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 <linux/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_fragsize(struct msghdr *msg, struct sk_buff *skb)
104 if (IPCB(skb)->frag_max_size == 0)
107 val = IPCB(skb)->frag_max_size;
108 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
111 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
112 int tlen, int offset)
114 __wsum csum = skb->csum;
116 if (skb->ip_summed != CHECKSUM_COMPLETE)
120 int tend_off = skb_transport_offset(skb) + tlen;
121 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
124 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
127 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
133 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
137 err = security_secid_to_secctx(secid, &secdata, &seclen);
141 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
142 security_release_secctx(secdata, seclen);
145 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
147 struct sockaddr_in sin;
148 const struct iphdr *iph = ip_hdr(skb);
149 __be16 *ports = (__be16 *)skb_transport_header(skb);
151 if (skb_transport_offset(skb) + 4 > (int)skb->len)
154 /* All current transport protocols have the port numbers in the
155 * first four bytes of the transport header and this function is
156 * written with this assumption in mind.
159 sin.sin_family = AF_INET;
160 sin.sin_addr.s_addr = iph->daddr;
161 sin.sin_port = ports[1];
162 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
164 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
167 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
168 struct sk_buff *skb, int tlen, int offset)
170 struct inet_sock *inet = inet_sk(sk);
171 unsigned int flags = inet->cmsg_flags;
173 /* Ordered by supposed usage frequency */
174 if (flags & IP_CMSG_PKTINFO) {
175 ip_cmsg_recv_pktinfo(msg, skb);
177 flags &= ~IP_CMSG_PKTINFO;
182 if (flags & IP_CMSG_TTL) {
183 ip_cmsg_recv_ttl(msg, skb);
185 flags &= ~IP_CMSG_TTL;
190 if (flags & IP_CMSG_TOS) {
191 ip_cmsg_recv_tos(msg, skb);
193 flags &= ~IP_CMSG_TOS;
198 if (flags & IP_CMSG_RECVOPTS) {
199 ip_cmsg_recv_opts(msg, skb);
201 flags &= ~IP_CMSG_RECVOPTS;
206 if (flags & IP_CMSG_RETOPTS) {
207 ip_cmsg_recv_retopts(msg, skb);
209 flags &= ~IP_CMSG_RETOPTS;
214 if (flags & IP_CMSG_PASSSEC) {
215 ip_cmsg_recv_security(msg, skb);
217 flags &= ~IP_CMSG_PASSSEC;
222 if (flags & IP_CMSG_ORIGDSTADDR) {
223 ip_cmsg_recv_dstaddr(msg, skb);
225 flags &= ~IP_CMSG_ORIGDSTADDR;
230 if (flags & IP_CMSG_CHECKSUM)
231 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
233 if (flags & IP_CMSG_RECVFRAGSIZE)
234 ip_cmsg_recv_fragsize(msg, skb);
236 EXPORT_SYMBOL(ip_cmsg_recv_offset);
238 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
242 struct cmsghdr *cmsg;
243 struct net *net = sock_net(sk);
245 for_each_cmsghdr(cmsg, msg) {
246 if (!CMSG_OK(msg, cmsg))
248 #if IS_ENABLED(CONFIG_IPV6)
250 cmsg->cmsg_level == SOL_IPV6 &&
251 cmsg->cmsg_type == IPV6_PKTINFO) {
252 struct in6_pktinfo *src_info;
254 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
256 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
257 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
259 ipc->oif = src_info->ipi6_ifindex;
260 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
264 if (cmsg->cmsg_level == SOL_SOCKET) {
265 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
271 if (cmsg->cmsg_level != SOL_IP)
273 switch (cmsg->cmsg_type) {
275 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
277 /* Our caller is responsible for freeing ipc->opt */
278 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
279 err < 40 ? err : 40);
285 struct in_pktinfo *info;
286 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
288 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
289 ipc->oif = info->ipi_ifindex;
290 ipc->addr = info->ipi_spec_dst.s_addr;
294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
296 val = *(int *)CMSG_DATA(cmsg);
297 if (val < 1 || val > 255)
302 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
303 val = *(int *)CMSG_DATA(cmsg);
304 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
305 val = *(u8 *)CMSG_DATA(cmsg);
308 if (val < 0 || val > 255)
311 ipc->priority = rt_tos2priority(ipc->tos);
322 /* Special input handler for packets caught by router alert option.
