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 unsigned long flags = inet_cmsg_flags(inet_sk(sk));
179 /* Ordered by supposed usage frequency */
180 if (flags & IP_CMSG_PKTINFO) {
181 ip_cmsg_recv_pktinfo(msg, skb);
183 flags &= ~IP_CMSG_PKTINFO;
188 if (flags & IP_CMSG_TTL) {
189 ip_cmsg_recv_ttl(msg, skb);
191 flags &= ~IP_CMSG_TTL;
196 if (flags & IP_CMSG_TOS) {
197 ip_cmsg_recv_tos(msg, skb);
199 flags &= ~IP_CMSG_TOS;
204 if (flags & IP_CMSG_RECVOPTS) {
205 ip_cmsg_recv_opts(msg, skb);
207 flags &= ~IP_CMSG_RECVOPTS;
212 if (flags & IP_CMSG_RETOPTS) {
213 ip_cmsg_recv_retopts(sock_net(sk), msg, skb);
215 flags &= ~IP_CMSG_RETOPTS;
220 if (flags & IP_CMSG_PASSSEC) {
221 ip_cmsg_recv_security(msg, skb);
223 flags &= ~IP_CMSG_PASSSEC;
228 if (flags & IP_CMSG_ORIGDSTADDR) {
229 ip_cmsg_recv_dstaddr(msg, skb);
231 flags &= ~IP_CMSG_ORIGDSTADDR;
236 if (flags & IP_CMSG_CHECKSUM)
237 ip_cmsg_recv_checksum(msg, skb, tlen, offset);
239 if (flags & IP_CMSG_RECVFRAGSIZE)
240 ip_cmsg_recv_fragsize(msg, skb);
242 EXPORT_SYMBOL(ip_cmsg_recv_offset);
244 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
248 struct cmsghdr *cmsg;
249 struct net *net = sock_net(sk);
251 for_each_cmsghdr(cmsg, msg) {
252 if (!CMSG_OK(msg, cmsg))
254 #if IS_ENABLED(CONFIG_IPV6)
256 cmsg->cmsg_level == SOL_IPV6 &&
257 cmsg->cmsg_type == IPV6_PKTINFO) {
258 struct in6_pktinfo *src_info;
260 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
262 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
263 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
265 if (src_info->ipi6_ifindex)
266 ipc->oif = src_info->ipi6_ifindex;
267 ipc->addr = src_info->ipi6_addr.s6_addr32[3];
271 if (cmsg->cmsg_level == SOL_SOCKET) {
272 err = __sock_cmsg_send(sk, cmsg, &ipc->sockc);
278 if (cmsg->cmsg_level != SOL_IP)
280 switch (cmsg->cmsg_type) {
282 err = cmsg->cmsg_len - sizeof(struct cmsghdr);
284 /* Our caller is responsible for freeing ipc->opt */
285 err = ip_options_get(net, &ipc->opt,
286 KERNEL_SOCKPTR(CMSG_DATA(cmsg)),
287 err < 40 ? err : 40);
293 struct in_pktinfo *info;
294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
296 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
297 if (info->ipi_ifindex)
298 ipc->oif = info->ipi_ifindex;
299 ipc->addr = info->ipi_spec_dst.s_addr;
303 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
305 val = *(int *)CMSG_DATA(cmsg);
306 if (val < 1 || val > 255)
311 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
312 val = *(int *)CMSG_DATA(cmsg);
313 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
314 val = *(u8 *)CMSG_DATA(cmsg);
317 if (val < 0 || val > 255)
320 ipc->priority = rt_tos2priority(ipc->tos);
323 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
325 val = *(int *)CMSG_DATA(cmsg);
326 if (val < 1 || val > 255)
337 static void ip_ra_destroy_rcu(struct rcu_head *head)
339 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
341 sock_put(ra->saved_sk);
345 int ip_ra_control(struct sock *sk, unsigned char on,
346 void (*destructor)(struct sock *))
348 struct ip_ra_chain *ra, *new_ra;
349 struct ip_ra_chain __rcu **rap;
350 struct net *net = sock_net(sk);
352 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
355 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
359 mutex_lock(&net->ipv4.ra_mutex);
360 for (rap = &net->ipv4.ra_chain;
361 (ra = rcu_dereference_protected(*rap,
362 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
366 mutex_unlock(&net->ipv4.ra_mutex);
370 /* dont let ip_call_ra_chain() use sk again */
372 RCU_INIT_POINTER(*rap, ra->next);
373 mutex_unlock(&net->ipv4.ra_mutex);
378 * Delay sock_put(sk) and kfree(ra) after one rcu grace
379 * period. This guarantee ip_call_ra_chain() dont need
380 * to mess with socket refcounts.
