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>
51 * SOL_IP control messages.
54 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
56 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
58 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
60 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
63 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
65 int ttl = ip_hdr(skb)->ttl;
66 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
69 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
71 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
74 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
76 if (IPCB(skb)->opt.optlen == 0)
79 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
84 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg,
87 unsigned char optbuf[sizeof(struct ip_options) + 40];
88 struct ip_options *opt = (struct ip_options *)optbuf;
90 if (IPCB(skb)->opt.optlen == 0)
93 if (ip_options_echo(net, opt, skb)) {
94 msg->msg_flags |= MSG_CTRUNC;
99 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
102 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb)
106 if (IPCB(skb)->frag_max_size == 0)
109 val = IPCB(skb)->frag_max_size;
110 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val);
113 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
114 int tlen, int offset)
116 __wsum csum = skb->csum;
118 if (skb->ip_summed != CHECKSUM_COMPLETE)
122 int tend_off = skb_transport_offset(skb) + tlen;
123 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0));
126 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
129 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
135 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
139 err = security_secid_to_secctx(secid, &secdata, &seclen);
143 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
144 security_release_secctx(secdata, seclen);
147 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
149 __be16 _ports[2], *ports;
150 struct sockaddr_in sin;
152 /* All current transport protocols have the port numbers in the
153 * first four bytes of the transport header and this function is
154 * written with this assumption in mind.
156 ports = skb_header_pointer(skb, skb_transport_offset(skb),
157 sizeof(_ports), &_ports);
161 sin.sin_family = AF_INET;
162 sin.sin_addr.s_addr = ip_hdr(skb)->daddr;
163 sin.sin_port = ports[1];
164 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
166 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
169 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
170 struct sk_buff *skb, int tlen, int offset)
172 unsigned long flags = inet_cmsg_flags(inet_sk(sk));
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, 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,
284 KERNEL_SOCKPTR(CMSG_DATA(cmsg)),
285 err < 40 ? err : 40);
291 struct in_pktinfo *info;
292 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
294 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
295 if (info->ipi_ifindex)
296 ipc->oif = info->ipi_ifindex;
297 ipc->addr = info->ipi_spec_dst.s_addr;
301 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
303 val = *(int *)CMSG_DATA(cmsg);
304 if (val < 1 || val > 255)
309 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
310 val = *(int *)CMSG_DATA(cmsg);
311 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
312 val = *(u8 *)CMSG_DATA(cmsg);
315 if (val < 0 || val > 255)
318 ipc->priority = rt_tos2priority(ipc->tos);
321 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
323 val = *(int *)CMSG_DATA(cmsg);
324 if (val < 1 || val > 255)
335 static void ip_ra_destroy_rcu(struct rcu_head *head)
337 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
339 sock_put(ra->saved_sk);
343 int ip_ra_control(struct sock *sk, unsigned char on,
344 void (*destructor)(struct sock *))
346 struct ip_ra_chain *ra, *new_ra;
347 struct ip_ra_chain __rcu **rap;
348 struct net *net = sock_net(sk);
350 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
353 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
357 mutex_lock(&net->ipv4.ra_mutex);
358 for (rap = &net->ipv4.ra_chain;
359 (ra = rcu_dereference_protected(*rap,
360 lockdep_is_held(&net->ipv4.ra_mutex))) != NULL;
364 mutex_unlock(&net->ipv4.ra_mutex);
368 /* dont let ip_call_ra_chain() use sk again */
370 RCU_INIT_POINTER(*rap, ra->next);
371 mutex_unlock(&net->ipv4.ra_mutex);
376 * Delay sock_put(sk) and kfree(ra) after one rcu grace
377 * period. This guarantee ip_call_ra_chain() dont need
378 * to mess with socket refcounts.
