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 csum = csum_sub(csum,
110 csum_partial(skb->data + tlen,
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 struct sockaddr_in sin;
137 const struct iphdr *iph = ip_hdr(skb);
138 __be16 *ports = (__be16 *)skb_transport_header(skb);
140 if (skb_transport_offset(skb) + 4 > skb->len)
143 /* All current transport protocols have the port numbers in the
144 * first four bytes of the transport header and this function is
145 * written with this assumption in mind.
148 sin.sin_family = AF_INET;
149 sin.sin_addr.s_addr = iph->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 net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
228 struct cmsghdr *cmsg;
230 for_each_cmsghdr(cmsg, msg) {
231 if (!CMSG_OK(msg, cmsg))
233 #if IS_ENABLED(CONFIG_IPV6)
235 cmsg->cmsg_level == SOL_IPV6 &&
236 cmsg->cmsg_type == IPV6_PKTINFO) {
237 struct in6_pktinfo *src_info;
239 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
241 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
242 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
244 ipc->oif = src_info->ipi6_ifindex;
245 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
249 if (cmsg->cmsg_level != SOL_IP)
251 switch (cmsg->cmsg_type) {
253 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
255 /* Our caller is responsible for freeing ipc->opt */
256 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
257 err < 40 ? err : 40);
263 struct in_pktinfo *info;
264 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
266 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
267 ipc->oif = info->ipi_ifindex;
268 ipc->addr = info->ipi_spec_dst.s_addr;
272 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
274 val = *(int *)CMSG_DATA(cmsg);
275 if (val < 1 || val > 255)
280 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
282 val = *(int *)CMSG_DATA(cmsg);
283 if (val < 0 || val > 255)
286 ipc->priority = rt_tos2priority(ipc->tos);
297 /* Special input handler for packets caught by router alert option.
298 They are selected only by protocol field, and then processed likely
299 local ones; but only if someone wants them! Otherwise, router
300 not running rsvpd will kill RSVP.
302 It is user level problem, what it will make with them.
303 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
304 but receiver should be enough clever f.e. to forward mtrace requests,
305 sent to multicast group to reach destination designated router.
307 struct ip_ra_chain __rcu *ip_ra_chain;
308 static DEFINE_SPINLOCK(ip_ra_lock);
311 static void ip_ra_destroy_rcu(struct rcu_head *head)
313 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
315 sock_put(ra->saved_sk);
319 int ip_ra_control(struct sock *sk, unsigned char on,
320 void (*destructor)(struct sock *))
322 struct ip_ra_chain *ra, *new_ra;
323 struct ip_ra_chain __rcu **rap;
325 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
328 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
330 spin_lock_bh(&ip_ra_lock);
331 for (rap = &ip_ra_chain;
332 (ra = rcu_dereference_protected(*rap,
333 lockdep_is_held(&ip_ra_lock))) != NULL;
337 spin_unlock_bh(&ip_ra_lock);
341 /* dont let ip_call_ra_chain() use sk again */
343 RCU_INIT_POINTER(*rap, ra->next);
344 spin_unlock_bh(&ip_ra_lock);
349 * Delay sock_put(sk) and kfree(ra) after one rcu grace
350 * period. This guarantee ip_call_ra_chain() dont need
351 * to mess with socket refcounts.
354 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
359 spin_unlock_bh(&ip_ra_lock);
363 new_ra->destructor = destructor;
365 RCU_INIT_POINTER(new_ra->next, ra);
366 rcu_assign_pointer(*rap, new_ra);
368 spin_unlock_bh(&ip_ra_lock);
373 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
374 __be16 port, u32 info, u8 *payload)
376 struct sock_exterr_skb *serr;
378 skb = skb_clone(skb, GFP_ATOMIC);
382 serr = SKB_EXT_ERR(skb);
383 serr->ee.ee_errno = err;
384 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
385 serr->ee.ee_type = icmp_hdr(skb)->type;
386 serr->ee.ee_code = icmp_hdr(skb)->code;
388 serr->ee.ee_info = info;
389 serr->ee.ee_data = 0;
390 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
391 skb_network_header(skb);
394 if (skb_pull(skb, payload - skb->data)) {
395 skb_reset_transport_header(skb);
396 if (sock_queue_err_skb(sk, skb) == 0)
402 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
404 struct inet_sock *inet = inet_sk(sk);
405 struct sock_exterr_skb *serr;
412 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
416 skb_put(skb, sizeof(struct iphdr));
417 skb_reset_network_header(skb);
421 serr = SKB_EXT_ERR(skb);
422 serr->ee.ee_errno = err;
423 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
424 serr->ee.ee_type = 0;
425 serr->ee.ee_code = 0;
427 serr->ee.ee_info = info;
428 serr->ee.ee_data = 0;
429 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
432 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
433 skb_reset_transport_header(skb);
435 if (sock_queue_err_skb(sk, skb))
439 /* For some errors we have valid addr_offset even with zero payload and
440 * zero port. Also, addr_offset should be supported if port is set.