383 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
388 mutex_unlock(&net->ipv4.ra_mutex);
392 new_ra->destructor = destructor;
394 RCU_INIT_POINTER(new_ra->next, ra);
395 rcu_assign_pointer(*rap, new_ra);
397 mutex_unlock(&net->ipv4.ra_mutex);
402 static void ipv4_icmp_error_rfc4884(const struct sk_buff *skb,
403 struct sock_ee_data_rfc4884 *out)
405 switch (icmp_hdr(skb)->type) {
406 case ICMP_DEST_UNREACH:
407 case ICMP_TIME_EXCEEDED:
408 case ICMP_PARAMETERPROB:
409 ip_icmp_error_rfc4884(skb, out, sizeof(struct icmphdr),
410 icmp_hdr(skb)->un.reserved[1] * 4);
414 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
415 __be16 port, u32 info, u8 *payload)
417 struct sock_exterr_skb *serr;
419 skb = skb_clone(skb, GFP_ATOMIC);
423 serr = SKB_EXT_ERR(skb);
424 serr->ee.ee_errno = err;
425 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
426 serr->ee.ee_type = icmp_hdr(skb)->type;
427 serr->ee.ee_code = icmp_hdr(skb)->code;
429 serr->ee.ee_info = info;
430 serr->ee.ee_data = 0;
431 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
432 skb_network_header(skb);
435 if (skb_pull(skb, payload - skb->data)) {
436 if (inet_test_bit(RECVERR_RFC4884, sk))
437 ipv4_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
439 skb_reset_transport_header(skb);
440 if (sock_queue_err_skb(sk, skb) == 0)
445 EXPORT_SYMBOL_GPL(ip_icmp_error);
447 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
449 struct sock_exterr_skb *serr;
453 if (!inet_test_bit(RECVERR, sk))
456 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
460 skb_put(skb, sizeof(struct iphdr));
461 skb_reset_network_header(skb);
465 serr = SKB_EXT_ERR(skb);
466 serr->ee.ee_errno = err;
467 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
468 serr->ee.ee_type = 0;
469 serr->ee.ee_code = 0;
471 serr->ee.ee_info = info;
472 serr->ee.ee_data = 0;
473 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
476 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
477 skb_reset_transport_header(skb);
479 if (sock_queue_err_skb(sk, skb))
483 /* For some errors we have valid addr_offset even with zero payload and
484 * zero port. Also, addr_offset should be supported if port is set.
486 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
488 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
489 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
492 /* IPv4 supports cmsg on all imcp errors and some timestamps
494 * Timestamp code paths do not initialize the fields expected by cmsg:
495 * the PKTINFO fields in skb->cb[]. Fill those in here.
497 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
501 struct in_pktinfo *info;
503 if (ee_origin == SO_EE_ORIGIN_ICMP)
506 if (ee_origin == SO_EE_ORIGIN_LOCAL)
509 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
510 * timestamp with egress dev. Not possible for packets without iif
511 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
513 info = PKTINFO_SKB_CB(skb);
514 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) ||
518 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
523 * Handle MSG_ERRQUEUE
525 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
527 struct sock_exterr_skb *serr;
529 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
531 struct sock_extended_err ee;
532 struct sockaddr_in offender;
538 skb = sock_dequeue_err_skb(sk);
544 msg->msg_flags |= MSG_TRUNC;
547 err = skb_copy_datagram_msg(skb, 0, msg, copied);
552 sock_recv_timestamp(msg, sk, skb);
554 serr = SKB_EXT_ERR(skb);
556 if (sin && ipv4_datagram_support_addr(serr)) {
557 sin->sin_family = AF_INET;
558 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
560 sin->sin_port = serr->port;
561 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
562 *addr_len = sizeof(*sin);
565 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
566 sin = &errhdr.offender;
567 memset(sin, 0, sizeof(*sin));
569 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
570 sin->sin_family = AF_INET;
571 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
572 if (inet_cmsg_flags(inet_sk(sk)))
573 ip_cmsg_recv(msg, skb);
576 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
578 /* Now we could try to dump offended packet options */
580 msg->msg_flags |= MSG_ERRQUEUE;
588 void __ip_sock_set_tos(struct sock *sk, int val)
590 if (sk->sk_type == SOCK_STREAM) {
591 val &= ~INET_ECN_MASK;
592 val |= inet_sk(sk)->tos & INET_ECN_MASK;
594 if (inet_sk(sk)->tos != val) {
595 inet_sk(sk)->tos = val;
596 WRITE_ONCE(sk->sk_priority, rt_tos2priority(val));
601 void ip_sock_set_tos(struct sock *sk, int val)
604 __ip_sock_set_tos(sk, val);
607 EXPORT_SYMBOL(ip_sock_set_tos);
609 void ip_sock_set_freebind(struct sock *sk)
612 inet_sk(sk)->freebind = true;
615 EXPORT_SYMBOL(ip_sock_set_freebind);
617 void ip_sock_set_recverr(struct sock *sk)
619 inet_set_bit(RECVERR, sk);
621 EXPORT_SYMBOL(ip_sock_set_recverr);
623 int ip_sock_set_mtu_discover(struct sock *sk, int val)
625 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
628 inet_sk(sk)->pmtudisc = val;
632 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
634 void ip_sock_set_pktinfo(struct sock *sk)
636 inet_set_bit(PKTINFO, sk);
638 EXPORT_SYMBOL(ip_sock_set_pktinfo);
641 * Socket option code for IP. This is the end of the line after any
642 * TCP,UDP etc options on an IP socket.
