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 #include <net/checksum.h>
41 #if IS_ENABLED(CONFIG_IPV6)
42 #include <net/transp_v6.h>
44 #include <net/ip_fib.h>
46 #include <linux/errqueue.h>
47 #include <linux/uaccess.h>
50 * SOL_IP control messages.
53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
55 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
57 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
59 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
64 int ttl = ip_hdr(skb)->ttl;
65 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
70 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
75 if (IPCB(skb)->opt.optlen == 0)
78 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
85 unsigned char optbuf[sizeof(struct ip_options) + 40];
86 struct ip_options *opt = (struct ip_options *)optbuf;
88 if (IPCB(skb)->opt.optlen == 0)
91 if (ip_options_echo(opt, skb)) {
92 msg->msg_flags |= MSG_CTRUNC;
97 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
100 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
104 if (IPCB(skb)->frag_max_size == 0)
107 val = IPCB(skb)->frag_max_size;
108 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
111 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
112 int tlen, int offset)
114 __wsum csum = skb->csum;
116 if (skb->ip_summed != CHECKSUM_COMPLETE)
120 int tend_off = skb_transport_offset(skb) + tlen;
121 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
124 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
127 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
133 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
137 err = security_secid_to_secctx(secid, &secdata, &seclen);
141 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
142 security_release_secctx(secdata, seclen);
145 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
147 struct sockaddr_in sin;
148 const struct iphdr *iph = ip_hdr(skb);
149 __be16 *ports = (__be16 *)skb_transport_header(skb);
151 if (skb_transport_offset(skb) + 4 > (int)skb->len)
154 /* All current transport protocols have the port numbers in the
155 * first four bytes of the transport header and this function is
156 * written with this assumption in mind.
159 sin.sin_family = AF_INET;
160 sin.sin_addr.s_addr = iph->daddr;
161 sin.sin_port = ports[1];
162 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
164 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
167 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
168 struct sk_buff *skb, int tlen, int offset)
170 struct inet_sock *inet = inet_sk(sk);
171 unsigned int flags = inet->cmsg_flags;
173 /* Ordered by supposed usage frequency */
174 if (flags & IP_CMSG_PKTINFO) {
175 ip_cmsg_recv_pktinfo(msg, skb);
177 flags &= ~IP_CMSG_PKTINFO;
182 if (flags & IP_CMSG_TTL) {
183 ip_cmsg_recv_ttl(msg, skb);
185 flags &= ~IP_CMSG_TTL;
190 if (flags & IP_CMSG_TOS) {
191 ip_cmsg_recv_tos(msg, skb);
193 flags &= ~IP_CMSG_TOS;
198 if (flags & IP_CMSG_RECVOPTS) {
199 ip_cmsg_recv_opts(msg, skb);
201 flags &= ~IP_CMSG_RECVOPTS;
206 if (flags & IP_CMSG_RETOPTS) {
207 ip_cmsg_recv_retopts(msg, skb);
209 flags &= ~IP_CMSG_RETOPTS;
214 if (flags & IP_CMSG_PASSSEC) {
215 ip_cmsg_recv_security(msg, skb);
217 flags &= ~IP_CMSG_PASSSEC;
222 if (flags & IP_CMSG_ORIGDSTADDR) {
223 ip_cmsg_recv_dstaddr(msg, skb);
225 flags &= ~IP_CMSG_ORIGDSTADDR;
230 if (flags & IP_CMSG_CHECKSUM)
231 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
233 if (flags & IP_CMSG_RECVFRAGSIZE)
234 ip_cmsg_recv_fragsize(msg, skb);
236 EXPORT_SYMBOL(ip_cmsg_recv_offset);
238 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
242 struct cmsghdr *cmsg;
243 struct net *net = sock_net(sk);
245 for_each_cmsghdr(cmsg, msg) {
246 if (!CMSG_OK(msg, cmsg))
248 #if IS_ENABLED(CONFIG_IPV6)
250 cmsg->cmsg_level == SOL_IPV6 &&
251 cmsg->cmsg_type == IPV6_PKTINFO) {
252 struct in6_pktinfo *src_info;
254 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
256 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
257 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
259 ipc->oif = src_info->ipi6_ifindex;
260 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