323 They are selected only by protocol field, and then processed likely
324 local ones; but only if someone wants them! Otherwise, router
325 not running rsvpd will kill RSVP.
327 It is user level problem, what it will make with them.
328 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
329 but receiver should be enough clever f.e. to forward mtrace requests,
330 sent to multicast group to reach destination designated router.
332 struct ip_ra_chain __rcu *ip_ra_chain;
335 static void ip_ra_destroy_rcu(struct rcu_head *head)
337 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
339 sock_put(ra->saved_sk);
343 int ip_ra_control(struct sock *sk, unsigned char on,
344 void (*destructor)(struct sock *))
346 struct ip_ra_chain *ra, *new_ra;
347 struct ip_ra_chain __rcu **rap;
349 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
352 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
354 for (rap = &ip_ra_chain;
355 (ra = rtnl_dereference(*rap)) != NULL;
362 /* dont let ip_call_ra_chain() use sk again */
364 RCU_INIT_POINTER(*rap, ra->next);
369 * Delay sock_put(sk) and kfree(ra) after one rcu grace
370 * period. This guarantee ip_call_ra_chain() dont need
371 * to mess with socket refcounts.
374 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
381 new_ra->destructor = destructor;
383 RCU_INIT_POINTER(new_ra->next, ra);
384 rcu_assign_pointer(*rap, new_ra);
390 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
391 __be16 port, u32 info, u8 *payload)
393 struct sock_exterr_skb *serr;
395 skb = skb_clone(skb, GFP_ATOMIC);
399 serr = SKB_EXT_ERR(skb);
400 serr->ee.ee_errno = err;
401 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
402 serr->ee.ee_type = icmp_hdr(skb)->type;
403 serr->ee.ee_code = icmp_hdr(skb)->code;
405 serr->ee.ee_info = info;
406 serr->ee.ee_data = 0;
407 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
408 skb_network_header(skb);
411 if (skb_pull(skb, payload - skb->data)) {
412 skb_reset_transport_header(skb);
413 if (sock_queue_err_skb(sk, skb) == 0)
419 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
421 struct inet_sock *inet = inet_sk(sk);
422 struct sock_exterr_skb *serr;
429 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
433 skb_put(skb, sizeof(struct iphdr));
434 skb_reset_network_header(skb);
438 serr = SKB_EXT_ERR(skb);
439 serr->ee.ee_errno = err;
440 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
441 serr->ee.ee_type = 0;
442 serr->ee.ee_code = 0;
444 serr->ee.ee_info = info;
445 serr->ee.ee_data = 0;
446 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
449 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
450 skb_reset_transport_header(skb);
452 if (sock_queue_err_skb(sk, skb))
456 /* For some errors we have valid addr_offset even with zero payload and
457 * zero port. Also, addr_offset should be supported if port is set.
459 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
461 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
462 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
465 /* IPv4 supports cmsg on all imcp errors and some timestamps
467 * Timestamp code paths do not initialize the fields expected by cmsg:
468 * the PKTINFO fields in skb->cb[]. Fill those in here.
470 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
474 struct in_pktinfo *info;
476 if (ee_origin == SO_EE_ORIGIN_ICMP)
479 if (ee_origin == SO_EE_ORIGIN_LOCAL)
482 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
483 * timestamp with egress dev. Not possible for packets without iif
484 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
486 info = PKTINFO_SKB_CB(skb);
487 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
491 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
496 * Handle MSG_ERRQUEUE
498 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
500 struct sock_exterr_skb *serr;
502 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
504 struct sock_extended_err ee;
505 struct sockaddr_in offender;
510 WARN_ON_ONCE(sk->sk_family == AF_INET6);
513 skb = sock_dequeue_err_skb(sk);
519 msg->msg_flags |= MSG_TRUNC;
522 err = skb_copy_datagram_msg(skb, 0, msg, copied);
527 sock_recv_timestamp(msg, sk, skb);
529 serr = SKB_EXT_ERR(skb);
531 if (sin && ipv4_datagram_support_addr(serr)) {
532 sin->sin_family = AF_INET;
533 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
535 sin->sin_port = serr->port;
536 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
537 *addr_len = sizeof(*sin);
540 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
541 sin = &errhdr.offender;
542 memset(sin, 0, sizeof(*sin));
544 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
545 sin->sin_family = AF_INET;
546 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
547 if (inet_sk(sk)->cmsg_flags)
548 ip_cmsg_recv(msg, skb);
551 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
553 /* Now we could try to dump offended packet options */
555 msg->msg_flags |= MSG_ERRQUEUE;
565 * Socket option code for IP. This is the end of the line after any
566 * TCP,UDP etc options on an IP socket.
