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, int *addr_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));
426 *addr_len = sizeof(*sin);
429 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
430 sin = &errhdr.offender;
431 sin->sin_family = AF_UNSPEC;
432 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
433 struct inet_sock *inet = inet_sk(sk);
435 sin->sin_family = AF_INET;
436 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
438 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
439 if (inet->cmsg_flags)
440 ip_cmsg_recv(msg, skb);
443 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
445 /* Now we could try to dump offended packet options */
447 msg->msg_flags |= MSG_ERRQUEUE;
450 /* Reset and regenerate socket error */
451 spin_lock_bh(&sk->sk_error_queue.lock);
453 skb2 = skb_peek(&sk->sk_error_queue);
455 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
456 spin_unlock_bh(&sk->sk_error_queue.lock);
457 sk->sk_error_report(sk);
459 spin_unlock_bh(&sk->sk_error_queue.lock);
469 * Socket option code for IP. This is the end of the line after any
470 * TCP,UDP etc options on an IP socket.
473 static int do_ip_setsockopt(struct sock *sk, int level,
474 int optname, char __user *optval, unsigned int optlen)
476 struct inet_sock *inet = inet_sk(sk);
488 case IP_MTU_DISCOVER:
490 case IP_ROUTER_ALERT:
497 case IP_MULTICAST_TTL:
498 case IP_MULTICAST_ALL:
499 case IP_MULTICAST_LOOP:
500 case IP_RECVORIGDSTADDR:
501 if (optlen >= sizeof(int)) {
502 if (get_user(val, (int __user *) optval))
504 } else if (optlen >= sizeof(char)) {
507 if (get_user(ucval, (unsigned char __user *) optval))
513 /* If optlen==0, it is equivalent to val == 0 */
515 if (ip_mroute_opt(optname))
516 return ip_mroute_setsockopt(sk, optname, optval, optlen);
524 struct ip_options_rcu *old, *opt = NULL;
528 err = ip_options_get_from_user(sock_net(sk), &opt,
532 old = rcu_dereference_protected(inet->inet_opt,
533 sock_owned_by_user(sk));
535 struct inet_connection_sock *icsk = inet_csk(sk);
536 #if IS_ENABLED(CONFIG_IPV6)
537 if (sk->sk_family == PF_INET ||
538 (!((1 << sk->sk_state) &
539 (TCPF_LISTEN | TCPF_CLOSE)) &&
540 inet->inet_daddr != LOOPBACK4_IPV6)) {
543 icsk->icsk_ext_hdr_len -= old->opt.optlen;
545 icsk->icsk_ext_hdr_len += opt->opt.optlen;
546 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
547 #if IS_ENABLED(CONFIG_IPV6)
551 rcu_assign_pointer(inet->inet_opt, opt);
558 inet->cmsg_flags |= IP_CMSG_PKTINFO;
560 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
564 inet->cmsg_flags |= IP_CMSG_TTL;
566 inet->cmsg_flags &= ~IP_CMSG_TTL;
570 inet->cmsg_flags |= IP_CMSG_TOS;
572 inet->cmsg_flags &= ~IP_CMSG_TOS;
576 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
578 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
582 inet->cmsg_flags |= IP_CMSG_RETOPTS;
584 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
588 inet->cmsg_flags |= IP_CMSG_PASSSEC;
590 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
592 case IP_RECVORIGDSTADDR:
594 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
596 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
598 case IP_TOS: /* This sets both TOS and Precedence */
599 if (sk->sk_type == SOCK_STREAM) {
600 val &= ~INET_ECN_MASK;
601 val |= inet->tos & INET_ECN_MASK;
603 if (inet->tos != val) {
605 sk->sk_priority = rt_tos2priority(val);
612 if (val != -1 && (val < 1 || val > 255))
617 if (sk->sk_type != SOCK_RAW) {
621 inet->hdrincl = val ? 1 : 0;
