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 if (inet_sk(sk)->recverr_rfc4884)
415 ip_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
417 skb_reset_transport_header(skb);
418 if (sock_queue_err_skb(sk, skb) == 0)
424 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
426 struct inet_sock *inet = inet_sk(sk);
427 struct sock_exterr_skb *serr;
434 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
438 skb_put(skb, sizeof(struct iphdr));
439 skb_reset_network_header(skb);
443 serr = SKB_EXT_ERR(skb);
444 serr->ee.ee_errno = err;
445 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
446 serr->ee.ee_type = 0;
447 serr->ee.ee_code = 0;
449 serr->ee.ee_info = info;
450 serr->ee.ee_data = 0;
451 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
454 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
455 skb_reset_transport_header(skb);
457 if (sock_queue_err_skb(sk, skb))
461 /* For some errors we have valid addr_offset even with zero payload and
462 * zero port. Also, addr_offset should be supported if port is set.
464 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
466 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
467 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
470 /* IPv4 supports cmsg on all imcp errors and some timestamps
472 * Timestamp code paths do not initialize the fields expected by cmsg:
473 * the PKTINFO fields in skb->cb[]. Fill those in here.
475 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
479 struct in_pktinfo *info;
481 if (ee_origin == SO_EE_ORIGIN_ICMP)
484 if (ee_origin == SO_EE_ORIGIN_LOCAL)
487 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
488 * timestamp with egress dev. Not possible for packets without iif
489 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
491 info = PKTINFO_SKB_CB(skb);
492 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
496 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
501 * Handle MSG_ERRQUEUE
503 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
505 struct sock_exterr_skb *serr;
507 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
509 struct sock_extended_err ee;
510 struct sockaddr_in offender;
516 skb = sock_dequeue_err_skb(sk);
522 msg->msg_flags |= MSG_TRUNC;
525 err = skb_copy_datagram_msg(skb, 0, msg, copied);
530 sock_recv_timestamp(msg, sk, skb);
532 serr = SKB_EXT_ERR(skb);
534 if (sin && ipv4_datagram_support_addr(serr)) {
535 sin->sin_family = AF_INET;
536 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
538 sin->sin_port = serr->port;
539 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
540 *addr_len = sizeof(*sin);
543 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
544 sin = &errhdr.offender;
545 memset(sin, 0, sizeof(*sin));
547 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
548 sin->sin_family = AF_INET;
549 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
550 if (inet_sk(sk)->cmsg_flags)
551 ip_cmsg_recv(msg, skb);
554 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
556 /* Now we could try to dump offended packet options */
558 msg->msg_flags |= MSG_ERRQUEUE;
566 static void __ip_sock_set_tos(struct sock *sk, int val)
568 if (sk->sk_type == SOCK_STREAM) {
569 val &= ~INET_ECN_MASK;
570 val |= inet_sk(sk)->tos & INET_ECN_MASK;
572 if (inet_sk(sk)->tos != val) {
573 inet_sk(sk)->tos = val;
574 sk->sk_priority = rt_tos2priority(val);
579 void ip_sock_set_tos(struct sock *sk, int val)
582 __ip_sock_set_tos(sk, val);
585 EXPORT_SYMBOL(ip_sock_set_tos);
587 void ip_sock_set_freebind(struct sock *sk)
590 inet_sk(sk)->freebind = true;
593 EXPORT_SYMBOL(ip_sock_set_freebind);
595 void ip_sock_set_recverr(struct sock *sk)
598 inet_sk(sk)->recverr = true;
601 EXPORT_SYMBOL(ip_sock_set_recverr);
603 int ip_sock_set_mtu_discover(struct sock *sk, int val)
605 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
608 inet_sk(sk)->pmtudisc = val;
612 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
614 void ip_sock_set_pktinfo(struct sock *sk)
617 inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
620 EXPORT_SYMBOL(ip_sock_set_pktinfo);
623 * Socket option code for IP. This is the end of the line after any
624 * TCP,UDP etc options on an IP socket.
