1 // SPDX-License-Identifier: GPL-2.0
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
12 * Many : Split from ip.c , see ip.c for history.
13 * Martin Mares : TOS setting fixed.
14 * Alan Cox : Fixed a couple of oopses in Martin's
16 * Mike McLagan : Routing by source
19 #include <linux/module.h>
20 #include <linux/types.h>
22 #include <linux/skbuff.h>
24 #include <linux/icmp.h>
25 #include <linux/inetdevice.h>
26 #include <linux/netdevice.h>
27 #include <linux/slab.h>
31 #include <net/tcp_states.h>
32 #include <linux/udp.h>
33 #include <linux/igmp.h>
34 #include <linux/netfilter.h>
35 #include <linux/route.h>
36 #include <linux/mroute.h>
37 #include <net/inet_ecn.h>
38 #include <net/route.h>
40 #include <net/compat.h>
41 #include <net/checksum.h>
42 #if IS_ENABLED(CONFIG_IPV6)
43 #include <net/transp_v6.h>
45 #include <net/ip_fib.h>
47 #include <linux/errqueue.h>
48 #include <linux/uaccess.h>
50 #include <linux/bpfilter.h>
53 * SOL_IP control messages.
56 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
58 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
60 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
62 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
65 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
67 int ttl = ip_hdr(skb)->ttl;
68 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
71 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
73 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
76 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
78 if (IPCB(skb)->opt.optlen == 0)
81 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
86 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
89 unsigned char optbuf[sizeof(struct ip_options) + 40];
90 struct ip_options *opt = (struct ip_options *)optbuf;
92 if (IPCB(skb)->opt.optlen == 0)
95 if (ip_options_echo(net, opt, skb)) {
96 msg->msg_flags |= MSG_CTRUNC;
101 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
104 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
108 if (IPCB(skb)->frag_max_size == 0)
111 val = IPCB(skb)->frag_max_size;
112 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
115 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
116 int tlen, int offset)
118 __wsum csum = skb->csum;
120 if (skb->ip_summed != CHECKSUM_COMPLETE)
124 int tend_off = skb_transport_offset(skb) + tlen;
125 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
128 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
131 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
137 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
141 err = security_secid_to_secctx(secid, &secdata, &seclen);
145 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
146 security_release_secctx(secdata, seclen);
149 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
151 struct sockaddr_in sin;
152 const struct iphdr *iph = ip_hdr(skb);
156 end = skb_transport_offset(skb) + 4;
157 if (end > 0 && !pskb_may_pull(skb, end))
160 /* All current transport protocols have the port numbers in the
161 * first four bytes of the transport header and this function is
162 * written with this assumption in mind.
164 ports = (__be16 *)skb_transport_header(skb);
166 sin.sin_family = AF_INET;
167 sin.sin_addr.s_addr = iph->daddr;
168 sin.sin_port = ports[1];
169 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
171 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
174 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
175 struct sk_buff *skb, int tlen, int offset)
177 struct inet_sock *inet = inet_sk(sk);
178 unsigned int flags = inet->cmsg_flags;
