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));
210 /* Our caller is responsible for freeing ipc->opt */
211 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
212 err < 40 ? err : 40);
218 struct in_pktinfo *info;
219 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
221 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
222 ipc->oif = info->ipi_ifindex;
223 ipc->addr = info->ipi_spec_dst.s_addr;
234 /* Special input handler for packets caught by router alert option.
235 They are selected only by protocol field, and then processed likely
236 local ones; but only if someone wants them! Otherwise, router
237 not running rsvpd will kill RSVP.
239 It is user level problem, what it will make with them.
240 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
241 but receiver should be enough clever f.e. to forward mtrace requests,
242 sent to multicast group to reach destination designated router.
244 struct ip_ra_chain __rcu *ip_ra_chain;
245 static DEFINE_SPINLOCK(ip_ra_lock);
248 static void ip_ra_destroy_rcu(struct rcu_head *head)
250 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
252 sock_put(ra->saved_sk);
256 int ip_ra_control(struct sock *sk, unsigned char on,
257 void (*destructor)(struct sock *))
259 struct ip_ra_chain *ra, *new_ra;
260 struct ip_ra_chain __rcu **rap;
262 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
265 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
267 spin_lock_bh(&ip_ra_lock);
268 for (rap = &ip_ra_chain;
269 (ra = rcu_dereference_protected(*rap,
270 lockdep_is_held(&ip_ra_lock))) != NULL;
274 spin_unlock_bh(&ip_ra_lock);
278 /* dont let ip_call_ra_chain() use sk again */
280 rcu_assign_pointer(*rap, ra->next);
281 spin_unlock_bh(&ip_ra_lock);
286 * Delay sock_put(sk) and kfree(ra) after one rcu grace
287 * period. This guarantee ip_call_ra_chain() dont need
288 * to mess with socket refcounts.
291 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
295 if (new_ra == NULL) {
296 spin_unlock_bh(&ip_ra_lock);
300 new_ra->destructor = destructor;
303 rcu_assign_pointer(*rap, new_ra);
305 spin_unlock_bh(&ip_ra_lock);
310 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
311 __be16 port, u32 info, u8 *payload)
313 struct sock_exterr_skb *serr;
315 skb = skb_clone(skb, GFP_ATOMIC);
319 serr = SKB_EXT_ERR(skb);
320 serr->ee.ee_errno = err;
321 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
322 serr->ee.ee_type = icmp_hdr(skb)->type;
323 serr->ee.ee_code = icmp_hdr(skb)->code;
325 serr->ee.ee_info = info;
326 serr->ee.ee_data = 0;
327 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
328 skb_network_header(skb);
331 if (skb_pull(skb, payload - skb->data) != NULL) {
332 skb_reset_transport_header(skb);
333 if (sock_queue_err_skb(sk, skb) == 0)
339 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
341 struct inet_sock *inet = inet_sk(sk);
342 struct sock_exterr_skb *serr;
349 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
353 skb_put(skb, sizeof(struct iphdr));
354 skb_reset_network_header(skb);
358 serr = SKB_EXT_ERR(skb);
359 serr->ee.ee_errno = err;
360 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
361 serr->ee.ee_type = 0;
362 serr->ee.ee_code = 0;
364 serr->ee.ee_info = info;
365 serr->ee.ee_data = 0;
366 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
369 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
370 skb_reset_transport_header(skb);
372 if (sock_queue_err_skb(sk, skb))
377 * Handle MSG_ERRQUEUE
379 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
381 struct sock_exterr_skb *serr;
382 struct sk_buff *skb, *skb2;
383 struct sockaddr_in *sin;
385 struct sock_extended_err ee;
386 struct sockaddr_in offender;
392 skb = skb_dequeue(&sk->sk_error_queue);
398 msg->msg_flags |= MSG_TRUNC;
401 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
405 sock_recv_timestamp(msg, sk, skb);
407 serr = SKB_EXT_ERR(skb);
409 sin = (struct sockaddr_in *)msg->msg_name;
411 sin->sin_family = AF_INET;
412 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
414 sin->sin_port = serr->port;
415 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
416 *addr_len = sizeof(*sin);
419 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
420 sin = &errhdr.offender;
421 memset(sin, 0, sizeof(*sin));
422 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
423 sin->sin_family = AF_INET;
424 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
425 if (inet_sk(sk)->cmsg_flags)
426 ip_cmsg_recv(msg, skb);
429 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
431 /* Now we could try to dump offended packet options */
433 msg->msg_flags |= MSG_ERRQUEUE;
436 /* Reset and regenerate socket error */
437 spin_lock_bh(&sk->sk_error_queue.lock);
439 skb2 = skb_peek(&sk->sk_error_queue);
441 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
442 spin_unlock_bh(&sk->sk_error_queue.lock);
443 sk->sk_error_report(sk);
445 spin_unlock_bh(&sk->sk_error_queue.lock);
454 static void opt_kfree_rcu(struct rcu_head *head)
456 kfree(container_of(head, struct ip_options_rcu, rcu));
460 * Socket option code for IP. This is the end of the line after any
461 * TCP,UDP etc options on an IP socket.
