2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
11 * Many : Split from ip.c , see ip.c for history.
12 * Martin Mares : TOS setting fixed.
13 * Alan Cox : Fixed a couple of oopses in Martin's
15 * Mike McLagan : Routing by source
18 #include <linux/module.h>
19 #include <linux/types.h>
21 #include <linux/skbuff.h>
23 #include <linux/icmp.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/slab.h>
30 #include <net/tcp_states.h>
31 #include <linux/udp.h>
32 #include <linux/igmp.h>
33 #include <linux/netfilter.h>
34 #include <linux/route.h>
35 #include <linux/mroute.h>
36 #include <net/inet_ecn.h>
37 #include <net/route.h>
39 #include <net/compat.h>
40 #if IS_ENABLED(CONFIG_IPV6)
41 #include <net/transp_v6.h>
43 #include <net/ip_fib.h>
45 #include <linux/errqueue.h>
46 #include <asm/uaccess.h>
48 #define IP_CMSG_PKTINFO 1
51 #define IP_CMSG_RECVOPTS 8
52 #define IP_CMSG_RETOPTS 16
53 #define IP_CMSG_PASSSEC 32
54 #define IP_CMSG_ORIGDSTADDR 64
57 * SOL_IP control messages.
59 #define PKTINFO_SKB_CB(__skb) ((struct in_pktinfo *)((__skb)->cb))
61 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
63 struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
65 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
67 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
70 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
72 int ttl = ip_hdr(skb)->ttl;
73 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
76 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
78 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
81 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
83 if (IPCB(skb)->opt.optlen == 0)
86 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
91 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
93 unsigned char optbuf[sizeof(struct ip_options) + 40];
94 struct ip_options *opt = (struct ip_options *)optbuf;
96 if (IPCB(skb)->opt.optlen == 0)
99 if (ip_options_echo(opt, skb)) {
100 msg->msg_flags |= MSG_CTRUNC;
103 ip_options_undo(opt);
105 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
108 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
114 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
118 err = security_secid_to_secctx(secid, &secdata, &seclen);
122 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
123 security_release_secctx(secdata, seclen);
126 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
128 struct sockaddr_in sin;
129 const struct iphdr *iph = ip_hdr(skb);
130 __be16 *ports = (__be16 *)skb_transport_header(skb);
132 if (skb_transport_offset(skb) + 4 > skb->len)
135 /* All current transport protocols have the port numbers in the
136 * first four bytes of the transport header and this function is
137 * written with this assumption in mind.
140 sin.sin_family = AF_INET;
141 sin.sin_addr.s_addr = iph->daddr;
142 sin.sin_port = ports[1];
143 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
145 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
148 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
150 struct inet_sock *inet = inet_sk(skb->sk);
151 unsigned int flags = inet->cmsg_flags;
153 /* Ordered by supposed usage frequency */
155 ip_cmsg_recv_pktinfo(msg, skb);
156 if ((flags >>= 1) == 0)
160 ip_cmsg_recv_ttl(msg, skb);
161 if ((flags >>= 1) == 0)
165 ip_cmsg_recv_tos(msg, skb);
166 if ((flags >>= 1) == 0)
170 ip_cmsg_recv_opts(msg, skb);
171 if ((flags >>= 1) == 0)
175 ip_cmsg_recv_retopts(msg, skb);
176 if ((flags >>= 1) == 0)
180 ip_cmsg_recv_security(msg, skb);
182 if ((flags >>= 1) == 0)
185 ip_cmsg_recv_dstaddr(msg, skb);
188 EXPORT_SYMBOL(ip_cmsg_recv);
190 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
193 struct cmsghdr *cmsg;
195 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
196 if (!CMSG_OK(msg, cmsg))
198 if (cmsg->cmsg_level != SOL_IP)
200 switch (cmsg->cmsg_type) {
202 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
203 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
204 err < 40 ? err : 40);
210 struct in_pktinfo *info;
211 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
213 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
214 ipc->oif = info->ipi_ifindex;
215 ipc->addr = info->ipi_spec_dst.s_addr;
226 /* Special input handler for packets caught by router alert option.
