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;
349 mutex_lock(&net->ipv4.ra_mutex);
350 for (rap = &net->ipv4.ra_chain;
351 (ra = rcu_dereference_protected(*rap,
352 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
356 mutex_unlock(&net->ipv4.ra_mutex);
360 /* dont let ip_call_ra_chain() use sk again */
362 RCU_INIT_POINTER(*rap, ra->next);
363 mutex_unlock(&net->ipv4.ra_mutex);
368 * Delay sock_put(sk) and kfree(ra) after one rcu grace
369 * period. This guarantee ip_call_ra_chain() dont need
370 * to mess with socket refcounts.
373 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
378 mutex_unlock(&net->ipv4.ra_mutex);
382 new_ra->destructor = destructor;
384 RCU_INIT_POINTER(new_ra->next, ra);
385 rcu_assign_pointer(*rap, new_ra);
387 mutex_unlock(&net->ipv4.ra_mutex);
392 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
393 __be16 port, u32 info, u8 *payload)
395 struct sock_exterr_skb *serr;
397 skb = skb_clone(skb, GFP_ATOMIC);
401 serr = SKB_EXT_ERR(skb);
402 serr->ee.ee_errno = err;
403 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
404 serr->ee.ee_type = icmp_hdr(skb)->type;
405 serr->ee.ee_code = icmp_hdr(skb)->code;
407 serr->ee.ee_info = info;
408 serr->ee.ee_data = 0;
409 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
410 skb_network_header(skb);
413 if (skb_pull(skb, payload - skb->data)) {
414 skb_reset_transport_header(skb);
415 if (sock_queue_err_skb(sk, skb) == 0)
421 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
423 struct inet_sock *inet = inet_sk(sk);
424 struct sock_exterr_skb *serr;
431 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
435 skb_put(skb, sizeof(struct iphdr));
436 skb_reset_network_header(skb);
440 serr = SKB_EXT_ERR(skb);
441 serr->ee.ee_errno = err;
442 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
443 serr->ee.ee_type = 0;
444 serr->ee.ee_code = 0;
446 serr->ee.ee_info = info;
447 serr->ee.ee_data = 0;
448 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
451 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
452 skb_reset_transport_header(skb);
454 if (sock_queue_err_skb(sk, skb))
458 /* For some errors we have valid addr_offset even with zero payload and
459 * zero port. Also, addr_offset should be supported if port is set.
461 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
463 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
464 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
467 /* IPv4 supports cmsg on all imcp errors and some timestamps
469 * Timestamp code paths do not initialize the fields expected by cmsg:
470 * the PKTINFO fields in skb->cb[]. Fill those in here.
472 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
476 struct in_pktinfo *info;
478 if (ee_origin == SO_EE_ORIGIN_ICMP)
481 if (ee_origin == SO_EE_ORIGIN_LOCAL)
484 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
485 * timestamp with egress dev. Not possible for packets without iif
486 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
488 info = PKTINFO_SKB_CB(skb);
489 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
493 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
498 * Handle MSG_ERRQUEUE
500 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
502 struct sock_exterr_skb *serr;
504 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
506 struct sock_extended_err ee;
507 struct sockaddr_in offender;
513 skb = sock_dequeue_err_skb(sk);
519 msg->msg_flags |= MSG_TRUNC;
522 err = skb_copy_datagram_msg(skb, 0, msg, copied);
527 sock_recv_timestamp(msg, sk, skb);
529 serr = SKB_EXT_ERR(skb);
531 if (sin && ipv4_datagram_support_addr(serr)) {
532 sin->sin_family = AF_INET;
533 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
535 sin->sin_port = serr->port;
536 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
537 *addr_len = sizeof(*sin);
540 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
541 sin = &errhdr.offender;
542 memset(sin, 0, sizeof(*sin));
544 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
545 sin->sin_family = AF_INET;
546 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
547 if (inet_sk(sk)->cmsg_flags)
548 ip_cmsg_recv(msg, skb);
551 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
553 /* Now we could try to dump offended packet options */
555 msg->msg_flags |= MSG_ERRQUEUE;
565 * Socket option code for IP. This is the end of the line after any
566 * TCP,UDP etc options on an IP socket.