568 static bool setsockopt_needs_rtnl(int optname)
571 case IP_ADD_MEMBERSHIP:
572 case IP_ADD_SOURCE_MEMBERSHIP:
573 case IP_BLOCK_SOURCE:
574 case IP_DROP_MEMBERSHIP:
575 case IP_DROP_SOURCE_MEMBERSHIP:
577 case IP_UNBLOCK_SOURCE:
578 case MCAST_BLOCK_SOURCE:
580 case MCAST_JOIN_GROUP:
581 case MCAST_JOIN_SOURCE_GROUP:
582 case MCAST_LEAVE_GROUP:
583 case MCAST_LEAVE_SOURCE_GROUP:
584 case MCAST_UNBLOCK_SOURCE:
585 case IP_ROUTER_ALERT:
591 static int do_ip_setsockopt(struct sock *sk, int level,
592 int optname, char __user *optval, unsigned int optlen)
594 struct inet_sock *inet = inet_sk(sk);
595 struct net *net = sock_net(sk);
597 bool needs_rtnl = setsockopt_needs_rtnl(optname);
608 case IP_MTU_DISCOVER:
610 case IP_ROUTER_ALERT:
616 case IP_BIND_ADDRESS_NO_PORT:
618 case IP_MULTICAST_TTL:
619 case IP_MULTICAST_ALL:
620 case IP_MULTICAST_LOOP:
621 case IP_RECVORIGDSTADDR:
623 case IP_RECVFRAGSIZE:
624 if (optlen >= sizeof(int)) {
625 if (get_user(val, (int __user *) optval))
627 } else if (optlen >= sizeof(char)) {
630 if (get_user(ucval, (unsigned char __user *) optval))
636 /* If optlen==0, it is equivalent to val == 0 */
638 if (ip_mroute_opt(optname))
639 return ip_mroute_setsockopt(sk, optname, optval, optlen);
649 struct ip_options_rcu *old, *opt = NULL;
653 err = ip_options_get_from_user(sock_net(sk), &opt,
657 old = rcu_dereference_protected(inet->inet_opt,
658 lockdep_sock_is_held(sk));
660 struct inet_connection_sock *icsk = inet_csk(sk);
661 #if IS_ENABLED(CONFIG_IPV6)
662 if (sk->sk_family == PF_INET ||
663 (!((1 << sk->sk_state) &
664 (TCPF_LISTEN | TCPF_CLOSE)) &&
665 inet->inet_daddr != LOOPBACK4_IPV6)) {
668 icsk->icsk_ext_hdr_len -= old->opt.optlen;
670 icsk->icsk_ext_hdr_len += opt->opt.optlen;
671 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
672 #if IS_ENABLED(CONFIG_IPV6)
676 rcu_assign_pointer(inet->inet_opt, opt);
683 inet->cmsg_flags |= IP_CMSG_PKTINFO;
685 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
689 inet->cmsg_flags |= IP_CMSG_TTL;
691 inet->cmsg_flags &= ~IP_CMSG_TTL;
695 inet->cmsg_flags |= IP_CMSG_TOS;
697 inet->cmsg_flags &= ~IP_CMSG_TOS;
701 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
703 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
707 inet->cmsg_flags |= IP_CMSG_RETOPTS;
709 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
713 inet->cmsg_flags |= IP_CMSG_PASSSEC;
715 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
717 case IP_RECVORIGDSTADDR:
719 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
721 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
725 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
726 inet_inc_convert_csum(sk);
727 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
730 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
731 inet_dec_convert_csum(sk);
732 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
736 case IP_RECVFRAGSIZE:
737 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
740 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
742 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
744 case IP_TOS: /* This sets both TOS and Precedence */
745 if (sk->sk_type == SOCK_STREAM) {
746 val &= ~INET_ECN_MASK;
747 val |= inet->tos & INET_ECN_MASK;
749 if (inet->tos != val) {
751 sk->sk_priority = rt_tos2priority(val);
758 if (val != -1 && (val < 1 || val > 255))
763 if (sk->sk_type != SOCK_RAW) {
767 inet->hdrincl = val ? 1 : 0;
770 if (sk->sk_type != SOCK_RAW) {
774 inet->nodefrag = val ? 1 : 0;
776 case IP_BIND_ADDRESS_NO_PORT:
777 inet->bind_address_no_port = val ? 1 : 0;
779 case IP_MTU_DISCOVER:
780 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
