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 defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
41 #include <net/transp_v6.h>
44 #include <linux/errqueue.h>
45 #include <asm/uaccess.h>
47 #define IP_CMSG_PKTINFO 1
50 #define IP_CMSG_RECVOPTS 8
51 #define IP_CMSG_RETOPTS 16
52 #define IP_CMSG_PASSSEC 32
53 #define IP_CMSG_ORIGDSTADDR 64
56 * SOL_IP control messages.
59 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
61 struct in_pktinfo info;
62 struct rtable *rt = skb_rtable(skb);
64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
66 info.ipi_ifindex = rt->rt_iif;
67 info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
70 info.ipi_spec_dst.s_addr = 0;
73 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
76 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
78 int ttl = ip_hdr(skb)->ttl;
79 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
82 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
84 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
87 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
89 if (IPCB(skb)->opt.optlen == 0)
92 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
97 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
99 unsigned char optbuf[sizeof(struct ip_options) + 40];
100 struct ip_options * opt = (struct ip_options *)optbuf;
102 if (IPCB(skb)->opt.optlen == 0)
105 if (ip_options_echo(opt, skb)) {
106 msg->msg_flags |= MSG_CTRUNC;
109 ip_options_undo(opt);
111 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
114 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
120 err = security_socket_getpeersec_dgram(NULL, skb, &secid);
124 err = security_secid_to_secctx(secid, &secdata, &seclen);
128 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
129 security_release_secctx(secdata, seclen);
132 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
134 struct sockaddr_in sin;
135 const struct iphdr *iph = ip_hdr(skb);
136 __be16 *ports = (__be16 *)skb_transport_header(skb);
138 if (skb_transport_offset(skb) + 4 > skb->len)
141 /* All current transport protocols have the port numbers in the
142 * first four bytes of the transport header and this function is
143 * written with this assumption in mind.
146 sin.sin_family = AF_INET;
147 sin.sin_addr.s_addr = iph->daddr;
148 sin.sin_port = ports[1];
149 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
151 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
154 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
156 struct inet_sock *inet = inet_sk(skb->sk);
157 unsigned flags = inet->cmsg_flags;
159 /* Ordered by supposed usage frequency */
161 ip_cmsg_recv_pktinfo(msg, skb);
162 if ((flags >>= 1) == 0)
166 ip_cmsg_recv_ttl(msg, skb);
167 if ((flags >>= 1) == 0)
171 ip_cmsg_recv_tos(msg, skb);
172 if ((flags >>= 1) == 0)
176 ip_cmsg_recv_opts(msg, skb);
177 if ((flags >>= 1) == 0)
181 ip_cmsg_recv_retopts(msg, skb);
182 if ((flags >>= 1) == 0)
186 ip_cmsg_recv_security(msg, skb);
188 if ((flags >>= 1) == 0)
191 ip_cmsg_recv_dstaddr(msg, skb);
194 EXPORT_SYMBOL(ip_cmsg_recv);
196 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
199 struct cmsghdr *cmsg;
201 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
202 if (!CMSG_OK(msg, cmsg))
204 if (cmsg->cmsg_level != SOL_IP)
206 switch (cmsg->cmsg_type) {
208 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
209 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
210 err < 40 ? err : 40);
216 struct in_pktinfo *info;
217 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
219 info = (struct in_pktinfo *)CMSG_DATA(cmsg);
220 ipc->oif = info->ipi_ifindex;
221 ipc->addr = info->ipi_spec_dst.s_addr;
232 /* Special input handler for packets caught by router alert option.
233 They are selected only by protocol field, and then processed likely
234 local ones; but only if someone wants them! Otherwise, router
235 not running rsvpd will kill RSVP.
237 It is user level problem, what it will make with them.
238 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
239 but receiver should be enough clever f.e. to forward mtrace requests,
240 sent to multicast group to reach destination designated router.
