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/config.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>
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>
63 #include <asm/uaccess.h>
64 #include <asm/system.h>
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
70 int sysctl_ipv6_bindv6only;
72 /* The inetsw table contains everything that inet_create needs to
75 static struct list_head inetsw6[SOCK_MAX];
76 static DEFINE_SPINLOCK(inetsw6_lock);
78 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
80 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
82 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
85 static int inet6_create(struct socket *sock, int protocol)
87 struct inet_sock *inet;
88 struct ipv6_pinfo *np;
91 struct inet_protosw *answer;
92 struct proto *answer_prot;
93 unsigned char answer_flags;
95 int try_loading_module = 0;
98 /* Look for the requested type/protocol pair. */
101 err = -ESOCKTNOSUPPORT;
103 list_for_each_rcu(p, &inetsw6[sock->type]) {
104 answer = list_entry(p, struct inet_protosw, list);
106 /* Check the non-wild match. */
107 if (protocol == answer->protocol) {
108 if (protocol != IPPROTO_IP)
111 /* Check for the two wild cases. */
112 if (IPPROTO_IP == protocol) {
113 protocol = answer->protocol;
116 if (IPPROTO_IP == answer->protocol)
119 err = -EPROTONOSUPPORT;
124 if (try_loading_module < 2) {
127 * Be more specific, e.g. net-pf-10-proto-132-type-1
128 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
130 if (++try_loading_module == 1)
131 request_module("net-pf-%d-proto-%d-type-%d",
132 PF_INET6, protocol, sock->type);
134 * Fall back to generic, e.g. net-pf-10-proto-132
135 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
138 request_module("net-pf-%d-proto-%d",
140 goto lookup_protocol;
146 if (answer->capability > 0 && !capable(answer->capability))
149 sock->ops = answer->ops;
150 answer_prot = answer->prot;
151 answer_no_check = answer->no_check;
152 answer_flags = answer->flags;
155 BUG_TRAP(answer_prot->slab != NULL);
158 sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
162 sock_init_data(sock, sk);
165 sk->sk_no_check = answer_no_check;
166 if (INET_PROTOSW_REUSE & answer_flags)
170 inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
172 if (SOCK_RAW == sock->type) {
173 inet->num = protocol;
174 if (IPPROTO_RAW == protocol)
178 sk->sk_destruct = inet_sock_destruct;
179 sk->sk_family = PF_INET6;
180 sk->sk_protocol = protocol;
182 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
184 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
188 np->pmtudisc = IPV6_PMTUDISC_WANT;
189 np->ipv6only = sysctl_ipv6_bindv6only;
191 /* Init the ipv4 part of the socket since we can have sockets
192 * using v6 API for ipv4.
199 inet->mc_list = NULL;
201 if (ipv4_config.no_pmtu_disc)
202 inet->pmtudisc = IP_PMTUDISC_DONT;
204 inet->pmtudisc = IP_PMTUDISC_WANT;
206 * Increment only the relevant sk_prot->socks debug field, this changes
207 * the previous behaviour of incrementing both the equivalent to
208 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
210 * This allows better debug granularity as we'll know exactly how many
211 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
212 * transport protocol socks. -acme
214 sk_refcnt_debug_inc(sk);
217 /* It assumes that any protocol which allows
218 * the user to assign a number at socket
219 * creation time automatically shares.
221 inet->sport = ntohs(inet->num);
222 sk->sk_prot->hash(sk);
224 if (sk->sk_prot->init) {
225 err = sk->sk_prot->init(sk);
227 sk_common_release(sk);
239 /* bind for INET6 API */
240 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
242 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
243 struct sock *sk = sock->sk;
244 struct inet_sock *inet = inet_sk(sk);
245 struct ipv6_pinfo *np = inet6_sk(sk);
251 /* If the socket has its own bind function then use it. */
252 if (sk->sk_prot->bind)
253 return sk->sk_prot->bind(sk, uaddr, addr_len);
255 if (addr_len < SIN6_LEN_RFC2133)
257 addr_type = ipv6_addr_type(&addr->sin6_addr);
258 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
261 snum = ntohs(addr->sin6_port);
262 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
267 /* Check these errors (active socket, double bind). */
268 if (sk->sk_state != TCP_CLOSE || inet->num) {
273 /* Check if the address belongs to the host. */
274 if (addr_type == IPV6_ADDR_MAPPED) {
275 v4addr = addr->sin6_addr.s6_addr32[3];
276 if (inet_addr_type(v4addr) != RTN_LOCAL) {
277 err = -EADDRNOTAVAIL;
281 if (addr_type != IPV6_ADDR_ANY) {
282 struct net_device *dev = NULL;
284 if (addr_type & IPV6_ADDR_LINKLOCAL) {
285 if (addr_len >= sizeof(struct sockaddr_in6) &&
286 addr->sin6_scope_id) {
287 /* Override any existing binding, if another one
288 * is supplied by user.
