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);
327 static void ip_ra_destroy_rcu(struct rcu_head *head)
329 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
331 sock_put(ra->saved_sk);
335 int ip_ra_control(struct sock *sk, unsigned char on,
336 void (*destructor)(struct sock *))
338 struct ip_ra_chain *ra, *new_ra;
339 struct ip_ra_chain __rcu **rap;
340 struct net *net = sock_net(sk);
342 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
345 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
347 mutex_lock(&net->ipv4.ra_mutex);
348 for (rap = &net->ipv4.ra_chain;
349 (ra = rcu_dereference_protected(*rap,
350 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
354 mutex_unlock(&net->ipv4.ra_mutex);
358 /* dont let ip_call_ra_chain() use sk again */
360 RCU_INIT_POINTER(*rap, ra->next);
361 mutex_unlock(&net->ipv4.ra_mutex);
366 * Delay sock_put(sk) and kfree(ra) after one rcu grace
367 * period. This guarantee ip_call_ra_chain() dont need
368 * to mess with socket refcounts.
371 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
376 mutex_unlock(&net->ipv4.ra_mutex);
380 new_ra->destructor = destructor;
382 RCU_INIT_POINTER(new_ra->next, ra);
383 rcu_assign_pointer(*rap, new_ra);
385 mutex_unlock(&net->ipv4.ra_mutex);
390 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
391 __be16 port, u32 info, u8 *payload)
393 struct sock_exterr_skb *serr;
395 skb = skb_clone(skb, GFP_ATOMIC);
399 serr = SKB_EXT_ERR(skb);
400 serr->ee.ee_errno = err;
401 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
402 serr->ee.ee_type = icmp_hdr(skb)->type;
403 serr->ee.ee_code = icmp_hdr(skb)->code;
405 serr->ee.ee_info = info;
406 serr->ee.ee_data = 0;
407 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
408 skb_network_header(skb);
411 if (skb_pull(skb, payload - skb->data)) {
412 skb_reset_transport_header(skb);
413 if (sock_queue_err_skb(sk, skb) == 0)
419 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
421 struct inet_sock *inet = inet_sk(sk);
422 struct sock_exterr_skb *serr;
429 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
433 skb_put(skb, sizeof(struct iphdr));
434 skb_reset_network_header(skb);
438 serr = SKB_EXT_ERR(skb);
439 serr->ee.ee_errno = err;
440 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
441 serr->ee.ee_type = 0;
442 serr->ee.ee_code = 0;
444 serr->ee.ee_info = info;
445 serr->ee.ee_data = 0;
446 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
449 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
450 skb_reset_transport_header(skb);
452 if (sock_queue_err_skb(sk, skb))
456 /* For some errors we have valid addr_offset even with zero payload and
457 * zero port. Also, addr_offset should be supported if port is set.
459 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
461 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
462 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
465 /* IPv4 supports cmsg on all imcp errors and some timestamps
467 * Timestamp code paths do not initialize the fields expected by cmsg:
468 * the PKTINFO fields in skb->cb[]. Fill those in here.
470 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
474 struct in_pktinfo *info;
476 if (ee_origin == SO_EE_ORIGIN_ICMP)
479 if (ee_origin == SO_EE_ORIGIN_LOCAL)
482 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
483 * timestamp with egress dev. Not possible for packets without iif
484 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
486 info = PKTINFO_SKB_CB(skb);
487 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
491 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
496 * Handle MSG_ERRQUEUE
498 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
500 struct sock_exterr_skb *serr;
502 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
504 struct sock_extended_err ee;
505 struct sockaddr_in offender;
511 skb = sock_dequeue_err_skb(sk);
517 msg->msg_flags |= MSG_TRUNC;
520 err = skb_copy_datagram_msg(skb, 0, msg, copied);
525 sock_recv_timestamp(msg, sk, skb);
527 serr = SKB_EXT_ERR(skb);
529 if (sin && ipv4_datagram_support_addr(serr)) {
530 sin->sin_family = AF_INET;
531 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
533 sin->sin_port = serr->port;
534 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
535 *addr_len = sizeof(*sin);
538 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
539 sin = &errhdr.offender;
540 memset(sin, 0, sizeof(*sin));
542 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
543 sin->sin_family = AF_INET;
544 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
545 if (inet_sk(sk)->cmsg_flags)
546 ip_cmsg_recv(msg, skb);
549 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
551 /* Now we could try to dump offended packet options */
553 msg->msg_flags |= MSG_ERRQUEUE;
563 * Socket option code for IP. This is the end of the line after any
564 * TCP,UDP etc options on an IP socket.
