2 * PF_INET6 socket protocol family
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/af_inet.c
10 * $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
13 * piggy, Karl Knutson : Socket protocol table
14 * Hideaki YOSHIFUJI : sin6_scope_id support
15 * Arnaldo Melo : check proc_net_create return, cleanups
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
24 #include <linux/module.h>
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
30 #include <linux/kernel.h>
31 #include <linux/timer.h>
32 #include <linux/string.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/fcntl.h>
37 #include <linux/interrupt.h>
38 #include <linux/proc_fs.h>
39 #include <linux/stat.h>
40 #include <linux/init.h>
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/smp_lock.h>
46 #include <linux/netfilter_ipv6.h>
51 #include <net/udplite.h>
54 #include <net/protocol.h>
55 #include <net/inet_common.h>
56 #include <net/transp_v6.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #ifdef CONFIG_IPV6_TUNNEL
60 #include <net/ip6_tunnel.h>
62 #ifdef CONFIG_IPV6_MIP6
66 #include <asm/uaccess.h>
67 #include <asm/system.h>
69 MODULE_AUTHOR("Cast of dozens");
70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
71 MODULE_LICENSE("GPL");
73 int sysctl_ipv6_bindv6only __read_mostly;
75 /* The inetsw table contains everything that inet_create needs to
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
81 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
83 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
85 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
88 static int inet6_create(struct socket *sock, int protocol)
90 struct inet_sock *inet;
91 struct ipv6_pinfo *np;
94 struct inet_protosw *answer;
95 struct proto *answer_prot;
96 unsigned char answer_flags;
98 int try_loading_module = 0;
101 /* Look for the requested type/protocol pair. */
104 err = -ESOCKTNOSUPPORT;
106 list_for_each_rcu(p, &inetsw6[sock->type]) {
107 answer = list_entry(p, struct inet_protosw, list);
109 /* Check the non-wild match. */
110 if (protocol == answer->protocol) {
111 if (protocol != IPPROTO_IP)
114 /* Check for the two wild cases. */
115 if (IPPROTO_IP == protocol) {
116 protocol = answer->protocol;
119 if (IPPROTO_IP == answer->protocol)
122 err = -EPROTONOSUPPORT;
127 if (try_loading_module < 2) {
130 * Be more specific, e.g. net-pf-10-proto-132-type-1
131 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
133 if (++try_loading_module == 1)
134 request_module("net-pf-%d-proto-%d-type-%d",
135 PF_INET6, protocol, sock->type);
137 * Fall back to generic, e.g. net-pf-10-proto-132
138 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
141 request_module("net-pf-%d-proto-%d",
143 goto lookup_protocol;
149 if (answer->capability > 0 && !capable(answer->capability))
152 sock->ops = answer->ops;
153 answer_prot = answer->prot;
154 answer_no_check = answer->no_check;
155 answer_flags = answer->flags;
158 BUG_TRAP(answer_prot->slab != NULL);
161 sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
165 sock_init_data(sock, sk);
168 sk->sk_no_check = answer_no_check;
169 if (INET_PROTOSW_REUSE & answer_flags)
173 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
175 if (SOCK_RAW == sock->type) {
176 inet->num = protocol;
177 if (IPPROTO_RAW == protocol)
181 sk->sk_destruct = inet_sock_destruct;
182 sk->sk_family = PF_INET6;
183 sk->sk_protocol = protocol;
185 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
187 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
191 np->pmtudisc = IPV6_PMTUDISC_WANT;
192 np->ipv6only = sysctl_ipv6_bindv6only;
194 /* Init the ipv4 part of the socket since we can have sockets
195 * using v6 API for ipv4.
202 inet->mc_list = NULL;
204 if (ipv4_config.no_pmtu_disc)
205 inet->pmtudisc = IP_PMTUDISC_DONT;
207 inet->pmtudisc = IP_PMTUDISC_WANT;
209 * Increment only the relevant sk_prot->socks debug field, this changes
210 * the previous behaviour of incrementing both the equivalent to
211 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
213 * This allows better debug granularity as we'll know exactly how many
214 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
215 * transport protocol socks. -acme
217 sk_refcnt_debug_inc(sk);
220 /* It assumes that any protocol which allows
221 * the user to assign a number at socket
222 * creation time automatically shares.