464 static int do_ip_setsockopt(struct sock *sk, int level,
465 int optname, char __user *optval, unsigned int optlen)
467 struct inet_sock *inet = inet_sk(sk);
479 case IP_MTU_DISCOVER:
481 case IP_ROUTER_ALERT:
487 case IP_MULTICAST_TTL:
488 case IP_MULTICAST_ALL:
489 case IP_MULTICAST_LOOP:
490 case IP_RECVORIGDSTADDR:
491 if (optlen >= sizeof(int)) {
492 if (get_user(val, (int __user *) optval))
494 } else if (optlen >= sizeof(char)) {
497 if (get_user(ucval, (unsigned char __user *) optval))
503 /* If optlen==0, it is equivalent to val == 0 */
505 if (ip_mroute_opt(optname))
506 return ip_mroute_setsockopt(sk, optname, optval, optlen);
514 struct ip_options_rcu *old, *opt = NULL;
518 err = ip_options_get_from_user(sock_net(sk), &opt,
522 old = rcu_dereference_protected(inet->inet_opt,
523 sock_owned_by_user(sk));
525 struct inet_connection_sock *icsk = inet_csk(sk);
526 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
527 if (sk->sk_family == PF_INET ||
528 (!((1 << sk->sk_state) &
529 (TCPF_LISTEN | TCPF_CLOSE)) &&
530 inet->inet_daddr != LOOPBACK4_IPV6)) {
533 icsk->icsk_ext_hdr_len -= old->opt.optlen;
535 icsk->icsk_ext_hdr_len += opt->opt.optlen;
536 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
537 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
541 rcu_assign_pointer(inet->inet_opt, opt);
543 call_rcu(&old->rcu, opt_kfree_rcu);
548 inet->cmsg_flags |= IP_CMSG_PKTINFO;
550 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
554 inet->cmsg_flags |= IP_CMSG_TTL;
556 inet->cmsg_flags &= ~IP_CMSG_TTL;
560 inet->cmsg_flags |= IP_CMSG_TOS;
562 inet->cmsg_flags &= ~IP_CMSG_TOS;
566 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
568 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
572 inet->cmsg_flags |= IP_CMSG_RETOPTS;
574 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
578 inet->cmsg_flags |= IP_CMSG_PASSSEC;
580 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
582 case IP_RECVORIGDSTADDR:
584 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
586 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
588 case IP_TOS: /* This sets both TOS and Precedence */
589 if (sk->sk_type == SOCK_STREAM) {
590 val &= ~INET_ECN_MASK;
591 val |= inet->tos & INET_ECN_MASK;
593 if (inet->tos != val) {
595 sk->sk_priority = rt_tos2priority(val);
602 if (val != -1 && (val < 1 || val > 255))
607 if (sk->sk_type != SOCK_RAW) {
611 inet->hdrincl = val ? 1 : 0;
614 if (sk->sk_type != SOCK_RAW) {
618 inet->nodefrag = val ? 1 : 0;
620 case IP_MTU_DISCOVER:
621 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
623 inet->pmtudisc = val;
626 inet->recverr = !!val;
628 skb_queue_purge(&sk->sk_error_queue);
630 case IP_MULTICAST_TTL:
631 if (sk->sk_type == SOCK_STREAM)
637 if (val < 0 || val > 255)
641 case IP_MULTICAST_LOOP:
644 inet->mc_loop = !!val;
646 case IP_MULTICAST_IF:
648 struct ip_mreqn mreq;
649 struct net_device *dev = NULL;
651 if (sk->sk_type == SOCK_STREAM)
654 * Check the arguments are allowable
657 if (optlen < sizeof(struct in_addr))
661 if (optlen >= sizeof(struct ip_mreqn)) {
662 if (copy_from_user(&mreq, optval, sizeof(mreq)))
665 memset(&mreq, 0, sizeof(mreq));
666 if (optlen >= sizeof(struct in_addr) &&
667 copy_from_user(&mreq.imr_address, optval,
668 sizeof(struct in_addr)))
672 if (!mreq.imr_ifindex) {
673 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
679 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
681 mreq.imr_ifindex = dev->ifindex;
683 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
686 err = -EADDRNOTAVAIL;
692 if (sk->sk_bound_dev_if &&
693 mreq.imr_ifindex != sk->sk_bound_dev_if)