381 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
386 mutex_unlock(&net->ipv4.ra_mutex);
390 new_ra->destructor = destructor;
392 RCU_INIT_POINTER(new_ra->next, ra);
393 rcu_assign_pointer(*rap, new_ra);
395 mutex_unlock(&net->ipv4.ra_mutex);
400 static void ipv4_icmp_error_rfc4884(const struct sk_buff *skb,
401 struct sock_ee_data_rfc4884 *out)
403 switch (icmp_hdr(skb)->type) {
404 case ICMP_DEST_UNREACH:
405 case ICMP_TIME_EXCEEDED:
406 case ICMP_PARAMETERPROB:
407 ip_icmp_error_rfc4884(skb, out, sizeof(struct icmphdr),
408 icmp_hdr(skb)->un.reserved[1] * 4);
412 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
413 __be16 port, u32 info, u8 *payload)
415 struct sock_exterr_skb *serr;
417 skb = skb_clone(skb, GFP_ATOMIC);
421 serr = SKB_EXT_ERR(skb);
422 serr->ee.ee_errno = err;
423 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
424 serr->ee.ee_type = icmp_hdr(skb)->type;
425 serr->ee.ee_code = icmp_hdr(skb)->code;
427 serr->ee.ee_info = info;
428 serr->ee.ee_data = 0;
429 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
430 skb_network_header(skb);
433 if (skb_pull(skb, payload - skb->data)) {
434 if (inet_test_bit(RECVERR_RFC4884, sk))
435 ipv4_icmp_error_rfc4884(skb, &serr->ee.ee_rfc4884);
437 skb_reset_transport_header(skb);
438 if (sock_queue_err_skb(sk, skb) == 0)
443 EXPORT_SYMBOL_GPL(ip_icmp_error);
445 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
447 struct sock_exterr_skb *serr;
451 if (!inet_test_bit(RECVERR, sk))
454 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
458 skb_put(skb, sizeof(struct iphdr));
459 skb_reset_network_header(skb);
463 serr = SKB_EXT_ERR(skb);
464 serr->ee.ee_errno = err;
465 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
466 serr->ee.ee_type = 0;
467 serr->ee.ee_code = 0;
469 serr->ee.ee_info = info;
470 serr->ee.ee_data = 0;
471 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
474 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
475 skb_reset_transport_header(skb);
477 if (sock_queue_err_skb(sk, skb))
481 /* For some errors we have valid addr_offset even with zero payload and
482 * zero port. Also, addr_offset should be supported if port is set.
484 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
486 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
487 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
490 /* IPv4 supports cmsg on all imcp errors and some timestamps
492 * Timestamp code paths do not initialize the fields expected by cmsg:
493 * the PKTINFO fields in skb->cb[]. Fill those in here.
495 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
499 struct in_pktinfo *info;
501 if (ee_origin == SO_EE_ORIGIN_ICMP)
504 if (ee_origin == SO_EE_ORIGIN_LOCAL)
507 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
508 * timestamp with egress dev. Not possible for packets without iif
509 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
511 info = PKTINFO_SKB_CB(skb);
512 if (!(READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_CMSG) ||
516 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
521 * Handle MSG_ERRQUEUE
523 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
525 struct sock_exterr_skb *serr;
527 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
529 struct sock_extended_err ee;
530 struct sockaddr_in offender;
536 skb = sock_dequeue_err_skb(sk);
542 msg->msg_flags |= MSG_TRUNC;
545 err = skb_copy_datagram_msg(skb, 0, msg, copied);
550 sock_recv_timestamp(msg, sk, skb);
552 serr = SKB_EXT_ERR(skb);
554 if (sin && ipv4_datagram_support_addr(serr)) {
555 sin->sin_family = AF_INET;
556 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
558 sin->sin_port = serr->port;
559 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
560 *addr_len = sizeof(*sin);
563 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
564 sin = &errhdr.offender;
565 memset(sin, 0, sizeof(*sin));
567 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
568 sin->sin_family = AF_INET;
569 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
570 if (inet_cmsg_flags(inet_sk(sk)))
571 ip_cmsg_recv(msg, skb);
574 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
576 /* Now we could try to dump offended packet options */
578 msg->msg_flags |= MSG_ERRQUEUE;
586 void __ip_sock_set_tos(struct sock *sk, int val)
588 u8 old_tos = inet_sk(sk)->tos;
590 if (sk->sk_type == SOCK_STREAM) {
591 val &= ~INET_ECN_MASK;
592 val |= old_tos & INET_ECN_MASK;
594 if (old_tos != val) {
595 WRITE_ONCE(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)
603 sockopt_lock_sock(sk);
604 __ip_sock_set_tos(sk, val);
605 sockopt_release_sock(sk);
607 EXPORT_SYMBOL(ip_sock_set_tos);
609 void ip_sock_set_freebind(struct sock *sk)
611 inet_set_bit(FREEBIND, sk);
613 EXPORT_SYMBOL(ip_sock_set_freebind);
615 void ip_sock_set_recverr(struct sock *sk)
617 inet_set_bit(RECVERR, sk);
619 EXPORT_SYMBOL(ip_sock_set_recverr);
621 int ip_sock_set_mtu_discover(struct sock *sk, int val)
623 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
625 WRITE_ONCE(inet_sk(sk)->pmtudisc, val);
628 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
630 void ip_sock_set_pktinfo(struct sock *sk)
632 inet_set_bit(PKTINFO, sk);
634 EXPORT_SYMBOL(ip_sock_set_pktinfo);
637 * Socket option code for IP. This is the end of the line after any
638 * TCP,UDP etc options on an IP socket.