224 inet->sport = htons(inet->num);
225 sk->sk_prot->hash(sk);
227 if (sk->sk_prot->init) {
228 err = sk->sk_prot->init(sk);
230 sk_common_release(sk);
242 /* bind for INET6 API */
243 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
245 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
246 struct sock *sk = sock->sk;
247 struct inet_sock *inet = inet_sk(sk);
248 struct ipv6_pinfo *np = inet6_sk(sk);
254 /* If the socket has its own bind function then use it. */
255 if (sk->sk_prot->bind)
256 return sk->sk_prot->bind(sk, uaddr, addr_len);
258 if (addr_len < SIN6_LEN_RFC2133)
260 addr_type = ipv6_addr_type(&addr->sin6_addr);
261 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
264 snum = ntohs(addr->sin6_port);
265 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
270 /* Check these errors (active socket, double bind). */
271 if (sk->sk_state != TCP_CLOSE || inet->num) {
276 /* Check if the address belongs to the host. */
277 if (addr_type == IPV6_ADDR_MAPPED) {
278 v4addr = addr->sin6_addr.s6_addr32[3];
279 if (inet_addr_type(v4addr) != RTN_LOCAL) {
280 err = -EADDRNOTAVAIL;
284 if (addr_type != IPV6_ADDR_ANY) {
285 struct net_device *dev = NULL;
287 if (addr_type & IPV6_ADDR_LINKLOCAL) {
288 if (addr_len >= sizeof(struct sockaddr_in6) &&
289 addr->sin6_scope_id) {
290 /* Override any existing binding, if another one
291 * is supplied by user.
293 sk->sk_bound_dev_if = addr->sin6_scope_id;
296 /* Binding to link-local address requires an interface */
297 if (!sk->sk_bound_dev_if) {
301 dev = dev_get_by_index(sk->sk_bound_dev_if);
308 /* ipv4 addr of the socket is invalid. Only the
309 * unspecified and mapped address have a v4 equivalent.
311 v4addr = LOOPBACK4_IPV6;
312 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
313 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
316 err = -EADDRNOTAVAIL;
325 inet->rcv_saddr = v4addr;
326 inet->saddr = v4addr;
328 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
330 if (!(addr_type & IPV6_ADDR_MULTICAST))
331 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
333 /* Make sure we are allowed to bind here. */
334 if (sk->sk_prot->get_port(sk, snum)) {
335 inet_reset_saddr(sk);
340 if (addr_type != IPV6_ADDR_ANY)
341 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
343 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
344 inet->sport = htons(inet->num);
352 int inet6_release(struct socket *sock)
354 struct sock *sk = sock->sk;
360 ipv6_sock_mc_close(sk);
363 ipv6_sock_ac_close(sk);
365 return inet_release(sock);
368 int inet6_destroy_sock(struct sock *sk)
370 struct ipv6_pinfo *np = inet6_sk(sk);
372 struct ipv6_txoptions *opt;
374 /* Release rx options */
376 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
379 /* Free flowlabels */
380 fl6_free_socklist(sk);
382 /* Free tx options */
384 if ((opt = xchg(&np->opt, NULL)) != NULL)
385 sock_kfree_s(sk, opt, opt->tot_len);
390 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
393 * This does both peername and sockname.