624 if (sk->sk_type != SOCK_RAW) {
628 inet->nodefrag = val ? 1 : 0;
630 case IP_MTU_DISCOVER:
631 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_INTERFACE)
633 inet->pmtudisc = val;
636 inet->recverr = !!val;
638 skb_queue_purge(&sk->sk_error_queue);
640 case IP_MULTICAST_TTL:
641 if (sk->sk_type == SOCK_STREAM)
647 if (val < 0 || val > 255)
651 case IP_MULTICAST_LOOP:
654 inet->mc_loop = !!val;
658 struct net_device *dev = NULL;
661 if (optlen != sizeof(int))
664 ifindex = (__force int)ntohl((__force __be32)val);
671 dev = dev_get_by_index(sock_net(sk), ifindex);
672 err = -EADDRNOTAVAIL;
678 if (sk->sk_bound_dev_if)
681 inet->uc_index = ifindex;
685 case IP_MULTICAST_IF:
687 struct ip_mreqn mreq;
688 struct net_device *dev = NULL;
690 if (sk->sk_type == SOCK_STREAM)
693 * Check the arguments are allowable
696 if (optlen < sizeof(struct in_addr))
700 if (optlen >= sizeof(struct ip_mreqn)) {
701 if (copy_from_user(&mreq, optval, sizeof(mreq)))
704 memset(&mreq, 0, sizeof(mreq));
705 if (optlen >= sizeof(struct ip_mreq)) {
706 if (copy_from_user(&mreq, optval,
707 sizeof(struct ip_mreq)))
709 } else if (optlen >= sizeof(struct in_addr)) {
710 if (copy_from_user(&mreq.imr_address, optval,
711 sizeof(struct in_addr)))
716 if (!mreq.imr_ifindex) {
717 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
723 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
725 mreq.imr_ifindex = dev->ifindex;
727 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
730 err = -EADDRNOTAVAIL;
736 if (sk->sk_bound_dev_if &&
737 mreq.imr_ifindex != sk->sk_bound_dev_if)
740 inet->mc_index = mreq.imr_ifindex;
741 inet->mc_addr = mreq.imr_address.s_addr;
746 case IP_ADD_MEMBERSHIP:
747 case IP_DROP_MEMBERSHIP:
749 struct ip_mreqn mreq;
752 if (inet_sk(sk)->is_icsk)
755 if (optlen < sizeof(struct ip_mreq))
758 if (optlen >= sizeof(struct ip_mreqn)) {
759 if (copy_from_user(&mreq, optval, sizeof(mreq)))
762 memset(&mreq, 0, sizeof(mreq));
763 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
767 if (optname == IP_ADD_MEMBERSHIP)
768 err = ip_mc_join_group(sk, &mreq);
770 err = ip_mc_leave_group(sk, &mreq);
775 struct ip_msfilter *msf;
777 if (optlen < IP_MSFILTER_SIZE(0))
779 if (optlen > sysctl_optmem_max) {
783 msf = kmalloc(optlen, GFP_KERNEL);
789 if (copy_from_user(msf, optval, optlen)) {
793 /* numsrc >= (1G-4) overflow in 32 bits */
794 if (msf->imsf_numsrc >= 0x3ffffffcU ||
795 msf->imsf_numsrc > sysctl_igmp_max_msf) {
800 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
805 err = ip_mc_msfilter(sk, msf, 0);
809 case IP_BLOCK_SOURCE:
810 case IP_UNBLOCK_SOURCE:
811 case IP_ADD_SOURCE_MEMBERSHIP:
812 case IP_DROP_SOURCE_MEMBERSHIP:
814 struct ip_mreq_source mreqs;
817 if (optlen != sizeof(struct ip_mreq_source))
819 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
823 if (optname == IP_BLOCK_SOURCE) {
824 omode = MCAST_EXCLUDE;
826 } else if (optname == IP_UNBLOCK_SOURCE) {
827 omode = MCAST_EXCLUDE;
829 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
830 struct ip_mreqn mreq;
832 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
833 mreq.imr_address.s_addr = mreqs.imr_interface;
834 mreq.imr_ifindex = 0;
835 err = ip_mc_join_group(sk, &mreq);
836 if (err && err != -EADDRINUSE)
838 omode = MCAST_INCLUDE;
840 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
841 omode = MCAST_INCLUDE;
844 err = ip_mc_source(add, omode, sk, &mreqs, 0);
847 case MCAST_JOIN_GROUP:
848 case MCAST_LEAVE_GROUP:
850 struct group_req greq;
