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;
563 static void __ip_sock_set_tos(struct sock *sk, int val)
565 if (sk->sk_type == SOCK_STREAM) {
566 val &= ~INET_ECN_MASK;
567 val |= inet_sk(sk)->tos & INET_ECN_MASK;
569 if (inet_sk(sk)->tos != val) {
570 inet_sk(sk)->tos = val;
571 sk->sk_priority = rt_tos2priority(val);
576 void ip_sock_set_tos(struct sock *sk, int val)
579 __ip_sock_set_tos(sk, val);
582 EXPORT_SYMBOL(ip_sock_set_tos);
584 void ip_sock_set_freebind(struct sock *sk)
587 inet_sk(sk)->freebind = true;
590 EXPORT_SYMBOL(ip_sock_set_freebind);
592 void ip_sock_set_recverr(struct sock *sk)
595 inet_sk(sk)->recverr = true;
598 EXPORT_SYMBOL(ip_sock_set_recverr);
600 int ip_sock_set_mtu_discover(struct sock *sk, int val)
602 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
605 inet_sk(sk)->pmtudisc = val;
609 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
611 void ip_sock_set_pktinfo(struct sock *sk)
614 inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
617 EXPORT_SYMBOL(ip_sock_set_pktinfo);
620 * Socket option code for IP. This is the end of the line after any
621 * TCP,UDP etc options on an IP socket.
623 static bool setsockopt_needs_rtnl(int optname)
626 case IP_ADD_MEMBERSHIP:
627 case IP_ADD_SOURCE_MEMBERSHIP:
628 case IP_BLOCK_SOURCE:
629 case IP_DROP_MEMBERSHIP:
630 case IP_DROP_SOURCE_MEMBERSHIP:
632 case IP_UNBLOCK_SOURCE:
633 case MCAST_BLOCK_SOURCE:
635 case MCAST_JOIN_GROUP:
636 case MCAST_JOIN_SOURCE_GROUP:
637 case MCAST_LEAVE_GROUP:
638 case MCAST_LEAVE_SOURCE_GROUP:
639 case MCAST_UNBLOCK_SOURCE:
645 static int set_mcast_msfilter(struct sock *sk, int ifindex,
646 int numsrc, int fmode,
647 struct sockaddr_storage *group,
648 struct sockaddr_storage *list)
650 int msize = IP_MSFILTER_SIZE(numsrc);
651 struct ip_msfilter *msf;
652 struct sockaddr_in *psin;
655 msf = kmalloc(msize, GFP_KERNEL);
659 psin = (struct sockaddr_in *)group;
660 if (psin->sin_family != AF_INET)
662 msf->imsf_multiaddr = psin->sin_addr.s_addr;
663 msf->imsf_interface = 0;
664 msf->imsf_fmode = fmode;
665 msf->imsf_numsrc = numsrc;
666 for (i = 0; i < numsrc; ++i) {
667 psin = (struct sockaddr_in *)&list[i];
669 if (psin->sin_family != AF_INET)
671 msf->imsf_slist[i] = psin->sin_addr.s_addr;
673 err = ip_mc_msfilter(sk, msf, ifindex);
679 return -EADDRNOTAVAIL;
682 static int copy_group_source_from_user(struct group_source_req *greqs,
683 void __user *optval, int optlen)
685 if (in_compat_syscall()) {
686 struct compat_group_source_req gr32;
688 if (optlen != sizeof(gr32))
690 if (copy_from_user(&gr32, optval, sizeof(gr32)))
692 greqs->gsr_interface = gr32.gsr_interface;
693 greqs->gsr_group = gr32.gsr_group;
694 greqs->gsr_source = gr32.gsr_source;
696 if (optlen != sizeof(*greqs))
698 if (copy_from_user(greqs, optval, sizeof(*greqs)))
705 static int do_mcast_group_source(struct sock *sk, int optname,
706 void __user *optval, int optlen)
708 struct group_source_req greqs;
709 struct ip_mreq_source mreqs;
710 struct sockaddr_in *psin;
713 err = copy_group_source_from_user(&greqs, optval, optlen);
717 if (greqs.gsr_group.ss_family != AF_INET ||
718 greqs.gsr_source.ss_family != AF_INET)
719 return -EADDRNOTAVAIL;
721 psin = (struct sockaddr_in *)&greqs.gsr_group;
722 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
723 psin = (struct sockaddr_in *)&greqs.gsr_source;
724 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
725 mreqs.imr_interface = 0; /* use index for mc_source */
727 if (optname == MCAST_BLOCK_SOURCE) {
728 omode = MCAST_EXCLUDE;
730 } else if (optname == MCAST_UNBLOCK_SOURCE) {
731 omode = MCAST_EXCLUDE;