782 inet->pmtudisc = val;
785 inet->recverr = !!val;
787 skb_queue_purge(&sk->sk_error_queue);
789 case IP_MULTICAST_TTL:
790 if (sk->sk_type == SOCK_STREAM)
796 if (val < 0 || val > 255)
800 case IP_MULTICAST_LOOP:
803 inet->mc_loop = !!val;
807 struct net_device *dev = NULL;
810 if (optlen != sizeof(int))
813 ifindex = (__force int)ntohl((__force __be32)val);
820 dev = dev_get_by_index(sock_net(sk), ifindex);
821 err = -EADDRNOTAVAIL;
827 if (sk->sk_bound_dev_if)
830 inet->uc_index = ifindex;
834 case IP_MULTICAST_IF:
836 struct ip_mreqn mreq;
837 struct net_device *dev = NULL;
840 if (sk->sk_type == SOCK_STREAM)
843 * Check the arguments are allowable
846 if (optlen < sizeof(struct in_addr))
850 if (optlen >= sizeof(struct ip_mreqn)) {
851 if (copy_from_user(&mreq, optval, sizeof(mreq)))
854 memset(&mreq, 0, sizeof(mreq));
855 if (optlen >= sizeof(struct ip_mreq)) {
856 if (copy_from_user(&mreq, optval,
857 sizeof(struct ip_mreq)))
859 } else if (optlen >= sizeof(struct in_addr)) {
860 if (copy_from_user(&mreq.imr_address, optval,
861 sizeof(struct in_addr)))
866 if (!mreq.imr_ifindex) {
867 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
873 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
875 mreq.imr_ifindex = dev->ifindex;
877 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
880 err = -EADDRNOTAVAIL;
884 midx = l3mdev_master_ifindex(dev);
889 if (sk->sk_bound_dev_if &&
890 mreq.imr_ifindex != sk->sk_bound_dev_if &&
891 (!midx || midx != sk->sk_bound_dev_if))
894 inet->mc_index = mreq.imr_ifindex;
895 inet->mc_addr = mreq.imr_address.s_addr;
900 case IP_ADD_MEMBERSHIP:
901 case IP_DROP_MEMBERSHIP:
903 struct ip_mreqn mreq;
906 if (inet_sk(sk)->is_icsk)
909 if (optlen < sizeof(struct ip_mreq))
912 if (optlen >= sizeof(struct ip_mreqn)) {
913 if (copy_from_user(&mreq, optval, sizeof(mreq)))
916 memset(&mreq, 0, sizeof(mreq));
917 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
921 if (optname == IP_ADD_MEMBERSHIP)
922 err = ip_mc_join_group(sk, &mreq);
924 err = ip_mc_leave_group(sk, &mreq);
929 struct ip_msfilter *msf;
931 if (optlen < IP_MSFILTER_SIZE(0))
933 if (optlen > sysctl_optmem_max) {
937 msf = kmalloc(optlen, GFP_KERNEL);
943 if (copy_from_user(msf, optval, optlen)) {
947 /* numsrc >= (1G-4) overflow in 32 bits */
948 if (msf->imsf_numsrc >= 0x3ffffffcU ||
949 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
954 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
959 err = ip_mc_msfilter(sk, msf, 0);
963 case IP_BLOCK_SOURCE:
964 case IP_UNBLOCK_SOURCE:
965 case IP_ADD_SOURCE_MEMBERSHIP:
966 case IP_DROP_SOURCE_MEMBERSHIP:
968 struct ip_mreq_source mreqs;
971 if (optlen != sizeof(struct ip_mreq_source))
973 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
977 if (optname == IP_BLOCK_SOURCE) {
978 omode = MCAST_EXCLUDE;
980 } else if (optname == IP_UNBLOCK_SOURCE) {
981 omode = MCAST_EXCLUDE;
983 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
984 struct ip_mreqn mreq;
986 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
987 mreq.imr_address.s_addr = mreqs.imr_interface;
988 mreq.imr_ifindex = 0;
989 err = ip_mc_join_group(sk, &mreq);
990 if (err && err != -EADDRINUSE)
992 omode = MCAST_INCLUDE;
994 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
995 omode = MCAST_INCLUDE;
998 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1001 case MCAST_JOIN_GROUP:
1002 case MCAST_LEAVE_GROUP:
1004 struct group_req greq;
1005 struct sockaddr_in *psin;
1006 struct ip_mreqn mreq;
1008 if (optlen < sizeof(struct group_req))
1011 if (copy_from_user(&greq, optval, sizeof(greq)))
1013 psin = (struct sockaddr_in *)&greq.gr_group;
1014 if (psin->sin_family != AF_INET)