644 static bool setsockopt_needs_rtnl(int optname)
647 case IP_ADD_MEMBERSHIP:
648 case IP_ADD_SOURCE_MEMBERSHIP:
649 case IP_BLOCK_SOURCE:
650 case IP_DROP_MEMBERSHIP:
651 case IP_DROP_SOURCE_MEMBERSHIP:
653 case IP_UNBLOCK_SOURCE:
654 case MCAST_BLOCK_SOURCE:
656 case MCAST_JOIN_GROUP:
657 case MCAST_JOIN_SOURCE_GROUP:
658 case MCAST_LEAVE_GROUP:
659 case MCAST_LEAVE_SOURCE_GROUP:
660 case MCAST_UNBLOCK_SOURCE:
666 static int set_mcast_msfilter(struct sock *sk, int ifindex,
667 int numsrc, int fmode,
668 struct sockaddr_storage *group,
669 struct sockaddr_storage *list)
671 struct ip_msfilter *msf;
672 struct sockaddr_in *psin;
675 msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
679 psin = (struct sockaddr_in *)group;
680 if (psin->sin_family != AF_INET)
682 msf->imsf_multiaddr = psin->sin_addr.s_addr;
683 msf->imsf_interface = 0;
684 msf->imsf_fmode = fmode;
685 msf->imsf_numsrc = numsrc;
686 for (i = 0; i < numsrc; ++i) {
687 psin = (struct sockaddr_in *)&list[i];
689 if (psin->sin_family != AF_INET)
691 msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
693 err = ip_mc_msfilter(sk, msf, ifindex);
699 return -EADDRNOTAVAIL;
702 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
703 sockptr_t optval, int optlen)
705 if (in_compat_syscall()) {
706 struct compat_group_source_req gr32;
708 if (optlen != sizeof(gr32))
710 if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
712 greqs->gsr_interface = gr32.gsr_interface;
713 greqs->gsr_group = gr32.gsr_group;
714 greqs->gsr_source = gr32.gsr_source;
716 if (optlen != sizeof(*greqs))
718 if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
725 static int do_mcast_group_source(struct sock *sk, int optname,
726 sockptr_t optval, int optlen)
728 struct group_source_req greqs;
729 struct ip_mreq_source mreqs;
730 struct sockaddr_in *psin;
733 err = copy_group_source_from_sockptr(&greqs, optval, optlen);
737 if (greqs.gsr_group.ss_family != AF_INET ||
738 greqs.gsr_source.ss_family != AF_INET)
739 return -EADDRNOTAVAIL;
741 psin = (struct sockaddr_in *)&greqs.gsr_group;
742 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
743 psin = (struct sockaddr_in *)&greqs.gsr_source;
744 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
745 mreqs.imr_interface = 0; /* use index for mc_source */
747 if (optname == MCAST_BLOCK_SOURCE) {
748 omode = MCAST_EXCLUDE;
750 } else if (optname == MCAST_UNBLOCK_SOURCE) {
751 omode = MCAST_EXCLUDE;
753 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
754 struct ip_mreqn mreq;
756 psin = (struct sockaddr_in *)&greqs.gsr_group;
757 mreq.imr_multiaddr = psin->sin_addr;
758 mreq.imr_address.s_addr = 0;
759 mreq.imr_ifindex = greqs.gsr_interface;
760 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
761 if (err && err != -EADDRINUSE)
763 greqs.gsr_interface = mreq.imr_ifindex;
764 omode = MCAST_INCLUDE;
766 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
767 omode = MCAST_INCLUDE;
770 return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
773 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
775 struct group_filter *gsf = NULL;
778 if (optlen < GROUP_FILTER_SIZE(0))
780 if (optlen > READ_ONCE(sysctl_optmem_max))
783 gsf = memdup_sockptr(optval, optlen);
787 /* numsrc >= (4G-140)/128 overflow in 32 bits */
789 if (gsf->gf_numsrc >= 0x1ffffff ||
790 gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
794 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
797 err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
798 gsf->gf_fmode, &gsf->gf_group,
805 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