640 static bool setsockopt_needs_rtnl(int optname)
643 case IP_ADD_MEMBERSHIP:
644 case IP_ADD_SOURCE_MEMBERSHIP:
645 case IP_BLOCK_SOURCE:
646 case IP_DROP_MEMBERSHIP:
647 case IP_DROP_SOURCE_MEMBERSHIP:
649 case IP_UNBLOCK_SOURCE:
650 case MCAST_BLOCK_SOURCE:
652 case MCAST_JOIN_GROUP:
653 case MCAST_JOIN_SOURCE_GROUP:
654 case MCAST_LEAVE_GROUP:
655 case MCAST_LEAVE_SOURCE_GROUP:
656 case MCAST_UNBLOCK_SOURCE:
662 static int set_mcast_msfilter(struct sock *sk, int ifindex,
663 int numsrc, int fmode,
664 struct sockaddr_storage *group,
665 struct sockaddr_storage *list)
667 struct ip_msfilter *msf;
668 struct sockaddr_in *psin;
671 msf = kmalloc(IP_MSFILTER_SIZE(numsrc), GFP_KERNEL);
675 psin = (struct sockaddr_in *)group;
676 if (psin->sin_family != AF_INET)
678 msf->imsf_multiaddr = psin->sin_addr.s_addr;
679 msf->imsf_interface = 0;
680 msf->imsf_fmode = fmode;
681 msf->imsf_numsrc = numsrc;
682 for (i = 0; i < numsrc; ++i) {
683 psin = (struct sockaddr_in *)&list[i];
685 if (psin->sin_family != AF_INET)
687 msf->imsf_slist_flex[i] = psin->sin_addr.s_addr;
689 err = ip_mc_msfilter(sk, msf, ifindex);
695 return -EADDRNOTAVAIL;
698 static int copy_group_source_from_sockptr(struct group_source_req *greqs,
699 sockptr_t optval, int optlen)
701 if (in_compat_syscall()) {
702 struct compat_group_source_req gr32;
704 if (optlen != sizeof(gr32))
706 if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
708 greqs->gsr_interface = gr32.gsr_interface;
709 greqs->gsr_group = gr32.gsr_group;
710 greqs->gsr_source = gr32.gsr_source;
712 if (optlen != sizeof(*greqs))
714 if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
721 static int do_mcast_group_source(struct sock *sk, int optname,
722 sockptr_t optval, int optlen)
724 struct group_source_req greqs;
725 struct ip_mreq_source mreqs;
726 struct sockaddr_in *psin;
729 err = copy_group_source_from_sockptr(&greqs, optval, optlen);
733 if (greqs.gsr_group.ss_family != AF_INET ||
734 greqs.gsr_source.ss_family != AF_INET)
735 return -EADDRNOTAVAIL;
737 psin = (struct sockaddr_in *)&greqs.gsr_group;
738 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
739 psin = (struct sockaddr_in *)&greqs.gsr_source;
740 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
741 mreqs.imr_interface = 0; /* use index for mc_source */
743 if (optname == MCAST_BLOCK_SOURCE) {
744 omode = MCAST_EXCLUDE;
746 } else if (optname == MCAST_UNBLOCK_SOURCE) {
747 omode = MCAST_EXCLUDE;
749 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
750 struct ip_mreqn mreq;
752 psin = (struct sockaddr_in *)&greqs.gsr_group;
753 mreq.imr_multiaddr = psin->sin_addr;
754 mreq.imr_address.s_addr = 0;
755 mreq.imr_ifindex = greqs.gsr_interface;
756 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
757 if (err && err != -EADDRINUSE)
759 greqs.gsr_interface = mreq.imr_ifindex;
760 omode = MCAST_INCLUDE;
762 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
763 omode = MCAST_INCLUDE;
766 return ip_mc_source(add, omode, sk, &mreqs, greqs.gsr_interface);
769 static int ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval, int optlen)
771 struct group_filter *gsf = NULL;
774 if (optlen < GROUP_FILTER_SIZE(0))
776 if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max))
779 gsf = memdup_sockptr(optval, optlen);
783 /* numsrc >= (4G-140)/128 overflow in 32 bits */
785 if (gsf->gf_numsrc >= 0x1ffffff ||
786 gsf->gf_numsrc > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
790 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
793 err = set_mcast_msfilter(sk, gsf->gf_interface, gsf->gf_numsrc,
794 gsf->gf_fmode, &gsf->gf_group,
801 static int compat_ip_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