180 /* Ordered by supposed usage frequency */
181 if (flags & IP_CMSG_PKTINFO) {
182 ip_cmsg_recv_pktinfo(msg, skb);
184 flags &= ~IP_CMSG_PKTINFO;
189 if (flags & IP_CMSG_TTL) {
190 ip_cmsg_recv_ttl(msg, skb);
192 flags &= ~IP_CMSG_TTL;
197 if (flags & IP_CMSG_TOS) {
198 ip_cmsg_recv_tos(msg, skb);
200 flags &= ~IP_CMSG_TOS;
205 if (flags & IP_CMSG_RECVOPTS) {
206 ip_cmsg_recv_opts(msg, skb);
208 flags &= ~IP_CMSG_RECVOPTS;
213 if (flags & IP_CMSG_RETOPTS) {
214 ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
216 flags &= ~IP_CMSG_RETOPTS;
221 if (flags & IP_CMSG_PASSSEC) {
222 ip_cmsg_recv_security(msg, skb);
224 flags &= ~IP_CMSG_PASSSEC;
229 if (flags & IP_CMSG_ORIGDSTADDR) {
230 ip_cmsg_recv_dstaddr(msg, skb);
232 flags &= ~IP_CMSG_ORIGDSTADDR;
237 if (flags & IP_CMSG_CHECKSUM)
238 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
240 if (flags & IP_CMSG_RECVFRAGSIZE)
241 ip_cmsg_recv_fragsize(msg, skb);
243 EXPORT_SYMBOL(ip_cmsg_recv_offset);
245 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
249 struct cmsghdr *cmsg;
250 struct net *net = sock_net(sk);
252 for_each_cmsghdr(cmsg, msg) {
253 if (!CMSG_OK(msg, cmsg))
255 #if IS_ENABLED(CONFIG_IPV6)
257 cmsg->cmsg_level == SOL_IPV6 &&
258 cmsg->cmsg_type == IPV6_PKTINFO) {
259 struct in6_pktinfo *src_info;
261 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
263 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
264 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
266 if (src_info->ipi6_ifindex)
267 ipc->oif = src_info->ipi6_ifindex;
268 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
272 if (cmsg->cmsg_level == SOL_SOCKET) {
273 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
279 if (cmsg->cmsg_level != SOL_IP)
281 switch (cmsg->cmsg_type) {
283 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
285 /* Our caller is responsible for freeing ipc->opt */
286 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
287 err < 40 ? err : 40);
293 struct in_pktinfo *info;
294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
296 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
297 if (info->ipi_ifindex)
298 ipc->oif = info->ipi_ifindex;
299 ipc->addr = info->ipi_spec_dst.s_addr;
303 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
305 val = *(int *)CMSG_DATA(cmsg);
306 if (val < 1 || val > 255)
311 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
312 val = *(int *)CMSG_DATA(cmsg);
313 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
314 val = *(u8 *)CMSG_DATA(cmsg);
317 if (val < 0 || val > 255)
320 ipc->priority = rt_tos2priority(ipc->tos);
330 static void ip_ra_destroy_rcu(struct rcu_head *head)
332 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
334 sock_put(ra->saved_sk);
338 int ip_ra_control(struct sock *sk, unsigned char on,
339 void (*destructor)(struct sock *))
341 struct ip_ra_chain *ra, *new_ra;
342 struct ip_ra_chain __rcu **rap;
343 struct net *net = sock_net(sk);
345 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
348 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
350 mutex_lock(&net->ipv4.ra_mutex);
351 for (rap = &net->ipv4.ra_chain;
352 (ra = rcu_dereference_protected(*rap,
353 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
357 mutex_unlock(&net->ipv4.ra_mutex);
361 /* dont let ip_call_ra_chain() use sk again */
363 RCU_INIT_POINTER(*rap, ra->next);
364 mutex_unlock(&net->ipv4.ra_mutex);
369 * Delay sock_put(sk) and kfree(ra) after one rcu grace
370 * period. This guarantee ip_call_ra_chain() dont need
371 * to mess with socket refcounts.
