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/config.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
31 #include <linux/kernel.h>
32 #include <linux/sched.h>
33 #include <linux/timer.h>
34 #include <linux/string.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/fcntl.h>
39 #include <linux/interrupt.h>
40 #include <linux/proc_fs.h>
41 #include <linux/stat.h>
42 #include <linux/init.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <linux/icmpv6.h>
47 #include <linux/smp_lock.h>
48 #include <linux/netfilter_ipv6.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>
64 #include <asm/uaccess.h>
65 #include <asm/system.h>
67 MODULE_AUTHOR("Cast of dozens");
68 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
69 MODULE_LICENSE("GPL");
71 int sysctl_ipv6_bindv6only;
73 /* The inetsw table contains everything that inet_create needs to
76 static struct list_head inetsw6[SOCK_MAX];
77 static DEFINE_SPINLOCK(inetsw6_lock);
79 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
81 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
83 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
86 static int inet6_create(struct socket *sock, int protocol)
88 struct inet_sock *inet;
89 struct ipv6_pinfo *np;
92 struct inet_protosw *answer;
93 struct proto *answer_prot;
94 unsigned char answer_flags;
96 int try_loading_module = 0;
99 /* Look for the requested type/protocol pair. */
102 err = -ESOCKTNOSUPPORT;
104 list_for_each_rcu(p, &inetsw6[sock->type]) {
105 answer = list_entry(p, struct inet_protosw, list);
107 /* Check the non-wild match. */
108 if (protocol == answer->protocol) {
109 if (protocol != IPPROTO_IP)
112 /* Check for the two wild cases. */
113 if (IPPROTO_IP == protocol) {
114 protocol = answer->protocol;
117 if (IPPROTO_IP == answer->protocol)
120 err = -EPROTONOSUPPORT;
125 if (try_loading_module < 2) {
128 * Be more specific, e.g. net-pf-10-proto-132-type-1
129 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
131 if (++try_loading_module == 1)
132 request_module("net-pf-%d-proto-%d-type-%d",
133 PF_INET6, protocol, sock->type);
135 * Fall back to generic, e.g. net-pf-10-proto-132
136 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
139 request_module("net-pf-%d-proto-%d",
141 goto lookup_protocol;
147 if (answer->capability > 0 && !capable(answer->capability))
150 sock->ops = answer->ops;
151 answer_prot = answer->prot;
152 answer_no_check = answer->no_check;
153 answer_flags = answer->flags;
156 BUG_TRAP(answer_prot->slab != NULL);
159 sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
163 sock_init_data(sock, sk);
166 sk->sk_no_check = answer_no_check;
167 if (INET_PROTOSW_REUSE & answer_flags)
171 inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
173 if (SOCK_RAW == sock->type) {
174 inet->num = protocol;
175 if (IPPROTO_RAW == protocol)
179 sk->sk_destruct = inet_sock_destruct;
180 sk->sk_family = PF_INET6;
181 sk->sk_protocol = protocol;
183 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
185 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
189 np->pmtudisc = IPV6_PMTUDISC_WANT;
190 np->ipv6only = sysctl_ipv6_bindv6only;
192 /* Init the ipv4 part of the socket since we can have sockets
193 * using v6 API for ipv4.
200 inet->mc_list = NULL;
202 if (ipv4_config.no_pmtu_disc)
203 inet->pmtudisc = IP_PMTUDISC_DONT;
205 inet->pmtudisc = IP_PMTUDISC_WANT;
207 * Increment only the relevant sk_prot->socks debug field, this changes
208 * the previous behaviour of incrementing both the equivalent to
209 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
211 * This allows better debug granularity as we'll know exactly how many
212 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
213 * transport protocol socks. -acme
215 sk_refcnt_debug_inc(sk);
218 /* It assumes that any protocol which allows
219 * the user to assign a number at socket
220 * creation time automatically shares.