851 struct sockaddr_in *psin;
852 struct ip_mreqn mreq;
854 if (optlen < sizeof(struct group_req))
857 if (copy_from_user(&greq, optval, sizeof(greq)))
859 psin = (struct sockaddr_in *)&greq.gr_group;
860 if (psin->sin_family != AF_INET)
862 memset(&mreq, 0, sizeof(mreq));
863 mreq.imr_multiaddr = psin->sin_addr;
864 mreq.imr_ifindex = greq.gr_interface;
866 if (optname == MCAST_JOIN_GROUP)
867 err = ip_mc_join_group(sk, &mreq);
869 err = ip_mc_leave_group(sk, &mreq);
872 case MCAST_JOIN_SOURCE_GROUP:
873 case MCAST_LEAVE_SOURCE_GROUP:
874 case MCAST_BLOCK_SOURCE:
875 case MCAST_UNBLOCK_SOURCE:
877 struct group_source_req greqs;
878 struct ip_mreq_source mreqs;
879 struct sockaddr_in *psin;
882 if (optlen != sizeof(struct group_source_req))
884 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
888 if (greqs.gsr_group.ss_family != AF_INET ||
889 greqs.gsr_source.ss_family != AF_INET) {
890 err = -EADDRNOTAVAIL;
893 psin = (struct sockaddr_in *)&greqs.gsr_group;
894 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
895 psin = (struct sockaddr_in *)&greqs.gsr_source;
896 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
897 mreqs.imr_interface = 0; /* use index for mc_source */
899 if (optname == MCAST_BLOCK_SOURCE) {
900 omode = MCAST_EXCLUDE;
902 } else if (optname == MCAST_UNBLOCK_SOURCE) {
903 omode = MCAST_EXCLUDE;
905 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
906 struct ip_mreqn mreq;
908 psin = (struct sockaddr_in *)&greqs.gsr_group;
909 mreq.imr_multiaddr = psin->sin_addr;
910 mreq.imr_address.s_addr = 0;
911 mreq.imr_ifindex = greqs.gsr_interface;
912 err = ip_mc_join_group(sk, &mreq);
913 if (err && err != -EADDRINUSE)
915 greqs.gsr_interface = mreq.imr_ifindex;
916 omode = MCAST_INCLUDE;
918 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
919 omode = MCAST_INCLUDE;
922 err = ip_mc_source(add, omode, sk, &mreqs,
923 greqs.gsr_interface);
928 struct sockaddr_in *psin;
929 struct ip_msfilter *msf = NULL;
930 struct group_filter *gsf = NULL;
931 int msize, i, ifindex;
933 if (optlen < GROUP_FILTER_SIZE(0))
935 if (optlen > sysctl_optmem_max) {
939 gsf = kmalloc(optlen, GFP_KERNEL);
945 if (copy_from_user(gsf, optval, optlen))
948 /* numsrc >= (4G-140)/128 overflow in 32 bits */
949 if (gsf->gf_numsrc >= 0x1ffffff ||
950 gsf->gf_numsrc > sysctl_igmp_max_msf) {
954 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
958 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
959 msf = kmalloc(msize, GFP_KERNEL);
964 ifindex = gsf->gf_interface;
965 psin = (struct sockaddr_in *)&gsf->gf_group;
966 if (psin->sin_family != AF_INET) {
967 err = -EADDRNOTAVAIL;
970 msf->imsf_multiaddr = psin->sin_addr.s_addr;
971 msf->imsf_interface = 0;
972 msf->imsf_fmode = gsf->gf_fmode;
973 msf->imsf_numsrc = gsf->gf_numsrc;
974 err = -EADDRNOTAVAIL;
975 for (i = 0; i < gsf->gf_numsrc; ++i) {
976 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
978 if (psin->sin_family != AF_INET)
980 msf->imsf_slist[i] = psin->sin_addr.s_addr;
985 err = ip_mc_msfilter(sk, msf, ifindex);
991 case IP_MULTICAST_ALL:
994 if (val != 0 && val != 1)
998 case IP_ROUTER_ALERT:
999 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1005 inet->freebind = !!val;
1008 case IP_IPSEC_POLICY:
1009 case IP_XFRM_POLICY:
1011 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1013 err = xfrm_user_policy(sk, optname, optval, optlen);
1016 case IP_TRANSPARENT:
1017 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1018 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1024 inet->transparent = !!val;
