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_transport_header(skb) + 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 sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
228 struct cmsghdr *cmsg;
229 struct net *net = sock_net(sk);
231 for_each_cmsghdr(cmsg, msg) {
232 if (!CMSG_OK(msg, cmsg))
234 #if IS_ENABLED(CONFIG_IPV6)
236 cmsg->cmsg_level == SOL_IPV6 &&
237 cmsg->cmsg_type == IPV6_PKTINFO) {
238 struct in6_pktinfo *src_info;
240 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
242 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
243 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
245 ipc->oif = src_info->ipi6_ifindex;
246 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
250 if (cmsg->cmsg_level == SOL_SOCKET) {
251 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
257 if (cmsg->cmsg_level != SOL_IP)
259 switch (cmsg->cmsg_type) {
261 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
263 /* Our caller is responsible for freeing ipc->opt */
264 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
265 err < 40 ? err : 40);
271 struct in_pktinfo *info;
272 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
274 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
275 ipc->oif = info->ipi_ifindex;
276 ipc->addr = info->ipi_spec_dst.s_addr;
280 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
282 val = *(int *)CMSG_DATA(cmsg);
283 if (val < 1 || val > 255)
288 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
289 val = *(int *)CMSG_DATA(cmsg);
290 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
291 val = *(u8 *)CMSG_DATA(cmsg);
294 if (val < 0 || val > 255)
297 ipc->priority = rt_tos2priority(ipc->tos);
308 /* Special input handler for packets caught by router alert option.
309 They are selected only by protocol field, and then processed likely
310 local ones; but only if someone wants them! Otherwise, router
311 not running rsvpd will kill RSVP.
313 It is user level problem, what it will make with them.
314 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
315 but receiver should be enough clever f.e. to forward mtrace requests,
316 sent to multicast group to reach destination designated router.
318 struct ip_ra_chain __rcu *ip_ra_chain;
319 static DEFINE_SPINLOCK(ip_ra_lock);
322 static void ip_ra_destroy_rcu(struct rcu_head *head)
324 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
326 sock_put(ra->saved_sk);
330 int ip_ra_control(struct sock *sk, unsigned char on,
331 void (*destructor)(struct sock *))
333 struct ip_ra_chain *ra, *new_ra;
334 struct ip_ra_chain __rcu **rap;
336 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
339 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
341 spin_lock_bh(&ip_ra_lock);
342 for (rap = &ip_ra_chain;
343 (ra = rcu_dereference_protected(*rap,
344 lockdep_is_held(&ip_ra_lock))) != NULL;
348 spin_unlock_bh(&ip_ra_lock);
352 /* dont let ip_call_ra_chain() use sk again */
354 RCU_INIT_POINTER(*rap, ra->next);
355 spin_unlock_bh(&ip_ra_lock);
360 * Delay sock_put(sk) and kfree(ra) after one rcu grace
361 * period. This guarantee ip_call_ra_chain() dont need
362 * to mess with socket refcounts.
365 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
370 spin_unlock_bh(&ip_ra_lock);
374 new_ra->destructor = destructor;
376 RCU_INIT_POINTER(new_ra->next, ra);
377 rcu_assign_pointer(*rap, new_ra);
379 spin_unlock_bh(&ip_ra_lock);
384 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
385 __be16 port, u32 info, u8 *payload)
387 struct sock_exterr_skb *serr;
389 skb = skb_clone(skb, GFP_ATOMIC);
393 serr = SKB_EXT_ERR(skb);
394 serr->ee.ee_errno = err;
395 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
396 serr->ee.ee_type = icmp_hdr(skb)->type;
397 serr->ee.ee_code = icmp_hdr(skb)->code;
399 serr->ee.ee_info = info;
400 serr->ee.ee_data = 0;
401 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
402 skb_network_header(skb);
405 if (skb_pull(skb, payload - skb->data)) {
406 skb_reset_transport_header(skb);
407 if (sock_queue_err_skb(sk, skb) == 0)
413 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
415 struct inet_sock *inet = inet_sk(sk);
416 struct sock_exterr_skb *serr;
423 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
427 skb_put(skb, sizeof(struct iphdr));
428 skb_reset_network_header(skb);
432 serr = SKB_EXT_ERR(skb);
433 serr->ee.ee_errno = err;
434 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
435 serr->ee.ee_type = 0;
436 serr->ee.ee_code = 0;
438 serr->ee.ee_info = info;
439 serr->ee.ee_data = 0;
440 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
443 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
444 skb_reset_transport_header(skb);
446 if (sock_queue_err_skb(sk, skb))
450 /* For some errors we have valid addr_offset even with zero payload and
451 * zero port. Also, addr_offset should be supported if port is set.