568 static bool setsockopt_needs_rtnl(int optname)
571 case IP_ADD_MEMBERSHIP:
572 case IP_ADD_SOURCE_MEMBERSHIP:
573 case IP_BLOCK_SOURCE:
574 case IP_DROP_MEMBERSHIP:
575 case IP_DROP_SOURCE_MEMBERSHIP:
577 case IP_UNBLOCK_SOURCE:
578 case MCAST_BLOCK_SOURCE:
580 case MCAST_JOIN_GROUP:
581 case MCAST_JOIN_SOURCE_GROUP:
582 case MCAST_LEAVE_GROUP:
583 case MCAST_LEAVE_SOURCE_GROUP:
584 case MCAST_UNBLOCK_SOURCE:
590 static int set_mcast_msfilter(struct sock *sk, int ifindex,
591 int numsrc, int fmode,
592 struct sockaddr_storage *group,
593 struct sockaddr_storage *list)
595 int msize = IP_MSFILTER_SIZE(numsrc);
596 struct ip_msfilter *msf;
597 struct sockaddr_in *psin;
600 msf = kmalloc(msize, GFP_KERNEL);
604 psin = (struct sockaddr_in *)group;
605 if (psin->sin_family != AF_INET)
607 msf->imsf_multiaddr = psin->sin_addr.s_addr;
608 msf->imsf_interface = 0;
609 msf->imsf_fmode = fmode;
610 msf->imsf_numsrc = numsrc;
611 for (i = 0; i < numsrc; ++i) {
612 psin = (struct sockaddr_in *)&list[i];
614 if (psin->sin_family != AF_INET)
616 msf->imsf_slist[i] = psin->sin_addr.s_addr;
618 err = ip_mc_msfilter(sk, msf, ifindex);
624 return -EADDRNOTAVAIL;
627 static int do_ip_setsockopt(struct sock *sk, int level,
628 int optname, char __user *optval, unsigned int optlen)
630 struct inet_sock *inet = inet_sk(sk);
631 struct net *net = sock_net(sk);
633 bool needs_rtnl = setsockopt_needs_rtnl(optname);
644 case IP_MTU_DISCOVER:
646 case IP_ROUTER_ALERT:
652 case IP_BIND_ADDRESS_NO_PORT:
654 case IP_MULTICAST_TTL:
655 case IP_MULTICAST_ALL:
656 case IP_MULTICAST_LOOP:
657 case IP_RECVORIGDSTADDR:
659 case IP_RECVFRAGSIZE:
660 if (optlen >= sizeof(int)) {
661 if (get_user(val, (int __user *) optval))
663 } else if (optlen >= sizeof(char)) {
666 if (get_user(ucval, (unsigned char __user *) optval))
672 /* If optlen==0, it is equivalent to val == 0 */
674 if (optname == IP_ROUTER_ALERT)
675 return ip_ra_control(sk, val ? 1 : 0, NULL);
676 if (ip_mroute_opt(optname))
677 return ip_mroute_setsockopt(sk, optname, optval, optlen);
687 struct ip_options_rcu *old, *opt = NULL;
691 err = ip_options_get_from_user(sock_net(sk), &opt,
695 old = rcu_dereference_protected(inet->inet_opt,
696 lockdep_sock_is_held(sk));
698 struct inet_connection_sock *icsk = inet_csk(sk);
699 #if IS_ENABLED(CONFIG_IPV6)
700 if (sk->sk_family == PF_INET ||
701 (!((1 << sk->sk_state) &
702 (TCPF_LISTEN | TCPF_CLOSE)) &&
703 inet->inet_daddr != LOOPBACK4_IPV6)) {
706 icsk->icsk_ext_hdr_len -= old->opt.optlen;
708 icsk->icsk_ext_hdr_len += opt->opt.optlen;
709 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
710 #if IS_ENABLED(CONFIG_IPV6)
714 rcu_assign_pointer(inet->inet_opt, opt);
721 inet->cmsg_flags |= IP_CMSG_PKTINFO;
723 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
727 inet->cmsg_flags |= IP_CMSG_TTL;
729 inet->cmsg_flags &= ~IP_CMSG_TTL;
733 inet->cmsg_flags |= IP_CMSG_TOS;
735 inet->cmsg_flags &= ~IP_CMSG_TOS;
739 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
741 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
745 inet->cmsg_flags |= IP_CMSG_RETOPTS;
747 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
751 inet->cmsg_flags |= IP_CMSG_PASSSEC;
753 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
755 case IP_RECVORIGDSTADDR:
757 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
759 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
763 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
764 inet_inc_convert_csum(sk);
765 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
