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 static bool ipv4_pktinfo_prepare_errqueue(const struct sock *sk,
436 const struct sk_buff *skb,
439 struct in_pktinfo *info = PKTINFO_SKB_CB(skb);
441 if ((ee_origin != SO_EE_ORIGIN_TIMESTAMPING) ||
442 (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
446 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
447 info->ipi_ifindex = skb->dev->ifindex;
452 * Handle MSG_ERRQUEUE
454 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
456 struct sock_exterr_skb *serr;
458 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
460 struct sock_extended_err ee;
461 struct sockaddr_in offender;
466 WARN_ON_ONCE(sk->sk_family == AF_INET6);
469 skb = sock_dequeue_err_skb(sk);
475 msg->msg_flags |= MSG_TRUNC;
478 err = skb_copy_datagram_msg(skb, 0, msg, copied);
482 sock_recv_timestamp(msg, sk, skb);
484 serr = SKB_EXT_ERR(skb);
487 sin->sin_family = AF_INET;
488 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
490 sin->sin_port = serr->port;
491 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
492 *addr_len = sizeof(*sin);
495 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
496 sin = &errhdr.offender;
497 memset(sin, 0, sizeof(*sin));
499 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
500 ipv4_pktinfo_prepare_errqueue(sk, skb, serr->ee.ee_origin)) {
501 sin->sin_family = AF_INET;
502 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
503 if (inet_sk(sk)->cmsg_flags)
504 ip_cmsg_recv(msg, skb);
507 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
509 /* Now we could try to dump offended packet options */
511 msg->msg_flags |= MSG_ERRQUEUE;
522 * Socket option code for IP. This is the end of the line after any
523 * TCP,UDP etc options on an IP socket.
526 static int do_ip_setsockopt(struct sock *sk, int level,
527 int optname, char __user *optval, unsigned int optlen)
529 struct inet_sock *inet = inet_sk(sk);
541 case IP_MTU_DISCOVER:
543 case IP_ROUTER_ALERT:
550 case IP_MULTICAST_TTL:
551 case IP_MULTICAST_ALL:
552 case IP_MULTICAST_LOOP:
553 case IP_RECVORIGDSTADDR:
555 if (optlen >= sizeof(int)) {
556 if (get_user(val, (int __user *) optval))
558 } else if (optlen >= sizeof(char)) {
561 if (get_user(ucval, (unsigned char __user *) optval))
567 /* If optlen==0, it is equivalent to val == 0 */
569 if (ip_mroute_opt(optname))
570 return ip_mroute_setsockopt(sk, optname, optval, optlen);
578 struct ip_options_rcu *old, *opt = NULL;
582 err = ip_options_get_from_user(sock_net(sk), &opt,
586 old = rcu_dereference_protected(inet->inet_opt,
587 sock_owned_by_user(sk));
589 struct inet_connection_sock *icsk = inet_csk(sk);
590 #if IS_ENABLED(CONFIG_IPV6)
591 if (sk->sk_family == PF_INET ||
592 (!((1 << sk->sk_state) &
593 (TCPF_LISTEN | TCPF_CLOSE)) &&
594 inet->inet_daddr != LOOPBACK4_IPV6)) {
597 icsk->icsk_ext_hdr_len -= old->opt.optlen;
599 icsk->icsk_ext_hdr_len += opt->opt.optlen;
600 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
601 #if IS_ENABLED(CONFIG_IPV6)
605 rcu_assign_pointer(inet->inet_opt, opt);
612 inet->cmsg_flags |= IP_CMSG_PKTINFO;
614 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
618 inet->cmsg_flags |= IP_CMSG_TTL;
620 inet->cmsg_flags &= ~IP_CMSG_TTL;
624 inet->cmsg_flags |= IP_CMSG_TOS;
626 inet->cmsg_flags &= ~IP_CMSG_TOS;
630 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
632 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
636 inet->cmsg_flags |= IP_CMSG_RETOPTS;
638 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
642 inet->cmsg_flags |= IP_CMSG_PASSSEC;
644 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
646 case IP_RECVORIGDSTADDR:
648 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
650 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
