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);
354 if (new_ra == NULL) {
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) != NULL) {
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 /* IPv4 supports cmsg on all imcp errors and some timestamps
437 * Timestamp code paths do not initialize the fields expected by cmsg:
438 * the PKTINFO fields in skb->cb[]. Fill those in here.
440 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
444 struct in_pktinfo *info;
446 if (ee_origin == SO_EE_ORIGIN_ICMP)
449 if (ee_origin == SO_EE_ORIGIN_LOCAL)
452 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
453 * timestamp with egress dev. Not possible for packets without dev
454 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
456 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
460 info = PKTINFO_SKB_CB(skb);
461 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
462 info->ipi_ifindex = skb->dev->ifindex;
467 * Handle MSG_ERRQUEUE
469 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
471 struct sock_exterr_skb *serr;
473 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
475 struct sock_extended_err ee;
476 struct sockaddr_in offender;
481 WARN_ON_ONCE(sk->sk_family == AF_INET6);
484 skb = sock_dequeue_err_skb(sk);
490 msg->msg_flags |= MSG_TRUNC;
493 err = skb_copy_datagram_msg(skb, 0, msg, copied);
497 sock_recv_timestamp(msg, sk, skb);
499 serr = SKB_EXT_ERR(skb);
501 if (sin && serr->port) {
502 sin->sin_family = AF_INET;
503 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
505 sin->sin_port = serr->port;
506 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
507 *addr_len = sizeof(*sin);
510 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
511 sin = &errhdr.offender;
512 memset(sin, 0, sizeof(*sin));
514 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
515 sin->sin_family = AF_INET;
516 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
517 if (inet_sk(sk)->cmsg_flags)
518 ip_cmsg_recv(msg, skb);
521 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
523 /* Now we could try to dump offended packet options */
525 msg->msg_flags |= MSG_ERRQUEUE;
536 * Socket option code for IP. This is the end of the line after any
537 * TCP,UDP etc options on an IP socket.
540 static int do_ip_setsockopt(struct sock *sk, int level,
541 int optname, char __user *optval, unsigned int optlen)
543 struct inet_sock *inet = inet_sk(sk);
555 case IP_MTU_DISCOVER:
557 case IP_ROUTER_ALERT:
564 case IP_MULTICAST_TTL:
565 case IP_MULTICAST_ALL:
566 case IP_MULTICAST_LOOP:
567 case IP_RECVORIGDSTADDR:
569 if (optlen >= sizeof(int)) {
570 if (get_user(val, (int __user *) optval))
572 } else if (optlen >= sizeof(char)) {
575 if (get_user(ucval, (unsigned char __user *) optval))
581 /* If optlen==0, it is equivalent to val == 0 */
583 if (ip_mroute_opt(optname))
584 return ip_mroute_setsockopt(sk, optname, optval, optlen);
592 struct ip_options_rcu *old, *opt = NULL;
596 err = ip_options_get_from_user(sock_net(sk), &opt,
600 old = rcu_dereference_protected(inet->inet_opt,
601 sock_owned_by_user(sk));
603 struct inet_connection_sock *icsk = inet_csk(sk);
604 #if IS_ENABLED(CONFIG_IPV6)
605 if (sk->sk_family == PF_INET ||
606 (!((1 << sk->sk_state) &
607 (TCPF_LISTEN | TCPF_CLOSE)) &&
608 inet->inet_daddr != LOOPBACK4_IPV6)) {
611 icsk->icsk_ext_hdr_len -= old->opt.optlen;
613 icsk->icsk_ext_hdr_len += opt->opt.optlen;
614 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
615 #if IS_ENABLED(CONFIG_IPV6)
619 rcu_assign_pointer(inet->inet_opt, opt);
626 inet->cmsg_flags |= IP_CMSG_PKTINFO;
628 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
632 inet->cmsg_flags |= IP_CMSG_TTL;
634 inet->cmsg_flags &= ~IP_CMSG_TTL;
638 inet->cmsg_flags |= IP_CMSG_TOS;
640 inet->cmsg_flags &= ~IP_CMSG_TOS;
644 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
646 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
