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,
101 int tlen, int offset)
103 __wsum csum = skb->csum;
105 if (skb->ip_summed != CHECKSUM_COMPLETE)
109 int tend_off = skb_transport_offset(skb) + tlen;
110 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
113 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
116 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
122 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
126 err = security_secid_to_secctx(secid, &secdata, &seclen);
130 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
131 security_release_secctx(secdata, seclen);
134 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
136 __be16 _ports[2], *ports;
137 struct sockaddr_in sin;
139 /* All current transport protocols have the port numbers in the
140 * first four bytes of the transport header and this function is
141 * written with this assumption in mind.
143 ports = skb_header_pointer(skb, skb_transport_offset(skb),
144 sizeof(_ports), &_ports);
148 sin.sin_family = AF_INET;
149 sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
150 sin.sin_port = ports[1];
151 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
153 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
156 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
157 int tlen, int offset)
159 struct inet_sock *inet = inet_sk(skb->sk);
160 unsigned int flags = inet->cmsg_flags;
162 /* Ordered by supposed usage frequency */
163 if (flags & IP_CMSG_PKTINFO) {
164 ip_cmsg_recv_pktinfo(msg, skb);
166 flags &= ~IP_CMSG_PKTINFO;
171 if (flags & IP_CMSG_TTL) {
172 ip_cmsg_recv_ttl(msg, skb);
174 flags &= ~IP_CMSG_TTL;
179 if (flags & IP_CMSG_TOS) {
180 ip_cmsg_recv_tos(msg, skb);
182 flags &= ~IP_CMSG_TOS;
187 if (flags & IP_CMSG_RECVOPTS) {
188 ip_cmsg_recv_opts(msg, skb);
190 flags &= ~IP_CMSG_RECVOPTS;
195 if (flags & IP_CMSG_RETOPTS) {
196 ip_cmsg_recv_retopts(msg, skb);
198 flags &= ~IP_CMSG_RETOPTS;
203 if (flags & IP_CMSG_PASSSEC) {
204 ip_cmsg_recv_security(msg, skb);
206 flags &= ~IP_CMSG_PASSSEC;
211 if (flags & IP_CMSG_ORIGDSTADDR) {
212 ip_cmsg_recv_dstaddr(msg, skb);
214 flags &= ~IP_CMSG_ORIGDSTADDR;
219 if (flags & IP_CMSG_CHECKSUM)
220 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
222 EXPORT_SYMBOL(ip_cmsg_recv_offset);
224 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
228 struct cmsghdr *cmsg;
229 struct net *net = sock_net(sk);
231 for_each_cmsghdr(cmsg, msg) {
232 if (!CMSG_OK(msg, cmsg))
234 #if IS_ENABLED(CONFIG_IPV6)
236 cmsg->cmsg_level == SOL_IPV6 &&
237 cmsg->cmsg_type == IPV6_PKTINFO) {
238 struct in6_pktinfo *src_info;
240 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
242 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
243 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
245 if (src_info->ipi6_ifindex)
246 ipc->oif = src_info->ipi6_ifindex;
247 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
251 if (cmsg->cmsg_level == SOL_SOCKET) {
252 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
258 if (cmsg->cmsg_level != SOL_IP)
260 switch (cmsg->cmsg_type) {
262 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
264 /* Our caller is responsible for freeing ipc->opt */
265 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
266 err < 40 ? err : 40);
272 struct in_pktinfo *info;
273 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
275 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
276 if (info->ipi_ifindex)
277 ipc->oif = info->ipi_ifindex;
278 ipc->addr = info->ipi_spec_dst.s_addr;
282 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
284 val = *(int *)CMSG_DATA(cmsg);
285 if (val < 1 || val > 255)
290 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
291 val = *(int *)CMSG_DATA(cmsg);
292 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
293 val = *(u8 *)CMSG_DATA(cmsg);
296 if (val < 0 || val > 255)
299 ipc->priority = rt_tos2priority(ipc->tos);
310 /* Special input handler for packets caught by router alert option.
