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 #if IS_ENABLED(CONFIG_IPV6)
41 #include <net/transp_v6.h>
43 #include <net/ip_fib.h>
45 #include <linux/errqueue.h>
46 #include <asm/uaccess.h>
48 #define IP_CMSG_PKTINFO 1
51 #define IP_CMSG_RECVOPTS 8
52 #define IP_CMSG_RETOPTS 16
53 #define IP_CMSG_PASSSEC 32
54 #define IP_CMSG_ORIGDSTADDR 64
57 * SOL_IP control messages.
59 #define PKTINFO_SKB_CB(__skb) ((struct in_pktinfo *)((__skb)->cb))
61 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
63 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
65 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
67 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
70 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
72 int ttl = ip_hdr(skb)->ttl;
73 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
76 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
78 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
81 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
83 if (IPCB(skb)->opt.optlen == 0)
86 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
91 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
93 unsigned char optbuf[sizeof(struct ip_options) + 40];
94 struct ip_options *opt = (struct ip_options *)optbuf;
96 if (IPCB(skb)->opt.optlen == 0)
99 if (ip_options_echo(opt, skb)) {
100 msg->msg_flags |= MSG_CTRUNC;
103 ip_options_undo(opt);
105 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
108 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
114 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
118 err = security_secid_to_secctx(secid, &secdata, &seclen);
122 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
123 security_release_secctx(secdata, seclen);
126 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
128 struct sockaddr_in sin;
129 const struct iphdr *iph = ip_hdr(skb);
130 __be16 *ports = (__be16 *)skb_transport_header(skb);
132 if (skb_transport_offset(skb) + 4 > skb->len)
135 /* All current transport protocols have the port numbers in the
136 * first four bytes of the transport header and this function is
137 * written with this assumption in mind.
140 sin.sin_family = AF_INET;
141 sin.sin_addr.s_addr = iph->daddr;
142 sin.sin_port = ports[1];
143 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
145 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
148 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
150 struct inet_sock *inet = inet_sk(skb->sk);
151 unsigned int flags = inet->cmsg_flags;
153 /* Ordered by supposed usage frequency */
155 ip_cmsg_recv_pktinfo(msg, skb);
156 if ((flags >>= 1) == 0)
160 ip_cmsg_recv_ttl(msg, skb);
161 if ((flags >>= 1) == 0)
165 ip_cmsg_recv_tos(msg, skb);
166 if ((flags >>= 1) == 0)
170 ip_cmsg_recv_opts(msg, skb);
171 if ((flags >>= 1) == 0)
175 ip_cmsg_recv_retopts(msg, skb);
176 if ((flags >>= 1) == 0)
180 ip_cmsg_recv_security(msg, skb);
182 if ((flags >>= 1) == 0)
185 ip_cmsg_recv_dstaddr(msg, skb);
188 EXPORT_SYMBOL(ip_cmsg_recv);
190 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
193 struct cmsghdr *cmsg;
195 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
196 if (!CMSG_OK(msg, cmsg))
198 if (cmsg->cmsg_level != SOL_IP)
200 switch (cmsg->cmsg_type) {
202 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
203 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
204 err < 40 ? err : 40);
210 struct in_pktinfo *info;
211 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
213 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
214 ipc->oif = info->ipi_ifindex;
215 ipc->addr = info->ipi_spec_dst.s_addr;
219 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
221 val = *(int *)CMSG_DATA(cmsg);
222 if (val < 1 || val > 255)
227 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
229 val = *(int *)CMSG_DATA(cmsg);
230 if (val < 0 || val > 255)
233 ipc->priority = rt_tos2priority(ipc->tos);
244 /* Special input handler for packets caught by router alert option.
245 They are selected only by protocol field, and then processed likely
246 local ones; but only if someone wants them! Otherwise, router
247 not running rsvpd will kill RSVP.
249 It is user level problem, what it will make with them.
250 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
251 but receiver should be enough clever f.e. to forward mtrace requests,
252 sent to multicast group to reach destination designated router.
