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:
594 case IP_BIND_ADDRESS_NO_PORT:
596 case IP_MULTICAST_TTL:
597 case IP_MULTICAST_ALL:
598 case IP_MULTICAST_LOOP:
599 case IP_RECVORIGDSTADDR:
601 if (optlen >= sizeof(int)) {
602 if (get_user(val, (int __user *) optval))
604 } else if (optlen >= sizeof(char)) {
607 if (get_user(ucval, (unsigned char __user *) optval))
613 /* If optlen==0, it is equivalent to val == 0 */
615 if (ip_mroute_opt(optname))
616 return ip_mroute_setsockopt(sk, optname, optval, optlen);
626 struct ip_options_rcu *old, *opt = NULL;
630 err = ip_options_get_from_user(sock_net(sk), &opt,
634 old = rcu_dereference_protected(inet->inet_opt,
635 sock_owned_by_user(sk));
637 struct inet_connection_sock *icsk = inet_csk(sk);
638 #if IS_ENABLED(CONFIG_IPV6)
639 if (sk->sk_family == PF_INET ||
640 (!((1 << sk->sk_state) &
641 (TCPF_LISTEN | TCPF_CLOSE)) &&
642 inet->inet_daddr != LOOPBACK4_IPV6)) {
645 icsk->icsk_ext_hdr_len -= old->opt.optlen;
647 icsk->icsk_ext_hdr_len += opt->opt.optlen;
648 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
649 #if IS_ENABLED(CONFIG_IPV6)
653 rcu_assign_pointer(inet->inet_opt, opt);
660 inet->cmsg_flags |= IP_CMSG_PKTINFO;
662 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
666 inet->cmsg_flags |= IP_CMSG_TTL;
668 inet->cmsg_flags &= ~IP_CMSG_TTL;
672 inet->cmsg_flags |= IP_CMSG_TOS;
674 inet->cmsg_flags &= ~IP_CMSG_TOS;
678 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
680 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
684 inet->cmsg_flags |= IP_CMSG_RETOPTS;
686 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
690 inet->cmsg_flags |= IP_CMSG_PASSSEC;
692 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
694 case IP_RECVORIGDSTADDR:
696 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
698 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
702 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
703 inet_inc_convert_csum(sk);
704 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
707 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
708 inet_dec_convert_csum(sk);
709 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
713 case IP_TOS: /* This sets both TOS and Precedence */
714 if (sk->sk_type == SOCK_STREAM) {
715 val &= ~INET_ECN_MASK;
716 val |= inet->tos & INET_ECN_MASK;
718 if (inet->tos != val) {
720 sk->sk_priority = rt_tos2priority(val);
727 if (val != -1 && (val < 1 || val > 255))
732 if (sk->sk_type != SOCK_RAW) {
736 inet->hdrincl = val ? 1 : 0;
739 if (sk->sk_type != SOCK_RAW) {
743 inet->nodefrag = val ? 1 : 0;
745 case IP_BIND_ADDRESS_NO_PORT:
746 inet->bind_address_no_port = val ? 1 : 0;
748 case IP_MTU_DISCOVER:
749 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
751 inet->pmtudisc = val;
754 inet->recverr = !!val;
756 skb_queue_purge(&sk->sk_error_queue);
758 case IP_MULTICAST_TTL:
759 if (sk->sk_type == SOCK_STREAM)
765 if (val < 0 || val > 255)
769 case IP_MULTICAST_LOOP:
772 inet->mc_loop = !!val;
776 struct net_device *dev = NULL;
779 if (optlen != sizeof(int))
782 ifindex = (__force int)ntohl((__force __be32)val);
789 dev = dev_get_by_index(sock_net(sk), ifindex);
790 err = -EADDRNOTAVAIL;
796 if (sk->sk_bound_dev_if)
799 inet->uc_index = ifindex;
803 case IP_MULTICAST_IF:
805 struct ip_mreqn mreq;
806 struct net_device *dev = NULL;
808 if (sk->sk_type == SOCK_STREAM)
811 * Check the arguments are allowable
814 if (optlen < sizeof(struct in_addr))
818 if (optlen >= sizeof(struct ip_mreqn)) {
819 if (copy_from_user(&mreq, optval, sizeof(mreq)))
822 memset(&mreq, 0, sizeof(mreq));
823 if (optlen >= sizeof(struct ip_mreq)) {
824 if (copy_from_user(&mreq, optval,
825 sizeof(struct ip_mreq)))
827 } else if (optlen >= sizeof(struct in_addr)) {
828 if (copy_from_user(&mreq.imr_address, optval,
829 sizeof(struct in_addr)))
834 if (!mreq.imr_ifindex) {
835 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
841 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
