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.
60 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
62 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
66 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
69 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
71 int ttl = ip_hdr(skb)->ttl;
72 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
75 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
77 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
80 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
82 if (IPCB(skb)->opt.optlen == 0)
85 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
90 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
92 unsigned char optbuf[sizeof(struct ip_options) + 40];
93 struct ip_options *opt = (struct ip_options *)optbuf;
95 if (IPCB(skb)->opt.optlen == 0)
98 if (ip_options_echo(opt, skb)) {
99 msg->msg_flags |= MSG_CTRUNC;
102 ip_options_undo(opt);
104 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
107 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
113 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
117 err = security_secid_to_secctx(secid, &secdata, &seclen);
121 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
122 security_release_secctx(secdata, seclen);
125 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
127 struct sockaddr_in sin;
128 const struct iphdr *iph = ip_hdr(skb);
129 __be16 *ports = (__be16 *)skb_transport_header(skb);
131 if (skb_transport_offset(skb) + 4 > skb->len)
134 /* All current transport protocols have the port numbers in the
135 * first four bytes of the transport header and this function is
136 * written with this assumption in mind.
139 sin.sin_family = AF_INET;
140 sin.sin_addr.s_addr = iph->daddr;
141 sin.sin_port = ports[1];
142 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
144 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
147 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
149 struct inet_sock *inet = inet_sk(skb->sk);
150 unsigned int flags = inet->cmsg_flags;
152 /* Ordered by supposed usage frequency */
154 ip_cmsg_recv_pktinfo(msg, skb);
155 if ((flags >>= 1) == 0)
159 ip_cmsg_recv_ttl(msg, skb);
160 if ((flags >>= 1) == 0)
164 ip_cmsg_recv_tos(msg, skb);
165 if ((flags >>= 1) == 0)
169 ip_cmsg_recv_opts(msg, skb);
170 if ((flags >>= 1) == 0)
174 ip_cmsg_recv_retopts(msg, skb);
175 if ((flags >>= 1) == 0)
179 ip_cmsg_recv_security(msg, skb);
181 if ((flags >>= 1) == 0)
184 ip_cmsg_recv_dstaddr(msg, skb);
187 EXPORT_SYMBOL(ip_cmsg_recv);
189 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
193 struct cmsghdr *cmsg;
195 for_each_cmsghdr(cmsg, msg) {
196 if (!CMSG_OK(msg, cmsg))
198 #if IS_ENABLED(CONFIG_IPV6)
200 cmsg->cmsg_level == SOL_IPV6 &&
201 cmsg->cmsg_type == IPV6_PKTINFO) {
202 struct in6_pktinfo *src_info;
204 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
206 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
207 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
209 ipc->oif = src_info->ipi6_ifindex;
210 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
214 if (cmsg->cmsg_level != SOL_IP)
216 switch (cmsg->cmsg_type) {
218 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
219 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
220 err < 40 ? err : 40);
226 struct in_pktinfo *info;
227 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
229 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
230 ipc->oif = info->ipi_ifindex;
231 ipc->addr = info->ipi_spec_dst.s_addr;
235 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
237 val = *(int *)CMSG_DATA(cmsg);
238 if (val < 1 || val > 255)
243 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
245 val = *(int *)CMSG_DATA(cmsg);
246 if (val < 0 || val > 255)
249 ipc->priority = rt_tos2priority(ipc->tos);
260 /* Special input handler for packets caught by router alert option.
261 They are selected only by protocol field, and then processed likely
262 local ones; but only if someone wants them! Otherwise, router
263 not running rsvpd will kill RSVP.
265 It is user level problem, what it will make with them.
266 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
267 but receiver should be enough clever f.e. to forward mtrace requests,
268 sent to multicast group to reach destination designated router.
270 struct ip_ra_chain __rcu *ip_ra_chain;
271 static DEFINE_SPINLOCK(ip_ra_lock);
274 static void ip_ra_destroy_rcu(struct rcu_head *head)
276 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
278 sock_put(ra->saved_sk);
282 int ip_ra_control(struct sock *sk, unsigned char on,
283 void (*destructor)(struct sock *))
285 struct ip_ra_chain *ra, *new_ra;
286 struct ip_ra_chain __rcu **rap;
288 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
291 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
293 spin_lock_bh(&ip_ra_lock);
294 for (rap = &ip_ra_chain;
295 (ra = rcu_dereference_protected(*rap,
296 lockdep_is_held(&ip_ra_lock))) != NULL;
300 spin_unlock_bh(&ip_ra_lock);
304 /* dont let ip_call_ra_chain() use sk again */
306 RCU_INIT_POINTER(*rap, ra->next);
307 spin_unlock_bh(&ip_ra_lock);
312 * Delay sock_put(sk) and kfree(ra) after one rcu grace
313 * period. This guarantee ip_call_ra_chain() dont need
314 * to mess with socket refcounts.