733 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
734 struct ip_mreqn mreq;
736 psin = (struct sockaddr_in *)&greqs.gsr_group;
737 mreq.imr_multiaddr = psin->sin_addr;
738 mreq.imr_address.s_addr = 0;
739 mreq.imr_ifindex = greqs.gsr_interface;
740 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
741 if (err && err != -EADDRINUSE)
743 greqs.gsr_interface = mreq.imr_ifindex;
744 omode = MCAST_INCLUDE;
746 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
747 omode = MCAST_INCLUDE;
750 return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
753 static int ip_set_mcast_msfilter(struct sock *sk, void __user *optval,
756 struct group_filter *gsf = NULL;
759 if (optlen < GROUP_FILTER_SIZE(0))
761 if (optlen > sysctl_optmem_max)
764 gsf = memdup_user(optval, optlen);
768 /* numsrc >= (4G-140)/128 overflow in 32 bits */
770 if (gsf->gf_numsrc >= 0x1ffffff ||
771 gsf->gf_numsrc > sock_net(sk)->ipv4.sysctl_igmp_max_msf)
775 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
778 err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
779 gsf->gf_fmode, &gsf->gf_group, gsf->gf_slist);
785 static int compat_ip_set_mcast_msfilter(struct sock *sk, void __user *optval,
788 const int size0 = offsetof(struct compat_group_filter, gf_slist);
789 struct compat_group_filter *gf32;
796 if (optlen > sysctl_optmem_max - 4)
799 p = kmalloc(optlen + 4, GFP_KERNEL);
802 gf32 = p + 4; /* we want ->gf_group and ->gf_slist aligned */
805 if (copy_from_user(gf32, optval, optlen))
808 /* numsrc >= (4G-140)/128 overflow in 32 bits */
815 if (offsetof(struct compat_group_filter, gf_slist[n]) > optlen)
818 /* numsrc >= (4G-140)/128 overflow in 32 bits */
820 if (n > sock_net(sk)->ipv4.sysctl_igmp_max_msf)
822 err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
823 &gf32->gf_group, gf32->gf_slist);
829 static int ip_mcast_join_leave(struct sock *sk, int optname,
830 void __user *optval, int optlen)
832 struct ip_mreqn mreq = { };
833 struct sockaddr_in *psin;
834 struct group_req greq;
836 if (optlen < sizeof(struct group_req))
838 if (copy_from_user(&greq, optval, sizeof(greq)))
841 psin = (struct sockaddr_in *)&greq.gr_group;
842 if (psin->sin_family != AF_INET)
844 mreq.imr_multiaddr = psin->sin_addr;
845 mreq.imr_ifindex = greq.gr_interface;
846 if (optname == MCAST_JOIN_GROUP)
847 return ip_mc_join_group(sk, &mreq);
848 return ip_mc_leave_group(sk, &mreq);
851 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
852 void __user *optval, int optlen)
854 struct compat_group_req greq;
855 struct ip_mreqn mreq = { };
856 struct sockaddr_in *psin;
858 if (optlen < sizeof(struct compat_group_req))
860 if (copy_from_user(&greq, optval, sizeof(greq)))
863 psin = (struct sockaddr_in *)&greq.gr_group;
864 if (psin->sin_family != AF_INET)
866 mreq.imr_multiaddr = psin->sin_addr;
867 mreq.imr_ifindex = greq.gr_interface;
869 if (optname == MCAST_JOIN_GROUP)
870 return ip_mc_join_group(sk, &mreq);
871 return ip_mc_leave_group(sk, &mreq);
874 static int do_ip_setsockopt(struct sock *sk, int level,
875 int optname, char __user *optval, unsigned int optlen)
877 struct inet_sock *inet = inet_sk(sk);
878 struct net *net = sock_net(sk);
880 bool needs_rtnl = setsockopt_needs_rtnl(optname);
891 case IP_MTU_DISCOVER:
893 case IP_ROUTER_ALERT:
899 case IP_BIND_ADDRESS_NO_PORT:
901 case IP_MULTICAST_TTL:
902 case IP_MULTICAST_ALL:
903 case IP_MULTICAST_LOOP:
904 case IP_RECVORIGDSTADDR:
906 case IP_RECVFRAGSIZE:
907 if (optlen >= sizeof(int)) {
908 if (get_user(val, (int __user *) optval))
910 } else if (optlen >= sizeof(char)) {
913 if (get_user(ucval, (unsigned char __user *) optval))
