags |= IP_CMSG_PKTINFO;
673 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
677 inet->cmsg_flags |= IP_CMSG_TTL;
679 inet->cmsg_flags &= ~IP_CMSG_TTL;
683 inet->cmsg_flags |= IP_CMSG_TOS;
685 inet->cmsg_flags &= ~IP_CMSG_TOS;
689 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
691 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
695 inet->cmsg_flags |= IP_CMSG_RETOPTS;
697 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
701 inet->cmsg_flags |= IP_CMSG_PASSSEC;
703 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
705 case IP_RECVORIGDSTADDR:
707 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
709 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
713 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
714 inet_inc_convert_csum(sk);
715 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
718 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
719 inet_dec_convert_csum(sk);
720 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
724 case IP_TOS: /* This sets both TOS and Precedence */
725 if (sk->sk_type == SOCK_STREAM) {
726 val &= ~INET_ECN_MASK;
727 val |= inet->tos & INET_ECN_MASK;
729 if (inet->tos != val) {
731 sk->sk_priority = rt_tos2priority(val);
738 if (val != -1 && (val < 1 || val > 255))
743 if (sk->sk_type != SOCK_RAW) {
747 inet->hdrincl = val ? 1 : 0;
750 if (sk->sk_type != SOCK_RAW) {
754 inet->nodefrag = val ? 1 : 0;
756 case IP_BIND_ADDRESS_NO_PORT:
757 inet->bind_address_no_port = val ? 1 : 0;
759 case IP_MTU_DISCOVER:
760 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
762 inet->pmtudisc = val;
765 inet->recverr = !!val;
767 skb_queue_purge(&sk->sk_error_queue);
769 case IP_MULTICAST_TTL:
770 if (sk->sk_type == SOCK_STREAM)
776 if (val < 0 || val > 255)
780 case IP_MULTICAST_LOOP:
783 inet->mc_loop = !!val;
787 struct net_device *dev = NULL;
790 if (optlen != sizeof(int))
793 ifindex = (__force int)ntohl((__force __be32)val);
800 dev = dev_get_by_index(sock_net(sk), ifindex);
801 err = -EADDRNOTAVAIL;
807 if (sk->sk_bound_dev_if)
810 inet->uc_index = ifindex;
814 case IP_MULTICAST_IF:
816 struct ip_mreqn mreq;
817 struct net_device *dev = NULL;
819 if (sk->sk_type == SOCK_STREAM)
822 * Check the arguments are allowable
825 if (optlen < sizeof(struct in_addr))
829 if (optlen >= sizeof(struct ip_mreqn)) {
830 if (copy_from_user(&mreq, optval, sizeof(mreq)))
833 memset(&mreq, 0, sizeof(mreq));
834 if (optlen >= sizeof(struct ip_mreq)) {
835 if (copy_from_user(&mreq, optval,
836 sizeof(struct ip_mreq)))
838 } else if (optlen >= sizeof(struct in_addr)) {
839 if (copy_from_user(&mreq.imr_address, optval,
840 sizeof(struct in_addr)))
845 if (!mreq.imr_ifindex) {
846 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
852 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
854 mreq.imr_ifindex = dev->ifindex;
856 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
859 err = -EADDRNOTAVAIL;
865 if (sk->sk_bound_dev_if &&
866 mreq.imr_ifindex != sk->sk_bound_dev_if)
869 inet->mc_index = mreq.imr_ifindex;
870 inet->mc_addr = mreq.imr_address.s_addr;
875 case IP_ADD_MEMBERSHIP:
876 case IP_DROP_MEMBERSHIP:
878 struct ip_mreqn mreq;
881 if (inet_sk(sk)->is_icsk)
884 if (optlen < sizeof(struct ip_mreq))
887 if (optlen >= sizeof(struct ip_mreqn)) {
888 if (copy_from_user(&mreq, optval, sizeof(mreq)))
891 memset(&mreq, 0, sizeof(mreq));
892 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
896 if (optname == IP_ADD_MEMBERSHIP)
897 err = ip_mc_join_group(sk, &mreq);
899 err = ip_mc_leave_group(sk, &mreq);
904 struct ip_msfilter *msf;
906 if (optlen < IP_MSFILTER_SIZE(0))
908 if (optlen > sysctl_optmem_max) {
912 msf = kmalloc(optlen, GFP_KERNEL);
918 if (copy_from_user(msf, optval, optlen)) {
922 /* numsrc >= (1G-4) overflow in 32 bits */
923 if (msf->imsf_numsrc >= 0x3ffffffcU ||
924 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
929 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
934 err = ip_mc_msfilter(sk, msf, 0);
938 case IP_BLOCK_SOURCE:
939 case IP_UNBLOCK_SOURCE:
940 case IP_ADD_SOURCE_MEMBERSHIP:
941 case IP_DROP_SOURCE_MEMBERSHIP:
943 struct ip_mreq_source mreqs;
946 if (optlen != sizeof(struct ip_mreq_source))
948 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
952 if (optname == IP_BLOCK_SOURCE) {
953 omode = MCAST_EXCLUDE;
955 } else if (optname == IP_UNBLOCK_SOURCE) {
956 omode = MCAST_EXCLUDE;
958 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
959 struct ip_mreqn mreq;
961 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
962 mreq.imr_address.s_addr = mreqs.imr_interface;
963 mreq.imr_ifindex = 0;
964 err = ip_mc_join_group(sk, &mreq);
965 if (err && err != -EADDRINUSE)
967 omode = MCAST_INCLUDE;
969 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
970 omode = MCAST_INCLUDE;
973 err = ip_mc_source(add, omode, sk, &mreqs, 0);
976 case MCAST_JOIN_GROUP:
977 case MCAST_LEAVE_GROUP:
979 struct group_req greq;
980 struct sockaddr_in *psin;
981 struct ip_mreqn mreq;
983 if (optlen < sizeof(struct group_req))
986 if (copy_from_user(&greq, optval, sizeof(greq)))
988 psin = (struct sockaddr_in *)&greq.gr_group;
989 if (psin->sin_family != AF_INET)
991 memset(&mreq, 0, sizeof(mreq));
992 mreq.imr_multiaddr = psin->sin_addr;
993 mreq.imr_ifindex = greq.gr_interface;
995 if (optname == MCAST_JOIN_GROUP)
996 err = ip_mc_join_group(sk, &mreq);
998 err = ip_mc_leave_group(sk, &mreq);
1001 case MCAST_JOIN_SOURCE_GROUP:
1002 case MCAST_LEAVE_SOURCE_GROUP:
1003 case MCAST_BLOCK_SOURCE:
1004 case MCAST_UNBLOCK_SOURCE:
1006 struct group_source_req greqs;
1007 struct ip_mreq_source mreqs;
1008 struct sockaddr_in *psin;
1011 if (optlen != sizeof(struct group_source_req))
1013 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1017 if (greqs.gsr_group.ss_family != AF_INET ||
1018 greqs.gsr_source.ss_family != AF_INET) {
1019 err = -EADDRNOTAVAIL;
1022 psin = (struct sockaddr_in *)&greqs.gsr_group;
1023 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1024 psin = (struct sockaddr_in *)&greqs.gsr_source;
1025 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1026 mreqs.imr_interface = 0; /* use index for mc_source */
1028 if (optname == MCAST_BLOCK_SOURCE) {
1029 omode = MCAST_EXCLUDE;
1031 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1032 omode = MCAST_EXCLUDE;
1034 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1035 struct ip_mreqn mreq;
1037 psin = (struct sockaddr_in *)&greqs.gsr_group;
1038 mreq.imr_multiaddr = psin->sin_addr;
1039 mreq.imr_address.s_addr = 0;
1040 mreq.imr_ifindex = greqs.gsr_interface;
1041 err = ip_mc_join_group(sk, &mreq);
1042 if (err && err != -EADDRINUSE)
1044 greqs.gsr_interface = mreq.imr_ifindex;
1045 omode = MCAST_INCLUDE;
1047 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1048 omode = MCAST_INCLUDE;
1051 err = ip_mc_source(add, omode, sk, &mreqs,
1052 greqs.gsr_interface);
1055 case MCAST_MSFILTER:
1057 struct sockaddr_in *psin;
1058 struct ip_msfilter *msf = NULL;
1059 struct group_filter *gsf = NULL;
1060 int msize, i, ifindex;
1062 if (optlen < GROUP_FILTER_SIZE(0))
1064 if (optlen > sysctl_optmem_max) {
1068 gsf = kmalloc(optlen, GFP_KERNEL);
1074 if (copy_from_user(gsf, optval, optlen))
1077 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1078 if (gsf->gf_numsrc >= 0x1ffffff ||
1079 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1083 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1087 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1088 msf = kmalloc(msize, GFP_KERNEL);
1093 ifindex = gsf->gf_interface;
1094 psin = (struct sockaddr_in *)&gsf->gf_group;
1095 if (psin->sin_family != AF_INET) {
1096 err = -EADDRNOTAVAIL;
1099 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1100 msf->imsf_interface = 0;
1101 msf->imsf_fmode = gsf->gf_fmode;
1102 msf->imsf_numsrc = gsf->gf_numsrc;
1103 err = -EADDRNOTAVAIL;
1104 for (i = 0; i < gsf->gf_numsrc; ++i) {
1105 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1107 if (psin->sin_family != AF_INET)
1109 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1114 err = ip_mc_msfilter(sk, msf, ifindex);
1120 case IP_MULTICAST_ALL:
1123 if (val != 0 && val != 1)
1127 case IP_ROUTER_ALERT:
1128 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1134 inet->freebind = !!val;
1137 case IP_IPSEC_POLICY:
1138 case IP_XFRM_POLICY:
1140 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1142 err = xfrm_user_policy(sk, optname, optval, optlen);
1145 case IP_TRANSPARENT:
1146 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1147 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1153 inet->transparent = !!val;
1159 if (val < 0 || val > 255)
1161 inet->min_ttl = val;
1181 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1185 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1186 * destination in skb->cb[] before dst drop.
1187 * This way, receiver doesn't make cache line misses to read rtable.
1189 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1191 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1192 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1195 if (prepare && skb_rtable(skb)) {
1196 pktinfo->ipi_ifindex = inet_iif(skb);
1197 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1199 pktinfo->ipi_ifindex = 0;
1200 pktinfo->ipi_spec_dst.s_addr = 0;
1205 int ip_setsockopt(struct sock *sk, int level,
1206 int optname, char __user *optval, unsigned int optlen)
1210 if (level != SOL_IP)
1211 return -ENOPROTOOPT;
1213 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1214 #ifdef CONFIG_NETFILTER
1215 /* we need to exclude all possible ENOPROTOOPTs except default case */
1216 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1217 optname != IP_IPSEC_POLICY &&
1218 optname != IP_XFRM_POLICY &&
1219 !ip_mroute_opt(optname)) {
1221 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1227 EXPORT_SYMBOL(ip_setsockopt);
1229 #ifdef CONFIG_COMPAT
1230 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1231 char __user *optval, unsigned int optlen)
1235 if (level != SOL_IP)
1236 return -ENOPROTOOPT;
1238 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1239 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1242 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1243 #ifdef CONFIG_NETFILTER
1244 /* we need to exclude all possible ENOPROTOOPTs except default case */
1245 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1246 optname != IP_IPSEC_POLICY &&
1247 optname != IP_XFRM_POLICY &&
1248 !ip_mroute_opt(optname)) {
1250 err = compat_nf_setsockopt(sk, PF_INET, optname,
1257 EXPORT_SYMBOL(compat_ip_setsockopt);
1261 * Get the options. Note for future reference. The GET of IP options gets
1262 * the _received_ ones. The set sets the _sent_ ones.
1265 static bool getsockopt_needs_rtnl(int optname)
1269 case MCAST_MSFILTER:
1275 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1276 char __user *optval, int __user *optlen, unsigned int flags)
1278 struct inet_sock *inet = inet_sk(sk);
1279 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1283 if (level != SOL_IP)
1286 if (ip_mroute_opt(optname))
1287 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1289 if (get_user(len, optlen))
1301 unsigned char optbuf[sizeof(struct ip_options)+40];
1302 struct ip_options *opt = (struct ip_options *)optbuf;
1303 struct ip_options_rcu *inet_opt;
1305 inet_opt = rcu_dereference_protected(inet->inet_opt,
1306 lockdep_sock_is_held(sk));
1309 memcpy(optbuf, &inet_opt->opt,
1310 sizeof(struct ip_options) +
1311 inet_opt->opt.optlen);
1314 if (opt->optlen == 0)
1315 return put_user(0, optlen);
1317 ip_options_undo(opt);
1319 len = min_t(unsigned int, len, opt->optlen);
1320 if (put_user(len, optlen))
1322 if (copy_to_user(optval, opt->__data, len))
1327 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1330 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1333 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1336 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1339 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1342 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1344 case IP_RECVORIGDSTADDR:
1345 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1348 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1355 struct net *net = sock_net(sk);
1356 val = (inet->uc_ttl == -1 ?