453 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
455 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
456 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
459 /* IPv4 supports cmsg on all imcp errors and some timestamps
461 * Timestamp code paths do not initialize the fields expected by cmsg:
462 * the PKTINFO fields in skb->cb[]. Fill those in here.
464 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
468 struct in_pktinfo *info;
470 if (ee_origin == SO_EE_ORIGIN_ICMP)
473 if (ee_origin == SO_EE_ORIGIN_LOCAL)
476 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
477 * timestamp with egress dev. Not possible for packets without dev
478 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
480 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
484 info = PKTINFO_SKB_CB(skb);
485 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
486 info->ipi_ifindex = skb->dev->ifindex;
491 * Handle MSG_ERRQUEUE
493 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
495 struct sock_exterr_skb *serr;
497 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
499 struct sock_extended_err ee;
500 struct sockaddr_in offender;
505 WARN_ON_ONCE(sk->sk_family == AF_INET6);
508 skb = sock_dequeue_err_skb(sk);
514 msg->msg_flags |= MSG_TRUNC;
517 err = skb_copy_datagram_msg(skb, 0, msg, copied);
522 sock_recv_timestamp(msg, sk, skb);
524 serr = SKB_EXT_ERR(skb);
526 if (sin && ipv4_datagram_support_addr(serr)) {
527 sin->sin_family = AF_INET;
528 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
530 sin->sin_port = serr->port;
531 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
532 *addr_len = sizeof(*sin);
535 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
536 sin = &errhdr.offender;
537 memset(sin, 0, sizeof(*sin));
539 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
540 sin->sin_family = AF_INET;
541 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
542 if (inet_sk(sk)->cmsg_flags)
543 ip_cmsg_recv(msg, skb);
546 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
548 /* Now we could try to dump offended packet options */
550 msg->msg_flags |= MSG_ERRQUEUE;
560 * Socket option code for IP. This is the end of the line after any
561 * TCP,UDP etc options on an IP socket.
563 static bool setsockopt_needs_rtnl(int optname)
566 case IP_ADD_MEMBERSHIP:
567 case IP_ADD_SOURCE_MEMBERSHIP:
568 case IP_BLOCK_SOURCE:
569 case IP_DROP_MEMBERSHIP:
570 case IP_DROP_SOURCE_MEMBERSHIP:
572 case IP_UNBLOCK_SOURCE:
573 case MCAST_BLOCK_SOURCE:
575 case MCAST_JOIN_GROUP:
576 case MCAST_JOIN_SOURCE_GROUP:
577 case MCAST_LEAVE_GROUP:
578 case MCAST_LEAVE_SOURCE_GROUP:
579 case MCAST_UNBLOCK_SOURCE:
585 static int do_ip_setsockopt(struct sock *sk, int level,
586 int optname, char __user *optval, unsigned int optlen)
588 struct inet_sock *inet = inet_sk(sk);
589 struct net *net = sock_net(sk);
591 bool needs_rtnl = setsockopt_needs_rtnl(optname);
602 case IP_MTU_DISCOVER:
604 case IP_ROUTER_ALERT:
610 case IP_BIND_ADDRESS_NO_PORT:
612 case IP_MULTICAST_TTL:
613 case IP_MULTICAST_ALL:
614 case IP_MULTICAST_LOOP:
615 case IP_RECVORIGDSTADDR:
617 if (optlen >= sizeof(int)) {
618 if (get_user(val, (int __user *) optval))
620 } else if (optlen >= sizeof(char)) {
623 if (get_user(ucval, (unsigned char __user *) optval))
629 /* If optlen==0, it is equivalent to val == 0 */
631 if (ip_mroute_opt(optname))
632 return ip_mroute_setsockopt(sk, optname, optval, optlen);
642 struct ip_options_rcu *old, *opt = NULL;
646 err = ip_options_get_from_user(sock_net(sk), &opt,
650 old = rcu_dereference_protected(inet->inet_opt,
651 lockdep_sock_is_held(sk));
653 struct inet_connection_sock *icsk = inet_csk(sk);
654 #if IS_ENABLED(CONFIG_IPV6)
655 if (sk->sk_family == PF_INET ||
656 (!((1 << sk->sk_state) &
657 (TCPF_LISTEN | TCPF_CLOSE)) &&
658 inet->inet_daddr != LOOPBACK4_IPV6)) {
661 icsk->icsk_ext_hdr_len -= old->opt.optlen;
663 icsk->icsk_ext_hdr_len += opt->opt.optlen;
664 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
665 #if IS_ENABLED(CONFIG_IPV6)
669 rcu_assign_pointer(inet->inet_opt, opt);
676 inet->cmsg_flags |= IP_CMSG_PKTINFO;
678 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