654 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
655 inet_inc_convert_csum(sk);
656 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
659 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
660 inet_dec_convert_csum(sk);
661 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
665 case IP_TOS: /* This sets both TOS and Precedence */
666 if (sk->sk_type == SOCK_STREAM) {
667 val &= ~INET_ECN_MASK;
668 val |= inet->tos & INET_ECN_MASK;
670 if (inet->tos != val) {
672 sk->sk_priority = rt_tos2priority(val);
679 if (val != -1 && (val < 1 || val > 255))
684 if (sk->sk_type != SOCK_RAW) {
688 inet->hdrincl = val ? 1 : 0;
691 if (sk->sk_type != SOCK_RAW) {
695 inet->nodefrag = val ? 1 : 0;
697 case IP_MTU_DISCOVER:
698 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
700 inet->pmtudisc = val;
703 inet->recverr = !!val;
705 skb_queue_purge(&sk->sk_error_queue);
707 case IP_MULTICAST_TTL:
708 if (sk->sk_type == SOCK_STREAM)
714 if (val < 0 || val > 255)
718 case IP_MULTICAST_LOOP:
721 inet->mc_loop = !!val;
725 struct net_device *dev = NULL;
728 if (optlen != sizeof(int))
731 ifindex = (__force int)ntohl((__force __be32)val);
738 dev = dev_get_by_index(sock_net(sk), ifindex);
739 err = -EADDRNOTAVAIL;
745 if (sk->sk_bound_dev_if)
748 inet->uc_index = ifindex;
752 case IP_MULTICAST_IF:
754 struct ip_mreqn mreq;
755 struct net_device *dev = NULL;
757 if (sk->sk_type == SOCK_STREAM)
760 * Check the arguments are allowable
763 if (optlen < sizeof(struct in_addr))
767 if (optlen >= sizeof(struct ip_mreqn)) {
768 if (copy_from_user(&mreq, optval, sizeof(mreq)))
771 memset(&mreq, 0, sizeof(mreq));
772 if (optlen >= sizeof(struct ip_mreq)) {
773 if (copy_from_user(&mreq, optval,
774 sizeof(struct ip_mreq)))
776 } else if (optlen >= sizeof(struct in_addr)) {
777 if (copy_from_user(&mreq.imr_address, optval,
778 sizeof(struct in_addr)))
783 if (!mreq.imr_ifindex) {
784 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
790 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
792 mreq.imr_ifindex = dev->ifindex;
794 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
797 err = -EADDRNOTAVAIL;
803 if (sk->sk_bound_dev_if &&
804 mreq.imr_ifindex != sk->sk_bound_dev_if)
807 inet->mc_index = mreq.imr_ifindex;
808 inet->mc_addr = mreq.imr_address.s_addr;
813 case IP_ADD_MEMBERSHIP:
814 case IP_DROP_MEMBERSHIP:
816 struct ip_mreqn mreq;
819 if (inet_sk(sk)->is_icsk)
822 if (optlen < sizeof(struct ip_mreq))
825 if (optlen >= sizeof(struct ip_mreqn)) {
826 if (copy_from_user(&mreq, optval, sizeof(mreq)))
829 memset(&mreq, 0, sizeof(mreq));
830 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
834 if (optname == IP_ADD_MEMBERSHIP)
835 err = ip_mc_join_group(sk, &mreq);
837 err = ip_mc_leave_group(sk, &mreq);
842 struct ip_msfilter *msf;
844 if (optlen < IP_MSFILTER_SIZE(0))
846 if (optlen > sysctl_optmem_max) {
850 msf = kmalloc(optlen, GFP_KERNEL);
856 if (copy_from_user(msf, optval, optlen)) {
860 /* numsrc >= (1G-4) overflow in 32 bits */
861 if (msf->imsf_numsrc >= 0x3ffffffcU ||
862 msf->imsf_numsrc > sysctl_igmp_max_msf) {
867 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
872 err = ip_mc_msfilter(sk, msf, 0);
876 case IP_BLOCK_SOURCE:
877 case IP_UNBLOCK_SOURCE:
878 case IP_ADD_SOURCE_MEMBERSHIP:
879 case IP_DROP_SOURCE_MEMBERSHIP:
881 struct ip_mreq_source mreqs;
884 if (optlen != sizeof(struct ip_mreq_source))
886 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
890 if (optname == IP_BLOCK_SOURCE) {
891 omode = MCAST_EXCLUDE;
893 } else if (optname == IP_UNBLOCK_SOURCE) {
894 omode = MCAST_EXCLUDE;
896 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
897 struct ip_mreqn mreq;
899 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