1016 memset(&mreq, 0, sizeof(mreq));
1017 mreq.imr_multiaddr = psin->sin_addr;
1018 mreq.imr_ifindex = greq.gr_interface;
1020 if (optname == MCAST_JOIN_GROUP)
1021 err = ip_mc_join_group(sk, &mreq);
1023 err = ip_mc_leave_group(sk, &mreq);
1026 case MCAST_JOIN_SOURCE_GROUP:
1027 case MCAST_LEAVE_SOURCE_GROUP:
1028 case MCAST_BLOCK_SOURCE:
1029 case MCAST_UNBLOCK_SOURCE:
1031 struct group_source_req greqs;
1032 struct ip_mreq_source mreqs;
1033 struct sockaddr_in *psin;
1036 if (optlen != sizeof(struct group_source_req))
1038 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1042 if (greqs.gsr_group.ss_family != AF_INET ||
1043 greqs.gsr_source.ss_family != AF_INET) {
1044 err = -EADDRNOTAVAIL;
1047 psin = (struct sockaddr_in *)&greqs.gsr_group;
1048 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1049 psin = (struct sockaddr_in *)&greqs.gsr_source;
1050 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1051 mreqs.imr_interface = 0; /* use index for mc_source */
1053 if (optname == MCAST_BLOCK_SOURCE) {
1054 omode = MCAST_EXCLUDE;
1056 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1057 omode = MCAST_EXCLUDE;
1059 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1060 struct ip_mreqn mreq;
1062 psin = (struct sockaddr_in *)&greqs.gsr_group;
1063 mreq.imr_multiaddr = psin->sin_addr;
1064 mreq.imr_address.s_addr = 0;
1065 mreq.imr_ifindex = greqs.gsr_interface;
1066 err = ip_mc_join_group(sk, &mreq);
1067 if (err && err != -EADDRINUSE)
1069 greqs.gsr_interface = mreq.imr_ifindex;
1070 omode = MCAST_INCLUDE;
1072 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1073 omode = MCAST_INCLUDE;
1076 err = ip_mc_source(add, omode, sk, &mreqs,
1077 greqs.gsr_interface);
1080 case MCAST_MSFILTER:
1082 struct sockaddr_in *psin;
1083 struct ip_msfilter *msf = NULL;
1084 struct group_filter *gsf = NULL;
1085 int msize, i, ifindex;
1087 if (optlen < GROUP_FILTER_SIZE(0))
1089 if (optlen > sysctl_optmem_max) {
1093 gsf = kmalloc(optlen, GFP_KERNEL);
1099 if (copy_from_user(gsf, optval, optlen))
1102 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1103 if (gsf->gf_numsrc >= 0x1ffffff ||
1104 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1108 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1112 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1113 msf = kmalloc(msize, GFP_KERNEL);
1118 ifindex = gsf->gf_interface;
1119 psin = (struct sockaddr_in *)&gsf->gf_group;
1120 if (psin->sin_family != AF_INET) {
1121 err = -EADDRNOTAVAIL;
1124 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1125 msf->imsf_interface = 0;
1126 msf->imsf_fmode = gsf->gf_fmode;
1127 msf->imsf_numsrc = gsf->gf_numsrc;
1128 err = -EADDRNOTAVAIL;
1129 for (i = 0; i < gsf->gf_numsrc; ++i) {
1130 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1132 if (psin->sin_family != AF_INET)
1134 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1139 err = ip_mc_msfilter(sk, msf, ifindex);
1145 case IP_MULTICAST_ALL:
1148 if (val != 0 && val != 1)
1152 case IP_ROUTER_ALERT:
1153 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1159 inet->freebind = !!val;
1162 case IP_IPSEC_POLICY:
1163 case IP_XFRM_POLICY:
1165 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1167 err = xfrm_user_policy(sk, optname, optval, optlen);
1170 case IP_TRANSPARENT:
1171 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1172 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1178 inet->transparent = !!val;
1184 if (val < 0 || val > 255)
1186 inet->min_ttl = val;
1206 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1210 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1211 * destination in skb->cb[] before dst drop.