242 struct ip_ra_chain __rcu *ip_ra_chain;
243 static DEFINE_SPINLOCK(ip_ra_lock);
246 static void ip_ra_destroy_rcu(struct rcu_head *head)
248 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
250 sock_put(ra->saved_sk);
254 int ip_ra_control(struct sock *sk, unsigned char on,
255 void (*destructor)(struct sock *))
257 struct ip_ra_chain *ra, *new_ra;
258 struct ip_ra_chain __rcu **rap;
260 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
263 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
265 spin_lock_bh(&ip_ra_lock);
266 for (rap = &ip_ra_chain;
267 (ra = rcu_dereference_protected(*rap,
268 lockdep_is_held(&ip_ra_lock))) != NULL;
272 spin_unlock_bh(&ip_ra_lock);
276 /* dont let ip_call_ra_chain() use sk again */
278 rcu_assign_pointer(*rap, ra->next);
279 spin_unlock_bh(&ip_ra_lock);
284 * Delay sock_put(sk) and kfree(ra) after one rcu grace
285 * period. This guarantee ip_call_ra_chain() dont need
286 * to mess with socket refcounts.
289 call_rcu(&ra->rcu, ip_ra_destroy_rcu);
293 if (new_ra == NULL) {
294 spin_unlock_bh(&ip_ra_lock);
298 new_ra->destructor = destructor;
301 rcu_assign_pointer(*rap, new_ra);
303 spin_unlock_bh(&ip_ra_lock);
308 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
309 __be16 port, u32 info, u8 *payload)
311 struct sock_exterr_skb *serr;
313 skb = skb_clone(skb, GFP_ATOMIC);
317 serr = SKB_EXT_ERR(skb);
318 serr->ee.ee_errno = err;
319 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
320 serr->ee.ee_type = icmp_hdr(skb)->type;
321 serr->ee.ee_code = icmp_hdr(skb)->code;
323 serr->ee.ee_info = info;
324 serr->ee.ee_data = 0;
325 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
326 skb_network_header(skb);
329 if (skb_pull(skb, payload - skb->data) != NULL) {
330 skb_reset_transport_header(skb);
331 if (sock_queue_err_skb(sk, skb) == 0)
337 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
339 struct inet_sock *inet = inet_sk(sk);
340 struct sock_exterr_skb *serr;
347 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
351 skb_put(skb, sizeof(struct iphdr));
352 skb_reset_network_header(skb);
356 serr = SKB_EXT_ERR(skb);
357 serr->ee.ee_errno = err;
358 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
359 serr->ee.ee_type = 0;
360 serr->ee.ee_code = 0;
362 serr->ee.ee_info = info;
363 serr->ee.ee_data = 0;
364 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
367 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
368 skb_reset_transport_header(skb);
370 if (sock_queue_err_skb(sk, skb))
375 * Handle MSG_ERRQUEUE
377 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
379 struct sock_exterr_skb *serr;
380 struct sk_buff *skb, *skb2;
381 struct sockaddr_in *sin;
383 struct sock_extended_err ee;
384 struct sockaddr_in offender;
390 skb = skb_dequeue(&sk->sk_error_queue);
396 msg->msg_flags |= MSG_TRUNC;
399 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
403 sock_recv_timestamp(msg, sk, skb);
405 serr = SKB_EXT_ERR(skb);
407 sin = (struct sockaddr_in *)msg->msg_name;
409 sin->sin_family = AF_INET;
410 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
412 sin->sin_port = serr->port;
413 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
416 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
417 sin = &errhdr.offender;
418 sin->sin_family = AF_UNSPEC;
419 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
420 struct inet_sock *inet = inet_sk(sk);
422 sin->sin_family = AF_INET;
423 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
425 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
426 if (inet->cmsg_flags)
427 ip_cmsg_recv(msg, skb);
430 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
432 /* Now we could try to dump offended packet options */
434 msg->msg_flags |= MSG_ERRQUEUE;
437 /* Reset and regenerate socket error */
438 spin_lock_bh(&sk->sk_error_queue.lock);
440 skb2 = skb_peek(&sk->sk_error_queue);
442 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
443 spin_unlock_bh(&sk->sk_error_queue.lock);
444 sk->sk_error_report(sk);
446 spin_unlock_bh(&sk->sk_error_queue.lock);
455 static void opt_kfree_rcu(struct rcu_head *head)
457 kfree(container_of(head, struct ip_options_rcu, rcu));
461 * Socket option code for IP. This is the end of the line after any
462 * TCP,UDP etc options on an IP socket.