227 They are selected only by protocol field, and then processed likely
228 local ones; but only if someone wants them! Otherwise, router
229 not running rsvpd will kill RSVP.
231 It is user level problem, what it will make with them.
232 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
233 but receiver should be enough clever f.e. to forward mtrace requests,
234 sent to multicast group to reach destination designated router.
236 struct ip_ra_chain __rcu *ip_ra_chain;
237 static DEFINE_SPINLOCK(ip_ra_lock);
240 static void ip_ra_destroy_rcu(struct rcu_head *head)
242 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
244 sock_put(ra->saved_sk);
248 int ip_ra_control(struct sock *sk, unsigned char on,
249 void (*destructor)(struct sock *))
251 struct ip_ra_chain *ra, *new_ra;
252 struct ip_ra_chain __rcu **rap;
254 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
257 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
259 spin_lock_bh(&ip_ra_lock);
260 for (rap = &ip_ra_chain;
261 (ra = rcu_dereference_protected(*rap,
262 lockdep_is_held(&ip_ra_lock))) != NULL;
266 spin_unlock_bh(&ip_ra_lock);
270 /* dont let ip_call_ra_chain() use sk again */
272 rcu_assign_pointer(*rap, ra->next);
273 spin_unlock_bh(&ip_ra_lock);
278 * Delay sock_put(sk) and kfree(ra) after one rcu grace
279 * period. This guarantee ip_call_ra_chain() dont need
280 * to mess with socket refcounts.
283 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
287 if (new_ra == NULL) {
288 spin_unlock_bh(&ip_ra_lock);
292 new_ra->destructor = destructor;
295 rcu_assign_pointer(*rap, new_ra);
297 spin_unlock_bh(&ip_ra_lock);
302 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
303 __be16 port, u32 info, u8 *payload)
305 struct sock_exterr_skb *serr;
307 skb = skb_clone(skb, GFP_ATOMIC);
311 serr = SKB_EXT_ERR(skb);
312 serr->ee.ee_errno = err;
313 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
314 serr->ee.ee_type = icmp_hdr(skb)->type;
315 serr->ee.ee_code = icmp_hdr(skb)->code;
317 serr->ee.ee_info = info;
318 serr->ee.ee_data = 0;
319 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
320 skb_network_header(skb);
323 if (skb_pull(skb, payload - skb->data) != NULL) {
324 skb_reset_transport_header(skb);
325 if (sock_queue_err_skb(sk, skb) == 0)
331 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
333 struct inet_sock *inet = inet_sk(sk);
334 struct sock_exterr_skb *serr;
341 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
345 skb_put(skb, sizeof(struct iphdr));
346 skb_reset_network_header(skb);
350 serr = SKB_EXT_ERR(skb);
351 serr->ee.ee_errno = err;
352 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
353 serr->ee.ee_type = 0;
354 serr->ee.ee_code = 0;
356 serr->ee.ee_info = info;
357 serr->ee.ee_data = 0;
358 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
361 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
362 skb_reset_transport_header(skb);
364 if (sock_queue_err_skb(sk, skb))
369 * Handle MSG_ERRQUEUE
371 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
373 struct sock_exterr_skb *serr;
374 struct sk_buff *skb, *skb2;
375 struct sockaddr_in *sin;
377 struct sock_extended_err ee;
378 struct sockaddr_in offender;
384 skb = skb_dequeue(&sk->sk_error_queue);
390 msg->msg_flags |= MSG_TRUNC;
393 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
397 sock_recv_timestamp(msg, sk, skb);
399 serr = SKB_EXT_ERR(skb);
401 sin = (struct sockaddr_in *)msg->msg_name;
403 sin->sin_family = AF_INET;
404 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
406 sin->sin_port = serr->port;
407 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
410 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
411 sin = &errhdr.offender;
412 sin->sin_family = AF_UNSPEC;
413 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
414 struct inet_sock *inet = inet_sk(sk);
416 sin->sin_family = AF_INET;
417 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
419 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
420 if (inet->cmsg_flags)
421 ip_cmsg_recv(msg, skb);
424 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
426 /* Now we could try to dump offended packet options */
428 msg->msg_flags |= MSG_ERRQUEUE;
431 /* Reset and regenerate socket error */
432 spin_lock_bh(&sk->sk_error_queue.lock);
434 skb2 = skb_peek(&sk->sk_error_queue);
436 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
437 spin_unlock_bh(&sk->sk_error_queue.lock);
438 sk->sk_error_report(sk);
440 spin_unlock_bh(&sk->sk_error_queue.lock);
450 * Socket option code for IP. This is the end of the line after any
451 * TCP,UDP etc options on an IP socket.