264 if (cmsg->cmsg_level == SOL_SOCKET) {
265 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
271 if (cmsg->cmsg_level != SOL_IP)
273 switch (cmsg->cmsg_type) {
275 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
277 /* Our caller is responsible for freeing ipc->opt */
278 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
279 err < 40 ? err : 40);
285 struct in_pktinfo *info;
286 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
288 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
289 ipc->oif = info->ipi_ifindex;
290 ipc->addr = info->ipi_spec_dst.s_addr;
294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
296 val = *(int *)CMSG_DATA(cmsg);
297 if (val < 1 || val > 255)
302 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
303 val = *(int *)CMSG_DATA(cmsg);
304 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
305 val = *(u8 *)CMSG_DATA(cmsg);
308 if (val < 0 || val > 255)
311 ipc->priority = rt_tos2priority(ipc->tos);
322 /* Special input handler for packets caught by router alert option.
323 They are selected only by protocol field, and then processed likely
324 local ones; but only if someone wants them! Otherwise, router
325 not running rsvpd will kill RSVP.
327 It is user level problem, what it will make with them.
328 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
329 but receiver should be enough clever f.e. to forward mtrace requests,
330 sent to multicast group to reach destination designated router.
332 struct ip_ra_chain __rcu *ip_ra_chain;
333 static DEFINE_SPINLOCK(ip_ra_lock);
336 static void ip_ra_destroy_rcu(struct rcu_head *head)
338 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
340 sock_put(ra->saved_sk);
344 int ip_ra_control(struct sock *sk, unsigned char on,
345 void (*destructor)(struct sock *))
347 struct ip_ra_chain *ra, *new_ra;
348 struct ip_ra_chain __rcu **rap;
350 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
353 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
355 spin_lock_bh(&ip_ra_lock);
356 for (rap = &ip_ra_chain;
357 (ra = rcu_dereference_protected(*rap,
358 lockdep_is_held(&ip_ra_lock))) != NULL;
362 spin_unlock_bh(&ip_ra_lock);
366 /* dont let ip_call_ra_chain() use sk again */
368 RCU_INIT_POINTER(*rap, ra->next);
369 spin_unlock_bh(&ip_ra_lock);
374 * Delay sock_put(sk) and kfree(ra) after one rcu grace
375 * period. This guarantee ip_call_ra_chain() dont need
376 * to mess with socket refcounts.
379 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
384 spin_unlock_bh(&ip_ra_lock);
388 new_ra->destructor = destructor;
390 RCU_INIT_POINTER(new_ra->next, ra);
391 rcu_assign_pointer(*rap, new_ra);
393 spin_unlock_bh(&ip_ra_lock);
398 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
399 __be16 port, u32 info, u8 *payload)
401 struct sock_exterr_skb *serr;
403 skb = skb_clone(skb, GFP_ATOMIC);
407 serr = SKB_EXT_ERR(skb);
408 serr->ee.ee_errno = err;
409 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
410 serr->ee.ee_type = icmp_hdr(skb)->type;
411 serr->ee.ee_code = icmp_hdr(skb)->code;
413 serr->ee.ee_info = info;
414 serr->ee.ee_data = 0;
415 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
416 skb_network_header(skb);
419 if (skb_pull(skb, payload - skb->data)) {
420 skb_reset_transport_header(skb);
421 if (sock_queue_err_skb(sk, skb) == 0)
427 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
429 struct inet_sock *inet = inet_sk(sk);
430 struct sock_exterr_skb *serr;
437 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
441 skb_put(skb, sizeof(struct iphdr));
442 skb_reset_network_header(skb);
446 serr = SKB_EXT_ERR(skb);
447 serr->ee.ee_errno = err;
448 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
449 serr->ee.ee_type = 0;
450 serr->ee.ee_code = 0;
452 serr->ee.ee_info = info;
453 serr->ee.ee_data = 0;
454 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
457 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
458 skb_reset_transport_header(skb);
460 if (sock_queue_err_skb(sk, skb))
464 /* For some errors we have valid addr_offset even with zero payload and
465 * zero port. Also, addr_offset should be supported if port is set.