768 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
769 inet_dec_convert_csum(sk);
770 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
774 case IP_RECVFRAGSIZE:
775 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
778 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
780 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
782 case IP_TOS: /* This sets both TOS and Precedence */
783 if (sk->sk_type == SOCK_STREAM) {
784 val &= ~INET_ECN_MASK;
785 val |= inet->tos & INET_ECN_MASK;
787 if (inet->tos != val) {
789 sk->sk_priority = rt_tos2priority(val);
796 if (val != -1 && (val < 1 || val > 255))
801 if (sk->sk_type != SOCK_RAW) {
805 inet->hdrincl = val ? 1 : 0;
808 if (sk->sk_type != SOCK_RAW) {
812 inet->nodefrag = val ? 1 : 0;
814 case IP_BIND_ADDRESS_NO_PORT:
815 inet->bind_address_no_port = val ? 1 : 0;
817 case IP_MTU_DISCOVER:
818 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
820 inet->pmtudisc = val;
823 inet->recverr = !!val;
825 skb_queue_purge(&sk->sk_error_queue);
827 case IP_MULTICAST_TTL:
828 if (sk->sk_type == SOCK_STREAM)
834 if (val < 0 || val > 255)
838 case IP_MULTICAST_LOOP:
841 inet->mc_loop = !!val;
845 struct net_device *dev = NULL;
849 if (optlen != sizeof(int))
852 ifindex = (__force int)ntohl((__force __be32)val);
859 dev = dev_get_by_index(sock_net(sk), ifindex);
860 err = -EADDRNOTAVAIL;
864 midx = l3mdev_master_ifindex(dev);
868 if (sk->sk_bound_dev_if &&
869 (!midx || midx != sk->sk_bound_dev_if))
872 inet->uc_index = ifindex;
876 case IP_MULTICAST_IF:
878 struct ip_mreqn mreq;
879 struct net_device *dev = NULL;
882 if (sk->sk_type == SOCK_STREAM)
885 * Check the arguments are allowable
888 if (optlen < sizeof(struct in_addr))
892 if (optlen >= sizeof(struct ip_mreqn)) {
893 if (copy_from_user(&mreq, optval, sizeof(mreq)))
896 memset(&mreq, 0, sizeof(mreq));
897 if (optlen >= sizeof(struct ip_mreq)) {
898 if (copy_from_user(&mreq, optval,
899 sizeof(struct ip_mreq)))
901 } else if (optlen >= sizeof(struct in_addr)) {
902 if (copy_from_user(&mreq.imr_address, optval,
903 sizeof(struct in_addr)))
908 if (!mreq.imr_ifindex) {
909 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
915 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
917 mreq.imr_ifindex = dev->ifindex;
919 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
922 err = -EADDRNOTAVAIL;
926 midx = l3mdev_master_ifindex(dev);
931 if (sk->sk_bound_dev_if &&
932 mreq.imr_ifindex != sk->sk_bound_dev_if &&
933 (!midx || midx != sk->sk_bound_dev_if))
936 inet->mc_index = mreq.imr_ifindex;
937 inet->mc_addr = mreq.imr_address.s_addr;
942 case IP_ADD_MEMBERSHIP:
943 case IP_DROP_MEMBERSHIP:
945 struct ip_mreqn mreq;
948 if (inet_sk(sk)->is_icsk)
951 if (optlen < sizeof(struct ip_mreq))
954 if (optlen >= sizeof(struct ip_mreqn)) {
955 if (copy_from_user(&mreq, optval, sizeof(mreq)))
958 memset(&mreq, 0, sizeof(mreq));
959 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
963 if (optname == IP_ADD_MEMBERSHIP)
964 err = ip_mc_join_group(sk, &mreq);
966 err = ip_mc_leave_group(sk, &mreq);
971 struct ip_msfilter *msf;
973 if (optlen < IP_MSFILTER_SIZE(0))
975 if (optlen > sysctl_optmem_max) {
979 msf = memdup_user(optval, optlen);
984 /* numsrc >= (1G-4) overflow in 32 bits */
985 if (msf->imsf_numsrc >= 0x3ffffffcU ||
986 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
991 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
996 err = ip_mc_msfilter(sk, msf, 0);
1000 case IP_BLOCK_SOURCE:
1001 case IP_UNBLOCK_SOURCE:
1002 case IP_ADD_SOURCE_MEMBERSHIP:
1003 case IP_DROP_SOURCE_MEMBERSHIP:
1005 struct ip_mreq_source mreqs;
1008 if (optlen != sizeof(struct ip_mreq_source))
1010 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
1014 if (optname == IP_BLOCK_SOURCE) {
1015 omode = MCAST_EXCLUDE;
1017 } else if (optname == IP_UNBLOCK_SOURCE) {
1018 omode = MCAST_EXCLUDE;
1020 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1021 struct ip_mreqn mreq;
1023 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1024 mreq.imr_address.s_addr = mreqs.imr_interface;
1025 mreq.imr_ifindex = 0;
1026 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1027 if (err && err != -EADDRINUSE)
1029 omode = MCAST_INCLUDE;
1031 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1032 omode = MCAST_INCLUDE;
1035 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1038 case MCAST_JOIN_GROUP:
1039 case MCAST_LEAVE_GROUP:
1041 struct group_req greq;
1042 struct sockaddr_in *psin;
1043 struct ip_mreqn mreq;
1045 if (optlen < sizeof(struct group_req))
1048 if (copy_from_user(&greq, optval, sizeof(greq)))
1050 psin = (struct sockaddr_in *)&greq.gr_group;
1051 if (psin->sin_family != AF_INET)
1053 memset(&mreq, 0, sizeof(mreq));
1054 mreq.imr_multiaddr = psin->sin_addr;
1055 mreq.imr_ifindex = greq.gr_interface;
1057 if (optname == MCAST_JOIN_GROUP)
1058 err = ip_mc_join_group(sk, &mreq);
1060 err = ip_mc_leave_group(sk, &mreq);
1063 case MCAST_JOIN_SOURCE_GROUP:
1064 case MCAST_LEAVE_SOURCE_GROUP:
1065 case MCAST_BLOCK_SOURCE:
1066 case MCAST_UNBLOCK_SOURCE:
1068 struct group_source_req greqs;
1069 struct ip_mreq_source mreqs;
1070 struct sockaddr_in *psin;
1073 if (optlen != sizeof(struct group_source_req))
1075 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1079 if (greqs.gsr_group.ss_family != AF_INET ||
1080 greqs.gsr_source.ss_family != AF_INET) {
1081 err = -EADDRNOTAVAIL;
1084 psin = (struct sockaddr_in *)&greqs.gsr_group;
1085 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1086 psin = (struct sockaddr_in *)&greqs.gsr_source;
1087 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1088 mreqs.imr_interface = 0; /* use index for mc_source */
1090 if (optname == MCAST_BLOCK_SOURCE) {
1091 omode = MCAST_EXCLUDE;
1093 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1094 omode = MCAST_EXCLUDE;
1096 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1097 struct ip_mreqn mreq;
1099 psin = (struct sockaddr_in *)&greqs.gsr_group;
1100 mreq.imr_multiaddr = psin->sin_addr;
1101 mreq.imr_address.s_addr = 0;
1102 mreq.imr_ifindex = greqs.gsr_interface;
1103 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1104 if (err && err != -EADDRINUSE)
1106 greqs.gsr_interface = mreq.imr_ifindex;
1107 omode = MCAST_INCLUDE;
1109 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1110 omode = MCAST_INCLUDE;
1113 err = ip_mc_source(add, omode, sk, &mreqs,
1114 greqs.gsr_interface);
1117 case MCAST_MSFILTER:
1119 struct group_filter *gsf = NULL;
1121 if (optlen < GROUP_FILTER_SIZE(0))
1123 if (optlen > sysctl_optmem_max) {
1127 gsf = memdup_user(optval, optlen);
1132 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1133 if (gsf->gf_numsrc >= 0x1ffffff ||
1134 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1138 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1142 err = set_mcast_msfilter(sk, gsf->gf_interface,
1143 gsf->gf_numsrc, gsf->gf_fmode,
1144 &gsf->gf_group, gsf->gf_slist);
1149 case IP_MULTICAST_ALL:
1152 if (val != 0 && val != 1)
1160 inet->freebind = !!val;
1163 case IP_IPSEC_POLICY:
1164 case IP_XFRM_POLICY:
1166 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1168 err = xfrm_user_policy(sk, optname, optval, optlen);
1171 case IP_TRANSPARENT:
1172 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1173 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1179 inet->transparent = !!val;
1185 if (val < 0 || val > 255)
1187 inet->min_ttl = val;
1207 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1211 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1212 * destination in skb->cb[] before dst drop.
1213 * This way, receiver doesn't make cache line misses to read rtable.
1215 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1217 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1218 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1221 if (prepare && skb_rtable(skb)) {
1222 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1223 * which has interface index (iif) as the first member of the
1224 * underlying inet{6}_skb_parm struct. This code then overlays
1225 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1226 * element so the iif is picked up from the prior IPCB. If iif
1227 * is the loopback interface, then return the sending interface
1228 * (e.g., process binds socket to eth0 for Tx which is
1229 * redirected to loopback in the rtable/dst).
1231 struct rtable *rt = skb_rtable(skb);
1232 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1234 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1235 pktinfo->ipi_ifindex = inet_iif(skb);
1236 else if (l3slave && rt && rt->rt_iif)
1237 pktinfo->ipi_ifindex = rt->rt_iif;
1239 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1241 pktinfo->ipi_ifindex = 0;
1242 pktinfo->ipi_spec_dst.s_addr = 0;
1247 int ip_setsockopt(struct sock *sk, int level,
1248 int optname, char __user *optval, unsigned int optlen)
1252 if (level != SOL_IP)
1253 return -ENOPROTOOPT;
1255 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1256 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1257 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1258 optname < BPFILTER_IPT_SET_MAX)
1259 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1261 #ifdef CONFIG_NETFILTER
1262 /* we need to exclude all possible ENOPROTOOPTs except default case */
1263 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1264 optname != IP_IPSEC_POLICY &&
1265 optname != IP_XFRM_POLICY &&
1266 !ip_mroute_opt(optname))
1267 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1271 EXPORT_SYMBOL(ip_setsockopt);
1273 #ifdef CONFIG_COMPAT
1274 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1275 char __user *optval, unsigned int optlen)
1279 if (level != SOL_IP)
1280 return -ENOPROTOOPT;
1283 case MCAST_JOIN_GROUP:
1284 case MCAST_LEAVE_GROUP:
1285 case MCAST_JOIN_SOURCE_GROUP:
1286 case MCAST_LEAVE_SOURCE_GROUP:
1287 case MCAST_BLOCK_SOURCE:
1288 case MCAST_UNBLOCK_SOURCE:
1289 case MCAST_MSFILTER:
1290 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1294 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1295 #ifdef CONFIG_NETFILTER
1296 /* we need to exclude all possible ENOPROTOOPTs except default case */
1297 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1298 optname != IP_IPSEC_POLICY &&
1299 optname != IP_XFRM_POLICY &&
1300 !ip_mroute_opt(optname))
1301 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1306 EXPORT_SYMBOL(compat_ip_setsockopt);
1310 * Get the options. Note for future reference. The GET of IP options gets
1311 * the _received_ ones. The set sets the _sent_ ones.