454 static int do_ip_setsockopt(struct sock *sk, int level,
455 int optname, char __user *optval, unsigned int optlen)
457 struct inet_sock *inet = inet_sk(sk);
460 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
461 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
462 (1<<IP_RETOPTS) | (1<<IP_TOS) |
463 (1<<IP_TTL) | (1<<IP_HDRINCL) |
464 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
465 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
466 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) |
467 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) ||
468 optname == IP_UNICAST_IF ||
469 optname == IP_MULTICAST_TTL ||
470 optname == IP_MULTICAST_ALL ||
471 optname == IP_MULTICAST_LOOP ||
472 optname == IP_RECVORIGDSTADDR) {
473 if (optlen >= sizeof(int)) {
474 if (get_user(val, (int __user *) optval))
476 } else if (optlen >= sizeof(char)) {
479 if (get_user(ucval, (unsigned char __user *) optval))
485 /* If optlen==0, it is equivalent to val == 0 */
487 if (ip_mroute_opt(optname))
488 return ip_mroute_setsockopt(sk, optname, optval, optlen);
496 struct ip_options_rcu *old, *opt = NULL;
500 err = ip_options_get_from_user(sock_net(sk), &opt,
504 old = rcu_dereference_protected(inet->inet_opt,
505 sock_owned_by_user(sk));
507 struct inet_connection_sock *icsk = inet_csk(sk);
508 #if IS_ENABLED(CONFIG_IPV6)
509 if (sk->sk_family == PF_INET ||
510 (!((1 << sk->sk_state) &
511 (TCPF_LISTEN | TCPF_CLOSE)) &&
512 inet->inet_daddr != LOOPBACK4_IPV6)) {
515 icsk->icsk_ext_hdr_len -= old->opt.optlen;
517 icsk->icsk_ext_hdr_len += opt->opt.optlen;
518 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
519 #if IS_ENABLED(CONFIG_IPV6)
523 rcu_assign_pointer(inet->inet_opt, opt);
530 inet->cmsg_flags |= IP_CMSG_PKTINFO;
532 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
536 inet->cmsg_flags |= IP_CMSG_TTL;
538 inet->cmsg_flags &= ~IP_CMSG_TTL;
542 inet->cmsg_flags |= IP_CMSG_TOS;
544 inet->cmsg_flags &= ~IP_CMSG_TOS;
548 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
550 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
554 inet->cmsg_flags |= IP_CMSG_RETOPTS;
556 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
560 inet->cmsg_flags |= IP_CMSG_PASSSEC;
562 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
564 case IP_RECVORIGDSTADDR:
566 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
568 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
570 case IP_TOS: /* This sets both TOS and Precedence */
571 if (sk->sk_type == SOCK_STREAM) {
572 val &= ~INET_ECN_MASK;
573 val |= inet->tos & INET_ECN_MASK;
575 if (inet->tos != val) {
577 sk->sk_priority = rt_tos2priority(val);
584 if (val != -1 && (val < 0 || val > 255))
589 if (sk->sk_type != SOCK_RAW) {
593 inet->hdrincl = val ? 1 : 0;
596 if (sk->sk_type != SOCK_RAW) {
600 inet->nodefrag = val ? 1 : 0;
602 case IP_MTU_DISCOVER:
603 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
605 inet->pmtudisc = val;
608 inet->recverr = !!val;
610 skb_queue_purge(&sk->sk_error_queue);
612 case IP_MULTICAST_TTL:
613 if (sk->sk_type == SOCK_STREAM)
619 if (val < 0 || val > 255)
623 case IP_MULTICAST_LOOP:
626 inet->mc_loop = !!val;
630 struct net_device *dev = NULL;
633 if (optlen != sizeof(int))