804 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
805 struct compat_group_filter *gf32;
812 if (optlen > READ_ONCE(sock_net(sk)->core.sysctl_optmem_max) - 4)
815 p = kmalloc(optlen + 4, GFP_KERNEL);
818 gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
821 if (copy_from_sockptr(gf32, optval, optlen))
824 /* numsrc >= (4G-140)/128 overflow in 32 bits */
831 if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
834 /* numsrc >= (4G-140)/128 overflow in 32 bits */
836 if (n > READ_ONCE(sock_net(sk)->ipv4.sysctl_igmp_max_msf))
838 err = set_mcast_msfilter(sk, gf32->gf_interface, n, gf32->gf_fmode,
839 &gf32->gf_group, gf32->gf_slist_flex);
845 static int ip_mcast_join_leave(struct sock *sk, int optname,
846 sockptr_t optval, int optlen)
848 struct ip_mreqn mreq = { };
849 struct sockaddr_in *psin;
850 struct group_req greq;
852 if (optlen < sizeof(struct group_req))
854 if (copy_from_sockptr(&greq, optval, sizeof(greq)))
857 psin = (struct sockaddr_in *)&greq.gr_group;
858 if (psin->sin_family != AF_INET)
860 mreq.imr_multiaddr = psin->sin_addr;
861 mreq.imr_ifindex = greq.gr_interface;
862 if (optname == MCAST_JOIN_GROUP)
863 return ip_mc_join_group(sk, &mreq);
864 return ip_mc_leave_group(sk, &mreq);
867 static int compat_ip_mcast_join_leave(struct sock *sk, int optname,
868 sockptr_t optval, int optlen)
870 struct compat_group_req greq;
871 struct ip_mreqn mreq = { };
872 struct sockaddr_in *psin;
874 if (optlen < sizeof(struct compat_group_req))
876 if (copy_from_sockptr(&greq, optval, sizeof(greq)))
879 psin = (struct sockaddr_in *)&greq.gr_group;
880 if (psin->sin_family != AF_INET)
882 mreq.imr_multiaddr = psin->sin_addr;
883 mreq.imr_ifindex = greq.gr_interface;
885 if (optname == MCAST_JOIN_GROUP)
886 return ip_mc_join_group(sk, &mreq);
887 return ip_mc_leave_group(sk, &mreq);
890 DEFINE_STATIC_KEY_FALSE(ip4_min_ttl);
892 int do_ip_setsockopt(struct sock *sk, int level, int optname,
893 sockptr_t optval, unsigned int optlen)
895 struct inet_sock *inet = inet_sk(sk);
896 struct net *net = sock_net(sk);
898 bool needs_rtnl = setsockopt_needs_rtnl(optname);
909 case IP_MTU_DISCOVER:
911 case IP_ROUTER_ALERT:
917 case IP_BIND_ADDRESS_NO_PORT:
919 case IP_MULTICAST_TTL:
920 case IP_MULTICAST_ALL:
921 case IP_MULTICAST_LOOP:
922 case IP_RECVORIGDSTADDR:
924 case IP_RECVFRAGSIZE:
925 case IP_RECVERR_RFC4884:
926 case IP_LOCAL_PORT_RANGE:
927 if (optlen >= sizeof(int)) {
928 if (copy_from_sockptr(&val, optval, sizeof(val)))
930 } else if (optlen >= sizeof(char)) {
933 if (copy_from_sockptr(&ucval, optval, sizeof(ucval)))
939 /* If optlen==0, it is equivalent to val == 0 */
941 if (optname == IP_ROUTER_ALERT)
942 return ip_ra_control(sk, val ? 1 : 0, NULL);
943 if (ip_mroute_opt(optname))
944 return ip_mroute_setsockopt(sk, optname, optval, optlen);
946 /* Handle options that can be set without locking the socket. */
949 inet_assign_bit(PKTINFO, sk, val);
952 inet_assign_bit(TTL, sk, val);
955 inet_assign_bit(TOS, sk, val);
958 inet_assign_bit(RECVOPTS, sk, val);
961 inet_assign_bit(RETOPTS, sk, val);
964 inet_assign_bit(PASSSEC, sk, val);
966 case IP_RECVORIGDSTADDR:
967 inet_assign_bit(ORIGDSTADDR, sk, val);
969 case IP_RECVFRAGSIZE:
970 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM)
972 inet_assign_bit(RECVFRAGSIZE, sk, val);
975 inet_assign_bit(RECVERR, sk, val);
977 skb_errqueue_purge(&sk->sk_error_queue);
979 case IP_RECVERR_RFC4884:
980 if (val < 0 || val > 1)
982 inet_assign_bit(RECVERR_RFC4884, sk, val);
987 inet_assign_bit(FREEBIND, sk, val);
990 if (sk->sk_type != SOCK_RAW)
992 inet_assign_bit(HDRINCL, sk, val);
994 case IP_MULTICAST_LOOP:
997 inet_assign_bit(MC_LOOP, sk, val);