317 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
321 if (new_ra == NULL) {
322 spin_unlock_bh(&ip_ra_lock);
326 new_ra->destructor = destructor;
328 RCU_INIT_POINTER(new_ra->next, ra);
329 rcu_assign_pointer(*rap, new_ra);
331 spin_unlock_bh(&ip_ra_lock);
336 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
337 __be16 port, u32 info, u8 *payload)
339 struct sock_exterr_skb *serr;
341 skb = skb_clone(skb, GFP_ATOMIC);
345 serr = SKB_EXT_ERR(skb);
346 serr->ee.ee_errno = err;
347 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
348 serr->ee.ee_type = icmp_hdr(skb)->type;
349 serr->ee.ee_code = icmp_hdr(skb)->code;
351 serr->ee.ee_info = info;
352 serr->ee.ee_data = 0;
353 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
354 skb_network_header(skb);
357 if (skb_pull(skb, payload - skb->data) != NULL) {
358 skb_reset_transport_header(skb);
359 if (sock_queue_err_skb(sk, skb) == 0)
365 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
367 struct inet_sock *inet = inet_sk(sk);
368 struct sock_exterr_skb *serr;
375 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
379 skb_put(skb, sizeof(struct iphdr));
380 skb_reset_network_header(skb);
384 serr = SKB_EXT_ERR(skb);
385 serr->ee.ee_errno = err;
386 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
387 serr->ee.ee_type = 0;
388 serr->ee.ee_code = 0;
390 serr->ee.ee_info = info;
391 serr->ee.ee_data = 0;
392 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
395 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
396 skb_reset_transport_header(skb);
398 if (sock_queue_err_skb(sk, skb))
402 static bool ipv4_pktinfo_prepare_errqueue(const struct sock *sk,
403 const struct sk_buff *skb,
406 struct in_pktinfo *info = PKTINFO_SKB_CB(skb);
408 if ((ee_origin != SO_EE_ORIGIN_TIMESTAMPING) ||
409 (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
413 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
414 info->ipi_ifindex = skb->dev->ifindex;
419 * Handle MSG_ERRQUEUE
421 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
423 struct sock_exterr_skb *serr;
425 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
427 struct sock_extended_err ee;
428 struct sockaddr_in offender;
433 WARN_ON_ONCE(sk->sk_family == AF_INET6);
436 skb = sock_dequeue_err_skb(sk);
442 msg->msg_flags |= MSG_TRUNC;
445 err = skb_copy_datagram_msg(skb, 0, msg, copied);
449 sock_recv_timestamp(msg, sk, skb);
451 serr = SKB_EXT_ERR(skb);
454 sin->sin_family = AF_INET;
455 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
457 sin->sin_port = serr->port;
458 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
459 *addr_len = sizeof(*sin);
462 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
463 sin = &errhdr.offender;
464 memset(sin, 0, sizeof(*sin));
466 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
467 ipv4_pktinfo_prepare_errqueue(sk, skb, serr->ee.ee_origin)) {
468 sin->sin_family = AF_INET;
469 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
470 if (inet_sk(sk)->cmsg_flags)
471 ip_cmsg_recv(msg, skb);
474 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
476 /* Now we could try to dump offended packet options */
478 msg->msg_flags |= MSG_ERRQUEUE;
489 * Socket option code for IP. This is the end of the line after any
490 * TCP,UDP etc options on an IP socket.