1314 static bool getsockopt_needs_rtnl(int optname)
1318 case MCAST_MSFILTER:
1324 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1325 char __user *optval, int __user *optlen, unsigned int flags)
1327 struct inet_sock *inet = inet_sk(sk);
1328 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1332 if (level != SOL_IP)
1335 if (ip_mroute_opt(optname))
1336 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1338 if (get_user(len, optlen))
1350 unsigned char optbuf[sizeof(struct ip_options)+40];
1351 struct ip_options *opt = (struct ip_options *)optbuf;
1352 struct ip_options_rcu *inet_opt;
1354 inet_opt = rcu_dereference_protected(inet->inet_opt,
1355 lockdep_sock_is_held(sk));
1358 memcpy(optbuf, &inet_opt->opt,
1359 sizeof(struct ip_options) +
1360 inet_opt->opt.optlen);
1363 if (opt->optlen == 0)
1364 return put_user(0, optlen);
1366 ip_options_undo(opt);
1368 len = min_t(unsigned int, len, opt->optlen);
1369 if (put_user(len, optlen))
1371 if (copy_to_user(optval, opt->__data, len))
1376 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1379 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1382 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1385 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1388 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1391 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1393 case IP_RECVORIGDSTADDR:
1394 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1397 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1399 case IP_RECVFRAGSIZE:
1400 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1407 struct net *net = sock_net(sk);
1408 val = (inet->uc_ttl == -1 ?
1409 net->ipv4.sysctl_ip_default_ttl :
1414 val = inet->hdrincl;
1417 val = inet->nodefrag;
1419 case IP_BIND_ADDRESS_NO_PORT:
1420 val = inet->bind_address_no_port;
1422 case IP_MTU_DISCOVER:
1423 val = inet->pmtudisc;
1427 struct dst_entry *dst;
1429 dst = sk_dst_get(sk);
1441 val = inet->recverr;
1443 case IP_MULTICAST_TTL:
1446 case IP_MULTICAST_LOOP:
1447 val = inet->mc_loop;
1450 val = (__force int)htonl((__u32) inet->uc_index);
1452 case IP_MULTICAST_IF:
1454 struct in_addr addr;
1455 len = min_t(unsigned int, len, sizeof(struct in_addr));
1456 addr.s_addr = inet->mc_addr;
1459 if (put_user(len, optlen))
1461 if (copy_to_user(optval, &addr, len))
1467 struct ip_msfilter msf;
1469 if (len < IP_MSFILTER_SIZE(0)) {
1473 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1477 err = ip_mc_msfget(sk, &msf,
1478 (struct ip_msfilter __user *)optval, optlen);
1481 case MCAST_MSFILTER:
1483 struct group_filter __user *p = (void __user *)optval;
1484 struct group_filter gsf;
1485 const int size0 = offsetof(struct group_filter, gf_slist);
1492 if (copy_from_user(&gsf, p, size0)) {
1496 num = gsf.gf_numsrc;
1497 err = ip_mc_gsfget(sk, &gsf, p->gf_slist);
1500 if (gsf.gf_numsrc < num)
1501 num = gsf.gf_numsrc;
1502 if (put_user(GROUP_FILTER_SIZE(num), optlen) ||
1503 copy_to_user(p, &gsf, size0))
1507 case IP_MULTICAST_ALL:
1516 if (sk->sk_type != SOCK_STREAM)
1517 return -ENOPROTOOPT;
1519 msg.msg_control_is_user = true;
1520 msg.msg_control_user = optval;
1521 msg.msg_controllen = len;
1522 msg.msg_flags = flags;
1524 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1525 struct in_pktinfo info;