650 inet->cmsg_flags |= IP_CMSG_RETOPTS;
652 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
656 inet->cmsg_flags |= IP_CMSG_PASSSEC;
658 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
660 case IP_RECVORIGDSTADDR:
662 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
664 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
668 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
669 inet_inc_convert_csum(sk);
670 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
673 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
674 inet_dec_convert_csum(sk);
675 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
679 case IP_TOS: /* This sets both TOS and Precedence */
680 if (sk->sk_type == SOCK_STREAM) {
681 val &= ~INET_ECN_MASK;
682 val |= inet->tos & INET_ECN_MASK;
684 if (inet->tos != val) {
686 sk->sk_priority = rt_tos2priority(val);
693 if (val != -1 && (val < 1 || val > 255))
698 if (sk->sk_type != SOCK_RAW) {
702 inet->hdrincl = val ? 1 : 0;
705 if (sk->sk_type != SOCK_RAW) {
709 inet->nodefrag = val ? 1 : 0;
711 case IP_MTU_DISCOVER:
712 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
714 inet->pmtudisc = val;
717 inet->recverr = !!val;
719 skb_queue_purge(&sk->sk_error_queue);
721 case IP_MULTICAST_TTL:
722 if (sk->sk_type == SOCK_STREAM)
728 if (val < 0 || val > 255)
732 case IP_MULTICAST_LOOP:
735 inet->mc_loop = !!val;
739 struct net_device *dev = NULL;
742 if (optlen != sizeof(int))
745 ifindex = (__force int)ntohl((__force __be32)val);
752 dev = dev_get_by_index(sock_net(sk), ifindex);
753 err = -EADDRNOTAVAIL;
759 if (sk->sk_bound_dev_if)
762 inet->uc_index = ifindex;
766 case IP_MULTICAST_IF:
768 struct ip_mreqn mreq;
769 struct net_device *dev = NULL;
771 if (sk->sk_type == SOCK_STREAM)
774 * Check the arguments are allowable
777 if (optlen < sizeof(struct in_addr))
781 if (optlen >= sizeof(struct ip_mreqn)) {
782 if (copy_from_user(&mreq, optval, sizeof(mreq)))
785 memset(&mreq, 0, sizeof(mreq));
786 if (optlen >= sizeof(struct ip_mreq)) {
787 if (copy_from_user(&mreq, optval,
788 sizeof(struct ip_mreq)))
790 } else if (optlen >= sizeof(struct in_addr)) {
791 if (copy_from_user(&mreq.imr_address, optval,
792 sizeof(struct in_addr)))
797 if (!mreq.imr_ifindex) {
798 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
804 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
806 mreq.imr_ifindex = dev->ifindex;
808 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
811 err = -EADDRNOTAVAIL;
817 if (sk->sk_bound_dev_if &&
818 mreq.imr_ifindex != sk->sk_bound_dev_if)
821 inet->mc_index = mreq.imr_ifindex;
822 inet->mc_addr = mreq.imr_address.s_addr;
827 case IP_ADD_MEMBERSHIP:
828 case IP_DROP_MEMBERSHIP:
830 struct ip_mreqn mreq;
833 if (inet_sk(sk)->is_icsk)
836 if (optlen < sizeof(struct ip_mreq))
839 if (optlen >= sizeof(struct ip_mreqn)) {
840 if (copy_from_user(&mreq, optval, sizeof(mreq)))
843 memset(&mreq, 0, sizeof(mreq));
844 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
848 if (optname == IP_ADD_MEMBERSHIP)
849 err = ip_mc_join_group(sk, &mreq);
851 err = ip_mc_leave_group(sk, &mreq);
856 struct ip_msfilter *msf;
858 if (optlen < IP_MSFILTER_SIZE(0))
860 if (optlen > sysctl_optmem_max) {
864 msf = kmalloc(optlen, GFP_KERNEL);
870 if (copy_from_user(msf, optval, optlen)) {
874 /* numsrc >= (1G-4) overflow in 32 bits */
875 if (msf->imsf_numsrc >= 0x3ffffffcU ||
876 msf->imsf_numsrc > sysctl_igmp_max_msf) {
881 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
886 err = ip_mc_msfilter(sk, msf, 0);
890 case IP_BLOCK_SOURCE:
891 case IP_UNBLOCK_SOURCE:
892 case IP_ADD_SOURCE_MEMBERSHIP:
893 case IP_DROP_SOURCE_MEMBERSHIP:
895 struct ip_mreq_source mreqs;
898 if (optlen != sizeof(struct ip_mreq_source))
900 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
904 if (optname == IP_BLOCK_SOURCE) {
905 omode = MCAST_EXCLUDE;