1030 if (val < 0 || val > 255)
1032 inet->min_ttl = val;
1048 * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1052 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1053 * destination in skb->cb[] before dst drop.
1054 * This way, receiver doesnt make cache line misses to read rtable.
1056 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1058 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1060 if ((inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) &&
1062 pktinfo->ipi_ifindex = inet_iif(skb);
1063 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1065 pktinfo->ipi_ifindex = 0;
1066 pktinfo->ipi_spec_dst.s_addr = 0;
1071 int ip_setsockopt(struct sock *sk, int level,
1072 int optname, char __user *optval, unsigned int optlen)
1076 if (level != SOL_IP)
1077 return -ENOPROTOOPT;
1079 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1080 #ifdef CONFIG_NETFILTER
1081 /* we need to exclude all possible ENOPROTOOPTs except default case */
1082 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1083 optname != IP_IPSEC_POLICY &&
1084 optname != IP_XFRM_POLICY &&
1085 !ip_mroute_opt(optname)) {
1087 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1093 EXPORT_SYMBOL(ip_setsockopt);
1095 #ifdef CONFIG_COMPAT
1096 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1097 char __user *optval, unsigned int optlen)
1101 if (level != SOL_IP)
1102 return -ENOPROTOOPT;
1104 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1105 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1108 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1109 #ifdef CONFIG_NETFILTER
1110 /* we need to exclude all possible ENOPROTOOPTs except default case */
1111 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1112 optname != IP_IPSEC_POLICY &&
1113 optname != IP_XFRM_POLICY &&
1114 !ip_mroute_opt(optname)) {
1116 err = compat_nf_setsockopt(sk, PF_INET, optname,
1123 EXPORT_SYMBOL(compat_ip_setsockopt);
1127 * Get the options. Note for future reference. The GET of IP options gets
1128 * the _received_ ones. The set sets the _sent_ ones.
1131 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1132 char __user *optval, int __user *optlen, unsigned int flags)
1134 struct inet_sock *inet = inet_sk(sk);
1138 if (level != SOL_IP)
1141 if (ip_mroute_opt(optname))
1142 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1144 if (get_user(len, optlen))
1154 unsigned char optbuf[sizeof(struct ip_options)+40];
1155 struct ip_options *opt = (struct ip_options *)optbuf;
1156 struct ip_options_rcu *inet_opt;
1158 inet_opt = rcu_dereference_protected(inet->inet_opt,
1159 sock_owned_by_user(sk));
1162 memcpy(optbuf, &inet_opt->opt,
1163 sizeof(struct ip_options) +
1164 inet_opt->opt.optlen);
1167 if (opt->optlen == 0)
1168 return put_user(0, optlen);
1170 ip_options_undo(opt);
1172 len = min_t(unsigned int, len, opt->optlen);
1173 if (put_user(len, optlen))
1175 if (copy_to_user(optval, opt->__data, len))
1180 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1183 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1186 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1189 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1192 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1195 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1197 case IP_RECVORIGDSTADDR:
1198 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1204 val = (inet->uc_ttl == -1 ?