808 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
809 struct compat_group_filter *gf32;
816 if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
819 p = kmalloc(optlen + 4, GFP_KERNEL);
822 gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
825 if (copy_from_sockptr(gf32, optval, optlen))
828 /* numsrc >= (4G-140)/128 overflow in 32 bits */
835 if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
838 /* numsrc >= (4G-140)/128 overflow in 32 bits */
840 if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
842 err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
843 &gf32->gf_group, gf32->gf_slist_flex);
849 static int ip_mcast_join_leave(struct sock *sk, int optname,
850 sockptr_t optval, int optlen)
852 struct ip_mreqn mreq = { };
853 struct sockaddr_in *psin;
854 struct group_req greq;
856 if (optlen < sizeof(struct group_req))
858 if (copy_from_sockptr(&greq, optval, sizeof(greq)))
861 psin = (struct sockaddr_in *)&greq.gr_group;
862 if (psin->sin_family != AF_INET)
864 mreq.imr_multiaddr = psin->sin_addr;
865 mreq.imr_ifindex = greq.gr_interface;
866 if (optname == MCAST_JOIN_GROUP)
867 return ip_mc_join_group(sk, &mreq);
868 return ip_mc_leave_group(sk, &mreq);
871 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
872 sockptr_t optval, int optlen)
874 struct compat_group_req greq;
875 struct ip_mreqn mreq = { };
876 struct sockaddr_in *psin;
878 if (optlen < sizeof(struct compat_group_req))
880 if (copy_from_sockptr(&greq, optval, sizeof(greq)))
883 psin = (struct sockaddr_in *)&greq.gr_group;
884 if (psin->sin_family != AF_INET)
886 mreq.imr_multiaddr = psin->sin_addr;
887 mreq.imr_ifindex = greq.gr_interface;
889 if (optname == MCAST_JOIN_GROUP)
890 return ip_mc_join_group(sk, &mreq);
891 return ip_mc_leave_group(sk, &mreq);
894 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
896 int do_ip_setsockopt(struct sock *sk, int level, int optname,
897 sockptr_t optval, unsigned int optlen)
899 struct inet_sock *inet = inet_sk(sk);
900 struct net *net = sock_net(sk);
902 bool needs_rtnl = setsockopt_needs_rtnl(optname);
913 case IP_MTU_DISCOVER:
915 case IP_ROUTER_ALERT:
921 case IP_BIND_ADDRESS_NO_PORT:
923 case IP_MULTICAST_TTL:
924 case IP_MULTICAST_ALL:
925 case IP_MULTICAST_LOOP:
926 case IP_RECVORIGDSTADDR:
928 case IP_RECVFRAGSIZE:
929 case IP_RECVERR_RFC4884:
930 case IP_LOCAL_PORT_RANGE:
931 if (optlen >= sizeof(int)) {
932 if (copy_from_sockptr(&val, optval, sizeof(val)))
934 } else if (optlen >= sizeof(char)) {
937 if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
943 /* If optlen==0, it is equivalent to val == 0 */
945 if (optname == IP_ROUTER_ALERT)
946 return ip_ra_control(sk, val ? 1 : 0, NULL);
947 if (ip_mroute_opt(optname))
948 return ip_mroute_setsockopt(sk, optname, optval, optlen);
950 /* Handle options that can be set without locking the socket. */
953 inet_assign_bit(PKTINFO, sk, val);
956 inet_assign_bit(TTL, sk, val);
959 inet_assign_bit(TOS, sk, val);
962 inet_assign_bit(RECVOPTS, sk, val);
965 inet_assign_bit(RETOPTS, sk, val);
968 inet_assign_bit(PASSSEC, sk, val);
970 case IP_RECVORIGDSTADDR:
971 inet_assign_bit(ORIGDSTADDR, sk, val);
973 case IP_RECVFRAGSIZE:
974 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
976 inet_assign_bit(RECVFRAGSIZE, sk, val);
979 inet_assign_bit(RECVERR, sk, val);
981 skb_queue_purge(&sk->sk_error_queue);
983 case IP_RECVERR_RFC4884:
984 if (val < 0 || val > 1)
986 inet_assign_bit(RECVERR_RFC4884, sk, val);
993 sockopt_lock_sock(sk);
998 struct ip_options_rcu *old, *opt = NULL;
1002 err = ip_options_get(sock_net(sk), &opt, optval, optlen);