254 struct ip_ra_chain __rcu *ip_ra_chain;
255 static DEFINE_SPINLOCK(ip_ra_lock);
258 static void ip_ra_destroy_rcu(struct rcu_head *head)
260 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
262 sock_put(ra->saved_sk);
266 int ip_ra_control(struct sock *sk, unsigned char on,
267 void (*destructor)(struct sock *))
269 struct ip_ra_chain *ra, *new_ra;
270 struct ip_ra_chain __rcu **rap;
272 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
275 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
277 spin_lock_bh(&ip_ra_lock);
278 for (rap = &ip_ra_chain;
279 (ra = rcu_dereference_protected(*rap,
280 lockdep_is_held(&ip_ra_lock))) != NULL;
284 spin_unlock_bh(&ip_ra_lock);
288 /* dont let ip_call_ra_chain() use sk again */
290 rcu_assign_pointer(*rap, ra->next);
291 spin_unlock_bh(&ip_ra_lock);
296 * Delay sock_put(sk) and kfree(ra) after one rcu grace
297 * period. This guarantee ip_call_ra_chain() dont need
298 * to mess with socket refcounts.
301 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
305 if (new_ra == NULL) {
306 spin_unlock_bh(&ip_ra_lock);
310 new_ra->destructor = destructor;
313 rcu_assign_pointer(*rap, new_ra);
315 spin_unlock_bh(&ip_ra_lock);
320 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
321 __be16 port, u32 info, u8 *payload)
323 struct sock_exterr_skb *serr;
325 skb = skb_clone(skb, GFP_ATOMIC);
329 serr = SKB_EXT_ERR(skb);
330 serr->ee.ee_errno = err;
331 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
332 serr->ee.ee_type = icmp_hdr(skb)->type;
333 serr->ee.ee_code = icmp_hdr(skb)->code;
335 serr->ee.ee_info = info;
336 serr->ee.ee_data = 0;
337 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
338 skb_network_header(skb);
341 if (skb_pull(skb, payload - skb->data) != NULL) {
342 skb_reset_transport_header(skb);
343 if (sock_queue_err_skb(sk, skb) == 0)
349 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
351 struct inet_sock *inet = inet_sk(sk);
352 struct sock_exterr_skb *serr;
359 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
363 skb_put(skb, sizeof(struct iphdr));
364 skb_reset_network_header(skb);
368 serr = SKB_EXT_ERR(skb);
369 serr->ee.ee_errno = err;
370 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
371 serr->ee.ee_type = 0;
372 serr->ee.ee_code = 0;
374 serr->ee.ee_info = info;
375 serr->ee.ee_data = 0;
376 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
379 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
380 skb_reset_transport_header(skb);
382 if (sock_queue_err_skb(sk, skb))
387 * Handle MSG_ERRQUEUE
389 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
391 struct sock_exterr_skb *serr;
392 struct sk_buff *skb, *skb2;
393 struct sockaddr_in *sin;
395 struct sock_extended_err ee;
396 struct sockaddr_in offender;
402 skb = skb_dequeue(&sk->sk_error_queue);
408 msg->msg_flags |= MSG_TRUNC;
411 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
415 sock_recv_timestamp(msg, sk, skb);
417 serr = SKB_EXT_ERR(skb);
419 sin = (struct sockaddr_in *)msg->msg_name;
421 sin->sin_family = AF_INET;
422 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
424 sin->sin_port = serr->port;
425 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
428 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
429 sin = &errhdr.offender;
430 sin->sin_family = AF_UNSPEC;
431 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
432 struct inet_sock *inet = inet_sk(sk);
434 sin->sin_family = AF_INET;
435 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
437 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
438 if (inet->cmsg_flags)
439 ip_cmsg_recv(msg, skb);
442 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
444 /* Now we could try to dump offended packet options */
446 msg->msg_flags |= MSG_ERRQUEUE;
449 /* Reset and regenerate socket error */
450 spin_lock_bh(&sk->sk_error_queue.lock);
452 skb2 = skb_peek(&sk->sk_error_queue);
454 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
455 spin_unlock_bh(&sk->sk_error_queue.lock);
456 sk->sk_error_report(sk);
458 spin_unlock_bh(&sk->sk_error_queue.lock);
468 * Socket option code for IP. This is the end of the line after any
469 * TCP,UDP etc options on an IP socket.