1205 sysctl_ip_default_ttl :
1209 val = inet->hdrincl;
1212 val = inet->nodefrag;
1214 case IP_MTU_DISCOVER:
1215 val = inet->pmtudisc;
1219 struct dst_entry *dst;
1221 dst = sk_dst_get(sk);
1233 val = inet->recverr;
1235 case IP_MULTICAST_TTL:
1238 case IP_MULTICAST_LOOP:
1239 val = inet->mc_loop;
1242 val = (__force int)htonl((__u32) inet->uc_index);
1244 case IP_MULTICAST_IF:
1246 struct in_addr addr;
1247 len = min_t(unsigned int, len, sizeof(struct in_addr));
1248 addr.s_addr = inet->mc_addr;
1251 if (put_user(len, optlen))
1253 if (copy_to_user(optval, &addr, len))
1259 struct ip_msfilter msf;
1262 if (len < IP_MSFILTER_SIZE(0)) {
1266 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1270 err = ip_mc_msfget(sk, &msf,
1271 (struct ip_msfilter __user *)optval, optlen);
1275 case MCAST_MSFILTER:
1277 struct group_filter gsf;
1280 if (len < GROUP_FILTER_SIZE(0)) {
1284 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1288 err = ip_mc_gsfget(sk, &gsf,
1289 (struct group_filter __user *)optval,
1294 case IP_MULTICAST_ALL:
1303 if (sk->sk_type != SOCK_STREAM)
1304 return -ENOPROTOOPT;
1306 msg.msg_control = optval;
1307 msg.msg_controllen = len;
1308 msg.msg_flags = flags;
1310 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1311 struct in_pktinfo info;
1313 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1314 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1315 info.ipi_ifindex = inet->mc_index;
1316 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1318 if (inet->cmsg_flags & IP_CMSG_TTL) {
1319 int hlim = inet->mc_ttl;
1320 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1322 if (inet->cmsg_flags & IP_CMSG_TOS) {
1323 int tos = inet->rcv_tos;
1324 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1326 len -= msg.msg_controllen;
1327 return put_user(len, optlen);
1330 val = inet->freebind;
1332 case IP_TRANSPARENT:
1333 val = inet->transparent;
1336 val = inet->min_ttl;
1340 return -ENOPROTOOPT;
1344 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1345 unsigned char ucval = (unsigned char)val;
1347 if (put_user(len, optlen))
1349 if (copy_to_user(optval, &ucval, 1))
1352 len = min_t(unsigned int, sizeof(int), len);
1353 if (put_user(len, optlen))
1355 if (copy_to_user(optval, &val, len))
1361 int ip_getsockopt(struct sock *sk, int level,
1362 int optname, char __user *optval, int __user *optlen)
1366 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1367 #ifdef CONFIG_NETFILTER
1368 /* we need to exclude all possible ENOPROTOOPTs except default case */
1369 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1370 !ip_mroute_opt(optname)) {
1373 if (get_user(len, optlen))
1377 err = nf_getsockopt(sk, PF_INET, optname, optval,
1381 err = put_user(len, optlen);
1387 EXPORT_SYMBOL(ip_getsockopt);
1389 #ifdef CONFIG_COMPAT
1390 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1391 char __user *optval, int __user *optlen)
1395 if (optname == MCAST_MSFILTER)
1396 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1399 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1402 #ifdef CONFIG_NETFILTER
1403 /* we need to exclude all possible ENOPROTOOPTs except default case */
1404 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1405 !ip_mroute_opt(optname)) {
1408 if (get_user(len, optlen))
1412 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1415 err = put_user(len, optlen);
1421 EXPORT_SYMBOL(compat_ip_getsockopt);