566 static bool setsockopt_needs_rtnl(int optname)
569 case IP_ADD_MEMBERSHIP:
570 case IP_ADD_SOURCE_MEMBERSHIP:
571 case IP_BLOCK_SOURCE:
572 case IP_DROP_MEMBERSHIP:
573 case IP_DROP_SOURCE_MEMBERSHIP:
575 case IP_UNBLOCK_SOURCE:
576 case MCAST_BLOCK_SOURCE:
578 case MCAST_JOIN_GROUP:
579 case MCAST_JOIN_SOURCE_GROUP:
580 case MCAST_LEAVE_GROUP:
581 case MCAST_LEAVE_SOURCE_GROUP:
582 case MCAST_UNBLOCK_SOURCE:
588 static int do_ip_setsockopt(struct sock *sk, int level,
589 int optname, char __user *optval, unsigned int optlen)
591 struct inet_sock *inet = inet_sk(sk);
592 struct net *net = sock_net(sk);
594 bool needs_rtnl = setsockopt_needs_rtnl(optname);
605 case IP_MTU_DISCOVER:
607 case IP_ROUTER_ALERT:
613 case IP_BIND_ADDRESS_NO_PORT:
615 case IP_MULTICAST_TTL:
616 case IP_MULTICAST_ALL:
617 case IP_MULTICAST_LOOP:
618 case IP_RECVORIGDSTADDR:
620 case IP_RECVFRAGSIZE:
621 if (optlen >= sizeof(int)) {
622 if (get_user(val, (int __user *) optval))
624 } else if (optlen >= sizeof(char)) {
627 if (get_user(ucval, (unsigned char __user *) optval))
633 /* If optlen==0, it is equivalent to val == 0 */
635 if (optname == IP_ROUTER_ALERT)
636 return ip_ra_control(sk, val ? 1 : 0, NULL);
637 if (ip_mroute_opt(optname))
638 return ip_mroute_setsockopt(sk, optname, optval, optlen);
648 struct ip_options_rcu *old, *opt = NULL;
652 err = ip_options_get_from_user(sock_net(sk), &opt,
656 old = rcu_dereference_protected(inet->inet_opt,
657 lockdep_sock_is_held(sk));
659 struct inet_connection_sock *icsk = inet_csk(sk);
660 #if IS_ENABLED(CONFIG_IPV6)
661 if (sk->sk_family == PF_INET ||
662 (!((1 << sk->sk_state) &
663 (TCPF_LISTEN | TCPF_CLOSE)) &&
664 inet->inet_daddr != LOOPBACK4_IPV6)) {
667 icsk->icsk_ext_hdr_len -= old->opt.optlen;
669 icsk->icsk_ext_hdr_len += opt->opt.optlen;
670 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
671 #if IS_ENABLED(CONFIG_IPV6)
675 rcu_assign_pointer(inet->inet_opt, opt);
682 inet->cmsg_flags |= IP_CMSG_PKTINFO;
684 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
688 inet->cmsg_flags |= IP_CMSG_TTL;
690 inet->cmsg_flags &= ~IP_CMSG_TTL;
694 inet->cmsg_flags |= IP_CMSG_TOS;
696 inet->cmsg_flags &= ~IP_CMSG_TOS;
700 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
702 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
706 inet->cmsg_flags |= IP_CMSG_RETOPTS;
708 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
712 inet->cmsg_flags |= IP_CMSG_PASSSEC;
714 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
716 case IP_RECVORIGDSTADDR:
718 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
720 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
724 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
725 inet_inc_convert_csum(sk);
726 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
729 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
730 inet_dec_convert_csum(sk);
731 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
735 case IP_RECVFRAGSIZE:
736 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
739 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
741 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
743 case IP_TOS: /* This sets both TOS and Precedence */
744 if (sk->sk_type == SOCK_STREAM) {
745 val &= ~INET_ECN_MASK;
746 val |= inet->tos & INET_ECN_MASK;
748 if (inet->tos != val) {
750 sk->sk_priority = rt_tos2priority(val);
757 if (val != -1 && (val < 1 || val > 255))
762 if (sk->sk_type != SOCK_RAW) {
766 inet->hdrincl = val ? 1 : 0;
769 if (sk->sk_type != SOCK_RAW) {
773 inet->nodefrag = val ? 1 : 0;
775 case IP_BIND_ADDRESS_NO_PORT:
776 inet->bind_address_no_port = val ? 1 : 0;
778 case IP_MTU_DISCOVER:
779 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
781 inet->pmtudisc = val;
784 inet->recverr = !!val;
786 skb_queue_purge(&sk->sk_error_queue);
788 case IP_MULTICAST_TTL:
789 if (sk->sk_type == SOCK_STREAM)
795 if (val < 0 || val > 255)
799 case IP_MULTICAST_LOOP:
802 inet->mc_loop = !!val;
806 struct net_device *dev = NULL;
810 if (optlen != sizeof(int))
813 ifindex = (__force int)ntohl((__force __be32)val);
820 dev = dev_get_by_index(sock_net(sk), ifindex);
821 err = -EADDRNOTAVAIL;
825 midx = l3mdev_master_ifindex(dev);
829 if (sk->sk_bound_dev_if &&
830 (!midx || midx != sk->sk_bound_dev_if))
833 inet->uc_index = ifindex;
837 case IP_MULTICAST_IF:
839 struct ip_mreqn mreq;
840 struct net_device *dev = NULL;
843 if (sk->sk_type == SOCK_STREAM)
846 * Check the arguments are allowable
849 if (optlen < sizeof(struct in_addr))
853 if (optlen >= sizeof(struct ip_mreqn)) {
854 if (copy_from_user(&mreq, optval, sizeof(mreq)))
857 memset(&mreq, 0, sizeof(mreq));
858 if (optlen >= sizeof(struct ip_mreq)) {
859 if (copy_from_user(&mreq, optval,
860 sizeof(struct ip_mreq)))
862 } else if (optlen >= sizeof(struct in_addr)) {
863 if (copy_from_user(&mreq.imr_address, optval,
864 sizeof(struct in_addr)))
869 if (!mreq.imr_ifindex) {
870 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
876 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
878 mreq.imr_ifindex = dev->ifindex;
880 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
883 err = -EADDRNOTAVAIL;
887 midx = l3mdev_master_ifindex(dev);
892 if (sk->sk_bound_dev_if &&
893 mreq.imr_ifindex != sk->sk_bound_dev_if &&
894 (!midx || midx != sk->sk_bound_dev_if))
897 inet->mc_index = mreq.imr_ifindex;
898 inet->mc_addr = mreq.imr_address.s_addr;
903 case IP_ADD_MEMBERSHIP:
904 case IP_DROP_MEMBERSHIP:
906 struct ip_mreqn mreq;
909 if (inet_sk(sk)->is_icsk)
912 if (optlen < sizeof(struct ip_mreq))
915 if (optlen >= sizeof(struct ip_mreqn)) {
916 if (copy_from_user(&mreq, optval, sizeof(mreq)))
919 memset(&mreq, 0, sizeof(mreq));
920 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
924 if (optname == IP_ADD_MEMBERSHIP)
925 err = ip_mc_join_group(sk, &mreq);
927 err = ip_mc_leave_group(sk, &mreq);
932 struct ip_msfilter *msf;
934 if (optlen < IP_MSFILTER_SIZE(0))
936 if (optlen > sysctl_optmem_max) {
940 msf = memdup_user(optval, optlen);
945 /* numsrc >= (1G-4) overflow in 32 bits */
946 if (msf->imsf_numsrc >= 0x3ffffffcU ||
947 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
952 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
957 err = ip_mc_msfilter(sk, msf, 0);
961 case IP_BLOCK_SOURCE:
962 case IP_UNBLOCK_SOURCE:
963 case IP_ADD_SOURCE_MEMBERSHIP:
964 case IP_DROP_SOURCE_MEMBERSHIP:
966 struct ip_mreq_source mreqs;
969 if (optlen != sizeof(struct ip_mreq_source))
971 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
975 if (optname == IP_BLOCK_SOURCE) {
976 omode = MCAST_EXCLUDE;
978 } else if (optname == IP_UNBLOCK_SOURCE) {
979 omode = MCAST_EXCLUDE;
981 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
982 struct ip_mreqn mreq;
984 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
985 mreq.imr_address.s_addr = mreqs.imr_interface;
986 mreq.imr_ifindex = 0;
987 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
988 if (err && err != -EADDRINUSE)
990 omode = MCAST_INCLUDE;
992 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
993 omode = MCAST_INCLUDE;
996 err = ip_mc_source(add, omode, sk, &mreqs, 0);
999 case MCAST_JOIN_GROUP:
1000 case MCAST_LEAVE_GROUP:
1002 struct group_req greq;
1003 struct sockaddr_in *psin;
1004 struct ip_mreqn mreq;
1006 if (optlen < sizeof(struct group_req))
1009 if (copy_from_user(&greq, optval, sizeof(greq)))