396 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
397 int *uaddr_len, int peer)
399 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
400 struct sock *sk = sock->sk;
401 struct inet_sock *inet = inet_sk(sk);
402 struct ipv6_pinfo *np = inet6_sk(sk);
404 sin->sin6_family = AF_INET6;
405 sin->sin6_flowinfo = 0;
406 sin->sin6_scope_id = 0;
410 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
413 sin->sin6_port = inet->dport;
414 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
416 sin->sin6_flowinfo = np->flow_label;
418 if (ipv6_addr_any(&np->rcv_saddr))
419 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
421 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
423 sin->sin6_port = inet->sport;
425 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
426 sin->sin6_scope_id = sk->sk_bound_dev_if;
427 *uaddr_len = sizeof(*sin);
431 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
433 struct sock *sk = sock->sk;
438 return sock_get_timestamp(sk, (struct timeval __user *)arg);
443 return(ipv6_route_ioctl(cmd,(void __user *)arg));
446 return addrconf_add_ifaddr((void __user *) arg);
448 return addrconf_del_ifaddr((void __user *) arg);
450 return addrconf_set_dstaddr((void __user *) arg);
452 if (!sk->sk_prot->ioctl)
454 return sk->sk_prot->ioctl(sk, cmd, arg);
460 const struct proto_ops inet6_stream_ops = {
462 .owner = THIS_MODULE,
463 .release = inet6_release,
465 .connect = inet_stream_connect, /* ok */
466 .socketpair = sock_no_socketpair, /* a do nothing */
467 .accept = inet_accept, /* ok */
468 .getname = inet6_getname,
469 .poll = tcp_poll, /* ok */
470 .ioctl = inet6_ioctl, /* must change */
471 .listen = inet_listen, /* ok */
472 .shutdown = inet_shutdown, /* ok */
473 .setsockopt = sock_common_setsockopt, /* ok */
474 .getsockopt = sock_common_getsockopt, /* ok */
475 .sendmsg = inet_sendmsg, /* ok */
476 .recvmsg = sock_common_recvmsg, /* ok */
477 .mmap = sock_no_mmap,
478 .sendpage = tcp_sendpage,
480 .compat_setsockopt = compat_sock_common_setsockopt,
481 .compat_getsockopt = compat_sock_common_getsockopt,
485 const struct proto_ops inet6_dgram_ops = {
487 .owner = THIS_MODULE,
488 .release = inet6_release,
490 .connect = inet_dgram_connect, /* ok */
491 .socketpair = sock_no_socketpair, /* a do nothing */
492 .accept = sock_no_accept, /* a do nothing */
493 .getname = inet6_getname,
494 .poll = udp_poll, /* ok */
495 .ioctl = inet6_ioctl, /* must change */
496 .listen = sock_no_listen, /* ok */
497 .shutdown = inet_shutdown, /* ok */
498 .setsockopt = sock_common_setsockopt, /* ok */
499 .getsockopt = sock_common_getsockopt, /* ok */
500 .sendmsg = inet_sendmsg, /* ok */
501 .recvmsg = sock_common_recvmsg, /* ok */
502 .mmap = sock_no_mmap,
503 .sendpage = sock_no_sendpage,
505 .compat_setsockopt = compat_sock_common_setsockopt,
506 .compat_getsockopt = compat_sock_common_getsockopt,
510 static struct net_proto_family inet6_family_ops = {
512 .create = inet6_create,
513 .owner = THIS_MODULE,
516 /* Same as inet6_dgram_ops, sans udp_poll. */
517 static const struct proto_ops inet6_sockraw_ops = {
519 .owner = THIS_MODULE,
520 .release = inet6_release,
522 .connect = inet_dgram_connect, /* ok */
523 .socketpair = sock_no_socketpair, /* a do nothing */
524 .accept = sock_no_accept, /* a do nothing */
525 .getname = inet6_getname,
526 .poll = datagram_poll, /* ok */
527 .ioctl = inet6_ioctl, /* must change */
528 .listen = sock_no_listen, /* ok */
529 .shutdown = inet_shutdown, /* ok */
530 .setsockopt = sock_common_setsockopt, /* ok */
531 .getsockopt = sock_common_getsockopt, /* ok */
532 .sendmsg = inet_sendmsg, /* ok */
533 .recvmsg = sock_common_recvmsg, /* ok */
534 .mmap = sock_no_mmap,
535 .sendpage = sock_no_sendpage,
537 .compat_setsockopt = compat_sock_common_setsockopt,
538 .compat_getsockopt = compat_sock_common_getsockopt,
542 static struct inet_protosw rawv6_protosw = {
544 .protocol = IPPROTO_IP, /* wild card */
546 .ops = &inet6_sockraw_ops,
547 .capability = CAP_NET_RAW,
548 .no_check = UDP_CSUM_DEFAULT,
549 .flags = INET_PROTOSW_REUSE,
553 inet6_register_protosw(struct inet_protosw *p)
555 struct list_head *lh;
556 struct inet_protosw *answer;
557 int protocol = p->protocol;
558 struct list_head *last_perm;
560 spin_lock_bh(&inetsw6_lock);
562 if (p->type >= SOCK_MAX)
565 /* If we are trying to override a permanent protocol, bail. */
567 last_perm = &inetsw6[p->type];
568 list_for_each(lh, &inetsw6[p->type]) {
569 answer = list_entry(lh, struct inet_protosw, list);
571 /* Check only the non-wild match. */
572 if (INET_PROTOSW_PERMANENT & answer->flags) {
573 if (protocol == answer->protocol)
583 /* Add the new entry after the last permanent entry if any, so that
584 * the new entry does not override a permanent entry when matched with
585 * a wild-card protocol. But it is allowed to override any existing
586 * non-permanent entry. This means that when we remove this entry, the
587 * system automatically returns to the old behavior.
