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->data, offset, 0));
111 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
114 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
120 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
124 err = security_secid_to_secctx(secid, &secdata, &seclen);
128 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
129 security_release_secctx(secdata, seclen);
132 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
134 struct sockaddr_in sin;
135 const struct iphdr *iph = ip_hdr(skb);
136 __be16 *ports = (__be16 *)skb_transport_header(skb);
138 if (skb_transport_offset(skb) + 4 > skb->len)
141 /* All current transport protocols have the port numbers in the
142 * first four bytes of the transport header and this function is
143 * written with this assumption in mind.
146 sin.sin_family = AF_INET;
147 sin.sin_addr.s_addr = iph->daddr;
148 sin.sin_port = ports[1];
149 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
151 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
154 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
157 struct inet_sock *inet = inet_sk(skb->sk);
158 unsigned int flags = inet->cmsg_flags;
160 /* Ordered by supposed usage frequency */
161 if (flags & IP_CMSG_PKTINFO) {
162 ip_cmsg_recv_pktinfo(msg, skb);
164 flags &= ~IP_CMSG_PKTINFO;
169 if (flags & IP_CMSG_TTL) {
170 ip_cmsg_recv_ttl(msg, skb);
172 flags &= ~IP_CMSG_TTL;
177 if (flags & IP_CMSG_TOS) {
178 ip_cmsg_recv_tos(msg, skb);
180 flags &= ~IP_CMSG_TOS;
185 if (flags & IP_CMSG_RECVOPTS) {
186 ip_cmsg_recv_opts(msg, skb);
188 flags &= ~IP_CMSG_RECVOPTS;
193 if (flags & IP_CMSG_RETOPTS) {
194 ip_cmsg_recv_retopts(msg, skb);
196 flags &= ~IP_CMSG_RETOPTS;
201 if (flags & IP_CMSG_PASSSEC) {
202 ip_cmsg_recv_security(msg, skb);
204 flags &= ~IP_CMSG_PASSSEC;
209 if (flags & IP_CMSG_ORIGDSTADDR) {
210 ip_cmsg_recv_dstaddr(msg, skb);
212 flags &= ~IP_CMSG_ORIGDSTADDR;
217 if (flags & IP_CMSG_CHECKSUM)
218 ip_cmsg_recv_checksum(msg, skb, offset);
220 EXPORT_SYMBOL(ip_cmsg_recv_offset);
222 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
226 struct cmsghdr *cmsg;
228 for_each_cmsghdr(cmsg, msg) {
229 if (!CMSG_OK(msg, cmsg))
231 #if IS_ENABLED(CONFIG_IPV6)
233 cmsg->cmsg_level == SOL_IPV6 &&
234 cmsg->cmsg_type == IPV6_PKTINFO) {
235 struct in6_pktinfo *src_info;
237 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
239 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
240 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
242 ipc->oif = src_info->ipi6_ifindex;
243 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
247 if (cmsg->cmsg_level != SOL_IP)
249 switch (cmsg->cmsg_type) {
251 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
252 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
253 err < 40 ? err : 40);
259 struct in_pktinfo *info;
260 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
262 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
263 ipc->oif = info->ipi_ifindex;
264 ipc->addr = info->ipi_spec_dst.s_addr;
268 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
270 val = *(int *)CMSG_DATA(cmsg);
271 if (val < 1 || val > 255)
276 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
278 val = *(int *)CMSG_DATA(cmsg);
279 if (val < 0 || val > 255)
282 ipc->priority = rt_tos2priority(ipc->tos);
293 /* Special input handler for packets caught by router alert option.
294 They are selected only by protocol field, and then processed likely
295 local ones; but only if someone wants them! Otherwise, router
296 not running rsvpd will kill RSVP.
298 It is user level problem, what it will make with them.
299 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
300 but receiver should be enough clever f.e. to forward mtrace requests,
301 sent to multicast group to reach destination designated router.