900 mreq.imr_address.s_addr = mreqs.imr_interface;
901 mreq.imr_ifindex = 0;
902 err = ip_mc_join_group(sk, &mreq);
903 if (err && err != -EADDRINUSE)
905 omode = MCAST_INCLUDE;
907 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
908 omode = MCAST_INCLUDE;
911 err = ip_mc_source(add, omode, sk, &mreqs, 0);
914 case MCAST_JOIN_GROUP:
915 case MCAST_LEAVE_GROUP:
917 struct group_req greq;
918 struct sockaddr_in *psin;
919 struct ip_mreqn mreq;
921 if (optlen < sizeof(struct group_req))
924 if (copy_from_user(&greq, optval, sizeof(greq)))
926 psin = (struct sockaddr_in *)&greq.gr_group;
927 if (psin->sin_family != AF_INET)
929 memset(&mreq, 0, sizeof(mreq));
930 mreq.imr_multiaddr = psin->sin_addr;
931 mreq.imr_ifindex = greq.gr_interface;
933 if (optname == MCAST_JOIN_GROUP)
934 err = ip_mc_join_group(sk, &mreq);
936 err = ip_mc_leave_group(sk, &mreq);
939 case MCAST_JOIN_SOURCE_GROUP:
940 case MCAST_LEAVE_SOURCE_GROUP:
941 case MCAST_BLOCK_SOURCE:
942 case MCAST_UNBLOCK_SOURCE:
944 struct group_source_req greqs;
945 struct ip_mreq_source mreqs;
946 struct sockaddr_in *psin;
949 if (optlen != sizeof(struct group_source_req))
951 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
955 if (greqs.gsr_group.ss_family != AF_INET ||
956 greqs.gsr_source.ss_family != AF_INET) {
957 err = -EADDRNOTAVAIL;
960 psin = (struct sockaddr_in *)&greqs.gsr_group;
961 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
962 psin = (struct sockaddr_in *)&greqs.gsr_source;
963 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
964 mreqs.imr_interface = 0; /* use index for mc_source */
966 if (optname == MCAST_BLOCK_SOURCE) {
967 omode = MCAST_EXCLUDE;
969 } else if (optname == MCAST_UNBLOCK_SOURCE) {
970 omode = MCAST_EXCLUDE;
972 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
973 struct ip_mreqn mreq;
975 psin = (struct sockaddr_in *)&greqs.gsr_group;
976 mreq.imr_multiaddr = psin->sin_addr;
977 mreq.imr_address.s_addr = 0;
978 mreq.imr_ifindex = greqs.gsr_interface;
979 err = ip_mc_join_group(sk, &mreq);
980 if (err && err != -EADDRINUSE)
982 greqs.gsr_interface = mreq.imr_ifindex;
983 omode = MCAST_INCLUDE;
985 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
986 omode = MCAST_INCLUDE;
989 err = ip_mc_source(add, omode, sk, &mreqs,
990 greqs.gsr_interface);
995 struct sockaddr_in *psin;
996 struct ip_msfilter *msf = NULL;
997 struct group_filter *gsf = NULL;
998 int msize, i, ifindex;
1000 if (optlen < GROUP_FILTER_SIZE(0))
1002 if (optlen > sysctl_optmem_max) {
1006 gsf = kmalloc(optlen, GFP_KERNEL);
1012 if (copy_from_user(gsf, optval, optlen))
1015 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1016 if (gsf->gf_numsrc >= 0x1ffffff ||
1017 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1021 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1025 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1026 msf = kmalloc(msize, GFP_KERNEL);
1031 ifindex = gsf->gf_interface;
1032 psin = (struct sockaddr_in *)&gsf->gf_group;
1033 if (psin->sin_family != AF_INET) {
1034 err = -EADDRNOTAVAIL;
1037 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1038 msf->imsf_interface = 0;
1039 msf->imsf_fmode = gsf->gf_fmode;
1040 msf->imsf_numsrc = gsf->gf_numsrc;
1041 err = -EADDRNOTAVAIL;
1042 for (i = 0; i < gsf->gf_numsrc; ++i) {
1043 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1045 if (psin->sin_family != AF_INET)
1047 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1052 err = ip_mc_msfilter(sk, msf, ifindex);
1058 case IP_MULTICAST_ALL:
1061 if (val != 0 && val != 1)
1065 case IP_ROUTER_ALERT:
1066 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1072 inet->freebind = !!val;
1075 case IP_IPSEC_POLICY:
1076 case IP_XFRM_POLICY:
1078 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1080 err = xfrm_user_policy(sk, optname, optval, optlen);
1083 case IP_TRANSPARENT:
1084 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1085 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1091 inet->transparent = !!val;
1097 if (val < 0 || val > 255)
1099 inet->min_ttl = val;
1115 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1119 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1120 * destination in skb->cb[] before dst drop.
1121 * This way, receiver doesn't make cache line misses to read rtable.
1123 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1125 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1126 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1129 if (prepare && skb_rtable(skb)) {
1130 pktinfo->ipi_ifindex = inet_iif(skb);
1131 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1133 pktinfo->ipi_ifindex = 0;
1134 pktinfo->ipi_spec_dst.s_addr = 0;
1139 int ip_setsockopt(struct sock *sk, int level,
1140 int optname, char __user *optval, unsigned int optlen)
1144 if (level != SOL_IP)
1145 return -ENOPROTOOPT;
1147 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1148 #ifdef CONFIG_NETFILTER
1149 /* we need to exclude all possible ENOPROTOOPTs except default case */
1150 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1151 optname != IP_IPSEC_POLICY &&
1152 optname != IP_XFRM_POLICY &&
1153 !ip_mroute_opt(optname)) {
1155 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1161 EXPORT_SYMBOL(ip_setsockopt);
1163 #ifdef CONFIG_COMPAT
1164 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1165 char __user *optval, unsigned int optlen)
1169 if (level != SOL_IP)
1170 return -ENOPROTOOPT;
1172 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1173 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1176 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1177 #ifdef CONFIG_NETFILTER
1178 /* we need to exclude all possible ENOPROTOOPTs except default case */
1179 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1180 optname != IP_IPSEC_POLICY &&
1181 optname != IP_XFRM_POLICY &&
1182 !ip_mroute_opt(optname)) {
1184 err = compat_nf_setsockopt(sk, PF_INET, optname,
1191 EXPORT_SYMBOL(compat_ip_setsockopt);
1195 * Get the options. Note for future reference. The GET of IP options gets
1196 * the _received_ ones. The set sets the _sent_ ones.
1199 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1200 char __user *optval, int __user *optlen, unsigned int flags)
1202 struct inet_sock *inet = inet_sk(sk);
1206 if (level != SOL_IP)
1209 if (ip_mroute_opt(optname))
1210 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1212 if (get_user(len, optlen))
1222 unsigned char optbuf[sizeof(struct ip_options)+40];
1223 struct ip_options *opt = (struct ip_options *)optbuf;
1224 struct ip_options_rcu *inet_opt;
1226 inet_opt = rcu_dereference_protected(inet->inet_opt,
1227 sock_owned_by_user(sk));
1230 memcpy(optbuf, &inet_opt->opt,
1231 sizeof(struct ip_options) +
1232 inet_opt->opt.optlen);
1235 if (opt->optlen == 0)
1236 return put_user(0, optlen);
1238 ip_options_undo(opt);
1240 len = min_t(unsigned int, len, opt->optlen);
1241 if (put_user(len, optlen))
1243 if (copy_to_user(optval, opt->__data, len))
1248 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1251 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1254 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1257 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1260 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1263 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1265 case IP_RECVORIGDSTADDR:
1266 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1269 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1275 val = (inet->uc_ttl == -1 ?