442 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
444 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
445 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
448 /* IPv4 supports cmsg on all imcp errors and some timestamps
450 * Timestamp code paths do not initialize the fields expected by cmsg:
451 * the PKTINFO fields in skb->cb[]. Fill those in here.
453 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
457 struct in_pktinfo *info;
459 if (ee_origin == SO_EE_ORIGIN_ICMP)
462 if (ee_origin == SO_EE_ORIGIN_LOCAL)
465 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
466 * timestamp with egress dev. Not possible for packets without dev
467 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
469 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
473 info = PKTINFO_SKB_CB(skb);
474 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
475 info->ipi_ifindex = skb->dev->ifindex;
480 * Handle MSG_ERRQUEUE
482 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
484 struct sock_exterr_skb *serr;
486 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
488 struct sock_extended_err ee;
489 struct sockaddr_in offender;
494 WARN_ON_ONCE(sk->sk_family == AF_INET6);
497 skb = sock_dequeue_err_skb(sk);
503 msg->msg_flags |= MSG_TRUNC;
506 err = skb_copy_datagram_msg(skb, 0, msg, copied);
510 sock_recv_timestamp(msg, sk, skb);
512 serr = SKB_EXT_ERR(skb);
514 if (sin && ipv4_datagram_support_addr(serr)) {
515 sin->sin_family = AF_INET;
516 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
518 sin->sin_port = serr->port;
519 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
520 *addr_len = sizeof(*sin);
523 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
524 sin = &errhdr.offender;
525 memset(sin, 0, sizeof(*sin));
527 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
528 sin->sin_family = AF_INET;
529 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
530 if (inet_sk(sk)->cmsg_flags)
531 ip_cmsg_recv(msg, skb);
534 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
536 /* Now we could try to dump offended packet options */
538 msg->msg_flags |= MSG_ERRQUEUE;
549 * Socket option code for IP. This is the end of the line after any
550 * TCP,UDP etc options on an IP socket.
552 static bool setsockopt_needs_rtnl(int optname)
555 case IP_ADD_MEMBERSHIP:
556 case IP_ADD_SOURCE_MEMBERSHIP:
557 case IP_BLOCK_SOURCE:
558 case IP_DROP_MEMBERSHIP:
559 case IP_DROP_SOURCE_MEMBERSHIP:
561 case IP_UNBLOCK_SOURCE:
562 case MCAST_BLOCK_SOURCE:
564 case MCAST_JOIN_GROUP:
565 case MCAST_JOIN_SOURCE_GROUP:
566 case MCAST_LEAVE_GROUP:
567 case MCAST_LEAVE_SOURCE_GROUP:
568 case MCAST_UNBLOCK_SOURCE:
574 static int do_ip_setsockopt(struct sock *sk, int level,
575 int optname, char __user *optval, unsigned int optlen)
577 struct inet_sock *inet = inet_sk(sk);
579 bool needs_rtnl = setsockopt_needs_rtnl(optname);
590 case IP_MTU_DISCOVER:
592 case IP_ROUTER_ALERT:
598 case IP_BIND_ADDRESS_NO_PORT:
600 case IP_MULTICAST_TTL:
601 case IP_MULTICAST_ALL:
602 case IP_MULTICAST_LOOP:
603 case IP_RECVORIGDSTADDR:
605 if (optlen >= sizeof(int)) {
606 if (get_user(val, (int __user *) optval))
608 } else if (optlen >= sizeof(char)) {
611 if (get_user(ucval, (unsigned char __user *) optval))
617 /* If optlen==0, it is equivalent to val == 0 */
619 if (ip_mroute_opt(optname))
620 return ip_mroute_setsockopt(sk, optname, optval, optlen);
630 struct ip_options_rcu *old, *opt = NULL;
634 err = ip_options_get_from_user(sock_net(sk), &opt,
638 old = rcu_dereference_protected(inet->inet_opt,
639 sock_owned_by_user(sk));
641 struct inet_connection_sock *icsk = inet_csk(sk);
642 #if IS_ENABLED(CONFIG_IPV6)
643 if (sk->sk_family == PF_INET ||
644 (!((1 << sk->sk_state) &