1011 psin = (struct sockaddr_in *)&greq.gr_group;
1012 if (psin->sin_family != AF_INET)
1014 memset(&mreq, 0, sizeof(mreq));
1015 mreq.imr_multiaddr = psin->sin_addr;
1016 mreq.imr_ifindex = greq.gr_interface;
1018 if (optname == MCAST_JOIN_GROUP)
1019 err = ip_mc_join_group(sk, &mreq);
1021 err = ip_mc_leave_group(sk, &mreq);
1024 case MCAST_JOIN_SOURCE_GROUP:
1025 case MCAST_LEAVE_SOURCE_GROUP:
1026 case MCAST_BLOCK_SOURCE:
1027 case MCAST_UNBLOCK_SOURCE:
1029 struct group_source_req greqs;
1030 struct ip_mreq_source mreqs;
1031 struct sockaddr_in *psin;
1034 if (optlen != sizeof(struct group_source_req))
1036 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1040 if (greqs.gsr_group.ss_family != AF_INET ||
1041 greqs.gsr_source.ss_family != AF_INET) {
1042 err = -EADDRNOTAVAIL;
1045 psin = (struct sockaddr_in *)&greqs.gsr_group;
1046 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1047 psin = (struct sockaddr_in *)&greqs.gsr_source;
1048 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1049 mreqs.imr_interface = 0; /* use index for mc_source */
1051 if (optname == MCAST_BLOCK_SOURCE) {
1052 omode = MCAST_EXCLUDE;
1054 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1055 omode = MCAST_EXCLUDE;
1057 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1058 struct ip_mreqn mreq;
1060 psin = (struct sockaddr_in *)&greqs.gsr_group;
1061 mreq.imr_multiaddr = psin->sin_addr;
1062 mreq.imr_address.s_addr = 0;
1063 mreq.imr_ifindex = greqs.gsr_interface;
1064 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1065 if (err && err != -EADDRINUSE)
1067 greqs.gsr_interface = mreq.imr_ifindex;
1068 omode = MCAST_INCLUDE;
1070 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1071 omode = MCAST_INCLUDE;
1074 err = ip_mc_source(add, omode, sk, &mreqs,
1075 greqs.gsr_interface);
1078 case MCAST_MSFILTER:
1080 struct sockaddr_in *psin;
1081 struct ip_msfilter *msf = NULL;
1082 struct group_filter *gsf = NULL;
1083 int msize, i, ifindex;
1085 if (optlen < GROUP_FILTER_SIZE(0))
1087 if (optlen > sysctl_optmem_max) {
1091 gsf = memdup_user(optval, optlen);
1097 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1098 if (gsf->gf_numsrc >= 0x1ffffff ||
1099 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1103 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1107 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1108 msf = kmalloc(msize, GFP_KERNEL);
1113 ifindex = gsf->gf_interface;
1114 psin = (struct sockaddr_in *)&gsf->gf_group;
1115 if (psin->sin_family != AF_INET) {
1116 err = -EADDRNOTAVAIL;
1119 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1120 msf->imsf_interface = 0;
1121 msf->imsf_fmode = gsf->gf_fmode;
1122 msf->imsf_numsrc = gsf->gf_numsrc;
1123 err = -EADDRNOTAVAIL;
1124 for (i = 0; i < gsf->gf_numsrc; ++i) {
1125 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1127 if (psin->sin_family != AF_INET)
1129 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1134 err = ip_mc_msfilter(sk, msf, ifindex);
1140 case IP_MULTICAST_ALL:
1143 if (val != 0 && val != 1)
1151 inet->freebind = !!val;
1154 case IP_IPSEC_POLICY:
1155 case IP_XFRM_POLICY:
1157 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1159 err = xfrm_user_policy(sk, optname, optval, optlen);
1162 case IP_TRANSPARENT:
1163 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1164 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1170 inet->transparent = !!val;
1176 if (val < 0 || val > 255)
1178 inet->min_ttl = val;
1198 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1202 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1203 * destination in skb->cb[] before dst drop.
