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/netfilter_ipv6.h>
50 #include <net/udplite.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/transp_v6.h>
56 #include <net/ip6_route.h>
57 #include <net/addrconf.h>
58 #ifdef CONFIG_IPV6_TUNNEL
59 #include <net/ip6_tunnel.h>
62 #include <asm/uaccess.h>
63 #include <asm/system.h>
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
69 int sysctl_ipv6_bindv6only __read_mostly;
71 /* The inetsw table contains everything that inet_create needs to
74 static struct list_head inetsw6[SOCK_MAX];
75 static DEFINE_SPINLOCK(inetsw6_lock);
77 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
79 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
81 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
84 static int inet6_create(struct socket *sock, int protocol)
86 struct inet_sock *inet;
87 struct ipv6_pinfo *np;
90 struct inet_protosw *answer;
91 struct proto *answer_prot;
92 unsigned char answer_flags;
94 int try_loading_module = 0;
97 if (sock->type != SOCK_RAW &&
98 sock->type != SOCK_DGRAM &&
100 build_ehash_secret();
102 /* Look for the requested type/protocol pair. */
105 err = -ESOCKTNOSUPPORT;
107 list_for_each_rcu(p, &inetsw6[sock->type]) {
108 answer = list_entry(p, struct inet_protosw, list);
110 /* Check the non-wild match. */
111 if (protocol == answer->protocol) {
112 if (protocol != IPPROTO_IP)
115 /* Check for the two wild cases. */
116 if (IPPROTO_IP == protocol) {
117 protocol = answer->protocol;
120 if (IPPROTO_IP == answer->protocol)
123 err = -EPROTONOSUPPORT;
128 if (try_loading_module < 2) {
131 * Be more specific, e.g. net-pf-10-proto-132-type-1
132 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
134 if (++try_loading_module == 1)
135 request_module("net-pf-%d-proto-%d-type-%d",
136 PF_INET6, protocol, sock->type);
138 * Fall back to generic, e.g. net-pf-10-proto-132
139 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
142 request_module("net-pf-%d-proto-%d",
144 goto lookup_protocol;
150 if (answer->capability > 0 && !capable(answer->capability))
153 sock->ops = answer->ops;
154 answer_prot = answer->prot;
155 answer_no_check = answer->no_check;
156 answer_flags = answer->flags;
159 BUG_TRAP(answer_prot->slab != NULL);
162 sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
166 sock_init_data(sock, sk);
169 sk->sk_no_check = answer_no_check;
170 if (INET_PROTOSW_REUSE & answer_flags)
174 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
176 if (SOCK_RAW == sock->type) {
177 inet->num = protocol;
178 if (IPPROTO_RAW == protocol)
182 sk->sk_destruct = inet_sock_destruct;
183 sk->sk_family = PF_INET6;
184 sk->sk_protocol = protocol;
186 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
188 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
192 np->pmtudisc = IPV6_PMTUDISC_WANT;
193 np->ipv6only = sysctl_ipv6_bindv6only;
195 /* Init the ipv4 part of the socket since we can have sockets
196 * using v6 API for ipv4.
203 inet->mc_list = NULL;
205 if (ipv4_config.no_pmtu_disc)
206 inet->pmtudisc = IP_PMTUDISC_DONT;
208 inet->pmtudisc = IP_PMTUDISC_WANT;
210 * Increment only the relevant sk_prot->socks debug field, this changes
211 * the previous behaviour of incrementing both the equivalent to
212 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
214 * This allows better debug granularity as we'll know exactly how many
215 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
216 * transport protocol socks. -acme
218 sk_refcnt_debug_inc(sk);
221 /* It assumes that any protocol which allows
222 * the user to assign a number at socket
223 * creation time automatically shares.