493 static int do_ip_setsockopt(struct sock *sk, int level,
494 int optname, char __user *optval, unsigned int optlen)
496 struct inet_sock *inet = inet_sk(sk);
508 case IP_MTU_DISCOVER:
510 case IP_ROUTER_ALERT:
517 case IP_MULTICAST_TTL:
518 case IP_MULTICAST_ALL:
519 case IP_MULTICAST_LOOP:
520 case IP_RECVORIGDSTADDR:
521 if (optlen >= sizeof(int)) {
522 if (get_user(val, (int __user *) optval))
524 } else if (optlen >= sizeof(char)) {
527 if (get_user(ucval, (unsigned char __user *) optval))
533 /* If optlen==0, it is equivalent to val == 0 */
535 if (ip_mroute_opt(optname))
536 return ip_mroute_setsockopt(sk, optname, optval, optlen);
544 struct ip_options_rcu *old, *opt = NULL;
548 err = ip_options_get_from_user(sock_net(sk), &opt,
552 old = rcu_dereference_protected(inet->inet_opt,
553 sock_owned_by_user(sk));
555 struct inet_connection_sock *icsk = inet_csk(sk);
556 #if IS_ENABLED(CONFIG_IPV6)
557 if (sk->sk_family == PF_INET ||
558 (!((1 << sk->sk_state) &
559 (TCPF_LISTEN | TCPF_CLOSE)) &&
560 inet->inet_daddr != LOOPBACK4_IPV6)) {
563 icsk->icsk_ext_hdr_len -= old->opt.optlen;
565 icsk->icsk_ext_hdr_len += opt->opt.optlen;
566 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
567 #if IS_ENABLED(CONFIG_IPV6)
571 rcu_assign_pointer(inet->inet_opt, opt);
578 inet->cmsg_flags |= IP_CMSG_PKTINFO;
580 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
584 inet->cmsg_flags |= IP_CMSG_TTL;
586 inet->cmsg_flags &= ~IP_CMSG_TTL;
590 inet->cmsg_flags |= IP_CMSG_TOS;
592 inet->cmsg_flags &= ~IP_CMSG_TOS;
596 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
598 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
602 inet->cmsg_flags |= IP_CMSG_RETOPTS;
604 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
608 inet->cmsg_flags |= IP_CMSG_PASSSEC;
610 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
612 case IP_RECVORIGDSTADDR:
614 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
616 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
618 case IP_TOS: /* This sets both TOS and Precedence */
619 if (sk->sk_type == SOCK_STREAM) {
620 val &= ~INET_ECN_MASK;
621 val |= inet->tos & INET_ECN_MASK;
623 if (inet->tos != val) {
625 sk->sk_priority = rt_tos2priority(val);
632 if (val != -1 && (val < 1 || val > 255))
637 if (sk->sk_type != SOCK_RAW) {
641 inet->hdrincl = val ? 1 : 0;
644 if (sk->sk_type != SOCK_RAW) {
648 inet->nodefrag = val ? 1 : 0;
650 case IP_MTU_DISCOVER:
651 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
653 inet->pmtudisc = val;
656 inet->recverr = !!val;
658 skb_queue_purge(&sk->sk_error_queue);
660 case IP_MULTICAST_TTL:
661 if (sk->sk_type == SOCK_STREAM)
667 if (val < 0 || val > 255)
671 case IP_MULTICAST_LOOP:
674 inet->mc_loop = !!val;
678 struct net_device *dev = NULL;
681 if (optlen != sizeof(int))
684 ifindex = (__force int)ntohl((__force __be32)val);
691 dev = dev_get_by_index(sock_net(sk), ifindex);
692 err = -EADDRNOTAVAIL;
698 if (sk->sk_bound_dev_if)
701 inet->uc_index = ifindex;
705 case IP_MULTICAST_IF:
707 struct ip_mreqn mreq;
708 struct net_device *dev = NULL;
710 if (sk->sk_type == SOCK_STREAM)
713 * Check the arguments are allowable
716 if (optlen < sizeof(struct in_addr))
720 if (optlen >= sizeof(struct ip_mreqn)) {
721 if (copy_from_user(&mreq, optval, sizeof(mreq)))
724 memset(&mreq, 0, sizeof(mreq));
725 if (optlen >= sizeof(struct ip_mreq)) {
726 if (copy_from_user(&mreq, optval,
727 sizeof(struct ip_mreq)))
729 } else if (optlen >= sizeof(struct in_addr)) {
730 if (copy_from_user(&mreq.imr_address, optval,
731 sizeof(struct in_addr)))
736 if (!mreq.imr_ifindex) {
737 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
743 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
745 mreq.imr_ifindex = dev->ifindex;
747 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
750 err = -EADDRNOTAVAIL;