472 static int do_ip_setsockopt(struct sock *sk, int level,
473 int optname, char __user *optval, unsigned int optlen)
475 struct inet_sock *inet = inet_sk(sk);
487 case IP_MTU_DISCOVER:
489 case IP_ROUTER_ALERT:
496 case IP_MULTICAST_TTL:
497 case IP_MULTICAST_ALL:
498 case IP_MULTICAST_LOOP:
499 case IP_RECVORIGDSTADDR:
500 if (optlen >= sizeof(int)) {
501 if (get_user(val, (int __user *) optval))
503 } else if (optlen >= sizeof(char)) {
506 if (get_user(ucval, (unsigned char __user *) optval))
512 /* If optlen==0, it is equivalent to val == 0 */
514 if (ip_mroute_opt(optname))
515 return ip_mroute_setsockopt(sk, optname, optval, optlen);
523 struct ip_options_rcu *old, *opt = NULL;
527 err = ip_options_get_from_user(sock_net(sk), &opt,
531 old = rcu_dereference_protected(inet->inet_opt,
532 sock_owned_by_user(sk));
534 struct inet_connection_sock *icsk = inet_csk(sk);
535 #if IS_ENABLED(CONFIG_IPV6)
536 if (sk->sk_family == PF_INET ||
537 (!((1 << sk->sk_state) &
538 (TCPF_LISTEN | TCPF_CLOSE)) &&
539 inet->inet_daddr != LOOPBACK4_IPV6)) {
542 icsk->icsk_ext_hdr_len -= old->opt.optlen;
544 icsk->icsk_ext_hdr_len += opt->opt.optlen;
545 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
546 #if IS_ENABLED(CONFIG_IPV6)
550 rcu_assign_pointer(inet->inet_opt, opt);
557 inet->cmsg_flags |= IP_CMSG_PKTINFO;
559 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
563 inet->cmsg_flags |= IP_CMSG_TTL;
565 inet->cmsg_flags &= ~IP_CMSG_TTL;
569 inet->cmsg_flags |= IP_CMSG_TOS;
571 inet->cmsg_flags &= ~IP_CMSG_TOS;
575 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
577 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
581 inet->cmsg_flags |= IP_CMSG_RETOPTS;
583 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
587 inet->cmsg_flags |= IP_CMSG_PASSSEC;
589 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
591 case IP_RECVORIGDSTADDR:
593 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
595 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
597 case IP_TOS: /* This sets both TOS and Precedence */
598 if (sk->sk_type == SOCK_STREAM) {
599 val &= ~INET_ECN_MASK;
600 val |= inet->tos & INET_ECN_MASK;
602 if (inet->tos != val) {
604 sk->sk_priority = rt_tos2priority(val);
611 if (val != -1 && (val < 1 || val > 255))
616 if (sk->sk_type != SOCK_RAW) {
620 inet->hdrincl = val ? 1 : 0;
623 if (sk->sk_type != SOCK_RAW) {
627 inet->nodefrag = val ? 1 : 0;
629 case IP_MTU_DISCOVER:
630 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_INTERFACE)
632 inet->pmtudisc = val;
635 inet->recverr = !!val;
637 skb_queue_purge(&sk->sk_error_queue);
639 case IP_MULTICAST_TTL:
640 if (sk->sk_type == SOCK_STREAM)
646 if (val < 0 || val > 255)
650 case IP_MULTICAST_LOOP:
653 inet->mc_loop = !!val;
657 struct net_device *dev = NULL;
660 if (optlen != sizeof(int))
663 ifindex = (__force int)ntohl((__force __be32)val);
670 dev = dev_get_by_index(sock_net(sk), ifindex);
671 err = -EADDRNOTAVAIL;
677 if (sk->sk_bound_dev_if)
680 inet->uc_index = ifindex;
684 case IP_MULTICAST_IF:
686 struct ip_mreqn mreq;
687 struct net_device *dev = NULL;
689 if (sk->sk_type == SOCK_STREAM)
692 * Check the arguments are allowable
695 if (optlen < sizeof(struct in_addr))
699 if (optlen >= sizeof(struct ip_mreqn)) {
700 if (copy_from_user(&mreq, optval, sizeof(mreq)))
703 memset(&mreq, 0, sizeof(mreq));
704 if (optlen >= sizeof(struct ip_mreq)) {
705 if (copy_from_user(&mreq, optval,
706 sizeof(struct ip_mreq)))
708 } else if (optlen >= sizeof(struct in_addr)) {
709 if (copy_from_user(&mreq.imr_address, optval,
710 sizeof(struct in_addr)))
715 if (!mreq.imr_ifindex) {
716 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
722 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
724 mreq.imr_ifindex = dev->ifindex;
726 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