1276 sysctl_ip_default_ttl :
1280 val = inet->hdrincl;
1283 val = inet->nodefrag;
1285 case IP_MTU_DISCOVER:
1286 val = inet->pmtudisc;
1290 struct dst_entry *dst;
1292 dst = sk_dst_get(sk);
1304 val = inet->recverr;
1306 case IP_MULTICAST_TTL:
1309 case IP_MULTICAST_LOOP:
1310 val = inet->mc_loop;
1313 val = (__force int)htonl((__u32) inet->uc_index);
1315 case IP_MULTICAST_IF:
1317 struct in_addr addr;
1318 len = min_t(unsigned int, len, sizeof(struct in_addr));
1319 addr.s_addr = inet->mc_addr;
1322 if (put_user(len, optlen))
1324 if (copy_to_user(optval, &addr, len))
1330 struct ip_msfilter msf;
1333 if (len < IP_MSFILTER_SIZE(0)) {
1337 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1341 err = ip_mc_msfget(sk, &msf,
1342 (struct ip_msfilter __user *)optval, optlen);
1346 case MCAST_MSFILTER:
1348 struct group_filter gsf;
1351 if (len < GROUP_FILTER_SIZE(0)) {
1355 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1359 err = ip_mc_gsfget(sk, &gsf,
1360 (struct group_filter __user *)optval,
1365 case IP_MULTICAST_ALL:
1374 if (sk->sk_type != SOCK_STREAM)
1375 return -ENOPROTOOPT;
1377 msg.msg_control = (__force void *) optval;
1378 msg.msg_controllen = len;
1379 msg.msg_flags = flags;
1381 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1382 struct in_pktinfo info;
1384 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1385 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1386 info.ipi_ifindex = inet->mc_index;
1387 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1389 if (inet->cmsg_flags & IP_CMSG_TTL) {
1390 int hlim = inet->mc_ttl;
1391 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1393 if (inet->cmsg_flags & IP_CMSG_TOS) {
1394 int tos = inet->rcv_tos;
1395 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1397 len -= msg.msg_controllen;
1398 return put_user(len, optlen);
1401 val = inet->freebind;
1403 case IP_TRANSPARENT:
1404 val = inet->transparent;
1407 val = inet->min_ttl;
1411 return -ENOPROTOOPT;
1415 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1416 unsigned char ucval = (unsigned char)val;
1418 if (put_user(len, optlen))
1420 if (copy_to_user(optval, &ucval, 1))
1423 len = min_t(unsigned int, sizeof(int), len);
1424 if (put_user(len, optlen))
1426 if (copy_to_user(optval, &val, len))
1432 int ip_getsockopt(struct sock *sk, int level,
1433 int optname, char __user *optval, int __user *optlen)
1437 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1438 #ifdef CONFIG_NETFILTER
1439 /* we need to exclude all possible ENOPROTOOPTs except default case */
1440 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1441 !ip_mroute_opt(optname)) {
1444 if (get_user(len, optlen))
1448 err = nf_getsockopt(sk, PF_INET, optname, optval,
1452 err = put_user(len, optlen);
1458 EXPORT_SYMBOL(ip_getsockopt);
1460 #ifdef CONFIG_COMPAT
1461 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1462 char __user *optval, int __user *optlen)
1466 if (optname == MCAST_MSFILTER)
1467 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1470 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1473 #ifdef CONFIG_NETFILTER
1474 /* we need to exclude all possible ENOPROTOOPTs except default case */
1475 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1476 !ip_mroute_opt(optname)) {
1479 if (get_user(len, optlen))
1483 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1486 err = put_user(len, optlen);
1492 EXPORT_SYMBOL(compat_ip_getsockopt);