290 sk->sk_bound_dev_if = addr->sin6_scope_id;
293 /* Binding to link-local address requires an interface */
294 if (!sk->sk_bound_dev_if) {
298 dev = dev_get_by_index(sk->sk_bound_dev_if);
305 /* ipv4 addr of the socket is invalid. Only the
306 * unspecified and mapped address have a v4 equivalent.
308 v4addr = LOOPBACK4_IPV6;
309 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
310 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
313 err = -EADDRNOTAVAIL;
322 inet->rcv_saddr = v4addr;
323 inet->saddr = v4addr;
325 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
327 if (!(addr_type & IPV6_ADDR_MULTICAST))
328 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
330 /* Make sure we are allowed to bind here. */
331 if (sk->sk_prot->get_port(sk, snum)) {
332 inet_reset_saddr(sk);
337 if (addr_type != IPV6_ADDR_ANY)
338 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
340 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
341 inet->sport = ntohs(inet->num);
349 int inet6_release(struct socket *sock)
351 struct sock *sk = sock->sk;
357 ipv6_sock_mc_close(sk);
360 ipv6_sock_ac_close(sk);
362 return inet_release(sock);
365 int inet6_destroy_sock(struct sock *sk)
367 struct ipv6_pinfo *np = inet6_sk(sk);
369 struct ipv6_txoptions *opt;
372 * Release destination entry
377 /* Release rx options */
379 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
382 /* Free flowlabels */
383 fl6_free_socklist(sk);
385 /* Free tx options */
387 if ((opt = xchg(&np->opt, NULL)) != NULL)
388 sock_kfree_s(sk, opt, opt->tot_len);
393 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
396 * This does both peername and sockname.
399 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
400 int *uaddr_len, int peer)
402 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
403 struct sock *sk = sock->sk;
404 struct inet_sock *inet = inet_sk(sk);
405 struct ipv6_pinfo *np = inet6_sk(sk);
407 sin->sin6_family = AF_INET6;
408 sin->sin6_flowinfo = 0;
409 sin->sin6_scope_id = 0;
413 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
416 sin->sin6_port = inet->dport;
417 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
419 sin->sin6_flowinfo = np->flow_label;
421 if (ipv6_addr_any(&np->rcv_saddr))
422 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
424 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
426 sin->sin6_port = inet->sport;
428 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
429 sin->sin6_scope_id = sk->sk_bound_dev_if;
430 *uaddr_len = sizeof(*sin);
434 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
436 struct sock *sk = sock->sk;
441 return sock_get_timestamp(sk, (struct timeval __user *)arg);
446 return(ipv6_route_ioctl(cmd,(void __user *)arg));
449 return addrconf_add_ifaddr((void __user *) arg);
451 return addrconf_del_ifaddr((void __user *) arg);
453 return addrconf_set_dstaddr((void __user *) arg);
455 if (!sk->sk_prot->ioctl)
457 return sk->sk_prot->ioctl(sk, cmd, arg);
463 const struct proto_ops inet6_stream_ops = {
465 .owner = THIS_MODULE,
466 .release = inet6_release,
468 .connect = inet_stream_connect, /* ok */
469 .socketpair = sock_no_socketpair, /* a do nothing */
470 .accept = inet_accept, /* ok */
471 .getname = inet6_getname,
472 .poll = tcp_poll, /* ok */
473 .ioctl = inet6_ioctl, /* must change */
474 .listen = inet_listen, /* ok */
475 .shutdown = inet_shutdown, /* ok */
476 .setsockopt = sock_common_setsockopt, /* ok */
477 .getsockopt = sock_common_getsockopt, /* ok */
478 .sendmsg = inet_sendmsg, /* ok */
479 .recvmsg = sock_common_recvmsg, /* ok */
480 .mmap = sock_no_mmap,
481 .sendpage = tcp_sendpage
484 const struct proto_ops inet6_dgram_ops = {
486 .owner = THIS_MODULE,
487 .release = inet6_release,
489 .connect = inet_dgram_connect, /* ok */
490 .socketpair = sock_no_socketpair, /* a do nothing */
491 .accept = sock_no_accept, /* a do nothing */
492 .getname = inet6_getname,
493 .poll = udp_poll, /* ok */
494 .ioctl = inet6_ioctl, /* must change */