1212 * This way, receiver doesn't make cache line misses to read rtable.
1214 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1216 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1217 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1220 if (prepare && skb_rtable(skb)) {
1221 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1222 * which has interface index (iif) as the first member of the
1223 * underlying inet{6}_skb_parm struct. This code then overlays
1224 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1225 * element so the iif is picked up from the prior IPCB. If iif
1226 * is the loopback interface, then return the sending interface
1227 * (e.g., process binds socket to eth0 for Tx which is
1228 * redirected to loopback in the rtable/dst).
1230 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1231 pktinfo->ipi_ifindex = inet_iif(skb);
1233 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1235 pktinfo->ipi_ifindex = 0;
1236 pktinfo->ipi_spec_dst.s_addr = 0;
1238 /* We need to keep the dst for __ip_options_echo()
1239 * We could restrict the test to opt.ts_needtime || opt.srr,
1240 * but the following is good enough as IP options are not often used.
1242 if (unlikely(IPCB(skb)->opt.optlen))
1248 int ip_setsockopt(struct sock *sk, int level,
1249 int optname, char __user *optval, unsigned int optlen)
1253 if (level != SOL_IP)
1254 return -ENOPROTOOPT;
1256 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1257 #ifdef CONFIG_NETFILTER
1258 /* we need to exclude all possible ENOPROTOOPTs except default case */
1259 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1260 optname != IP_IPSEC_POLICY &&
1261 optname != IP_XFRM_POLICY &&
1262 !ip_mroute_opt(optname)) {
1264 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1270 EXPORT_SYMBOL(ip_setsockopt);
1272 #ifdef CONFIG_COMPAT
1273 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1274 char __user *optval, unsigned int optlen)
1278 if (level != SOL_IP)
1279 return -ENOPROTOOPT;
1281 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1282 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1285 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1286 #ifdef CONFIG_NETFILTER
1287 /* we need to exclude all possible ENOPROTOOPTs except default case */
1288 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1289 optname != IP_IPSEC_POLICY &&
1290 optname != IP_XFRM_POLICY &&
1291 !ip_mroute_opt(optname)) {
1293 err = compat_nf_setsockopt(sk, PF_INET, optname,
1300 EXPORT_SYMBOL(compat_ip_setsockopt);
1304 * Get the options. Note for future reference. The GET of IP options gets
1305 * the _received_ ones. The set sets the _sent_ ones.
1308 static bool getsockopt_needs_rtnl(int optname)
1312 case MCAST_MSFILTER:
1318 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1319 char __user *optval, int __user *optlen, unsigned int flags)
1321 struct inet_sock *inet = inet_sk(sk);
1322 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1326 if (level != SOL_IP)
1329 if (ip_mroute_opt(optname))
1330 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1332 if (get_user(len, optlen))
1344 unsigned char optbuf[sizeof(struct ip_options)+40];
1345 struct ip_options *opt = (struct ip_options *)optbuf;
1346 struct ip_options_rcu *inet_opt;
1348 inet_opt = rcu_dereference_protected(inet->inet_opt,
1349 lockdep_sock_is_held(sk));
1352 memcpy(optbuf, &inet_opt->opt,
1353 sizeof(struct ip_options) +
1354 inet_opt->opt.optlen);
1357 if (opt->optlen == 0)
1358 return put_user(0, optlen);
1360 ip_options_undo(opt);
1362 len = min_t(unsigned int, len, opt->optlen);
1363 if (put_user(len, optlen))
1365 if (copy_to_user(optval, opt->__data, len))
1370 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1373 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1376 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1379 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1382 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1385 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1387 case IP_RECVORIGDSTADDR:
1388 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1391 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1393 case IP_RECVFRAGSIZE:
1394 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1401 struct net *net = sock_net(sk);
1402 val = (inet->uc_ttl == -1 ?