999 case IP_MULTICAST_ALL:
1002 if (val != 0 && val != 1)
1004 inet_assign_bit(MC_ALL, sk, val);
1006 case IP_TRANSPARENT:
1007 if (!!val && !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1008 !sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1012 inet_assign_bit(TRANSPARENT, sk, val);
1015 if (sk->sk_type != SOCK_RAW)
1016 return -ENOPROTOOPT;
1017 inet_assign_bit(NODEFRAG, sk, val);
1019 case IP_BIND_ADDRESS_NO_PORT:
1020 inet_assign_bit(BIND_ADDRESS_NO_PORT, sk, val);
1025 if (val != -1 && (val < 1 || val > 255))
1027 WRITE_ONCE(inet->uc_ttl, val);
1032 if (val < 0 || val > 255)
1036 static_branch_enable(&ip4_min_ttl);
1038 WRITE_ONCE(inet->min_ttl, val);
1040 case IP_MULTICAST_TTL:
1041 if (sk->sk_type == SOCK_STREAM)
1047 if (val < 0 || val > 255)
1049 WRITE_ONCE(inet->mc_ttl, val);
1051 case IP_MTU_DISCOVER:
1052 return ip_sock_set_mtu_discover(sk, val);
1053 case IP_TOS: /* This sets both TOS and Precedence */
1054 ip_sock_set_tos(sk, val);
1056 case IP_LOCAL_PORT_RANGE:
1061 if (optlen != sizeof(u32))
1063 if (lo != 0 && hi != 0 && lo > hi)
1066 WRITE_ONCE(inet->local_port_range, val);
1074 sockopt_lock_sock(sk);
1079 struct ip_options_rcu *old, *opt = NULL;
1083 err = ip_options_get(sock_net(sk), &opt, optval, optlen);
1086 old = rcu_dereference_protected(inet->inet_opt,
1087 lockdep_sock_is_held(sk));
1088 if (inet_test_bit(IS_ICSK, sk)) {
1089 struct inet_connection_sock *icsk = inet_csk(sk);
1090 #if IS_ENABLED(CONFIG_IPV6)
1091 if (sk->sk_family == PF_INET ||
1092 (!((1 << sk->sk_state) &
1093 (TCPF_LISTEN | TCPF_CLOSE)) &&
1094 inet->inet_daddr != LOOPBACK4_IPV6)) {
1097 icsk->icsk_ext_hdr_len -= old->opt.optlen;
1099 icsk->icsk_ext_hdr_len += opt->opt.optlen;
1100 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
1101 #if IS_ENABLED(CONFIG_IPV6)
1105 rcu_assign_pointer(inet->inet_opt, opt);
1107 kfree_rcu(old, rcu);
1112 if (!(inet_test_bit(CHECKSUM, sk))) {
1113 inet_inc_convert_csum(sk);
1114 inet_set_bit(CHECKSUM, sk);
1117 if (inet_test_bit(CHECKSUM, sk)) {
1118 inet_dec_convert_csum(sk);
1119 inet_clear_bit(CHECKSUM, sk);
1125 struct net_device *dev = NULL;
1129 if (optlen != sizeof(int))
1132 ifindex = (__force int)ntohl((__force __be32)val);
1134 WRITE_ONCE(inet->uc_index, 0);
1139 dev = dev_get_by_index(sock_net(sk), ifindex);
1140 err = -EADDRNOTAVAIL;
1144 midx = l3mdev_master_ifindex(dev);
1148 if (sk->sk_bound_dev_if && midx != sk->sk_bound_dev_if)
1151 WRITE_ONCE(inet->uc_index, ifindex);
1155 case IP_MULTICAST_IF:
1157 struct ip_mreqn mreq;
1158 struct net_device *dev = NULL;
1161 if (sk->sk_type == SOCK_STREAM)
1164 * Check the arguments are allowable
1167 if (optlen < sizeof(struct in_addr))
1171 if (optlen >= sizeof(struct ip_mreqn)) {
1172 if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1175 memset(&mreq, 0, sizeof(mreq));
1176 if (optlen >= sizeof(struct ip_mreq)) {
1177 if (copy_from_sockptr(&mreq, optval,
1178 sizeof(struct ip_mreq)))
1180 } else if (optlen >= sizeof(struct in_addr)) {
1181 if (copy_from_sockptr(&mreq.imr_address, optval,
1182 sizeof(struct in_addr)))
1187 if (!mreq.imr_ifindex) {
1188 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
1189 WRITE_ONCE(inet->mc_index, 0);
1190 WRITE_ONCE(inet->mc_addr, 0);
1194 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
1196 mreq.imr_ifindex = dev->ifindex;
1198 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
1201 err = -EADDRNOTAVAIL;
1205 midx = l3mdev_master_ifindex(dev);
1210 if (sk->sk_bound_dev_if &&
1211 mreq.imr_ifindex != sk->sk_bound_dev_if &&
1212 midx != sk->sk_bound_dev_if)
1215 WRITE_ONCE(inet->mc_index, mreq.imr_ifindex);
1216 WRITE_ONCE(inet->mc_addr, mreq.imr_address.s_addr);