626 static bool setsockopt_needs_rtnl(int optname)
629 case IP_ADD_MEMBERSHIP:
630 case IP_ADD_SOURCE_MEMBERSHIP:
631 case IP_BLOCK_SOURCE:
632 case IP_DROP_MEMBERSHIP:
633 case IP_DROP_SOURCE_MEMBERSHIP:
635 case IP_UNBLOCK_SOURCE:
636 case MCAST_BLOCK_SOURCE:
638 case MCAST_JOIN_GROUP:
639 case MCAST_JOIN_SOURCE_GROUP:
640 case MCAST_LEAVE_GROUP:
641 case MCAST_LEAVE_SOURCE_GROUP:
642 case MCAST_UNBLOCK_SOURCE:
648 static int set_mcast_msfilter(struct sock *sk, int ifindex,
649 int numsrc, int fmode,
650 struct sockaddr_storage *group,
651 struct sockaddr_storage *list)
653 int msize = IP_MSFILTER_SIZE(numsrc);
654 struct ip_msfilter *msf;
655 struct sockaddr_in *psin;
658 msf = kmalloc(msize, GFP_KERNEL);
662 psin = (struct sockaddr_in *)group;
663 if (psin->sin_family != AF_INET)
665 msf->imsf_multiaddr = psin->sin_addr.s_addr;
666 msf->imsf_interface = 0;
667 msf->imsf_fmode = fmode;
668 msf->imsf_numsrc = numsrc;
669 for (i = 0; i < numsrc; ++i) {
670 psin = (struct sockaddr_in *)&list[i];
672 if (psin->sin_family != AF_INET)
674 msf->imsf_slist[i] = psin->sin_addr.s_addr;
676 err = ip_mc_msfilter(sk, msf, ifindex);
682 return -EADDRNOTAVAIL;
685 static int copy_group_source_from_user(struct group_source_req *greqs,
686 void __user *optval, int optlen)
688 if (in_compat_syscall()) {
689 struct compat_group_source_req gr32;
691 if (optlen != sizeof(gr32))
693 if (copy_from_user(&gr32, optval, sizeof(gr32)))
695 greqs->gsr_interface = gr32.gsr_interface;
696 greqs->gsr_group = gr32.gsr_group;
697 greqs->gsr_source = gr32.gsr_source;
699 if (optlen != sizeof(*greqs))
701 if (copy_from_user(greqs, optval, sizeof(*greqs)))
708 static int do_mcast_group_source(struct sock *sk, int optname,
709 void __user *optval, int optlen)
711 struct group_source_req greqs;
712 struct ip_mreq_source mreqs;
713 struct sockaddr_in *psin;
716 err = copy_group_source_from_user(&greqs, optval, optlen);
720 if (greqs.gsr_group.ss_family != AF_INET ||
721 greqs.gsr_source.ss_family != AF_INET)
722 return -EADDRNOTAVAIL;
724 psin = (struct sockaddr_in *)&greqs.gsr_group;
725 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
726 psin = (struct sockaddr_in *)&greqs.gsr_source;
727 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
728 mreqs.imr_interface = 0; /* use index for mc_source */
730 if (optname == MCAST_BLOCK_SOURCE) {
731 omode = MCAST_EXCLUDE;
733 } else if (optname == MCAST_UNBLOCK_SOURCE) {
734 omode = MCAST_EXCLUDE;
736 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
737 struct ip_mreqn mreq;
739 psin = (struct sockaddr_in *)&greqs.gsr_group;
740 mreq.imr_multiaddr = psin->sin_addr;
741 mreq.imr_address.s_addr = 0;
742 mreq.imr_ifindex = greqs.gsr_interface;
743 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
744 if (err && err != -EADDRINUSE)
746 greqs.gsr_interface = mreq.imr_ifindex;
747 omode = MCAST_INCLUDE;
749 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
750 omode = MCAST_INCLUDE;
753 return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
756 static int ip_set_mcast_msfilter(struct sock *sk, void __user *optval,
759 struct group_filter *gsf = NULL;
762 if (optlen < GROUP_FILTER_SIZE(0))
764 if (optlen > sysctl_optmem_max)
767 gsf = memdup_user(optval, optlen);
771 /* numsrc >= (4G-140)/128 overflow in 32 bits */
773 if (gsf->gf_numsrc >= 0x1ffffff ||
