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));
253 /* Our caller is responsible for freeing ipc->opt */
254 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
255 err < 40 ? err : 40);
261 struct in_pktinfo *info;
262 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
264 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
265 ipc->oif = info->ipi_ifindex;
266 ipc->addr = info->ipi_spec_dst.s_addr;
270 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
272 val = *(int *)CMSG_DATA(cmsg);
273 if (val < 1 || val > 255)
278 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
280 val = *(int *)CMSG_DATA(cmsg);
281 if (val < 0 || val > 255)
284 ipc->priority = rt_tos2priority(ipc->tos);
295 /* Special input handler for packets caught by router alert option.
296 They are selected only by protocol field, and then processed likely
297 local ones; but only if someone wants them! Otherwise, router
298 not running rsvpd will kill RSVP.
300 It is user level problem, what it will make with them.
301 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
302 but receiver should be enough clever f.e. to forward mtrace requests,
303 sent to multicast group to reach destination designated router.
305 struct ip_ra_chain __rcu *ip_ra_chain;
306 static DEFINE_SPINLOCK(ip_ra_lock);
309 static void ip_ra_destroy_rcu(struct rcu_head *head)
311 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
313 sock_put(ra->saved_sk);
317 int ip_ra_control(struct sock *sk, unsigned char on,
318 void (*destructor)(struct sock *))
320 struct ip_ra_chain *ra, *new_ra;
321 struct ip_ra_chain __rcu **rap;
323 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
326 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
328 spin_lock_bh(&ip_ra_lock);
329 for (rap = &ip_ra_chain;
330 (ra = rcu_dereference_protected(*rap,
331 lockdep_is_held(&ip_ra_lock))) != NULL;
335 spin_unlock_bh(&ip_ra_lock);
339 /* dont let ip_call_ra_chain() use sk again */
341 RCU_INIT_POINTER(*rap, ra->next);
342 spin_unlock_bh(&ip_ra_lock);
347 * Delay sock_put(sk) and kfree(ra) after one rcu grace
348 * period. This guarantee ip_call_ra_chain() dont need
349 * to mess with socket refcounts.
352 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
357 spin_unlock_bh(&ip_ra_lock);
361 new_ra->destructor = destructor;
363 RCU_INIT_POINTER(new_ra->next, ra);
364 rcu_assign_pointer(*rap, new_ra);
366 spin_unlock_bh(&ip_ra_lock);
371 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
372 __be16 port, u32 info, u8 *payload)
374 struct sock_exterr_skb *serr;
376 skb = skb_clone(skb, GFP_ATOMIC);
380 serr = SKB_EXT_ERR(skb);
381 serr->ee.ee_errno = err;
382 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
383 serr->ee.ee_type = icmp_hdr(skb)->type;
384 serr->ee.ee_code = icmp_hdr(skb)->code;
386 serr->ee.ee_info = info;
387 serr->ee.ee_data = 0;
388 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
389 skb_network_header(skb);
392 if (skb_pull(skb, payload - skb->data)) {
393 skb_reset_transport_header(skb);
394 if (sock_queue_err_skb(sk, skb) == 0)
400 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
402 struct inet_sock *inet = inet_sk(sk);
403 struct sock_exterr_skb *serr;
410 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
414 skb_put(skb, sizeof(struct iphdr));
415 skb_reset_network_header(skb);
419 serr = SKB_EXT_ERR(skb);
420 serr->ee.ee_errno = err;
421 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
422 serr->ee.ee_type = 0;
423 serr->ee.ee_code = 0;
425 serr->ee.ee_info = info;
426 serr->ee.ee_data = 0;
427 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
430 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
431 skb_reset_transport_header(skb);
433 if (sock_queue_err_skb(sk, skb))
437 /* For some errors we have valid addr_offset even with zero payload and
438 * zero port. Also, addr_offset should be supported if port is set.
440 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
442 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
443 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
446 /* IPv4 supports cmsg on all imcp errors and some timestamps
448 * Timestamp code paths do not initialize the fields expected by cmsg:
449 * the PKTINFO fields in skb->cb[]. Fill those in here.
