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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
41 #include <net/transp_v6.h>
44 #include <linux/errqueue.h>
45 #include <asm/uaccess.h>
47 #define IP_CMSG_PKTINFO 1
50 #define IP_CMSG_RECVOPTS 8
51 #define IP_CMSG_RETOPTS 16
52 #define IP_CMSG_PASSSEC 32
53 #define IP_CMSG_ORIGDSTADDR 64
56 * SOL_IP control messages.
59 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
61 struct in_pktinfo info;
62 struct rtable *rt = skb_rtable(skb);
64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
66 info.ipi_ifindex = rt->rt_iif;
67 info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
70 info.ipi_spec_dst.s_addr = 0;
73 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
76 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
78 int ttl = ip_hdr(skb)->ttl;
79 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
82 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
84 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
87 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
89 if (IPCB(skb)->opt.optlen == 0)
92 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
97 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
99 unsigned char optbuf[sizeof(struct ip_options) + 40];
100 struct ip_options * opt = (struct ip_options *)optbuf;
102 if (IPCB(skb)->opt.optlen == 0)
105 if (ip_options_echo(opt, skb)) {
106 msg->msg_flags |= MSG_CTRUNC;
109 ip_options_undo(opt);
111 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
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(struct msghdr *msg, struct sk_buff *skb)
156 struct inet_sock *inet = inet_sk(skb->sk);
157 unsigned flags = inet->cmsg_flags;
159 /* Ordered by supposed usage frequency */
161 ip_cmsg_recv_pktinfo(msg, skb);
162 if ((flags >>= 1) == 0)
166 ip_cmsg_recv_ttl(msg, skb);
167 if ((flags >>= 1) == 0)
171 ip_cmsg_recv_tos(msg, skb);
172 if ((flags >>= 1) == 0)
176 ip_cmsg_recv_opts(msg, skb);
177 if ((flags >>= 1) == 0)
181 ip_cmsg_recv_retopts(msg, skb);
182 if ((flags >>= 1) == 0)
186 ip_cmsg_recv_security(msg, skb);
188 if ((flags >>= 1) == 0)
191 ip_cmsg_recv_dstaddr(msg, skb);
194 EXPORT_SYMBOL(ip_cmsg_recv);
196 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
199 struct cmsghdr *cmsg;
201 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
202 if (!CMSG_OK(msg, cmsg))
204 if (cmsg->cmsg_level != SOL_IP)
206 switch (cmsg->cmsg_type) {
208 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
209 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
210 err < 40 ? err : 40);
216 struct in_pktinfo *info;
217 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
219 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
220 ipc->oif = info->ipi_ifindex;
221 ipc->addr = info->ipi_spec_dst.s_addr;
232 /* Special input handler for packets caught by router alert option.
233 They are selected only by protocol field, and then processed likely
234 local ones; but only if someone wants them! Otherwise, router
235 not running rsvpd will kill RSVP.
237 It is user level problem, what it will make with them.
238 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
239 but receiver should be enough clever f.e. to forward mtrace requests,
240 sent to multicast group to reach destination designated router.
242 struct ip_ra_chain __rcu *ip_ra_chain;
243 static DEFINE_SPINLOCK(ip_ra_lock);
246 static void ip_ra_destroy_rcu(struct rcu_head *head)
248 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
250 sock_put(ra->saved_sk);
254 int ip_ra_control(struct sock *sk, unsigned char on,
255 void (*destructor)(struct sock *))
257 struct ip_ra_chain *ra, *new_ra;
258 struct ip_ra_chain __rcu **rap;
260 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
263 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
265 spin_lock_bh(&ip_ra_lock);
266 for (rap = &ip_ra_chain;
267 (ra = rcu_dereference_protected(*rap,
268 lockdep_is_held(&ip_ra_lock))) != NULL;
272 spin_unlock_bh(&ip_ra_lock);
276 /* dont let ip_call_ra_chain() use sk again */
278 rcu_assign_pointer(*rap, ra->next);
279 spin_unlock_bh(&ip_ra_lock);
284 * Delay sock_put(sk) and kfree(ra) after one rcu grace
285 * period. This guarantee ip_call_ra_chain() dont need
286 * to mess with socket refcounts.