303 struct ip_ra_chain __rcu *ip_ra_chain;
304 static DEFINE_SPINLOCK(ip_ra_lock);
307 static void ip_ra_destroy_rcu(struct rcu_head *head)
309 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
311 sock_put(ra->saved_sk);
315 int ip_ra_control(struct sock *sk, unsigned char on,
316 void (*destructor)(struct sock *))
318 struct ip_ra_chain *ra, *new_ra;
319 struct ip_ra_chain __rcu **rap;
321 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
324 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
326 spin_lock_bh(&ip_ra_lock);
327 for (rap = &ip_ra_chain;
328 (ra = rcu_dereference_protected(*rap,
329 lockdep_is_held(&ip_ra_lock))) != NULL;
333 spin_unlock_bh(&ip_ra_lock);
337 /* dont let ip_call_ra_chain() use sk again */
339 RCU_INIT_POINTER(*rap, ra->next);
340 spin_unlock_bh(&ip_ra_lock);
345 * Delay sock_put(sk) and kfree(ra) after one rcu grace
346 * period. This guarantee ip_call_ra_chain() dont need
347 * to mess with socket refcounts.
350 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
355 spin_unlock_bh(&ip_ra_lock);
359 new_ra->destructor = destructor;
361 RCU_INIT_POINTER(new_ra->next, ra);
362 rcu_assign_pointer(*rap, new_ra);
364 spin_unlock_bh(&ip_ra_lock);
369 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
370 __be16 port, u32 info, u8 *payload)
372 struct sock_exterr_skb *serr;
374 skb = skb_clone(skb, GFP_ATOMIC);
378 serr = SKB_EXT_ERR(skb);
379 serr->ee.ee_errno = err;
380 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
381 serr->ee.ee_type = icmp_hdr(skb)->type;
382 serr->ee.ee_code = icmp_hdr(skb)->code;
384 serr->ee.ee_info = info;
385 serr->ee.ee_data = 0;
386 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
387 skb_network_header(skb);
390 if (skb_pull(skb, payload - skb->data)) {
391 skb_reset_transport_header(skb);
392 if (sock_queue_err_skb(sk, skb) == 0)
398 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
400 struct inet_sock *inet = inet_sk(sk);
401 struct sock_exterr_skb *serr;
408 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
412 skb_put(skb, sizeof(struct iphdr));
413 skb_reset_network_header(skb);
417 serr = SKB_EXT_ERR(skb);
418 serr->ee.ee_errno = err;
419 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
420 serr->ee.ee_type = 0;
421 serr->ee.ee_code = 0;
423 serr->ee.ee_info = info;
424 serr->ee.ee_data = 0;
425 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
428 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
429 skb_reset_transport_header(skb);
431 if (sock_queue_err_skb(sk, skb))
435 /* For some errors we have valid addr_offset even with zero payload and
436 * zero port. Also, addr_offset should be supported if port is set.
438 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
440 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
441 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
444 /* IPv4 supports cmsg on all imcp errors and some timestamps
446 * Timestamp code paths do not initialize the fields expected by cmsg:
447 * the PKTINFO fields in skb->cb[]. Fill those in here.
449 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
453 struct in_pktinfo *info;
455 if (ee_origin == SO_EE_ORIGIN_ICMP)
458 if (ee_origin == SO_EE_ORIGIN_LOCAL)
461 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
462 * timestamp with egress dev. Not possible for packets without dev
463 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
465 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
469 info = PKTINFO_SKB_CB(skb);
470 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
471 info->ipi_ifindex = skb->dev->ifindex;
476 * Handle MSG_ERRQUEUE
478 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
480 struct sock_exterr_skb *serr;
482 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
484 struct sock_extended_err ee;
485 struct sockaddr_in offender;
490 WARN_ON_ONCE(sk->sk_family == AF_INET6);
493 skb = sock_dequeue_err_skb(sk);
499 msg->msg_flags |= MSG_TRUNC;
502 err = skb_copy_datagram_msg(skb, 0, msg, copied);
506 sock_recv_timestamp(msg, sk, skb);
508 serr = SKB_EXT_ERR(skb);
510 if (sin && ipv4_datagram_support_addr(serr)) {
511 sin->sin_family = AF_INET;
512 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
514 sin->sin_port = serr->port;
515 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
516 *addr_len = sizeof(*sin);
519 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
520 sin = &errhdr.offender;
521 memset(sin, 0, sizeof(*sin));
523 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
524 sin->sin_family = AF_INET;
525 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
526 if (inet_sk(sk)->cmsg_flags)
527 ip_cmsg_recv(msg, skb);
530 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
532 /* Now we could try to dump offended packet options */
534 msg->msg_flags |= MSG_ERRQUEUE;
545 * Socket option code for IP. This is the end of the line after any
546 * TCP,UDP etc options on an IP socket.