919 /* If optlen==0, it is equivalent to val == 0 */
921 if (optname == IP_ROUTER_ALERT)
922 return ip_ra_control(sk, val ? 1 : 0, NULL);
923 if (ip_mroute_opt(optname))
924 return ip_mroute_setsockopt(sk, optname, optval, optlen);
934 struct ip_options_rcu *old, *opt = NULL;
938 err = ip_options_get_from_user(sock_net(sk), &opt,
942 old = rcu_dereference_protected(inet->inet_opt,
943 lockdep_sock_is_held(sk));
945 struct inet_connection_sock *icsk = inet_csk(sk);
946 #if IS_ENABLED(CONFIG_IPV6)
947 if (sk->sk_family == PF_INET ||
948 (!((1 << sk->sk_state) &
949 (TCPF_LISTEN | TCPF_CLOSE)) &&
950 inet->inet_daddr != LOOPBACK4_IPV6)) {
953 icsk->icsk_ext_hdr_len -= old->opt.optlen;
955 icsk->icsk_ext_hdr_len += opt->opt.optlen;
956 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
957 #if IS_ENABLED(CONFIG_IPV6)
961 rcu_assign_pointer(inet->inet_opt, opt);
968 inet->cmsg_flags |= IP_CMSG_PKTINFO;
970 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
974 inet->cmsg_flags |= IP_CMSG_TTL;
976 inet->cmsg_flags &= ~IP_CMSG_TTL;
980 inet->cmsg_flags |= IP_CMSG_TOS;
982 inet->cmsg_flags &= ~IP_CMSG_TOS;
986 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
988 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
992 inet->cmsg_flags |= IP_CMSG_RETOPTS;
994 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
998 inet->cmsg_flags |= IP_CMSG_PASSSEC;
1000 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
1002 case IP_RECVORIGDSTADDR:
1004 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
1006 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
1010 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
1011 inet_inc_convert_csum(sk);
1012 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
1015 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
1016 inet_dec_convert_csum(sk);
1017 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
1021 case IP_RECVFRAGSIZE:
1022 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
1025 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE;
1027 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
1029 case IP_TOS: /* This sets both TOS and Precedence */
1030 __ip_sock_set_tos(sk, val);
1035 if (val != -1 && (val < 1 || val > 255))
1040 if (sk->sk_type != SOCK_RAW) {
1044 inet->hdrincl = val ? 1 : 0;
1047 if (sk->sk_type != SOCK_RAW) {
1051 inet->nodefrag = val ? 1 : 0;
1053 case IP_BIND_ADDRESS_NO_PORT:
1054 inet->bind_address_no_port = val ? 1 : 0;
1056 case IP_MTU_DISCOVER:
1057 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1059 inet->pmtudisc = val;
1062 inet->recverr = !!val;
1064 skb_queue_purge(&sk->sk_error_queue);
1066 case IP_MULTICAST_TTL:
1067 if (sk->sk_type == SOCK_STREAM)
1073 if (val < 0 || val > 255)
1077 case IP_MULTICAST_LOOP:
1080 inet->mc_loop = !!val;
1084 struct net_device *dev = NULL;
1088 if (optlen != sizeof(int))
1091 ifindex = (__force int)ntohl((__force __be32)val);
1098 dev = dev_get_by_index(sock_net(sk), ifindex);
1099 err = -EADDRNOTAVAIL;
1103 midx = l3mdev_master_ifindex(dev);
1107 if (sk->sk_bound_dev_if &&
1108 (!midx || midx != sk->sk_bound_dev_if))
1111 inet->uc_index = ifindex;
1115 case IP_MULTICAST_IF:
1117 struct ip_mreqn mreq;
1118 struct net_device *dev = NULL;
1121 if (sk->sk_type == SOCK_STREAM)
1124 * Check the arguments are allowable
1127 if (optlen < sizeof(struct in_addr))
1131 if (optlen >= sizeof(struct ip_mreqn)) {
1132 if (copy_from_user(&mreq, optval, sizeof(mreq)))
1135 memset(&mreq, 0, sizeof(mreq));
1136 if (optlen >= sizeof(struct ip_mreq)) {
1137 if (copy_from_user(&mreq, optval,
1138 sizeof(struct ip_mreq)))
1140 } else if (optlen >= sizeof(struct in_addr)) {