645 (TCPF_LISTEN | TCPF_CLOSE)) &&
646 inet->inet_daddr != LOOPBACK4_IPV6)) {
649 icsk->icsk_ext_hdr_len -= old->opt.optlen;
651 icsk->icsk_ext_hdr_len += opt->opt.optlen;
652 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
653 #if IS_ENABLED(CONFIG_IPV6)
657 rcu_assign_pointer(inet->inet_opt, opt);
664 inet->cmsg_flags |= IP_CMSG_PKTINFO;
666 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
670 inet->cmsg_flags |= IP_CMSG_TTL;
672 inet->cmsg_flags &= ~IP_CMSG_TTL;
676 inet->cmsg_flags |= IP_CMSG_TOS;
678 inet->cmsg_flags &= ~IP_CMSG_TOS;
682 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
684 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
688 inet->cmsg_flags |= IP_CMSG_RETOPTS;
690 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
694 inet->cmsg_flags |= IP_CMSG_PASSSEC;
696 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
698 case IP_RECVORIGDSTADDR:
700 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
702 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
706 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
707 inet_inc_convert_csum(sk);
708 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
711 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
712 inet_dec_convert_csum(sk);
713 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
717 case IP_TOS: /* This sets both TOS and Precedence */
718 if (sk->sk_type == SOCK_STREAM) {
719 val &= ~INET_ECN_MASK;
720 val |= inet->tos & INET_ECN_MASK;
722 if (inet->tos != val) {
724 sk->sk_priority = rt_tos2priority(val);
731 if (val != -1 && (val < 1 || val > 255))
736 if (sk->sk_type != SOCK_RAW) {
740 inet->hdrincl = val ? 1 : 0;
743 if (sk->sk_type != SOCK_RAW) {
747 inet->nodefrag = val ? 1 : 0;
749 case IP_BIND_ADDRESS_NO_PORT:
750 inet->bind_address_no_port = val ? 1 : 0;
752 case IP_MTU_DISCOVER:
753 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
755 inet->pmtudisc = val;
758 inet->recverr = !!val;
760 skb_queue_purge(&sk->sk_error_queue);
762 case IP_MULTICAST_TTL:
763 if (sk->sk_type == SOCK_STREAM)
769 if (val < 0 || val > 255)
773 case IP_MULTICAST_LOOP:
776 inet->mc_loop = !!val;
780 struct net_device *dev = NULL;
783 if (optlen != sizeof(int))
786 ifindex = (__force int)ntohl((__force __be32)val);
793 dev = dev_get_by_index(sock_net(sk), ifindex);
794 err = -EADDRNOTAVAIL;
800 if (sk->sk_bound_dev_if)
803 inet->uc_index = ifindex;
807 case IP_MULTICAST_IF:
809 struct ip_mreqn mreq;
810 struct net_device *dev = NULL;
812 if (sk->sk_type == SOCK_STREAM)
815 * Check the arguments are allowable
818 if (optlen < sizeof(struct in_addr))
822 if (optlen >= sizeof(struct ip_mreqn)) {
823 if (copy_from_user(&mreq, optval, sizeof(mreq)))
826 memset(&mreq, 0, sizeof(mreq));
827 if (optlen >= sizeof(struct ip_mreq)) {
828 if (copy_from_user(&mreq, optval,
829 sizeof(struct ip_mreq)))
831 } else if (optlen >= sizeof(struct in_addr)) {
832 if (copy_from_user(&mreq.imr_address, optval,
833 sizeof(struct in_addr)))
838 if (!mreq.imr_ifindex) {
839 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
845 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
847 mreq.imr_ifindex = dev->ifindex;
849 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
852 err = -EADDRNOTAVAIL;
858 if (sk->sk_bound_dev_if &&
859 mreq.imr_ifindex != sk->sk_bound_dev_if)
862 inet->mc_index = mreq.imr_ifindex;
863 inet->mc_addr = mreq.imr_address.s_addr;
868 case IP_ADD_MEMBERSHIP:
869 case IP_DROP_MEMBERSHIP:
871 struct ip_mreqn mreq;
874 if (inet_sk(sk)->is_icsk)
877 if (optlen < sizeof(struct ip_mreq))
880 if (optlen >= sizeof(struct ip_mreqn)) {
881 if (copy_from_user(&mreq, optval, sizeof(mreq)))
884 memset(&mreq, 0, sizeof(mreq));
885 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