636 ifindex = (__force int)ntohl((__force __be32)val);
643 dev = dev_get_by_index(sock_net(sk), ifindex);
644 err = -EADDRNOTAVAIL;
650 if (sk->sk_bound_dev_if)
653 inet->uc_index = ifindex;
657 case IP_MULTICAST_IF:
659 struct ip_mreqn mreq;
660 struct net_device *dev = NULL;
662 if (sk->sk_type == SOCK_STREAM)
665 * Check the arguments are allowable
668 if (optlen < sizeof(struct in_addr))
672 if (optlen >= sizeof(struct ip_mreqn)) {
673 if (copy_from_user(&mreq, optval, sizeof(mreq)))
676 memset(&mreq, 0, sizeof(mreq));
677 if (optlen >= sizeof(struct ip_mreq)) {
678 if (copy_from_user(&mreq, optval,
679 sizeof(struct ip_mreq)))
681 } else if (optlen >= sizeof(struct in_addr)) {
682 if (copy_from_user(&mreq.imr_address, optval,
683 sizeof(struct in_addr)))
688 if (!mreq.imr_ifindex) {
689 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
695 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
697 mreq.imr_ifindex = dev->ifindex;
699 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
702 err = -EADDRNOTAVAIL;
708 if (sk->sk_bound_dev_if &&
709 mreq.imr_ifindex != sk->sk_bound_dev_if)
712 inet->mc_index = mreq.imr_ifindex;
713 inet->mc_addr = mreq.imr_address.s_addr;
718 case IP_ADD_MEMBERSHIP:
719 case IP_DROP_MEMBERSHIP:
721 struct ip_mreqn mreq;
724 if (inet_sk(sk)->is_icsk)
727 if (optlen < sizeof(struct ip_mreq))
730 if (optlen >= sizeof(struct ip_mreqn)) {
731 if (copy_from_user(&mreq, optval, sizeof(mreq)))
734 memset(&mreq, 0, sizeof(mreq));
735 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
739 if (optname == IP_ADD_MEMBERSHIP)
740 err = ip_mc_join_group(sk, &mreq);
742 err = ip_mc_leave_group(sk, &mreq);
747 struct ip_msfilter *msf;
749 if (optlen < IP_MSFILTER_SIZE(0))
751 if (optlen > sysctl_optmem_max) {
755 msf = kmalloc(optlen, GFP_KERNEL);
761 if (copy_from_user(msf, optval, optlen)) {
765 /* numsrc >= (1G-4) overflow in 32 bits */
766 if (msf->imsf_numsrc >= 0x3ffffffcU ||
767 msf->imsf_numsrc > sysctl_igmp_max_msf) {
772 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
777 err = ip_mc_msfilter(sk, msf, 0);
781 case IP_BLOCK_SOURCE:
782 case IP_UNBLOCK_SOURCE:
783 case IP_ADD_SOURCE_MEMBERSHIP:
784 case IP_DROP_SOURCE_MEMBERSHIP:
786 struct ip_mreq_source mreqs;
789 if (optlen != sizeof(struct ip_mreq_source))
791 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
795 if (optname == IP_BLOCK_SOURCE) {
796 omode = MCAST_EXCLUDE;
798 } else if (optname == IP_UNBLOCK_SOURCE) {
799 omode = MCAST_EXCLUDE;
801 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
802 struct ip_mreqn mreq;
804 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
805 mreq.imr_address.s_addr = mreqs.imr_interface;
806 mreq.imr_ifindex = 0;
807 err = ip_mc_join_group(sk, &mreq);
808 if (err && err != -EADDRINUSE)
810 omode = MCAST_INCLUDE;
812 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
813 omode = MCAST_INCLUDE;
816 err = ip_mc_source(add, omode, sk, &mreqs, 0);
819 case MCAST_JOIN_GROUP:
820 case MCAST_LEAVE_GROUP:
822 struct group_req greq;
823 struct sockaddr_in *psin;
824 struct ip_mreqn mreq;
826 if (optlen < sizeof(struct group_req))
829 if (copy_from_user(&greq, optval, sizeof(greq)))
831 psin = (struct sockaddr_in *)&greq.gr_group;
832 if (psin->sin_family != AF_INET)
834 memset(&mreq, 0, sizeof(mreq));