548 static bool setsockopt_needs_rtnl(int optname)
551 case IP_ADD_MEMBERSHIP:
552 case IP_ADD_SOURCE_MEMBERSHIP:
553 case IP_BLOCK_SOURCE:
554 case IP_DROP_MEMBERSHIP:
555 case IP_DROP_SOURCE_MEMBERSHIP:
557 case IP_UNBLOCK_SOURCE:
558 case MCAST_BLOCK_SOURCE:
560 case MCAST_JOIN_GROUP:
561 case MCAST_JOIN_SOURCE_GROUP:
562 case MCAST_LEAVE_GROUP:
563 case MCAST_LEAVE_SOURCE_GROUP:
564 case MCAST_UNBLOCK_SOURCE:
570 static int do_ip_setsockopt(struct sock *sk, int level,
571 int optname, char __user *optval, unsigned int optlen)
573 struct inet_sock *inet = inet_sk(sk);
575 bool needs_rtnl = setsockopt_needs_rtnl(optname);
586 case IP_MTU_DISCOVER:
588 case IP_ROUTER_ALERT:
595 case IP_MULTICAST_TTL:
596 case IP_MULTICAST_ALL:
597 case IP_MULTICAST_LOOP:
598 case IP_RECVORIGDSTADDR:
600 if (optlen >= sizeof(int)) {
601 if (get_user(val, (int __user *) optval))
603 } else if (optlen >= sizeof(char)) {
606 if (get_user(ucval, (unsigned char __user *) optval))
612 /* If optlen==0, it is equivalent to val == 0 */
614 if (ip_mroute_opt(optname))
615 return ip_mroute_setsockopt(sk, optname, optval, optlen);
625 struct ip_options_rcu *old, *opt = NULL;
629 err = ip_options_get_from_user(sock_net(sk), &opt,
633 old = rcu_dereference_protected(inet->inet_opt,
634 sock_owned_by_user(sk));
636 struct inet_connection_sock *icsk = inet_csk(sk);
637 #if IS_ENABLED(CONFIG_IPV6)
638 if (sk->sk_family == PF_INET ||
639 (!((1 << sk->sk_state) &
640 (TCPF_LISTEN | TCPF_CLOSE)) &&
641 inet->inet_daddr != LOOPBACK4_IPV6)) {
644 icsk->icsk_ext_hdr_len -= old->opt.optlen;
646 icsk->icsk_ext_hdr_len += opt->opt.optlen;
647 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
648 #if IS_ENABLED(CONFIG_IPV6)
652 rcu_assign_pointer(inet->inet_opt, opt);
659 inet->cmsg_flags |= IP_CMSG_PKTINFO;
661 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
665 inet->cmsg_flags |= IP_CMSG_TTL;
667 inet->cmsg_flags &= ~IP_CMSG_TTL;
671 inet->cmsg_flags |= IP_CMSG_TOS;
673 inet->cmsg_flags &= ~IP_CMSG_TOS;
677 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
679 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
683 inet->cmsg_flags |= IP_CMSG_RETOPTS;
685 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
689 inet->cmsg_flags |= IP_CMSG_PASSSEC;
691 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
693 case IP_RECVORIGDSTADDR:
695 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
697 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
701 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
702 inet_inc_convert_csum(sk);
703 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
706 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
707 inet_dec_convert_csum(sk);
708 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
712 case IP_TOS: /* This sets both TOS and Precedence */
713 if (sk->sk_type == SOCK_STREAM) {
714 val &= ~INET_ECN_MASK;
715 val |= inet->tos & INET_ECN_MASK;
717 if (inet->tos != val) {
719 sk->sk_priority = rt_tos2priority(val);
726 if (val != -1 && (val < 1 || val > 255))
731 if (sk->sk_type != SOCK_RAW) {
735 inet->hdrincl = val ? 1 : 0;
738 if (sk->sk_type != SOCK_RAW) {
742 inet->nodefrag = val ? 1 : 0;
744 case IP_MTU_DISCOVER:
745 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
747 inet->pmtudisc = val;
750 inet->recverr = !!val;
752 skb_queue_purge(&sk->sk_error_queue);
754 case IP_MULTICAST_TTL:
755 if (sk->sk_type == SOCK_STREAM)
761 if (val < 0 || val > 255)