729 err = -EADDRNOTAVAIL;
735 if (sk->sk_bound_dev_if &&
736 mreq.imr_ifindex != sk->sk_bound_dev_if)
739 inet->mc_index = mreq.imr_ifindex;
740 inet->mc_addr = mreq.imr_address.s_addr;
745 case IP_ADD_MEMBERSHIP:
746 case IP_DROP_MEMBERSHIP:
748 struct ip_mreqn mreq;
751 if (inet_sk(sk)->is_icsk)
754 if (optlen < sizeof(struct ip_mreq))
757 if (optlen >= sizeof(struct ip_mreqn)) {
758 if (copy_from_user(&mreq, optval, sizeof(mreq)))
761 memset(&mreq, 0, sizeof(mreq));
762 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
766 if (optname == IP_ADD_MEMBERSHIP)
767 err = ip_mc_join_group(sk, &mreq);
769 err = ip_mc_leave_group(sk, &mreq);
774 struct ip_msfilter *msf;
776 if (optlen < IP_MSFILTER_SIZE(0))
778 if (optlen > sysctl_optmem_max) {
782 msf = kmalloc(optlen, GFP_KERNEL);
788 if (copy_from_user(msf, optval, optlen)) {
792 /* numsrc >= (1G-4) overflow in 32 bits */
793 if (msf->imsf_numsrc >= 0x3ffffffcU ||
794 msf->imsf_numsrc > sysctl_igmp_max_msf) {
799 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
804 err = ip_mc_msfilter(sk, msf, 0);
808 case IP_BLOCK_SOURCE:
809 case IP_UNBLOCK_SOURCE:
810 case IP_ADD_SOURCE_MEMBERSHIP:
811 case IP_DROP_SOURCE_MEMBERSHIP:
813 struct ip_mreq_source mreqs;
816 if (optlen != sizeof(struct ip_mreq_source))
818 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
822 if (optname == IP_BLOCK_SOURCE) {
823 omode = MCAST_EXCLUDE;
825 } else if (optname == IP_UNBLOCK_SOURCE) {
826 omode = MCAST_EXCLUDE;
828 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
829 struct ip_mreqn mreq;
831 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
832 mreq.imr_address.s_addr = mreqs.imr_interface;
833 mreq.imr_ifindex = 0;
834 err = ip_mc_join_group(sk, &mreq);
835 if (err && err != -EADDRINUSE)
837 omode = MCAST_INCLUDE;
839 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
840 omode = MCAST_INCLUDE;
843 err = ip_mc_source(add, omode, sk, &mreqs, 0);
846 case MCAST_JOIN_GROUP:
847 case MCAST_LEAVE_GROUP:
849 struct group_req greq;
850 struct sockaddr_in *psin;
851 struct ip_mreqn mreq;
853 if (optlen < sizeof(struct group_req))
856 if (copy_from_user(&greq, optval, sizeof(greq)))
858 psin = (struct sockaddr_in *)&greq.gr_group;
859 if (psin->sin_family != AF_INET)
861 memset(&mreq, 0, sizeof(mreq));
862 mreq.imr_multiaddr = psin->sin_addr;
863 mreq.imr_ifindex = greq.gr_interface;
865 if (optname == MCAST_JOIN_GROUP)
866 err = ip_mc_join_group(sk, &mreq);
868 err = ip_mc_leave_group(sk, &mreq);
871 case MCAST_JOIN_SOURCE_GROUP:
872 case MCAST_LEAVE_SOURCE_GROUP:
873 case MCAST_BLOCK_SOURCE:
874 case MCAST_UNBLOCK_SOURCE:
876 struct group_source_req greqs;
877 struct ip_mreq_source mreqs;
878 struct sockaddr_in *psin;
881 if (optlen != sizeof(struct group_source_req))
883 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
887 if (greqs.gsr_group.ss_family != AF_INET ||
888 greqs.gsr_source.ss_family != AF_INET) {
889 err = -EADDRNOTAVAIL;
892 psin = (struct sockaddr_in *)&greqs.gsr_group;
893 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
894 psin = (struct sockaddr_in *)&greqs.gsr_source;
895 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
896 mreqs.imr_interface = 0; /* use index for mc_source */
898 if (optname == MCAST_BLOCK_SOURCE) {
899 omode = MCAST_EXCLUDE;
901 } else if (optname == MCAST_UNBLOCK_SOURCE) {
902 omode = MCAST_EXCLUDE;
904 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
905 struct ip_mreqn mreq;
907 psin = (struct sockaddr_in *)&greqs.gsr_group;
908 mreq.imr_multiaddr = psin->sin_addr;
909 mreq.imr_address.s_addr = 0;
910 mreq.imr_ifindex = greqs.gsr_interface;
911 err = ip_mc_join_group(sk, &mreq);
912 if (err && err != -EADDRINUSE)
914 greqs.gsr_interface = mreq.imr_ifindex;
915 omode = MCAST_INCLUDE;
917 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