289 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
293 if (new_ra == NULL) {
294 spin_unlock_bh(&ip_ra_lock);
298 new_ra->destructor = destructor;
301 rcu_assign_pointer(*rap, new_ra);
303 spin_unlock_bh(&ip_ra_lock);
308 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
309 __be16 port, u32 info, u8 *payload)
311 struct sock_exterr_skb *serr;
313 skb = skb_clone(skb, GFP_ATOMIC);
317 serr = SKB_EXT_ERR(skb);
318 serr->ee.ee_errno = err;
319 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
320 serr->ee.ee_type = icmp_hdr(skb)->type;
321 serr->ee.ee_code = icmp_hdr(skb)->code;
323 serr->ee.ee_info = info;
324 serr->ee.ee_data = 0;
325 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
326 skb_network_header(skb);
329 if (skb_pull(skb, payload - skb->data) != NULL) {
330 skb_reset_transport_header(skb);
331 if (sock_queue_err_skb(sk, skb) == 0)
337 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
339 struct inet_sock *inet = inet_sk(sk);
340 struct sock_exterr_skb *serr;
347 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
351 skb_put(skb, sizeof(struct iphdr));
352 skb_reset_network_header(skb);
356 serr = SKB_EXT_ERR(skb);
357 serr->ee.ee_errno = err;
358 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
359 serr->ee.ee_type = 0;
360 serr->ee.ee_code = 0;
362 serr->ee.ee_info = info;
363 serr->ee.ee_data = 0;
364 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
367 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
368 skb_reset_transport_header(skb);
370 if (sock_queue_err_skb(sk, skb))
375 * Handle MSG_ERRQUEUE
377 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
379 struct sock_exterr_skb *serr;
380 struct sk_buff *skb, *skb2;
381 struct sockaddr_in *sin;
383 struct sock_extended_err ee;
384 struct sockaddr_in offender;
390 skb = skb_dequeue(&sk->sk_error_queue);
396 msg->msg_flags |= MSG_TRUNC;
399 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
403 sock_recv_timestamp(msg, sk, skb);
405 serr = SKB_EXT_ERR(skb);
407 sin = (struct sockaddr_in *)msg->msg_name;
409 sin->sin_family = AF_INET;
410 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
412 sin->sin_port = serr->port;
413 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
414 *addr_len = sizeof(*sin);
417 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
418 sin = &errhdr.offender;
419 sin->sin_family = AF_UNSPEC;
420 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
421 struct inet_sock *inet = inet_sk(sk);
423 sin->sin_family = AF_INET;
424 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
426 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
427 if (inet->cmsg_flags)
428 ip_cmsg_recv(msg, skb);
431 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
433 /* Now we could try to dump offended packet options */
435 msg->msg_flags |= MSG_ERRQUEUE;
438 /* Reset and regenerate socket error */
439 spin_lock_bh(&sk->sk_error_queue.lock);
441 skb2 = skb_peek(&sk->sk_error_queue);
443 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
444 spin_unlock_bh(&sk->sk_error_queue.lock);
445 sk->sk_error_report(sk);
447 spin_unlock_bh(&sk->sk_error_queue.lock);
456 static void opt_kfree_rcu(struct rcu_head *head)
458 kfree(container_of(head, struct ip_options_rcu, rcu));
462 * Socket option code for IP. This is the end of the line after any
463 * TCP,UDP etc options on an IP socket.
