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/sched.h>
32 #include <linux/timer.h>
33 #include <linux/string.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/fcntl.h>
38 #include <linux/interrupt.h>
39 #include <linux/proc_fs.h>
40 #include <linux/stat.h>
41 #include <linux/init.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <linux/icmpv6.h>
46 #include <linux/smp_lock.h>
47 #include <linux/netfilter_ipv6.h>
52 #include <net/udplite.h>
55 #include <net/protocol.h>
56 #include <net/inet_common.h>
57 #include <net/transp_v6.h>
58 #include <net/ip6_route.h>
59 #include <net/addrconf.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
63 #ifdef CONFIG_IPV6_MIP6
67 #include <asm/uaccess.h>
68 #include <asm/system.h>
70 MODULE_AUTHOR("Cast of dozens");
71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
72 MODULE_LICENSE("GPL");
74 int sysctl_ipv6_bindv6only __read_mostly;
76 /* The inetsw table contains everything that inet_create needs to
79 static struct list_head inetsw6[SOCK_MAX];
80 static DEFINE_SPINLOCK(inetsw6_lock);
82 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
84 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
86 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
89 static int inet6_create(struct socket *sock, int protocol)
91 struct inet_sock *inet;
92 struct ipv6_pinfo *np;
95 struct inet_protosw *answer;
96 struct proto *answer_prot;
97 unsigned char answer_flags;
99 int try_loading_module = 0;
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) == INET_PROTOSW_ICSK;
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 int inet6_release(struct socket *sock)
355 struct sock *sk = sock->sk;
361 ipv6_sock_mc_close(sk);
364 ipv6_sock_ac_close(sk);
366 return inet_release(sock);
369 int inet6_destroy_sock(struct sock *sk)
371 struct ipv6_pinfo *np = inet6_sk(sk);
373 struct ipv6_txoptions *opt;
375 /* Release rx options */
377 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
380 /* Free flowlabels */
381 fl6_free_socklist(sk);
383 /* Free tx options */
385 if ((opt = xchg(&np->opt, NULL)) != NULL)
386 sock_kfree_s(sk, opt, opt->tot_len);
391 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
394 * This does both peername and sockname.
397 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
398 int *uaddr_len, int peer)
400 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
401 struct sock *sk = sock->sk;
402 struct inet_sock *inet = inet_sk(sk);
403 struct ipv6_pinfo *np = inet6_sk(sk);
405 sin->sin6_family = AF_INET6;
406 sin->sin6_flowinfo = 0;
407 sin->sin6_scope_id = 0;
411 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
414 sin->sin6_port = inet->dport;
415 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
417 sin->sin6_flowinfo = np->flow_label;
419 if (ipv6_addr_any(&np->rcv_saddr))
420 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
422 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
424 sin->sin6_port = inet->sport;
426 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
427 sin->sin6_scope_id = sk->sk_bound_dev_if;
428 *uaddr_len = sizeof(*sin);
432 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
434 struct sock *sk = sock->sk;
439 return sock_get_timestamp(sk, (struct timeval __user *)arg);
444 return(ipv6_route_ioctl(cmd,(void __user *)arg));
447 return addrconf_add_ifaddr((void __user *) arg);
449 return addrconf_del_ifaddr((void __user *) arg);
451 return addrconf_set_dstaddr((void __user *) arg);
453 if (!sk->sk_prot->ioctl)
455 return sk->sk_prot->ioctl(sk, cmd, arg);
461 const struct proto_ops inet6_stream_ops = {
463 .owner = THIS_MODULE,
464 .release = inet6_release,
466 .connect = inet_stream_connect, /* ok */
467 .socketpair = sock_no_socketpair, /* a do nothing */
468 .accept = inet_accept, /* ok */
469 .getname = inet6_getname,
470 .poll = tcp_poll, /* ok */
471 .ioctl = inet6_ioctl, /* must change */
472 .listen = inet_listen, /* ok */
473 .shutdown = inet_shutdown, /* ok */
474 .setsockopt = sock_common_setsockopt, /* ok */
475 .getsockopt = sock_common_getsockopt, /* ok */
476 .sendmsg = inet_sendmsg, /* ok */
477 .recvmsg = sock_common_recvmsg, /* ok */
478 .mmap = sock_no_mmap,
479 .sendpage = tcp_sendpage,
481 .compat_setsockopt = compat_sock_common_setsockopt,
482 .compat_getsockopt = compat_sock_common_getsockopt,
486 const struct proto_ops inet6_dgram_ops = {
488 .owner = THIS_MODULE,
489 .release = inet6_release,
491 .connect = inet_dgram_connect, /* ok */