843 mreq.imr_ifindex = dev->ifindex;
845 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
848 err = -EADDRNOTAVAIL;
854 if (sk->sk_bound_dev_if &&
855 mreq.imr_ifindex != sk->sk_bound_dev_if)
858 inet->mc_index = mreq.imr_ifindex;
859 inet->mc_addr = mreq.imr_address.s_addr;
864 case IP_ADD_MEMBERSHIP:
865 case IP_DROP_MEMBERSHIP:
867 struct ip_mreqn mreq;
870 if (inet_sk(sk)->is_icsk)
873 if (optlen < sizeof(struct ip_mreq))
876 if (optlen >= sizeof(struct ip_mreqn)) {
877 if (copy_from_user(&mreq, optval, sizeof(mreq)))
880 memset(&mreq, 0, sizeof(mreq));
881 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
885 if (optname == IP_ADD_MEMBERSHIP)
886 err = ip_mc_join_group(sk, &mreq);
888 err = ip_mc_leave_group(sk, &mreq);
893 struct ip_msfilter *msf;
895 if (optlen < IP_MSFILTER_SIZE(0))
897 if (optlen > sysctl_optmem_max) {
901 msf = kmalloc(optlen, GFP_KERNEL);
907 if (copy_from_user(msf, optval, optlen)) {
911 /* numsrc >= (1G-4) overflow in 32 bits */
912 if (msf->imsf_numsrc >= 0x3ffffffcU ||
913 msf->imsf_numsrc > sysctl_igmp_max_msf) {
918 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
923 err = ip_mc_msfilter(sk, msf, 0);
927 case IP_BLOCK_SOURCE:
928 case IP_UNBLOCK_SOURCE:
929 case IP_ADD_SOURCE_MEMBERSHIP:
930 case IP_DROP_SOURCE_MEMBERSHIP:
932 struct ip_mreq_source mreqs;
935 if (optlen != sizeof(struct ip_mreq_source))
937 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
941 if (optname == IP_BLOCK_SOURCE) {
942 omode = MCAST_EXCLUDE;
944 } else if (optname == IP_UNBLOCK_SOURCE) {
945 omode = MCAST_EXCLUDE;
947 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
948 struct ip_mreqn mreq;
950 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
951 mreq.imr_address.s_addr = mreqs.imr_interface;
952 mreq.imr_ifindex = 0;
953 err = ip_mc_join_group(sk, &mreq);
954 if (err && err != -EADDRINUSE)
956 omode = MCAST_INCLUDE;
958 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
959 omode = MCAST_INCLUDE;
962 err = ip_mc_source(add, omode, sk, &mreqs, 0);
965 case MCAST_JOIN_GROUP:
966 case MCAST_LEAVE_GROUP:
968 struct group_req greq;
969 struct sockaddr_in *psin;
970 struct ip_mreqn mreq;
972 if (optlen < sizeof(struct group_req))
975 if (copy_from_user(&greq, optval, sizeof(greq)))
977 psin = (struct sockaddr_in *)&greq.gr_group;
978 if (psin->sin_family != AF_INET)
980 memset(&mreq, 0, sizeof(mreq));
981 mreq.imr_multiaddr = psin->sin_addr;
982 mreq.imr_ifindex = greq.gr_interface;
984 if (optname == MCAST_JOIN_GROUP)
985 err = ip_mc_join_group(sk, &mreq);
987 err = ip_mc_leave_group(sk, &mreq);
990 case MCAST_JOIN_SOURCE_GROUP:
991 case MCAST_LEAVE_SOURCE_GROUP:
992 case MCAST_BLOCK_SOURCE:
993 case MCAST_UNBLOCK_SOURCE:
995 struct group_source_req greqs;
996 struct ip_mreq_source mreqs;
997 struct sockaddr_in *psin;
1000 if (optlen != sizeof(struct group_source_req))
1002 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1006 if (greqs.gsr_group.ss_family != AF_INET ||
1007 greqs.gsr_source.ss_family != AF_INET) {
1008 err = -EADDRNOTAVAIL;
1011 psin = (struct sockaddr_in *)&greqs.gsr_group;
1012 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1013 psin = (struct sockaddr_in *)&greqs.gsr_source;
1014 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1015 mreqs.imr_interface = 0; /* use index for mc_source */
1017 if (optname == MCAST_BLOCK_SOURCE) {
1018 omode = MCAST_EXCLUDE;
1020 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1021 omode = MCAST_EXCLUDE;
1023 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1024 struct ip_mreqn mreq;
1026 psin = (struct sockaddr_in *)&greqs.gsr_group;
1027 mreq.imr_multiaddr = psin->sin_addr;
1028 mreq.imr_address.s_addr = 0;
1029 mreq.imr_ifindex = greqs.gsr_interface;
1030 err = ip_mc_join_group(sk, &mreq);
1031 if (err && err != -EADDRINUSE)
1033 greqs.gsr_interface = mreq.imr_ifindex;
1034 omode = MCAST_INCLUDE;
1036 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1037 omode = MCAST_INCLUDE;