225 inet->sport = htons(inet->num);
226 sk->sk_prot->hash(sk);
228 if (sk->sk_prot->init) {
229 err = sk->sk_prot->init(sk);
231 sk_common_release(sk);
243 /* bind for INET6 API */
244 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
246 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
247 struct sock *sk = sock->sk;
248 struct inet_sock *inet = inet_sk(sk);
249 struct ipv6_pinfo *np = inet6_sk(sk);
255 /* If the socket has its own bind function then use it. */
256 if (sk->sk_prot->bind)
257 return sk->sk_prot->bind(sk, uaddr, addr_len);
259 if (addr_len < SIN6_LEN_RFC2133)
261 addr_type = ipv6_addr_type(&addr->sin6_addr);
262 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
265 snum = ntohs(addr->sin6_port);
266 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
271 /* Check these errors (active socket, double bind). */
272 if (sk->sk_state != TCP_CLOSE || inet->num) {
277 /* Check if the address belongs to the host. */
278 if (addr_type == IPV6_ADDR_MAPPED) {
279 v4addr = addr->sin6_addr.s6_addr32[3];
280 if (inet_addr_type(v4addr) != RTN_LOCAL) {
281 err = -EADDRNOTAVAIL;
285 if (addr_type != IPV6_ADDR_ANY) {
286 struct net_device *dev = NULL;
288 if (addr_type & IPV6_ADDR_LINKLOCAL) {
289 if (addr_len >= sizeof(struct sockaddr_in6) &&
290 addr->sin6_scope_id) {
291 /* Override any existing binding, if another one
292 * is supplied by user.
294 sk->sk_bound_dev_if = addr->sin6_scope_id;
297 /* Binding to link-local address requires an interface */
298 if (!sk->sk_bound_dev_if) {
302 dev = dev_get_by_index(sk->sk_bound_dev_if);
309 /* ipv4 addr of the socket is invalid. Only the
310 * unspecified and mapped address have a v4 equivalent.
312 v4addr = LOOPBACK4_IPV6;
313 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
314 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
317 err = -EADDRNOTAVAIL;
326 inet->rcv_saddr = v4addr;
327 inet->saddr = v4addr;
329 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
331 if (!(addr_type & IPV6_ADDR_MULTICAST))
332 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
334 /* Make sure we are allowed to bind here. */
335 if (sk->sk_prot->get_port(sk, snum)) {
336 inet_reset_saddr(sk);
341 if (addr_type != IPV6_ADDR_ANY)
342 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
344 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
345 inet->sport = htons(inet->num);
353 EXPORT_SYMBOL(inet6_bind);
355 int inet6_release(struct socket *sock)
357 struct sock *sk = sock->sk;
363 ipv6_sock_mc_close(sk);
366 ipv6_sock_ac_close(sk);
368 return inet_release(sock);
371 EXPORT_SYMBOL(inet6_release);
373 int inet6_destroy_sock(struct sock *sk)
375 struct ipv6_pinfo *np = inet6_sk(sk);
377 struct ipv6_txoptions *opt;
379 /* Release rx options */
381 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
384 /* Free flowlabels */
385 fl6_free_socklist(sk);
387 /* Free tx options */
389 if ((opt = xchg(&np->opt, NULL)) != NULL)
390 sock_kfree_s(sk, opt, opt->tot_len);
395 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
398 * This does both peername and sockname.