451 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
455 struct in_pktinfo *info;
457 if (ee_origin == SO_EE_ORIGIN_ICMP)
460 if (ee_origin == SO_EE_ORIGIN_LOCAL)
463 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
464 * timestamp with egress dev. Not possible for packets without dev
465 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
467 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
471 info = PKTINFO_SKB_CB(skb);
472 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
473 info->ipi_ifindex = skb->dev->ifindex;
478 * Handle MSG_ERRQUEUE
480 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
482 struct sock_exterr_skb *serr;
484 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
486 struct sock_extended_err ee;
487 struct sockaddr_in offender;
492 WARN_ON_ONCE(sk->sk_family == AF_INET6);
495 skb = sock_dequeue_err_skb(sk);
501 msg->msg_flags |= MSG_TRUNC;
504 err = skb_copy_datagram_msg(skb, 0, msg, copied);
508 sock_recv_timestamp(msg, sk, skb);
510 serr = SKB_EXT_ERR(skb);
512 if (sin && ipv4_datagram_support_addr(serr)) {
513 sin->sin_family = AF_INET;
514 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
516 sin->sin_port = serr->port;
517 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
518 *addr_len = sizeof(*sin);
521 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
522 sin = &errhdr.offender;
523 memset(sin, 0, sizeof(*sin));
525 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
526 sin->sin_family = AF_INET;
527 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
528 if (inet_sk(sk)->cmsg_flags)
529 ip_cmsg_recv(msg, skb);
532 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
534 /* Now we could try to dump offended packet options */
536 msg->msg_flags |= MSG_ERRQUEUE;
547 * Socket option code for IP. This is the end of the line after any
548 * TCP,UDP etc options on an IP socket.
550 static bool setsockopt_needs_rtnl(int optname)
553 case IP_ADD_MEMBERSHIP:
554 case IP_ADD_SOURCE_MEMBERSHIP:
555 case IP_BLOCK_SOURCE:
556 case IP_DROP_MEMBERSHIP:
557 case IP_DROP_SOURCE_MEMBERSHIP:
559 case IP_UNBLOCK_SOURCE:
560 case MCAST_BLOCK_SOURCE:
562 case MCAST_JOIN_GROUP:
563 case MCAST_JOIN_SOURCE_GROUP:
564 case MCAST_LEAVE_GROUP:
565 case MCAST_LEAVE_SOURCE_GROUP:
566 case MCAST_UNBLOCK_SOURCE:
572 static int do_ip_setsockopt(struct sock *sk, int level,
573 int optname, char __user *optval, unsigned int optlen)
575 struct inet_sock *inet = inet_sk(sk);
577 bool needs_rtnl = setsockopt_needs_rtnl(optname);
588 case IP_MTU_DISCOVER:
590 case IP_ROUTER_ALERT:
596 case IP_BIND_ADDRESS_NO_PORT:
598 case IP_MULTICAST_TTL:
599 case IP_MULTICAST_ALL:
600 case IP_MULTICAST_LOOP:
601 case IP_RECVORIGDSTADDR:
603 if (optlen >= sizeof(int)) {
604 if (get_user(val, (int __user *) optval))
606 } else if (optlen >= sizeof(char)) {
609 if (get_user(ucval, (unsigned char __user *) optval))
615 /* If optlen==0, it is equivalent to val == 0 */
617 if (ip_mroute_opt(optname))
618 return ip_mroute_setsockopt(sk, optname, optval, optlen);
628 struct ip_options_rcu *old, *opt = NULL;
632 err = ip_options_get_from_user(sock_net(sk), &opt,
636 old = rcu_dereference_protected(inet->inet_opt,
637 sock_owned_by_user(sk));
639 struct inet_connection_sock *icsk = inet_csk(sk);
640 #if IS_ENABLED(CONFIG_IPV6)
641 if (sk->sk_family == PF_INET ||
642 (!((1 << sk->sk_state) &
643 (TCPF_LISTEN | TCPF_CLOSE)) &&
644 inet->inet_daddr != LOOPBACK4_IPV6)) {
647 icsk->icsk_ext_hdr_len -= old->opt.optlen;
649 icsk->icsk_ext_hdr_len += opt->opt.optlen;
650 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
651 #if IS_ENABLED(CONFIG_IPV6)
655 rcu_assign_pointer(inet->inet_opt, opt);
662 inet->cmsg_flags |= IP_CMSG_PKTINFO;
664 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
668 inet->cmsg_flags |= IP_CMSG_TTL;
670 inet->cmsg_flags &= ~IP_CMSG_TTL;
674 inet->cmsg_flags |= IP_CMSG_TOS;
676 inet->cmsg_flags &= ~IP_CMSG_TOS;
680 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