589 list_add_rcu(&p->list, last_perm);
591 spin_unlock_bh(&inetsw6_lock);
595 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
601 "Ignoring attempt to register invalid socket type %d.\n",
607 inet6_unregister_protosw(struct inet_protosw *p)
609 if (INET_PROTOSW_PERMANENT & p->flags) {
611 "Attempt to unregister permanent protocol %d.\n",
614 spin_lock_bh(&inetsw6_lock);
615 list_del_rcu(&p->list);
616 spin_unlock_bh(&inetsw6_lock);
622 int inet6_sk_rebuild_header(struct sock *sk)
625 struct dst_entry *dst;
626 struct ipv6_pinfo *np = inet6_sk(sk);
628 dst = __sk_dst_check(sk, np->dst_cookie);
631 struct inet_sock *inet = inet_sk(sk);
632 struct in6_addr *final_p = NULL, final;
635 memset(&fl, 0, sizeof(fl));
636 fl.proto = sk->sk_protocol;
637 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
638 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
639 fl.fl6_flowlabel = np->flow_label;
640 fl.oif = sk->sk_bound_dev_if;
641 fl.fl_ip_dport = inet->dport;
642 fl.fl_ip_sport = inet->sport;
643 security_sk_classify_flow(sk, &fl);
645 if (np->opt && np->opt->srcrt) {
646 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
647 ipv6_addr_copy(&final, &fl.fl6_dst);
648 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
652 err = ip6_dst_lookup(sk, &dst, &fl);
654 sk->sk_route_caps = 0;
658 ipv6_addr_copy(&fl.fl6_dst, final_p);
660 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
661 sk->sk_err_soft = -err;
665 __ip6_dst_store(sk, dst, NULL, NULL);
671 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
673 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
675 struct ipv6_pinfo *np = inet6_sk(sk);
676 struct inet6_skb_parm *opt = IP6CB(skb);
679 if ((opt->hop && (np->rxopt.bits.hopopts ||
680 np->rxopt.bits.ohopopts)) ||
681 ((IPV6_FLOWINFO_MASK & *(__be32*)skb->nh.raw) &&
682 np->rxopt.bits.rxflow) ||
683 (opt->srcrt && (np->rxopt.bits.srcrt ||
684 np->rxopt.bits.osrcrt)) ||
685 ((opt->dst1 || opt->dst0) &&
686 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
692 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
695 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
700 ptr[0] = __alloc_percpu(mibsize);
704 ptr[1] = __alloc_percpu(mibsize);
718 snmp6_mib_free(void *ptr[2])
724 ptr[0] = ptr[1] = NULL;
727 static int __init init_ipv6_mibs(void)
729 if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
730 __alignof__(struct ipstats_mib)) < 0)
732 if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
733 __alignof__(struct icmpv6_mib)) < 0)
735 if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
736 __alignof__(struct udp_mib)) < 0)
738 if (snmp6_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
739 __alignof__(struct udp_mib)) < 0)
740 goto err_udplite_mib;
744 snmp6_mib_free((void **)udp_stats_in6);
746 snmp6_mib_free((void **)icmpv6_statistics);
748 snmp6_mib_free((void **)ipv6_statistics);
754 static void cleanup_ipv6_mibs(void)
756 snmp6_mib_free((void **)ipv6_statistics);
757 snmp6_mib_free((void **)icmpv6_statistics);
758 snmp6_mib_free((void **)udp_stats_in6);
759 snmp6_mib_free((void **)udplite_stats_in6);
762 static int __init inet6_init(void)
764 struct sk_buff *dummy_skb;
768 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
771 #if 0 /* FIXME --RR */
772 if (!mod_member_present(&__this_module, can_unload))
775 __this_module.can_unload = &ipv6_unload;
779 err = proto_register(&tcpv6_prot, 1);
783 err = proto_register(&udpv6_prot, 1);
785 goto out_unregister_tcp_proto;
787 err = proto_register(&udplitev6_prot, 1);
789 goto out_unregister_udp_proto;
791 err = proto_register(&rawv6_prot, 1);
793 goto out_unregister_udplite_proto;
796 /* Register the socket-side information for inet6_create. */
797 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
800 /* We MUST register RAW sockets before we create the ICMP6,
801 * IGMP6, or NDISC control sockets.
