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;
356 mutex_lock(&net->ipv4.ra_mutex);
357 for (rap = &net->ipv4.ra_chain;
358 (ra = rcu_dereference_protected(*rap,
359 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
363 mutex_unlock(&net->ipv4.ra_mutex);
367 /* dont let ip_call_ra_chain() use sk again */
369 RCU_INIT_POINTER(*rap, ra->next);
370 mutex_unlock(&net->ipv4.ra_mutex);
375 * Delay sock_put(sk) and kfree(ra) after one rcu grace
376 * period. This guarantee ip_call_ra_chain() dont need
377 * to mess with socket refcounts.
380 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
385 mutex_unlock(&net->ipv4.ra_mutex);
389 new_ra->destructor = destructor;
391 RCU_INIT_POINTER(new_ra->next, ra);
392 rcu_assign_pointer(*rap, new_ra);
394 mutex_unlock(&net->ipv4.ra_mutex);
399 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
400 __be16 port, u32 info, u8 *payload)
402 struct sock_exterr_skb *serr;
404 skb = skb_clone(skb, GFP_ATOMIC);
408 serr = SKB_EXT_ERR(skb);
409 serr->ee.ee_errno = err;
410 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
411 serr->ee.ee_type = icmp_hdr(skb)->type;
412 serr->ee.ee_code = icmp_hdr(skb)->code;
414 serr->ee.ee_info = info;
415 serr->ee.ee_data = 0;
416 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
417 skb_network_header(skb);
420 if (skb_pull(skb, payload - skb->data)) {
421 skb_reset_transport_header(skb);
422 if (sock_queue_err_skb(sk, skb) == 0)
428 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
430 struct inet_sock *inet = inet_sk(sk);
431 struct sock_exterr_skb *serr;
438 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
442 skb_put(skb, sizeof(struct iphdr));
443 skb_reset_network_header(skb);
447 serr = SKB_EXT_ERR(skb);
448 serr->ee.ee_errno = err;
449 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
450 serr->ee.ee_type = 0;
451 serr->ee.ee_code = 0;
453 serr->ee.ee_info = info;
454 serr->ee.ee_data = 0;
455 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
458 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
459 skb_reset_transport_header(skb);
461 if (sock_queue_err_skb(sk, skb))
465 /* For some errors we have valid addr_offset even with zero payload and
466 * zero port. Also, addr_offset should be supported if port is set.
468 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
470 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
471 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
474 /* IPv4 supports cmsg on all imcp errors and some timestamps
476 * Timestamp code paths do not initialize the fields expected by cmsg:
477 * the PKTINFO fields in skb->cb[]. Fill those in here.
479 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
483 struct in_pktinfo *info;
485 if (ee_origin == SO_EE_ORIGIN_ICMP)
488 if (ee_origin == SO_EE_ORIGIN_LOCAL)
491 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
492 * timestamp with egress dev. Not possible for packets without iif
493 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
495 info = PKTINFO_SKB_CB(skb);
496 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
500 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
505 * Handle MSG_ERRQUEUE
507 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
509 struct sock_exterr_skb *serr;
511 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
513 struct sock_extended_err ee;
514 struct sockaddr_in offender;
520 skb = sock_dequeue_err_skb(sk);
526 msg->msg_flags |= MSG_TRUNC;
529 err = skb_copy_datagram_msg(skb, 0, msg, copied);
534 sock_recv_timestamp(msg, sk, skb);
536 serr = SKB_EXT_ERR(skb);
538 if (sin && ipv4_datagram_support_addr(serr)) {
539 sin->sin_family = AF_INET;
540 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
542 sin->sin_port = serr->port;
543 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
544 *addr_len = sizeof(*sin);
547 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
548 sin = &errhdr.offender;
549 memset(sin, 0, sizeof(*sin));
551 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
552 sin->sin_family = AF_INET;
553 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
554 if (inet_sk(sk)->cmsg_flags)
555 ip_cmsg_recv(msg, skb);
558 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
560 /* Now we could try to dump offended packet options */
562 msg->msg_flags |= MSG_ERRQUEUE;
572 * Socket option code for IP. This is the end of the line after any
573 * TCP,UDP etc options on an IP socket.
