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 __be16 _ports[2], *ports;
152 struct sockaddr_in sin;
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.
158 ports = skb_header_pointer(skb, skb_transport_offset(skb),
159 sizeof(_ports), &_ports);
163 sin.sin_family = AF_INET;
164 sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
165 sin.sin_port = ports[1];
166 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
168 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
171 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
172 struct sk_buff *skb, int tlen, int offset)
174 struct inet_sock *inet = inet_sk(sk);
175 unsigned int flags = inet->cmsg_flags;
177 /* Ordered by supposed usage frequency */
178 if (flags & IP_CMSG_PKTINFO) {
179 ip_cmsg_recv_pktinfo(msg, skb);
181 flags &= ~IP_CMSG_PKTINFO;
186 if (flags & IP_CMSG_TTL) {
187 ip_cmsg_recv_ttl(msg, skb);
189 flags &= ~IP_CMSG_TTL;
194 if (flags & IP_CMSG_TOS) {
195 ip_cmsg_recv_tos(msg, skb);
197 flags &= ~IP_CMSG_TOS;
202 if (flags & IP_CMSG_RECVOPTS) {
203 ip_cmsg_recv_opts(msg, skb);
205 flags &= ~IP_CMSG_RECVOPTS;
210 if (flags & IP_CMSG_RETOPTS) {
211 ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
213 flags &= ~IP_CMSG_RETOPTS;
218 if (flags & IP_CMSG_PASSSEC) {
219 ip_cmsg_recv_security(msg, skb);
221 flags &= ~IP_CMSG_PASSSEC;
226 if (flags & IP_CMSG_ORIGDSTADDR) {
227 ip_cmsg_recv_dstaddr(msg, skb);
229 flags &= ~IP_CMSG_ORIGDSTADDR;
234 if (flags & IP_CMSG_CHECKSUM)
235 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
237 if (flags & IP_CMSG_RECVFRAGSIZE)
238 ip_cmsg_recv_fragsize(msg, skb);
240 EXPORT_SYMBOL(ip_cmsg_recv_offset);
242 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
246 struct cmsghdr *cmsg;
247 struct net *net = sock_net(sk);
249 for_each_cmsghdr(cmsg, msg) {
250 if (!CMSG_OK(msg, cmsg))
252 #if IS_ENABLED(CONFIG_IPV6)
254 cmsg->cmsg_level == SOL_IPV6 &&
255 cmsg->cmsg_type == IPV6_PKTINFO) {
256 struct in6_pktinfo *src_info;
258 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
260 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
261 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
263 if (src_info->ipi6_ifindex)
264 ipc->oif = src_info->ipi6_ifindex;
265 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
269 if (cmsg->cmsg_level == SOL_SOCKET) {
270 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
276 if (cmsg->cmsg_level != SOL_IP)
278 switch (cmsg->cmsg_type) {
280 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
282 /* Our caller is responsible for freeing ipc->opt */
283 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
284 err < 40 ? err : 40);
290 struct in_pktinfo *info;
291 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
293 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
294 if (info->ipi_ifindex)
295 ipc->oif = info->ipi_ifindex;
296 ipc->addr = info->ipi_spec_dst.s_addr;
300 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
302 val = *(int *)CMSG_DATA(cmsg);
303 if (val < 1 || val > 255)
308 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
309 val = *(int *)CMSG_DATA(cmsg);
310 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
311 val = *(u8 *)CMSG_DATA(cmsg);
314 if (val < 0 || val > 255)
317 ipc->priority = rt_tos2priority(ipc->tos);
320 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
322 val = *(int *)CMSG_DATA(cmsg);
323 if (val < 1 || val > 255)
334 static void ip_ra_destroy_rcu(struct rcu_head *head)
336 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
338 sock_put(ra->saved_sk);
342 int ip_ra_control(struct sock *sk, unsigned char on,
343 void (*destructor)(struct sock *))
345 struct ip_ra_chain *ra, *new_ra;
346 struct ip_ra_chain __rcu **rap;
347 struct net *net = sock_net(sk);
349 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
352 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
354 mutex_lock(&net->ipv4.ra_mutex);
355 for (rap = &net->ipv4.ra_chain;
356 (ra = rcu_dereference_protected(*rap,
357 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
361 mutex_unlock(&net->ipv4.ra_mutex);
365 /* dont let ip_call_ra_chain() use sk again */
367 RCU_INIT_POINTER(*rap, ra->next);
368 mutex_unlock(&net->ipv4.ra_mutex);
373 * Delay sock_put(sk) and kfree(ra) after one rcu grace
374 * period. This guarantee ip_call_ra_chain() dont need
375 * to mess with socket refcounts.