1005 old = rcu_dereference_protected(inet->inet_opt,
1006 lockdep_sock_is_held(sk));
1007 if (inet->is_icsk) {
1008 struct inet_connection_sock *icsk = inet_csk(sk);
1009 #if IS_ENABLED(CONFIG_IPV6)
1010 if (sk->sk_family == PF_INET ||
1011 (!((1 << sk->sk_state) &
1012 (TCPF_LISTEN | TCPF_CLOSE)) &&
1013 inet->inet_daddr != LOOPBACK4_IPV6)) {
1016 icsk->icsk_ext_hdr_len -= old->opt.optlen;
1018 icsk->icsk_ext_hdr_len += opt->opt.optlen;
1019 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
1020 #if IS_ENABLED(CONFIG_IPV6)
1024 rcu_assign_pointer(inet->inet_opt, opt);
1026 kfree_rcu(old, rcu);
1031 if (!(inet_test_bit(CHECKSUM, sk))) {
1032 inet_inc_convert_csum(sk);
1033 inet_set_bit(CHECKSUM, sk);
1036 if (inet_test_bit(CHECKSUM, sk)) {
1037 inet_dec_convert_csum(sk);
1038 inet_clear_bit(CHECKSUM, sk);
1042 case IP_TOS: /* This sets both TOS and Precedence */
1043 __ip_sock_set_tos(sk, val);
1048 if (val != -1 && (val < 1 || val > 255))
1053 if (sk->sk_type != SOCK_RAW) {
1057 inet->hdrincl = val ? 1 : 0;
1060 if (sk->sk_type != SOCK_RAW) {
1064 inet->nodefrag = val ? 1 : 0;
1066 case IP_BIND_ADDRESS_NO_PORT:
1067 inet->bind_address_no_port = val ? 1 : 0;
1069 case IP_MTU_DISCOVER:
1070 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
1072 inet->pmtudisc = val;
1074 case IP_MULTICAST_TTL:
1075 if (sk->sk_type == SOCK_STREAM)
1081 if (val < 0 || val > 255)
1085 case IP_MULTICAST_LOOP:
1088 inet->mc_loop = !!val;
1092 struct net_device *dev = NULL;
1096 if (optlen != sizeof(int))
1099 ifindex = (__force int)ntohl((__force __be32)val);
1106 dev = dev_get_by_index(sock_net(sk), ifindex);
1107 err = -EADDRNOTAVAIL;
1111 midx = l3mdev_master_ifindex(dev);
1115 if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1118 inet->uc_index = ifindex;
1122 case IP_MULTICAST_IF:
1124 struct ip_mreqn mreq;
1125 struct net_device *dev = NULL;
1128 if (sk->sk_type == SOCK_STREAM)
1131 * Check the arguments are allowable
1134 if (optlen < sizeof(struct in_addr))
1138 if (optlen >= sizeof(struct ip_mreqn)) {
1139 if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1142 memset(&mreq, 0, sizeof(mreq));
1143 if (optlen >= sizeof(struct ip_mreq)) {
1144 if (copy_from_sockptr(&mreq, optval,
1145 sizeof(struct ip_mreq)))
1147 } else if (optlen >= sizeof(struct in_addr)) {
1148 if (copy_from_sockptr(&mreq.imr_address, optval,
1149 sizeof(struct in_addr)))
1154 if (!mreq.imr_ifindex) {
1155 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1161 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1163 mreq.imr_ifindex = dev->ifindex;
1165 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1168 err = -EADDRNOTAVAIL;
1172 midx = l3mdev_master_ifindex(dev);
1177 if (sk->sk_bound_dev_if &&
1178 mreq.imr_ifindex != sk->sk_bound_dev_if &&
1179 midx != sk->sk_bound_dev_if)
1182 inet->mc_index = mreq.imr_ifindex;
1183 inet->mc_addr = mreq.imr_address.s_addr;
1188 case IP_ADD_MEMBERSHIP:
1189 case IP_DROP_MEMBERSHIP:
1191 struct ip_mreqn mreq;
1194 if (inet_sk(sk)->is_icsk)
1197 if (optlen < sizeof(struct ip_mreq))
1200 if (optlen >= sizeof(struct ip_mreqn)) {
1201 if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1204 memset(&mreq, 0, sizeof(mreq));
1205 if (copy_from_sockptr(&mreq, optval,
1206 sizeof(struct ip_mreq)))
1210 if (optname == IP_ADD_MEMBERSHIP)
1211 err = ip_mc_join_group(sk, &mreq);
1213 err = ip_mc_leave_group(sk, &mreq);
1218 struct ip_msfilter *msf;
1220 if (optlen < IP_MSFILTER_SIZE(0))
1222 if (optlen > READ_ONCE(sysctl_optmem_max)) {
1226 msf = memdup_sockptr(optval, optlen);