575 static bool setsockopt_needs_rtnl(int optname)
578 case IP_ADD_MEMBERSHIP:
579 case IP_ADD_SOURCE_MEMBERSHIP:
580 case IP_BLOCK_SOURCE:
581 case IP_DROP_MEMBERSHIP:
582 case IP_DROP_SOURCE_MEMBERSHIP:
584 case IP_UNBLOCK_SOURCE:
585 case MCAST_BLOCK_SOURCE:
587 case MCAST_JOIN_GROUP:
588 case MCAST_JOIN_SOURCE_GROUP:
589 case MCAST_LEAVE_GROUP:
590 case MCAST_LEAVE_SOURCE_GROUP:
591 case MCAST_UNBLOCK_SOURCE:
597 static int do_ip_setsockopt(struct sock *sk, int level,
598 int optname, char __user *optval, unsigned int optlen)
600 struct inet_sock *inet = inet_sk(sk);
601 struct net *net = sock_net(sk);
603 bool needs_rtnl = setsockopt_needs_rtnl(optname);
614 case IP_MTU_DISCOVER:
616 case IP_ROUTER_ALERT:
622 case IP_BIND_ADDRESS_NO_PORT:
624 case IP_MULTICAST_TTL:
625 case IP_MULTICAST_ALL:
626 case IP_MULTICAST_LOOP:
627 case IP_RECVORIGDSTADDR:
629 case IP_RECVFRAGSIZE:
630 if (optlen >= sizeof(int)) {
631 if (get_user(val, (int __user *) optval))
633 } else if (optlen >= sizeof(char)) {
636 if (get_user(ucval, (unsigned char __user *) optval))
642 /* If optlen==0, it is equivalent to val == 0 */
644 if (optname == IP_ROUTER_ALERT)
645 return ip_ra_control(sk, val ? 1 : 0, NULL);
646 if (ip_mroute_opt(optname))
647 return ip_mroute_setsockopt(sk, optname, optval, optlen);
657 struct ip_options_rcu *old, *opt = NULL;
661 err = ip_options_get_from_user(sock_net(sk), &opt,
665 old = rcu_dereference_protected(inet->inet_opt,
666 lockdep_sock_is_held(sk));
668 struct inet_connection_sock *icsk = inet_csk(sk);
669 #if IS_ENABLED(CONFIG_IPV6)
670 if (sk->sk_family == PF_INET ||
671 (!((1 << sk->sk_state) &
672 (TCPF_LISTEN | TCPF_CLOSE)) &&
673 inet->inet_daddr != LOOPBACK4_IPV6)) {
676 icsk->icsk_ext_hdr_len -= old->opt.optlen;
678 icsk->icsk_ext_hdr_len += opt->opt.optlen;
679 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
680 #if IS_ENABLED(CONFIG_IPV6)
684 rcu_assign_pointer(inet->inet_opt, opt);
691 inet->cmsg_flags |= IP_CMSG_PKTINFO;
693 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
697 inet->cmsg_flags |= IP_CMSG_TTL;
699 inet->cmsg_flags &= ~IP_CMSG_TTL;
703 inet->cmsg_flags |= IP_CMSG_TOS;
705 inet->cmsg_flags &= ~IP_CMSG_TOS;
709 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
711 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
715 inet->cmsg_flags |= IP_CMSG_RETOPTS;
717 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
721 inet->cmsg_flags |= IP_CMSG_PASSSEC;
723 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
725 case IP_RECVORIGDSTADDR:
727 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
729 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
733 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
734 inet_inc_convert_csum(sk);
735 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
738 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
739 inet_dec_convert_csum(sk);
740 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
744 case IP_RECVFRAGSIZE:
745 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
748 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
750 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
752 case IP_TOS: /* This sets both TOS and Precedence */
753 if (sk->sk_type == SOCK_STREAM) {
754 val &= ~INET_ECN_MASK;
755 val |= inet->tos & INET_ECN_MASK;
757 if (inet->tos != val) {
759 sk->sk_priority = rt_tos2priority(val);
766 if (val != -1 && (val < 1 || val > 255))
771 if (sk->sk_type != SOCK_RAW) {
775 inet->hdrincl = val ? 1 : 0;
778 if (sk->sk_type != SOCK_RAW) {
782 inet->nodefrag = val ? 1 : 0;
784 case IP_BIND_ADDRESS_NO_PORT:
785 inet->bind_address_no_port = val ? 1 : 0;
787 case IP_MTU_DISCOVER:
788 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
790 inet->pmtudisc = val;
793 inet->recverr = !!val;
795 skb_queue_purge(&sk->sk_error_queue);
797 case IP_MULTICAST_TTL:
798 if (sk->sk_type == SOCK_STREAM)
804 if (val < 0 || val > 255)
808 case IP_MULTICAST_LOOP:
811 inet->mc_loop = !!val;
815 struct net_device *dev = NULL;
819 if (optlen != sizeof(int))
822 ifindex = (__force int)ntohl((__force __be32)val);
829 dev = dev_get_by_index(sock_net(sk), ifindex);
830 err = -EADDRNOTAVAIL;
834 midx = l3mdev_master_ifindex(dev);
838 if (sk->sk_bound_dev_if &&