378 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
383 mutex_unlock(&net->ipv4.ra_mutex);
387 new_ra->destructor = destructor;
389 RCU_INIT_POINTER(new_ra->next, ra);
390 rcu_assign_pointer(*rap, new_ra);
392 mutex_unlock(&net->ipv4.ra_mutex);
397 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
398 __be16 port, u32 info, u8 *payload)
400 struct sock_exterr_skb *serr;
402 skb = skb_clone(skb, GFP_ATOMIC);
406 serr = SKB_EXT_ERR(skb);
407 serr->ee.ee_errno = err;
408 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
409 serr->ee.ee_type = icmp_hdr(skb)->type;
410 serr->ee.ee_code = icmp_hdr(skb)->code;
412 serr->ee.ee_info = info;
413 serr->ee.ee_data = 0;
414 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
415 skb_network_header(skb);
418 if (skb_pull(skb, payload - skb->data)) {
419 skb_reset_transport_header(skb);
420 if (sock_queue_err_skb(sk, skb) == 0)
426 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
428 struct inet_sock *inet = inet_sk(sk);
429 struct sock_exterr_skb *serr;
436 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
440 skb_put(skb, sizeof(struct iphdr));
441 skb_reset_network_header(skb);
445 serr = SKB_EXT_ERR(skb);
446 serr->ee.ee_errno = err;
447 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
448 serr->ee.ee_type = 0;
449 serr->ee.ee_code = 0;
451 serr->ee.ee_info = info;
452 serr->ee.ee_data = 0;
453 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
456 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
457 skb_reset_transport_header(skb);
459 if (sock_queue_err_skb(sk, skb))
463 /* For some errors we have valid addr_offset even with zero payload and
464 * zero port. Also, addr_offset should be supported if port is set.
466 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
468 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
469 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
472 /* IPv4 supports cmsg on all imcp errors and some timestamps
474 * Timestamp code paths do not initialize the fields expected by cmsg:
475 * the PKTINFO fields in skb->cb[]. Fill those in here.
477 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
481 struct in_pktinfo *info;
483 if (ee_origin == SO_EE_ORIGIN_ICMP)
486 if (ee_origin == SO_EE_ORIGIN_LOCAL)
489 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
490 * timestamp with egress dev. Not possible for packets without iif
491 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
493 info = PKTINFO_SKB_CB(skb);
494 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
498 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
503 * Handle MSG_ERRQUEUE
505 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
507 struct sock_exterr_skb *serr;
509 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
511 struct sock_extended_err ee;
512 struct sockaddr_in offender;
518 skb = sock_dequeue_err_skb(sk);
524 msg->msg_flags |= MSG_TRUNC;
527 err = skb_copy_datagram_msg(skb, 0, msg, copied);
532 sock_recv_timestamp(msg, sk, skb);
534 serr = SKB_EXT_ERR(skb);
536 if (sin && ipv4_datagram_support_addr(serr)) {
537 sin->sin_family = AF_INET;
538 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
540 sin->sin_port = serr->port;
541 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
542 *addr_len = sizeof(*sin);
545 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
546 sin = &errhdr.offender;
547 memset(sin, 0, sizeof(*sin));
549 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
550 sin->sin_family = AF_INET;
551 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
552 if (inet_sk(sk)->cmsg_flags)
553 ip_cmsg_recv(msg, skb);
556 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
558 /* Now we could try to dump offended packet options */
560 msg->msg_flags |= MSG_ERRQUEUE;
570 * Socket option code for IP. This is the end of the line after any
571 * TCP,UDP etc options on an IP socket.
