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 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
256 if (cmsg->cmsg_level != SOL_IP)
258 switch (cmsg->cmsg_type) {
260 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
262 /* Our caller is responsible for freeing ipc->opt */
263 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
264 err < 40 ? err : 40);
270 struct in_pktinfo *info;
271 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
273 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
274 ipc->oif = info->ipi_ifindex;
275 ipc->addr = info->ipi_spec_dst.s_addr;
279 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
281 val = *(int *)CMSG_DATA(cmsg);
282 if (val < 1 || val > 255)
287 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
288 val = *(int *)CMSG_DATA(cmsg);
289 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
290 val = *(u8 *)CMSG_DATA(cmsg);
293 if (val < 0 || val > 255)
296 ipc->priority = rt_tos2priority(ipc->tos);
307 /* Special input handler for packets caught by router alert option.
308 They are selected only by protocol field, and then processed likely
309 local ones; but only if someone wants them! Otherwise, router
310 not running rsvpd will kill RSVP.
312 It is user level problem, what it will make with them.
313 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
314 but receiver should be enough clever f.e. to forward mtrace requests,
315 sent to multicast group to reach destination designated router.
317 struct ip_ra_chain __rcu *ip_ra_chain;
318 static DEFINE_SPINLOCK(ip_ra_lock);
321 static void ip_ra_destroy_rcu(struct rcu_head *head)
323 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
325 sock_put(ra->saved_sk);
329 int ip_ra_control(struct sock *sk, unsigned char on,
330 void (*destructor)(struct sock *))
332 struct ip_ra_chain *ra, *new_ra;
333 struct ip_ra_chain __rcu **rap;
335 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
338 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
340 spin_lock_bh(&ip_ra_lock);
341 for (rap = &ip_ra_chain;
342 (ra = rcu_dereference_protected(*rap,
343 lockdep_is_held(&ip_ra_lock))) != NULL;
347 spin_unlock_bh(&ip_ra_lock);
351 /* dont let ip_call_ra_chain() use sk again */
353 RCU_INIT_POINTER(*rap, ra->next);
354 spin_unlock_bh(&ip_ra_lock);
359 * Delay sock_put(sk) and kfree(ra) after one rcu grace
360 * period. This guarantee ip_call_ra_chain() dont need
361 * to mess with socket refcounts.
364 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
369 spin_unlock_bh(&ip_ra_lock);
373 new_ra->destructor = destructor;
375 RCU_INIT_POINTER(new_ra->next, ra);
376 rcu_assign_pointer(*rap, new_ra);
378 spin_unlock_bh(&ip_ra_lock);
383 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
384 __be16 port, u32 info, u8 *payload)
386 struct sock_exterr_skb *serr;
388 skb = skb_clone(skb, GFP_ATOMIC);
392 serr = SKB_EXT_ERR(skb);
393 serr->ee.ee_errno = err;
394 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
395 serr->ee.ee_type = icmp_hdr(skb)->type;
396 serr->ee.ee_code = icmp_hdr(skb)->code;
398 serr->ee.ee_info = info;
399 serr->ee.ee_data = 0;
400 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
401 skb_network_header(skb);
404 if (skb_pull(skb, payload - skb->data)) {
405 skb_reset_transport_header(skb);
406 if (sock_queue_err_skb(sk, skb) == 0)
412 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
414 struct inet_sock *inet = inet_sk(sk);
415 struct sock_exterr_skb *serr;
422 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
426 skb_put(skb, sizeof(struct iphdr));
427 skb_reset_network_header(skb);
431 serr = SKB_EXT_ERR(skb);
432 serr->ee.ee_errno = err;
433 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
434 serr->ee.ee_type = 0;
435 serr->ee.ee_code = 0;
437 serr->ee.ee_info = info;
438 serr->ee.ee_data = 0;
439 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
442 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
443 skb_reset_transport_header(skb);
445 if (sock_queue_err_skb(sk, skb))
449 /* For some errors we have valid addr_offset even with zero payload and
450 * zero port. Also, addr_offset should be supported if port is set.
452 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
454 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
455 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
458 /* IPv4 supports cmsg on all imcp errors and some timestamps
460 * Timestamp code paths do not initialize the fields expected by cmsg:
461 * the PKTINFO fields in skb->cb[]. Fill those in here.