682 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
686 inet->cmsg_flags |= IP_CMSG_RETOPTS;
688 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
692 inet->cmsg_flags |= IP_CMSG_PASSSEC;
694 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
696 case IP_RECVORIGDSTADDR:
698 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
700 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
704 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
705 inet_inc_convert_csum(sk);
706 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
709 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
710 inet_dec_convert_csum(sk);
711 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
715 case IP_TOS: /* This sets both TOS and Precedence */
716 if (sk->sk_type == SOCK_STREAM) {
717 val &= ~INET_ECN_MASK;
718 val |= inet->tos & INET_ECN_MASK;
720 if (inet->tos != val) {
722 sk->sk_priority = rt_tos2priority(val);
729 if (val != -1 && (val < 1 || val > 255))
734 if (sk->sk_type != SOCK_RAW) {
738 inet->hdrincl = val ? 1 : 0;
741 if (sk->sk_type != SOCK_RAW) {
745 inet->nodefrag = val ? 1 : 0;
747 case IP_BIND_ADDRESS_NO_PORT:
748 inet->bind_address_no_port = val ? 1 : 0;
750 case IP_MTU_DISCOVER:
751 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
753 inet->pmtudisc = val;
756 inet->recverr = !!val;
758 skb_queue_purge(&sk->sk_error_queue);
760 case IP_MULTICAST_TTL:
761 if (sk->sk_type == SOCK_STREAM)
767 if (val < 0 || val > 255)
771 case IP_MULTICAST_LOOP:
774 inet->mc_loop = !!val;
778 struct net_device *dev = NULL;
781 if (optlen != sizeof(int))
784 ifindex = (__force int)ntohl((__force __be32)val);
791 dev = dev_get_by_index(sock_net(sk), ifindex);
792 err = -EADDRNOTAVAIL;
798 if (sk->sk_bound_dev_if)
801 inet->uc_index = ifindex;
805 case IP_MULTICAST_IF:
807 struct ip_mreqn mreq;
808 struct net_device *dev = NULL;
810 if (sk->sk_type == SOCK_STREAM)
813 * Check the arguments are allowable
816 if (optlen < sizeof(struct in_addr))
820 if (optlen >= sizeof(struct ip_mreqn)) {
821 if (copy_from_user(&mreq, optval, sizeof(mreq)))
824 memset(&mreq, 0, sizeof(mreq));
825 if (optlen >= sizeof(struct ip_mreq)) {
826 if (copy_from_user(&mreq, optval,
827 sizeof(struct ip_mreq)))
829 } else if (optlen >= sizeof(struct in_addr)) {
830 if (copy_from_user(&mreq.imr_address, optval,
831 sizeof(struct in_addr)))
836 if (!mreq.imr_ifindex) {
837 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
843 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
845 mreq.imr_ifindex = dev->ifindex;
847 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
850 err = -EADDRNOTAVAIL;
856 if (sk->sk_bound_dev_if &&
857 mreq.imr_ifindex != sk->sk_bound_dev_if)
860 inet->mc_index = mreq.imr_ifindex;
861 inet->mc_addr = mreq.imr_address.s_addr;
866 case IP_ADD_MEMBERSHIP:
867 case IP_DROP_MEMBERSHIP:
869 struct ip_mreqn mreq;
872 if (inet_sk(sk)->is_icsk)
875 if (optlen < sizeof(struct ip_mreq))
878 if (optlen >= sizeof(struct ip_mreqn)) {
879 if (copy_from_user(&mreq, optval, sizeof(mreq)))
882 memset(&mreq, 0, sizeof(mreq));
883 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
887 if (optname == IP_ADD_MEMBERSHIP)
888 err = ip_mc_join_group(sk, &mreq);
890 err = ip_mc_leave_group(sk, &mreq);
895 struct ip_msfilter *msf;
897 if (optlen < IP_MSFILTER_SIZE(0))
899 if (optlen > sysctl_optmem_max) {
903 msf = kmalloc(optlen, GFP_KERNEL);
909 if (copy_from_user(msf, optval, optlen)) {
913 /* numsrc >= (1G-4) overflow in 32 bits */
914 if (msf->imsf_numsrc >= 0x3ffffffcU ||
915 msf->imsf_numsrc > sysctl_igmp_max_msf) {
920 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
925 err = ip_mc_msfilter(sk, msf, 0);
929 case IP_BLOCK_SOURCE:
930 case IP_UNBLOCK_SOURCE:
931 case IP_ADD_SOURCE_MEMBERSHIP:
932 case IP_DROP_SOURCE_MEMBERSHIP:
934 struct ip_mreq_source mreqs;
937 if (optlen != sizeof(struct ip_mreq_source))
939 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
943 if (optname == IP_BLOCK_SOURCE) {