756 if (sk->sk_bound_dev_if &&
757 mreq.imr_ifindex != sk->sk_bound_dev_if)
760 inet->mc_index = mreq.imr_ifindex;
761 inet->mc_addr = mreq.imr_address.s_addr;
766 case IP_ADD_MEMBERSHIP:
767 case IP_DROP_MEMBERSHIP:
769 struct ip_mreqn mreq;
772 if (inet_sk(sk)->is_icsk)
775 if (optlen < sizeof(struct ip_mreq))
778 if (optlen >= sizeof(struct ip_mreqn)) {
779 if (copy_from_user(&mreq, optval, sizeof(mreq)))
782 memset(&mreq, 0, sizeof(mreq));
783 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
787 if (optname == IP_ADD_MEMBERSHIP)
788 err = ip_mc_join_group(sk, &mreq);
790 err = ip_mc_leave_group(sk, &mreq);
795 struct ip_msfilter *msf;
797 if (optlen < IP_MSFILTER_SIZE(0))
799 if (optlen > sysctl_optmem_max) {
803 msf = kmalloc(optlen, GFP_KERNEL);
809 if (copy_from_user(msf, optval, optlen)) {
813 /* numsrc >= (1G-4) overflow in 32 bits */
814 if (msf->imsf_numsrc >= 0x3ffffffcU ||
815 msf->imsf_numsrc > sysctl_igmp_max_msf) {
820 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
825 err = ip_mc_msfilter(sk, msf, 0);
829 case IP_BLOCK_SOURCE:
830 case IP_UNBLOCK_SOURCE:
831 case IP_ADD_SOURCE_MEMBERSHIP:
832 case IP_DROP_SOURCE_MEMBERSHIP:
834 struct ip_mreq_source mreqs;
837 if (optlen != sizeof(struct ip_mreq_source))
839 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
843 if (optname == IP_BLOCK_SOURCE) {
844 omode = MCAST_EXCLUDE;
846 } else if (optname == IP_UNBLOCK_SOURCE) {
847 omode = MCAST_EXCLUDE;
849 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
850 struct ip_mreqn mreq;
852 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
853 mreq.imr_address.s_addr = mreqs.imr_interface;
854 mreq.imr_ifindex = 0;
855 err = ip_mc_join_group(sk, &mreq);
856 if (err && err != -EADDRINUSE)
858 omode = MCAST_INCLUDE;
860 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
861 omode = MCAST_INCLUDE;
864 err = ip_mc_source(add, omode, sk, &mreqs, 0);
867 case MCAST_JOIN_GROUP:
868 case MCAST_LEAVE_GROUP:
870 struct group_req greq;
871 struct sockaddr_in *psin;
872 struct ip_mreqn mreq;
874 if (optlen < sizeof(struct group_req))
877 if (copy_from_user(&greq, optval, sizeof(greq)))
879 psin = (struct sockaddr_in *)&greq.gr_group;
880 if (psin->sin_family != AF_INET)
882 memset(&mreq, 0, sizeof(mreq));
883 mreq.imr_multiaddr = psin->sin_addr;
884 mreq.imr_ifindex = greq.gr_interface;
886 if (optname == MCAST_JOIN_GROUP)
887 err = ip_mc_join_group(sk, &mreq);
889 err = ip_mc_leave_group(sk, &mreq);
892 case MCAST_JOIN_SOURCE_GROUP:
893 case MCAST_LEAVE_SOURCE_GROUP:
894 case MCAST_BLOCK_SOURCE:
895 case MCAST_UNBLOCK_SOURCE:
897 struct group_source_req greqs;
898 struct ip_mreq_source mreqs;
899 struct sockaddr_in *psin;
902 if (optlen != sizeof(struct group_source_req))
904 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
908 if (greqs.gsr_group.ss_family != AF_INET ||
909 greqs.gsr_source.ss_family != AF_INET) {
910 err = -EADDRNOTAVAIL;
913 psin = (struct sockaddr_in *)&greqs.gsr_group;
914 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
915 psin = (struct sockaddr_in *)&greqs.gsr_source;
916 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
917 mreqs.imr_interface = 0; /* use index for mc_source */
919 if (optname == MCAST_BLOCK_SOURCE) {
920 omode = MCAST_EXCLUDE;
922 } else if (optname == MCAST_UNBLOCK_SOURCE) {
923 omode = MCAST_EXCLUDE;
925 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
926 struct ip_mreqn mreq;
928 psin = (struct sockaddr_in *)&greqs.gsr_group;
929 mreq.imr_multiaddr = psin->sin_addr;
930 mreq.imr_address.s_addr = 0;
931 mreq.imr_ifindex = greqs.gsr_interface;
932 err = ip_mc_join_group(sk, &mreq);
933 if (err && err != -EADDRINUSE)
935 greqs.gsr_interface = mreq.imr_ifindex;
936 omode = MCAST_INCLUDE;
938 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
939 omode = MCAST_INCLUDE;