918 omode = MCAST_INCLUDE;
921 err = ip_mc_source(add, omode, sk, &mreqs,
922 greqs.gsr_interface);
927 struct sockaddr_in *psin;
928 struct ip_msfilter *msf = NULL;
929 struct group_filter *gsf = NULL;
930 int msize, i, ifindex;
932 if (optlen < GROUP_FILTER_SIZE(0))
934 if (optlen > sysctl_optmem_max) {
938 gsf = kmalloc(optlen, GFP_KERNEL);
944 if (copy_from_user(gsf, optval, optlen))
947 /* numsrc >= (4G-140)/128 overflow in 32 bits */
948 if (gsf->gf_numsrc >= 0x1ffffff ||
949 gsf->gf_numsrc > sysctl_igmp_max_msf) {
953 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
957 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
958 msf = kmalloc(msize, GFP_KERNEL);
963 ifindex = gsf->gf_interface;
964 psin = (struct sockaddr_in *)&gsf->gf_group;
965 if (psin->sin_family != AF_INET) {
966 err = -EADDRNOTAVAIL;
969 msf->imsf_multiaddr = psin->sin_addr.s_addr;
970 msf->imsf_interface = 0;
971 msf->imsf_fmode = gsf->gf_fmode;
972 msf->imsf_numsrc = gsf->gf_numsrc;
973 err = -EADDRNOTAVAIL;
974 for (i = 0; i < gsf->gf_numsrc; ++i) {
975 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
977 if (psin->sin_family != AF_INET)
979 msf->imsf_slist[i] = psin->sin_addr.s_addr;
984 err = ip_mc_msfilter(sk, msf, ifindex);
990 case IP_MULTICAST_ALL:
993 if (val != 0 && val != 1)
997 case IP_ROUTER_ALERT:
998 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1004 inet->freebind = !!val;
1007 case IP_IPSEC_POLICY:
1008 case IP_XFRM_POLICY:
1010 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1012 err = xfrm_user_policy(sk, optname, optval, optlen);
1015 case IP_TRANSPARENT:
1016 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1017 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1023 inet->transparent = !!val;
1029 if (val < 0 || val > 255)
1031 inet->min_ttl = val;
1047 * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1051 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1052 * destination in skb->cb[] before dst drop.
1053 * This way, receiver doesnt make cache line misses to read rtable.
1055 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1057 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1059 if ((inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) &&
1061 pktinfo->ipi_ifindex = inet_iif(skb);
1062 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1064 pktinfo->ipi_ifindex = 0;
1065 pktinfo->ipi_spec_dst.s_addr = 0;
1070 int ip_setsockopt(struct sock *sk, int level,
1071 int optname, char __user *optval, unsigned int optlen)
1075 if (level != SOL_IP)
1076 return -ENOPROTOOPT;
1078 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1079 #ifdef CONFIG_NETFILTER
1080 /* we need to exclude all possible ENOPROTOOPTs except default case */
1081 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1082 optname != IP_IPSEC_POLICY &&
1083 optname != IP_XFRM_POLICY &&
1084 !ip_mroute_opt(optname)) {
1086 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1092 EXPORT_SYMBOL(ip_setsockopt);
1094 #ifdef CONFIG_COMPAT
1095 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1096 char __user *optval, unsigned int optlen)
1100 if (level != SOL_IP)
1101 return -ENOPROTOOPT;
1103 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1104 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1107 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1108 #ifdef CONFIG_NETFILTER
1109 /* we need to exclude all possible ENOPROTOOPTs except default case */
1110 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1111 optname != IP_IPSEC_POLICY &&
1112 optname != IP_XFRM_POLICY &&
1113 !ip_mroute_opt(optname)) {
1115 err = compat_nf_setsockopt(sk, PF_INET, optname,
1122 EXPORT_SYMBOL(compat_ip_setsockopt);
1126 * Get the options. Note for future reference. The GET of IP options gets
1127 * the _received_ ones. The set sets the _sent_ ones.