839 (!midx || midx != sk->sk_bound_dev_if))
842 inet->uc_index = ifindex;
846 case IP_MULTICAST_IF:
848 struct ip_mreqn mreq;
849 struct net_device *dev = NULL;
852 if (sk->sk_type == SOCK_STREAM)
855 * Check the arguments are allowable
858 if (optlen < sizeof(struct in_addr))
862 if (optlen >= sizeof(struct ip_mreqn)) {
863 if (copy_from_user(&mreq, optval, sizeof(mreq)))
866 memset(&mreq, 0, sizeof(mreq));
867 if (optlen >= sizeof(struct ip_mreq)) {
868 if (copy_from_user(&mreq, optval,
869 sizeof(struct ip_mreq)))
871 } else if (optlen >= sizeof(struct in_addr)) {
872 if (copy_from_user(&mreq.imr_address, optval,
873 sizeof(struct in_addr)))
878 if (!mreq.imr_ifindex) {
879 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
885 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
887 mreq.imr_ifindex = dev->ifindex;
889 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
892 err = -EADDRNOTAVAIL;
896 midx = l3mdev_master_ifindex(dev);
901 if (sk->sk_bound_dev_if &&
902 mreq.imr_ifindex != sk->sk_bound_dev_if &&
903 (!midx || midx != sk->sk_bound_dev_if))
906 inet->mc_index = mreq.imr_ifindex;
907 inet->mc_addr = mreq.imr_address.s_addr;
912 case IP_ADD_MEMBERSHIP:
913 case IP_DROP_MEMBERSHIP:
915 struct ip_mreqn mreq;
918 if (inet_sk(sk)->is_icsk)
921 if (optlen < sizeof(struct ip_mreq))
924 if (optlen >= sizeof(struct ip_mreqn)) {
925 if (copy_from_user(&mreq, optval, sizeof(mreq)))
928 memset(&mreq, 0, sizeof(mreq));
929 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
933 if (optname == IP_ADD_MEMBERSHIP)
934 err = ip_mc_join_group(sk, &mreq);
936 err = ip_mc_leave_group(sk, &mreq);
941 struct ip_msfilter *msf;
943 if (optlen < IP_MSFILTER_SIZE(0))
945 if (optlen > sysctl_optmem_max) {
949 msf = memdup_user(optval, optlen);
954 /* numsrc >= (1G-4) overflow in 32 bits */
955 if (msf->imsf_numsrc >= 0x3ffffffcU ||
956 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
961 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
966 err = ip_mc_msfilter(sk, msf, 0);
970 case IP_BLOCK_SOURCE:
971 case IP_UNBLOCK_SOURCE:
972 case IP_ADD_SOURCE_MEMBERSHIP:
973 case IP_DROP_SOURCE_MEMBERSHIP:
975 struct ip_mreq_source mreqs;
978 if (optlen != sizeof(struct ip_mreq_source))
980 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
984 if (optname == IP_BLOCK_SOURCE) {
985 omode = MCAST_EXCLUDE;
987 } else if (optname == IP_UNBLOCK_SOURCE) {
988 omode = MCAST_EXCLUDE;
990 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
991 struct ip_mreqn mreq;
993 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
994 mreq.imr_address.s_addr = mreqs.imr_interface;
995 mreq.imr_ifindex = 0;
996 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
997 if (err && err != -EADDRINUSE)
999 omode = MCAST_INCLUDE;
1001 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1002 omode = MCAST_INCLUDE;
1005 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1008 case MCAST_JOIN_GROUP:
1009 case MCAST_LEAVE_GROUP:
1011 struct group_req greq;
1012 struct sockaddr_in *psin;
1013 struct ip_mreqn mreq;
1015 if (optlen < sizeof(struct group_req))
1018 if (copy_from_user(&greq, optval, sizeof(greq)))
1020 psin = (struct sockaddr_in *)&greq.gr_group;
1021 if (psin->sin_family != AF_INET)
1023 memset(&mreq, 0, sizeof(mreq));
1024 mreq.imr_multiaddr = psin->sin_addr;
1025 mreq.imr_ifindex = greq.gr_interface;
1027 if (optname == MCAST_JOIN_GROUP)
1028 err = ip_mc_join_group(sk, &mreq);
1030 err = ip_mc_leave_group(sk, &mreq);
1033 case MCAST_JOIN_SOURCE_GROUP:
1034 case MCAST_LEAVE_SOURCE_GROUP:
1035 case MCAST_BLOCK_SOURCE:
1036 case MCAST_UNBLOCK_SOURCE:
1038 struct group_source_req greqs;
1039 struct ip_mreq_source mreqs;
1040 struct sockaddr_in *psin;
1043 if (optlen != sizeof(struct group_source_req))
1045 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1049 if (greqs.gsr_group.ss_family != AF_INET ||
1050 greqs.gsr_source.ss_family != AF_INET) {
1051 err = -EADDRNOTAVAIL;
1054 psin = (struct sockaddr_in *)&greqs.gsr_group;
1055 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1056 psin = (struct sockaddr_in *)&greqs.gsr_source;
1057 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1058 mreqs.imr_interface = 0; /* use index for mc_source */
1060 if (optname == MCAST_BLOCK_SOURCE) {