463 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
467 struct in_pktinfo *info;
469 if (ee_origin == SO_EE_ORIGIN_ICMP)
472 if (ee_origin == SO_EE_ORIGIN_LOCAL)
475 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
476 * timestamp with egress dev. Not possible for packets without dev
477 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
479 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
483 info = PKTINFO_SKB_CB(skb);
484 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
485 info->ipi_ifindex = skb->dev->ifindex;
490 * Handle MSG_ERRQUEUE
492 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
494 struct sock_exterr_skb *serr;
496 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
498 struct sock_extended_err ee;
499 struct sockaddr_in offender;
504 WARN_ON_ONCE(sk->sk_family == AF_INET6);
507 skb = sock_dequeue_err_skb(sk);
513 msg->msg_flags |= MSG_TRUNC;
516 err = skb_copy_datagram_msg(skb, 0, msg, copied);
521 sock_recv_timestamp(msg, sk, skb);
523 serr = SKB_EXT_ERR(skb);
525 if (sin && ipv4_datagram_support_addr(serr)) {
526 sin->sin_family = AF_INET;
527 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
529 sin->sin_port = serr->port;
530 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
531 *addr_len = sizeof(*sin);
534 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
535 sin = &errhdr.offender;
536 memset(sin, 0, sizeof(*sin));
538 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
539 sin->sin_family = AF_INET;
540 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
541 if (inet_sk(sk)->cmsg_flags)
542 ip_cmsg_recv(msg, skb);
545 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
547 /* Now we could try to dump offended packet options */
549 msg->msg_flags |= MSG_ERRQUEUE;
559 * Socket option code for IP. This is the end of the line after any
560 * TCP,UDP etc options on an IP socket.
562 static bool setsockopt_needs_rtnl(int optname)
565 case IP_ADD_MEMBERSHIP:
566 case IP_ADD_SOURCE_MEMBERSHIP:
567 case IP_BLOCK_SOURCE:
568 case IP_DROP_MEMBERSHIP:
569 case IP_DROP_SOURCE_MEMBERSHIP:
571 case IP_UNBLOCK_SOURCE:
572 case MCAST_BLOCK_SOURCE:
574 case MCAST_JOIN_GROUP:
575 case MCAST_JOIN_SOURCE_GROUP:
576 case MCAST_LEAVE_GROUP:
577 case MCAST_LEAVE_SOURCE_GROUP:
578 case MCAST_UNBLOCK_SOURCE:
584 static int do_ip_setsockopt(struct sock *sk, int level,
585 int optname, char __user *optval, unsigned int optlen)
587 struct inet_sock *inet = inet_sk(sk);
588 struct net *net = sock_net(sk);
590 bool needs_rtnl = setsockopt_needs_rtnl(optname);
601 case IP_MTU_DISCOVER:
603 case IP_ROUTER_ALERT:
609 case IP_BIND_ADDRESS_NO_PORT:
611 case IP_MULTICAST_TTL:
612 case IP_MULTICAST_ALL:
613 case IP_MULTICAST_LOOP:
614 case IP_RECVORIGDSTADDR:
616 if (optlen >= sizeof(int)) {
617 if (get_user(val, (int __user *) optval))
619 } else if (optlen >= sizeof(char)) {
622 if (get_user(ucval, (unsigned char __user *) optval))
628 /* If optlen==0, it is equivalent to val == 0 */
630 if (ip_mroute_opt(optname))
631 return ip_mroute_setsockopt(sk, optname, optval, optlen);
641 struct ip_options_rcu *old, *opt = NULL;
645 err = ip_options_get_from_user(sock_net(sk), &opt,
649 old = rcu_dereference_protected(inet->inet_opt,
650 lockdep_sock_is_held(sk));
652 struct inet_connection_sock *icsk = inet_csk(sk);
653 #if IS_ENABLED(CONFIG_IPV6)
654 if (sk->sk_family == PF_INET ||
655 (!((1 << sk->sk_state) &
656 (TCPF_LISTEN | TCPF_CLOSE)) &&
657 inet->inet_daddr != LOOPBACK4_IPV6)) {
660 icsk->icsk_ext_hdr_len -= old->opt.optlen;
662 icsk->icsk_ext_hdr_len += opt->opt.optlen;
663 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
664 #if IS_ENABLED(CONFIG_IPV6)
668 rcu_assign_pointer(inet->inet_opt, opt);
675 inet->cmsg_flags |= IP_CMSG_PKTINFO;
677 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
681 inet->cmsg_flags |= IP_CMSG_TTL;
683 inet->cmsg_flags &= ~IP_CMSG_TTL;
687 inet->cmsg_flags |= IP_CMSG_TOS;
689 inet->cmsg_flags &= ~IP_CMSG_TOS;
693 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
695 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