774 gsf->gf_numsrc > sock_net(sk)->ipv4.sysctl_igmp_max_msf)
778 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
781 err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
782 gsf->gf_fmode, &gsf->gf_group, gsf->gf_slist);
788 static int compat_ip_set_mcast_msfilter(struct sock *sk, void __user *optval,
791 const int size0 = offsetof(struct compat_group_filter, gf_slist);
792 struct compat_group_filter *gf32;
799 if (optlen > sysctl_optmem_max - 4)
802 p = kmalloc(optlen + 4, GFP_KERNEL);
805 gf32 = p + 4; /* we want ->gf_group and ->gf_slist aligned */
808 if (copy_from_user(gf32, optval, optlen))
811 /* numsrc >= (4G-140)/128 overflow in 32 bits */
818 if (offsetof(struct compat_group_filter, gf_slist[n]) > optlen)
821 /* numsrc >= (4G-140)/128 overflow in 32 bits */
823 if (n > sock_net(sk)->ipv4.sysctl_igmp_max_msf)
825 err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
826 &gf32->gf_group, gf32->gf_slist);
832 static int ip_mcast_join_leave(struct sock *sk, int optname,
833 void __user *optval, int optlen)
835 struct ip_mreqn mreq = { };
836 struct sockaddr_in *psin;
837 struct group_req greq;
839 if (optlen < sizeof(struct group_req))
841 if (copy_from_user(&greq, optval, sizeof(greq)))
844 psin = (struct sockaddr_in *)&greq.gr_group;
845 if (psin->sin_family != AF_INET)
847 mreq.imr_multiaddr = psin->sin_addr;
848 mreq.imr_ifindex = greq.gr_interface;
849 if (optname == MCAST_JOIN_GROUP)
850 return ip_mc_join_group(sk, &mreq);
851 return ip_mc_leave_group(sk, &mreq);
854 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
855 void __user *optval, int optlen)
857 struct compat_group_req greq;
858 struct ip_mreqn mreq = { };
859 struct sockaddr_in *psin;
861 if (optlen < sizeof(struct compat_group_req))
863 if (copy_from_user(&greq, optval, sizeof(greq)))
866 psin = (struct sockaddr_in *)&greq.gr_group;
867 if (psin->sin_family != AF_INET)
869 mreq.imr_multiaddr = psin->sin_addr;
870 mreq.imr_ifindex = greq.gr_interface;
872 if (optname == MCAST_JOIN_GROUP)
873 return ip_mc_join_group(sk, &mreq);
874 return ip_mc_leave_group(sk, &mreq);
877 static int do_ip_setsockopt(struct sock *sk, int level,
878 int optname, char __user *optval, unsigned int optlen)
880 struct inet_sock *inet = inet_sk(sk);
881 struct net *net = sock_net(sk);
883 bool needs_rtnl = setsockopt_needs_rtnl(optname);
894 case IP_MTU_DISCOVER:
896 case IP_ROUTER_ALERT:
902 case IP_BIND_ADDRESS_NO_PORT:
904 case IP_MULTICAST_TTL:
905 case IP_MULTICAST_ALL:
906 case IP_MULTICAST_LOOP:
907 case IP_RECVORIGDSTADDR:
909 case IP_RECVFRAGSIZE:
910 case IP_RECVERR_RFC4884:
911 if (optlen >= sizeof(int)) {
912 if (get_user(val, (int __user *) optval))
914 } else if (optlen >= sizeof(char)) {
917 if (get_user(ucval, (unsigned char __user *) optval))
923 /* If optlen==0, it is equivalent to val == 0 */
925 if (optname == IP_ROUTER_ALERT)
926 return ip_ra_control(sk, val ? 1 : 0, NULL);
927 if (ip_mroute_opt(optname))
928 return ip_mroute_setsockopt(sk, optname, optval, optlen);
938 struct ip_options_rcu *old, *opt = NULL;
942 err = ip_options_get_from_user(sock_net(sk), &opt,
946 old = rcu_dereference_protected(inet->inet_opt,
947 lockdep_sock_is_held(sk));
949 struct inet_connection_sock *icsk = inet_csk(sk);
950 #if IS_ENABLED(CONFIG_IPV6)
951 if (sk->sk_family == PF_INET ||
952 (!((1 << sk->sk_state) &
953 (TCPF_LISTEN | TCPF_CLOSE)) &&