573 static bool setsockopt_needs_rtnl(int optname)
576 case IP_ADD_MEMBERSHIP:
577 case IP_ADD_SOURCE_MEMBERSHIP:
578 case IP_BLOCK_SOURCE:
579 case IP_DROP_MEMBERSHIP:
580 case IP_DROP_SOURCE_MEMBERSHIP:
582 case IP_UNBLOCK_SOURCE:
583 case MCAST_BLOCK_SOURCE:
585 case MCAST_JOIN_GROUP:
586 case MCAST_JOIN_SOURCE_GROUP:
587 case MCAST_LEAVE_GROUP:
588 case MCAST_LEAVE_SOURCE_GROUP:
589 case MCAST_UNBLOCK_SOURCE:
595 static int do_ip_setsockopt(struct sock *sk, int level,
596 int optname, char __user *optval, unsigned int optlen)
598 struct inet_sock *inet = inet_sk(sk);
599 struct net *net = sock_net(sk);
601 bool needs_rtnl = setsockopt_needs_rtnl(optname);
612 case IP_MTU_DISCOVER:
614 case IP_ROUTER_ALERT:
620 case IP_BIND_ADDRESS_NO_PORT:
622 case IP_MULTICAST_TTL:
623 case IP_MULTICAST_ALL:
624 case IP_MULTICAST_LOOP:
625 case IP_RECVORIGDSTADDR:
627 case IP_RECVFRAGSIZE:
628 if (optlen >= sizeof(int)) {
629 if (get_user(val, (int __user *) optval))
631 } else if (optlen >= sizeof(char)) {
634 if (get_user(ucval, (unsigned char __user *) optval))
640 /* If optlen==0, it is equivalent to val == 0 */
642 if (optname == IP_ROUTER_ALERT)
643 return ip_ra_control(sk, val ? 1 : 0, NULL);
644 if (ip_mroute_opt(optname))
645 return ip_mroute_setsockopt(sk, optname, optval, optlen);
655 struct ip_options_rcu *old, *opt = NULL;
659 err = ip_options_get_from_user(sock_net(sk), &opt,
663 old = rcu_dereference_protected(inet->inet_opt,
664 lockdep_sock_is_held(sk));
666 struct inet_connection_sock *icsk = inet_csk(sk);
667 #if IS_ENABLED(CONFIG_IPV6)
668 if (sk->sk_family == PF_INET ||
669 (!((1 << sk->sk_state) &
670 (TCPF_LISTEN | TCPF_CLOSE)) &&
671 inet->inet_daddr != LOOPBACK4_IPV6)) {
674 icsk->icsk_ext_hdr_len -= old->opt.optlen;
676 icsk->icsk_ext_hdr_len += opt->opt.optlen;
677 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
678 #if IS_ENABLED(CONFIG_IPV6)
682 rcu_assign_pointer(inet->inet_opt, opt);
689 inet->cmsg_flags |= IP_CMSG_PKTINFO;
691 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
695 inet->cmsg_flags |= IP_CMSG_TTL;
697 inet->cmsg_flags &= ~IP_CMSG_TTL;
701 inet->cmsg_flags |= IP_CMSG_TOS;
703 inet->cmsg_flags &= ~IP_CMSG_TOS;
707 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
709 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
713 inet->cmsg_flags |= IP_CMSG_RETOPTS;
715 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
719 inet->cmsg_flags |= IP_CMSG_PASSSEC;
721 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
723 case IP_RECVORIGDSTADDR:
725 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
727 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
731 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
732 inet_inc_convert_csum(sk);
733 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
736 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
737 inet_dec_convert_csum(sk);
738 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
742 case IP_RECVFRAGSIZE:
743 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
746 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
748 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
750 case IP_TOS: /* This sets both TOS and Precedence */
751 if (sk->sk_type == SOCK_STREAM) {
752 val &= ~INET_ECN_MASK;
753 val |= inet->tos & INET_ECN_MASK;
755 if (inet->tos != val) {
757 sk->sk_priority = rt_tos2priority(val);
764 if (val != -1 && (val < 1 || val > 255))
769 if (sk->sk_type != SOCK_RAW) {
773 inet->hdrincl = val ? 1 : 0;
776 if (sk->sk_type != SOCK_RAW) {
780 inet->nodefrag = val ? 1 : 0;
782 case IP_BIND_ADDRESS_NO_PORT:
783 inet->bind_address_no_port = val ? 1 : 0;
785 case IP_MTU_DISCOVER:
786 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
788 inet->pmtudisc = val;
791 inet->recverr = !!val;
793 skb_queue_purge(&sk->sk_error_queue);
795 case IP_MULTICAST_TTL:
796 if (sk->sk_type == SOCK_STREAM)
802 if (val < 0 || val > 255)
806 case IP_MULTICAST_LOOP:
809 inet->mc_loop = !!val;
813 struct net_device *dev = NULL;
817 if (optlen != sizeof(int))
820 ifindex = (__force int)ntohl((__force __be32)val);
827 dev = dev_get_by_index(sock_net(sk), ifindex);
828 err = -EADDRNOTAVAIL;
832 midx = l3mdev_master_ifindex(dev);
836 if (sk->sk_bound_dev_if &&
837 (!midx || midx != sk->sk_bound_dev_if))
840 inet->uc_index = ifindex;
844 case IP_MULTICAST_IF:
846 struct ip_mreqn mreq;
847 struct net_device *dev = NULL;
850 if (sk->sk_type == SOCK_STREAM)
853 * Check the arguments are allowable
856 if (optlen < sizeof(struct in_addr))
860 if (optlen >= sizeof(struct ip_mreqn)) {
861 if (copy_from_user(&mreq, optval, sizeof(mreq)))
864 memset(&mreq, 0, sizeof(mreq));
865 if (optlen >= sizeof(struct ip_mreq)) {
866 if (copy_from_user(&mreq, optval,
867 sizeof(struct ip_mreq)))
869 } else if (optlen >= sizeof(struct in_addr)) {
870 if (copy_from_user(&mreq.imr_address, optval,
871 sizeof(struct in_addr)))
876 if (!mreq.imr_ifindex) {
877 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
883 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
885 mreq.imr_ifindex = dev->ifindex;
887 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
890 err = -EADDRNOTAVAIL;
894 midx = l3mdev_master_ifindex(dev);
899 if (sk->sk_bound_dev_if &&
900 mreq.imr_ifindex != sk->sk_bound_dev_if &&
901 (!midx || midx != sk->sk_bound_dev_if))
904 inet->mc_index = mreq.imr_ifindex;
905 inet->mc_addr = mreq.imr_address.s_addr;
910 case IP_ADD_MEMBERSHIP:
911 case IP_DROP_MEMBERSHIP:
913 struct ip_mreqn mreq;
916 if (inet_sk(sk)->is_icsk)
919 if (optlen < sizeof(struct ip_mreq))
922 if (optlen >= sizeof(struct ip_mreqn)) {
923 if (copy_from_user(&mreq, optval, sizeof(mreq)))
926 memset(&mreq, 0, sizeof(mreq));
927 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
931 if (optname == IP_ADD_MEMBERSHIP)
932 err = ip_mc_join_group(sk, &mreq);
934 err = ip_mc_leave_group(sk, &mreq);
939 struct ip_msfilter *msf;
941 if (optlen < IP_MSFILTER_SIZE(0))
943 if (optlen > sysctl_optmem_max) {
947 msf = memdup_user(optval, optlen);
952 /* numsrc >= (1G-4) overflow in 32 bits */
953 if (msf->imsf_numsrc >= 0x3ffffffcU ||
954 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
959 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
964 err = ip_mc_msfilter(sk, msf, 0);
968 case IP_BLOCK_SOURCE:
969 case IP_UNBLOCK_SOURCE:
970 case IP_ADD_SOURCE_MEMBERSHIP:
971 case IP_DROP_SOURCE_MEMBERSHIP:
973 struct ip_mreq_source mreqs;
976 if (optlen != sizeof(struct ip_mreq_source))
978 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
982 if (optname == IP_BLOCK_SOURCE) {
983 omode = MCAST_EXCLUDE;
985 } else if (optname == IP_UNBLOCK_SOURCE) {
986 omode = MCAST_EXCLUDE;
988 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
989 struct ip_mreqn mreq;
991 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
992 mreq.imr_address.s_addr = mreqs.imr_interface;
993 mreq.imr_ifindex = 0;
994 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
995 if (err && err != -EADDRINUSE)
997 omode = MCAST_INCLUDE;
999 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1000 omode = MCAST_INCLUDE;
1003 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1006 case MCAST_JOIN_GROUP:
1007 case MCAST_LEAVE_GROUP:
1009 struct group_req greq;
1010 struct sockaddr_in *psin;
1011 struct ip_mreqn mreq;
1013 if (optlen < sizeof(struct group_req))
1016 if (copy_from_user(&greq, optval, sizeof(greq)))