401 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
402 int *uaddr_len, int peer)
404 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
405 struct sock *sk = sock->sk;
406 struct inet_sock *inet = inet_sk(sk);
407 struct ipv6_pinfo *np = inet6_sk(sk);
409 sin->sin6_family = AF_INET6;
410 sin->sin6_flowinfo = 0;
411 sin->sin6_scope_id = 0;
415 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
418 sin->sin6_port = inet->dport;
419 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
421 sin->sin6_flowinfo = np->flow_label;
423 if (ipv6_addr_any(&np->rcv_saddr))
424 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
426 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
428 sin->sin6_port = inet->sport;
430 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
431 sin->sin6_scope_id = sk->sk_bound_dev_if;
432 *uaddr_len = sizeof(*sin);
436 EXPORT_SYMBOL(inet6_getname);
438 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
440 struct sock *sk = sock->sk;
445 return sock_get_timestamp(sk, (struct timeval __user *)arg);
448 return sock_get_timestampns(sk, (struct timespec __user *)arg);
453 return(ipv6_route_ioctl(cmd,(void __user *)arg));
456 return addrconf_add_ifaddr((void __user *) arg);
458 return addrconf_del_ifaddr((void __user *) arg);
460 return addrconf_set_dstaddr((void __user *) arg);
462 if (!sk->sk_prot->ioctl)
464 return sk->sk_prot->ioctl(sk, cmd, arg);
470 EXPORT_SYMBOL(inet6_ioctl);
472 const struct proto_ops inet6_stream_ops = {
474 .owner = THIS_MODULE,
475 .release = inet6_release,
477 .connect = inet_stream_connect, /* ok */
478 .socketpair = sock_no_socketpair, /* a do nothing */
479 .accept = inet_accept, /* ok */
480 .getname = inet6_getname,
481 .poll = tcp_poll, /* ok */
482 .ioctl = inet6_ioctl, /* must change */
483 .listen = inet_listen, /* ok */
484 .shutdown = inet_shutdown, /* ok */
485 .setsockopt = sock_common_setsockopt, /* ok */
486 .getsockopt = sock_common_getsockopt, /* ok */
487 .sendmsg = inet_sendmsg, /* ok */
488 .recvmsg = sock_common_recvmsg, /* ok */
489 .mmap = sock_no_mmap,
490 .sendpage = tcp_sendpage,
492 .compat_setsockopt = compat_sock_common_setsockopt,
493 .compat_getsockopt = compat_sock_common_getsockopt,
497 const struct proto_ops inet6_dgram_ops = {
499 .owner = THIS_MODULE,
500 .release = inet6_release,
502 .connect = inet_dgram_connect, /* ok */
503 .socketpair = sock_no_socketpair, /* a do nothing */
504 .accept = sock_no_accept, /* a do nothing */
505 .getname = inet6_getname,
506 .poll = udp_poll, /* ok */
507 .ioctl = inet6_ioctl, /* must change */
508 .listen = sock_no_listen, /* ok */
509 .shutdown = inet_shutdown, /* ok */
510 .setsockopt = sock_common_setsockopt, /* ok */
511 .getsockopt = sock_common_getsockopt, /* ok */
512 .sendmsg = inet_sendmsg, /* ok */
513 .recvmsg = sock_common_recvmsg, /* ok */
514 .mmap = sock_no_mmap,
515 .sendpage = sock_no_sendpage,
517 .compat_setsockopt = compat_sock_common_setsockopt,
518 .compat_getsockopt = compat_sock_common_getsockopt,
522 static struct net_proto_family inet6_family_ops = {
524 .create = inet6_create,
525 .owner = THIS_MODULE,
528 /* Same as inet6_dgram_ops, sans udp_poll. */
529 static const struct proto_ops inet6_sockraw_ops = {
531 .owner = THIS_MODULE,
532 .release = inet6_release,
534 .connect = inet_dgram_connect, /* ok */
535 .socketpair = sock_no_socketpair, /* a do nothing */
536 .accept = sock_no_accept, /* a do nothing */
537 .getname = inet6_getname,
538 .poll = datagram_poll, /* ok */
539 .ioctl = inet6_ioctl, /* must change */
540 .listen = sock_no_listen, /* ok */
541 .shutdown = inet_shutdown, /* ok */
542 .setsockopt = sock_common_setsockopt, /* ok */
543 .getsockopt = sock_common_getsockopt, /* ok */
544 .sendmsg = inet_sendmsg, /* ok */
545 .recvmsg = sock_common_recvmsg, /* ok */
546 .mmap = sock_no_mmap,
547 .sendpage = sock_no_sendpage,
549 .compat_setsockopt = compat_sock_common_setsockopt,
550 .compat_getsockopt = compat_sock_common_getsockopt,
554 static struct inet_protosw rawv6_protosw = {
556 .protocol = IPPROTO_IP, /* wild card */
558 .ops = &inet6_sockraw_ops,
559 .capability = CAP_NET_RAW,
560 .no_check = UDP_CSUM_DEFAULT,
561 .flags = INET_PROTOSW_REUSE,
565 inet6_register_protosw(struct inet_protosw *p)
567 struct list_head *lh;
568 struct inet_protosw *answer;
569 int protocol = p->protocol;
570 struct list_head *last_perm;
572 spin_lock_bh(&inetsw6_lock);
574 if (p->type >= SOCK_MAX)
577 /* If we are trying to override a permanent protocol, bail. */
579 last_perm = &inetsw6[p->type];
580 list_for_each(lh, &inetsw6[p->type]) {
581 answer = list_entry(lh, struct inet_protosw, list);
583 /* Check only the non-wild match. */
584 if (INET_PROTOSW_PERMANENT & answer->flags) {
585 if (protocol == answer->protocol)
595 /* Add the new entry after the last permanent entry if any, so that
596 * the new entry does not override a permanent entry when matched with
597 * a wild-card protocol. But it is allowed to override any existing
598 * non-permanent entry. This means that when we remove this entry, the
599 * system automatically returns to the old behavior.