1231 /* numsrc >= (1G-4) overflow in 32 bits */
1232 if (msf->imsf_numsrc >= 0x3ffffffcU ||
1233 msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1238 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1243 err = ip_mc_msfilter(sk, msf, 0);
1247 case IP_BLOCK_SOURCE:
1248 case IP_UNBLOCK_SOURCE:
1249 case IP_ADD_SOURCE_MEMBERSHIP:
1250 case IP_DROP_SOURCE_MEMBERSHIP:
1252 struct ip_mreq_source mreqs;
1255 if (optlen != sizeof(struct ip_mreq_source))
1257 if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1261 if (optname == IP_BLOCK_SOURCE) {
1262 omode = MCAST_EXCLUDE;
1264 } else if (optname == IP_UNBLOCK_SOURCE) {
1265 omode = MCAST_EXCLUDE;
1267 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1268 struct ip_mreqn mreq;
1270 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1271 mreq.imr_address.s_addr = mreqs.imr_interface;
1272 mreq.imr_ifindex = 0;
1273 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1274 if (err && err != -EADDRINUSE)
1276 omode = MCAST_INCLUDE;
1278 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1279 omode = MCAST_INCLUDE;
1282 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1285 case MCAST_JOIN_GROUP:
1286 case MCAST_LEAVE_GROUP:
1287 if (in_compat_syscall())
1288 err = compat_ip_mcast_join_leave(sk, optname, optval,
1291 err = ip_mcast_join_leave(sk, optname, optval, optlen);
1293 case MCAST_JOIN_SOURCE_GROUP:
1294 case MCAST_LEAVE_SOURCE_GROUP:
1295 case MCAST_BLOCK_SOURCE:
1296 case MCAST_UNBLOCK_SOURCE:
1297 err = do_mcast_group_source(sk, optname, optval, optlen);
1299 case MCAST_MSFILTER:
1300 if (in_compat_syscall())
1301 err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1303 err = ip_set_mcast_msfilter(sk, optval, optlen);
1305 case IP_MULTICAST_ALL:
1308 if (val != 0 && val != 1)
1316 inet->freebind = !!val;
1319 case IP_IPSEC_POLICY:
1320 case IP_XFRM_POLICY:
1322 if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1324 err = xfrm_user_policy(sk, optname, optval, optlen);
1327 case IP_TRANSPARENT:
1328 if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1329 !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1335 inet->transparent = !!val;
1341 if (val < 0 || val > 255)
1345 static_branch_enable(&ip4_min_ttl);
1347 /* tcp_v4_err() and tcp_v4_rcv() might read min_ttl
1348 * while we are changint it.
1350 WRITE_ONCE(inet->min_ttl, val);
1353 case IP_LOCAL_PORT_RANGE:
1355 const __u16 lo = val;
1356 const __u16 hi = val >> 16;
1358 if (optlen != sizeof(__u32))
1360 if (lo != 0 && hi != 0 && lo > hi)
1363 inet->local_port_range.lo = lo;
1364 inet->local_port_range.hi = hi;
1371 sockopt_release_sock(sk);
1377 sockopt_release_sock(sk);
1384 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1388 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1389 * destination in skb->cb[] before dst drop.
1390 * This way, receiver doesn't make cache line misses to read rtable.
1392 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1394 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1395 bool prepare = inet_test_bit(PKTINFO, sk) ||
1398 if (prepare && skb_rtable(skb)) {
1399 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1400 * which has interface index (iif) as the first member of the
1401 * underlying inet{6}_skb_parm struct. This code then overlays
1402 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1403 * element so the iif is picked up from the prior IPCB. If iif
1404 * is the loopback interface, then return the sending interface
1405 * (e.g., process binds socket to eth0 for Tx which is
1406 * redirected to loopback in the rtable/dst).