682 inet->cmsg_flags |= IP_CMSG_TTL;
684 inet->cmsg_flags &= ~IP_CMSG_TTL;
688 inet->cmsg_flags |= IP_CMSG_TOS;
690 inet->cmsg_flags &= ~IP_CMSG_TOS;
694 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
696 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
700 inet->cmsg_flags |= IP_CMSG_RETOPTS;
702 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
706 inet->cmsg_flags |= IP_CMSG_PASSSEC;
708 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
710 case IP_RECVORIGDSTADDR:
712 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
714 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
718 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
719 inet_inc_convert_csum(sk);
720 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
723 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
724 inet_dec_convert_csum(sk);
725 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
729 case IP_TOS: /* This sets both TOS and Precedence */
730 if (sk->sk_type == SOCK_STREAM) {
731 val &= ~INET_ECN_MASK;
732 val |= inet->tos & INET_ECN_MASK;
734 if (inet->tos != val) {
736 sk->sk_priority = rt_tos2priority(val);
743 if (val != -1 && (val < 1 || val > 255))
748 if (sk->sk_type != SOCK_RAW) {
752 inet->hdrincl = val ? 1 : 0;
755 if (sk->sk_type != SOCK_RAW) {
759 inet->nodefrag = val ? 1 : 0;
761 case IP_BIND_ADDRESS_NO_PORT:
762 inet->bind_address_no_port = val ? 1 : 0;
764 case IP_MTU_DISCOVER:
765 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
767 inet->pmtudisc = val;
770 inet->recverr = !!val;
772 skb_queue_purge(&sk->sk_error_queue);
774 case IP_MULTICAST_TTL:
775 if (sk->sk_type == SOCK_STREAM)
781 if (val < 0 || val > 255)
785 case IP_MULTICAST_LOOP:
788 inet->mc_loop = !!val;
792 struct net_device *dev = NULL;
795 if (optlen != sizeof(int))
798 ifindex = (__force int)ntohl((__force __be32)val);
805 dev = dev_get_by_index(sock_net(sk), ifindex);
806 err = -EADDRNOTAVAIL;
812 if (sk->sk_bound_dev_if)
815 inet->uc_index = ifindex;
819 case IP_MULTICAST_IF:
821 struct ip_mreqn mreq;
822 struct net_device *dev = NULL;
824 if (sk->sk_type == SOCK_STREAM)
827 * Check the arguments are allowable
830 if (optlen < sizeof(struct in_addr))
834 if (optlen >= sizeof(struct ip_mreqn)) {
835 if (copy_from_user(&mreq, optval, sizeof(mreq)))
838 memset(&mreq, 0, sizeof(mreq));
839 if (optlen >= sizeof(struct ip_mreq)) {
840 if (copy_from_user(&mreq, optval,
841 sizeof(struct ip_mreq)))
843 } else if (optlen >= sizeof(struct in_addr)) {
844 if (copy_from_user(&mreq.imr_address, optval,
845 sizeof(struct in_addr)))
850 if (!mreq.imr_ifindex) {
851 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
857 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
859 mreq.imr_ifindex = dev->ifindex;
861 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
864 err = -EADDRNOTAVAIL;
870 if (sk->sk_bound_dev_if &&
871 mreq.imr_ifindex != sk->sk_bound_dev_if)
874 inet->mc_index = mreq.imr_ifindex;
875 inet->mc_addr = mreq.imr_address.s_addr;
880 case IP_ADD_MEMBERSHIP:
881 case IP_DROP_MEMBERSHIP:
883 struct ip_mreqn mreq;
886 if (inet_sk(sk)->is_icsk)
889 if (optlen < sizeof(struct ip_mreq))
892 if (optlen >= sizeof(struct ip_mreqn)) {
893 if (copy_from_user(&mreq, optval, sizeof(mreq)))
896 memset(&mreq, 0, sizeof(mreq));
897 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
901 if (optname == IP_ADD_MEMBERSHIP)
902 err = ip_mc_join_group(sk, &mreq);
904 err = ip_mc_leave_group(sk, &mreq);
909 struct ip_msfilter *msf;
911 if (optlen < IP_MSFILTER_SIZE(0))
913 if (optlen > sysctl_optmem_max) {
917 msf = kmalloc(optlen, GFP_KERNEL);
923 if (copy_from_user(msf, optval, optlen)) {
927 /* numsrc >= (1G-4) overflow in 32 bits */
928 if (msf->imsf_numsrc >= 0x3ffffffcU ||
929 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
934 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
939 err = ip_mc_msfilter(sk, msf, 0);
943 case IP_BLOCK_SOURCE:
944 case IP_UNBLOCK_SOURCE:
945 case IP_ADD_SOURCE_MEMBERSHIP:
946 case IP_DROP_SOURCE_MEMBERSHIP:
948 struct ip_mreq_source mreqs;
951 if (optlen != sizeof(struct ip_mreq_source))
953 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