835 mreq.imr_multiaddr = psin->sin_addr;
836 mreq.imr_ifindex = greq.gr_interface;
838 if (optname == MCAST_JOIN_GROUP)
839 err = ip_mc_join_group(sk, &mreq);
841 err = ip_mc_leave_group(sk, &mreq);
844 case MCAST_JOIN_SOURCE_GROUP:
845 case MCAST_LEAVE_SOURCE_GROUP:
846 case MCAST_BLOCK_SOURCE:
847 case MCAST_UNBLOCK_SOURCE:
849 struct group_source_req greqs;
850 struct ip_mreq_source mreqs;
851 struct sockaddr_in *psin;
854 if (optlen != sizeof(struct group_source_req))
856 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
860 if (greqs.gsr_group.ss_family != AF_INET ||
861 greqs.gsr_source.ss_family != AF_INET) {
862 err = -EADDRNOTAVAIL;
865 psin = (struct sockaddr_in *)&greqs.gsr_group;
866 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
867 psin = (struct sockaddr_in *)&greqs.gsr_source;
868 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
869 mreqs.imr_interface = 0; /* use index for mc_source */
871 if (optname == MCAST_BLOCK_SOURCE) {
872 omode = MCAST_EXCLUDE;
874 } else if (optname == MCAST_UNBLOCK_SOURCE) {
875 omode = MCAST_EXCLUDE;
877 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
878 struct ip_mreqn mreq;
880 psin = (struct sockaddr_in *)&greqs.gsr_group;
881 mreq.imr_multiaddr = psin->sin_addr;
882 mreq.imr_address.s_addr = 0;
883 mreq.imr_ifindex = greqs.gsr_interface;
884 err = ip_mc_join_group(sk, &mreq);
885 if (err && err != -EADDRINUSE)
887 greqs.gsr_interface = mreq.imr_ifindex;
888 omode = MCAST_INCLUDE;
890 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
891 omode = MCAST_INCLUDE;
894 err = ip_mc_source(add, omode, sk, &mreqs,
895 greqs.gsr_interface);
900 struct sockaddr_in *psin;
901 struct ip_msfilter *msf = NULL;
902 struct group_filter *gsf = NULL;
903 int msize, i, ifindex;
905 if (optlen < GROUP_FILTER_SIZE(0))
907 if (optlen > sysctl_optmem_max) {
911 gsf = kmalloc(optlen, GFP_KERNEL);
917 if (copy_from_user(gsf, optval, optlen))
920 /* numsrc >= (4G-140)/128 overflow in 32 bits */
921 if (gsf->gf_numsrc >= 0x1ffffff ||
922 gsf->gf_numsrc > sysctl_igmp_max_msf) {
926 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
930 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
931 msf = kmalloc(msize, GFP_KERNEL);
936 ifindex = gsf->gf_interface;
937 psin = (struct sockaddr_in *)&gsf->gf_group;
938 if (psin->sin_family != AF_INET) {
939 err = -EADDRNOTAVAIL;
942 msf->imsf_multiaddr = psin->sin_addr.s_addr;
943 msf->imsf_interface = 0;
944 msf->imsf_fmode = gsf->gf_fmode;
945 msf->imsf_numsrc = gsf->gf_numsrc;
946 err = -EADDRNOTAVAIL;
947 for (i = 0; i < gsf->gf_numsrc; ++i) {
948 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
950 if (psin->sin_family != AF_INET)
952 msf->imsf_slist[i] = psin->sin_addr.s_addr;
957 err = ip_mc_msfilter(sk, msf, ifindex);
963 case IP_MULTICAST_ALL:
966 if (val != 0 && val != 1)
970 case IP_ROUTER_ALERT:
971 err = ip_ra_control(sk, val ? 1 : 0, NULL);
977 inet->freebind = !!val;
980 case IP_IPSEC_POLICY:
983 if (!capable(CAP_NET_ADMIN))
985 err = xfrm_user_policy(sk, optname, optval, optlen);
989 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) {
995 inet->transparent = !!val;
1001 if (val < 0 || val > 255)
1003 inet->min_ttl = val;