942 err = ip_mc_source(add, omode, sk, &mreqs,
943 greqs.gsr_interface);
948 struct sockaddr_in *psin;
949 struct ip_msfilter *msf = NULL;
950 struct group_filter *gsf = NULL;
951 int msize, i, ifindex;
953 if (optlen < GROUP_FILTER_SIZE(0))
955 if (optlen > sysctl_optmem_max) {
959 gsf = kmalloc(optlen, GFP_KERNEL);
965 if (copy_from_user(gsf, optval, optlen))
968 /* numsrc >= (4G-140)/128 overflow in 32 bits */
969 if (gsf->gf_numsrc >= 0x1ffffff ||
970 gsf->gf_numsrc > sysctl_igmp_max_msf) {
974 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
978 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
979 msf = kmalloc(msize, GFP_KERNEL);
984 ifindex = gsf->gf_interface;
985 psin = (struct sockaddr_in *)&gsf->gf_group;
986 if (psin->sin_family != AF_INET) {
987 err = -EADDRNOTAVAIL;
990 msf->imsf_multiaddr = psin->sin_addr.s_addr;
991 msf->imsf_interface = 0;
992 msf->imsf_fmode = gsf->gf_fmode;
993 msf->imsf_numsrc = gsf->gf_numsrc;
994 err = -EADDRNOTAVAIL;
995 for (i = 0; i < gsf->gf_numsrc; ++i) {
996 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
998 if (psin->sin_family != AF_INET)
1000 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1005 err = ip_mc_msfilter(sk, msf, ifindex);
1011 case IP_MULTICAST_ALL:
1014 if (val != 0 && val != 1)
1018 case IP_ROUTER_ALERT:
1019 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1025 inet->freebind = !!val;
1028 case IP_IPSEC_POLICY:
1029 case IP_XFRM_POLICY:
1031 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1033 err = xfrm_user_policy(sk, optname, optval, optlen);
1036 case IP_TRANSPARENT:
1037 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1038 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1044 inet->transparent = !!val;
1050 if (val < 0 || val > 255)
1052 inet->min_ttl = val;
1068 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1072 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1073 * destination in skb->cb[] before dst drop.
1074 * This way, receiver doesn't make cache line misses to read rtable.
1076 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1078 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1079 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1082 if (prepare && skb_rtable(skb)) {
1083 pktinfo->ipi_ifindex = inet_iif(skb);
1084 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1086 pktinfo->ipi_ifindex = 0;
1087 pktinfo->ipi_spec_dst.s_addr = 0;
1092 int ip_setsockopt(struct sock *sk, int level,
1093 int optname, char __user *optval, unsigned int optlen)
1097 if (level != SOL_IP)
1098 return -ENOPROTOOPT;
1100 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1101 #ifdef CONFIG_NETFILTER
1102 /* we need to exclude all possible ENOPROTOOPTs except default case */
1103 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1104 optname != IP_IPSEC_POLICY &&
1105 optname != IP_XFRM_POLICY &&
1106 !ip_mroute_opt(optname)) {
1108 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1114 EXPORT_SYMBOL(ip_setsockopt);
1116 #ifdef CONFIG_COMPAT
1117 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1118 char __user *optval, unsigned int optlen)
1122 if (level != SOL_IP)
1123 return -ENOPROTOOPT;
1125 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1126 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1129 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1130 #ifdef CONFIG_NETFILTER
1131 /* we need to exclude all possible ENOPROTOOPTs except default case */
1132 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1133 optname != IP_IPSEC_POLICY &&
1134 optname != IP_XFRM_POLICY &&
1135 !ip_mroute_opt(optname)) {
1137 err = compat_nf_setsockopt(sk, PF_INET, optname,
1144 EXPORT_SYMBOL(compat_ip_setsockopt);
1148 * Get the options. Note for future reference. The GET of IP options gets
1149 * the _received_ ones. The set sets the _sent_ ones.