374 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
379 mutex_unlock(&net->ipv4.ra_mutex);
383 new_ra->destructor = destructor;
385 RCU_INIT_POINTER(new_ra->next, ra);
386 rcu_assign_pointer(*rap, new_ra);
388 mutex_unlock(&net->ipv4.ra_mutex);
393 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
394 __be16 port, u32 info, u8 *payload)
396 struct sock_exterr_skb *serr;
398 skb = skb_clone(skb, GFP_ATOMIC);
402 serr = SKB_EXT_ERR(skb);
403 serr->ee.ee_errno = err;
404 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
405 serr->ee.ee_type = icmp_hdr(skb)->type;
406 serr->ee.ee_code = icmp_hdr(skb)->code;
408 serr->ee.ee_info = info;
409 serr->ee.ee_data = 0;
410 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
411 skb_network_header(skb);
414 if (skb_pull(skb, payload - skb->data)) {
415 skb_reset_transport_header(skb);
416 if (sock_queue_err_skb(sk, skb) == 0)
422 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
424 struct inet_sock *inet = inet_sk(sk);
425 struct sock_exterr_skb *serr;
432 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
436 skb_put(skb, sizeof(struct iphdr));
437 skb_reset_network_header(skb);
441 serr = SKB_EXT_ERR(skb);
442 serr->ee.ee_errno = err;
443 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
444 serr->ee.ee_type = 0;
445 serr->ee.ee_code = 0;
447 serr->ee.ee_info = info;
448 serr->ee.ee_data = 0;
449 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
452 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
453 skb_reset_transport_header(skb);
455 if (sock_queue_err_skb(sk, skb))
459 /* For some errors we have valid addr_offset even with zero payload and
460 * zero port. Also, addr_offset should be supported if port is set.
462 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
464 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
465 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
468 /* IPv4 supports cmsg on all imcp errors and some timestamps
470 * Timestamp code paths do not initialize the fields expected by cmsg:
471 * the PKTINFO fields in skb->cb[]. Fill those in here.
473 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
477 struct in_pktinfo *info;
479 if (ee_origin == SO_EE_ORIGIN_ICMP)
482 if (ee_origin == SO_EE_ORIGIN_LOCAL)
485 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
486 * timestamp with egress dev. Not possible for packets without iif
487 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
489 info = PKTINFO_SKB_CB(skb);
490 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
494 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
499 * Handle MSG_ERRQUEUE
501 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
503 struct sock_exterr_skb *serr;
505 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
507 struct sock_extended_err ee;
508 struct sockaddr_in offender;
514 skb = sock_dequeue_err_skb(sk);
520 msg->msg_flags |= MSG_TRUNC;
523 err = skb_copy_datagram_msg(skb, 0, msg, copied);
528 sock_recv_timestamp(msg, sk, skb);
530 serr = SKB_EXT_ERR(skb);
532 if (sin && ipv4_datagram_support_addr(serr)) {
533 sin->sin_family = AF_INET;
534 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
536 sin->sin_port = serr->port;
537 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
538 *addr_len = sizeof(*sin);
541 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
542 sin = &errhdr.offender;
543 memset(sin, 0, sizeof(*sin));
545 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
546 sin->sin_family = AF_INET;
547 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
548 if (inet_sk(sk)->cmsg_flags)
549 ip_cmsg_recv(msg, skb);
552 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
554 /* Now we could try to dump offended packet options */
556 msg->msg_flags |= MSG_ERRQUEUE;
566 * Socket option code for IP. This is the end of the line after any
567 * TCP,UDP etc options on an IP socket.