1040 err = ip_mc_source(add, omode, sk, &mreqs,
1041 greqs.gsr_interface);
1044 case MCAST_MSFILTER:
1046 struct sockaddr_in *psin;
1047 struct ip_msfilter *msf = NULL;
1048 struct group_filter *gsf = NULL;
1049 int msize, i, ifindex;
1051 if (optlen < GROUP_FILTER_SIZE(0))
1053 if (optlen > sysctl_optmem_max) {
1057 gsf = kmalloc(optlen, GFP_KERNEL);
1063 if (copy_from_user(gsf, optval, optlen))
1066 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1067 if (gsf->gf_numsrc >= 0x1ffffff ||
1068 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1072 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1076 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1077 msf = kmalloc(msize, GFP_KERNEL);
1082 ifindex = gsf->gf_interface;
1083 psin = (struct sockaddr_in *)&gsf->gf_group;
1084 if (psin->sin_family != AF_INET) {
1085 err = -EADDRNOTAVAIL;
1088 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1089 msf->imsf_interface = 0;
1090 msf->imsf_fmode = gsf->gf_fmode;
1091 msf->imsf_numsrc = gsf->gf_numsrc;
1092 err = -EADDRNOTAVAIL;
1093 for (i = 0; i < gsf->gf_numsrc; ++i) {
1094 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1096 if (psin->sin_family != AF_INET)
1098 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1103 err = ip_mc_msfilter(sk, msf, ifindex);
1109 case IP_MULTICAST_ALL:
1112 if (val != 0 && val != 1)
1116 case IP_ROUTER_ALERT:
1117 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1123 inet->freebind = !!val;
1126 case IP_IPSEC_POLICY:
1127 case IP_XFRM_POLICY:
1129 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1131 err = xfrm_user_policy(sk, optname, optval, optlen);
1134 case IP_TRANSPARENT:
1135 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1136 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1142 inet->transparent = !!val;
1148 if (val < 0 || val > 255)
1150 inet->min_ttl = val;
1170 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1174 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1175 * destination in skb->cb[] before dst drop.
1176 * This way, receiver doesn't make cache line misses to read rtable.
1178 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1180 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1181 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1184 if (prepare && skb_rtable(skb)) {
1185 pktinfo->ipi_ifindex = inet_iif(skb);
1186 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1188 pktinfo->ipi_ifindex = 0;
1189 pktinfo->ipi_spec_dst.s_addr = 0;
1194 int ip_setsockopt(struct sock *sk, int level,
1195 int optname, char __user *optval, unsigned int optlen)
1199 if (level != SOL_IP)
1200 return -ENOPROTOOPT;
1202 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1203 #ifdef CONFIG_NETFILTER
1204 /* we need to exclude all possible ENOPROTOOPTs except default case */
1205 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1206 optname != IP_IPSEC_POLICY &&
1207 optname != IP_XFRM_POLICY &&
1208 !ip_mroute_opt(optname)) {
1210 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1216 EXPORT_SYMBOL(ip_setsockopt);
1218 #ifdef CONFIG_COMPAT
1219 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1220 char __user *optval, unsigned int optlen)
1224 if (level != SOL_IP)
1225 return -ENOPROTOOPT;
1227 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1228 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1231 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1232 #ifdef CONFIG_NETFILTER
1233 /* we need to exclude all possible ENOPROTOOPTs except default case */
1234 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1235 optname != IP_IPSEC_POLICY &&
1236 optname != IP_XFRM_POLICY &&
1237 !ip_mroute_opt(optname)) {
1239 err = compat_nf_setsockopt(sk, PF_INET, optname,
1246 EXPORT_SYMBOL(compat_ip_setsockopt);
1250 * Get the options. Note for future reference. The GET of IP options gets
1251 * the _received_ ones. The set sets the _sent_ ones.