1204 * This way, receiver doesn't make cache line misses to read rtable.
1206 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1208 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1209 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1212 if (prepare && skb_rtable(skb)) {
1213 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1214 * which has interface index (iif) as the first member of the
1215 * underlying inet{6}_skb_parm struct. This code then overlays
1216 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1217 * element so the iif is picked up from the prior IPCB. If iif
1218 * is the loopback interface, then return the sending interface
1219 * (e.g., process binds socket to eth0 for Tx which is
1220 * redirected to loopback in the rtable/dst).
1222 struct rtable *rt = skb_rtable(skb);
1223 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1225 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1226 pktinfo->ipi_ifindex = inet_iif(skb);
1227 else if (l3slave && rt && rt->rt_iif)
1228 pktinfo->ipi_ifindex = rt->rt_iif;
1230 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1232 pktinfo->ipi_ifindex = 0;
1233 pktinfo->ipi_spec_dst.s_addr = 0;
1238 int ip_setsockopt(struct sock *sk, int level,
1239 int optname, char __user *optval, unsigned int optlen)
1243 if (level != SOL_IP)
1244 return -ENOPROTOOPT;
1246 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1247 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1248 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1249 optname < BPFILTER_IPT_SET_MAX)
1250 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1252 #ifdef CONFIG_NETFILTER
1253 /* we need to exclude all possible ENOPROTOOPTs except default case */
1254 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1255 optname != IP_IPSEC_POLICY &&
1256 optname != IP_XFRM_POLICY &&
1257 !ip_mroute_opt(optname))
1258 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1262 EXPORT_SYMBOL(ip_setsockopt);
1264 #ifdef CONFIG_COMPAT
1265 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1266 char __user *optval, unsigned int optlen)
1270 if (level != SOL_IP)
1271 return -ENOPROTOOPT;
1273 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1274 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1277 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1278 #ifdef CONFIG_NETFILTER
1279 /* we need to exclude all possible ENOPROTOOPTs except default case */
1280 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1281 optname != IP_IPSEC_POLICY &&
1282 optname != IP_XFRM_POLICY &&
1283 !ip_mroute_opt(optname))
1284 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1289 EXPORT_SYMBOL(compat_ip_setsockopt);
1293 * Get the options. Note for future reference. The GET of IP options gets
1294 * the _received_ ones. The set sets the _sent_ ones.
1297 static bool getsockopt_needs_rtnl(int optname)
1301 case MCAST_MSFILTER:
1307 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1308 char __user *optval, int __user *optlen, unsigned int flags)
1310 struct inet_sock *inet = inet_sk(sk);
1311 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1315 if (level != SOL_IP)
1318 if (ip_mroute_opt(optname))
1319 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1321 if (get_user(len, optlen))
1333 unsigned char optbuf[sizeof(struct ip_options)+40];
1334 struct ip_options *opt = (struct ip_options *)optbuf;
1335 struct ip_options_rcu *inet_opt;
1337 inet_opt = rcu_dereference_protected(inet->inet_opt,
1338 lockdep_sock_is_held(sk));
1341 memcpy(optbuf, &inet_opt->opt,
1342 sizeof(struct ip_options) +
1343 inet_opt->opt.optlen);
1346 if (opt->optlen == 0)
1347 return put_user(0, optlen);
1349 ip_options_undo(opt);
1351 len = min_t(unsigned int, len, opt->optlen);
1352 if (put_user(len, optlen))
1354 if (copy_to_user(optval, opt->__data, len))
1359 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1362 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1365 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1368 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1371 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1374 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1376 case IP_RECVORIGDSTADDR:
1377 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1380 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1382 case IP_RECVFRAGSIZE:
1383 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1390 struct net *net = sock_net(sk);
1391 val = (inet->uc_ttl == -1 ?