1141 if (copy_from_user(&mreq.imr_address, optval,
1142 sizeof(struct in_addr)))
1147 if (!mreq.imr_ifindex) {
1148 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1154 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1156 mreq.imr_ifindex = dev->ifindex;
1158 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1161 err = -EADDRNOTAVAIL;
1165 midx = l3mdev_master_ifindex(dev);
1170 if (sk->sk_bound_dev_if &&
1171 mreq.imr_ifindex != sk->sk_bound_dev_if &&
1172 (!midx || midx != sk->sk_bound_dev_if))
1175 inet->mc_index = mreq.imr_ifindex;
1176 inet->mc_addr = mreq.imr_address.s_addr;
1181 case IP_ADD_MEMBERSHIP:
1182 case IP_DROP_MEMBERSHIP:
1184 struct ip_mreqn mreq;
1187 if (inet_sk(sk)->is_icsk)
1190 if (optlen < sizeof(struct ip_mreq))
1193 if (optlen >= sizeof(struct ip_mreqn)) {
1194 if (copy_from_user(&mreq, optval, sizeof(mreq)))
1197 memset(&mreq, 0, sizeof(mreq));
1198 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
1202 if (optname == IP_ADD_MEMBERSHIP)
1203 err = ip_mc_join_group(sk, &mreq);
1205 err = ip_mc_leave_group(sk, &mreq);
1210 struct ip_msfilter *msf;
1212 if (optlen < IP_MSFILTER_SIZE(0))
1214 if (optlen > sysctl_optmem_max) {
1218 msf = memdup_user(optval, optlen);
1223 /* numsrc >= (1G-4) overflow in 32 bits */
1224 if (msf->imsf_numsrc >= 0x3ffffffcU ||
1225 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1230 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1235 err = ip_mc_msfilter(sk, msf, 0);
1239 case IP_BLOCK_SOURCE:
1240 case IP_UNBLOCK_SOURCE:
1241 case IP_ADD_SOURCE_MEMBERSHIP:
1242 case IP_DROP_SOURCE_MEMBERSHIP:
1244 struct ip_mreq_source mreqs;
1247 if (optlen != sizeof(struct ip_mreq_source))
1249 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
1253 if (optname == IP_BLOCK_SOURCE) {
1254 omode = MCAST_EXCLUDE;
1256 } else if (optname == IP_UNBLOCK_SOURCE) {
1257 omode = MCAST_EXCLUDE;
1259 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1260 struct ip_mreqn mreq;
1262 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1263 mreq.imr_address.s_addr = mreqs.imr_interface;
1264 mreq.imr_ifindex = 0;
1265 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1266 if (err && err != -EADDRINUSE)
1268 omode = MCAST_INCLUDE;
1270 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1271 omode = MCAST_INCLUDE;
1274 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1277 case MCAST_JOIN_GROUP:
1278 case MCAST_LEAVE_GROUP:
1279 if (in_compat_syscall())
1280 err = compat_ip_mcast_join_leave(sk, optname, optval,
1283 err = ip_mcast_join_leave(sk, optname, optval, optlen);
1285 case MCAST_JOIN_SOURCE_GROUP:
1286 case MCAST_LEAVE_SOURCE_GROUP:
1287 case MCAST_BLOCK_SOURCE:
1288 case MCAST_UNBLOCK_SOURCE:
1289 err = do_mcast_group_source(sk, optname, optval, optlen);
1291 case MCAST_MSFILTER:
1292 if (in_compat_syscall())
1293 err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1295 err = ip_set_mcast_msfilter(sk, optval, optlen);
1297 case IP_MULTICAST_ALL:
1300 if (val != 0 && val != 1)
1308 inet->freebind = !!val;
1311 case IP_IPSEC_POLICY:
1312 case IP_XFRM_POLICY:
1314 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1316 err = xfrm_user_policy(sk, optname, optval, optlen);
1319 case IP_TRANSPARENT:
1320 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1321 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1327 inet->transparent = !!val;
1333 if (val < 0 || val > 255)
1335 inet->min_ttl = val;
1355 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1359 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1360 * destination in skb->cb[] before dst drop.