467 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
469 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
470 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
473 /* IPv4 supports cmsg on all imcp errors and some timestamps
475 * Timestamp code paths do not initialize the fields expected by cmsg:
476 * the PKTINFO fields in skb->cb[]. Fill those in here.
478 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
482 struct in_pktinfo *info;
484 if (ee_origin == SO_EE_ORIGIN_ICMP)
487 if (ee_origin == SO_EE_ORIGIN_LOCAL)
490 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
491 * timestamp with egress dev. Not possible for packets without dev
492 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
494 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
498 info = PKTINFO_SKB_CB(skb);
499 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
500 info->ipi_ifindex = skb->dev->ifindex;
505 * Handle MSG_ERRQUEUE
507 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
509 struct sock_exterr_skb *serr;
511 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
513 struct sock_extended_err ee;
514 struct sockaddr_in offender;
519 WARN_ON_ONCE(sk->sk_family == AF_INET6);
522 skb = sock_dequeue_err_skb(sk);
528 msg->msg_flags |= MSG_TRUNC;
531 err = skb_copy_datagram_msg(skb, 0, msg, copied);
536 sock_recv_timestamp(msg, sk, skb);
538 serr = SKB_EXT_ERR(skb);
540 if (sin && ipv4_datagram_support_addr(serr)) {
541 sin->sin_family = AF_INET;
542 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
544 sin->sin_port = serr->port;
545 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
546 *addr_len = sizeof(*sin);
549 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
550 sin = &errhdr.offender;
551 memset(sin, 0, sizeof(*sin));
553 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
554 sin->sin_family = AF_INET;
555 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
556 if (inet_sk(sk)->cmsg_flags)
557 ip_cmsg_recv(msg, skb);
560 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
562 /* Now we could try to dump offended packet options */
564 msg->msg_flags |= MSG_ERRQUEUE;
574 * Socket option code for IP. This is the end of the line after any
575 * TCP,UDP etc options on an IP socket.
577 static bool setsockopt_needs_rtnl(int optname)
580 case IP_ADD_MEMBERSHIP:
581 case IP_ADD_SOURCE_MEMBERSHIP:
582 case IP_BLOCK_SOURCE:
583 case IP_DROP_MEMBERSHIP:
584 case IP_DROP_SOURCE_MEMBERSHIP:
586 case IP_UNBLOCK_SOURCE:
587 case MCAST_BLOCK_SOURCE:
589 case MCAST_JOIN_GROUP:
590 case MCAST_JOIN_SOURCE_GROUP:
591 case MCAST_LEAVE_GROUP:
592 case MCAST_LEAVE_SOURCE_GROUP:
593 case MCAST_UNBLOCK_SOURCE:
599 static int do_ip_setsockopt(struct sock *sk, int level,
600 int optname, char __user *optval, unsigned int optlen)
602 struct inet_sock *inet = inet_sk(sk);
603 struct net *net = sock_net(sk);
605 bool needs_rtnl = setsockopt_needs_rtnl(optname);
616 case IP_MTU_DISCOVER:
618 case IP_ROUTER_ALERT:
624 case IP_BIND_ADDRESS_NO_PORT:
626 case IP_MULTICAST_TTL:
627 case IP_MULTICAST_ALL:
628 case IP_MULTICAST_LOOP:
629 case IP_RECVORIGDSTADDR:
631 case IP_RECVFRAGSIZE:
632 if (optlen >= sizeof(int)) {
633 if (get_user(val, (int __user *) optval))
635 } else if (optlen >= sizeof(char)) {
638 if (get_user(ucval, (unsigned char __user *) optval))