601 list_add_rcu(&p->list, last_perm);
603 spin_unlock_bh(&inetsw6_lock);
607 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
613 "Ignoring attempt to register invalid socket type %d.\n",
618 EXPORT_SYMBOL(inet6_register_protosw);
621 inet6_unregister_protosw(struct inet_protosw *p)
623 if (INET_PROTOSW_PERMANENT & p->flags) {
625 "Attempt to unregister permanent protocol %d.\n",
628 spin_lock_bh(&inetsw6_lock);
629 list_del_rcu(&p->list);
630 spin_unlock_bh(&inetsw6_lock);
636 EXPORT_SYMBOL(inet6_unregister_protosw);
638 int inet6_sk_rebuild_header(struct sock *sk)
641 struct dst_entry *dst;
642 struct ipv6_pinfo *np = inet6_sk(sk);
644 dst = __sk_dst_check(sk, np->dst_cookie);
647 struct inet_sock *inet = inet_sk(sk);
648 struct in6_addr *final_p = NULL, final;
651 memset(&fl, 0, sizeof(fl));
652 fl.proto = sk->sk_protocol;
653 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
654 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
655 fl.fl6_flowlabel = np->flow_label;
656 fl.oif = sk->sk_bound_dev_if;
657 fl.fl_ip_dport = inet->dport;
658 fl.fl_ip_sport = inet->sport;
659 security_sk_classify_flow(sk, &fl);
661 if (np->opt && np->opt->srcrt) {
662 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
663 ipv6_addr_copy(&final, &fl.fl6_dst);
664 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
668 err = ip6_dst_lookup(sk, &dst, &fl);
670 sk->sk_route_caps = 0;
674 ipv6_addr_copy(&fl.fl6_dst, final_p);
676 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
677 sk->sk_err_soft = -err;
681 __ip6_dst_store(sk, dst, NULL, NULL);
687 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
689 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
691 struct ipv6_pinfo *np = inet6_sk(sk);
692 struct inet6_skb_parm *opt = IP6CB(skb);
695 if ((opt->hop && (np->rxopt.bits.hopopts ||
696 np->rxopt.bits.ohopopts)) ||
697 ((IPV6_FLOWINFO_MASK &
698 *(__be32 *)skb_network_header(skb)) &&
699 np->rxopt.bits.rxflow) ||
700 (opt->srcrt && (np->rxopt.bits.srcrt ||
701 np->rxopt.bits.osrcrt)) ||
702 ((opt->dst1 || opt->dst0) &&
703 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
709 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
711 static int __init init_ipv6_mibs(void)
713 if (snmp_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
714 __alignof__(struct ipstats_mib)) < 0)
716 if (snmp_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
717 __alignof__(struct icmpv6_mib)) < 0)
719 if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
720 __alignof__(struct udp_mib)) < 0)
722 if (snmp_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
723 __alignof__(struct udp_mib)) < 0)
724 goto err_udplite_mib;
728 snmp_mib_free((void **)udp_stats_in6);
730 snmp_mib_free((void **)icmpv6_statistics);
732 snmp_mib_free((void **)ipv6_statistics);
738 static void cleanup_ipv6_mibs(void)
740 snmp_mib_free((void **)ipv6_statistics);
741 snmp_mib_free((void **)icmpv6_statistics);
742 snmp_mib_free((void **)udp_stats_in6);
743 snmp_mib_free((void **)udplite_stats_in6);
746 static int __init inet6_init(void)
748 struct sk_buff *dummy_skb;
752 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
755 #if 0 /* FIXME --RR */
756 if (!mod_member_present(&__this_module, can_unload))
759 __this_module.can_unload = &ipv6_unload;
763 err = proto_register(&tcpv6_prot, 1);
767 err = proto_register(&udpv6_prot, 1);
769 goto out_unregister_tcp_proto;
771 err = proto_register(&udplitev6_prot, 1);
773 goto out_unregister_udp_proto;
775 err = proto_register(&rawv6_prot, 1);
777 goto out_unregister_udplite_proto;
780 /* Register the socket-side information for inet6_create. */
781 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
784 /* We MUST register RAW sockets before we create the ICMP6,
785 * IGMP6, or NDISC control sockets.