954 inet->inet_daddr != LOOPBACK4_IPV6)) {
957 icsk->icsk_ext_hdr_len -= old->opt.optlen;
959 icsk->icsk_ext_hdr_len += opt->opt.optlen;
960 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
961 #if IS_ENABLED(CONFIG_IPV6)
965 rcu_assign_pointer(inet->inet_opt, opt);
972 inet->cmsg_flags |= IP_CMSG_PKTINFO;
974 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
978 inet->cmsg_flags |= IP_CMSG_TTL;
980 inet->cmsg_flags &= ~IP_CMSG_TTL;
984 inet->cmsg_flags |= IP_CMSG_TOS;
986 inet->cmsg_flags &= ~IP_CMSG_TOS;
990 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
992 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
996 inet->cmsg_flags |= IP_CMSG_RETOPTS;
998 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
1002 inet->cmsg_flags |= IP_CMSG_PASSSEC;
1004 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
1006 case IP_RECVORIGDSTADDR:
1008 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
1010 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
1014 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
1015 inet_inc_convert_csum(sk);
1016 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
1019 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
1020 inet_dec_convert_csum(sk);
1021 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
1025 case IP_RECVFRAGSIZE:
1026 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
1029 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
1031 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
1033 case IP_TOS: /* This sets both TOS and Precedence */
1034 __ip_sock_set_tos(sk, val);
1039 if (val != -1 && (val < 1 || val > 255))
1044 if (sk->sk_type != SOCK_RAW) {
1048 inet->hdrincl = val ? 1 : 0;
1051 if (sk->sk_type != SOCK_RAW) {
1055 inet->nodefrag = val ? 1 : 0;
1057 case IP_BIND_ADDRESS_NO_PORT:
1058 inet->bind_address_no_port = val ? 1 : 0;
1060 case IP_MTU_DISCOVER:
1061 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1063 inet->pmtudisc = val;
1066 inet->recverr = !!val;
1068 skb_queue_purge(&sk->sk_error_queue);
1070 case IP_RECVERR_RFC4884:
1071 if (val < 0 || val > 1)
1073 inet->recverr_rfc4884 = !!val;
1075 case IP_MULTICAST_TTL:
1076 if (sk->sk_type == SOCK_STREAM)
1082 if (val < 0 || val > 255)
1086 case IP_MULTICAST_LOOP:
1089 inet->mc_loop = !!val;
1093 struct net_device *dev = NULL;
1097 if (optlen != sizeof(int))
1100 ifindex = (__force int)ntohl((__force __be32)val);
1107 dev = dev_get_by_index(sock_net(sk), ifindex);
1108 err = -EADDRNOTAVAIL;
1112 midx = l3mdev_master_ifindex(dev);
1116 if (sk->sk_bound_dev_if &&
1117 (!midx || midx != sk->sk_bound_dev_if))
1120 inet->uc_index = ifindex;
1124 case IP_MULTICAST_IF:
1126 struct ip_mreqn mreq;
1127 struct net_device *dev = NULL;
1130 if (sk->sk_type == SOCK_STREAM)
1133 * Check the arguments are allowable
1136 if (optlen < sizeof(struct in_addr))
1140 if (optlen >= sizeof(struct ip_mreqn)) {
1141 if (copy_from_user(&mreq, optval, sizeof(mreq)))
1144 memset(&mreq, 0, sizeof(mreq));
1145 if (optlen >= sizeof(struct ip_mreq)) {
1146 if (copy_from_user(&mreq, optval,
1147 sizeof(struct ip_mreq)))
1149 } else if (optlen >= sizeof(struct in_addr)) {
1150 if (copy_from_user(&mreq.imr_address, optval,
1151 sizeof(struct in_addr)))
1156 if (!mreq.imr_ifindex) {
1157 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1163 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1165 mreq.imr_ifindex = dev->ifindex;