569 static bool setsockopt_needs_rtnl(int optname)
572 case IP_ADD_MEMBERSHIP:
573 case IP_ADD_SOURCE_MEMBERSHIP:
574 case IP_BLOCK_SOURCE:
575 case IP_DROP_MEMBERSHIP:
576 case IP_DROP_SOURCE_MEMBERSHIP:
578 case IP_UNBLOCK_SOURCE:
579 case MCAST_BLOCK_SOURCE:
581 case MCAST_JOIN_GROUP:
582 case MCAST_JOIN_SOURCE_GROUP:
583 case MCAST_LEAVE_GROUP:
584 case MCAST_LEAVE_SOURCE_GROUP:
585 case MCAST_UNBLOCK_SOURCE:
591 static int do_ip_setsockopt(struct sock *sk, int level,
592 int optname, char __user *optval, unsigned int optlen)
594 struct inet_sock *inet = inet_sk(sk);
595 struct net *net = sock_net(sk);
597 bool needs_rtnl = setsockopt_needs_rtnl(optname);
608 case IP_MTU_DISCOVER:
610 case IP_ROUTER_ALERT:
616 case IP_BIND_ADDRESS_NO_PORT:
618 case IP_MULTICAST_TTL:
619 case IP_MULTICAST_ALL:
620 case IP_MULTICAST_LOOP:
621 case IP_RECVORIGDSTADDR:
623 case IP_RECVFRAGSIZE:
624 if (optlen >= sizeof(int)) {
625 if (get_user(val, (int __user *) optval))
627 } else if (optlen >= sizeof(char)) {
630 if (get_user(ucval, (unsigned char __user *) optval))
636 /* If optlen==0, it is equivalent to val == 0 */
638 if (optname == IP_ROUTER_ALERT)
639 return ip_ra_control(sk, val ? 1 : 0, NULL);
640 if (ip_mroute_opt(optname))
641 return ip_mroute_setsockopt(sk, optname, optval, optlen);
651 struct ip_options_rcu *old, *opt = NULL;
655 err = ip_options_get_from_user(sock_net(sk), &opt,
659 old = rcu_dereference_protected(inet->inet_opt,
660 lockdep_sock_is_held(sk));
662 struct inet_connection_sock *icsk = inet_csk(sk);
663 #if IS_ENABLED(CONFIG_IPV6)
664 if (sk->sk_family == PF_INET ||
665 (!((1 << sk->sk_state) &
666 (TCPF_LISTEN | TCPF_CLOSE)) &&
667 inet->inet_daddr != LOOPBACK4_IPV6)) {
670 icsk->icsk_ext_hdr_len -= old->opt.optlen;
672 icsk->icsk_ext_hdr_len += opt->opt.optlen;
673 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
674 #if IS_ENABLED(CONFIG_IPV6)
678 rcu_assign_pointer(inet->inet_opt, opt);
685 inet->cmsg_flags |= IP_CMSG_PKTINFO;
687 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
691 inet->cmsg_flags |= IP_CMSG_TTL;
693 inet->cmsg_flags &= ~IP_CMSG_TTL;
697 inet->cmsg_flags |= IP_CMSG_TOS;
699 inet->cmsg_flags &= ~IP_CMSG_TOS;
703 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
705 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
709 inet->cmsg_flags |= IP_CMSG_RETOPTS;
711 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
715 inet->cmsg_flags |= IP_CMSG_PASSSEC;
717 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
719 case IP_RECVORIGDSTADDR:
721 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
723 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
727 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
728 inet_inc_convert_csum(sk);
729 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
732 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
733 inet_dec_convert_csum(sk);
734 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
738 case IP_RECVFRAGSIZE:
739 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
742 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
744 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
746 case IP_TOS: /* This sets both TOS and Precedence */
747 if (sk->sk_type == SOCK_STREAM) {
748 val &= ~INET_ECN_MASK;
749 val |= inet->tos & INET_ECN_MASK;
751 if (inet->tos != val) {
753 sk->sk_priority = rt_tos2priority(val);
760 if (val != -1 && (val < 1 || val > 255))
765 if (sk->sk_type != SOCK_RAW) {
769 inet->hdrincl = val ? 1 : 0;
772 if (sk->sk_type != SOCK_RAW) {
776 inet->nodefrag = val ? 1 : 0;
778 case IP_BIND_ADDRESS_NO_PORT:
779 inet->bind_address_no_port = val ? 1 : 0;
781 case IP_MTU_DISCOVER:
782 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
784 inet->pmtudisc = val;
787 inet->recverr = !!val;
789 skb_queue_purge(&sk->sk_error_queue);
791 case IP_MULTICAST_TTL:
792 if (sk->sk_type == SOCK_STREAM)
798 if (val < 0 || val > 255)
802 case IP_MULTICAST_LOOP:
805 inet->mc_loop = !!val;
809 struct net_device *dev = NULL;
813 if (optlen != sizeof(int))
816 ifindex = (__force int)ntohl((__force __be32)val);
823 dev = dev_get_by_index(sock_net(sk), ifindex);
824 err = -EADDRNOTAVAIL;
828 midx = l3mdev_master_ifindex(dev);
832 if (sk->sk_bound_dev_if &&
833 (!midx || midx != sk->sk_bound_dev_if))
836 inet->uc_index = ifindex;
840 case IP_MULTICAST_IF:
842 struct ip_mreqn mreq;
843 struct net_device *dev = NULL;
846 if (sk->sk_type == SOCK_STREAM)
849 * Check the arguments are allowable
852 if (optlen < sizeof(struct in_addr))
856 if (optlen >= sizeof(struct ip_mreqn)) {
857 if (copy_from_user(&mreq, optval, sizeof(mreq)))
860 memset(&mreq, 0, sizeof(mreq));
861 if (optlen >= sizeof(struct ip_mreq)) {
862 if (copy_from_user(&mreq, optval,
863 sizeof(struct ip_mreq)))
865 } else if (optlen >= sizeof(struct in_addr)) {
866 if (copy_from_user(&mreq.imr_address, optval,
867 sizeof(struct in_addr)))
872 if (!mreq.imr_ifindex) {
873 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
879 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
881 mreq.imr_ifindex = dev->ifindex;
883 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
886 err = -EADDRNOTAVAIL;
890 midx = l3mdev_master_ifindex(dev);
895 if (sk->sk_bound_dev_if &&
896 mreq.imr_ifindex != sk->sk_bound_dev_if &&
897 (!midx || midx != sk->sk_bound_dev_if))
900 inet->mc_index = mreq.imr_ifindex;
901 inet->mc_addr = mreq.imr_address.s_addr;
906 case IP_ADD_MEMBERSHIP:
907 case IP_DROP_MEMBERSHIP:
909 struct ip_mreqn mreq;
912 if (inet_sk(sk)->is_icsk)
915 if (optlen < sizeof(struct ip_mreq))
918 if (optlen >= sizeof(struct ip_mreqn)) {
919 if (copy_from_user(&mreq, optval, sizeof(mreq)))
922 memset(&mreq, 0, sizeof(mreq));
923 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
927 if (optname == IP_ADD_MEMBERSHIP)
928 err = ip_mc_join_group(sk, &mreq);
930 err = ip_mc_leave_group(sk, &mreq);
935 struct ip_msfilter *msf;
937 if (optlen < IP_MSFILTER_SIZE(0))
939 if (optlen > sysctl_optmem_max) {
943 msf = memdup_user(optval, optlen);
948 /* numsrc >= (1G-4) overflow in 32 bits */
949 if (msf->imsf_numsrc >= 0x3ffffffcU ||
950 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
955 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
960 err = ip_mc_msfilter(sk, msf, 0);
964 case IP_BLOCK_SOURCE:
965 case IP_UNBLOCK_SOURCE:
966 case IP_ADD_SOURCE_MEMBERSHIP:
967 case IP_DROP_SOURCE_MEMBERSHIP:
969 struct ip_mreq_source mreqs;
972 if (optlen != sizeof(struct ip_mreq_source))
974 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
978 if (optname == IP_BLOCK_SOURCE) {
979 omode = MCAST_EXCLUDE;
981 } else if (optname == IP_UNBLOCK_SOURCE) {
982 omode = MCAST_EXCLUDE;
984 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
985 struct ip_mreqn mreq;
987 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
988 mreq.imr_address.s_addr = mreqs.imr_interface;
989 mreq.imr_ifindex = 0;
990 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
991 if (err && err != -EADDRINUSE)
993 omode = MCAST_INCLUDE;
995 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
996 omode = MCAST_INCLUDE;
999 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1002 case MCAST_JOIN_GROUP:
1003 case MCAST_LEAVE_GROUP:
1005 struct group_req greq;
1006 struct sockaddr_in *psin;
1007 struct ip_mreqn mreq;
1009 if (optlen < sizeof(struct group_req))
1012 if (copy_from_user(&greq, optval, sizeof(greq)))