1018 psin = (struct sockaddr_in *)&greq.gr_group;
1019 if (psin->sin_family != AF_INET)
1021 memset(&mreq, 0, sizeof(mreq));
1022 mreq.imr_multiaddr = psin->sin_addr;
1023 mreq.imr_ifindex = greq.gr_interface;
1025 if (optname == MCAST_JOIN_GROUP)
1026 err = ip_mc_join_group(sk, &mreq);
1028 err = ip_mc_leave_group(sk, &mreq);
1031 case MCAST_JOIN_SOURCE_GROUP:
1032 case MCAST_LEAVE_SOURCE_GROUP:
1033 case MCAST_BLOCK_SOURCE:
1034 case MCAST_UNBLOCK_SOURCE:
1036 struct group_source_req greqs;
1037 struct ip_mreq_source mreqs;
1038 struct sockaddr_in *psin;
1041 if (optlen != sizeof(struct group_source_req))
1043 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1047 if (greqs.gsr_group.ss_family != AF_INET ||
1048 greqs.gsr_source.ss_family != AF_INET) {
1049 err = -EADDRNOTAVAIL;
1052 psin = (struct sockaddr_in *)&greqs.gsr_group;
1053 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1054 psin = (struct sockaddr_in *)&greqs.gsr_source;
1055 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1056 mreqs.imr_interface = 0; /* use index for mc_source */
1058 if (optname == MCAST_BLOCK_SOURCE) {
1059 omode = MCAST_EXCLUDE;
1061 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1062 omode = MCAST_EXCLUDE;
1064 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1065 struct ip_mreqn mreq;
1067 psin = (struct sockaddr_in *)&greqs.gsr_group;
1068 mreq.imr_multiaddr = psin->sin_addr;
1069 mreq.imr_address.s_addr = 0;
1070 mreq.imr_ifindex = greqs.gsr_interface;
1071 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1072 if (err && err != -EADDRINUSE)
1074 greqs.gsr_interface = mreq.imr_ifindex;
1075 omode = MCAST_INCLUDE;
1077 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1078 omode = MCAST_INCLUDE;
1081 err = ip_mc_source(add, omode, sk, &mreqs,
1082 greqs.gsr_interface);
1085 case MCAST_MSFILTER:
1087 struct sockaddr_in *psin;
1088 struct ip_msfilter *msf = NULL;
1089 struct group_filter *gsf = NULL;
1090 int msize, i, ifindex;
1092 if (optlen < GROUP_FILTER_SIZE(0))
1094 if (optlen > sysctl_optmem_max) {
1098 gsf = memdup_user(optval, optlen);
1104 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1105 if (gsf->gf_numsrc >= 0x1ffffff ||
1106 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1110 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1114 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1115 msf = kmalloc(msize, GFP_KERNEL);
1120 ifindex = gsf->gf_interface;
1121 psin = (struct sockaddr_in *)&gsf->gf_group;
1122 if (psin->sin_family != AF_INET) {
1123 err = -EADDRNOTAVAIL;
1126 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1127 msf->imsf_interface = 0;
1128 msf->imsf_fmode = gsf->gf_fmode;
1129 msf->imsf_numsrc = gsf->gf_numsrc;
1130 err = -EADDRNOTAVAIL;
1131 for (i = 0; i < gsf->gf_numsrc; ++i) {
1132 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1134 if (psin->sin_family != AF_INET)
1136 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1141 err = ip_mc_msfilter(sk, msf, ifindex);
1147 case IP_MULTICAST_ALL:
1150 if (val != 0 && val != 1)
1158 inet->freebind = !!val;
1161 case IP_IPSEC_POLICY:
1162 case IP_XFRM_POLICY:
1164 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1166 err = xfrm_user_policy(sk, optname, optval, optlen);
1169 case IP_TRANSPARENT:
1170 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1171 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1177 inet->transparent = !!val;
1183 if (val < 0 || val > 255)
1185 inet->min_ttl = val;
1205 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1209 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1210 * destination in skb->cb[] before dst drop.