1408 struct rtable *rt = skb_rtable(skb);
1409 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags);
1411 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX)
1412 pktinfo->ipi_ifindex = inet_iif(skb);
1413 else if (l3slave && rt && rt->rt_iif)
1414 pktinfo->ipi_ifindex = rt->rt_iif;
1416 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1418 pktinfo->ipi_ifindex = 0;
1419 pktinfo->ipi_spec_dst.s_addr = 0;
1424 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1425 unsigned int optlen)
1429 if (level != SOL_IP)
1430 return -ENOPROTOOPT;
1432 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1433 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1434 if (optname >= BPFILTER_IPT_SO_SET_REPLACE &&
1435 optname < BPFILTER_IPT_SET_MAX)
1436 err = bpfilter_ip_set_sockopt(sk, optname, optval, optlen);
1438 #ifdef CONFIG_NETFILTER
1439 /* we need to exclude all possible ENOPROTOOPTs except default case */
1440 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1441 optname != IP_IPSEC_POLICY &&
1442 optname != IP_XFRM_POLICY &&
1443 !ip_mroute_opt(optname))
1444 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1448 EXPORT_SYMBOL(ip_setsockopt);
1451 * Get the options. Note for future reference. The GET of IP options gets
1452 * the _received_ ones. The set sets the _sent_ ones.
1455 static bool getsockopt_needs_rtnl(int optname)
1459 case MCAST_MSFILTER:
1465 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1466 sockptr_t optlen, int len)
1468 const int size0 = offsetof(struct group_filter, gf_slist_flex);
1469 struct group_filter gsf;
1475 if (copy_from_sockptr(&gsf, optval, size0))
1478 num = gsf.gf_numsrc;
1479 err = ip_mc_gsfget(sk, &gsf, optval,
1480 offsetof(struct group_filter, gf_slist_flex));
1483 if (gsf.gf_numsrc < num)
1484 num = gsf.gf_numsrc;
1485 gsf_size = GROUP_FILTER_SIZE(num);
1486 if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1487 copy_to_sockptr(optval, &gsf, size0))
1492 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1493 sockptr_t optlen, int len)
1495 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1496 struct compat_group_filter gf32;
1497 struct group_filter gf;
1503 if (copy_from_sockptr(&gf32, optval, size0))
1506 gf.gf_interface = gf32.gf_interface;
1507 gf.gf_fmode = gf32.gf_fmode;
1508 num = gf.gf_numsrc = gf32.gf_numsrc;
1509 gf.gf_group = gf32.gf_group;
1511 err = ip_mc_gsfget(sk, &gf, optval,
1512 offsetof(struct compat_group_filter, gf_slist_flex));
1515 if (gf.gf_numsrc < num)
1517 len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1518 if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1519 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1520 &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1521 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1522 &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1527 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1528 sockptr_t optval, sockptr_t optlen)
1530 struct inet_sock *inet = inet_sk(sk);
1531 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1535 if (level != SOL_IP)
1538 if (ip_mroute_opt(optname))
1539 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1541 if (copy_from_sockptr(&len, optlen, sizeof(int)))
1546 /* Handle options that can be read without locking the socket. */
1549 val = inet_test_bit(PKTINFO, sk);
1552 val = inet_test_bit(TTL, sk);
1555 val = inet_test_bit(TOS, sk);
1558 val = inet_test_bit(RECVOPTS, sk);
1561 val = inet_test_bit(RETOPTS, sk);
1564 val = inet_test_bit(PASSSEC, sk);
1566 case IP_RECVORIGDSTADDR:
1567 val = inet_test_bit(ORIGDSTADDR, sk);
1570 val = inet_test_bit(CHECKSUM, sk);
1572 case IP_RECVFRAGSIZE:
1573 val = inet_test_bit(RECVFRAGSIZE, sk);
1576 val = inet_test_bit(RECVERR, sk);
1578 case IP_RECVERR_RFC4884:
1579 val = inet_test_bit(RECVERR_RFC4884, sk);
1585 sockopt_lock_sock(sk);
1590 unsigned char optbuf[sizeof(struct ip_options)+40];
1591 struct ip_options *opt = (struct ip_options *)optbuf;
1592 struct ip_options_rcu *inet_opt;
1594 inet_opt = rcu_dereference_protected(inet->inet_opt,
1595 lockdep_sock_is_held(sk));
1598 memcpy(optbuf, &inet_opt->opt,
1599 sizeof(struct ip_options) +
1600 inet_opt->opt.optlen);
1601 sockopt_release_sock(sk);
1603 if (opt->optlen == 0) {
1605 return copy_to_sockptr(optlen, &len, sizeof(int));
1608 ip_options_undo(opt);
1610 len = min_t(unsigned int, len, opt->optlen);
1611 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1613 if (copy_to_sockptr(optval, opt->__data, len))
1622 struct net *net = sock_net(sk);
1623 val = (inet->uc_ttl == -1 ?