1357 net->ipv4.sysctl_ip_default_ttl :
1362 val = inet->hdrincl;
1365 val = inet->nodefrag;
1367 case IP_BIND_ADDRESS_NO_PORT:
1368 val = inet->bind_address_no_port;
1370 case IP_MTU_DISCOVER:
1371 val = inet->pmtudisc;
1375 struct dst_entry *dst;
1377 dst = sk_dst_get(sk);
1389 val = inet->recverr;
1391 case IP_MULTICAST_TTL:
1394 case IP_MULTICAST_LOOP:
1395 val = inet->mc_loop;
1398 val = (__force int)htonl((__u32) inet->uc_index);
1400 case IP_MULTICAST_IF:
1402 struct in_addr addr;
1403 len = min_t(unsigned int, len, sizeof(struct in_addr));
1404 addr.s_addr = inet->mc_addr;
1407 if (put_user(len, optlen))
1409 if (copy_to_user(optval, &addr, len))
1415 struct ip_msfilter msf;
1417 if (len < IP_MSFILTER_SIZE(0)) {
1421 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1425 err = ip_mc_msfget(sk, &msf,
1426 (struct ip_msfilter __user *)optval, optlen);
1429 case MCAST_MSFILTER:
1431 struct group_filter gsf;
1433 if (len < GROUP_FILTER_SIZE(0)) {
1437 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1441 err = ip_mc_gsfget(sk, &gsf,
1442 (struct group_filter __user *)optval,
1446 case IP_MULTICAST_ALL:
1455 if (sk->sk_type != SOCK_STREAM)
1456 return -ENOPROTOOPT;
1458 msg.msg_control = (__force void *) optval;
1459 msg.msg_controllen = len;
1460 msg.msg_flags = flags;
1462 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1463 struct in_pktinfo info;
1465 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1466 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1467 info.ipi_ifindex = inet->mc_index;
1468 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1470 if (inet->cmsg_flags & IP_CMSG_TTL) {
1471 int hlim = inet->mc_ttl;
1472 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1474 if (inet->cmsg_flags & IP_CMSG_TOS) {
1475 int tos = inet->rcv_tos;
1476 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1478 len -= msg.msg_controllen;
1479 return put_user(len, optlen);
1482 val = inet->freebind;
1484 case IP_TRANSPARENT:
1485 val = inet->transparent;
1488 val = inet->min_ttl;
1492 return -ENOPROTOOPT;
1496 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1497 unsigned char ucval = (unsigned char)val;
1499 if (put_user(len, optlen))
1501 if (copy_to_user(optval, &ucval, 1))
1504 len = min_t(unsigned int, sizeof(int), len);
1505 if (put_user(len, optlen))
1507 if (copy_to_user(optval, &val, len))
1519 int ip_getsockopt(struct sock *sk, int level,
1520 int optname, char __user *optval, int __user *optlen)
1524 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1525 #ifdef CONFIG_NETFILTER
1526 /* we need to exclude all possible ENOPROTOOPTs except default case */
1527 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1528 !ip_mroute_opt(optname)) {
1531 if (get_user(len, optlen))
1535 err = nf_getsockopt(sk, PF_INET, optname, optval,
1539 err = put_user(len, optlen);
1545 EXPORT_SYMBOL(ip_getsockopt);
1547 #ifdef CONFIG_COMPAT
1548 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1549 char __user *optval, int __user *optlen)
1553 if (optname == MCAST_MSFILTER)
1554 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1557 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1560 #ifdef CONFIG_NETFILTER
1561 /* we need to exclude all possible ENOPROTOOPTs except default case */
1562 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1563 !ip_mroute_opt(optname)) {
1566 if (get_user(len, optlen))
1570 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1573 err = put_user(len, optlen);
1579 EXPORT_SYMBOL(compat_ip_getsockopt);