765 case IP_MULTICAST_LOOP:
768 inet->mc_loop = !!val;
772 struct net_device *dev = NULL;
775 if (optlen != sizeof(int))
778 ifindex = (__force int)ntohl((__force __be32)val);
785 dev = dev_get_by_index(sock_net(sk), ifindex);
786 err = -EADDRNOTAVAIL;
792 if (sk->sk_bound_dev_if)
795 inet->uc_index = ifindex;
799 case IP_MULTICAST_IF:
801 struct ip_mreqn mreq;
802 struct net_device *dev = NULL;
804 if (sk->sk_type == SOCK_STREAM)
807 * Check the arguments are allowable
810 if (optlen < sizeof(struct in_addr))
814 if (optlen >= sizeof(struct ip_mreqn)) {
815 if (copy_from_user(&mreq, optval, sizeof(mreq)))
818 memset(&mreq, 0, sizeof(mreq));
819 if (optlen >= sizeof(struct ip_mreq)) {
820 if (copy_from_user(&mreq, optval,
821 sizeof(struct ip_mreq)))
823 } else if (optlen >= sizeof(struct in_addr)) {
824 if (copy_from_user(&mreq.imr_address, optval,
825 sizeof(struct in_addr)))
830 if (!mreq.imr_ifindex) {
831 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
837 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
839 mreq.imr_ifindex = dev->ifindex;
841 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
844 err = -EADDRNOTAVAIL;
850 if (sk->sk_bound_dev_if &&
851 mreq.imr_ifindex != sk->sk_bound_dev_if)
854 inet->mc_index = mreq.imr_ifindex;
855 inet->mc_addr = mreq.imr_address.s_addr;
860 case IP_ADD_MEMBERSHIP:
861 case IP_DROP_MEMBERSHIP:
863 struct ip_mreqn mreq;
866 if (inet_sk(sk)->is_icsk)
869 if (optlen < sizeof(struct ip_mreq))
872 if (optlen >= sizeof(struct ip_mreqn)) {
873 if (copy_from_user(&mreq, optval, sizeof(mreq)))
876 memset(&mreq, 0, sizeof(mreq));
877 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
881 if (optname == IP_ADD_MEMBERSHIP)
882 err = ip_mc_join_group(sk, &mreq);
884 err = ip_mc_leave_group(sk, &mreq);
889 struct ip_msfilter *msf;
891 if (optlen < IP_MSFILTER_SIZE(0))
893 if (optlen > sysctl_optmem_max) {
897 msf = kmalloc(optlen, GFP_KERNEL);
903 if (copy_from_user(msf, optval, optlen)) {
907 /* numsrc >= (1G-4) overflow in 32 bits */
908 if (msf->imsf_numsrc >= 0x3ffffffcU ||
909 msf->imsf_numsrc > sysctl_igmp_max_msf) {
914 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
919 err = ip_mc_msfilter(sk, msf, 0);
923 case IP_BLOCK_SOURCE:
924 case IP_UNBLOCK_SOURCE:
925 case IP_ADD_SOURCE_MEMBERSHIP:
926 case IP_DROP_SOURCE_MEMBERSHIP:
928 struct ip_mreq_source mreqs;
931 if (optlen != sizeof(struct ip_mreq_source))
933 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
937 if (optname == IP_BLOCK_SOURCE) {
938 omode = MCAST_EXCLUDE;
940 } else if (optname == IP_UNBLOCK_SOURCE) {
941 omode = MCAST_EXCLUDE;
943 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
944 struct ip_mreqn mreq;
946 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
947 mreq.imr_address.s_addr = mreqs.imr_interface;
948 mreq.imr_ifindex = 0;
949 err = ip_mc_join_group(sk, &mreq);
950 if (err && err != -EADDRINUSE)
952 omode = MCAST_INCLUDE;
954 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
955 omode = MCAST_INCLUDE;
958 err = ip_mc_source(add, omode, sk, &mreqs, 0);
961 case MCAST_JOIN_GROUP:
962 case MCAST_LEAVE_GROUP:
964 struct group_req greq;
965 struct sockaddr_in *psin;
966 struct ip_mreqn mreq;
968 if (optlen < sizeof(struct group_req))
971 if (copy_from_user(&greq, optval, sizeof(greq)))
973 psin = (struct sockaddr_in *)&greq.gr_group;
974 if (psin->sin_family != AF_INET)
976 memset(&mreq, 0, sizeof(mreq));
977 mreq.imr_multiaddr = psin->sin_addr;