311 They are selected only by protocol field, and then processed likely
312 local ones; but only if someone wants them! Otherwise, router
313 not running rsvpd will kill RSVP.
315 It is user level problem, what it will make with them.
316 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
317 but receiver should be enough clever f.e. to forward mtrace requests,
318 sent to multicast group to reach destination designated router.
320 struct ip_ra_chain __rcu *ip_ra_chain;
321 static DEFINE_SPINLOCK(ip_ra_lock);
324 static void ip_ra_destroy_rcu(struct rcu_head *head)
326 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
328 sock_put(ra->saved_sk);
332 int ip_ra_control(struct sock *sk, unsigned char on,
333 void (*destructor)(struct sock *))
335 struct ip_ra_chain *ra, *new_ra;
336 struct ip_ra_chain __rcu **rap;
338 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
341 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
343 spin_lock_bh(&ip_ra_lock);
344 for (rap = &ip_ra_chain;
345 (ra = rcu_dereference_protected(*rap,
346 lockdep_is_held(&ip_ra_lock))) != NULL;
350 spin_unlock_bh(&ip_ra_lock);
354 /* dont let ip_call_ra_chain() use sk again */
356 RCU_INIT_POINTER(*rap, ra->next);
357 spin_unlock_bh(&ip_ra_lock);
362 * Delay sock_put(sk) and kfree(ra) after one rcu grace
363 * period. This guarantee ip_call_ra_chain() dont need
364 * to mess with socket refcounts.
367 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
372 spin_unlock_bh(&ip_ra_lock);
376 new_ra->destructor = destructor;
378 RCU_INIT_POINTER(new_ra->next, ra);
379 rcu_assign_pointer(*rap, new_ra);
381 spin_unlock_bh(&ip_ra_lock);
386 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
387 __be16 port, u32 info, u8 *payload)
389 struct sock_exterr_skb *serr;
391 skb = skb_clone(skb, GFP_ATOMIC);
395 serr = SKB_EXT_ERR(skb);
396 serr->ee.ee_errno = err;
397 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
398 serr->ee.ee_type = icmp_hdr(skb)->type;
399 serr->ee.ee_code = icmp_hdr(skb)->code;
401 serr->ee.ee_info = info;
402 serr->ee.ee_data = 0;
403 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
404 skb_network_header(skb);
407 if (skb_pull(skb, payload - skb->data)) {
408 skb_reset_transport_header(skb);
409 if (sock_queue_err_skb(sk, skb) == 0)
415 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
417 struct inet_sock *inet = inet_sk(sk);
418 struct sock_exterr_skb *serr;
425 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
429 skb_put(skb, sizeof(struct iphdr));
430 skb_reset_network_header(skb);
434 serr = SKB_EXT_ERR(skb);
435 serr->ee.ee_errno = err;
436 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
437 serr->ee.ee_type = 0;
438 serr->ee.ee_code = 0;
440 serr->ee.ee_info = info;
441 serr->ee.ee_data = 0;
442 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
445 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
446 skb_reset_transport_header(skb);
448 if (sock_queue_err_skb(sk, skb))
452 /* For some errors we have valid addr_offset even with zero payload and
453 * zero port. Also, addr_offset should be supported if port is set.
455 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
457 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
458 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
461 /* IPv4 supports cmsg on all imcp errors and some timestamps
463 * Timestamp code paths do not initialize the fields expected by cmsg:
464 * the PKTINFO fields in skb->cb[]. Fill those in here.