1221 case IP_ADD_MEMBERSHIP:
1222 case IP_DROP_MEMBERSHIP:
1224 struct ip_mreqn mreq;
1227 if (inet_test_bit(IS_ICSK, sk))
1230 if (optlen < sizeof(struct ip_mreq))
1233 if (optlen >= sizeof(struct ip_mreqn)) {
1234 if (copy_from_sockptr(&mreq, optval, sizeof(mreq)))
1237 memset(&mreq, 0, sizeof(mreq));
1238 if (copy_from_sockptr(&mreq, optval,
1239 sizeof(struct ip_mreq)))
1243 if (optname == IP_ADD_MEMBERSHIP)
1244 err = ip_mc_join_group(sk, &mreq);
1246 err = ip_mc_leave_group(sk, &mreq);
1251 struct ip_msfilter *msf;
1253 if (optlen < IP_MSFILTER_SIZE(0))
1255 if (optlen > READ_ONCE(net->core.sysctl_optmem_max)) {
1259 msf = memdup_sockptr(optval, optlen);
1264 /* numsrc >= (1G-4) overflow in 32 bits */
1265 if (msf->imsf_numsrc >= 0x3ffffffcU ||
1266 msf->imsf_numsrc > READ_ONCE(net->ipv4.sysctl_igmp_max_msf)) {
1271 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
1276 err = ip_mc_msfilter(sk, msf, 0);
1280 case IP_BLOCK_SOURCE:
1281 case IP_UNBLOCK_SOURCE:
1282 case IP_ADD_SOURCE_MEMBERSHIP:
1283 case IP_DROP_SOURCE_MEMBERSHIP:
1285 struct ip_mreq_source mreqs;
1288 if (optlen != sizeof(struct ip_mreq_source))
1290 if (copy_from_sockptr(&mreqs, optval, sizeof(mreqs))) {
1294 if (optname == IP_BLOCK_SOURCE) {
1295 omode = MCAST_EXCLUDE;
1297 } else if (optname == IP_UNBLOCK_SOURCE) {
1298 omode = MCAST_EXCLUDE;
1300 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
1301 struct ip_mreqn mreq;
1303 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
1304 mreq.imr_address.s_addr = mreqs.imr_interface;
1305 mreq.imr_ifindex = 0;
1306 err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
1307 if (err && err != -EADDRINUSE)
1309 omode = MCAST_INCLUDE;
1311 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
1312 omode = MCAST_INCLUDE;
1315 err = ip_mc_source(add, omode, sk, &mreqs, 0);
1318 case MCAST_JOIN_GROUP:
1319 case MCAST_LEAVE_GROUP:
1320 if (in_compat_syscall())
1321 err = compat_ip_mcast_join_leave(sk, optname, optval,
1324 err = ip_mcast_join_leave(sk, optname, optval, optlen);
1326 case MCAST_JOIN_SOURCE_GROUP:
1327 case MCAST_LEAVE_SOURCE_GROUP:
1328 case MCAST_BLOCK_SOURCE:
1329 case MCAST_UNBLOCK_SOURCE:
1330 err = do_mcast_group_source(sk, optname, optval, optlen);
1332 case MCAST_MSFILTER:
1333 if (in_compat_syscall())
1334 err = compat_ip_set_mcast_msfilter(sk, optval, optlen);
1336 err = ip_set_mcast_msfilter(sk, optval, optlen);
1338 case IP_IPSEC_POLICY:
1339 case IP_XFRM_POLICY:
1341 if (!sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1343 err = xfrm_user_policy(sk, optname, optval, optlen);
1350 sockopt_release_sock(sk);
1356 sockopt_release_sock(sk);
1363 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1366 * @drop_dst: if true, drops skb dst
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, bool drop_dst)
1374 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1375 bool prepare = inet_test_bit(PKTINFO, sk) ||
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;
1405 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1406 unsigned int optlen)
1410 if (level != SOL_IP)
1411 return -ENOPROTOOPT;
1413 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1414 #ifdef CONFIG_NETFILTER
1415 /* we need to exclude all possible ENOPROTOOPTs except default case */
1416 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1417 optname != IP_IPSEC_POLICY &&
1418 optname != IP_XFRM_POLICY &&
1419 !ip_mroute_opt(optname))
1420 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1424 EXPORT_SYMBOL(ip_setsockopt);
1427 * Get the options. Note for future reference. The GET of IP options gets
1428 * the _received_ ones. The set sets the _sent_ ones.