978 mreq.imr_ifindex = greq.gr_interface;
980 if (optname == MCAST_JOIN_GROUP)
981 err = ip_mc_join_group(sk, &mreq);
983 err = ip_mc_leave_group(sk, &mreq);
986 case MCAST_JOIN_SOURCE_GROUP:
987 case MCAST_LEAVE_SOURCE_GROUP:
988 case MCAST_BLOCK_SOURCE:
989 case MCAST_UNBLOCK_SOURCE:
991 struct group_source_req greqs;
992 struct ip_mreq_source mreqs;
993 struct sockaddr_in *psin;
996 if (optlen != sizeof(struct group_source_req))
998 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1002 if (greqs.gsr_group.ss_family != AF_INET ||
1003 greqs.gsr_source.ss_family != AF_INET) {
1004 err = -EADDRNOTAVAIL;
1007 psin = (struct sockaddr_in *)&greqs.gsr_group;
1008 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1009 psin = (struct sockaddr_in *)&greqs.gsr_source;
1010 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1011 mreqs.imr_interface = 0; /* use index for mc_source */
1013 if (optname == MCAST_BLOCK_SOURCE) {
1014 omode = MCAST_EXCLUDE;
1016 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1017 omode = MCAST_EXCLUDE;
1019 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1020 struct ip_mreqn mreq;
1022 psin = (struct sockaddr_in *)&greqs.gsr_group;
1023 mreq.imr_multiaddr = psin->sin_addr;
1024 mreq.imr_address.s_addr = 0;
1025 mreq.imr_ifindex = greqs.gsr_interface;
1026 err = ip_mc_join_group(sk, &mreq);
1027 if (err && err != -EADDRINUSE)
1029 greqs.gsr_interface = mreq.imr_ifindex;
1030 omode = MCAST_INCLUDE;
1032 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1033 omode = MCAST_INCLUDE;
1036 err = ip_mc_source(add, omode, sk, &mreqs,
1037 greqs.gsr_interface);
1040 case MCAST_MSFILTER:
1042 struct sockaddr_in *psin;
1043 struct ip_msfilter *msf = NULL;
1044 struct group_filter *gsf = NULL;
1045 int msize, i, ifindex;
1047 if (optlen < GROUP_FILTER_SIZE(0))
1049 if (optlen > sysctl_optmem_max) {
1053 gsf = kmalloc(optlen, GFP_KERNEL);
1059 if (copy_from_user(gsf, optval, optlen))
1062 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1063 if (gsf->gf_numsrc >= 0x1ffffff ||
1064 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1068 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1072 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1073 msf = kmalloc(msize, GFP_KERNEL);
1078 ifindex = gsf->gf_interface;
1079 psin = (struct sockaddr_in *)&gsf->gf_group;
1080 if (psin->sin_family != AF_INET) {
1081 err = -EADDRNOTAVAIL;
1084 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1085 msf->imsf_interface = 0;
1086 msf->imsf_fmode = gsf->gf_fmode;
1087 msf->imsf_numsrc = gsf->gf_numsrc;
1088 err = -EADDRNOTAVAIL;
1089 for (i = 0; i < gsf->gf_numsrc; ++i) {
1090 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1092 if (psin->sin_family != AF_INET)
1094 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1099 err = ip_mc_msfilter(sk, msf, ifindex);
1105 case IP_MULTICAST_ALL:
1108 if (val != 0 && val != 1)
1112 case IP_ROUTER_ALERT:
1113 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1119 inet->freebind = !!val;
1122 case IP_IPSEC_POLICY:
1123 case IP_XFRM_POLICY:
1125 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1127 err = xfrm_user_policy(sk, optname, optval, optlen);
1130 case IP_TRANSPARENT:
1131 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1132 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1138 inet->transparent = !!val;
1144 if (val < 0 || val > 255)
1146 inet->min_ttl = val;