466 static int do_ip_setsockopt(struct sock *sk, int level,
467 int optname, char __user *optval, unsigned int optlen)
469 struct inet_sock *inet = inet_sk(sk);
481 case IP_MTU_DISCOVER:
483 case IP_ROUTER_ALERT:
489 case IP_MULTICAST_TTL:
490 case IP_MULTICAST_ALL:
491 case IP_MULTICAST_LOOP:
492 case IP_RECVORIGDSTADDR:
493 if (optlen >= sizeof(int)) {
494 if (get_user(val, (int __user *) optval))
496 } else if (optlen >= sizeof(char)) {
499 if (get_user(ucval, (unsigned char __user *) optval))
505 /* If optlen==0, it is equivalent to val == 0 */
507 if (ip_mroute_opt(optname))
508 return ip_mroute_setsockopt(sk, optname, optval, optlen);
516 struct ip_options_rcu *old, *opt = NULL;
520 err = ip_options_get_from_user(sock_net(sk), &opt,
524 old = rcu_dereference_protected(inet->inet_opt,
525 sock_owned_by_user(sk));
527 struct inet_connection_sock *icsk = inet_csk(sk);
528 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
529 if (sk->sk_family == PF_INET ||
530 (!((1 << sk->sk_state) &
531 (TCPF_LISTEN | TCPF_CLOSE)) &&
532 inet->inet_daddr != LOOPBACK4_IPV6)) {
535 icsk->icsk_ext_hdr_len -= old->opt.optlen;
537 icsk->icsk_ext_hdr_len += opt->opt.optlen;
538 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
539 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
543 rcu_assign_pointer(inet->inet_opt, opt);
545 call_rcu(&old->rcu, opt_kfree_rcu);
550 inet->cmsg_flags |= IP_CMSG_PKTINFO;
552 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
556 inet->cmsg_flags |= IP_CMSG_TTL;
558 inet->cmsg_flags &= ~IP_CMSG_TTL;
562 inet->cmsg_flags |= IP_CMSG_TOS;
564 inet->cmsg_flags &= ~IP_CMSG_TOS;
568 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
570 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
574 inet->cmsg_flags |= IP_CMSG_RETOPTS;
576 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
580 inet->cmsg_flags |= IP_CMSG_PASSSEC;
582 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
584 case IP_RECVORIGDSTADDR:
586 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
588 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
590 case IP_TOS: /* This sets both TOS and Precedence */
591 if (sk->sk_type == SOCK_STREAM) {
592 val &= ~INET_ECN_MASK;
593 val |= inet->tos & INET_ECN_MASK;
595 if (inet->tos != val) {
597 sk->sk_priority = rt_tos2priority(val);
604 if (val != -1 && (val < 1 || val > 255))
609 if (sk->sk_type != SOCK_RAW) {
613 inet->hdrincl = val ? 1 : 0;
616 if (sk->sk_type != SOCK_RAW) {
620 inet->nodefrag = val ? 1 : 0;
622 case IP_MTU_DISCOVER:
623 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
625 inet->pmtudisc = val;
628 inet->recverr = !!val;
630 skb_queue_purge(&sk->sk_error_queue);
632 case IP_MULTICAST_TTL:
633 if (sk->sk_type == SOCK_STREAM)
639 if (val < 0 || val > 255)
643 case IP_MULTICAST_LOOP:
646 inet->mc_loop = !!val;
648 case IP_MULTICAST_IF:
650 struct ip_mreqn mreq;
651 struct net_device *dev = NULL;
653 if (sk->sk_type == SOCK_STREAM)
656 * Check the arguments are allowable
659 if (optlen < sizeof(struct in_addr))
663 if (optlen >= sizeof(struct ip_mreqn)) {
664 if (copy_from_user(&mreq, optval, sizeof(mreq)))
667 memset(&mreq, 0, sizeof(mreq));
668 if (optlen >= sizeof(struct in_addr) &&
669 copy_from_user(&mreq.imr_address, optval,
670 sizeof(struct in_addr)))
674 if (!mreq.imr_ifindex) {
675 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
681 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
683 mreq.imr_ifindex = dev->ifindex;
685 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
688 err = -EADDRNOTAVAIL;
694 if (sk->sk_bound_dev_if &&