466 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
470 struct in_pktinfo *info;
472 if (ee_origin == SO_EE_ORIGIN_ICMP)
475 if (ee_origin == SO_EE_ORIGIN_LOCAL)
478 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
479 * timestamp with egress dev. Not possible for packets without iif
480 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
482 info = PKTINFO_SKB_CB(skb);
483 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
487 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
492 * Handle MSG_ERRQUEUE
494 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
496 struct sock_exterr_skb *serr;
498 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
500 struct sock_extended_err ee;
501 struct sockaddr_in offender;
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;
824 if (sk->sk_type == SOCK_STREAM)
827 * Check the arguments are allowable
830 if (optlen < sizeof(struct in_addr))
834 if (optlen >= sizeof(struct ip_mreqn)) {
835 if (copy_from_user(&mreq, optval, sizeof(mreq)))
838 memset(&mreq, 0, sizeof(mreq));
839 if (optlen >= sizeof(struct ip_mreq)) {
840 if (copy_from_user(&mreq, optval,
841 sizeof(struct ip_mreq)))
843 } else if (optlen >= sizeof(struct in_addr)) {
844 if (copy_from_user(&mreq.imr_address, optval,
845 sizeof(struct in_addr)))
850 if (!mreq.imr_ifindex) {
851 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
857 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
859 mreq.imr_ifindex = dev->ifindex;
861 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
864 err = -EADDRNOTAVAIL;
868 midx = l3mdev_master_ifindex(dev);
873 if (sk->sk_bound_dev_if &&
874 mreq.imr_ifindex != sk->sk_bound_dev_if &&
875 (!midx || midx != sk->sk_bound_dev_if))
878 inet->mc_index = mreq.imr_ifindex;
879 inet->mc_addr = mreq.imr_address.s_addr;
884 case IP_ADD_MEMBERSHIP:
885 case IP_DROP_MEMBERSHIP:
887 struct ip_mreqn mreq;
890 if (inet_sk(sk)->is_icsk)
893 if (optlen < sizeof(struct ip_mreq))
896 if (optlen >= sizeof(struct ip_mreqn)) {
897 if (copy_from_user(&mreq, optval, sizeof(mreq)))
900 memset(&mreq, 0, sizeof(mreq));
901 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
905 if (optname == IP_ADD_MEMBERSHIP)
906 err = ip_mc_join_group(sk, &mreq);
908 err = ip_mc_leave_group(sk, &mreq);
913 struct ip_msfilter *msf;
915 if (optlen < IP_MSFILTER_SIZE(0))
917 if (optlen > sysctl_optmem_max) {
921 msf = kmalloc(optlen, GFP_KERNEL);
927 if (copy_from_user(msf, optval, optlen)) {
931 /* numsrc >= (1G-4) overflow in 32 bits */
932 if (msf->imsf_numsrc >= 0x3ffffffcU ||
933 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
938 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
943 err = ip_mc_msfilter(sk, msf, 0);
947 case IP_BLOCK_SOURCE:
948 case IP_UNBLOCK_SOURCE:
949 case IP_ADD_SOURCE_MEMBERSHIP:
950 case IP_DROP_SOURCE_MEMBERSHIP:
952 struct ip_mreq_source mreqs;
955 if (optlen != sizeof(struct ip_mreq_source))
957 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
961 if (optname == IP_BLOCK_SOURCE) {
962 omode = MCAST_EXCLUDE;
964 } else if (optname == IP_UNBLOCK_SOURCE) {
965 omode = MCAST_EXCLUDE;
967 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
968 struct ip_mreqn mreq;
970 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
971 mreq.imr_address.s_addr = mreqs.imr_interface;
972 mreq.imr_ifindex = 0;
973 err = ip_mc_join_group(sk, &mreq);
974 if (err && err != -EADDRINUSE)
976 omode = MCAST_INCLUDE;
978 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
979 omode = MCAST_INCLUDE;
982 err = ip_mc_source(add, omode, sk, &mreqs, 0);
985 case MCAST_JOIN_GROUP:
986 case MCAST_LEAVE_GROUP:
988 struct group_req greq;
989 struct sockaddr_in *psin;
990 struct ip_mreqn mreq;
992 if (optlen < sizeof(struct group_req))
995 if (copy_from_user(&greq, optval, sizeof(greq)))