1403 net->ipv4.sysctl_ip_default_ttl :
1408 val = inet->hdrincl;
1411 val = inet->nodefrag;
1413 case IP_BIND_ADDRESS_NO_PORT:
1414 val = inet->bind_address_no_port;
1416 case IP_MTU_DISCOVER:
1417 val = inet->pmtudisc;
1421 struct dst_entry *dst;
1423 dst = sk_dst_get(sk);
1435 val = inet->recverr;
1437 case IP_MULTICAST_TTL:
1440 case IP_MULTICAST_LOOP:
1441 val = inet->mc_loop;
1444 val = (__force int)htonl((__u32) inet->uc_index);
1446 case IP_MULTICAST_IF:
1448 struct in_addr addr;
1449 len = min_t(unsigned int, len, sizeof(struct in_addr));
1450 addr.s_addr = inet->mc_addr;
1453 if (put_user(len, optlen))
1455 if (copy_to_user(optval, &addr, len))
1461 struct ip_msfilter msf;
1463 if (len < IP_MSFILTER_SIZE(0)) {
1467 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1471 err = ip_mc_msfget(sk, &msf,
1472 (struct ip_msfilter __user *)optval, optlen);
1475 case MCAST_MSFILTER:
1477 struct group_filter gsf;
1479 if (len < GROUP_FILTER_SIZE(0)) {
1483 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1487 err = ip_mc_gsfget(sk, &gsf,
1488 (struct group_filter __user *)optval,
1492 case IP_MULTICAST_ALL:
1501 if (sk->sk_type != SOCK_STREAM)
1502 return -ENOPROTOOPT;
1504 msg.msg_control = (__force void *) optval;
1505 msg.msg_controllen = len;
1506 msg.msg_flags = flags;
1508 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1509 struct in_pktinfo info;
1511 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1512 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1513 info.ipi_ifindex = inet->mc_index;
1514 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1516 if (inet->cmsg_flags & IP_CMSG_TTL) {
1517 int hlim = inet->mc_ttl;
1518 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1520 if (inet->cmsg_flags & IP_CMSG_TOS) {
1521 int tos = inet->rcv_tos;
1522 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1524 len -= msg.msg_controllen;
1525 return put_user(len, optlen);
1528 val = inet->freebind;
1530 case IP_TRANSPARENT:
1531 val = inet->transparent;
1534 val = inet->min_ttl;
1538 return -ENOPROTOOPT;
1542 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1543 unsigned char ucval = (unsigned char)val;
1545 if (put_user(len, optlen))
1547 if (copy_to_user(optval, &ucval, 1))
1550 len = min_t(unsigned int, sizeof(int), len);
1551 if (put_user(len, optlen))
1553 if (copy_to_user(optval, &val, len))
1565 int ip_getsockopt(struct sock *sk, int level,
1566 int optname, char __user *optval, int __user *optlen)
1570 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1571 #ifdef CONFIG_NETFILTER
1572 /* we need to exclude all possible ENOPROTOOPTs except default case */
1573 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1574 !ip_mroute_opt(optname)) {
1577 if (get_user(len, optlen))
1581 err = nf_getsockopt(sk, PF_INET, optname, optval,
1585 err = put_user(len, optlen);
1591 EXPORT_SYMBOL(ip_getsockopt);
1593 #ifdef CONFIG_COMPAT
1594 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1595 char __user *optval, int __user *optlen)
1599 if (optname == MCAST_MSFILTER)
1600 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1603 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1606 #ifdef CONFIG_NETFILTER
1607 /* we need to exclude all possible ENOPROTOOPTs except default case */
1608 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1609 !ip_mroute_opt(optname)) {
1612 if (get_user(len, optlen))
1616 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1619 err = put_user(len, optlen);
1625 EXPORT_SYMBOL(compat_ip_getsockopt);