695 mreq.imr_ifindex != sk->sk_bound_dev_if)
698 inet->mc_index = mreq.imr_ifindex;
699 inet->mc_addr = mreq.imr_address.s_addr;
704 case IP_ADD_MEMBERSHIP:
705 case IP_DROP_MEMBERSHIP:
707 struct ip_mreqn mreq;
710 if (inet_sk(sk)->is_icsk)
713 if (optlen < sizeof(struct ip_mreq))
716 if (optlen >= sizeof(struct ip_mreqn)) {
717 if (copy_from_user(&mreq, optval, sizeof(mreq)))
720 memset(&mreq, 0, sizeof(mreq));
721 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
725 if (optname == IP_ADD_MEMBERSHIP)
726 err = ip_mc_join_group(sk, &mreq);
728 err = ip_mc_leave_group(sk, &mreq);
733 struct ip_msfilter *msf;
735 if (optlen < IP_MSFILTER_SIZE(0))
737 if (optlen > sysctl_optmem_max) {
741 msf = kmalloc(optlen, GFP_KERNEL);
747 if (copy_from_user(msf, optval, optlen)) {
751 /* numsrc >= (1G-4) overflow in 32 bits */
752 if (msf->imsf_numsrc >= 0x3ffffffcU ||
753 msf->imsf_numsrc > sysctl_igmp_max_msf) {
758 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
763 err = ip_mc_msfilter(sk, msf, 0);
767 case IP_BLOCK_SOURCE:
768 case IP_UNBLOCK_SOURCE:
769 case IP_ADD_SOURCE_MEMBERSHIP:
770 case IP_DROP_SOURCE_MEMBERSHIP:
772 struct ip_mreq_source mreqs;
775 if (optlen != sizeof(struct ip_mreq_source))
777 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
781 if (optname == IP_BLOCK_SOURCE) {
782 omode = MCAST_EXCLUDE;
784 } else if (optname == IP_UNBLOCK_SOURCE) {
785 omode = MCAST_EXCLUDE;
787 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
788 struct ip_mreqn mreq;
790 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
791 mreq.imr_address.s_addr = mreqs.imr_interface;
792 mreq.imr_ifindex = 0;
793 err = ip_mc_join_group(sk, &mreq);
794 if (err && err != -EADDRINUSE)
796 omode = MCAST_INCLUDE;
798 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
799 omode = MCAST_INCLUDE;
802 err = ip_mc_source(add, omode, sk, &mreqs, 0);
805 case MCAST_JOIN_GROUP:
806 case MCAST_LEAVE_GROUP:
808 struct group_req greq;
809 struct sockaddr_in *psin;
810 struct ip_mreqn mreq;
812 if (optlen < sizeof(struct group_req))
815 if (copy_from_user(&greq, optval, sizeof(greq)))
817 psin = (struct sockaddr_in *)&greq.gr_group;
818 if (psin->sin_family != AF_INET)
820 memset(&mreq, 0, sizeof(mreq));
821 mreq.imr_multiaddr = psin->sin_addr;
822 mreq.imr_ifindex = greq.gr_interface;
824 if (optname == MCAST_JOIN_GROUP)
825 err = ip_mc_join_group(sk, &mreq);
827 err = ip_mc_leave_group(sk, &mreq);
830 case MCAST_JOIN_SOURCE_GROUP:
831 case MCAST_LEAVE_SOURCE_GROUP:
832 case MCAST_BLOCK_SOURCE:
833 case MCAST_UNBLOCK_SOURCE:
835 struct group_source_req greqs;
836 struct ip_mreq_source mreqs;
837 struct sockaddr_in *psin;
840 if (optlen != sizeof(struct group_source_req))
842 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
846 if (greqs.gsr_group.ss_family != AF_INET ||
847 greqs.gsr_source.ss_family != AF_INET) {
848 err = -EADDRNOTAVAIL;
851 psin = (struct sockaddr_in *)&greqs.gsr_group;
852 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
853 psin = (struct sockaddr_in *)&greqs.gsr_source;
854 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
855 mreqs.imr_interface = 0; /* use index for mc_source */
857 if (optname == MCAST_BLOCK_SOURCE) {
858 omode = MCAST_EXCLUDE;
860 } else if (optname == MCAST_UNBLOCK_SOURCE) {
861 omode = MCAST_EXCLUDE;
863 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
864 struct ip_mreqn mreq;
866 psin = (struct sockaddr_in *)&greqs.gsr_group;
867 mreq.imr_multiaddr = psin->sin_addr;
868 mreq.imr_address.s_addr = 0;
869 mreq.imr_ifindex = greqs.gsr_interface;
870 err = ip_mc_join_group(sk, &mreq);
871 if (err && err != -EADDRINUSE)