997 psin = (struct sockaddr_in *)&greq.gr_group;
998 if (psin->sin_family != AF_INET)
1000 memset(&mreq, 0, sizeof(mreq));
1001 mreq.imr_multiaddr = psin->sin_addr;
1002 mreq.imr_ifindex = greq.gr_interface;
1004 if (optname == MCAST_JOIN_GROUP)
1005 err = ip_mc_join_group(sk, &mreq);
1007 err = ip_mc_leave_group(sk, &mreq);
1010 case MCAST_JOIN_SOURCE_GROUP:
1011 case MCAST_LEAVE_SOURCE_GROUP:
1012 case MCAST_BLOCK_SOURCE:
1013 case MCAST_UNBLOCK_SOURCE:
1015 struct group_source_req greqs;
1016 struct ip_mreq_source mreqs;
1017 struct sockaddr_in *psin;
1020 if (optlen != sizeof(struct group_source_req))
1022 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1026 if (greqs.gsr_group.ss_family != AF_INET ||
1027 greqs.gsr_source.ss_family != AF_INET) {
1028 err = -EADDRNOTAVAIL;
1031 psin = (struct sockaddr_in *)&greqs.gsr_group;
1032 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1033 psin = (struct sockaddr_in *)&greqs.gsr_source;
1034 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1035 mreqs.imr_interface = 0; /* use index for mc_source */
1037 if (optname == MCAST_BLOCK_SOURCE) {
1038 omode = MCAST_EXCLUDE;
1040 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1041 omode = MCAST_EXCLUDE;
1043 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1044 struct ip_mreqn mreq;
1046 psin = (struct sockaddr_in *)&greqs.gsr_group;
1047 mreq.imr_multiaddr = psin->sin_addr;
1048 mreq.imr_address.s_addr = 0;
1049 mreq.imr_ifindex = greqs.gsr_interface;
1050 err = ip_mc_join_group(sk, &mreq);
1051 if (err && err != -EADDRINUSE)
1053 greqs.gsr_interface = mreq.imr_ifindex;
1054 omode = MCAST_INCLUDE;
1056 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1057 omode = MCAST_INCLUDE;
1060 err = ip_mc_source(add, omode, sk, &mreqs,
1061 greqs.gsr_interface);
1064 case MCAST_MSFILTER:
1066 struct sockaddr_in *psin;
1067 struct ip_msfilter *msf = NULL;
1068 struct group_filter *gsf = NULL;
1069 int msize, i, ifindex;
1071 if (optlen < GROUP_FILTER_SIZE(0))
1073 if (optlen > sysctl_optmem_max) {
1077 gsf = kmalloc(optlen, GFP_KERNEL);
1083 if (copy_from_user(gsf, optval, optlen))
1086 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1087 if (gsf->gf_numsrc >= 0x1ffffff ||
1088 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1092 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1096 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1097 msf = kmalloc(msize, GFP_KERNEL);
1102 ifindex = gsf->gf_interface;
1103 psin = (struct sockaddr_in *)&gsf->gf_group;
1104 if (psin->sin_family != AF_INET) {
1105 err = -EADDRNOTAVAIL;
1108 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1109 msf->imsf_interface = 0;
1110 msf->imsf_fmode = gsf->gf_fmode;
1111 msf->imsf_numsrc = gsf->gf_numsrc;
1112 err = -EADDRNOTAVAIL;
1113 for (i = 0; i < gsf->gf_numsrc; ++i) {
1114 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1116 if (psin->sin_family != AF_INET)
1118 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1123 err = ip_mc_msfilter(sk, msf, ifindex);
1129 case IP_MULTICAST_ALL:
1132 if (val != 0 && val != 1)
1136 case IP_ROUTER_ALERT:
1137 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1143 inet->freebind = !!val;
1146 case IP_IPSEC_POLICY:
1147 case IP_XFRM_POLICY:
1149 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1151 err = xfrm_user_policy(sk, optname, optval, optlen);
1154 case IP_TRANSPARENT:
1155 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1156 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1162 inet->transparent = !!val;
1168 if (val < 0 || val > 255)
1170 inet->min_ttl = val;
1190 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1194 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1195 * destination in skb->cb[] before dst drop.