696 inet->mc_index = mreq.imr_ifindex;
697 inet->mc_addr = mreq.imr_address.s_addr;
702 case IP_ADD_MEMBERSHIP:
703 case IP_DROP_MEMBERSHIP:
705 struct ip_mreqn mreq;
708 if (inet_sk(sk)->is_icsk)
711 if (optlen < sizeof(struct ip_mreq))
714 if (optlen >= sizeof(struct ip_mreqn)) {
715 if (copy_from_user(&mreq, optval, sizeof(mreq)))
718 memset(&mreq, 0, sizeof(mreq));
719 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
723 if (optname == IP_ADD_MEMBERSHIP)
724 err = ip_mc_join_group(sk, &mreq);
726 err = ip_mc_leave_group(sk, &mreq);
731 struct ip_msfilter *msf;
733 if (optlen < IP_MSFILTER_SIZE(0))
735 if (optlen > sysctl_optmem_max) {
739 msf = kmalloc(optlen, GFP_KERNEL);
745 if (copy_from_user(msf, optval, optlen)) {
749 /* numsrc >= (1G-4) overflow in 32 bits */
750 if (msf->imsf_numsrc >= 0x3ffffffcU ||
751 msf->imsf_numsrc > sysctl_igmp_max_msf) {
756 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
761 err = ip_mc_msfilter(sk, msf, 0);
765 case IP_BLOCK_SOURCE:
766 case IP_UNBLOCK_SOURCE:
767 case IP_ADD_SOURCE_MEMBERSHIP:
768 case IP_DROP_SOURCE_MEMBERSHIP:
770 struct ip_mreq_source mreqs;
773 if (optlen != sizeof(struct ip_mreq_source))
775 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
779 if (optname == IP_BLOCK_SOURCE) {
780 omode = MCAST_EXCLUDE;
782 } else if (optname == IP_UNBLOCK_SOURCE) {
783 omode = MCAST_EXCLUDE;
785 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
786 struct ip_mreqn mreq;
788 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
789 mreq.imr_address.s_addr = mreqs.imr_interface;
790 mreq.imr_ifindex = 0;
791 err = ip_mc_join_group(sk, &mreq);
792 if (err && err != -EADDRINUSE)
794 omode = MCAST_INCLUDE;
796 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
797 omode = MCAST_INCLUDE;
800 err = ip_mc_source(add, omode, sk, &mreqs, 0);
803 case MCAST_JOIN_GROUP:
804 case MCAST_LEAVE_GROUP:
806 struct group_req greq;
807 struct sockaddr_in *psin;
808 struct ip_mreqn mreq;
810 if (optlen < sizeof(struct group_req))
813 if (copy_from_user(&greq, optval, sizeof(greq)))
815 psin = (struct sockaddr_in *)&greq.gr_group;
816 if (psin->sin_family != AF_INET)
818 memset(&mreq, 0, sizeof(mreq));
819 mreq.imr_multiaddr = psin->sin_addr;
820 mreq.imr_ifindex = greq.gr_interface;
822 if (optname == MCAST_JOIN_GROUP)
823 err = ip_mc_join_group(sk, &mreq);
825 err = ip_mc_leave_group(sk, &mreq);
828 case MCAST_JOIN_SOURCE_GROUP:
829 case MCAST_LEAVE_SOURCE_GROUP:
830 case MCAST_BLOCK_SOURCE:
831 case MCAST_UNBLOCK_SOURCE:
833 struct group_source_req greqs;
834 struct ip_mreq_source mreqs;
835 struct sockaddr_in *psin;
838 if (optlen != sizeof(struct group_source_req))
840 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
844 if (greqs.gsr_group.ss_family != AF_INET ||
845 greqs.gsr_source.ss_family != AF_INET) {
846 err = -EADDRNOTAVAIL;
849 psin = (struct sockaddr_in *)&greqs.gsr_group;
850 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
851 psin = (struct sockaddr_in *)&greqs.gsr_source;
852 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
853 mreqs.imr_interface = 0; /* use index for mc_source */
855 if (optname == MCAST_BLOCK_SOURCE) {
856 omode = MCAST_EXCLUDE;
858 } else if (optname == MCAST_UNBLOCK_SOURCE) {
859 omode = MCAST_EXCLUDE;
861 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
862 struct ip_mreqn mreq;
864 psin = (struct sockaddr_in *)&greqs.gsr_group;
865 mreq.imr_multiaddr = psin->sin_addr;
866 mreq.imr_address.s_addr = 0;
867 mreq.imr_ifindex = greqs.gsr_interface;
868 err = ip_mc_join_group(sk, &mreq);
869 if (err && err != -EADDRINUSE)