889 if (optname == IP_ADD_MEMBERSHIP)
890 err = ip_mc_join_group(sk, &mreq);
892 err = ip_mc_leave_group(sk, &mreq);
897 struct ip_msfilter *msf;
899 if (optlen < IP_MSFILTER_SIZE(0))
901 if (optlen > sysctl_optmem_max) {
905 msf = kmalloc(optlen, GFP_KERNEL);
911 if (copy_from_user(msf, optval, optlen)) {
915 /* numsrc >= (1G-4) overflow in 32 bits */
916 if (msf->imsf_numsrc >= 0x3ffffffcU ||
917 msf->imsf_numsrc > sysctl_igmp_max_msf) {
922 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
927 err = ip_mc_msfilter(sk, msf, 0);
931 case IP_BLOCK_SOURCE:
932 case IP_UNBLOCK_SOURCE:
933 case IP_ADD_SOURCE_MEMBERSHIP:
934 case IP_DROP_SOURCE_MEMBERSHIP:
936 struct ip_mreq_source mreqs;
939 if (optlen != sizeof(struct ip_mreq_source))
941 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
945 if (optname == IP_BLOCK_SOURCE) {
946 omode = MCAST_EXCLUDE;
948 } else if (optname == IP_UNBLOCK_SOURCE) {
949 omode = MCAST_EXCLUDE;
951 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
952 struct ip_mreqn mreq;
954 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
955 mreq.imr_address.s_addr = mreqs.imr_interface;
956 mreq.imr_ifindex = 0;
957 err = ip_mc_join_group(sk, &mreq);
958 if (err && err != -EADDRINUSE)
960 omode = MCAST_INCLUDE;
962 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
963 omode = MCAST_INCLUDE;
966 err = ip_mc_source(add, omode, sk, &mreqs, 0);
969 case MCAST_JOIN_GROUP:
970 case MCAST_LEAVE_GROUP:
972 struct group_req greq;
973 struct sockaddr_in *psin;
974 struct ip_mreqn mreq;
976 if (optlen < sizeof(struct group_req))
979 if (copy_from_user(&greq, optval, sizeof(greq)))
981 psin = (struct sockaddr_in *)&greq.gr_group;
982 if (psin->sin_family != AF_INET)
984 memset(&mreq, 0, sizeof(mreq));
985 mreq.imr_multiaddr = psin->sin_addr;
986 mreq.imr_ifindex = greq.gr_interface;
988 if (optname == MCAST_JOIN_GROUP)
989 err = ip_mc_join_group(sk, &mreq);
991 err = ip_mc_leave_group(sk, &mreq);
994 case MCAST_JOIN_SOURCE_GROUP:
995 case MCAST_LEAVE_SOURCE_GROUP:
996 case MCAST_BLOCK_SOURCE:
997 case MCAST_UNBLOCK_SOURCE:
999 struct group_source_req greqs;
1000 struct ip_mreq_source mreqs;
1001 struct sockaddr_in *psin;
1004 if (optlen != sizeof(struct group_source_req))
1006 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1010 if (greqs.gsr_group.ss_family != AF_INET ||
1011 greqs.gsr_source.ss_family != AF_INET) {
1012 err = -EADDRNOTAVAIL;
1015 psin = (struct sockaddr_in *)&greqs.gsr_group;
1016 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1017 psin = (struct sockaddr_in *)&greqs.gsr_source;
1018 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1019 mreqs.imr_interface = 0; /* use index for mc_source */
1021 if (optname == MCAST_BLOCK_SOURCE) {
1022 omode = MCAST_EXCLUDE;
1024 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1025 omode = MCAST_EXCLUDE;
1027 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1028 struct ip_mreqn mreq;
1030 psin = (struct sockaddr_in *)&greqs.gsr_group;
1031 mreq.imr_multiaddr = psin->sin_addr;
1032 mreq.imr_address.s_addr = 0;
1033 mreq.imr_ifindex = greqs.gsr_interface;
1034 err = ip_mc_join_group(sk, &mreq);
1035 if (err && err != -EADDRINUSE)
1037 greqs.gsr_interface = mreq.imr_ifindex;
1038 omode = MCAST_INCLUDE;
1040 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1041 omode = MCAST_INCLUDE;
1044 err = ip_mc_source(add, omode, sk, &mreqs,
1045 greqs.gsr_interface);
1048 case MCAST_MSFILTER:
1050 struct sockaddr_in *psin;
1051 struct ip_msfilter *msf = NULL;
1052 struct group_filter *gsf = NULL;
1053 int msize, i, ifindex;
1055 if (optlen < GROUP_FILTER_SIZE(0))
1057 if (optlen > sysctl_optmem_max) {
1061 gsf = kmalloc(optlen, GFP_KERNEL);
1067 if (copy_from_user(gsf, optval, optlen))