1196 * This way, receiver doesn't make cache line misses to read rtable.
1198 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1200 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1201 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1204 if (prepare && skb_rtable(skb)) {
1205 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1206 * which has interface index (iif) as the first member of the
1207 * underlying inet{6}_skb_parm struct. This code then overlays
1208 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1209 * element so the iif is picked up from the prior IPCB. If iif
1210 * is the loopback interface, then return the sending interface
1211 * (e.g., process binds socket to eth0 for Tx which is
1212 * redirected to loopback in the rtable/dst).
1214 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1215 pktinfo->ipi_ifindex = inet_iif(skb);
1217 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1219 pktinfo->ipi_ifindex = 0;
1220 pktinfo->ipi_spec_dst.s_addr = 0;
1222 /* We need to keep the dst for __ip_options_echo()
1223 * We could restrict the test to opt.ts_needtime || opt.srr,
1224 * but the following is good enough as IP options are not often used.
1226 if (unlikely(IPCB(skb)->opt.optlen))
1232 int ip_setsockopt(struct sock *sk, int level,
1233 int optname, char __user *optval, unsigned int optlen)
1237 if (level != SOL_IP)
1238 return -ENOPROTOOPT;
1240 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1241 #ifdef CONFIG_NETFILTER
1242 /* we need to exclude all possible ENOPROTOOPTs except default case */
1243 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1244 optname != IP_IPSEC_POLICY &&
1245 optname != IP_XFRM_POLICY &&
1246 !ip_mroute_opt(optname))
1247 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1251 EXPORT_SYMBOL(ip_setsockopt);
1253 #ifdef CONFIG_COMPAT
1254 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1255 char __user *optval, unsigned int optlen)
1259 if (level != SOL_IP)
1260 return -ENOPROTOOPT;
1262 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1263 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1266 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1267 #ifdef CONFIG_NETFILTER
1268 /* we need to exclude all possible ENOPROTOOPTs except default case */
1269 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1270 optname != IP_IPSEC_POLICY &&
1271 optname != IP_XFRM_POLICY &&
1272 !ip_mroute_opt(optname))
1273 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1278 EXPORT_SYMBOL(compat_ip_setsockopt);
1282 * Get the options. Note for future reference. The GET of IP options gets
1283 * the _received_ ones. The set sets the _sent_ ones.
1286 static bool getsockopt_needs_rtnl(int optname)
1290 case MCAST_MSFILTER:
1296 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1297 char __user *optval, int __user *optlen, unsigned int flags)
1299 struct inet_sock *inet = inet_sk(sk);
1300 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1304 if (level != SOL_IP)
1307 if (ip_mroute_opt(optname))
1308 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1310 if (get_user(len, optlen))
1322 unsigned char optbuf[sizeof(struct ip_options)+40];
1323 struct ip_options *opt = (struct ip_options *)optbuf;
1324 struct ip_options_rcu *inet_opt;
1326 inet_opt = rcu_dereference_protected(inet->inet_opt,
1327 lockdep_sock_is_held(sk));
1330 memcpy(optbuf, &inet_opt->opt,
1331 sizeof(struct ip_options) +
1332 inet_opt->opt.optlen);
1335 if (opt->optlen == 0)
1336 return put_user(0, optlen);
1338 ip_options_undo(opt);
1340 len = min_t(unsigned int, len, opt->optlen);
1341 if (put_user(len, optlen))
1343 if (copy_to_user(optval, opt->__data, len))
1348 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1351 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1354 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1357 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1360 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1363 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1365 case IP_RECVORIGDSTADDR:
1366 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1369 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1376 struct net *net = sock_net(sk);
1377 val = (inet->uc_ttl == -1 ?