871 greqs.gsr_interface = mreq.imr_ifindex;
872 omode = MCAST_INCLUDE;
874 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
875 omode = MCAST_INCLUDE;
878 err = ip_mc_source(add, omode, sk, &mreqs,
879 greqs.gsr_interface);
884 struct sockaddr_in *psin;
885 struct ip_msfilter *msf = NULL;
886 struct group_filter *gsf = NULL;
887 int msize, i, ifindex;
889 if (optlen < GROUP_FILTER_SIZE(0))
891 if (optlen > sysctl_optmem_max) {
895 gsf = kmalloc(optlen, GFP_KERNEL);
901 if (copy_from_user(gsf, optval, optlen))
904 /* numsrc >= (4G-140)/128 overflow in 32 bits */
905 if (gsf->gf_numsrc >= 0x1ffffff ||
906 gsf->gf_numsrc > sysctl_igmp_max_msf) {
910 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
914 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
915 msf = kmalloc(msize, GFP_KERNEL);
920 ifindex = gsf->gf_interface;
921 psin = (struct sockaddr_in *)&gsf->gf_group;
922 if (psin->sin_family != AF_INET) {
923 err = -EADDRNOTAVAIL;
926 msf->imsf_multiaddr = psin->sin_addr.s_addr;
927 msf->imsf_interface = 0;
928 msf->imsf_fmode = gsf->gf_fmode;
929 msf->imsf_numsrc = gsf->gf_numsrc;
930 err = -EADDRNOTAVAIL;
931 for (i = 0; i < gsf->gf_numsrc; ++i) {
932 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
934 if (psin->sin_family != AF_INET)
936 msf->imsf_slist[i] = psin->sin_addr.s_addr;
941 err = ip_mc_msfilter(sk, msf, ifindex);
947 case IP_MULTICAST_ALL:
950 if (val != 0 && val != 1)
954 case IP_ROUTER_ALERT:
955 err = ip_ra_control(sk, val ? 1 : 0, NULL);
961 inet->freebind = !!val;
964 case IP_IPSEC_POLICY:
967 if (!capable(CAP_NET_ADMIN))
969 err = xfrm_user_policy(sk, optname, optval, optlen);
973 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) {
979 inet->transparent = !!val;
985 if (val < 0 || val > 255)
1003 * ip_queue_rcv_skb - Queue an skb into sock receive queue
1007 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option
1008 * is not set, we drop skb dst entry now, while dst cache line is hot.
1010 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1012 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) &&
1013 !IPCB(skb)->opt.optlen)
1015 return sock_queue_rcv_skb(sk, skb);
1017 EXPORT_SYMBOL(ip_queue_rcv_skb);
1019 int ip_setsockopt(struct sock *sk, int level,
1020 int optname, char __user *optval, unsigned int optlen)
1024 if (level != SOL_IP)
1025 return -ENOPROTOOPT;
1027 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1028 #ifdef CONFIG_NETFILTER
1029 /* we need to exclude all possible ENOPROTOOPTs except default case */
1030 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1031 optname != IP_IPSEC_POLICY &&
1032 optname != IP_XFRM_POLICY &&
1033 !ip_mroute_opt(optname)) {
1035 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1041 EXPORT_SYMBOL(ip_setsockopt);
1043 #ifdef CONFIG_COMPAT
1044 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1045 char __user *optval, unsigned int optlen)
1049 if (level != SOL_IP)
1050 return -ENOPROTOOPT;
1052 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1053 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1056 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1057 #ifdef CONFIG_NETFILTER
1058 /* we need to exclude all possible ENOPROTOOPTs except default case */
1059 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1060 optname != IP_IPSEC_POLICY &&
1061 optname != IP_XFRM_POLICY &&
1062 !ip_mroute_opt(optname)) {
1064 err = compat_nf_setsockopt(sk, PF_INET, optname,
1071 EXPORT_SYMBOL(compat_ip_setsockopt);
1075 * Get the options. Note for future reference. The GET of IP options gets
1076 * the _received_ ones. The set sets the _sent_ ones.