1166 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1170 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1171 * destination in skb->cb[] before dst drop.
1172 * This way, receiver doesn't make cache line misses to read rtable.
1174 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1176 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1177 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1180 if (prepare && skb_rtable(skb)) {
1181 pktinfo->ipi_ifindex = inet_iif(skb);
1182 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1184 pktinfo->ipi_ifindex = 0;
1185 pktinfo->ipi_spec_dst.s_addr = 0;
1190 int ip_setsockopt(struct sock *sk, int level,
1191 int optname, char __user *optval, unsigned int optlen)
1195 if (level != SOL_IP)
1196 return -ENOPROTOOPT;
1198 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1199 #ifdef CONFIG_NETFILTER
1200 /* we need to exclude all possible ENOPROTOOPTs except default case */
1201 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1202 optname != IP_IPSEC_POLICY &&
1203 optname != IP_XFRM_POLICY &&
1204 !ip_mroute_opt(optname)) {
1206 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1212 EXPORT_SYMBOL(ip_setsockopt);
1214 #ifdef CONFIG_COMPAT
1215 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1216 char __user *optval, unsigned int optlen)
1220 if (level != SOL_IP)
1221 return -ENOPROTOOPT;
1223 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1224 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1227 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1228 #ifdef CONFIG_NETFILTER
1229 /* we need to exclude all possible ENOPROTOOPTs except default case */
1230 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1231 optname != IP_IPSEC_POLICY &&
1232 optname != IP_XFRM_POLICY &&
1233 !ip_mroute_opt(optname)) {
1235 err = compat_nf_setsockopt(sk, PF_INET, optname,
1242 EXPORT_SYMBOL(compat_ip_setsockopt);
1246 * Get the options. Note for future reference. The GET of IP options gets
1247 * the _received_ ones. The set sets the _sent_ ones.
1250 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1251 char __user *optval, int __user *optlen, unsigned int flags)
1253 struct inet_sock *inet = inet_sk(sk);
1257 if (level != SOL_IP)
1260 if (ip_mroute_opt(optname))
1261 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1263 if (get_user(len, optlen))
1273 unsigned char optbuf[sizeof(struct ip_options)+40];
1274 struct ip_options *opt = (struct ip_options *)optbuf;
1275 struct ip_options_rcu *inet_opt;
1277 inet_opt = rcu_dereference_protected(inet->inet_opt,
1278 sock_owned_by_user(sk));
1281 memcpy(optbuf, &inet_opt->opt,
1282 sizeof(struct ip_options) +
1283 inet_opt->opt.optlen);
1286 if (opt->optlen == 0)
1287 return put_user(0, optlen);
1289 ip_options_undo(opt);
1291 len = min_t(unsigned int, len, opt->optlen);
1292 if (put_user(len, optlen))
1294 if (copy_to_user(optval, opt->__data, len))
1299 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1302 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1305 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1308 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1311 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1314 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1316 case IP_RECVORIGDSTADDR:
1317 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1320 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1326 val = (inet->uc_ttl == -1 ?