1361 * This way, receiver doesn't make cache line misses to read rtable.
1363 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1365 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1366 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1369 if (prepare && skb_rtable(skb)) {
1370 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1371 * which has interface index (iif) as the first member of the
1372 * underlying inet{6}_skb_parm struct. This code then overlays
1373 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1374 * element so the iif is picked up from the prior IPCB. If iif
1375 * is the loopback interface, then return the sending interface
1376 * (e.g., process binds socket to eth0 for Tx which is
1377 * redirected to loopback in the rtable/dst).
1379 struct rtable *rt = skb_rtable(skb);
1380 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1382 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1383 pktinfo->ipi_ifindex = inet_iif(skb);
1384 else if (l3slave && rt && rt->rt_iif)
1385 pktinfo->ipi_ifindex = rt->rt_iif;
1387 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1389 pktinfo->ipi_ifindex = 0;
1390 pktinfo->ipi_spec_dst.s_addr = 0;
1395 int ip_setsockopt(struct sock *sk, int level,
1396 int optname, char __user *optval, unsigned int optlen)
1400 if (level != SOL_IP)
1401 return -ENOPROTOOPT;
1403 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1404 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1405 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1406 optname < BPFILTER_IPT_SET_MAX)
1407 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1409 #ifdef CONFIG_NETFILTER
1410 /* we need to exclude all possible ENOPROTOOPTs except default case */
1411 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1412 optname != IP_IPSEC_POLICY &&
1413 optname != IP_XFRM_POLICY &&
1414 !ip_mroute_opt(optname))
1415 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1419 EXPORT_SYMBOL(ip_setsockopt);
1422 * Get the options. Note for future reference. The GET of IP options gets
1423 * the _received_ ones. The set sets the _sent_ ones.
1426 static bool getsockopt_needs_rtnl(int optname)
1430 case MCAST_MSFILTER:
1436 static int ip_get_mcast_msfilter(struct sock *sk, void __user *optval,
1437 int __user *optlen, int len)
1439 const int size0 = offsetof(struct group_filter, gf_slist);
1440 struct group_filter __user *p = optval;
1441 struct group_filter gsf;
1447 if (copy_from_user(&gsf, p, size0))
1450 num = gsf.gf_numsrc;
1451 err = ip_mc_gsfget(sk, &gsf, p->gf_slist);
1454 if (gsf.gf_numsrc < num)
1455 num = gsf.gf_numsrc;
1456 if (put_user(GROUP_FILTER_SIZE(num), optlen) ||
1457 copy_to_user(p, &gsf, size0))
1462 static int compat_ip_get_mcast_msfilter(struct sock *sk, void __user *optval,
1463 int __user *optlen, int len)
1465 const int size0 = offsetof(struct compat_group_filter, gf_slist);
1466 struct compat_group_filter __user *p = optval;
1467 struct compat_group_filter gf32;
1468 struct group_filter gf;
1474 if (copy_from_user(&gf32, p, size0))
1477 gf.gf_interface = gf32.gf_interface;
1478 gf.gf_fmode = gf32.gf_fmode;
1479 num = gf.gf_numsrc = gf32.gf_numsrc;
1480 gf.gf_group = gf32.gf_group;
1482 err = ip_mc_gsfget(sk, &gf, p->gf_slist);
1485 if (gf.gf_numsrc < num)
1487 len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1488 if (put_user(len, optlen) ||
1489 put_user(gf.gf_fmode, &p->gf_fmode) ||
1490 put_user(gf.gf_numsrc, &p->gf_numsrc))
1495 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1496 char __user *optval, int __user *optlen)
1498 struct inet_sock *inet = inet_sk(sk);
1499 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1503 if (level != SOL_IP)
1506 if (ip_mroute_opt(optname))
1507 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1509 if (get_user(len, optlen))
1521 unsigned char optbuf[sizeof(struct ip_options)+40];
1522 struct ip_options *opt = (struct ip_options *)optbuf;
1523 struct ip_options_rcu *inet_opt;
1525 inet_opt = rcu_dereference_protected(inet->inet_opt,
1526 lockdep_sock_is_held(sk));
1529 memcpy(optbuf, &inet_opt->opt,
1530 sizeof(struct ip_options) +
1531 inet_opt->opt.optlen);
1534 if (opt->optlen == 0)
1535 return put_user(0, optlen);
1537 ip_options_undo(opt);
1539 len = min_t(unsigned int, len, opt->optlen);
1540 if (put_user(len, optlen))
1542 if (copy_to_user(optval, opt->__data, len))
1547 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1550 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1553 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1556 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1559 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1562 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1564 case IP_RECVORIGDSTADDR:
1565 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1568 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1570 case IP_RECVFRAGSIZE:
1571 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0;
1578 struct net *net = sock_net(sk);
1579 val = (inet->uc_ttl == -1 ?