907 } else if (optname == IP_UNBLOCK_SOURCE) {
908 omode = MCAST_EXCLUDE;
910 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
911 struct ip_mreqn mreq;
913 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
914 mreq.imr_address.s_addr = mreqs.imr_interface;
915 mreq.imr_ifindex = 0;
916 err = ip_mc_join_group(sk, &mreq);
917 if (err && err != -EADDRINUSE)
919 omode = MCAST_INCLUDE;
921 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
922 omode = MCAST_INCLUDE;
925 err = ip_mc_source(add, omode, sk, &mreqs, 0);
928 case MCAST_JOIN_GROUP:
929 case MCAST_LEAVE_GROUP:
931 struct group_req greq;
932 struct sockaddr_in *psin;
933 struct ip_mreqn mreq;
935 if (optlen < sizeof(struct group_req))
938 if (copy_from_user(&greq, optval, sizeof(greq)))
940 psin = (struct sockaddr_in *)&greq.gr_group;
941 if (psin->sin_family != AF_INET)
943 memset(&mreq, 0, sizeof(mreq));
944 mreq.imr_multiaddr = psin->sin_addr;
945 mreq.imr_ifindex = greq.gr_interface;
947 if (optname == MCAST_JOIN_GROUP)
948 err = ip_mc_join_group(sk, &mreq);
950 err = ip_mc_leave_group(sk, &mreq);
953 case MCAST_JOIN_SOURCE_GROUP:
954 case MCAST_LEAVE_SOURCE_GROUP:
955 case MCAST_BLOCK_SOURCE:
956 case MCAST_UNBLOCK_SOURCE:
958 struct group_source_req greqs;
959 struct ip_mreq_source mreqs;
960 struct sockaddr_in *psin;
963 if (optlen != sizeof(struct group_source_req))
965 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
969 if (greqs.gsr_group.ss_family != AF_INET ||
970 greqs.gsr_source.ss_family != AF_INET) {
971 err = -EADDRNOTAVAIL;
974 psin = (struct sockaddr_in *)&greqs.gsr_group;
975 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
976 psin = (struct sockaddr_in *)&greqs.gsr_source;
977 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
978 mreqs.imr_interface = 0; /* use index for mc_source */
980 if (optname == MCAST_BLOCK_SOURCE) {
981 omode = MCAST_EXCLUDE;
983 } else if (optname == MCAST_UNBLOCK_SOURCE) {
984 omode = MCAST_EXCLUDE;
986 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
987 struct ip_mreqn mreq;
989 psin = (struct sockaddr_in *)&greqs.gsr_group;
990 mreq.imr_multiaddr = psin->sin_addr;
991 mreq.imr_address.s_addr = 0;
992 mreq.imr_ifindex = greqs.gsr_interface;
993 err = ip_mc_join_group(sk, &mreq);
994 if (err && err != -EADDRINUSE)
996 greqs.gsr_interface = mreq.imr_ifindex;
997 omode = MCAST_INCLUDE;
999 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1000 omode = MCAST_INCLUDE;
1003 err = ip_mc_source(add, omode, sk, &mreqs,
1004 greqs.gsr_interface);
1007 case MCAST_MSFILTER:
1009 struct sockaddr_in *psin;
1010 struct ip_msfilter *msf = NULL;
1011 struct group_filter *gsf = NULL;
1012 int msize, i, ifindex;
1014 if (optlen < GROUP_FILTER_SIZE(0))
1016 if (optlen > sysctl_optmem_max) {
1020 gsf = kmalloc(optlen, GFP_KERNEL);
1026 if (copy_from_user(gsf, optval, optlen))
1029 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1030 if (gsf->gf_numsrc >= 0x1ffffff ||
1031 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1035 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1039 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1040 msf = kmalloc(msize, GFP_KERNEL);
1045 ifindex = gsf->gf_interface;
1046 psin = (struct sockaddr_in *)&gsf->gf_group;
1047 if (psin->sin_family != AF_INET) {
1048 err = -EADDRNOTAVAIL;
1051 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1052 msf->imsf_interface = 0;
1053 msf->imsf_fmode = gsf->gf_fmode;
1054 msf->imsf_numsrc = gsf->gf_numsrc;
1055 err = -EADDRNOTAVAIL;
1056 for (i = 0; i < gsf->gf_numsrc; ++i) {
1057 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1059 if (psin->sin_family != AF_INET)
1061 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1066 err = ip_mc_msfilter(sk, msf, ifindex);
1072 case IP_MULTICAST_ALL:
1075 if (val != 0 && val != 1)
1079 case IP_ROUTER_ALERT:
1080 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1086 inet->freebind = !!val;