944 omode = MCAST_EXCLUDE;
946 } else if (optname == IP_UNBLOCK_SOURCE) {
947 omode = MCAST_EXCLUDE;
949 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
950 struct ip_mreqn mreq;
952 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
953 mreq.imr_address.s_addr = mreqs.imr_interface;
954 mreq.imr_ifindex = 0;
955 err = ip_mc_join_group(sk, &mreq);
956 if (err && err != -EADDRINUSE)
958 omode = MCAST_INCLUDE;
960 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
961 omode = MCAST_INCLUDE;
964 err = ip_mc_source(add, omode, sk, &mreqs, 0);
967 case MCAST_JOIN_GROUP:
968 case MCAST_LEAVE_GROUP:
970 struct group_req greq;
971 struct sockaddr_in *psin;
972 struct ip_mreqn mreq;
974 if (optlen < sizeof(struct group_req))
977 if (copy_from_user(&greq, optval, sizeof(greq)))
979 psin = (struct sockaddr_in *)&greq.gr_group;
980 if (psin->sin_family != AF_INET)
982 memset(&mreq, 0, sizeof(mreq));
983 mreq.imr_multiaddr = psin->sin_addr;
984 mreq.imr_ifindex = greq.gr_interface;
986 if (optname == MCAST_JOIN_GROUP)
987 err = ip_mc_join_group(sk, &mreq);
989 err = ip_mc_leave_group(sk, &mreq);
992 case MCAST_JOIN_SOURCE_GROUP:
993 case MCAST_LEAVE_SOURCE_GROUP:
994 case MCAST_BLOCK_SOURCE:
995 case MCAST_UNBLOCK_SOURCE:
997 struct group_source_req greqs;
998 struct ip_mreq_source mreqs;
999 struct sockaddr_in *psin;
1002 if (optlen != sizeof(struct group_source_req))
1004 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1008 if (greqs.gsr_group.ss_family != AF_INET ||
1009 greqs.gsr_source.ss_family != AF_INET) {
1010 err = -EADDRNOTAVAIL;
1013 psin = (struct sockaddr_in *)&greqs.gsr_group;
1014 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1015 psin = (struct sockaddr_in *)&greqs.gsr_source;
1016 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1017 mreqs.imr_interface = 0; /* use index for mc_source */
1019 if (optname == MCAST_BLOCK_SOURCE) {
1020 omode = MCAST_EXCLUDE;
1022 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1023 omode = MCAST_EXCLUDE;
1025 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1026 struct ip_mreqn mreq;
1028 psin = (struct sockaddr_in *)&greqs.gsr_group;
1029 mreq.imr_multiaddr = psin->sin_addr;
1030 mreq.imr_address.s_addr = 0;
1031 mreq.imr_ifindex = greqs.gsr_interface;
1032 err = ip_mc_join_group(sk, &mreq);
1033 if (err && err != -EADDRINUSE)
1035 greqs.gsr_interface = mreq.imr_ifindex;
1036 omode = MCAST_INCLUDE;
1038 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1039 omode = MCAST_INCLUDE;
1042 err = ip_mc_source(add, omode, sk, &mreqs,
1043 greqs.gsr_interface);
1046 case MCAST_MSFILTER:
1048 struct sockaddr_in *psin;
1049 struct ip_msfilter *msf = NULL;
1050 struct group_filter *gsf = NULL;
1051 int msize, i, ifindex;
1053 if (optlen < GROUP_FILTER_SIZE(0))
1055 if (optlen > sysctl_optmem_max) {
1059 gsf = kmalloc(optlen, GFP_KERNEL);
1065 if (copy_from_user(gsf, optval, optlen))
1068 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1069 if (gsf->gf_numsrc >= 0x1ffffff ||
1070 gsf->gf_numsrc > sysctl_igmp_max_msf) {
1074 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1078 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1079 msf = kmalloc(msize, GFP_KERNEL);
1084 ifindex = gsf->gf_interface;
1085 psin = (struct sockaddr_in *)&gsf->gf_group;
1086 if (psin->sin_family != AF_INET) {
1087 err = -EADDRNOTAVAIL;
1090 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1091 msf->imsf_interface = 0;
1092 msf->imsf_fmode = gsf->gf_fmode;
1093 msf->imsf_numsrc = gsf->gf_numsrc;
1094 err = -EADDRNOTAVAIL;
1095 for (i = 0; i < gsf->gf_numsrc; ++i) {
1096 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1098 if (psin->sin_family != AF_INET)
1100 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1105 err = ip_mc_msfilter(sk, msf, ifindex);
1111 case IP_MULTICAST_ALL:
1114 if (val != 0 && val != 1)
1118 case IP_ROUTER_ALERT:
1119 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1125 inet->freebind = !!val;
1128 case IP_IPSEC_POLICY:
1129 case IP_XFRM_POLICY:
1131 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1133 err = xfrm_user_policy(sk, optname, optval, optlen);
1136 case IP_TRANSPARENT:
1137 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1138 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1144 inet->transparent = !!val;
1150 if (val < 0 || val > 255)
1152 inet->min_ttl = val;
1172 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1176 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1177 * destination in skb->cb[] before dst drop.
1178 * This way, receiver doesn't make cache line misses to read rtable.
1180 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1182 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1183 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1186 if (prepare && skb_rtable(skb)) {
1187 pktinfo->ipi_ifindex = inet_iif(skb);
1188 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1190 pktinfo->ipi_ifindex = 0;
1191 pktinfo->ipi_spec_dst.s_addr = 0;
1196 int ip_setsockopt(struct sock *sk, int level,
1197 int optname, char __user *optval, unsigned int optlen)
1201 if (level != SOL_IP)
1202 return -ENOPROTOOPT;
1204 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1205 #ifdef CONFIG_NETFILTER
1206 /* we need to exclude all possible ENOPROTOOPTs except default case */
1207 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1208 optname != IP_IPSEC_POLICY &&
1209 optname != IP_XFRM_POLICY &&
1210 !ip_mroute_opt(optname)) {
1212 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1218 EXPORT_SYMBOL(ip_setsockopt);
1220 #ifdef CONFIG_COMPAT
1221 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1222 char __user *optval, unsigned int optlen)
1226 if (level != SOL_IP)
1227 return -ENOPROTOOPT;
1229 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1230 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1233 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1234 #ifdef CONFIG_NETFILTER
1235 /* we need to exclude all possible ENOPROTOOPTs except default case */
1236 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1237 optname != IP_IPSEC_POLICY &&
1238 optname != IP_XFRM_POLICY &&
1239 !ip_mroute_opt(optname)) {
1241 err = compat_nf_setsockopt(sk, PF_INET, optname,
1248 EXPORT_SYMBOL(compat_ip_setsockopt);
1252 * Get the options. Note for future reference. The GET of IP options gets
1253 * the _received_ ones. The set sets the _sent_ ones.
1256 static bool getsockopt_needs_rtnl(int optname)
1260 case MCAST_MSFILTER:
1266 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1267 char __user *optval, int __user *optlen, unsigned int flags)
1269 struct inet_sock *inet = inet_sk(sk);
1270 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1274 if (level != SOL_IP)
1277 if (ip_mroute_opt(optname))
1278 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1280 if (get_user(len, optlen))
1292 unsigned char optbuf[sizeof(struct ip_options)+40];
1293 struct ip_options *opt = (struct ip_options *)optbuf;
1294 struct ip_options_rcu *inet_opt;
1296 inet_opt = rcu_dereference_protected(inet->inet_opt,
1297 sock_owned_by_user(sk));
1300 memcpy(optbuf, &inet_opt->opt,
1301 sizeof(struct ip_options) +
1302 inet_opt->opt.optlen);
1305 if (opt->optlen == 0)
1306 return put_user(0, optlen);
1308 ip_options_undo(opt);
1310 len = min_t(unsigned int, len, opt->optlen);
1311 if (put_user(len, optlen))
1313 if (copy_to_user(optval, opt->__data, len))
1318 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1321 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1324 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1327 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1330 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1333 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1335 case IP_RECVORIGDSTADDR:
1336 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1339 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1345 val = (inet->uc_ttl == -1 ?