1061 omode = MCAST_EXCLUDE;
1063 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1064 omode = MCAST_EXCLUDE;
1066 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1067 struct ip_mreqn mreq;
1069 psin = (struct sockaddr_in *)&greqs.gsr_group;
1070 mreq.imr_multiaddr = psin->sin_addr;
1071 mreq.imr_address.s_addr = 0;
1072 mreq.imr_ifindex = greqs.gsr_interface;
1073 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1074 if (err && err != -EADDRINUSE)
1076 greqs.gsr_interface = mreq.imr_ifindex;
1077 omode = MCAST_INCLUDE;
1079 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1080 omode = MCAST_INCLUDE;
1083 err = ip_mc_source(add, omode, sk, &mreqs,
1084 greqs.gsr_interface);
1087 case MCAST_MSFILTER:
1089 struct sockaddr_in *psin;
1090 struct ip_msfilter *msf = NULL;
1091 struct group_filter *gsf = NULL;
1092 int msize, i, ifindex;
1094 if (optlen < GROUP_FILTER_SIZE(0))
1096 if (optlen > sysctl_optmem_max) {
1100 gsf = memdup_user(optval, optlen);
1106 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1107 if (gsf->gf_numsrc >= 0x1ffffff ||
1108 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1112 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1116 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1117 msf = kmalloc(msize, GFP_KERNEL);
1122 ifindex = gsf->gf_interface;
1123 psin = (struct sockaddr_in *)&gsf->gf_group;
1124 if (psin->sin_family != AF_INET) {
1125 err = -EADDRNOTAVAIL;
1128 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1129 msf->imsf_interface = 0;
1130 msf->imsf_fmode = gsf->gf_fmode;
1131 msf->imsf_numsrc = gsf->gf_numsrc;
1132 err = -EADDRNOTAVAIL;
1133 for (i = 0; i < gsf->gf_numsrc; ++i) {
1134 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1136 if (psin->sin_family != AF_INET)
1138 msf->imsf_slist[i] = psin->sin_addr.s_addr;
1143 err = ip_mc_msfilter(sk, msf, ifindex);
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;
1282 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1283 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1286 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1287 #ifdef CONFIG_NETFILTER
1288 /* we need to exclude all possible ENOPROTOOPTs except default case */
1289 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1290 optname != IP_IPSEC_POLICY &&
1291 optname != IP_XFRM_POLICY &&
1292 !ip_mroute_opt(optname))
1293 err = compat_nf_setsockopt(sk, PF_INET, optname, optval,
1298 EXPORT_SYMBOL(compat_ip_setsockopt);
1302 * Get the options. Note for future reference. The GET of IP options gets
1303 * the _received_ ones. The set sets the _sent_ ones.
1306 static bool getsockopt_needs_rtnl(int optname)
1310 case MCAST_MSFILTER:
1316 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1317 char __user *optval, int __user *optlen, unsigned int flags)
1319 struct inet_sock *inet = inet_sk(sk);
1320 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1324 if (level != SOL_IP)
1327 if (ip_mroute_opt(optname))
1328 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1330 if (get_user(len, optlen))
1342 unsigned char optbuf[sizeof(struct ip_options)+40];
1343 struct ip_options *opt = (struct ip_options *)optbuf;
1344 struct ip_options_rcu *inet_opt;
1346 inet_opt = rcu_dereference_protected(inet->inet_opt,
1347 lockdep_sock_is_held(sk));
1350 memcpy(optbuf, &inet_opt->opt,
1351 sizeof(struct ip_options) +
1352 inet_opt->opt.optlen);
1355 if (opt->optlen == 0)
1356 return put_user(0, optlen);
1358 ip_options_undo(opt);
1360 len = min_t(unsigned int, len, opt->optlen);
1361 if (put_user(len, optlen))
1363 if (copy_to_user(optval, opt->__data, len))
1368 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1371 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1374 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1377 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1380 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1383 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1385 case IP_RECVORIGDSTADDR:
1386 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1389 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1391 case IP_RECVFRAGSIZE:
1392 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1399 struct net *net = sock_net(sk);
1400 val = (inet->uc_ttl == -1 ?