1254 static bool getsockopt_needs_rtnl(int optname)
1258 case MCAST_MSFILTER:
1264 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1265 char __user *optval, int __user *optlen, unsigned int flags)
1267 struct inet_sock *inet = inet_sk(sk);
1268 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1272 if (level != SOL_IP)
1275 if (ip_mroute_opt(optname))
1276 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1278 if (get_user(len, optlen))
1290 unsigned char optbuf[sizeof(struct ip_options)+40];
1291 struct ip_options *opt = (struct ip_options *)optbuf;
1292 struct ip_options_rcu *inet_opt;
1294 inet_opt = rcu_dereference_protected(inet->inet_opt,
1295 sock_owned_by_user(sk));
1298 memcpy(optbuf, &inet_opt->opt,
1299 sizeof(struct ip_options) +
1300 inet_opt->opt.optlen);
1303 if (opt->optlen == 0)
1304 return put_user(0, optlen);
1306 ip_options_undo(opt);
1308 len = min_t(unsigned int, len, opt->optlen);
1309 if (put_user(len, optlen))
1311 if (copy_to_user(optval, opt->__data, len))
1316 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1319 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1322 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1325 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1328 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1331 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1333 case IP_RECVORIGDSTADDR:
1334 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1337 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1343 val = (inet->uc_ttl == -1 ?
1344 sysctl_ip_default_ttl :
1348 val = inet->hdrincl;
1351 val = inet->nodefrag;
1353 case IP_BIND_ADDRESS_NO_PORT:
1354 val = inet->bind_address_no_port;
1356 case IP_MTU_DISCOVER:
1357 val = inet->pmtudisc;
1361 struct dst_entry *dst;
1363 dst = sk_dst_get(sk);
1375 val = inet->recverr;
1377 case IP_MULTICAST_TTL:
1380 case IP_MULTICAST_LOOP:
1381 val = inet->mc_loop;
1384 val = (__force int)htonl((__u32) inet->uc_index);
1386 case IP_MULTICAST_IF:
1388 struct in_addr addr;
1389 len = min_t(unsigned int, len, sizeof(struct in_addr));
1390 addr.s_addr = inet->mc_addr;
1393 if (put_user(len, optlen))
1395 if (copy_to_user(optval, &addr, len))
1401 struct ip_msfilter msf;
1403 if (len < IP_MSFILTER_SIZE(0)) {
1407 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1411 err = ip_mc_msfget(sk, &msf,
1412 (struct ip_msfilter __user *)optval, optlen);
1415 case MCAST_MSFILTER:
1417 struct group_filter gsf;
1419 if (len < GROUP_FILTER_SIZE(0)) {
1423 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1427 err = ip_mc_gsfget(sk, &gsf,
1428 (struct group_filter __user *)optval,
1432 case IP_MULTICAST_ALL:
1441 if (sk->sk_type != SOCK_STREAM)
1442 return -ENOPROTOOPT;
1444 msg.msg_control = (__force void *) optval;
1445 msg.msg_controllen = len;
1446 msg.msg_flags = flags;
1448 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1449 struct in_pktinfo info;
1451 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1452 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1453 info.ipi_ifindex = inet->mc_index;
1454 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1456 if (inet->cmsg_flags & IP_CMSG_TTL) {
1457 int hlim = inet->mc_ttl;
1458 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1460 if (inet->cmsg_flags & IP_CMSG_TOS) {
1461 int tos = inet->rcv_tos;
1462 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1464 len -= msg.msg_controllen;
1465 return put_user(len, optlen);
1468 val = inet->freebind;
1470 case IP_TRANSPARENT:
1471 val = inet->transparent;
1474 val = inet->min_ttl;
1478 return -ENOPROTOOPT;
1482 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1483 unsigned char ucval = (unsigned char)val;
1485 if (put_user(len, optlen))
1487 if (copy_to_user(optval, &ucval, 1))
1490 len = min_t(unsigned int, sizeof(int), len);
1491 if (put_user(len, optlen))
1493 if (copy_to_user(optval, &val, len))
1505 int ip_getsockopt(struct sock *sk, int level,
1506 int optname, char __user *optval, int __user *optlen)
1510 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1511 #ifdef CONFIG_NETFILTER
1512 /* we need to exclude all possible ENOPROTOOPTs except default case */
1513 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1514 !ip_mroute_opt(optname)) {
1517 if (get_user(len, optlen))
1521 err = nf_getsockopt(sk, PF_INET, optname, optval,
1525 err = put_user(len, optlen);
1531 EXPORT_SYMBOL(ip_getsockopt);
1533 #ifdef CONFIG_COMPAT
1534 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1535 char __user *optval, int __user *optlen)
1539 if (optname == MCAST_MSFILTER)
1540 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1543 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
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))
1556 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1559 err = put_user(len, optlen);
1565 EXPORT_SYMBOL(compat_ip_getsockopt);