1089 case IP_IPSEC_POLICY:
1090 case IP_XFRM_POLICY:
1092 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1094 err = xfrm_user_policy(sk, optname, optval, optlen);
1097 case IP_TRANSPARENT:
1098 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1099 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1105 inet->transparent = !!val;
1111 if (val < 0 || val > 255)
1113 inet->min_ttl = val;
1129 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1133 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1134 * destination in skb->cb[] before dst drop.
1135 * This way, receiver doesn't make cache line misses to read rtable.
1137 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1139 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1140 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1143 if (prepare && skb_rtable(skb)) {
1144 pktinfo->ipi_ifindex = inet_iif(skb);
1145 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1147 pktinfo->ipi_ifindex = 0;
1148 pktinfo->ipi_spec_dst.s_addr = 0;
1153 int ip_setsockopt(struct sock *sk, int level,
1154 int optname, char __user *optval, unsigned int optlen)
1158 if (level != SOL_IP)
1159 return -ENOPROTOOPT;
1161 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1162 #ifdef CONFIG_NETFILTER
1163 /* we need to exclude all possible ENOPROTOOPTs except default case */
1164 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1165 optname != IP_IPSEC_POLICY &&
1166 optname != IP_XFRM_POLICY &&
1167 !ip_mroute_opt(optname)) {
1169 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1175 EXPORT_SYMBOL(ip_setsockopt);
1177 #ifdef CONFIG_COMPAT
1178 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1179 char __user *optval, unsigned int optlen)
1183 if (level != SOL_IP)
1184 return -ENOPROTOOPT;
1186 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1187 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1190 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1191 #ifdef CONFIG_NETFILTER
1192 /* we need to exclude all possible ENOPROTOOPTs except default case */
1193 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1194 optname != IP_IPSEC_POLICY &&
1195 optname != IP_XFRM_POLICY &&
1196 !ip_mroute_opt(optname)) {
1198 err = compat_nf_setsockopt(sk, PF_INET, optname,
1205 EXPORT_SYMBOL(compat_ip_setsockopt);
1209 * Get the options. Note for future reference. The GET of IP options gets
1210 * the _received_ ones. The set sets the _sent_ ones.
1213 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1214 char __user *optval, int __user *optlen, unsigned int flags)
1216 struct inet_sock *inet = inet_sk(sk);
1220 if (level != SOL_IP)
1223 if (ip_mroute_opt(optname))
1224 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1226 if (get_user(len, optlen))
1236 unsigned char optbuf[sizeof(struct ip_options)+40];
1237 struct ip_options *opt = (struct ip_options *)optbuf;
1238 struct ip_options_rcu *inet_opt;
1240 inet_opt = rcu_dereference_protected(inet->inet_opt,
1241 sock_owned_by_user(sk));
1244 memcpy(optbuf, &inet_opt->opt,
1245 sizeof(struct ip_options) +
1246 inet_opt->opt.optlen);
1249 if (opt->optlen == 0)
1250 return put_user(0, optlen);
1252 ip_options_undo(opt);
1254 len = min_t(unsigned int, len, opt->optlen);
1255 if (put_user(len, optlen))
1257 if (copy_to_user(optval, opt->__data, len))
1262 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1265 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1268 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1271 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1274 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1277 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1279 case IP_RECVORIGDSTADDR:
1280 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1283 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1289 val = (inet->uc_ttl == -1 ?