1346 sysctl_ip_default_ttl :
1350 val = inet->hdrincl;
1353 val = inet->nodefrag;
1355 case IP_BIND_ADDRESS_NO_PORT:
1356 val = inet->bind_address_no_port;
1358 case IP_MTU_DISCOVER:
1359 val = inet->pmtudisc;
1363 struct dst_entry *dst;
1365 dst = sk_dst_get(sk);
1377 val = inet->recverr;
1379 case IP_MULTICAST_TTL:
1382 case IP_MULTICAST_LOOP:
1383 val = inet->mc_loop;
1386 val = (__force int)htonl((__u32) inet->uc_index);
1388 case IP_MULTICAST_IF:
1390 struct in_addr addr;
1391 len = min_t(unsigned int, len, sizeof(struct in_addr));
1392 addr.s_addr = inet->mc_addr;
1395 if (put_user(len, optlen))
1397 if (copy_to_user(optval, &addr, len))
1403 struct ip_msfilter msf;
1405 if (len < IP_MSFILTER_SIZE(0)) {
1409 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1413 err = ip_mc_msfget(sk, &msf,
1414 (struct ip_msfilter __user *)optval, optlen);
1417 case MCAST_MSFILTER:
1419 struct group_filter gsf;
1421 if (len < GROUP_FILTER_SIZE(0)) {
1425 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1429 err = ip_mc_gsfget(sk, &gsf,
1430 (struct group_filter __user *)optval,
1434 case IP_MULTICAST_ALL:
1443 if (sk->sk_type != SOCK_STREAM)
1444 return -ENOPROTOOPT;
1446 msg.msg_control = (__force void *) optval;
1447 msg.msg_controllen = len;
1448 msg.msg_flags = flags;
1450 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1451 struct in_pktinfo info;
1453 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1454 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1455 info.ipi_ifindex = inet->mc_index;
1456 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1458 if (inet->cmsg_flags & IP_CMSG_TTL) {
1459 int hlim = inet->mc_ttl;
1460 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1462 if (inet->cmsg_flags & IP_CMSG_TOS) {
1463 int tos = inet->rcv_tos;
1464 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1466 len -= msg.msg_controllen;
1467 return put_user(len, optlen);
1470 val = inet->freebind;
1472 case IP_TRANSPARENT:
1473 val = inet->transparent;
1476 val = inet->min_ttl;
1480 return -ENOPROTOOPT;
1484 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1485 unsigned char ucval = (unsigned char)val;
1487 if (put_user(len, optlen))
1489 if (copy_to_user(optval, &ucval, 1))
1492 len = min_t(unsigned int, sizeof(int), len);
1493 if (put_user(len, optlen))
1495 if (copy_to_user(optval, &val, len))
1507 int ip_getsockopt(struct sock *sk, int level,
1508 int optname, char __user *optval, int __user *optlen)
1512 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1513 #ifdef CONFIG_NETFILTER
1514 /* we need to exclude all possible ENOPROTOOPTs except default case */
1515 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1516 !ip_mroute_opt(optname)) {
1519 if (get_user(len, optlen))
1523 err = nf_getsockopt(sk, PF_INET, optname, optval,
1527 err = put_user(len, optlen);
1533 EXPORT_SYMBOL(ip_getsockopt);
1535 #ifdef CONFIG_COMPAT
1536 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1537 char __user *optval, int __user *optlen)
1541 if (optname == MCAST_MSFILTER)
1542 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1545 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1548 #ifdef CONFIG_NETFILTER
1549 /* we need to exclude all possible ENOPROTOOPTs except default case */
1550 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1551 !ip_mroute_opt(optname)) {
1554 if (get_user(len, optlen))
1558 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1561 err = put_user(len, optlen);
1567 EXPORT_SYMBOL(compat_ip_getsockopt);