1431 static bool getsockopt_needs_rtnl(int optname)
1435 case MCAST_MSFILTER:
1441 static int ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1442 sockptr_t optlen, int len)
1444 const int size0 = offsetof(struct group_filter, gf_slist_flex);
1445 struct group_filter gsf;
1451 if (copy_from_sockptr(&gsf, optval, size0))
1454 num = gsf.gf_numsrc;
1455 err = ip_mc_gsfget(sk, &gsf, optval,
1456 offsetof(struct group_filter, gf_slist_flex));
1459 if (gsf.gf_numsrc < num)
1460 num = gsf.gf_numsrc;
1461 gsf_size = GROUP_FILTER_SIZE(num);
1462 if (copy_to_sockptr(optlen, &gsf_size, sizeof(int)) ||
1463 copy_to_sockptr(optval, &gsf, size0))
1468 static int compat_ip_get_mcast_msfilter(struct sock *sk, sockptr_t optval,
1469 sockptr_t optlen, int len)
1471 const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1472 struct compat_group_filter gf32;
1473 struct group_filter gf;
1479 if (copy_from_sockptr(&gf32, optval, size0))
1482 gf.gf_interface = gf32.gf_interface;
1483 gf.gf_fmode = gf32.gf_fmode;
1484 num = gf.gf_numsrc = gf32.gf_numsrc;
1485 gf.gf_group = gf32.gf_group;
1487 err = ip_mc_gsfget(sk, &gf, optval,
1488 offsetof(struct compat_group_filter, gf_slist_flex));
1491 if (gf.gf_numsrc < num)
1493 len = GROUP_FILTER_SIZE(num) - (sizeof(gf) - sizeof(gf32));
1494 if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1495 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1496 &gf.gf_fmode, sizeof(gf.gf_fmode)) ||
1497 copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1498 &gf.gf_numsrc, sizeof(gf.gf_numsrc)))
1503 int do_ip_getsockopt(struct sock *sk, int level, int optname,
1504 sockptr_t optval, sockptr_t optlen)
1506 struct inet_sock *inet = inet_sk(sk);
1507 bool needs_rtnl = getsockopt_needs_rtnl(optname);
1511 if (level != SOL_IP)
1514 if (ip_mroute_opt(optname))
1515 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1517 if (copy_from_sockptr(&len, optlen, sizeof(int)))
1522 /* Handle options that can be read without locking the socket. */
1525 val = inet_test_bit(PKTINFO, sk);
1528 val = inet_test_bit(TTL, sk);
1531 val = inet_test_bit(TOS, sk);
1534 val = inet_test_bit(RECVOPTS, sk);
1537 val = inet_test_bit(RETOPTS, sk);
1540 val = inet_test_bit(PASSSEC, sk);
1542 case IP_RECVORIGDSTADDR:
1543 val = inet_test_bit(ORIGDSTADDR, sk);
1546 val = inet_test_bit(CHECKSUM, sk);
1548 case IP_RECVFRAGSIZE:
1549 val = inet_test_bit(RECVFRAGSIZE, sk);
1552 val = inet_test_bit(RECVERR, sk);
1554 case IP_RECVERR_RFC4884:
1555 val = inet_test_bit(RECVERR_RFC4884, sk);
1558 val = inet_test_bit(FREEBIND, sk);
1561 val = inet_test_bit(HDRINCL, sk);
1563 case IP_MULTICAST_LOOP:
1564 val = inet_test_bit(MC_LOOP, sk);
1566 case IP_MULTICAST_ALL:
1567 val = inet_test_bit(MC_ALL, sk);
1569 case IP_TRANSPARENT:
1570 val = inet_test_bit(TRANSPARENT, sk);
1573 val = inet_test_bit(NODEFRAG, sk);
1575 case IP_BIND_ADDRESS_NO_PORT:
1576 val = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
1579 val = READ_ONCE(inet->uc_ttl);
1581 val = READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_default_ttl);
1584 val = READ_ONCE(inet->min_ttl);
1586 case IP_MULTICAST_TTL:
1587 val = READ_ONCE(inet->mc_ttl);
1589 case IP_MTU_DISCOVER:
1590 val = READ_ONCE(inet->pmtudisc);
1593 val = READ_ONCE(inet->tos);
1597 unsigned char optbuf[sizeof(struct ip_options)+40];
1598 struct ip_options *opt = (struct ip_options *)optbuf;