1580 net->ipv4.sysctl_ip_default_ttl :
1585 val = inet->hdrincl;
1588 val = inet->nodefrag;
1590 case IP_BIND_ADDRESS_NO_PORT:
1591 val = inet->bind_address_no_port;
1593 case IP_MTU_DISCOVER:
1594 val = inet->pmtudisc;
1598 struct dst_entry *dst;
1600 dst = sk_dst_get(sk);
1612 val = inet->recverr;
1614 case IP_MULTICAST_TTL:
1617 case IP_MULTICAST_LOOP:
1618 val = inet->mc_loop;
1621 val = (__force int)htonl((__u32) inet->uc_index);
1623 case IP_MULTICAST_IF:
1625 struct in_addr addr;
1626 len = min_t(unsigned int, len, sizeof(struct in_addr));
1627 addr.s_addr = inet->mc_addr;
1630 if (put_user(len, optlen))
1632 if (copy_to_user(optval, &addr, len))
1638 struct ip_msfilter msf;
1640 if (len < IP_MSFILTER_SIZE(0)) {
1644 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1648 err = ip_mc_msfget(sk, &msf,
1649 (struct ip_msfilter __user *)optval, optlen);
1652 case MCAST_MSFILTER:
1653 if (in_compat_syscall())
1654 err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1657 err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1659 case IP_MULTICAST_ALL:
1668 if (sk->sk_type != SOCK_STREAM)
1669 return -ENOPROTOOPT;
1671 msg.msg_control_is_user = true;
1672 msg.msg_control_user = optval;
1673 msg.msg_controllen = len;
1674 msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1676 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1677 struct in_pktinfo info;
1679 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1680 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1681 info.ipi_ifindex = inet->mc_index;
1682 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1684 if (inet->cmsg_flags & IP_CMSG_TTL) {
1685 int hlim = inet->mc_ttl;
1686 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1688 if (inet->cmsg_flags & IP_CMSG_TOS) {
1689 int tos = inet->rcv_tos;
1690 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1692 len -= msg.msg_controllen;
1693 return put_user(len, optlen);
1696 val = inet->freebind;
1698 case IP_TRANSPARENT:
1699 val = inet->transparent;
1702 val = inet->min_ttl;
1706 return -ENOPROTOOPT;
1710 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1711 unsigned char ucval = (unsigned char)val;
1713 if (put_user(len, optlen))
1715 if (copy_to_user(optval, &ucval, 1))
1718 len = min_t(unsigned int, sizeof(int), len);
1719 if (put_user(len, optlen))
1721 if (copy_to_user(optval, &val, len))
1733 int ip_getsockopt(struct sock *sk, int level,
1734 int optname, char __user *optval, int __user *optlen)
1738 err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1740 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1741 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1742 optname < BPFILTER_IPT_GET_MAX)
1743 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1745 #ifdef CONFIG_NETFILTER
1746 /* we need to exclude all possible ENOPROTOOPTs except default case */
1747 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1748 !ip_mroute_opt(optname)) {
1751 if (get_user(len, optlen))
1754 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1756 err = put_user(len, optlen);
1762 EXPORT_SYMBOL(ip_getsockopt);