644 /* If optlen==0, it is equivalent to val == 0 */
646 if (ip_mroute_opt(optname))
647 return ip_mroute_setsockopt(sk, optname, optval, optlen);
657 struct ip_options_rcu *old, *opt = NULL;
661 err = ip_options_get_from_user(sock_net(sk), &opt,
665 old = rcu_dereference_protected(inet->inet_opt,
666 lockdep_sock_is_held(sk));
668 struct inet_connection_sock *icsk = inet_csk(sk);
669 #if IS_ENABLED(CONFIG_IPV6)
670 if (sk->sk_family == PF_INET ||
671 (!((1 << sk->sk_state) &
672 (TCPF_LISTEN | TCPF_CLOSE)) &&
673 inet->inet_daddr != LOOPBACK4_IPV6)) {
676 icsk->icsk_ext_hdr_len -= old->opt.optlen;
678 icsk->icsk_ext_hdr_len += opt->opt.optlen;
679 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
680 #if IS_ENABLED(CONFIG_IPV6)
684 rcu_assign_pointer(inet->inet_opt, opt);
691 inet->cmsg_flags |= IP_CMSG_PKTINFO;
693 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
697 inet->cmsg_flags |= IP_CMSG_TTL;
699 inet->cmsg_flags &= ~IP_CMSG_TTL;
703 inet->cmsg_flags |= IP_CMSG_TOS;
705 inet->cmsg_flags &= ~IP_CMSG_TOS;
709 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
711 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
715 inet->cmsg_flags |= IP_CMSG_RETOPTS;
717 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
721 inet->cmsg_flags |= IP_CMSG_PASSSEC;
723 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
725 case IP_RECVORIGDSTADDR:
727 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
729 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
733 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
734 inet_inc_convert_csum(sk);
735 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
738 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
739 inet_dec_convert_csum(sk);
740 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
744 case IP_RECVFRAGSIZE:
745 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
748 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
750 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
752 case IP_TOS: /* This sets both TOS and Precedence */
753 if (sk->sk_type == SOCK_STREAM) {
754 val &= ~INET_ECN_MASK;
755 val |= inet->tos & INET_ECN_MASK;
757 if (inet->tos != val) {
759 sk->sk_priority = rt_tos2priority(val);
766 if (val != -1 && (val < 1 || val > 255))
771 if (sk->sk_type != SOCK_RAW) {
775 inet->hdrincl = val ? 1 : 0;
778 if (sk->sk_type != SOCK_RAW) {
782 inet->nodefrag = val ? 1 : 0;
784 case IP_BIND_ADDRESS_NO_PORT:
785 inet->bind_address_no_port = val ? 1 : 0;
787 case IP_MTU_DISCOVER:
788 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
790 inet->pmtudisc = val;
793 inet->recverr = !!val;
795 skb_queue_purge(&sk->sk_error_queue);
797 case IP_MULTICAST_TTL:
798 if (sk->sk_type == SOCK_STREAM)
804 if (val < 0 || val > 255)
808 case IP_MULTICAST_LOOP:
811 inet->mc_loop = !!val;
815 struct net_device *dev = NULL;
818 if (optlen != sizeof(int))
821 ifindex = (__force int)ntohl((__force __be32)val);
828 dev = dev_get_by_index(sock_net(sk), ifindex);
829 err = -EADDRNOTAVAIL;
835 if (sk->sk_bound_dev_if)
838 inet->uc_index = ifindex;
842 case IP_MULTICAST_IF:
844 struct ip_mreqn mreq;
845 struct net_device *dev = NULL;
848 if (sk->sk_type == SOCK_STREAM)
851 * Check the arguments are allowable
854 if (optlen < sizeof(struct in_addr))
858 if (optlen >= sizeof(struct ip_mreqn)) {