492 .socketpair = sock_no_socketpair, /* a do nothing */
493 .accept = sock_no_accept, /* a do nothing */
494 .getname = inet6_getname,
495 .poll = udp_poll, /* ok */
496 .ioctl = inet6_ioctl, /* must change */
497 .listen = sock_no_listen, /* ok */
498 .shutdown = inet_shutdown, /* ok */
499 .setsockopt = sock_common_setsockopt, /* ok */
500 .getsockopt = sock_common_getsockopt, /* ok */
501 .sendmsg = inet_sendmsg, /* ok */
502 .recvmsg = sock_common_recvmsg, /* ok */
503 .mmap = sock_no_mmap,
504 .sendpage = sock_no_sendpage,
506 .compat_setsockopt = compat_sock_common_setsockopt,
507 .compat_getsockopt = compat_sock_common_getsockopt,
511 static struct net_proto_family inet6_family_ops = {
513 .create = inet6_create,
514 .owner = THIS_MODULE,
517 /* Same as inet6_dgram_ops, sans udp_poll. */
518 static const struct proto_ops inet6_sockraw_ops = {
520 .owner = THIS_MODULE,
521 .release = inet6_release,
523 .connect = inet_dgram_connect, /* ok */
524 .socketpair = sock_no_socketpair, /* a do nothing */
525 .accept = sock_no_accept, /* a do nothing */
526 .getname = inet6_getname,
527 .poll = datagram_poll, /* ok */
528 .ioctl = inet6_ioctl, /* must change */
529 .listen = sock_no_listen, /* ok */
530 .shutdown = inet_shutdown, /* ok */
531 .setsockopt = sock_common_setsockopt, /* ok */
532 .getsockopt = sock_common_getsockopt, /* ok */
533 .sendmsg = inet_sendmsg, /* ok */
534 .recvmsg = sock_common_recvmsg, /* ok */
535 .mmap = sock_no_mmap,
536 .sendpage = sock_no_sendpage,
538 .compat_setsockopt = compat_sock_common_setsockopt,
539 .compat_getsockopt = compat_sock_common_getsockopt,
543 static struct inet_protosw rawv6_protosw = {
545 .protocol = IPPROTO_IP, /* wild card */
547 .ops = &inet6_sockraw_ops,
548 .capability = CAP_NET_RAW,
549 .no_check = UDP_CSUM_DEFAULT,
550 .flags = INET_PROTOSW_REUSE,
554 inet6_register_protosw(struct inet_protosw *p)
556 struct list_head *lh;
557 struct inet_protosw *answer;
558 int protocol = p->protocol;
559 struct list_head *last_perm;
561 spin_lock_bh(&inetsw6_lock);
563 if (p->type >= SOCK_MAX)
566 /* If we are trying to override a permanent protocol, bail. */
568 last_perm = &inetsw6[p->type];
569 list_for_each(lh, &inetsw6[p->type]) {
570 answer = list_entry(lh, struct inet_protosw, list);
572 /* Check only the non-wild match. */
573 if (INET_PROTOSW_PERMANENT & answer->flags) {
574 if (protocol == answer->protocol)
584 /* Add the new entry after the last permanent entry if any, so that
585 * the new entry does not override a permanent entry when matched with
586 * a wild-card protocol. But it is allowed to override any existing
587 * non-permanent entry. This means that when we remove this entry, the
588 * system automatically returns to the old behavior.
590 list_add_rcu(&p->list, last_perm);
592 spin_unlock_bh(&inetsw6_lock);
596 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
602 "Ignoring attempt to register invalid socket type %d.\n",
608 inet6_unregister_protosw(struct inet_protosw *p)
610 if (INET_PROTOSW_PERMANENT & p->flags) {
612 "Attempt to unregister permanent protocol %d.\n",
615 spin_lock_bh(&inetsw6_lock);
616 list_del_rcu(&p->list);
617 spin_unlock_bh(&inetsw6_lock);
623 int inet6_sk_rebuild_header(struct sock *sk)
626 struct dst_entry *dst;
627 struct ipv6_pinfo *np = inet6_sk(sk);
629 dst = __sk_dst_check(sk, np->dst_cookie);
632 struct inet_sock *inet = inet_sk(sk);
633 struct in6_addr *final_p = NULL, final;
636 memset(&fl, 0, sizeof(fl));
637 fl.proto = sk->sk_protocol;
638 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
639 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
640 fl.fl6_flowlabel = np->flow_label;
641 fl.oif = sk->sk_bound_dev_if;
642 fl.fl_ip_dport = inet->dport;
643 fl.fl_ip_sport = inet->sport;
644 security_sk_classify_flow(sk, &fl);
646 if (np->opt && np->opt->srcrt) {
647 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
648 ipv6_addr_copy(&final, &fl.fl6_dst);
649 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
653 err = ip6_dst_lookup(sk, &dst, &fl);
655 sk->sk_route_caps = 0;
659 ipv6_addr_copy(&fl.fl6_dst, final_p);
661 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
662 sk->sk_err_soft = -err;
666 __ip6_dst_store(sk, dst, NULL, NULL);
672 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
674 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
676 struct ipv6_pinfo *np = inet6_sk(sk);
677 struct inet6_skb_parm *opt = IP6CB(skb);