1378 net->ipv4.sysctl_ip_default_ttl :
1383 val = inet->hdrincl;
1386 val = inet->nodefrag;
1388 case IP_BIND_ADDRESS_NO_PORT:
1389 val = inet->bind_address_no_port;
1391 case IP_MTU_DISCOVER:
1392 val = inet->pmtudisc;
1396 struct dst_entry *dst;
1398 dst = sk_dst_get(sk);
1410 val = inet->recverr;
1412 case IP_MULTICAST_TTL:
1415 case IP_MULTICAST_LOOP:
1416 val = inet->mc_loop;
1419 val = (__force int)htonl((__u32) inet->uc_index);
1421 case IP_MULTICAST_IF:
1423 struct in_addr addr;
1424 len = min_t(unsigned int, len, sizeof(struct in_addr));
1425 addr.s_addr = inet->mc_addr;
1428 if (put_user(len, optlen))
1430 if (copy_to_user(optval, &addr, len))
1436 struct ip_msfilter msf;
1438 if (len < IP_MSFILTER_SIZE(0)) {
1442 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1446 err = ip_mc_msfget(sk, &msf,
1447 (struct ip_msfilter __user *)optval, optlen);
1450 case MCAST_MSFILTER:
1452 struct group_filter gsf;
1454 if (len < GROUP_FILTER_SIZE(0)) {
1458 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1462 err = ip_mc_gsfget(sk, &gsf,
1463 (struct group_filter __user *)optval,
1467 case IP_MULTICAST_ALL:
1476 if (sk->sk_type != SOCK_STREAM)
1477 return -ENOPROTOOPT;
1479 msg.msg_control = (__force void *) optval;
1480 msg.msg_controllen = len;
1481 msg.msg_flags = flags;
1483 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1484 struct in_pktinfo info;
1486 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1487 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1488 info.ipi_ifindex = inet->mc_index;
1489 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1491 if (inet->cmsg_flags & IP_CMSG_TTL) {
1492 int hlim = inet->mc_ttl;
1493 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1495 if (inet->cmsg_flags & IP_CMSG_TOS) {
1496 int tos = inet->rcv_tos;
1497 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1499 len -= msg.msg_controllen;
1500 return put_user(len, optlen);
1503 val = inet->freebind;
1505 case IP_TRANSPARENT:
1506 val = inet->transparent;
1509 val = inet->min_ttl;
1513 return -ENOPROTOOPT;
1517 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1518 unsigned char ucval = (unsigned char)val;
1520 if (put_user(len, optlen))
1522 if (copy_to_user(optval, &ucval, 1))
1525 len = min_t(unsigned int, sizeof(int), len);
1526 if (put_user(len, optlen))
1528 if (copy_to_user(optval, &val, len))
1540 int ip_getsockopt(struct sock *sk, int level,
1541 int optname, char __user *optval, int __user *optlen)
1545 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1546 #ifdef CONFIG_NETFILTER
1547 /* we need to exclude all possible ENOPROTOOPTs except default case */
1548 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1549 !ip_mroute_opt(optname)) {
1552 if (get_user(len, optlen))
1555 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1557 err = put_user(len, optlen);
1563 EXPORT_SYMBOL(ip_getsockopt);
1565 #ifdef CONFIG_COMPAT
1566 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1567 char __user *optval, int __user *optlen)
1571 if (optname == MCAST_MSFILTER)
1572 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1575 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1578 #ifdef CONFIG_NETFILTER
1579 /* we need to exclude all possible ENOPROTOOPTs except default case */
1580 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1581 !ip_mroute_opt(optname)) {
1584 if (get_user(len, optlen))
1587 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1589 err = put_user(len, optlen);
1595 EXPORT_SYMBOL(compat_ip_getsockopt);