859 if (copy_from_user(&mreq, optval, sizeof(mreq)))
862 memset(&mreq, 0, sizeof(mreq));
863 if (optlen >= sizeof(struct ip_mreq)) {
864 if (copy_from_user(&mreq, optval,
865 sizeof(struct ip_mreq)))
867 } else if (optlen >= sizeof(struct in_addr)) {
868 if (copy_from_user(&mreq.imr_address, optval,
869 sizeof(struct in_addr)))
874 if (!mreq.imr_ifindex) {
875 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
881 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
883 mreq.imr_ifindex = dev->ifindex;
885 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
888 err = -EADDRNOTAVAIL;
892 midx = l3mdev_master_ifindex(dev);
897 if (sk->sk_bound_dev_if &&
898 mreq.imr_ifindex != sk->sk_bound_dev_if &&
899 (!midx || midx != sk->sk_bound_dev_if))
902 inet->mc_index = mreq.imr_ifindex;
903 inet->mc_addr = mreq.imr_address.s_addr;
908 case IP_ADD_MEMBERSHIP:
909 case IP_DROP_MEMBERSHIP:
911 struct ip_mreqn mreq;
914 if (inet_sk(sk)->is_icsk)
917 if (optlen < sizeof(struct ip_mreq))
920 if (optlen >= sizeof(struct ip_mreqn)) {
921 if (copy_from_user(&mreq, optval, sizeof(mreq)))
924 memset(&mreq, 0, sizeof(mreq));
925 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
929 if (optname == IP_ADD_MEMBERSHIP)
930 err = ip_mc_join_group(sk, &mreq);
932 err = ip_mc_leave_group(sk, &mreq);
937 struct ip_msfilter *msf;
939 if (optlen < IP_MSFILTER_SIZE(0))
941 if (optlen > sysctl_optmem_max) {
945 msf = kmalloc(optlen, GFP_KERNEL);
951 if (copy_from_user(msf, optval, optlen)) {
955 /* numsrc >= (1G-4) overflow in 32 bits */
956 if (msf->imsf_numsrc >= 0x3ffffffcU ||
957 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
962 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
967 err = ip_mc_msfilter(sk, msf, 0);
971 case IP_BLOCK_SOURCE:
972 case IP_UNBLOCK_SOURCE:
973 case IP_ADD_SOURCE_MEMBERSHIP:
974 case IP_DROP_SOURCE_MEMBERSHIP:
976 struct ip_mreq_source mreqs;
979 if (optlen != sizeof(struct ip_mreq_source))
981 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
985 if (optname == IP_BLOCK_SOURCE) {
986 omode = MCAST_EXCLUDE;
988 } else if (optname == IP_UNBLOCK_SOURCE) {
989 omode = MCAST_EXCLUDE;
991 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
992 struct ip_mreqn mreq;
994 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
995 mreq.imr_address.s_addr = mreqs.imr_interface;
996 mreq.imr_ifindex = 0;
997 err = ip_mc_join_group(sk, &mreq);
998 if (err && err != -EADDRINUSE)
1000 omode = MCAST_INCLUDE;
1002 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1003 omode = MCAST_INCLUDE;
1006 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1009 case MCAST_JOIN_GROUP:
1010 case MCAST_LEAVE_GROUP:
1012 struct group_req greq;
1013 struct sockaddr_in *psin;
1014 struct ip_mreqn mreq;
1016 if (optlen < sizeof(struct group_req))
1019 if (copy_from_user(&greq, optval, sizeof(greq)))
1021 psin = (struct sockaddr_in *)&greq.gr_group;
1022 if (psin->sin_family != AF_INET)
1024 memset(&mreq, 0, sizeof(mreq));
1025 mreq.imr_multiaddr = psin->sin_addr;
1026 mreq.imr_ifindex = greq.gr_interface;
1028 if (optname == MCAST_JOIN_GROUP)
1029 err = ip_mc_join_group(sk, &mreq);
1031 err = ip_mc_leave_group(sk, &mreq);
1034 case MCAST_JOIN_SOURCE_GROUP:
1035 case MCAST_LEAVE_SOURCE_GROUP:
1036 case MCAST_BLOCK_SOURCE:
1037 case MCAST_UNBLOCK_SOURCE:
1039 struct group_source_req greqs;
1040 struct ip_mreq_source mreqs;
1041 struct sockaddr_in *psin;
1044 if (optlen != sizeof(struct group_source_req))
1046 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1050 if (greqs.gsr_group.ss_family != AF_INET ||
1051 greqs.gsr_source.ss_family != AF_INET) {
1052 err = -EADDRNOTAVAIL;
1055 psin = (struct sockaddr_in *)&greqs.gsr_group;
1056 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1057 psin = (struct sockaddr_in *)&greqs.gsr_source;
1058 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1059 mreqs.imr_interface = 0; /* use index for mc_source */
1061 if (optname == MCAST_BLOCK_SOURCE) {
1062 omode = MCAST_EXCLUDE;
1064 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1065 omode = MCAST_EXCLUDE;
1067 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1068 struct ip_mreqn mreq;
1070 psin = (struct sockaddr_in *)&greqs.gsr_group;
1071 mreq.imr_multiaddr = psin->sin_addr;
1072 mreq.imr_address.s_addr = 0;
1073 mreq.imr_ifindex = greqs.gsr_interface;
1074 err = ip_mc_join_group(sk, &mreq);
1075 if (err && err != -EADDRINUSE)
1077 greqs.gsr_interface = mreq.imr_ifindex;
1078 omode = MCAST_INCLUDE;
1080 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1081 omode = MCAST_INCLUDE;
1084 err = ip_mc_source(add, omode, sk, &mreqs,
1085 greqs.gsr_interface);
1088 case MCAST_MSFILTER:
1090 struct sockaddr_in *psin;
1091 struct ip_msfilter *msf = NULL;
1092 struct group_filter *gsf = NULL;
1093 int msize, i, ifindex;
1095 if (optlen < GROUP_FILTER_SIZE(0))
1097 if (optlen > sysctl_optmem_max) {
1101 gsf = kmalloc(optlen, GFP_KERNEL);
1107 if (copy_from_user(gsf, optval, optlen))
1110 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1111 if (gsf->gf_numsrc >= 0x1ffffff ||
1112 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1116 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1120 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1121 msf = kmalloc(msize, GFP_KERNEL);
1126 ifindex = gsf->gf_interface;
1127 psin = (struct sockaddr_in *)&gsf->gf_group;
1128 if (psin->sin_family != AF_INET) {
1129 err = -EADDRNOTAVAIL;
1132 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1133 msf->imsf_interface = 0;
1134 msf->imsf_fmode = gsf->gf_fmode;
1135 msf->imsf_numsrc = gsf->gf_numsrc;
1136 err = -EADDRNOTAVAIL;
1137 for (i = 0; i < gsf->gf_numsrc; ++i) {
1138 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1140 if (psin->sin_family != AF_INET)
1142 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1147 err = ip_mc_msfilter(sk, msf, ifindex);
1153 case IP_MULTICAST_ALL:
1156 if (val != 0 && val != 1)
1160 case IP_ROUTER_ALERT:
1161 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1167 inet->freebind = !!val;
1170 case IP_IPSEC_POLICY:
1171 case IP_XFRM_POLICY:
1173 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1175 err = xfrm_user_policy(sk, optname, optval, optlen);
1178 case IP_TRANSPARENT:
1179 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1180 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1186 inet->transparent = !!val;
1192 if (val < 0 || val > 255)
1194 inet->min_ttl = val;
1214 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1218 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1219 * destination in skb->cb[] before dst drop.