957 if (optname == IP_BLOCK_SOURCE) {
958 omode = MCAST_EXCLUDE;
960 } else if (optname == IP_UNBLOCK_SOURCE) {
961 omode = MCAST_EXCLUDE;
963 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
964 struct ip_mreqn mreq;
966 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
967 mreq.imr_address.s_addr = mreqs.imr_interface;
968 mreq.imr_ifindex = 0;
969 err = ip_mc_join_group(sk, &mreq);
970 if (err && err != -EADDRINUSE)
972 omode = MCAST_INCLUDE;
974 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
975 omode = MCAST_INCLUDE;
978 err = ip_mc_source(add, omode, sk, &mreqs, 0);
981 case MCAST_JOIN_GROUP:
982 case MCAST_LEAVE_GROUP:
984 struct group_req greq;
985 struct sockaddr_in *psin;
986 struct ip_mreqn mreq;
988 if (optlen < sizeof(struct group_req))
991 if (copy_from_user(&greq, optval, sizeof(greq)))
993 psin = (struct sockaddr_in *)&greq.gr_group;
994 if (psin->sin_family != AF_INET)
996 memset(&mreq, 0, sizeof(mreq));
997 mreq.imr_multiaddr = psin->sin_addr;
998 mreq.imr_ifindex = greq.gr_interface;
1000 if (optname == MCAST_JOIN_GROUP)
1001 err = ip_mc_join_group(sk, &mreq);
1003 err = ip_mc_leave_group(sk, &mreq);
1006 case MCAST_JOIN_SOURCE_GROUP:
1007 case MCAST_LEAVE_SOURCE_GROUP:
1008 case MCAST_BLOCK_SOURCE:
1009 case MCAST_UNBLOCK_SOURCE:
1011 struct group_source_req greqs;
1012 struct ip_mreq_source mreqs;
1013 struct sockaddr_in *psin;
1016 if (optlen != sizeof(struct group_source_req))
1018 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1022 if (greqs.gsr_group.ss_family != AF_INET ||
1023 greqs.gsr_source.ss_family != AF_INET) {
1024 err = -EADDRNOTAVAIL;
1027 psin = (struct sockaddr_in *)&greqs.gsr_group;
1028 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1029 psin = (struct sockaddr_in *)&greqs.gsr_source;
1030 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1031 mreqs.imr_interface = 0; /* use index for mc_source */
1033 if (optname == MCAST_BLOCK_SOURCE) {
1034 omode = MCAST_EXCLUDE;
1036 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1037 omode = MCAST_EXCLUDE;
1039 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1040 struct ip_mreqn mreq;
1042 psin = (struct sockaddr_in *)&greqs.gsr_group;
1043 mreq.imr_multiaddr = psin->sin_addr;
1044 mreq.imr_address.s_addr = 0;
1045 mreq.imr_ifindex = greqs.gsr_interface;
1046 err = ip_mc_join_group(sk, &mreq);
1047 if (err && err != -EADDRINUSE)
1049 greqs.gsr_interface = mreq.imr_ifindex;
1050 omode = MCAST_INCLUDE;
1052 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1053 omode = MCAST_INCLUDE;
1056 err = ip_mc_source(add, omode, sk, &mreqs,
1057 greqs.gsr_interface);
1060 case MCAST_MSFILTER:
1062 struct sockaddr_in *psin;
1063 struct ip_msfilter *msf = NULL;
1064 struct group_filter *gsf = NULL;
1065 int msize, i, ifindex;
1067 if (optlen < GROUP_FILTER_SIZE(0))
1069 if (optlen > sysctl_optmem_max) {
1073 gsf = kmalloc(optlen, GFP_KERNEL);
1079 if (copy_from_user(gsf, optval, optlen))
1082 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1083 if (gsf->gf_numsrc >= 0x1ffffff ||
1084 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1088 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1092 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1093 msf = kmalloc(msize, GFP_KERNEL);
1098 ifindex = gsf->gf_interface;
1099 psin = (struct sockaddr_in *)&gsf->gf_group;
1100 if (psin->sin_family != AF_INET) {
1101 err = -EADDRNOTAVAIL;
1104 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1105 msf->imsf_interface = 0;
1106 msf->imsf_fmode = gsf->gf_fmode;
1107 msf->imsf_numsrc = gsf->gf_numsrc;
1108 err = -EADDRNOTAVAIL;
1109 for (i = 0; i < gsf->gf_numsrc; ++i) {
1110 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1112 if (psin->sin_family != AF_INET)
1114 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1119 err = ip_mc_msfilter(sk, msf, ifindex);
1125 case IP_MULTICAST_ALL:
1128 if (val != 0 && val != 1)
1132 case IP_ROUTER_ALERT:
1133 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1139 inet->freebind = !!val;
1142 case IP_IPSEC_POLICY:
1143 case IP_XFRM_POLICY:
1145 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1147 err = xfrm_user_policy(sk, optname, optval, optlen);