699 inet->cmsg_flags |= IP_CMSG_RETOPTS;
701 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
705 inet->cmsg_flags |= IP_CMSG_PASSSEC;
707 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
709 case IP_RECVORIGDSTADDR:
711 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
713 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
717 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
718 inet_inc_convert_csum(sk);
719 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
722 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
723 inet_dec_convert_csum(sk);
724 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
728 case IP_TOS: /* This sets both TOS and Precedence */
729 if (sk->sk_type == SOCK_STREAM) {
730 val &= ~INET_ECN_MASK;
731 val |= inet->tos & INET_ECN_MASK;
733 if (inet->tos != val) {
735 sk->sk_priority = rt_tos2priority(val);
742 if (val != -1 && (val < 1 || val > 255))
747 if (sk->sk_type != SOCK_RAW) {
751 inet->hdrincl = val ? 1 : 0;
754 if (sk->sk_type != SOCK_RAW) {
758 inet->nodefrag = val ? 1 : 0;
760 case IP_BIND_ADDRESS_NO_PORT:
761 inet->bind_address_no_port = val ? 1 : 0;
763 case IP_MTU_DISCOVER:
764 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
766 inet->pmtudisc = val;
769 inet->recverr = !!val;
771 skb_queue_purge(&sk->sk_error_queue);
773 case IP_MULTICAST_TTL:
774 if (sk->sk_type == SOCK_STREAM)
780 if (val < 0 || val > 255)
784 case IP_MULTICAST_LOOP:
787 inet->mc_loop = !!val;
791 struct net_device *dev = NULL;
794 if (optlen != sizeof(int))
797 ifindex = (__force int)ntohl((__force __be32)val);
804 dev = dev_get_by_index(sock_net(sk), ifindex);
805 err = -EADDRNOTAVAIL;
811 if (sk->sk_bound_dev_if)
814 inet->uc_index = ifindex;
818 case IP_MULTICAST_IF:
820 struct ip_mreqn mreq;
821 struct net_device *dev = NULL;
823 if (sk->sk_type == SOCK_STREAM)
826 * Check the arguments are allowable
829 if (optlen < sizeof(struct in_addr))
833 if (optlen >= sizeof(struct ip_mreqn)) {
834 if (copy_from_user(&mreq, optval, sizeof(mreq)))
837 memset(&mreq, 0, sizeof(mreq));
838 if (optlen >= sizeof(struct ip_mreq)) {
839 if (copy_from_user(&mreq, optval,
840 sizeof(struct ip_mreq)))
842 } else if (optlen >= sizeof(struct in_addr)) {
843 if (copy_from_user(&mreq.imr_address, optval,
844 sizeof(struct in_addr)))
849 if (!mreq.imr_ifindex) {
850 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
856 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
858 mreq.imr_ifindex = dev->ifindex;
860 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
863 err = -EADDRNOTAVAIL;
869 if (sk->sk_bound_dev_if &&
870 mreq.imr_ifindex != sk->sk_bound_dev_if)
873 inet->mc_index = mreq.imr_ifindex;
874 inet->mc_addr = mreq.imr_address.s_addr;
879 case IP_ADD_MEMBERSHIP:
880 case IP_DROP_MEMBERSHIP:
882 struct ip_mreqn mreq;
885 if (inet_sk(sk)->is_icsk)
888 if (optlen < sizeof(struct ip_mreq))
891 if (optlen >= sizeof(struct ip_mreqn)) {
892 if (copy_from_user(&mreq, optval, sizeof(mreq)))
895 memset(&mreq, 0, sizeof(mreq));
896 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
900 if (optname == IP_ADD_MEMBERSHIP)
901 err = ip_mc_join_group(sk, &mreq);
903 err = ip_mc_leave_group(sk, &mreq);
908 struct ip_msfilter *msf;
910 if (optlen < IP_MSFILTER_SIZE(0))
912 if (optlen > sysctl_optmem_max) {
916 msf = kmalloc(optlen, GFP_KERNEL);
922 if (copy_from_user(msf, optval, optlen)) {
926 /* numsrc >= (1G-4) overflow in 32 bits */
927 if (msf->imsf_numsrc >= 0x3ffffffcU ||
928 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
933 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
938 err = ip_mc_msfilter(sk, msf, 0);
942 case IP_BLOCK_SOURCE:
943 case IP_UNBLOCK_SOURCE:
944 case IP_ADD_SOURCE_MEMBERSHIP:
945 case IP_DROP_SOURCE_MEMBERSHIP:
947 struct ip_mreq_source mreqs;
950 if (optlen != sizeof(struct ip_mreq_source))
952 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
956 if (optname == IP_BLOCK_SOURCE) {
957 omode = MCAST_EXCLUDE;
959 } else if (optname == IP_UNBLOCK_SOURCE) {
960 omode = MCAST_EXCLUDE;
962 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