1152 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1153 char __user *optval, int __user *optlen, unsigned int flags)
1155 struct inet_sock *inet = inet_sk(sk);
1159 if (level != SOL_IP)
1162 if (ip_mroute_opt(optname))
1163 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1165 if (get_user(len, optlen))
1175 unsigned char optbuf[sizeof(struct ip_options)+40];
1176 struct ip_options *opt = (struct ip_options *)optbuf;
1177 struct ip_options_rcu *inet_opt;
1179 inet_opt = rcu_dereference_protected(inet->inet_opt,
1180 sock_owned_by_user(sk));
1183 memcpy(optbuf, &inet_opt->opt,
1184 sizeof(struct ip_options) +
1185 inet_opt->opt.optlen);
1188 if (opt->optlen == 0)
1189 return put_user(0, optlen);
1191 ip_options_undo(opt);
1193 len = min_t(unsigned int, len, opt->optlen);
1194 if (put_user(len, optlen))
1196 if (copy_to_user(optval, opt->__data, len))
1201 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1204 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1207 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1210 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1213 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1216 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1218 case IP_RECVORIGDSTADDR:
1219 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1225 val = (inet->uc_ttl == -1 ?
1226 sysctl_ip_default_ttl :
1230 val = inet->hdrincl;
1233 val = inet->nodefrag;
1235 case IP_MTU_DISCOVER:
1236 val = inet->pmtudisc;
1240 struct dst_entry *dst;
1242 dst = sk_dst_get(sk);
1254 val = inet->recverr;
1256 case IP_MULTICAST_TTL:
1259 case IP_MULTICAST_LOOP:
1260 val = inet->mc_loop;
1263 val = (__force int)htonl((__u32) inet->uc_index);
1265 case IP_MULTICAST_IF:
1267 struct in_addr addr;
1268 len = min_t(unsigned int, len, sizeof(struct in_addr));
1269 addr.s_addr = inet->mc_addr;
1272 if (put_user(len, optlen))
1274 if (copy_to_user(optval, &addr, len))
1280 struct ip_msfilter msf;
1283 if (len < IP_MSFILTER_SIZE(0)) {
1287 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1291 err = ip_mc_msfget(sk, &msf,
1292 (struct ip_msfilter __user *)optval, optlen);
1296 case MCAST_MSFILTER:
1298 struct group_filter gsf;
1301 if (len < GROUP_FILTER_SIZE(0)) {
1305 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1309 err = ip_mc_gsfget(sk, &gsf,
1310 (struct group_filter __user *)optval,
1315 case IP_MULTICAST_ALL:
1324 if (sk->sk_type != SOCK_STREAM)
1325 return -ENOPROTOOPT;
1327 msg.msg_control = (__force void *) optval;
1328 msg.msg_controllen = len;
1329 msg.msg_flags = flags;
1331 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1332 struct in_pktinfo info;
1334 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1335 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1336 info.ipi_ifindex = inet->mc_index;
1337 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1339 if (inet->cmsg_flags & IP_CMSG_TTL) {
1340 int hlim = inet->mc_ttl;
1341 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1343 if (inet->cmsg_flags & IP_CMSG_TOS) {
1344 int tos = inet->rcv_tos;
1345 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1347 len -= msg.msg_controllen;
1348 return put_user(len, optlen);
1351 val = inet->freebind;
1353 case IP_TRANSPARENT:
1354 val = inet->transparent;
1357 val = inet->min_ttl;
1361 return -ENOPROTOOPT;
1365 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1366 unsigned char ucval = (unsigned char)val;
1368 if (put_user(len, optlen))
1370 if (copy_to_user(optval, &ucval, 1))
1373 len = min_t(unsigned int, sizeof(int), len);
1374 if (put_user(len, optlen))
1376 if (copy_to_user(optval, &val, len))
1382 int ip_getsockopt(struct sock *sk, int level,
1383 int optname, char __user *optval, int __user *optlen)
1387 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1388 #ifdef CONFIG_NETFILTER
1389 /* we need to exclude all possible ENOPROTOOPTs except default case */
1390 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1391 !ip_mroute_opt(optname)) {
1394 if (get_user(len, optlen))
1398 err = nf_getsockopt(sk, PF_INET, optname, optval,
1402 err = put_user(len, optlen);
1408 EXPORT_SYMBOL(ip_getsockopt);
1410 #ifdef CONFIG_COMPAT
1411 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1412 char __user *optval, int __user *optlen)
1416 if (optname == MCAST_MSFILTER)
1417 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1420 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1423 #ifdef CONFIG_NETFILTER
1424 /* we need to exclude all possible ENOPROTOOPTs except default case */
1425 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1426 !ip_mroute_opt(optname)) {
1429 if (get_user(len, optlen))
1433 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1436 err = put_user(len, optlen);
1442 EXPORT_SYMBOL(compat_ip_getsockopt);