1150 case IP_TRANSPARENT:
1151 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1152 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1158 inet->transparent = !!val;
1164 if (val < 0 || val > 255)
1166 inet->min_ttl = val;
1186 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1190 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1191 * destination in skb->cb[] before dst drop.
1192 * This way, receiver doesn't make cache line misses to read rtable.
1194 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1196 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1197 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1200 if (prepare && skb_rtable(skb)) {
1201 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1202 * which has interface index (iif) as the first member of the
1203 * underlying inet{6}_skb_parm struct. This code then overlays
1204 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1205 * element so the iif is picked up from the prior IPCB
1207 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1209 pktinfo->ipi_ifindex = 0;
1210 pktinfo->ipi_spec_dst.s_addr = 0;
1215 int ip_setsockopt(struct sock *sk, int level,
1216 int optname, char __user *optval, unsigned int optlen)
1220 if (level != SOL_IP)
1221 return -ENOPROTOOPT;
1223 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1224 #ifdef CONFIG_NETFILTER
1225 /* we need to exclude all possible ENOPROTOOPTs except default case */
1226 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1227 optname != IP_IPSEC_POLICY &&
1228 optname != IP_XFRM_POLICY &&
1229 !ip_mroute_opt(optname)) {
1231 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1237 EXPORT_SYMBOL(ip_setsockopt);
1239 #ifdef CONFIG_COMPAT
1240 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1241 char __user *optval, unsigned int optlen)
1245 if (level != SOL_IP)
1246 return -ENOPROTOOPT;
1248 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1249 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1252 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1253 #ifdef CONFIG_NETFILTER
1254 /* we need to exclude all possible ENOPROTOOPTs except default case */
1255 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1256 optname != IP_IPSEC_POLICY &&
1257 optname != IP_XFRM_POLICY &&
1258 !ip_mroute_opt(optname)) {
1260 err = compat_nf_setsockopt(sk, PF_INET, optname,
1267 EXPORT_SYMBOL(compat_ip_setsockopt);
1271 * Get the options. Note for future reference. The GET of IP options gets
1272 * the _received_ ones. The set sets the _sent_ ones.
1275 static bool getsockopt_needs_rtnl(int optname)
1279 case MCAST_MSFILTER:
1285 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1286 char __user *optval, int __user *optlen, unsigned int flags)
1288 struct inet_sock *inet = inet_sk(sk);
1289 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1293 if (level != SOL_IP)
1296 if (ip_mroute_opt(optname))
1297 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1299 if (get_user(len, optlen))
1311 unsigned char optbuf[sizeof(struct ip_options)+40];
1312 struct ip_options *opt = (struct ip_options *)optbuf;
1313 struct ip_options_rcu *inet_opt;
1315 inet_opt = rcu_dereference_protected(inet->inet_opt,
1316 lockdep_sock_is_held(sk));
1319 memcpy(optbuf, &inet_opt->opt,
1320 sizeof(struct ip_options) +
1321 inet_opt->opt.optlen);
1324 if (opt->optlen == 0)
1325 return put_user(0, optlen);
1327 ip_options_undo(opt);
1329 len = min_t(unsigned int, len, opt->optlen);
1330 if (put_user(len, optlen))
1332 if (copy_to_user(optval, opt->__data, len))
1337 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1340 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1343 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1346 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1349 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1352 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1354 case IP_RECVORIGDSTADDR:
1355 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1358 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1365 struct net *net = sock_net(sk);
1366 val = (inet->uc_ttl == -1 ?