1290 sysctl_ip_default_ttl :
1294 val = inet->hdrincl;
1297 val = inet->nodefrag;
1299 case IP_MTU_DISCOVER:
1300 val = inet->pmtudisc;
1304 struct dst_entry *dst;
1306 dst = sk_dst_get(sk);
1318 val = inet->recverr;
1320 case IP_MULTICAST_TTL:
1323 case IP_MULTICAST_LOOP:
1324 val = inet->mc_loop;
1327 val = (__force int)htonl((__u32) inet->uc_index);
1329 case IP_MULTICAST_IF:
1331 struct in_addr addr;
1332 len = min_t(unsigned int, len, sizeof(struct in_addr));
1333 addr.s_addr = inet->mc_addr;
1336 if (put_user(len, optlen))
1338 if (copy_to_user(optval, &addr, len))
1344 struct ip_msfilter msf;
1347 if (len < IP_MSFILTER_SIZE(0)) {
1351 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1355 err = ip_mc_msfget(sk, &msf,
1356 (struct ip_msfilter __user *)optval, optlen);
1360 case MCAST_MSFILTER:
1362 struct group_filter gsf;
1365 if (len < GROUP_FILTER_SIZE(0)) {
1369 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1373 err = ip_mc_gsfget(sk, &gsf,
1374 (struct group_filter __user *)optval,
1379 case IP_MULTICAST_ALL:
1388 if (sk->sk_type != SOCK_STREAM)
1389 return -ENOPROTOOPT;
1391 msg.msg_control = (__force void *) optval;
1392 msg.msg_controllen = len;
1393 msg.msg_flags = flags;
1395 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1396 struct in_pktinfo info;
1398 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1399 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1400 info.ipi_ifindex = inet->mc_index;
1401 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1403 if (inet->cmsg_flags & IP_CMSG_TTL) {
1404 int hlim = inet->mc_ttl;
1405 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1407 if (inet->cmsg_flags & IP_CMSG_TOS) {
1408 int tos = inet->rcv_tos;
1409 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1411 len -= msg.msg_controllen;
1412 return put_user(len, optlen);
1415 val = inet->freebind;
1417 case IP_TRANSPARENT:
1418 val = inet->transparent;
1421 val = inet->min_ttl;
1425 return -ENOPROTOOPT;
1429 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1430 unsigned char ucval = (unsigned char)val;
1432 if (put_user(len, optlen))
1434 if (copy_to_user(optval, &ucval, 1))
1437 len = min_t(unsigned int, sizeof(int), len);
1438 if (put_user(len, optlen))
1440 if (copy_to_user(optval, &val, len))
1446 int ip_getsockopt(struct sock *sk, int level,
1447 int optname, char __user *optval, int __user *optlen)
1451 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1452 #ifdef CONFIG_NETFILTER
1453 /* we need to exclude all possible ENOPROTOOPTs except default case */
1454 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1455 !ip_mroute_opt(optname)) {
1458 if (get_user(len, optlen))
1462 err = nf_getsockopt(sk, PF_INET, optname, optval,
1466 err = put_user(len, optlen);
1472 EXPORT_SYMBOL(ip_getsockopt);
1474 #ifdef CONFIG_COMPAT
1475 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1476 char __user *optval, int __user *optlen)
1480 if (optname == MCAST_MSFILTER)
1481 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1484 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1487 #ifdef CONFIG_NETFILTER
1488 /* we need to exclude all possible ENOPROTOOPTs except default case */
1489 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1490 !ip_mroute_opt(optname)) {
1493 if (get_user(len, optlen))
1497 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1500 err = put_user(len, optlen);
1506 EXPORT_SYMBOL(compat_ip_getsockopt);