1167 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1170 err = -EADDRNOTAVAIL;
1174 midx = l3mdev_master_ifindex(dev);
1179 if (sk->sk_bound_dev_if &&
1180 mreq.imr_ifindex != sk->sk_bound_dev_if &&
1181 (!midx || midx != sk->sk_bound_dev_if))
1184 inet->mc_index = mreq.imr_ifindex;
1185 inet->mc_addr = mreq.imr_address.s_addr;
1190 case IP_ADD_MEMBERSHIP:
1191 case IP_DROP_MEMBERSHIP:
1193 struct ip_mreqn mreq;
1196 if (inet_sk(sk)->is_icsk)
1199 if (optlen < sizeof(struct ip_mreq))
1202 if (optlen >= sizeof(struct ip_mreqn)) {
1203 if (copy_from_user(&mreq, optval, sizeof(mreq)))
1206 memset(&mreq, 0, sizeof(mreq));
1207 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
1211 if (optname == IP_ADD_MEMBERSHIP)
1212 err = ip_mc_join_group(sk, &mreq);
1214 err = ip_mc_leave_group(sk, &mreq);
1219 struct ip_msfilter *msf;
1221 if (optlen < IP_MSFILTER_SIZE(0))
1223 if (optlen > sysctl_optmem_max) {
1227 msf = memdup_user(optval, optlen);
1232 /* numsrc >= (1G-4) overflow in 32 bits */
1233 if (msf->imsf_numsrc >= 0x3ffffffcU ||
1234 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1239 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1244 err = ip_mc_msfilter(sk, msf, 0);
1248 case IP_BLOCK_SOURCE:
1249 case IP_UNBLOCK_SOURCE:
1250 case IP_ADD_SOURCE_MEMBERSHIP:
1251 case IP_DROP_SOURCE_MEMBERSHIP:
1253 struct ip_mreq_source mreqs;
1256 if (optlen != sizeof(struct ip_mreq_source))
1258 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
1262 if (optname == IP_BLOCK_SOURCE) {
1263 omode = MCAST_EXCLUDE;
1265 } else if (optname == IP_UNBLOCK_SOURCE) {
1266 omode = MCAST_EXCLUDE;
1268 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1269 struct ip_mreqn mreq;
1271 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1272 mreq.imr_address.s_addr = mreqs.imr_interface;
1273 mreq.imr_ifindex = 0;
1274 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1275 if (err && err != -EADDRINUSE)
1277 omode = MCAST_INCLUDE;
1279 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1280 omode = MCAST_INCLUDE;
1283 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1286 case MCAST_JOIN_GROUP:
1287 case MCAST_LEAVE_GROUP:
1288 if (in_compat_syscall())
1289 err = compat_ip_mcast_join_leave(sk, optname, optval,
1292 err = ip_mcast_join_leave(sk, optname, optval, optlen);
1294 case MCAST_JOIN_SOURCE_GROUP:
1295 case MCAST_LEAVE_SOURCE_GROUP:
1296 case MCAST_BLOCK_SOURCE:
1297 case MCAST_UNBLOCK_SOURCE:
1298 err = do_mcast_group_source(sk, optname, optval, optlen);
1300 case MCAST_MSFILTER:
1301 if (in_compat_syscall())
1302 err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1304 err = ip_set_mcast_msfilter(sk, optval, optlen);
1306 case IP_MULTICAST_ALL:
1309 if (val != 0 && val != 1)
1317 inet->freebind = !!val;
1320 case IP_IPSEC_POLICY:
1321 case IP_XFRM_POLICY:
1323 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1325 err = xfrm_user_policy(sk, optname, optval, optlen);
1328 case IP_TRANSPARENT:
1329 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1330 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1336 inet->transparent = !!val;
1342 if (val < 0 || val > 255)
1344 inet->min_ttl = val;
1364 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1368 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1369 * destination in skb->cb[] before dst drop.