873 greqs.gsr_interface = mreq.imr_ifindex;
874 omode = MCAST_INCLUDE;
876 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
877 omode = MCAST_INCLUDE;
880 err = ip_mc_source(add, omode, sk, &mreqs,
881 greqs.gsr_interface);
886 struct sockaddr_in *psin;
887 struct ip_msfilter *msf = NULL;
888 struct group_filter *gsf = NULL;
889 int msize, i, ifindex;
891 if (optlen < GROUP_FILTER_SIZE(0))
893 if (optlen > sysctl_optmem_max) {
897 gsf = kmalloc(optlen, GFP_KERNEL);
903 if (copy_from_user(gsf, optval, optlen))
906 /* numsrc >= (4G-140)/128 overflow in 32 bits */
907 if (gsf->gf_numsrc >= 0x1ffffff ||
908 gsf->gf_numsrc > sysctl_igmp_max_msf) {
912 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
916 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
917 msf = kmalloc(msize, GFP_KERNEL);
922 ifindex = gsf->gf_interface;
923 psin = (struct sockaddr_in *)&gsf->gf_group;
924 if (psin->sin_family != AF_INET) {
925 err = -EADDRNOTAVAIL;
928 msf->imsf_multiaddr = psin->sin_addr.s_addr;
929 msf->imsf_interface = 0;
930 msf->imsf_fmode = gsf->gf_fmode;
931 msf->imsf_numsrc = gsf->gf_numsrc;
932 err = -EADDRNOTAVAIL;
933 for (i = 0; i < gsf->gf_numsrc; ++i) {
934 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
936 if (psin->sin_family != AF_INET)
938 msf->imsf_slist[i] = psin->sin_addr.s_addr;
943 err = ip_mc_msfilter(sk, msf, ifindex);
949 case IP_MULTICAST_ALL:
952 if (val != 0 && val != 1)
956 case IP_ROUTER_ALERT:
957 err = ip_ra_control(sk, val ? 1 : 0, NULL);
963 inet->freebind = !!val;
966 case IP_IPSEC_POLICY:
969 if (!capable(CAP_NET_ADMIN))
971 err = xfrm_user_policy(sk, optname, optval, optlen);
975 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) {
981 inet->transparent = !!val;
987 if (val < 0 || val > 255)
1005 * ip_queue_rcv_skb - Queue an skb into sock receive queue
1009 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option
1010 * is not set, we drop skb dst entry now, while dst cache line is hot.
1012 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1014 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO))
1016 return sock_queue_rcv_skb(sk, skb);
1018 EXPORT_SYMBOL(ip_queue_rcv_skb);
1020 int ip_setsockopt(struct sock *sk, int level,
1021 int optname, char __user *optval, unsigned int optlen)
1025 if (level != SOL_IP)
1026 return -ENOPROTOOPT;
1028 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1029 #ifdef CONFIG_NETFILTER
1030 /* we need to exclude all possible ENOPROTOOPTs except default case */
1031 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1032 optname != IP_IPSEC_POLICY &&
1033 optname != IP_XFRM_POLICY &&
1034 !ip_mroute_opt(optname)) {
1036 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1042 EXPORT_SYMBOL(ip_setsockopt);
1044 #ifdef CONFIG_COMPAT
1045 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1046 char __user *optval, unsigned int optlen)
1050 if (level != SOL_IP)
1051 return -ENOPROTOOPT;
1053 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1054 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1057 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1058 #ifdef CONFIG_NETFILTER
1059 /* we need to exclude all possible ENOPROTOOPTs except default case */
1060 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1061 optname != IP_IPSEC_POLICY &&
1062 optname != IP_XFRM_POLICY &&
1063 !ip_mroute_opt(optname)) {
1065 err = compat_nf_setsockopt(sk, PF_INET, optname,
1072 EXPORT_SYMBOL(compat_ip_setsockopt);
1076 * Get the options. Note for future reference. The GET of IP options gets
1077 * the _received_ ones. The set sets the _sent_ ones.