787 inet6_register_protosw(&rawv6_protosw);
789 /* Register the family here so that the init calls below will
790 * be able to create sockets. (?? is this dangerous ??)
792 err = sock_register(&inet6_family_ops);
794 goto out_unregister_raw_proto;
796 /* Initialise ipv6 mibs */
797 err = init_ipv6_mibs();
799 goto out_unregister_sock;
802 * ipngwg API draft makes clear that the correct semantics
803 * for TCP and UDP is to consider one TCP and UDP instance
804 * in a host availiable by both INET and INET6 APIs and
805 * able to communicate via both network protocols.
809 ipv6_sysctl_register();
811 err = icmpv6_init(&inet6_family_ops);
814 err = ndisc_init(&inet6_family_ops);
817 err = igmp6_init(&inet6_family_ops);
820 err = ipv6_netfilter_init();
823 /* Create /proc/foo6 entries. */
824 #ifdef CONFIG_PROC_FS
826 if (raw6_proc_init())
828 if (tcp6_proc_init())
830 if (udp6_proc_init())
832 if (udplite6_proc_init())
833 goto proc_udplite6_fail;
834 if (ipv6_misc_proc_init())
835 goto proc_misc6_fail;
838 goto proc_anycast6_fail;
843 ip6_flowlabel_init();
844 err = addrconf_init();
848 /* Init v6 extension headers. */
854 /* Init v6 transport protocols. */
865 ip6_flowlabel_cleanup();
867 #ifdef CONFIG_PROC_FS
872 ipv6_misc_proc_exit();
874 udplite6_proc_exit();
883 ipv6_netfilter_fini();
892 ipv6_sysctl_unregister();
896 sock_unregister(PF_INET6);
897 out_unregister_raw_proto:
898 proto_unregister(&rawv6_prot);
899 out_unregister_udplite_proto:
900 proto_unregister(&udplitev6_prot);
901 out_unregister_udp_proto:
902 proto_unregister(&udpv6_prot);
903 out_unregister_tcp_proto:
904 proto_unregister(&tcpv6_prot);
907 module_init(inet6_init);
909 static void __exit inet6_exit(void)
911 /* First of all disallow new sockets creation. */
912 sock_unregister(PF_INET6);
913 /* Disallow any further netlink messages */
914 rtnl_unregister_all(PF_INET6);
916 /* Cleanup code parts. */
917 ipv6_packet_cleanup();
920 ip6_flowlabel_cleanup();
922 #ifdef CONFIG_PROC_FS
924 /* Cleanup code parts. */
927 ipv6_misc_proc_exit();
928 udplite6_proc_exit();
933 ipv6_netfilter_fini();
938 ipv6_sysctl_unregister();
941 proto_unregister(&rawv6_prot);
942 proto_unregister(&udplitev6_prot);
943 proto_unregister(&udpv6_prot);
944 proto_unregister(&tcpv6_prot);
946 module_exit(inet6_exit);
948 MODULE_ALIAS_NETPROTO(PF_INET6);