1370 * This way, receiver doesn't make cache line misses to read rtable.
1372 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1374 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1375 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1378 if (prepare && skb_rtable(skb)) {
1379 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1380 * which has interface index (iif) as the first member of the
1381 * underlying inet{6}_skb_parm struct. This code then overlays
1382 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1383 * element so the iif is picked up from the prior IPCB. If iif
1384 * is the loopback interface, then return the sending interface
1385 * (e.g., process binds socket to eth0 for Tx which is
1386 * redirected to loopback in the rtable/dst).
1388 struct rtable *rt = skb_rtable(skb);
1389 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1391 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1392 pktinfo->ipi_ifindex = inet_iif(skb);
1393 else if (l3slave && rt && rt->rt_iif)
1394 pktinfo->ipi_ifindex = rt->rt_iif;
1396 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1398 pktinfo->ipi_ifindex = 0;
1399 pktinfo->ipi_spec_dst.s_addr = 0;
1404 int ip_setsockopt(struct sock *sk, int level,
1405 int optname, char __user *optval, unsigned int optlen)
1409 if (level != SOL_IP)
1410 return -ENOPROTOOPT;
1412 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1413 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1414 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1415 optname < BPFILTER_IPT_SET_MAX)
1416 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1418 #ifdef CONFIG_NETFILTER
1419 /* we need to exclude all possible ENOPROTOOPTs except default case */
1420 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1421 optname != IP_IPSEC_POLICY &&
1422 optname != IP_XFRM_POLICY &&
1423 !ip_mroute_opt(optname))
1424 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1428 EXPORT_SYMBOL(ip_setsockopt);
1431 * Get the options. Note for future reference. The GET of IP options gets
1432 * the _received_ ones. The set sets the _sent_ ones.
1435 static bool getsockopt_needs_rtnl(int optname)
1439 case MCAST_MSFILTER:
1445 static int ip_get_mcast_msfilter(struct sock *sk, void __user *optval,
1446 int __user *optlen, int len)
1448 const int size0 = offsetof(struct group_filter, gf_slist);
1449 struct group_filter __user *p = optval;
1450 struct group_filter gsf;
1456 if (copy_from_user(&gsf, p, size0))
1459 num = gsf.gf_numsrc;
1460 err = ip_mc_gsfget(sk, &gsf, p->gf_slist);
1463 if (gsf.gf_numsrc < num)
1464 num = gsf.gf_numsrc;
1465 if (put_user(GROUP_FILTER_SIZE(num), optlen) ||
1466 copy_to_user(p, &gsf, size0))
1471 static int compat_ip_get_mcast_msfilter(struct sock *sk, void __user *optval,
1472 int __user *optlen, int len)
1474 const int size0 = offsetof(struct compat_group_filter, gf_slist);
1475 struct compat_group_filter __user *p = optval;
1476 struct compat_group_filter gf32;
1477 struct group_filter gf;
1483 if (copy_from_user(&gf32, p, size0))
1486 gf.gf_interface = gf32.gf_interface;
1487 gf.gf_fmode = gf32.gf_fmode;
1488 num = gf.gf_numsrc = gf32.gf_numsrc;
1489 gf.gf_group = gf32.gf_group;
1491 err = ip_mc_gsfget(sk, &gf, p->gf_slist);
1494 if (gf.gf_numsrc < num)
1496 len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1497 if (put_user(len, optlen) ||
1498 put_user(gf.gf_fmode, &p->gf_fmode) ||
1499 put_user(gf.gf_numsrc, &p->gf_numsrc))
1504 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1505 char __user *optval, int __user *optlen)
1507 struct inet_sock *inet = inet_sk(sk);
1508 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1512 if (level != SOL_IP)
1515 if (ip_mroute_opt(optname))
1516 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1518 if (get_user(len, optlen))
1530 unsigned char optbuf[sizeof(struct ip_options)+40];
1531 struct ip_options *opt = (struct ip_options *)optbuf;
1532 struct ip_options_rcu *inet_opt;
1534 inet_opt = rcu_dereference_protected(inet->inet_opt,
1535 lockdep_sock_is_held(sk));
1538 memcpy(optbuf, &inet_opt->opt,
1539 sizeof(struct ip_options) +
1540 inet_opt->opt.optlen);
1543 if (opt->optlen == 0)
1544 return put_user(0, optlen);
1546 ip_options_undo(opt);
1548 len = min_t(unsigned int, len, opt->optlen);
1549 if (put_user(len, optlen))
1551 if (copy_to_user(optval, opt->__data, len))
1556 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1559 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1562 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1565 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1568 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1571 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1573 case IP_RECVORIGDSTADDR:
1574 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1577 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1579 case IP_RECVFRAGSIZE:
1580 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1587 struct net *net = sock_net(sk);
1588 val = (inet->uc_ttl == -1 ?