1130 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1131 char __user *optval, int __user *optlen, unsigned int flags)
1133 struct inet_sock *inet = inet_sk(sk);
1137 if (level != SOL_IP)
1140 if (ip_mroute_opt(optname))
1141 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1143 if (get_user(len, optlen))
1153 unsigned char optbuf[sizeof(struct ip_options)+40];
1154 struct ip_options *opt = (struct ip_options *)optbuf;
1155 struct ip_options_rcu *inet_opt;
1157 inet_opt = rcu_dereference_protected(inet->inet_opt,
1158 sock_owned_by_user(sk));
1161 memcpy(optbuf, &inet_opt->opt,
1162 sizeof(struct ip_options) +
1163 inet_opt->opt.optlen);
1166 if (opt->optlen == 0)
1167 return put_user(0, optlen);
1169 ip_options_undo(opt);
1171 len = min_t(unsigned int, len, opt->optlen);
1172 if (put_user(len, optlen))
1174 if (copy_to_user(optval, opt->__data, len))
1179 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1182 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1185 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1188 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1191 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1194 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1196 case IP_RECVORIGDSTADDR:
1197 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1203 val = (inet->uc_ttl == -1 ?
1204 sysctl_ip_default_ttl :
1208 val = inet->hdrincl;
1211 val = inet->nodefrag;
1213 case IP_MTU_DISCOVER:
1214 val = inet->pmtudisc;
1218 struct dst_entry *dst;
1220 dst = sk_dst_get(sk);
1232 val = inet->recverr;
1234 case IP_MULTICAST_TTL:
1237 case IP_MULTICAST_LOOP:
1238 val = inet->mc_loop;
1241 val = (__force int)htonl((__u32) inet->uc_index);
1243 case IP_MULTICAST_IF:
1245 struct in_addr addr;
1246 len = min_t(unsigned int, len, sizeof(struct in_addr));
1247 addr.s_addr = inet->mc_addr;
1250 if (put_user(len, optlen))
1252 if (copy_to_user(optval, &addr, len))
1258 struct ip_msfilter msf;
1261 if (len < IP_MSFILTER_SIZE(0)) {
1265 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1269 err = ip_mc_msfget(sk, &msf,
1270 (struct ip_msfilter __user *)optval, optlen);
1274 case MCAST_MSFILTER:
1276 struct group_filter gsf;
1279 if (len < GROUP_FILTER_SIZE(0)) {
1283 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1287 err = ip_mc_gsfget(sk, &gsf,
1288 (struct group_filter __user *)optval,
1293 case IP_MULTICAST_ALL:
1302 if (sk->sk_type != SOCK_STREAM)
1303 return -ENOPROTOOPT;
1305 msg.msg_control = optval;
1306 msg.msg_controllen = len;
1307 msg.msg_flags = flags;
1309 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1310 struct in_pktinfo info;
1312 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1313 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1314 info.ipi_ifindex = inet->mc_index;
1315 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1317 if (inet->cmsg_flags & IP_CMSG_TTL) {
1318 int hlim = inet->mc_ttl;
1319 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1321 if (inet->cmsg_flags & IP_CMSG_TOS) {
1322 int tos = inet->rcv_tos;
1323 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1325 len -= msg.msg_controllen;
1326 return put_user(len, optlen);
1329 val = inet->freebind;
1331 case IP_TRANSPARENT:
1332 val = inet->transparent;
1335 val = inet->min_ttl;
1339 return -ENOPROTOOPT;
1343 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1344 unsigned char ucval = (unsigned char)val;
1346 if (put_user(len, optlen))
1348 if (copy_to_user(optval, &ucval, 1))
1351 len = min_t(unsigned int, sizeof(int), len);
1352 if (put_user(len, optlen))
1354 if (copy_to_user(optval, &val, len))
1360 int ip_getsockopt(struct sock *sk, int level,
1361 int optname, char __user *optval, int __user *optlen)
1365 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1366 #ifdef CONFIG_NETFILTER
1367 /* we need to exclude all possible ENOPROTOOPTs except default case */
1368 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1369 !ip_mroute_opt(optname)) {
1372 if (get_user(len, optlen))
1376 err = nf_getsockopt(sk, PF_INET, optname, optval,
1380 err = put_user(len, optlen);
1386 EXPORT_SYMBOL(ip_getsockopt);
1388 #ifdef CONFIG_COMPAT
1389 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1390 char __user *optval, int __user *optlen)
1394 if (optname == MCAST_MSFILTER)
1395 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1398 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1401 #ifdef CONFIG_NETFILTER
1402 /* we need to exclude all possible ENOPROTOOPTs except default case */
1403 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1404 !ip_mroute_opt(optname)) {
1407 if (get_user(len, optlen))
1411 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1414 err = put_user(len, optlen);
1420 EXPORT_SYMBOL(compat_ip_getsockopt);