680 if ((opt->hop && (np->rxopt.bits.hopopts ||
681 np->rxopt.bits.ohopopts)) ||
682 ((IPV6_FLOWINFO_MASK & *(__be32*)skb->nh.raw) &&
683 np->rxopt.bits.rxflow) ||
684 (opt->srcrt && (np->rxopt.bits.srcrt ||
685 np->rxopt.bits.osrcrt)) ||
686 ((opt->dst1 || opt->dst0) &&
687 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
693 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
696 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
701 ptr[0] = __alloc_percpu(mibsize);
705 ptr[1] = __alloc_percpu(mibsize);
719 snmp6_mib_free(void *ptr[2])
725 ptr[0] = ptr[1] = NULL;
728 static int __init init_ipv6_mibs(void)
730 if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
731 __alignof__(struct ipstats_mib)) < 0)
733 if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
734 __alignof__(struct icmpv6_mib)) < 0)
736 if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
737 __alignof__(struct udp_mib)) < 0)
739 if (snmp6_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
740 __alignof__(struct udp_mib)) < 0)
741 goto err_udplite_mib;
745 snmp6_mib_free((void **)udp_stats_in6);
747 snmp6_mib_free((void **)icmpv6_statistics);
749 snmp6_mib_free((void **)ipv6_statistics);
755 static void cleanup_ipv6_mibs(void)
757 snmp6_mib_free((void **)ipv6_statistics);
758 snmp6_mib_free((void **)icmpv6_statistics);
759 snmp6_mib_free((void **)udp_stats_in6);
760 snmp6_mib_free((void **)udplite_stats_in6);
763 static int __init inet6_init(void)
765 struct sk_buff *dummy_skb;
769 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
772 #if 0 /* FIXME --RR */
773 if (!mod_member_present(&__this_module, can_unload))
776 __this_module.can_unload = &ipv6_unload;
780 err = proto_register(&tcpv6_prot, 1);
784 err = proto_register(&udpv6_prot, 1);
786 goto out_unregister_tcp_proto;
788 err = proto_register(&udplitev6_prot, 1);
790 goto out_unregister_udp_proto;
792 err = proto_register(&rawv6_prot, 1);
794 goto out_unregister_udplite_proto;
797 /* Register the socket-side information for inet6_create. */
798 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
801 /* We MUST register RAW sockets before we create the ICMP6,
802 * IGMP6, or NDISC control sockets.
804 inet6_register_protosw(&rawv6_protosw);
806 /* Register the family here so that the init calls below will
807 * be able to create sockets. (?? is this dangerous ??)
809 err = sock_register(&inet6_family_ops);
811 goto out_unregister_raw_proto;
813 /* Initialise ipv6 mibs */
814 err = init_ipv6_mibs();
816 goto out_unregister_sock;
819 * ipngwg API draft makes clear that the correct semantics
820 * for TCP and UDP is to consider one TCP and UDP instance
821 * in a host availiable by both INET and INET6 APIs and
822 * able to communicate via both network protocols.
826 ipv6_sysctl_register();
828 err = icmpv6_init(&inet6_family_ops);
831 err = ndisc_init(&inet6_family_ops);
834 err = igmp6_init(&inet6_family_ops);
837 err = ipv6_netfilter_init();
840 /* Create /proc/foo6 entries. */
841 #ifdef CONFIG_PROC_FS
843 if (raw6_proc_init())
845 if (tcp6_proc_init())
847 if (udp6_proc_init())
849 if (udplite6_proc_init())
850 goto proc_udplite6_fail;
851 if (ipv6_misc_proc_init())
852 goto proc_misc6_fail;
855 goto proc_anycast6_fail;
860 ip6_flowlabel_init();
861 err = addrconf_init();
865 /* Init v6 extension headers. */
870 #ifdef CONFIG_IPV6_MIP6
874 /* Init v6 transport protocols. */
885 ip6_flowlabel_cleanup();
887 #ifdef CONFIG_PROC_FS
892 ipv6_misc_proc_exit();
894 udplite6_proc_exit();
903 ipv6_netfilter_fini();
912 ipv6_sysctl_unregister();
916 sock_unregister(PF_INET6);
917 out_unregister_raw_proto:
918 proto_unregister(&rawv6_prot);
919 out_unregister_udplite_proto:
920 proto_unregister(&udplitev6_prot);
921 out_unregister_udp_proto:
922 proto_unregister(&udpv6_prot);
923 out_unregister_tcp_proto:
924 proto_unregister(&tcpv6_prot);
927 module_init(inet6_init);
929 static void __exit inet6_exit(void)
931 /* First of all disallow new sockets creation. */
932 sock_unregister(PF_INET6);
933 #ifdef CONFIG_PROC_FS
936 ipv6_misc_proc_exit();
938 udplite6_proc_exit();
942 #ifdef CONFIG_IPV6_MIP6
945 /* Cleanup code parts. */
946 ip6_flowlabel_cleanup();
949 ipv6_packet_cleanup();
951 ipv6_netfilter_fini();
955 ipv6_sysctl_unregister();
958 proto_unregister(&rawv6_prot);
959 proto_unregister(&udpv6_prot);
960 proto_unregister(&tcpv6_prot);
962 module_exit(inet6_exit);
964 MODULE_ALIAS_NETPROTO(PF_INET6);