1014 psin = (struct sockaddr_in *)&greq.gr_group;
1015 if (psin->sin_family != AF_INET)
1017 memset(&mreq, 0, sizeof(mreq));
1018 mreq.imr_multiaddr = psin->sin_addr;
1019 mreq.imr_ifindex = greq.gr_interface;
1021 if (optname == MCAST_JOIN_GROUP)
1022 err = ip_mc_join_group(sk, &mreq);
1024 err = ip_mc_leave_group(sk, &mreq);
1027 case MCAST_JOIN_SOURCE_GROUP:
1028 case MCAST_LEAVE_SOURCE_GROUP:
1029 case MCAST_BLOCK_SOURCE:
1030 case MCAST_UNBLOCK_SOURCE:
1032 struct group_source_req greqs;
1033 struct ip_mreq_source mreqs;
1034 struct sockaddr_in *psin;
1037 if (optlen != sizeof(struct group_source_req))
1039 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1043 if (greqs.gsr_group.ss_family != AF_INET ||
1044 greqs.gsr_source.ss_family != AF_INET) {
1045 err = -EADDRNOTAVAIL;
1048 psin = (struct sockaddr_in *)&greqs.gsr_group;
1049 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1050 psin = (struct sockaddr_in *)&greqs.gsr_source;
1051 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1052 mreqs.imr_interface = 0; /* use index for mc_source */
1054 if (optname == MCAST_BLOCK_SOURCE) {
1055 omode = MCAST_EXCLUDE;
1057 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1058 omode = MCAST_EXCLUDE;
1060 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1061 struct ip_mreqn mreq;
1063 psin = (struct sockaddr_in *)&greqs.gsr_group;
1064 mreq.imr_multiaddr = psin->sin_addr;
1065 mreq.imr_address.s_addr = 0;
1066 mreq.imr_ifindex = greqs.gsr_interface;
1067 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1068 if (err && err != -EADDRINUSE)
1070 greqs.gsr_interface = mreq.imr_ifindex;
1071 omode = MCAST_INCLUDE;
1073 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1074 omode = MCAST_INCLUDE;
1077 err = ip_mc_source(add, omode, sk, &mreqs,
1078 greqs.gsr_interface);
1081 case MCAST_MSFILTER:
1083 struct sockaddr_in *psin;
1084 struct ip_msfilter *msf = NULL;
1085 struct group_filter *gsf = NULL;
1086 int msize, i, ifindex;
1088 if (optlen < GROUP_FILTER_SIZE(0))
1090 if (optlen > sysctl_optmem_max) {
1094 gsf = memdup_user(optval, optlen);
1100 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1101 if (gsf->gf_numsrc >= 0x1ffffff ||
1102 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1106 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1110 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1111 msf = kmalloc(msize, GFP_KERNEL);
1116 ifindex = gsf->gf_interface;
1117 psin = (struct sockaddr_in *)&gsf->gf_group;
1118 if (psin->sin_family != AF_INET) {
1119 err = -EADDRNOTAVAIL;
1122 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1123 msf->imsf_interface = 0;
1124 msf->imsf_fmode = gsf->gf_fmode;
1125 msf->imsf_numsrc = gsf->gf_numsrc;
1126 err = -EADDRNOTAVAIL;
1127 for (i = 0; i < gsf->gf_numsrc; ++i) {
1128 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1130 if (psin->sin_family != AF_INET)
1132 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1137 err = ip_mc_msfilter(sk, msf, ifindex);
1143 case IP_MULTICAST_ALL:
1146 if (val != 0 && val != 1)
1154 inet->freebind = !!val;
1157 case IP_IPSEC_POLICY:
1158 case IP_XFRM_POLICY:
1160 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1162 err = xfrm_user_policy(sk, optname, optval, optlen);
1165 case IP_TRANSPARENT:
1166 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1167 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1173 inet->transparent = !!val;
1179 if (val < 0 || val > 255)
1181 inet->min_ttl = val;
1201 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1205 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1206 * destination in skb->cb[] before dst drop.