1589 net->ipv4.sysctl_ip_default_ttl :
1594 val = inet->hdrincl;
1597 val = inet->nodefrag;
1599 case IP_BIND_ADDRESS_NO_PORT:
1600 val = inet->bind_address_no_port;
1602 case IP_MTU_DISCOVER:
1603 val = inet->pmtudisc;
1607 struct dst_entry *dst;
1609 dst = sk_dst_get(sk);
1621 val = inet->recverr;
1623 case IP_RECVERR_RFC4884:
1624 val = inet->recverr_rfc4884;
1626 case IP_MULTICAST_TTL:
1629 case IP_MULTICAST_LOOP:
1630 val = inet->mc_loop;
1633 val = (__force int)htonl((__u32) inet->uc_index);
1635 case IP_MULTICAST_IF:
1637 struct in_addr addr;
1638 len = min_t(unsigned int, len, sizeof(struct in_addr));
1639 addr.s_addr = inet->mc_addr;
1642 if (put_user(len, optlen))
1644 if (copy_to_user(optval, &addr, len))
1650 struct ip_msfilter msf;
1652 if (len < IP_MSFILTER_SIZE(0)) {
1656 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1660 err = ip_mc_msfget(sk, &msf,
1661 (struct ip_msfilter __user *)optval, optlen);
1664 case MCAST_MSFILTER:
1665 if (in_compat_syscall())
1666 err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1669 err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1671 case IP_MULTICAST_ALL:
1680 if (sk->sk_type != SOCK_STREAM)
1681 return -ENOPROTOOPT;
1683 msg.msg_control_is_user = true;
1684 msg.msg_control_user = optval;
1685 msg.msg_controllen = len;
1686 msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1688 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1689 struct in_pktinfo info;
1691 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1692 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1693 info.ipi_ifindex = inet->mc_index;
1694 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1696 if (inet->cmsg_flags & IP_CMSG_TTL) {
1697 int hlim = inet->mc_ttl;
1698 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1700 if (inet->cmsg_flags & IP_CMSG_TOS) {
1701 int tos = inet->rcv_tos;
1702 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1704 len -= msg.msg_controllen;
1705 return put_user(len, optlen);
1708 val = inet->freebind;
1710 case IP_TRANSPARENT:
1711 val = inet->transparent;
1714 val = inet->min_ttl;
1718 return -ENOPROTOOPT;
1722 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1723 unsigned char ucval = (unsigned char)val;
1725 if (put_user(len, optlen))
1727 if (copy_to_user(optval, &ucval, 1))
1730 len = min_t(unsigned int, sizeof(int), len);
1731 if (put_user(len, optlen))
1733 if (copy_to_user(optval, &val, len))
1745 int ip_getsockopt(struct sock *sk, int level,
1746 int optname, char __user *optval, int __user *optlen)
1750 err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1752 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1753 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1754 optname < BPFILTER_IPT_GET_MAX)
1755 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1757 #ifdef CONFIG_NETFILTER
1758 /* we need to exclude all possible ENOPROTOOPTs except default case */
1759 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1760 !ip_mroute_opt(optname)) {
1763 if (get_user(len, optlen))
1766 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1768 err = put_user(len, optlen);
1774 EXPORT_SYMBOL(ip_getsockopt);