465 static int do_ip_setsockopt(struct sock *sk, int level,
466 int optname, char __user *optval, unsigned int optlen)
468 struct inet_sock *inet = inet_sk(sk);
471 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
472 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
473 (1<<IP_RETOPTS) | (1<<IP_TOS) |
474 (1<<IP_TTL) | (1<<IP_HDRINCL) |
475 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
476 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
477 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) |
478 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) ||
479 optname == IP_MULTICAST_TTL ||
480 optname == IP_MULTICAST_ALL ||
481 optname == IP_MULTICAST_LOOP ||
482 optname == IP_RECVORIGDSTADDR) {
483 if (optlen >= sizeof(int)) {
484 if (get_user(val, (int __user *) optval))
486 } else if (optlen >= sizeof(char)) {
489 if (get_user(ucval, (unsigned char __user *) optval))
495 /* If optlen==0, it is equivalent to val == 0 */
497 if (ip_mroute_opt(optname))
498 return ip_mroute_setsockopt(sk, optname, optval, optlen);
506 struct ip_options_rcu *old, *opt = NULL;
510 err = ip_options_get_from_user(sock_net(sk), &opt,
514 old = rcu_dereference_protected(inet->inet_opt,
515 sock_owned_by_user(sk));
517 struct inet_connection_sock *icsk = inet_csk(sk);
518 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
519 if (sk->sk_family == PF_INET ||
520 (!((1 << sk->sk_state) &
521 (TCPF_LISTEN | TCPF_CLOSE)) &&
522 inet->inet_daddr != LOOPBACK4_IPV6)) {
525 icsk->icsk_ext_hdr_len -= old->opt.optlen;
527 icsk->icsk_ext_hdr_len += opt->opt.optlen;
528 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
529 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
533 rcu_assign_pointer(inet->inet_opt, opt);
535 call_rcu(&old->rcu, opt_kfree_rcu);
540 inet->cmsg_flags |= IP_CMSG_PKTINFO;
542 inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
546 inet->cmsg_flags |= IP_CMSG_TTL;
548 inet->cmsg_flags &= ~IP_CMSG_TTL;
552 inet->cmsg_flags |= IP_CMSG_TOS;
554 inet->cmsg_flags &= ~IP_CMSG_TOS;
558 inet->cmsg_flags |= IP_CMSG_RECVOPTS;
560 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
564 inet->cmsg_flags |= IP_CMSG_RETOPTS;
566 inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
570 inet->cmsg_flags |= IP_CMSG_PASSSEC;
572 inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
574 case IP_RECVORIGDSTADDR:
576 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
578 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
580 case IP_TOS: /* This sets both TOS and Precedence */
581 if (sk->sk_type == SOCK_STREAM) {
582 val &= ~INET_ECN_MASK;
583 val |= inet->tos & INET_ECN_MASK;
585 if (inet->tos != val) {
587 sk->sk_priority = rt_tos2priority(val);
594 if (val != -1 && (val < 0 || val > 255))
599 if (sk->sk_type != SOCK_RAW) {
603 inet->hdrincl = val ? 1 : 0;
606 if (sk->sk_type != SOCK_RAW) {
610 inet->nodefrag = val ? 1 : 0;
612 case IP_MTU_DISCOVER:
613 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
615 inet->pmtudisc = val;
618 inet->recverr = !!val;
620 skb_queue_purge(&sk->sk_error_queue);
622 case IP_MULTICAST_TTL:
623 if (sk->sk_type == SOCK_STREAM)
629 if (val < 0 || val > 255)
633 case IP_MULTICAST_LOOP:
636 inet->mc_loop = !!val;
638 case IP_MULTICAST_IF:
640 struct ip_mreqn mreq;
641 struct net_device *dev = NULL;
643 if (sk->sk_type == SOCK_STREAM)
646 * Check the arguments are allowable
649 if (optlen < sizeof(struct in_addr))
653 if (optlen >= sizeof(struct ip_mreqn)) {
654 if (copy_from_user(&mreq, optval, sizeof(mreq)))
657 memset(&mreq, 0, sizeof(mreq));
658 if (optlen >= sizeof(struct in_addr) &&
659 copy_from_user(&mreq.imr_address, optval,
660 sizeof(struct in_addr)))
664 if (!mreq.imr_ifindex) {