495 .listen = sock_no_listen, /* ok */
496 .shutdown = inet_shutdown, /* ok */
497 .setsockopt = sock_common_setsockopt, /* ok */
498 .getsockopt = sock_common_getsockopt, /* ok */
499 .sendmsg = inet_sendmsg, /* ok */
500 .recvmsg = sock_common_recvmsg, /* ok */
501 .mmap = sock_no_mmap,
502 .sendpage = sock_no_sendpage,
505 static struct net_proto_family inet6_family_ops = {
507 .create = inet6_create,
508 .owner = THIS_MODULE,
511 /* Same as inet6_dgram_ops, sans udp_poll. */
512 static const struct proto_ops inet6_sockraw_ops = {
514 .owner = THIS_MODULE,
515 .release = inet6_release,
517 .connect = inet_dgram_connect, /* ok */
518 .socketpair = sock_no_socketpair, /* a do nothing */
519 .accept = sock_no_accept, /* a do nothing */
520 .getname = inet6_getname,
521 .poll = datagram_poll, /* ok */
522 .ioctl = inet6_ioctl, /* must change */
523 .listen = sock_no_listen, /* ok */
524 .shutdown = inet_shutdown, /* ok */
525 .setsockopt = sock_common_setsockopt, /* ok */
526 .getsockopt = sock_common_getsockopt, /* ok */
527 .sendmsg = inet_sendmsg, /* ok */
528 .recvmsg = sock_common_recvmsg, /* ok */
529 .mmap = sock_no_mmap,
530 .sendpage = sock_no_sendpage,
533 static struct inet_protosw rawv6_protosw = {
535 .protocol = IPPROTO_IP, /* wild card */
537 .ops = &inet6_sockraw_ops,
538 .capability = CAP_NET_RAW,
539 .no_check = UDP_CSUM_DEFAULT,
540 .flags = INET_PROTOSW_REUSE,
544 inet6_register_protosw(struct inet_protosw *p)
546 struct list_head *lh;
547 struct inet_protosw *answer;
548 int protocol = p->protocol;
549 struct list_head *last_perm;
551 spin_lock_bh(&inetsw6_lock);
553 if (p->type >= SOCK_MAX)
556 /* If we are trying to override a permanent protocol, bail. */
558 last_perm = &inetsw6[p->type];
559 list_for_each(lh, &inetsw6[p->type]) {
560 answer = list_entry(lh, struct inet_protosw, list);
562 /* Check only the non-wild match. */
563 if (INET_PROTOSW_PERMANENT & answer->flags) {
564 if (protocol == answer->protocol)
574 /* Add the new entry after the last permanent entry if any, so that
575 * the new entry does not override a permanent entry when matched with
576 * a wild-card protocol. But it is allowed to override any existing
577 * non-permanent entry. This means that when we remove this entry, the
578 * system automatically returns to the old behavior.
580 list_add_rcu(&p->list, last_perm);
582 spin_unlock_bh(&inetsw6_lock);
586 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
592 "Ignoring attempt to register invalid socket type %d.\n",
598 inet6_unregister_protosw(struct inet_protosw *p)
600 if (INET_PROTOSW_PERMANENT & p->flags) {
602 "Attempt to unregister permanent protocol %d.\n",
605 spin_lock_bh(&inetsw6_lock);
606 list_del_rcu(&p->list);
607 spin_unlock_bh(&inetsw6_lock);
613 int inet6_sk_rebuild_header(struct sock *sk)
616 struct dst_entry *dst;
617 struct ipv6_pinfo *np = inet6_sk(sk);
619 dst = __sk_dst_check(sk, np->dst_cookie);
622 struct inet_sock *inet = inet_sk(sk);
623 struct in6_addr *final_p = NULL, final;
626 memset(&fl, 0, sizeof(fl));
627 fl.proto = sk->sk_protocol;
628 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
629 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
630 fl.fl6_flowlabel = np->flow_label;
631 fl.oif = sk->sk_bound_dev_if;
632 fl.fl_ip_dport = inet->dport;
633 fl.fl_ip_sport = inet->sport;
635 if (np->opt && np->opt->srcrt) {
636 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
637 ipv6_addr_copy(&final, &fl.fl6_dst);
638 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
642 err = ip6_dst_lookup(sk, &dst, &fl);
644 sk->sk_route_caps = 0;
648 ipv6_addr_copy(&fl.fl6_dst, final_p);
650 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
651 sk->sk_err_soft = -err;
655 ip6_dst_store(sk, dst, NULL);
656 sk->sk_route_caps = dst->dev->features &
657 ~(NETIF_F_IP_CSUM | NETIF_F_TSO);
663 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
665 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
667 struct ipv6_pinfo *np = inet6_sk(sk);