1220 * This way, receiver doesn't make cache line misses to read rtable.
1222 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1224 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1225 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1228 if (prepare && skb_rtable(skb)) {
1229 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1230 * which has interface index (iif) as the first member of the
1231 * underlying inet{6}_skb_parm struct. This code then overlays
1232 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1233 * element so the iif is picked up from the prior IPCB. If iif
1234 * is the loopback interface, then return the sending interface
1235 * (e.g., process binds socket to eth0 for Tx which is
1236 * redirected to loopback in the rtable/dst).
1238 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1239 pktinfo->ipi_ifindex = inet_iif(skb);
1241 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1243 pktinfo->ipi_ifindex = 0;
1244 pktinfo->ipi_spec_dst.s_addr = 0;
1246 /* We need to keep the dst for __ip_options_echo()
1247 * We could restrict the test to opt.ts_needtime || opt.srr,
1248 * but the following is good enough as IP options are not often used.
1250 if (unlikely(IPCB(skb)->opt.optlen))
1256 int ip_setsockopt(struct sock *sk, int level,
1257 int optname, char __user *optval, unsigned int optlen)
1261 if (level != SOL_IP)
1262 return -ENOPROTOOPT;
1264 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1265 #ifdef CONFIG_NETFILTER
1266 /* we need to exclude all possible ENOPROTOOPTs except default case */
1267 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1268 optname != IP_IPSEC_POLICY &&
1269 optname != IP_XFRM_POLICY &&
1270 !ip_mroute_opt(optname)) {
1272 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1278 EXPORT_SYMBOL(ip_setsockopt);
1280 #ifdef CONFIG_COMPAT
1281 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1282 char __user *optval, unsigned int optlen)
1286 if (level != SOL_IP)
1287 return -ENOPROTOOPT;
1289 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1290 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1293 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1294 #ifdef CONFIG_NETFILTER
1295 /* we need to exclude all possible ENOPROTOOPTs except default case */
1296 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1297 optname != IP_IPSEC_POLICY &&
1298 optname != IP_XFRM_POLICY &&
1299 !ip_mroute_opt(optname)) {
1301 err = compat_nf_setsockopt(sk, PF_INET, optname,
1308 EXPORT_SYMBOL(compat_ip_setsockopt);
1312 * Get the options. Note for future reference. The GET of IP options gets
1313 * the _received_ ones. The set sets the _sent_ ones.
1316 static bool getsockopt_needs_rtnl(int optname)
1320 case MCAST_MSFILTER:
1326 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1327 char __user *optval, int __user *optlen, unsigned int flags)
1329 struct inet_sock *inet = inet_sk(sk);
1330 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1334 if (level != SOL_IP)
1337 if (ip_mroute_opt(optname))
1338 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1340 if (get_user(len, optlen))
1352 unsigned char optbuf[sizeof(struct ip_options)+40];
1353 struct ip_options *opt = (struct ip_options *)optbuf;
1354 struct ip_options_rcu *inet_opt;
1356 inet_opt = rcu_dereference_protected(inet->inet_opt,
1357 lockdep_sock_is_held(sk));
1360 memcpy(optbuf, &inet_opt->opt,
1361 sizeof(struct ip_options) +
1362 inet_opt->opt.optlen);
1365 if (opt->optlen == 0)
1366 return put_user(0, optlen);
1368 ip_options_undo(opt);
1370 len = min_t(unsigned int, len, opt->optlen);
1371 if (put_user(len, optlen))
1373 if (copy_to_user(optval, opt->__data, len))
1378 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1381 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1384 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1387 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1390 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1393 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1395 case IP_RECVORIGDSTADDR:
1396 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1399 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1401 case IP_RECVFRAGSIZE:
1402 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1409 struct net *net = sock_net(sk);
1410 val = (inet->uc_ttl == -1 ?