665 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
671 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
673 mreq.imr_ifindex = dev->ifindex;
675 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
678 err = -EADDRNOTAVAIL;
684 if (sk->sk_bound_dev_if &&
685 mreq.imr_ifindex != sk->sk_bound_dev_if)
688 inet->mc_index = mreq.imr_ifindex;
689 inet->mc_addr = mreq.imr_address.s_addr;
694 case IP_ADD_MEMBERSHIP:
695 case IP_DROP_MEMBERSHIP:
697 struct ip_mreqn mreq;
700 if (inet_sk(sk)->is_icsk)
703 if (optlen < sizeof(struct ip_mreq))
706 if (optlen >= sizeof(struct ip_mreqn)) {
707 if (copy_from_user(&mreq, optval, sizeof(mreq)))
710 memset(&mreq, 0, sizeof(mreq));
711 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
715 if (optname == IP_ADD_MEMBERSHIP)
716 err = ip_mc_join_group(sk, &mreq);
718 err = ip_mc_leave_group(sk, &mreq);
723 struct ip_msfilter *msf;
725 if (optlen < IP_MSFILTER_SIZE(0))
727 if (optlen > sysctl_optmem_max) {
731 msf = kmalloc(optlen, GFP_KERNEL);
737 if (copy_from_user(msf, optval, optlen)) {
741 /* numsrc >= (1G-4) overflow in 32 bits */
742 if (msf->imsf_numsrc >= 0x3ffffffcU ||
743 msf->imsf_numsrc > sysctl_igmp_max_msf) {
748 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
753 err = ip_mc_msfilter(sk, msf, 0);
757 case IP_BLOCK_SOURCE:
758 case IP_UNBLOCK_SOURCE:
759 case IP_ADD_SOURCE_MEMBERSHIP:
760 case IP_DROP_SOURCE_MEMBERSHIP:
762 struct ip_mreq_source mreqs;
765 if (optlen != sizeof(struct ip_mreq_source))
767 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
771 if (optname == IP_BLOCK_SOURCE) {
772 omode = MCAST_EXCLUDE;
774 } else if (optname == IP_UNBLOCK_SOURCE) {
775 omode = MCAST_EXCLUDE;
777 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
778 struct ip_mreqn mreq;
780 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
781 mreq.imr_address.s_addr = mreqs.imr_interface;
782 mreq.imr_ifindex = 0;
783 err = ip_mc_join_group(sk, &mreq);
784 if (err && err != -EADDRINUSE)
786 omode = MCAST_INCLUDE;
788 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
789 omode = MCAST_INCLUDE;
792 err = ip_mc_source(add, omode, sk, &mreqs, 0);
795 case MCAST_JOIN_GROUP:
796 case MCAST_LEAVE_GROUP:
798 struct group_req greq;
799 struct sockaddr_in *psin;
800 struct ip_mreqn mreq;
802 if (optlen < sizeof(struct group_req))
805 if (copy_from_user(&greq, optval, sizeof(greq)))
807 psin = (struct sockaddr_in *)&greq.gr_group;
808 if (psin->sin_family != AF_INET)
810 memset(&mreq, 0, sizeof(mreq));
811 mreq.imr_multiaddr = psin->sin_addr;
812 mreq.imr_ifindex = greq.gr_interface;
814 if (optname == MCAST_JOIN_GROUP)
815 err = ip_mc_join_group(sk, &mreq);
817 err = ip_mc_leave_group(sk, &mreq);
820 case MCAST_JOIN_SOURCE_GROUP:
821 case MCAST_LEAVE_SOURCE_GROUP:
822 case MCAST_BLOCK_SOURCE:
823 case MCAST_UNBLOCK_SOURCE:
825 struct group_source_req greqs;
826 struct ip_mreq_source mreqs;
827 struct sockaddr_in *psin;
830 if (optlen != sizeof(struct group_source_req))
832 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
836 if (greqs.gsr_group.ss_family != AF_INET ||
837 greqs.gsr_source.ss_family != AF_INET) {
838 err = -EADDRNOTAVAIL;
841 psin = (struct sockaddr_in *)&greqs.gsr_group;
842 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
843 psin = (struct sockaddr_in *)&greqs.gsr_source;
844 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
845 mreqs.imr_interface = 0; /* use index for mc_source */
847 if (optname == MCAST_BLOCK_SOURCE) {
848 omode = MCAST_EXCLUDE;
850 } else if (optname == MCAST_UNBLOCK_SOURCE) {
851 omode = MCAST_EXCLUDE;
853 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
854 struct ip_mreqn mreq;
856 psin = (struct sockaddr_in *)&greqs.gsr_group;
857 mreq.imr_multiaddr = psin->sin_addr;