1207 * This way, receiver doesn't make cache line misses to read rtable.
1209 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1211 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1212 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1215 if (prepare && skb_rtable(skb)) {
1216 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1217 * which has interface index (iif) as the first member of the
1218 * underlying inet{6}_skb_parm struct. This code then overlays
1219 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1220 * element so the iif is picked up from the prior IPCB. If iif
1221 * is the loopback interface, then return the sending interface
1222 * (e.g., process binds socket to eth0 for Tx which is
1223 * redirected to loopback in the rtable/dst).
1225 struct rtable *rt = skb_rtable(skb);
1226 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1228 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1229 pktinfo->ipi_ifindex = inet_iif(skb);
1230 else if (l3slave && rt && rt->rt_iif)
1231 pktinfo->ipi_ifindex = rt->rt_iif;
1233 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1235 pktinfo->ipi_ifindex = 0;
1236 pktinfo->ipi_spec_dst.s_addr = 0;
1241 int ip_setsockopt(struct sock *sk, int level,
1242 int optname, char __user *optval, unsigned int optlen)
1246 if (level != SOL_IP)
1247 return -ENOPROTOOPT;
1249 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1250 #ifdef CONFIG_BPFILTER
1251 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1252 optname < BPFILTER_IPT_SET_MAX)
1253 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1255 #ifdef CONFIG_NETFILTER
1256 /* we need to exclude all possible ENOPROTOOPTs except default case */
1257 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1258 optname != IP_IPSEC_POLICY &&
1259 optname != IP_XFRM_POLICY &&
1260 !ip_mroute_opt(optname))
1261 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1265 EXPORT_SYMBOL(ip_setsockopt);
1267 #ifdef CONFIG_COMPAT
1268 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1269 char __user *optval, unsigned int optlen)
1273 if (level != SOL_IP)
1274 return -ENOPROTOOPT;
1276 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1277 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1280 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1281 #ifdef CONFIG_NETFILTER
1282 /* we need to exclude all possible ENOPROTOOPTs except default case */
1283 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1284 optname != IP_IPSEC_POLICY &&
1285 optname != IP_XFRM_POLICY &&
1286 !ip_mroute_opt(optname))
1287 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1292 EXPORT_SYMBOL(compat_ip_setsockopt);
1296 * Get the options. Note for future reference. The GET of IP options gets
1297 * the _received_ ones. The set sets the _sent_ ones.
1300 static bool getsockopt_needs_rtnl(int optname)
1304 case MCAST_MSFILTER:
1310 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1311 char __user *optval, int __user *optlen, unsigned int flags)
1313 struct inet_sock *inet = inet_sk(sk);
1314 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1318 if (level != SOL_IP)
1321 if (ip_mroute_opt(optname))
1322 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1324 if (get_user(len, optlen))
1336 unsigned char optbuf[sizeof(struct ip_options)+40];
1337 struct ip_options *opt = (struct ip_options *)optbuf;
1338 struct ip_options_rcu *inet_opt;
1340 inet_opt = rcu_dereference_protected(inet->inet_opt,
1341 lockdep_sock_is_held(sk));
1344 memcpy(optbuf, &inet_opt->opt,
1345 sizeof(struct ip_options) +
1346 inet_opt->opt.optlen);
1349 if (opt->optlen == 0)
1350 return put_user(0, optlen);
1352 ip_options_undo(opt);
1354 len = min_t(unsigned int, len, opt->optlen);
1355 if (put_user(len, optlen))
1357 if (copy_to_user(optval, opt->__data, len))
1362 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1365 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1368 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1371 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1374 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1377 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1379 case IP_RECVORIGDSTADDR:
1380 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1383 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1385 case IP_RECVFRAGSIZE:
1386 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1393 struct net *net = sock_net(sk);
1394 val = (inet->uc_ttl == -1 ?