1070 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1071 if (gsf->gf_numsrc >= 0x1ffffff ||
1072 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1076 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1080 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1081 msf = kmalloc(msize, GFP_KERNEL);
1086 ifindex = gsf->gf_interface;
1087 psin = (struct sockaddr_in *)&gsf->gf_group;
1088 if (psin->sin_family != AF_INET) {
1089 err = -EADDRNOTAVAIL;
1092 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1093 msf->imsf_interface = 0;
1094 msf->imsf_fmode = gsf->gf_fmode;
1095 msf->imsf_numsrc = gsf->gf_numsrc;
1096 err = -EADDRNOTAVAIL;
1097 for (i = 0; i < gsf->gf_numsrc; ++i) {
1098 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1100 if (psin->sin_family != AF_INET)
1102 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1107 err = ip_mc_msfilter(sk, msf, ifindex);
1113 case IP_MULTICAST_ALL:
1116 if (val != 0 && val != 1)
1120 case IP_ROUTER_ALERT:
1121 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1127 inet->freebind = !!val;
1130 case IP_IPSEC_POLICY:
1131 case IP_XFRM_POLICY:
1133 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1135 err = xfrm_user_policy(sk, optname, optval, optlen);
1138 case IP_TRANSPARENT:
1139 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1140 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1146 inet->transparent = !!val;
1152 if (val < 0 || val > 255)
1154 inet->min_ttl = val;
1174 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1178 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1179 * destination in skb->cb[] before dst drop.
1180 * This way, receiver doesn't make cache line misses to read rtable.
1182 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1184 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1185 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1188 if (prepare && skb_rtable(skb)) {
1189 pktinfo->ipi_ifindex = inet_iif(skb);
1190 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1192 pktinfo->ipi_ifindex = 0;
1193 pktinfo->ipi_spec_dst.s_addr = 0;
1195 /* We need to keep the dst for __ip_options_echo()
1196 * We could restrict the test to opt.ts_needtime || opt.srr,
1197 * but the following is good enough as IP options are not often used.
1199 if (unlikely(IPCB(skb)->opt.optlen))
1205 int ip_setsockopt(struct sock *sk, int level,
1206 int optname, char __user *optval, unsigned int optlen)
1210 if (level != SOL_IP)
1211 return -ENOPROTOOPT;
1213 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1214 #ifdef CONFIG_NETFILTER
1215 /* we need to exclude all possible ENOPROTOOPTs except default case */
1216 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1217 optname != IP_IPSEC_POLICY &&
1218 optname != IP_XFRM_POLICY &&
1219 !ip_mroute_opt(optname)) {
1221 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1227 EXPORT_SYMBOL(ip_setsockopt);
1229 #ifdef CONFIG_COMPAT
1230 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1231 char __user *optval, unsigned int optlen)
1235 if (level != SOL_IP)
1236 return -ENOPROTOOPT;
1238 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1239 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1242 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1243 #ifdef CONFIG_NETFILTER
1244 /* we need to exclude all possible ENOPROTOOPTs except default case */
1245 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1246 optname != IP_IPSEC_POLICY &&
1247 optname != IP_XFRM_POLICY &&
1248 !ip_mroute_opt(optname)) {
1250 err = compat_nf_setsockopt(sk, PF_INET, optname,
1257 EXPORT_SYMBOL(compat_ip_setsockopt);
1261 * Get the options. Note for future reference. The GET of IP options gets
1262 * the _received_ ones. The set sets the _sent_ ones.