858 mreq.imr_address.s_addr = 0;
859 mreq.imr_ifindex = greqs.gsr_interface;
860 err = ip_mc_join_group(sk, &mreq);
861 if (err && err != -EADDRINUSE)
863 greqs.gsr_interface = mreq.imr_ifindex;
864 omode = MCAST_INCLUDE;
866 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
867 omode = MCAST_INCLUDE;
870 err = ip_mc_source(add, omode, sk, &mreqs,
871 greqs.gsr_interface);
876 struct sockaddr_in *psin;
877 struct ip_msfilter *msf = NULL;
878 struct group_filter *gsf = NULL;
879 int msize, i, ifindex;
881 if (optlen < GROUP_FILTER_SIZE(0))
883 if (optlen > sysctl_optmem_max) {
887 gsf = kmalloc(optlen, GFP_KERNEL);
893 if (copy_from_user(gsf, optval, optlen))
896 /* numsrc >= (4G-140)/128 overflow in 32 bits */
897 if (gsf->gf_numsrc >= 0x1ffffff ||
898 gsf->gf_numsrc > sysctl_igmp_max_msf) {
902 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
906 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
907 msf = kmalloc(msize, GFP_KERNEL);
912 ifindex = gsf->gf_interface;
913 psin = (struct sockaddr_in *)&gsf->gf_group;
914 if (psin->sin_family != AF_INET) {
915 err = -EADDRNOTAVAIL;
918 msf->imsf_multiaddr = psin->sin_addr.s_addr;
919 msf->imsf_interface = 0;
920 msf->imsf_fmode = gsf->gf_fmode;
921 msf->imsf_numsrc = gsf->gf_numsrc;
922 err = -EADDRNOTAVAIL;
923 for (i = 0; i < gsf->gf_numsrc; ++i) {
924 psin = (struct sockaddr_in *)&gsf->gf_slist[i];
926 if (psin->sin_family != AF_INET)
928 msf->imsf_slist[i] = psin->sin_addr.s_addr;
933 err = ip_mc_msfilter(sk, msf, ifindex);
939 case IP_MULTICAST_ALL:
942 if (val != 0 && val != 1)
946 case IP_ROUTER_ALERT:
947 err = ip_ra_control(sk, val ? 1 : 0, NULL);
953 inet->freebind = !!val;
956 case IP_IPSEC_POLICY:
959 if (!capable(CAP_NET_ADMIN))
961 err = xfrm_user_policy(sk, optname, optval, optlen);
965 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) {
971 inet->transparent = !!val;
977 if (val < 0 || val > 255)
995 * ip_queue_rcv_skb - Queue an skb into sock receive queue
999 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option
1000 * is not set, we drop skb dst entry now, while dst cache line is hot.
1002 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1004 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO))
1006 return sock_queue_rcv_skb(sk, skb);
1008 EXPORT_SYMBOL(ip_queue_rcv_skb);
1010 int ip_setsockopt(struct sock *sk, int level,
1011 int optname, char __user *optval, unsigned int optlen)
1015 if (level != SOL_IP)
1016 return -ENOPROTOOPT;
1018 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1019 #ifdef CONFIG_NETFILTER
1020 /* we need to exclude all possible ENOPROTOOPTs except default case */
1021 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1022 optname != IP_IPSEC_POLICY &&
1023 optname != IP_XFRM_POLICY &&
1024 !ip_mroute_opt(optname)) {
1026 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1032 EXPORT_SYMBOL(ip_setsockopt);
1034 #ifdef CONFIG_COMPAT
1035 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1036 char __user *optval, unsigned int optlen)
1040 if (level != SOL_IP)
1041 return -ENOPROTOOPT;
1043 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1044 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1047 err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1048 #ifdef CONFIG_NETFILTER
1049 /* we need to exclude all possible ENOPROTOOPTs except default case */
1050 if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1051 optname != IP_IPSEC_POLICY &&
1052 optname != IP_XFRM_POLICY &&
1053 !ip_mroute_opt(optname)) {
1055 err = compat_nf_setsockopt(sk, PF_INET, optname,
1062 EXPORT_SYMBOL(compat_ip_setsockopt);
1066 * Get the options. Note for future reference. The GET of IP options gets
1067 * the _received_ ones. The set sets the _sent_ ones.