1599 struct ip_options_rcu *inet_opt;
1602 inet_opt = rcu_dereference(inet->inet_opt);
1605 memcpy(optbuf, &inet_opt->opt,
1606 sizeof(struct ip_options) +
1607 inet_opt->opt.optlen);
1610 if (opt->optlen == 0) {
1612 return copy_to_sockptr(optlen, &len, sizeof(int));
1615 ip_options_undo(opt);
1617 len = min_t(unsigned int, len, opt->optlen);
1618 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1620 if (copy_to_sockptr(optval, opt->__data, len))
1626 struct dst_entry *dst;
1628 dst = sk_dst_get(sk);
1641 if (sk->sk_type != SOCK_STREAM)
1642 return -ENOPROTOOPT;
1644 if (optval.is_kernel) {
1645 msg.msg_control_is_user = false;
1646 msg.msg_control = optval.kernel;
1648 msg.msg_control_is_user = true;
1649 msg.msg_control_user = optval.user;
1651 msg.msg_controllen = len;
1652 msg.msg_flags = in_compat_syscall() ? MSG_CMSG_COMPAT : 0;
1654 if (inet_test_bit(PKTINFO, sk)) {
1655 struct in_pktinfo info;
1657 info.ipi_addr.s_addr = READ_ONCE(inet->inet_rcv_saddr);
1658 info.ipi_spec_dst.s_addr = READ_ONCE(inet->inet_rcv_saddr);
1659 info.ipi_ifindex = READ_ONCE(inet->mc_index);
1660 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1662 if (inet_test_bit(TTL, sk)) {
1663 int hlim = READ_ONCE(inet->mc_ttl);
1665 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1667 if (inet_test_bit(TOS, sk)) {
1668 int tos = READ_ONCE(inet->rcv_tos);
1669 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1671 len -= msg.msg_controllen;
1672 return copy_to_sockptr(optlen, &len, sizeof(int));
1675 val = (__force int)htonl((__u32) READ_ONCE(inet->uc_index));
1677 case IP_MULTICAST_IF:
1679 struct in_addr addr;
1680 len = min_t(unsigned int, len, sizeof(struct in_addr));
1681 addr.s_addr = READ_ONCE(inet->mc_addr);
1683 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1685 if (copy_to_sockptr(optval, &addr, len))
1689 case IP_LOCAL_PORT_RANGE:
1690 val = READ_ONCE(inet->local_port_range);
1696 sockopt_lock_sock(sk);
1701 struct ip_msfilter msf;
1703 if (len < IP_MSFILTER_SIZE(0)) {
1707 if (copy_from_sockptr(&msf, optval, IP_MSFILTER_SIZE(0))) {
1711 err = ip_mc_msfget(sk, &msf, optval, optlen);
1714 case MCAST_MSFILTER:
1715 if (in_compat_syscall())
1716 err = compat_ip_get_mcast_msfilter(sk, optval, optlen,
1719 err = ip_get_mcast_msfilter(sk, optval, optlen, len);
1722 val = inet_sk(sk)->inet_num;
1725 sockopt_release_sock(sk);
1726 return -ENOPROTOOPT;
1728 sockopt_release_sock(sk);
1730 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1731 unsigned char ucval = (unsigned char)val;
1733 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1735 if (copy_to_sockptr(optval, &ucval, 1))
1738 len = min_t(unsigned int, sizeof(int), len);
1739 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1741 if (copy_to_sockptr(optval, &val, len))
1747 sockopt_release_sock(sk);
1753 int ip_getsockopt(struct sock *sk, int level,
1754 int optname, char __user *optval, int __user *optlen)
1758 err = do_ip_getsockopt(sk, level, optname,
1759 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1761 #ifdef CONFIG_NETFILTER
1762 /* we need to exclude all possible ENOPROTOOPTs except default case */
1763 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1764 !ip_mroute_opt(optname)) {
1767 if (get_user(len, optlen))
1770 err = nf_getsockopt(sk, PF_INET, optname, optval, &len);
1772 err = put_user(len, optlen);
1778 EXPORT_SYMBOL(ip_getsockopt);