1327 sysctl_ip_default_ttl :
1331 val = inet->hdrincl;
1334 val = inet->nodefrag;
1336 case IP_MTU_DISCOVER:
1337 val = inet->pmtudisc;
1341 struct dst_entry *dst;
1343 dst = sk_dst_get(sk);
1355 val = inet->recverr;
1357 case IP_MULTICAST_TTL:
1360 case IP_MULTICAST_LOOP:
1361 val = inet->mc_loop;
1364 val = (__force int)htonl((__u32) inet->uc_index);
1366 case IP_MULTICAST_IF:
1368 struct in_addr addr;
1369 len = min_t(unsigned int, len, sizeof(struct in_addr));
1370 addr.s_addr = inet->mc_addr;
1373 if (put_user(len, optlen))
1375 if (copy_to_user(optval, &addr, len))
1381 struct ip_msfilter msf;
1384 if (len < IP_MSFILTER_SIZE(0)) {
1388 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1392 err = ip_mc_msfget(sk, &msf,
1393 (struct ip_msfilter __user *)optval, optlen);
1397 case MCAST_MSFILTER:
1399 struct group_filter gsf;
1402 if (len < GROUP_FILTER_SIZE(0)) {
1406 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1410 err = ip_mc_gsfget(sk, &gsf,
1411 (struct group_filter __user *)optval,
1416 case IP_MULTICAST_ALL:
1425 if (sk->sk_type != SOCK_STREAM)
1426 return -ENOPROTOOPT;
1428 msg.msg_control = (__force void *) optval;
1429 msg.msg_controllen = len;
1430 msg.msg_flags = flags;
1432 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1433 struct in_pktinfo info;
1435 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1436 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1437 info.ipi_ifindex = inet->mc_index;
1438 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1440 if (inet->cmsg_flags & IP_CMSG_TTL) {
1441 int hlim = inet->mc_ttl;
1442 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1444 if (inet->cmsg_flags & IP_CMSG_TOS) {
1445 int tos = inet->rcv_tos;
1446 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1448 len -= msg.msg_controllen;
1449 return put_user(len, optlen);
1452 val = inet->freebind;
1454 case IP_TRANSPARENT:
1455 val = inet->transparent;
1458 val = inet->min_ttl;
1462 return -ENOPROTOOPT;
1466 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1467 unsigned char ucval = (unsigned char)val;
1469 if (put_user(len, optlen))
1471 if (copy_to_user(optval, &ucval, 1))
1474 len = min_t(unsigned int, sizeof(int), len);
1475 if (put_user(len, optlen))
1477 if (copy_to_user(optval, &val, len))
1483 int ip_getsockopt(struct sock *sk, int level,
1484 int optname, char __user *optval, int __user *optlen)
1488 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1489 #ifdef CONFIG_NETFILTER
1490 /* we need to exclude all possible ENOPROTOOPTs except default case */
1491 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1492 !ip_mroute_opt(optname)) {
1495 if (get_user(len, optlen))
1499 err = nf_getsockopt(sk, PF_INET, optname, optval,
1503 err = put_user(len, optlen);
1509 EXPORT_SYMBOL(ip_getsockopt);
1511 #ifdef CONFIG_COMPAT
1512 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1513 char __user *optval, int __user *optlen)
1517 if (optname == MCAST_MSFILTER)
1518 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1521 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1524 #ifdef CONFIG_NETFILTER
1525 /* we need to exclude all possible ENOPROTOOPTs except default case */
1526 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1527 !ip_mroute_opt(optname)) {
1530 if (get_user(len, optlen))
1534 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1537 err = put_user(len, optlen);
1543 EXPORT_SYMBOL(compat_ip_getsockopt);