1211 * This way, receiver doesn't make cache line misses to read rtable.
1213 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1215 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1216 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1219 if (prepare && skb_rtable(skb)) {
1220 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1221 * which has interface index (iif) as the first member of the
1222 * underlying inet{6}_skb_parm struct. This code then overlays
1223 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1224 * element so the iif is picked up from the prior IPCB. If iif
1225 * is the loopback interface, then return the sending interface
1226 * (e.g., process binds socket to eth0 for Tx which is
1227 * redirected to loopback in the rtable/dst).
1229 struct rtable *rt = skb_rtable(skb);
1230 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1232 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1233 pktinfo->ipi_ifindex = inet_iif(skb);
1234 else if (l3slave && rt && rt->rt_iif)
1235 pktinfo->ipi_ifindex = rt->rt_iif;
1237 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1239 pktinfo->ipi_ifindex = 0;
1240 pktinfo->ipi_spec_dst.s_addr = 0;
1245 int ip_setsockopt(struct sock *sk, int level,
1246 int optname, char __user *optval, unsigned int optlen)
1250 if (level != SOL_IP)
1251 return -ENOPROTOOPT;
1253 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1254 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1255 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1256 optname < BPFILTER_IPT_SET_MAX)
1257 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1259 #ifdef CONFIG_NETFILTER
1260 /* we need to exclude all possible ENOPROTOOPTs except default case */
1261 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1262 optname != IP_IPSEC_POLICY &&
1263 optname != IP_XFRM_POLICY &&
1264 !ip_mroute_opt(optname))
1265 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1269 EXPORT_SYMBOL(ip_setsockopt);
1271 #ifdef CONFIG_COMPAT
1272 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1273 char __user *optval, unsigned int optlen)
1277 if (level != SOL_IP)
1278 return -ENOPROTOOPT;
1280 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1281 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1284 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1285 #ifdef CONFIG_NETFILTER
1286 /* we need to exclude all possible ENOPROTOOPTs except default case */
1287 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1288 optname != IP_IPSEC_POLICY &&
1289 optname != IP_XFRM_POLICY &&
1290 !ip_mroute_opt(optname))
1291 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1296 EXPORT_SYMBOL(compat_ip_setsockopt);
1300 * Get the options. Note for future reference. The GET of IP options gets
1301 * the _received_ ones. The set sets the _sent_ ones.
1304 static bool getsockopt_needs_rtnl(int optname)
1308 case MCAST_MSFILTER:
1314 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1315 char __user *optval, int __user *optlen, unsigned int flags)
1317 struct inet_sock *inet = inet_sk(sk);
1318 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1322 if (level != SOL_IP)
1325 if (ip_mroute_opt(optname))
1326 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1328 if (get_user(len, optlen))
1340 unsigned char optbuf[sizeof(struct ip_options)+40];
1341 struct ip_options *opt = (struct ip_options *)optbuf;
1342 struct ip_options_rcu *inet_opt;
1344 inet_opt = rcu_dereference_protected(inet->inet_opt,
1345 lockdep_sock_is_held(sk));
1348 memcpy(optbuf, &inet_opt->opt,
1349 sizeof(struct ip_options) +
1350 inet_opt->opt.optlen);
1353 if (opt->optlen == 0)
1354 return put_user(0, optlen);
1356 ip_options_undo(opt);
1358 len = min_t(unsigned int, len, opt->optlen);
1359 if (put_user(len, optlen))
1361 if (copy_to_user(optval, opt->__data, len))
1366 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1369 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1372 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1375 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1378 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1381 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1383 case IP_RECVORIGDSTADDR:
1384 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1387 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1389 case IP_RECVFRAGSIZE:
1390 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1397 struct net *net = sock_net(sk);
1398 val = (inet->uc_ttl == -1 ?