1265 static bool getsockopt_needs_rtnl(int optname)
1269 case MCAST_MSFILTER:
1275 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1276 char __user *optval, int __user *optlen, unsigned int flags)
1278 struct inet_sock *inet = inet_sk(sk);
1279 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1283 if (level != SOL_IP)
1286 if (ip_mroute_opt(optname))
1287 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1289 if (get_user(len, optlen))
1301 unsigned char optbuf[sizeof(struct ip_options)+40];
1302 struct ip_options *opt = (struct ip_options *)optbuf;
1303 struct ip_options_rcu *inet_opt;
1305 inet_opt = rcu_dereference_protected(inet->inet_opt,
1306 sock_owned_by_user(sk));
1309 memcpy(optbuf, &inet_opt->opt,
1310 sizeof(struct ip_options) +
1311 inet_opt->opt.optlen);
1314 if (opt->optlen == 0)
1315 return put_user(0, optlen);
1317 ip_options_undo(opt);
1319 len = min_t(unsigned int, len, opt->optlen);
1320 if (put_user(len, optlen))
1322 if (copy_to_user(optval, opt->__data, len))
1327 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1330 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1333 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1336 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1339 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1342 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1344 case IP_RECVORIGDSTADDR:
1345 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1348 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1354 val = (inet->uc_ttl == -1 ?
1355 sysctl_ip_default_ttl :
1359 val = inet->hdrincl;
1362 val = inet->nodefrag;
1364 case IP_BIND_ADDRESS_NO_PORT:
1365 val = inet->bind_address_no_port;
1367 case IP_MTU_DISCOVER:
1368 val = inet->pmtudisc;
1372 struct dst_entry *dst;
1374 dst = sk_dst_get(sk);
1386 val = inet->recverr;
1388 case IP_MULTICAST_TTL:
1391 case IP_MULTICAST_LOOP:
1392 val = inet->mc_loop;
1395 val = (__force int)htonl((__u32) inet->uc_index);
1397 case IP_MULTICAST_IF:
1399 struct in_addr addr;
1400 len = min_t(unsigned int, len, sizeof(struct in_addr));
1401 addr.s_addr = inet->mc_addr;
1404 if (put_user(len, optlen))
1406 if (copy_to_user(optval, &addr, len))
1412 struct ip_msfilter msf;
1414 if (len < IP_MSFILTER_SIZE(0)) {
1418 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1422 err = ip_mc_msfget(sk, &msf,
1423 (struct ip_msfilter __user *)optval, optlen);
1426 case MCAST_MSFILTER:
1428 struct group_filter gsf;
1430 if (len < GROUP_FILTER_SIZE(0)) {
1434 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1438 err = ip_mc_gsfget(sk, &gsf,
1439 (struct group_filter __user *)optval,
1443 case IP_MULTICAST_ALL:
1452 if (sk->sk_type != SOCK_STREAM)
1453 return -ENOPROTOOPT;
1455 msg.msg_control = (__force void *) optval;
1456 msg.msg_controllen = len;
1457 msg.msg_flags = flags;
1459 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1460 struct in_pktinfo info;
1462 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1463 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1464 info.ipi_ifindex = inet->mc_index;
1465 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1467 if (inet->cmsg_flags & IP_CMSG_TTL) {
1468 int hlim = inet->mc_ttl;
1469 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1471 if (inet->cmsg_flags & IP_CMSG_TOS) {
1472 int tos = inet->rcv_tos;
1473 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1475 len -= msg.msg_controllen;
1476 return put_user(len, optlen);
1479 val = inet->freebind;
1481 case IP_TRANSPARENT:
1482 val = inet->transparent;
1485 val = inet->min_ttl;
1489 return -ENOPROTOOPT;
1493 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1494 unsigned char ucval = (unsigned char)val;
1496 if (put_user(len, optlen))
1498 if (copy_to_user(optval, &ucval, 1))
1501 len = min_t(unsigned int, sizeof(int), len);
1502 if (put_user(len, optlen))
1504 if (copy_to_user(optval, &val, len))
1516 int ip_getsockopt(struct sock *sk, int level,
1517 int optname, char __user *optval, int __user *optlen)
1521 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1522 #ifdef CONFIG_NETFILTER
1523 /* we need to exclude all possible ENOPROTOOPTs except default case */
1524 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1525 !ip_mroute_opt(optname)) {
1528 if (get_user(len, optlen))
1532 err = nf_getsockopt(sk, PF_INET, optname, optval,
1536 err = put_user(len, optlen);
1542 EXPORT_SYMBOL(ip_getsockopt);
1544 #ifdef CONFIG_COMPAT
1545 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1546 char __user *optval, int __user *optlen)
1550 if (optname == MCAST_MSFILTER)
1551 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1554 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1557 #ifdef CONFIG_NETFILTER
1558 /* we need to exclude all possible ENOPROTOOPTs except default case */
1559 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1560 !ip_mroute_opt(optname)) {
1563 if (get_user(len, optlen))
1567 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1570 err = put_user(len, optlen);
1576 EXPORT_SYMBOL(compat_ip_getsockopt);