1392 net->ipv4.sysctl_ip_default_ttl :
1397 val = inet->hdrincl;
1400 val = inet->nodefrag;
1402 case IP_BIND_ADDRESS_NO_PORT:
1403 val = inet->bind_address_no_port;
1405 case IP_MTU_DISCOVER:
1406 val = inet->pmtudisc;
1410 struct dst_entry *dst;
1412 dst = sk_dst_get(sk);
1424 val = inet->recverr;
1426 case IP_MULTICAST_TTL:
1429 case IP_MULTICAST_LOOP:
1430 val = inet->mc_loop;
1433 val = (__force int)htonl((__u32) inet->uc_index);
1435 case IP_MULTICAST_IF:
1437 struct in_addr addr;
1438 len = min_t(unsigned int, len, sizeof(struct in_addr));
1439 addr.s_addr = inet->mc_addr;
1442 if (put_user(len, optlen))
1444 if (copy_to_user(optval, &addr, len))
1450 struct ip_msfilter msf;
1452 if (len < IP_MSFILTER_SIZE(0)) {
1456 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1460 err = ip_mc_msfget(sk, &msf,
1461 (struct ip_msfilter __user *)optval, optlen);
1464 case MCAST_MSFILTER:
1466 struct group_filter gsf;
1468 if (len < GROUP_FILTER_SIZE(0)) {
1472 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1476 err = ip_mc_gsfget(sk, &gsf,
1477 (struct group_filter __user *)optval,
1481 case IP_MULTICAST_ALL:
1490 if (sk->sk_type != SOCK_STREAM)
1491 return -ENOPROTOOPT;
1493 msg.msg_control = (__force void *) optval;
1494 msg.msg_controllen = len;
1495 msg.msg_flags = flags;
1497 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1498 struct in_pktinfo info;
1500 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1501 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1502 info.ipi_ifindex = inet->mc_index;
1503 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1505 if (inet->cmsg_flags & IP_CMSG_TTL) {
1506 int hlim = inet->mc_ttl;
1507 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1509 if (inet->cmsg_flags & IP_CMSG_TOS) {
1510 int tos = inet->rcv_tos;
1511 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1513 len -= msg.msg_controllen;
1514 return put_user(len, optlen);
1517 val = inet->freebind;
1519 case IP_TRANSPARENT:
1520 val = inet->transparent;
1523 val = inet->min_ttl;
1527 return -ENOPROTOOPT;
1531 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1532 unsigned char ucval = (unsigned char)val;
1534 if (put_user(len, optlen))
1536 if (copy_to_user(optval, &ucval, 1))
1539 len = min_t(unsigned int, sizeof(int), len);
1540 if (put_user(len, optlen))
1542 if (copy_to_user(optval, &val, len))
1554 int ip_getsockopt(struct sock *sk, int level,
1555 int optname, char __user *optval, int __user *optlen)
1559 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1560 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1561 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1562 optname < BPFILTER_IPT_GET_MAX)
1563 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1565 #ifdef CONFIG_NETFILTER
1566 /* we need to exclude all possible ENOPROTOOPTs except default case */
1567 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1568 !ip_mroute_opt(optname)) {
1571 if (get_user(len, optlen))
1574 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1576 err = put_user(len, optlen);
1582 EXPORT_SYMBOL(ip_getsockopt);
1584 #ifdef CONFIG_COMPAT
1585 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1586 char __user *optval, int __user *optlen)
1590 if (optname == MCAST_MSFILTER)
1591 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1594 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1597 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1598 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1599 optname < BPFILTER_IPT_GET_MAX)
1600 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1602 #ifdef CONFIG_NETFILTER
1603 /* we need to exclude all possible ENOPROTOOPTs except default case */
1604 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1605 !ip_mroute_opt(optname)) {
1608 if (get_user(len, optlen))
1611 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1613 err = put_user(len, optlen);
1619 EXPORT_SYMBOL(compat_ip_getsockopt);