1399 net->ipv4.sysctl_ip_default_ttl :
1404 val = inet->hdrincl;
1407 val = inet->nodefrag;
1409 case IP_BIND_ADDRESS_NO_PORT:
1410 val = inet->bind_address_no_port;
1412 case IP_MTU_DISCOVER:
1413 val = inet->pmtudisc;
1417 struct dst_entry *dst;
1419 dst = sk_dst_get(sk);
1431 val = inet->recverr;
1433 case IP_MULTICAST_TTL:
1436 case IP_MULTICAST_LOOP:
1437 val = inet->mc_loop;
1440 val = (__force int)htonl((__u32) inet->uc_index);
1442 case IP_MULTICAST_IF:
1444 struct in_addr addr;
1445 len = min_t(unsigned int, len, sizeof(struct in_addr));
1446 addr.s_addr = inet->mc_addr;
1449 if (put_user(len, optlen))
1451 if (copy_to_user(optval, &addr, len))
1457 struct ip_msfilter msf;
1459 if (len < IP_MSFILTER_SIZE(0)) {
1463 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1467 err = ip_mc_msfget(sk, &msf,
1468 (struct ip_msfilter __user *)optval, optlen);
1471 case MCAST_MSFILTER:
1473 struct group_filter gsf;
1475 if (len < GROUP_FILTER_SIZE(0)) {
1479 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1483 err = ip_mc_gsfget(sk, &gsf,
1484 (struct group_filter __user *)optval,
1488 case IP_MULTICAST_ALL:
1497 if (sk->sk_type != SOCK_STREAM)
1498 return -ENOPROTOOPT;
1500 msg.msg_control = (__force void *) optval;
1501 msg.msg_controllen = len;
1502 msg.msg_flags = flags;
1504 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1505 struct in_pktinfo info;
1507 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1508 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1509 info.ipi_ifindex = inet->mc_index;
1510 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1512 if (inet->cmsg_flags & IP_CMSG_TTL) {
1513 int hlim = inet->mc_ttl;
1514 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1516 if (inet->cmsg_flags & IP_CMSG_TOS) {
1517 int tos = inet->rcv_tos;
1518 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1520 len -= msg.msg_controllen;
1521 return put_user(len, optlen);
1524 val = inet->freebind;
1526 case IP_TRANSPARENT:
1527 val = inet->transparent;
1530 val = inet->min_ttl;
1533 val = inet_sk(sk)->inet_num;
1537 return -ENOPROTOOPT;
1541 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1542 unsigned char ucval = (unsigned char)val;
1544 if (put_user(len, optlen))
1546 if (copy_to_user(optval, &ucval, 1))
1549 len = min_t(unsigned int, sizeof(int), len);
1550 if (put_user(len, optlen))
1552 if (copy_to_user(optval, &val, len))
1564 int ip_getsockopt(struct sock *sk, int level,
1565 int optname, char __user *optval, int __user *optlen)
1569 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1570 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1571 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1572 optname < BPFILTER_IPT_GET_MAX)
1573 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1575 #ifdef CONFIG_NETFILTER
1576 /* we need to exclude all possible ENOPROTOOPTs except default case */
1577 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1578 !ip_mroute_opt(optname)) {
1581 if (get_user(len, optlen))
1584 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1586 err = put_user(len, optlen);
1592 EXPORT_SYMBOL(ip_getsockopt);
1594 #ifdef CONFIG_COMPAT
1595 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1596 char __user *optval, int __user *optlen)
1600 if (optname == MCAST_MSFILTER)
1601 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1604 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1607 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1608 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1609 optname < BPFILTER_IPT_GET_MAX)
1610 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1612 #ifdef CONFIG_NETFILTER
1613 /* we need to exclude all possible ENOPROTOOPTs except default case */
1614 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1615 !ip_mroute_opt(optname)) {
1618 if (get_user(len, optlen))
1621 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1623 err = put_user(len, optlen);
1629 EXPORT_SYMBOL(compat_ip_getsockopt);