1527 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1528 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1529 info.ipi_ifindex = inet->mc_index;
1530 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1532 if (inet->cmsg_flags & IP_CMSG_TTL) {
1533 int hlim = inet->mc_ttl;
1534 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1536 if (inet->cmsg_flags & IP_CMSG_TOS) {
1537 int tos = inet->rcv_tos;
1538 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1540 len -= msg.msg_controllen;
1541 return put_user(len, optlen);
1544 val = inet->freebind;
1546 case IP_TRANSPARENT:
1547 val = inet->transparent;
1550 val = inet->min_ttl;
1554 return -ENOPROTOOPT;
1558 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1559 unsigned char ucval = (unsigned char)val;
1561 if (put_user(len, optlen))
1563 if (copy_to_user(optval, &ucval, 1))
1566 len = min_t(unsigned int, sizeof(int), len);
1567 if (put_user(len, optlen))
1569 if (copy_to_user(optval, &val, len))
1581 int ip_getsockopt(struct sock *sk, int level,
1582 int optname, char __user *optval, int __user *optlen)
1586 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1587 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1588 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1589 optname < BPFILTER_IPT_GET_MAX)
1590 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1592 #ifdef CONFIG_NETFILTER
1593 /* we need to exclude all possible ENOPROTOOPTs except default case */
1594 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1595 !ip_mroute_opt(optname)) {
1598 if (get_user(len, optlen))
1601 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1603 err = put_user(len, optlen);
1609 EXPORT_SYMBOL(ip_getsockopt);
1611 #ifdef CONFIG_COMPAT
1612 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1613 char __user *optval, int __user *optlen)
1617 if (optname == MCAST_MSFILTER) {
1618 const int size0 = offsetof(struct compat_group_filter, gf_slist);
1619 struct compat_group_filter __user *p = (void __user *)optval;
1620 struct compat_group_filter gf32;
1621 struct group_filter gf;
1625 if (level != SOL_IP)
1628 if (get_user(ulen, optlen))
1634 if (copy_from_user(&gf32, p, size0))
1637 gf.gf_interface = gf32.gf_interface;
1638 gf.gf_fmode = gf32.gf_fmode;
1639 num = gf.gf_numsrc = gf32.gf_numsrc;
1640 gf.gf_group = gf32.gf_group;
1644 err = ip_mc_gsfget(sk, &gf, p->gf_slist);
1649 if (gf.gf_numsrc < num)
1651 ulen = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1652 if (put_user(ulen, optlen) ||
1653 put_user(gf.gf_fmode, &p->gf_fmode) ||
1654 put_user(gf.gf_numsrc, &p->gf_numsrc))
1659 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1662 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1663 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1664 optname < BPFILTER_IPT_GET_MAX)
1665 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1667 #ifdef CONFIG_NETFILTER
1668 /* we need to exclude all possible ENOPROTOOPTs except default case */
1669 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1670 !ip_mroute_opt(optname)) {
1673 if (get_user(len, optlen))
1676 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1678 err = put_user(len, optlen);
1684 EXPORT_SYMBOL(compat_ip_getsockopt);