1395 net->ipv4.sysctl_ip_default_ttl :
1400 val = inet->hdrincl;
1403 val = inet->nodefrag;
1405 case IP_BIND_ADDRESS_NO_PORT:
1406 val = inet->bind_address_no_port;
1408 case IP_MTU_DISCOVER:
1409 val = inet->pmtudisc;
1413 struct dst_entry *dst;
1415 dst = sk_dst_get(sk);
1427 val = inet->recverr;
1429 case IP_MULTICAST_TTL:
1432 case IP_MULTICAST_LOOP:
1433 val = inet->mc_loop;
1436 val = (__force int)htonl((__u32) inet->uc_index);
1438 case IP_MULTICAST_IF:
1440 struct in_addr addr;
1441 len = min_t(unsigned int, len, sizeof(struct in_addr));
1442 addr.s_addr = inet->mc_addr;
1445 if (put_user(len, optlen))
1447 if (copy_to_user(optval, &addr, len))
1453 struct ip_msfilter msf;
1455 if (len < IP_MSFILTER_SIZE(0)) {
1459 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1463 err = ip_mc_msfget(sk, &msf,
1464 (struct ip_msfilter __user *)optval, optlen);
1467 case MCAST_MSFILTER:
1469 struct group_filter gsf;
1471 if (len < GROUP_FILTER_SIZE(0)) {
1475 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1479 err = ip_mc_gsfget(sk, &gsf,
1480 (struct group_filter __user *)optval,
1484 case IP_MULTICAST_ALL:
1493 if (sk->sk_type != SOCK_STREAM)
1494 return -ENOPROTOOPT;
1496 msg.msg_control = (__force void *) optval;
1497 msg.msg_controllen = len;
1498 msg.msg_flags = flags;
1500 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1501 struct in_pktinfo info;
1503 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1504 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1505 info.ipi_ifindex = inet->mc_index;
1506 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1508 if (inet->cmsg_flags & IP_CMSG_TTL) {
1509 int hlim = inet->mc_ttl;
1510 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1512 if (inet->cmsg_flags & IP_CMSG_TOS) {
1513 int tos = inet->rcv_tos;
1514 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1516 len -= msg.msg_controllen;
1517 return put_user(len, optlen);
1520 val = inet->freebind;
1522 case IP_TRANSPARENT:
1523 val = inet->transparent;
1526 val = inet->min_ttl;
1530 return -ENOPROTOOPT;
1534 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1535 unsigned char ucval = (unsigned char)val;
1537 if (put_user(len, optlen))
1539 if (copy_to_user(optval, &ucval, 1))
1542 len = min_t(unsigned int, sizeof(int), len);
1543 if (put_user(len, optlen))
1545 if (copy_to_user(optval, &val, len))
1557 int ip_getsockopt(struct sock *sk, int level,
1558 int optname, char __user *optval, int __user *optlen)
1562 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1563 #ifdef CONFIG_BPFILTER
1564 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1565 optname < BPFILTER_IPT_GET_MAX)
1566 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1568 #ifdef CONFIG_NETFILTER
1569 /* we need to exclude all possible ENOPROTOOPTs except default case */
1570 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1571 !ip_mroute_opt(optname)) {
1574 if (get_user(len, optlen))
1577 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1579 err = put_user(len, optlen);
1585 EXPORT_SYMBOL(ip_getsockopt);
1587 #ifdef CONFIG_COMPAT
1588 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1589 char __user *optval, int __user *optlen)
1593 if (optname == MCAST_MSFILTER)
1594 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1597 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1600 #ifdef CONFIG_BPFILTER
1601 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1602 optname < BPFILTER_IPT_GET_MAX)
1603 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1605 #ifdef CONFIG_NETFILTER
1606 /* we need to exclude all possible ENOPROTOOPTs except default case */
1607 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1608 !ip_mroute_opt(optname)) {
1611 if (get_user(len, optlen))
1614 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1616 err = put_user(len, optlen);
1622 EXPORT_SYMBOL(compat_ip_getsockopt);