1367 net->ipv4.sysctl_ip_default_ttl :
1372 val = inet->hdrincl;
1375 val = inet->nodefrag;
1377 case IP_BIND_ADDRESS_NO_PORT:
1378 val = inet->bind_address_no_port;
1380 case IP_MTU_DISCOVER:
1381 val = inet->pmtudisc;
1385 struct dst_entry *dst;
1387 dst = sk_dst_get(sk);
1399 val = inet->recverr;
1401 case IP_MULTICAST_TTL:
1404 case IP_MULTICAST_LOOP:
1405 val = inet->mc_loop;
1408 val = (__force int)htonl((__u32) inet->uc_index);
1410 case IP_MULTICAST_IF:
1412 struct in_addr addr;
1413 len = min_t(unsigned int, len, sizeof(struct in_addr));
1414 addr.s_addr = inet->mc_addr;
1417 if (put_user(len, optlen))
1419 if (copy_to_user(optval, &addr, len))
1425 struct ip_msfilter msf;
1427 if (len < IP_MSFILTER_SIZE(0)) {
1431 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1435 err = ip_mc_msfget(sk, &msf,
1436 (struct ip_msfilter __user *)optval, optlen);
1439 case MCAST_MSFILTER:
1441 struct group_filter gsf;
1443 if (len < GROUP_FILTER_SIZE(0)) {
1447 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1451 err = ip_mc_gsfget(sk, &gsf,
1452 (struct group_filter __user *)optval,
1456 case IP_MULTICAST_ALL:
1465 if (sk->sk_type != SOCK_STREAM)
1466 return -ENOPROTOOPT;
1468 msg.msg_control = (__force void *) optval;
1469 msg.msg_controllen = len;
1470 msg.msg_flags = flags;
1472 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1473 struct in_pktinfo info;
1475 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1476 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1477 info.ipi_ifindex = inet->mc_index;
1478 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1480 if (inet->cmsg_flags & IP_CMSG_TTL) {
1481 int hlim = inet->mc_ttl;
1482 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1484 if (inet->cmsg_flags & IP_CMSG_TOS) {
1485 int tos = inet->rcv_tos;
1486 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1488 len -= msg.msg_controllen;
1489 return put_user(len, optlen);
1492 val = inet->freebind;
1494 case IP_TRANSPARENT:
1495 val = inet->transparent;
1498 val = inet->min_ttl;
1502 return -ENOPROTOOPT;
1506 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1507 unsigned char ucval = (unsigned char)val;
1509 if (put_user(len, optlen))
1511 if (copy_to_user(optval, &ucval, 1))
1514 len = min_t(unsigned int, sizeof(int), len);
1515 if (put_user(len, optlen))
1517 if (copy_to_user(optval, &val, len))
1529 int ip_getsockopt(struct sock *sk, int level,
1530 int optname, char __user *optval, int __user *optlen)
1534 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1535 #ifdef CONFIG_NETFILTER
1536 /* we need to exclude all possible ENOPROTOOPTs except default case */
1537 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1538 !ip_mroute_opt(optname)) {
1541 if (get_user(len, optlen))
1545 err = nf_getsockopt(sk, PF_INET, optname, optval,
1549 err = put_user(len, optlen);
1555 EXPORT_SYMBOL(ip_getsockopt);
1557 #ifdef CONFIG_COMPAT
1558 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1559 char __user *optval, int __user *optlen)
1563 if (optname == MCAST_MSFILTER)
1564 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1567 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1570 #ifdef CONFIG_NETFILTER
1571 /* we need to exclude all possible ENOPROTOOPTs except default case */
1572 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1573 !ip_mroute_opt(optname)) {
1576 if (get_user(len, optlen))
1580 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1583 err = put_user(len, optlen);
1589 EXPORT_SYMBOL(compat_ip_getsockopt);