1080 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1081 char __user *optval, int __user *optlen, unsigned flags)
1083 struct inet_sock *inet = inet_sk(sk);
1087 if (level != SOL_IP)
1090 if (ip_mroute_opt(optname))
1091 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1093 if (get_user(len, optlen))
1103 unsigned char optbuf[sizeof(struct ip_options)+40];
1104 struct ip_options *opt = (struct ip_options *)optbuf;
1105 struct ip_options_rcu *inet_opt;
1107 inet_opt = rcu_dereference_protected(inet->inet_opt,
1108 sock_owned_by_user(sk));
1111 memcpy(optbuf, &inet_opt->opt,
1112 sizeof(struct ip_options) +
1113 inet_opt->opt.optlen);
1116 if (opt->optlen == 0)
1117 return put_user(0, optlen);
1119 ip_options_undo(opt);
1121 len = min_t(unsigned int, len, opt->optlen);
1122 if (put_user(len, optlen))
1124 if (copy_to_user(optval, opt->__data, len))
1129 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1132 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1135 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1138 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1141 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1144 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1146 case IP_RECVORIGDSTADDR:
1147 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1153 val = (inet->uc_ttl == -1 ?
1154 sysctl_ip_default_ttl :
1158 val = inet->hdrincl;
1161 val = inet->nodefrag;
1163 case IP_MTU_DISCOVER:
1164 val = inet->pmtudisc;
1168 struct dst_entry *dst;
1170 dst = sk_dst_get(sk);
1182 val = inet->recverr;
1184 case IP_MULTICAST_TTL:
1187 case IP_MULTICAST_LOOP:
1188 val = inet->mc_loop;
1190 case IP_MULTICAST_IF:
1192 struct in_addr addr;
1193 len = min_t(unsigned int, len, sizeof(struct in_addr));
1194 addr.s_addr = inet->mc_addr;
1197 if (put_user(len, optlen))
1199 if (copy_to_user(optval, &addr, len))
1205 struct ip_msfilter msf;
1208 if (len < IP_MSFILTER_SIZE(0)) {
1212 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1216 err = ip_mc_msfget(sk, &msf,
1217 (struct ip_msfilter __user *)optval, optlen);
1221 case MCAST_MSFILTER:
1223 struct group_filter gsf;
1226 if (len < GROUP_FILTER_SIZE(0)) {
1230 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1234 err = ip_mc_gsfget(sk, &gsf,
1235 (struct group_filter __user *)optval,
1240 case IP_MULTICAST_ALL:
1249 if (sk->sk_type != SOCK_STREAM)
1250 return -ENOPROTOOPT;
1252 msg.msg_control = optval;
1253 msg.msg_controllen = len;
1254 msg.msg_flags = flags;
1256 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1257 struct in_pktinfo info;
1259 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1260 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1261 info.ipi_ifindex = inet->mc_index;
1262 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1264 if (inet->cmsg_flags & IP_CMSG_TTL) {
1265 int hlim = inet->mc_ttl;
1266 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1268 len -= msg.msg_controllen;
1269 return put_user(len, optlen);
1272 val = inet->freebind;
1274 case IP_TRANSPARENT:
1275 val = inet->transparent;
1278 val = inet->min_ttl;
1282 return -ENOPROTOOPT;
1286 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1287 unsigned char ucval = (unsigned char)val;
1289 if (put_user(len, optlen))
1291 if (copy_to_user(optval, &ucval, 1))
1294 len = min_t(unsigned int, sizeof(int), len);
1295 if (put_user(len, optlen))
1297 if (copy_to_user(optval, &val, len))
1303 int ip_getsockopt(struct sock *sk, int level,
1304 int optname, char __user *optval, int __user *optlen)
1308 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1309 #ifdef CONFIG_NETFILTER
1310 /* we need to exclude all possible ENOPROTOOPTs except default case */
1311 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1312 !ip_mroute_opt(optname)) {
1315 if (get_user(len, optlen))
1319 err = nf_getsockopt(sk, PF_INET, optname, optval,
1323 err = put_user(len, optlen);
1329 EXPORT_SYMBOL(ip_getsockopt);
1331 #ifdef CONFIG_COMPAT
1332 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1333 char __user *optval, int __user *optlen)
1337 if (optname == MCAST_MSFILTER)
1338 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1341 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1344 #ifdef CONFIG_NETFILTER
1345 /* we need to exclude all possible ENOPROTOOPTs except default case */
1346 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1347 !ip_mroute_opt(optname)) {
1350 if (get_user(len, optlen))
1354 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1357 err = put_user(len, optlen);
1363 EXPORT_SYMBOL(compat_ip_getsockopt);