1411 net->ipv4.sysctl_ip_default_ttl :
1416 val = inet->hdrincl;
1419 val = inet->nodefrag;
1421 case IP_BIND_ADDRESS_NO_PORT:
1422 val = inet->bind_address_no_port;
1424 case IP_MTU_DISCOVER:
1425 val = inet->pmtudisc;
1429 struct dst_entry *dst;
1431 dst = sk_dst_get(sk);
1443 val = inet->recverr;
1445 case IP_MULTICAST_TTL:
1448 case IP_MULTICAST_LOOP:
1449 val = inet->mc_loop;
1452 val = (__force int)htonl((__u32) inet->uc_index);
1454 case IP_MULTICAST_IF:
1456 struct in_addr addr;
1457 len = min_t(unsigned int, len, sizeof(struct in_addr));
1458 addr.s_addr = inet->mc_addr;
1461 if (put_user(len, optlen))
1463 if (copy_to_user(optval, &addr, len))
1469 struct ip_msfilter msf;
1471 if (len < IP_MSFILTER_SIZE(0)) {
1475 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1479 err = ip_mc_msfget(sk, &msf,
1480 (struct ip_msfilter __user *)optval, optlen);
1483 case MCAST_MSFILTER:
1485 struct group_filter gsf;
1487 if (len < GROUP_FILTER_SIZE(0)) {
1491 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1495 err = ip_mc_gsfget(sk, &gsf,
1496 (struct group_filter __user *)optval,
1500 case IP_MULTICAST_ALL:
1509 if (sk->sk_type != SOCK_STREAM)
1510 return -ENOPROTOOPT;
1512 msg.msg_control = (__force void *) optval;
1513 msg.msg_controllen = len;
1514 msg.msg_flags = flags;
1516 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1517 struct in_pktinfo info;
1519 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1520 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1521 info.ipi_ifindex = inet->mc_index;
1522 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1524 if (inet->cmsg_flags & IP_CMSG_TTL) {
1525 int hlim = inet->mc_ttl;
1526 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1528 if (inet->cmsg_flags & IP_CMSG_TOS) {
1529 int tos = inet->rcv_tos;
1530 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1532 len -= msg.msg_controllen;
1533 return put_user(len, optlen);
1536 val = inet->freebind;
1538 case IP_TRANSPARENT:
1539 val = inet->transparent;
1542 val = inet->min_ttl;
1546 return -ENOPROTOOPT;
1550 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1551 unsigned char ucval = (unsigned char)val;
1553 if (put_user(len, optlen))
1555 if (copy_to_user(optval, &ucval, 1))
1558 len = min_t(unsigned int, sizeof(int), len);
1559 if (put_user(len, optlen))
1561 if (copy_to_user(optval, &val, len))
1573 int ip_getsockopt(struct sock *sk, int level,
1574 int optname, char __user *optval, int __user *optlen)
1578 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1579 #ifdef CONFIG_NETFILTER
1580 /* we need to exclude all possible ENOPROTOOPTs except default case */
1581 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1582 !ip_mroute_opt(optname)) {
1585 if (get_user(len, optlen))
1589 err = nf_getsockopt(sk, PF_INET, optname, optval,
1593 err = put_user(len, optlen);
1599 EXPORT_SYMBOL(ip_getsockopt);
1601 #ifdef CONFIG_COMPAT
1602 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1603 char __user *optval, int __user *optlen)
1607 if (optname == MCAST_MSFILTER)
1608 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1611 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1614 #ifdef CONFIG_NETFILTER
1615 /* we need to exclude all possible ENOPROTOOPTs except default case */
1616 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1617 !ip_mroute_opt(optname)) {
1620 if (get_user(len, optlen))
1624 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1627 err = put_user(len, optlen);
1633 EXPORT_SYMBOL(compat_ip_getsockopt);