963 struct ip_mreqn mreq;
965 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
966 mreq.imr_address.s_addr = mreqs.imr_interface;
967 mreq.imr_ifindex = 0;
968 err = ip_mc_join_group(sk, &mreq);
969 if (err && err != -EADDRINUSE)
971 omode = MCAST_INCLUDE;
973 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
974 omode = MCAST_INCLUDE;
977 err = ip_mc_source(add, omode, sk, &mreqs, 0);
980 case MCAST_JOIN_GROUP:
981 case MCAST_LEAVE_GROUP:
983 struct group_req greq;
984 struct sockaddr_in *psin;
985 struct ip_mreqn mreq;
987 if (optlen < sizeof(struct group_req))
990 if (copy_from_user(&greq, optval, sizeof(greq)))
992 psin = (struct sockaddr_in *)&greq.gr_group;
993 if (psin->sin_family != AF_INET)
995 memset(&mreq, 0, sizeof(mreq));
996 mreq.imr_multiaddr = psin->sin_addr;
997 mreq.imr_ifindex = greq.gr_interface;
999 if (optname == MCAST_JOIN_GROUP)
1000 err = ip_mc_join_group(sk, &mreq);
1002 err = ip_mc_leave_group(sk, &mreq);
1005 case MCAST_JOIN_SOURCE_GROUP:
1006 case MCAST_LEAVE_SOURCE_GROUP:
1007 case MCAST_BLOCK_SOURCE:
1008 case MCAST_UNBLOCK_SOURCE:
1010 struct group_source_req greqs;
1011 struct ip_mreq_source mreqs;
1012 struct sockaddr_in *psin;
1015 if (optlen != sizeof(struct group_source_req))
1017 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1021 if (greqs.gsr_group.ss_family != AF_INET ||
1022 greqs.gsr_source.ss_family != AF_INET) {
1023 err = -EADDRNOTAVAIL;
1026 psin = (struct sockaddr_in *)&greqs.gsr_group;
1027 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1028 psin = (struct sockaddr_in *)&greqs.gsr_source;
1029 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1030 mreqs.imr_interface = 0; /* use index for mc_source */
1032 if (optname == MCAST_BLOCK_SOURCE) {
1033 omode = MCAST_EXCLUDE;
1035 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1036 omode = MCAST_EXCLUDE;
1038 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1039 struct ip_mreqn mreq;
1041 psin = (struct sockaddr_in *)&greqs.gsr_group;
1042 mreq.imr_multiaddr = psin->sin_addr;
1043 mreq.imr_address.s_addr = 0;
1044 mreq.imr_ifindex = greqs.gsr_interface;
1045 err = ip_mc_join_group(sk, &mreq);
1046 if (err && err != -EADDRINUSE)
1048 greqs.gsr_interface = mreq.imr_ifindex;
1049 omode = MCAST_INCLUDE;
1051 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1052 omode = MCAST_INCLUDE;
1055 err = ip_mc_source(add, omode, sk, &mreqs,
1056 greqs.gsr_interface);
1059 case MCAST_MSFILTER:
1061 struct sockaddr_in *psin;
1062 struct ip_msfilter *msf = NULL;
1063 struct group_filter *gsf = NULL;
1064 int msize, i, ifindex;
1066 if (optlen < GROUP_FILTER_SIZE(0))
1068 if (optlen > sysctl_optmem_max) {
1072 gsf = kmalloc(optlen, GFP_KERNEL);
1078 if (copy_from_user(gsf, optval, optlen))
1081 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1082 if (gsf->gf_numsrc >= 0x1ffffff ||
1083 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1087 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1091 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1092 msf = kmalloc(msize, GFP_KERNEL);
1097 ifindex = gsf->gf_interface;
1098 psin = (struct sockaddr_in *)&gsf->gf_group;
1099 if (psin->sin_family != AF_INET) {
1100 err = -EADDRNOTAVAIL;
1103 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1104 msf->imsf_interface = 0;
1105 msf->imsf_fmode = gsf->gf_fmode;
1106 msf->imsf_numsrc = gsf->gf_numsrc;
1107 err = -EADDRNOTAVAIL;
1108 for (i = 0; i < gsf->gf_numsrc; ++i) {
1109 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1111 if (psin->sin_family != AF_INET)
1113 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1118 err = ip_mc_msfilter(sk, msf, ifindex);
1124 case IP_MULTICAST_ALL:
1127 if (val != 0 && val != 1)
1131 case IP_ROUTER_ALERT:
1132 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1138 inet->freebind = !!val;
1141 case IP_IPSEC_POLICY:
1142 case IP_XFRM_POLICY:
1144 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1146 err = xfrm_user_policy(sk, optname, optval, optlen);
1149 case IP_TRANSPARENT:
1150 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1151 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1157 inet->transparent = !!val;
1163 if (val < 0 || val > 255)
1165 inet->min_ttl = val;
1185 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1189 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1190 * destination in skb->cb[] before dst drop.
1191 * This way, receiver doesn't make cache line misses to read rtable.
1193 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1195 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1196 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1199 if (prepare && skb_rtable(skb)) {
1200 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1201 * which has interface index (iif) as the first member of the
1202 * underlying inet{6}_skb_parm struct. This code then overlays
1203 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1204 * element so the iif is picked up from the prior IPCB
1206 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1208 pktinfo->ipi_ifindex = 0;
1209 pktinfo->ipi_spec_dst.s_addr = 0;
1214 int ip_setsockopt(struct sock *sk, int level,
1215 int optname, char __user *optval, unsigned int optlen)
1219 if (level != SOL_IP)
1220 return -ENOPROTOOPT;
1222 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1223 #ifdef CONFIG_NETFILTER
1224 /* we need to exclude all possible ENOPROTOOPTs except default case */
1225 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1226 optname != IP_IPSEC_POLICY &&
1227 optname != IP_XFRM_POLICY &&
1228 !ip_mroute_opt(optname)) {
1230 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1236 EXPORT_SYMBOL(ip_setsockopt);
1238 #ifdef CONFIG_COMPAT
1239 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1240 char __user *optval, unsigned int optlen)
1244 if (level != SOL_IP)
1245 return -ENOPROTOOPT;
1247 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1248 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1251 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1252 #ifdef CONFIG_NETFILTER
1253 /* we need to exclude all possible ENOPROTOOPTs except default case */
1254 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1255 optname != IP_IPSEC_POLICY &&
1256 optname != IP_XFRM_POLICY &&
1257 !ip_mroute_opt(optname)) {
1259 err = compat_nf_setsockopt(sk, PF_INET, optname,
1266 EXPORT_SYMBOL(compat_ip_setsockopt);
1270 * Get the options. Note for future reference. The GET of IP options gets
1271 * the _received_ ones. The set sets the _sent_ ones.
1274 static bool getsockopt_needs_rtnl(int optname)
1278 case MCAST_MSFILTER:
1284 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1285 char __user *optval, int __user *optlen, unsigned int flags)
1287 struct inet_sock *inet = inet_sk(sk);
1288 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1292 if (level != SOL_IP)
1295 if (ip_mroute_opt(optname))
1296 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1298 if (get_user(len, optlen))
1310 unsigned char optbuf[sizeof(struct ip_options)+40];
1311 struct ip_options *opt = (struct ip_options *)optbuf;
1312 struct ip_options_rcu *inet_opt;
1314 inet_opt = rcu_dereference_protected(inet->inet_opt,
1315 lockdep_sock_is_held(sk));
1318 memcpy(optbuf, &inet_opt->opt,
1319 sizeof(struct ip_options) +
1320 inet_opt->opt.optlen);
1323 if (opt->optlen == 0)
1324 return put_user(0, optlen);
1326 ip_options_undo(opt);
1328 len = min_t(unsigned int, len, opt->optlen);
1329 if (put_user(len, optlen))
1331 if (copy_to_user(optval, opt->__data, len))
1336 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1339 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1342 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1345 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1348 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1351 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1353 case IP_RECVORIGDSTADDR:
1354 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1357 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1364 struct net *net = sock_net(sk);
1365 val = (inet->uc_ttl == -1 ?