1401 net->ipv4.sysctl_ip_default_ttl :
1406 val = inet->hdrincl;
1409 val = inet->nodefrag;
1411 case IP_BIND_ADDRESS_NO_PORT:
1412 val = inet->bind_address_no_port;
1414 case IP_MTU_DISCOVER:
1415 val = inet->pmtudisc;
1419 struct dst_entry *dst;
1421 dst = sk_dst_get(sk);
1433 val = inet->recverr;
1435 case IP_MULTICAST_TTL:
1438 case IP_MULTICAST_LOOP:
1439 val = inet->mc_loop;
1442 val = (__force int)htonl((__u32) inet->uc_index);
1444 case IP_MULTICAST_IF:
1446 struct in_addr addr;
1447 len = min_t(unsigned int, len, sizeof(struct in_addr));
1448 addr.s_addr = inet->mc_addr;
1451 if (put_user(len, optlen))
1453 if (copy_to_user(optval, &addr, len))
1459 struct ip_msfilter msf;
1461 if (len < IP_MSFILTER_SIZE(0)) {
1465 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1469 err = ip_mc_msfget(sk, &msf,
1470 (struct ip_msfilter __user *)optval, optlen);
1473 case MCAST_MSFILTER:
1475 struct group_filter gsf;
1477 if (len < GROUP_FILTER_SIZE(0)) {
1481 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1485 err = ip_mc_gsfget(sk, &gsf,
1486 (struct group_filter __user *)optval,
1490 case IP_MULTICAST_ALL:
1499 if (sk->sk_type != SOCK_STREAM)
1500 return -ENOPROTOOPT;
1502 msg.msg_control = (__force void *) optval;
1503 msg.msg_controllen = len;
1504 msg.msg_flags = flags;
1506 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1507 struct in_pktinfo info;
1509 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1510 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1511 info.ipi_ifindex = inet->mc_index;
1512 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1514 if (inet->cmsg_flags & IP_CMSG_TTL) {
1515 int hlim = inet->mc_ttl;
1516 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1518 if (inet->cmsg_flags & IP_CMSG_TOS) {
1519 int tos = inet->rcv_tos;
1520 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1522 len -= msg.msg_controllen;
1523 return put_user(len, optlen);
1526 val = inet->freebind;
1528 case IP_TRANSPARENT:
1529 val = inet->transparent;
1532 val = inet->min_ttl;
1535 val = inet_sk(sk)->inet_num;
1539 return -ENOPROTOOPT;
1543 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1544 unsigned char ucval = (unsigned char)val;
1546 if (put_user(len, optlen))
1548 if (copy_to_user(optval, &ucval, 1))
1551 len = min_t(unsigned int, sizeof(int), len);
1552 if (put_user(len, optlen))
1554 if (copy_to_user(optval, &val, len))
1566 int ip_getsockopt(struct sock *sk, int level,
1567 int optname, char __user *optval, int __user *optlen)
1571 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1572 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1573 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1574 optname < BPFILTER_IPT_GET_MAX)
1575 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1577 #ifdef CONFIG_NETFILTER
1578 /* we need to exclude all possible ENOPROTOOPTs except default case */
1579 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1580 !ip_mroute_opt(optname)) {
1583 if (get_user(len, optlen))
1586 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1588 err = put_user(len, optlen);
1594 EXPORT_SYMBOL(ip_getsockopt);
1596 #ifdef CONFIG_COMPAT
1597 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1598 char __user *optval, int __user *optlen)
1602 if (optname == MCAST_MSFILTER)
1603 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1606 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1609 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1610 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1611 optname < BPFILTER_IPT_GET_MAX)
1612 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1614 #ifdef CONFIG_NETFILTER
1615 /* we need to exclude all possible ENOPROTOOPTs except default case */
1616 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1617 !ip_mroute_opt(optname)) {
1620 if (get_user(len, optlen))
1623 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1625 err = put_user(len, optlen);
1631 EXPORT_SYMBOL(compat_ip_getsockopt);