803 inet6_register_protosw(&rawv6_protosw);
805 /* Register the family here so that the init calls below will
806 * be able to create sockets. (?? is this dangerous ??)
808 err = sock_register(&inet6_family_ops);
810 goto out_unregister_raw_proto;
812 /* Initialise ipv6 mibs */
813 err = init_ipv6_mibs();
815 goto out_unregister_sock;
818 * ipngwg API draft makes clear that the correct semantics
819 * for TCP and UDP is to consider one TCP and UDP instance
820 * in a host availiable by both INET and INET6 APIs and
821 * able to communicate via both network protocols.
825 ipv6_sysctl_register();
827 err = icmpv6_init(&inet6_family_ops);
830 err = ndisc_init(&inet6_family_ops);
833 err = igmp6_init(&inet6_family_ops);
836 err = ipv6_netfilter_init();
839 /* Create /proc/foo6 entries. */
840 #ifdef CONFIG_PROC_FS
842 if (raw6_proc_init())
844 if (tcp6_proc_init())
846 if (udp6_proc_init())
848 if (udplite6_proc_init())
849 goto proc_udplite6_fail;
850 if (ipv6_misc_proc_init())
851 goto proc_misc6_fail;
854 goto proc_anycast6_fail;
859 ip6_flowlabel_init();
860 err = addrconf_init();
864 /* Init v6 extension headers. */
869 #ifdef CONFIG_IPV6_MIP6
873 /* Init v6 transport protocols. */
884 ip6_flowlabel_cleanup();
886 #ifdef CONFIG_PROC_FS
891 ipv6_misc_proc_exit();
893 udplite6_proc_exit();
902 ipv6_netfilter_fini();
911 ipv6_sysctl_unregister();
915 sock_unregister(PF_INET6);
916 out_unregister_raw_proto:
917 proto_unregister(&rawv6_prot);
918 out_unregister_udplite_proto:
919 proto_unregister(&udplitev6_prot);
920 out_unregister_udp_proto:
921 proto_unregister(&udpv6_prot);
922 out_unregister_tcp_proto:
923 proto_unregister(&tcpv6_prot);
926 module_init(inet6_init);
928 static void __exit inet6_exit(void)
930 /* First of all disallow new sockets creation. */
931 sock_unregister(PF_INET6);
933 /* Cleanup code parts. */
934 ipv6_packet_cleanup();
935 #ifdef CONFIG_IPV6_MIP6
939 ip6_flowlabel_cleanup();
941 #ifdef CONFIG_PROC_FS
943 /* Cleanup code parts. */
946 ipv6_misc_proc_exit();
947 udplite6_proc_exit();
952 ipv6_netfilter_fini();
957 ipv6_sysctl_unregister();
960 proto_unregister(&rawv6_prot);
961 proto_unregister(&udplitev6_prot);
962 proto_unregister(&udpv6_prot);
963 proto_unregister(&tcpv6_prot);
965 module_exit(inet6_exit);
967 MODULE_ALIAS_NETPROTO(PF_INET6);