1070 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1071 char __user *optval, int __user *optlen, unsigned flags)
1073 struct inet_sock *inet = inet_sk(sk);
1077 if (level != SOL_IP)
1080 if (ip_mroute_opt(optname))
1081 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1083 if (get_user(len, optlen))
1093 unsigned char optbuf[sizeof(struct ip_options)+40];
1094 struct ip_options *opt = (struct ip_options *)optbuf;
1095 struct ip_options_rcu *inet_opt;
1097 inet_opt = rcu_dereference_protected(inet->inet_opt,
1098 sock_owned_by_user(sk));
1101 memcpy(optbuf, &inet_opt->opt,
1102 sizeof(struct ip_options) +
1103 inet_opt->opt.optlen);
1106 if (opt->optlen == 0)
1107 return put_user(0, optlen);
1109 ip_options_undo(opt);
1111 len = min_t(unsigned int, len, opt->optlen);
1112 if (put_user(len, optlen))
1114 if (copy_to_user(optval, opt->__data, len))
1119 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1122 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1125 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1128 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1131 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1134 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1136 case IP_RECVORIGDSTADDR:
1137 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1143 val = (inet->uc_ttl == -1 ?
1144 sysctl_ip_default_ttl :
1148 val = inet->hdrincl;
1151 val = inet->nodefrag;
1153 case IP_MTU_DISCOVER:
1154 val = inet->pmtudisc;
1158 struct dst_entry *dst;
1160 dst = sk_dst_get(sk);
1172 val = inet->recverr;
1174 case IP_MULTICAST_TTL:
1177 case IP_MULTICAST_LOOP:
1178 val = inet->mc_loop;
1180 case IP_MULTICAST_IF:
1182 struct in_addr addr;
1183 len = min_t(unsigned int, len, sizeof(struct in_addr));
1184 addr.s_addr = inet->mc_addr;
1187 if (put_user(len, optlen))
1189 if (copy_to_user(optval, &addr, len))
1195 struct ip_msfilter msf;
1198 if (len < IP_MSFILTER_SIZE(0)) {
1202 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1206 err = ip_mc_msfget(sk, &msf,
1207 (struct ip_msfilter __user *)optval, optlen);
1211 case MCAST_MSFILTER:
1213 struct group_filter gsf;
1216 if (len < GROUP_FILTER_SIZE(0)) {
1220 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1224 err = ip_mc_gsfget(sk, &gsf,
1225 (struct group_filter __user *)optval,
1230 case IP_MULTICAST_ALL:
1239 if (sk->sk_type != SOCK_STREAM)
1240 return -ENOPROTOOPT;
1242 msg.msg_control = optval;
1243 msg.msg_controllen = len;
1244 msg.msg_flags = flags;
1246 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1247 struct in_pktinfo info;
1249 info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1250 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1251 info.ipi_ifindex = inet->mc_index;
1252 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1254 if (inet->cmsg_flags & IP_CMSG_TTL) {
1255 int hlim = inet->mc_ttl;
1256 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1258 len -= msg.msg_controllen;
1259 return put_user(len, optlen);
1262 val = inet->freebind;
1264 case IP_TRANSPARENT:
1265 val = inet->transparent;
1268 val = inet->min_ttl;
1272 return -ENOPROTOOPT;
1276 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1277 unsigned char ucval = (unsigned char)val;
1279 if (put_user(len, optlen))
1281 if (copy_to_user(optval, &ucval, 1))
1284 len = min_t(unsigned int, sizeof(int), len);
1285 if (put_user(len, optlen))
1287 if (copy_to_user(optval, &val, len))
1293 int ip_getsockopt(struct sock *sk, int level,
1294 int optname, char __user *optval, int __user *optlen)
1298 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1299 #ifdef CONFIG_NETFILTER
1300 /* we need to exclude all possible ENOPROTOOPTs except default case */
1301 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1302 !ip_mroute_opt(optname)) {
1305 if (get_user(len, optlen))
1309 err = nf_getsockopt(sk, PF_INET, optname, optval,
1313 err = put_user(len, optlen);
1319 EXPORT_SYMBOL(ip_getsockopt);
1321 #ifdef CONFIG_COMPAT
1322 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1323 char __user *optval, int __user *optlen)
1327 if (optname == MCAST_MSFILTER)
1328 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1331 err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1334 #ifdef CONFIG_NETFILTER
1335 /* we need to exclude all possible ENOPROTOOPTs except default case */
1336 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1337 !ip_mroute_opt(optname)) {
1340 if (get_user(len, optlen))
1344 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1347 err = put_user(len, optlen);
1353 EXPORT_SYMBOL(compat_ip_getsockopt);