1624 READ_ONCE(net->ipv4.sysctl_ip_default_ttl) :
1629 val = inet->hdrincl;
1632 val = inet->nodefrag;
1634 case IP_BIND_ADDRESS_NO_PORT:
1635 val = inet->bind_address_no_port;
1637 case IP_MTU_DISCOVER:
1638 val = inet->pmtudisc;
1642 struct dst_entry *dst;
1644 dst = sk_dst_get(sk);
1650 sockopt_release_sock(sk);
1655 case IP_MULTICAST_TTL:
1658 case IP_MULTICAST_LOOP:
1659 val = inet->mc_loop;
1662 val = (__force int)htonl((__u32) inet->uc_index);
1664 case IP_MULTICAST_IF:
1666 struct in_addr addr;
1667 len = min_t(unsigned int, len, sizeof(struct in_addr));
1668 addr.s_addr = inet->mc_addr;
1669 sockopt_release_sock(sk);
1671 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1673 if (copy_to_sockptr(optval, &addr, len))
1679 struct ip_msfilter msf;
1681 if (len < IP_MSFILTER_SIZE(0)) {
1685 if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1689 err = ip_mc_msfget(sk, &msf, optval, optlen);
1692 case MCAST_MSFILTER:
1693 if (in_compat_syscall())
1694 err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1697 err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1699 case IP_MULTICAST_ALL:
1706 sockopt_release_sock(sk);
1708 if (sk->sk_type != SOCK_STREAM)
1709 return -ENOPROTOOPT;
1711 if (optval.is_kernel) {
1712 msg.msg_control_is_user = false;
1713 msg.msg_control = optval.kernel;
1715 msg.msg_control_is_user = true;
1716 msg.msg_control_user = optval.user;
1718 msg.msg_controllen = len;
1719 msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1721 if (inet_test_bit(PKTINFO, sk)) {
1722 struct in_pktinfo info;
1724 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1725 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1726 info.ipi_ifindex = inet->mc_index;
1727 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1729 if (inet_test_bit(TTL, sk)) {
1730 int hlim = inet->mc_ttl;
1731 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1733 if (inet_test_bit(TOS, sk)) {
1734 int tos = inet->rcv_tos;
1735 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1737 len -= msg.msg_controllen;
1738 return copy_to_sockptr(optlen, &len, sizeof(int));
1741 val = inet->freebind;
1743 case IP_TRANSPARENT:
1744 val = inet->transparent;
1747 val = inet->min_ttl;
1749 case IP_LOCAL_PORT_RANGE:
1750 val = inet->local_port_range.hi << 16 | inet->local_port_range.lo;
1753 val = inet_sk(sk)->inet_num;
1756 sockopt_release_sock(sk);
1757 return -ENOPROTOOPT;
1759 sockopt_release_sock(sk);
1761 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1762 unsigned char ucval = (unsigned char)val;
1764 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1766 if (copy_to_sockptr(optval, &ucval, 1))
1769 len = min_t(unsigned int, sizeof(int), len);
1770 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1772 if (copy_to_sockptr(optval, &val, len))
1778 sockopt_release_sock(sk);
1784 int ip_getsockopt(struct sock *sk, int level,
1785 int optname, char __user *optval, int __user *optlen)
1789 err = do_ip_getsockopt(sk, level, optname,
1790 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1792 #if IS_ENABLED(CONFIG_BPFILTER_UMH)
1793 if (optname >= BPFILTER_IPT_SO_GET_INFO &&
1794 optname < BPFILTER_IPT_GET_MAX)
1795 err = bpfilter_ip_get_sockopt(sk, optname, optval, optlen);
1797 #ifdef CONFIG_NETFILTER
1798 /* we need to exclude all possible ENOPROTOOPTs except default case */
1799 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1800 !ip_mroute_opt(optname)) {
1803 if (get_user(len, optlen))
1806 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1808 err = put_user(len, optlen);
1814 EXPORT_SYMBOL(ip_getsockopt);