1366 net->ipv4.sysctl_ip_default_ttl :
1371 val = inet->hdrincl;
1374 val = inet->nodefrag;
1376 case IP_BIND_ADDRESS_NO_PORT:
1377 val = inet->bind_address_no_port;
1379 case IP_MTU_DISCOVER:
1380 val = inet->pmtudisc;
1384 struct dst_entry *dst;
1386 dst = sk_dst_get(sk);
1398 val = inet->recverr;
1400 case IP_MULTICAST_TTL:
1403 case IP_MULTICAST_LOOP:
1404 val = inet->mc_loop;
1407 val = (__force int)htonl((__u32) inet->uc_index);
1409 case IP_MULTICAST_IF:
1411 struct in_addr addr;
1412 len = min_t(unsigned int, len, sizeof(struct in_addr));
1413 addr.s_addr = inet->mc_addr;
1416 if (put_user(len, optlen))
1418 if (copy_to_user(optval, &addr, len))
1424 struct ip_msfilter msf;
1426 if (len < IP_MSFILTER_SIZE(0)) {
1430 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1434 err = ip_mc_msfget(sk, &msf,
1435 (struct ip_msfilter __user *)optval, optlen);
1438 case MCAST_MSFILTER:
1440 struct group_filter gsf;
1442 if (len < GROUP_FILTER_SIZE(0)) {
1446 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1450 err = ip_mc_gsfget(sk, &gsf,
1451 (struct group_filter __user *)optval,
1455 case IP_MULTICAST_ALL:
1464 if (sk->sk_type != SOCK_STREAM)
1465 return -ENOPROTOOPT;
1467 msg.msg_control = (__force void *) optval;
1468 msg.msg_controllen = len;
1469 msg.msg_flags = flags;
1471 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1472 struct in_pktinfo info;
1474 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1475 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1476 info.ipi_ifindex = inet->mc_index;
1477 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1479 if (inet->cmsg_flags & IP_CMSG_TTL) {
1480 int hlim = inet->mc_ttl;
1481 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1483 if (inet->cmsg_flags & IP_CMSG_TOS) {
1484 int tos = inet->rcv_tos;
1485 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1487 len -= msg.msg_controllen;
1488 return put_user(len, optlen);
1491 val = inet->freebind;
1493 case IP_TRANSPARENT:
1494 val = inet->transparent;
1497 val = inet->min_ttl;
1501 return -ENOPROTOOPT;
1505 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1506 unsigned char ucval = (unsigned char)val;
1508 if (put_user(len, optlen))
1510 if (copy_to_user(optval, &ucval, 1))
1513 len = min_t(unsigned int, sizeof(int), len);
1514 if (put_user(len, optlen))
1516 if (copy_to_user(optval, &val, len))
1528 int ip_getsockopt(struct sock *sk, int level,
1529 int optname, char __user *optval, int __user *optlen)
1533 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1534 #ifdef CONFIG_NETFILTER
1535 /* we need to exclude all possible ENOPROTOOPTs except default case */
1536 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1537 !ip_mroute_opt(optname)) {
1540 if (get_user(len, optlen))
1544 err = nf_getsockopt(sk, PF_INET, optname, optval,
1548 err = put_user(len, optlen);
1554 EXPORT_SYMBOL(ip_getsockopt);
1556 #ifdef CONFIG_COMPAT
1557 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1558 char __user *optval, int __user *optlen)
1562 if (optname == MCAST_MSFILTER)
1563 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1566 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1569 #ifdef CONFIG_NETFILTER
1570 /* we need to exclude all possible ENOPROTOOPTs except default case */
1571 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1572 !ip_mroute_opt(optname)) {
1575 if (get_user(len, optlen))
1579 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1582 err = put_user(len, optlen);
1588 EXPORT_SYMBOL(compat_ip_getsockopt);