1019 * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1023 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1024 * destination in skb->cb[] before dst drop.
1025 * This way, receiver doesnt make cache line misses to read rtable.
1027 void ipv4_pktinfo_prepare(struct sk_buff *skb)
1029 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1030 const struct rtable *rt = skb_rtable(skb);
1033 pktinfo->ipi_ifindex = rt->rt_iif;
1034 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1036 pktinfo->ipi_ifindex = 0;
1037 pktinfo->ipi_spec_dst.s_addr = 0;
1042 int ip_setsockopt(struct sock *sk, int level,
1043 int optname, char __user *optval, unsigned int optlen)
1047 if (level != SOL_IP)
1048 return -ENOPROTOOPT;
1050 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1051 #ifdef CONFIG_NETFILTER
1052 /* we need to exclude all possible ENOPROTOOPTs except default case */
1053 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1054 optname != IP_IPSEC_POLICY &&
1055 optname != IP_XFRM_POLICY &&
1056 !ip_mroute_opt(optname)) {
1058 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1064 EXPORT_SYMBOL(ip_setsockopt);
1066 #ifdef CONFIG_COMPAT
1067 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1068 char __user *optval, unsigned int optlen)
1072 if (level != SOL_IP)
1073 return -ENOPROTOOPT;
1075 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1076 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1079 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1080 #ifdef CONFIG_NETFILTER
1081 /* we need to exclude all possible ENOPROTOOPTs except default case */
1082 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1083 optname != IP_IPSEC_POLICY &&
1084 optname != IP_XFRM_POLICY &&
1085 !ip_mroute_opt(optname)) {
1087 err = compat_nf_setsockopt(sk, PF_INET, optname,
1094 EXPORT_SYMBOL(compat_ip_setsockopt);
1098 * Get the options. Note for future reference. The GET of IP options gets
1099 * the _received_ ones. The set sets the _sent_ ones.
1102 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1103 char __user *optval, int __user *optlen, unsigned int flags)
1105 struct inet_sock *inet = inet_sk(sk);
1109 if (level != SOL_IP)
1112 if (ip_mroute_opt(optname))
1113 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1115 if (get_user(len, optlen))
1125 unsigned char optbuf[sizeof(struct ip_options)+40];
1126 struct ip_options *opt = (struct ip_options *)optbuf;
1127 struct ip_options_rcu *inet_opt;
1129 inet_opt = rcu_dereference_protected(inet->inet_opt,
1130 sock_owned_by_user(sk));
1133 memcpy(optbuf, &inet_opt->opt,
1134 sizeof(struct ip_options) +
1135 inet_opt->opt.optlen);
1138 if (opt->optlen == 0)
1139 return put_user(0, optlen);
1141 ip_options_undo(opt);
1143 len = min_t(unsigned int, len, opt->optlen);
1144 if (put_user(len, optlen))
1146 if (copy_to_user(optval, opt->__data, len))
1151 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1154 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1157 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1160 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1163 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1166 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1168 case IP_RECVORIGDSTADDR:
1169 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1175 val = (inet->uc_ttl == -1 ?