1079 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1080 char __user *optval, int __user *optlen, unsigned flags)
1082 struct inet_sock *inet = inet_sk(sk);
1086 if (level != SOL_IP)
1089 if (ip_mroute_opt(optname))
1090 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1092 if (get_user(len, optlen))
1102 unsigned char optbuf[sizeof(struct ip_options)+40];
1103 struct ip_options *opt = (struct ip_options *)optbuf;
1104 struct ip_options_rcu *inet_opt;
1106 inet_opt = rcu_dereference_protected(inet->inet_opt,
1107 sock_owned_by_user(sk));
1110 memcpy(optbuf, &inet_opt->opt,
1111 sizeof(struct ip_options) +
1112 inet_opt->opt.optlen);
1115 if (opt->optlen == 0)
1116 return put_user(0, optlen);
1118 ip_options_undo(opt);
1120 len = min_t(unsigned int, len, opt->optlen);
1121 if (put_user(len, optlen))
1123 if (copy_to_user(optval, opt->__data, len))
1128 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1131 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1134 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1137 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1140 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1143 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1145 case IP_RECVORIGDSTADDR:
1146 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1152 val = (inet->uc_ttl == -1 ?
1153 sysctl_ip_default_ttl :
1157 val = inet->hdrincl;
1160 val = inet->nodefrag;
1162 case IP_MTU_DISCOVER:
1163 val = inet->pmtudisc;
1167 struct dst_entry *dst;
1169 dst = sk_dst_get(sk);
1181 val = inet->recverr;
1183 case IP_MULTICAST_TTL:
1186 case IP_MULTICAST_LOOP:
1187 val = inet->mc_loop;
1189 case IP_MULTICAST_IF:
1191 struct in_addr addr;
1192 len = min_t(unsigned int, len, sizeof(struct in_addr));
1193 addr.s_addr = inet->mc_addr;
1196 if (put_user(len, optlen))
1198 if (copy_to_user(optval, &addr, len))
1204 struct ip_msfilter msf;
1207 if (len < IP_MSFILTER_SIZE(0)) {
1211 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1215 err = ip_mc_msfget(sk, &msf,
1216 (struct ip_msfilter __user *)optval, optlen);
1220 case MCAST_MSFILTER:
1222 struct group_filter gsf;
1225 if (len < GROUP_FILTER_SIZE(0)) {
1229 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1233 err = ip_mc_gsfget(sk, &gsf,
1234 (struct group_filter __user *)optval,
1239 case IP_MULTICAST_ALL:
1248 if (sk->sk_type != SOCK_STREAM)
1249 return -ENOPROTOOPT;
1251 msg.msg_control = optval;
1252 msg.msg_controllen = len;
1253 msg.msg_flags = flags;
1255 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1256 struct in_pktinfo info;
1258 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1259 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1260 info.ipi_ifindex = inet->mc_index;
1261 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1263 if (inet->cmsg_flags & IP_CMSG_TTL) {
1264 int hlim = inet->mc_ttl;
1265 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1267 len -= msg.msg_controllen;
1268 return put_user(len, optlen);
1271 val = inet->freebind;
1273 case IP_TRANSPARENT:
1274 val = inet->transparent;
1277 val = inet->min_ttl;
1281 return -ENOPROTOOPT;
1285 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1286 unsigned char ucval = (unsigned char)val;
1288 if (put_user(len, optlen))
1290 if (copy_to_user(optval, &ucval, 1))
1293 len = min_t(unsigned int, sizeof(int), len);
1294 if (put_user(len, optlen))
1296 if (copy_to_user(optval, &val, len))
1302 int ip_getsockopt(struct sock *sk, int level,
1303 int optname, char __user *optval, int __user *optlen)
1307 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1308 #ifdef CONFIG_NETFILTER
1309 /* we need to exclude all possible ENOPROTOOPTs except default case */
1310 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1311 !ip_mroute_opt(optname)) {
1314 if (get_user(len, optlen))
1318 err = nf_getsockopt(sk, PF_INET, optname, optval,
1322 err = put_user(len, optlen);
1328 EXPORT_SYMBOL(ip_getsockopt);
1330 #ifdef CONFIG_COMPAT
1331 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1332 char __user *optval, int __user *optlen)
1336 if (optname == MCAST_MSFILTER)
1337 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1340 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1343 #ifdef CONFIG_NETFILTER
1344 /* we need to exclude all possible ENOPROTOOPTs except default case */
1345 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1346 !ip_mroute_opt(optname)) {
1349 if (get_user(len, optlen))
1353 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1356 err = put_user(len, optlen);
1362 EXPORT_SYMBOL(compat_ip_getsockopt);