668 struct inet6_skb_parm *opt = IP6CB(skb);
671 if ((opt->hop && (np->rxopt.bits.hopopts ||
672 np->rxopt.bits.ohopopts)) ||
673 ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
674 np->rxopt.bits.rxflow) ||
675 (opt->srcrt && (np->rxopt.bits.srcrt ||
676 np->rxopt.bits.osrcrt)) ||
677 ((opt->dst1 || opt->dst0) &&
678 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
684 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
687 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
692 ptr[0] = __alloc_percpu(mibsize, mibalign);
696 ptr[1] = __alloc_percpu(mibsize, mibalign);
710 snmp6_mib_free(void *ptr[2])
718 ptr[0] = ptr[1] = NULL;
721 static int __init init_ipv6_mibs(void)
723 if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
724 __alignof__(struct ipstats_mib)) < 0)
726 if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
727 __alignof__(struct icmpv6_mib)) < 0)
729 if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
730 __alignof__(struct udp_mib)) < 0)
735 snmp6_mib_free((void **)icmpv6_statistics);
737 snmp6_mib_free((void **)ipv6_statistics);
743 static void cleanup_ipv6_mibs(void)
745 snmp6_mib_free((void **)ipv6_statistics);
746 snmp6_mib_free((void **)icmpv6_statistics);
747 snmp6_mib_free((void **)udp_stats_in6);
750 static int __init inet6_init(void)
752 struct sk_buff *dummy_skb;
757 #if 0 /* FIXME --RR */
758 if (!mod_member_present(&__this_module, can_unload))
761 __this_module.can_unload = &ipv6_unload;
765 if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
766 printk(KERN_CRIT "inet6_proto_init: size fault\n");
770 err = proto_register(&tcpv6_prot, 1);
774 err = proto_register(&udpv6_prot, 1);
776 goto out_unregister_tcp_proto;
778 err = proto_register(&rawv6_prot, 1);
780 goto out_unregister_udp_proto;
783 /* Register the socket-side information for inet6_create. */
784 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
787 /* We MUST register RAW sockets before we create the ICMP6,
788 * IGMP6, or NDISC control sockets.
790 inet6_register_protosw(&rawv6_protosw);
792 /* Register the family here so that the init calls below will
793 * be able to create sockets. (?? is this dangerous ??)
795 err = sock_register(&inet6_family_ops);
797 goto out_unregister_raw_proto;
799 /* Initialise ipv6 mibs */
800 err = init_ipv6_mibs();
802 goto out_unregister_sock;
805 * ipngwg API draft makes clear that the correct semantics
806 * for TCP and UDP is to consider one TCP and UDP instance
807 * in a host availiable by both INET and INET6 APIs and
808 * able to communicate via both network protocols.
812 ipv6_sysctl_register();
814 err = icmpv6_init(&inet6_family_ops);
817 err = ndisc_init(&inet6_family_ops);
820 err = igmp6_init(&inet6_family_ops);
823 err = ipv6_netfilter_init();
826 /* Create /proc/foo6 entries. */
827 #ifdef CONFIG_PROC_FS
829 if (raw6_proc_init())
831 if (tcp6_proc_init())
833 if (udp6_proc_init())
835 if (ipv6_misc_proc_init())
836 goto proc_misc6_fail;
839 goto proc_anycast6_fail;
844 ip6_flowlabel_init();
845 err = addrconf_init();
850 /* Init v6 extension headers. */
856 /* Init v6 transport protocols. */
866 ip6_flowlabel_cleanup();
868 #ifdef CONFIG_PROC_FS
873 ipv6_misc_proc_exit();
882 ipv6_netfilter_fini();
891 ipv6_sysctl_unregister();
895 sock_unregister(PF_INET6);
896 out_unregister_raw_proto:
897 proto_unregister(&rawv6_prot);
898 out_unregister_udp_proto:
899 proto_unregister(&udpv6_prot);
900 out_unregister_tcp_proto:
901 proto_unregister(&tcpv6_prot);
904 module_init(inet6_init);
906 static void __exit inet6_exit(void)
908 /* First of all disallow new sockets creation. */
909 sock_unregister(PF_INET6);
910 #ifdef CONFIG_PROC_FS
913 ipv6_misc_proc_exit();
918 /* Cleanup code parts. */
920 ip6_flowlabel_cleanup();
923 ipv6_packet_cleanup();
925 ipv6_netfilter_fini();
929 ipv6_sysctl_unregister();
932 proto_unregister(&rawv6_prot);
933 proto_unregister(&udpv6_prot);
934 proto_unregister(&tcpv6_prot);
936 module_exit(inet6_exit);
938 MODULE_ALIAS_NETPROTO(PF_INET6);