1176 sysctl_ip_default_ttl :
1180 val = inet->hdrincl;
1183 val = inet->nodefrag;
1185 case IP_MTU_DISCOVER:
1186 val = inet->pmtudisc;
1190 struct dst_entry *dst;
1192 dst = sk_dst_get(sk);
1204 val = inet->recverr;
1206 case IP_MULTICAST_TTL:
1209 case IP_MULTICAST_LOOP:
1210 val = inet->mc_loop;
1213 val = (__force int)htonl((__u32) inet->uc_index);
1215 case IP_MULTICAST_IF:
1217 struct in_addr addr;
1218 len = min_t(unsigned int, len, sizeof(struct in_addr));
1219 addr.s_addr = inet->mc_addr;
1222 if (put_user(len, optlen))
1224 if (copy_to_user(optval, &addr, len))
1230 struct ip_msfilter msf;
1233 if (len < IP_MSFILTER_SIZE(0)) {
1237 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1241 err = ip_mc_msfget(sk, &msf,
1242 (struct ip_msfilter __user *)optval, optlen);
1246 case MCAST_MSFILTER:
1248 struct group_filter gsf;
1251 if (len < GROUP_FILTER_SIZE(0)) {
1255 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1259 err = ip_mc_gsfget(sk, &gsf,
1260 (struct group_filter __user *)optval,
1265 case IP_MULTICAST_ALL:
1274 if (sk->sk_type != SOCK_STREAM)
1275 return -ENOPROTOOPT;
1277 msg.msg_control = optval;
1278 msg.msg_controllen = len;
1279 msg.msg_flags = flags;
1281 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1282 struct in_pktinfo info;
1284 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1285 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1286 info.ipi_ifindex = inet->mc_index;
1287 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1289 if (inet->cmsg_flags & IP_CMSG_TTL) {
1290 int hlim = inet->mc_ttl;
1291 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1293 if (inet->cmsg_flags & IP_CMSG_TOS) {
1294 int tos = inet->rcv_tos;
1295 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1297 len -= msg.msg_controllen;
1298 return put_user(len, optlen);
1301 val = inet->freebind;
1303 case IP_TRANSPARENT:
1304 val = inet->transparent;
1307 val = inet->min_ttl;
1311 return -ENOPROTOOPT;
1315 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1316 unsigned char ucval = (unsigned char)val;
1318 if (put_user(len, optlen))
1320 if (copy_to_user(optval, &ucval, 1))
1323 len = min_t(unsigned int, sizeof(int), len);
1324 if (put_user(len, optlen))
1326 if (copy_to_user(optval, &val, len))
1332 int ip_getsockopt(struct sock *sk, int level,
1333 int optname, char __user *optval, int __user *optlen)
1337 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1338 #ifdef CONFIG_NETFILTER
1339 /* we need to exclude all possible ENOPROTOOPTs except default case */
1340 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1341 !ip_mroute_opt(optname)) {
1344 if (get_user(len, optlen))
1348 err = nf_getsockopt(sk, PF_INET, optname, optval,
1352 err = put_user(len, optlen);
1358 EXPORT_SYMBOL(ip_getsockopt);
1360 #ifdef CONFIG_COMPAT
1361 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1362 char __user *optval, int __user *optlen)
1366 if (optname == MCAST_MSFILTER)
1367 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1370 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1373 #ifdef CONFIG_NETFILTER
1374 /* we need to exclude all possible ENOPROTOOPTs except default case */
1375 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1376 !ip_mroute_opt(optname)) {
1379 if (get_user(len, optlen))
1383 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1386 err = put_user(len, optlen);
1392 EXPORT_SYMBOL(compat_ip_getsockopt);