222 inet->sport = ntohs(inet->num);
223 sk->sk_prot->hash(sk);
225 if (sk->sk_prot->init) {
226 err = sk->sk_prot->init(sk);
228 sk_common_release(sk);
240 /* bind for INET6 API */
241 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
243 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
244 struct sock *sk = sock->sk;
245 struct inet_sock *inet = inet_sk(sk);
246 struct ipv6_pinfo *np = inet6_sk(sk);
252 /* If the socket has its own bind function then use it. */
253 if (sk->sk_prot->bind)
254 return sk->sk_prot->bind(sk, uaddr, addr_len);
256 if (addr_len < SIN6_LEN_RFC2133)
258 addr_type = ipv6_addr_type(&addr->sin6_addr);
259 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
262 snum = ntohs(addr->sin6_port);
263 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
268 /* Check these errors (active socket, double bind). */
269 if (sk->sk_state != TCP_CLOSE || inet->num) {
274 /* Check if the address belongs to the host. */
275 if (addr_type == IPV6_ADDR_MAPPED) {
276 v4addr = addr->sin6_addr.s6_addr32[3];
277 if (inet_addr_type(v4addr) != RTN_LOCAL) {
278 err = -EADDRNOTAVAIL;
282 if (addr_type != IPV6_ADDR_ANY) {
283 struct net_device *dev = NULL;
285 if (addr_type & IPV6_ADDR_LINKLOCAL) {
286 if (addr_len >= sizeof(struct sockaddr_in6) &&
287 addr->sin6_scope_id) {
288 /* Override any existing binding, if another one
289 * is supplied by user.
291 sk->sk_bound_dev_if = addr->sin6_scope_id;
294 /* Binding to link-local address requires an interface */
295 if (!sk->sk_bound_dev_if) {
299 dev = dev_get_by_index(sk->sk_bound_dev_if);
306 /* ipv4 addr of the socket is invalid. Only the
307 * unspecified and mapped address have a v4 equivalent.
309 v4addr = LOOPBACK4_IPV6;
310 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
311 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
314 err = -EADDRNOTAVAIL;
323 inet->rcv_saddr = v4addr;
324 inet->saddr = v4addr;
326 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
328 if (!(addr_type & IPV6_ADDR_MULTICAST))
329 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
331 /* Make sure we are allowed to bind here. */
332 if (sk->sk_prot->get_port(sk, snum)) {
333 inet_reset_saddr(sk);
338 if (addr_type != IPV6_ADDR_ANY)
339 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
341 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
342 inet->sport = ntohs(inet->num);
350 int inet6_release(struct socket *sock)
352 struct sock *sk = sock->sk;
358 ipv6_sock_mc_close(sk);
361 ipv6_sock_ac_close(sk);
363 return inet_release(sock);
366 int inet6_destroy_sock(struct sock *sk)
368 struct ipv6_pinfo *np = inet6_sk(sk);
370 struct ipv6_txoptions *opt;
373 * Release destination entry
378 /* Release rx options */
380 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
383 /* Free flowlabels */
384 fl6_free_socklist(sk);
386 /* Free tx options */
388 if ((opt = xchg(&np->opt, NULL)) != NULL)
389 sock_kfree_s(sk, opt, opt->tot_len);
394 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
397 * This does both peername and sockname.
400 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
401 int *uaddr_len, int peer)
403 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
404 struct sock *sk = sock->sk;
405 struct inet_sock *inet = inet_sk(sk);
406 struct ipv6_pinfo *np = inet6_sk(sk);
408 sin->sin6_family = AF_INET6;
409 sin->sin6_flowinfo = 0;
410 sin->sin6_scope_id = 0;
414 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
417 sin->sin6_port = inet->dport;
418 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
420 sin->sin6_flowinfo = np->flow_label;
422 if (ipv6_addr_any(&np->rcv_saddr))
423 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
425 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
427 sin->sin6_port = inet->sport;
429 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
430 sin->sin6_scope_id = sk->sk_bound_dev_if;
431 *uaddr_len = sizeof(*sin);
435 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
437 struct sock *sk = sock->sk;
442 return sock_get_timestamp(sk, (struct timeval __user *)arg);
447 return(ipv6_route_ioctl(cmd,(void __user *)arg));
450 return addrconf_add_ifaddr((void __user *) arg);
452 return addrconf_del_ifaddr((void __user *) arg);
454 return addrconf_set_dstaddr((void __user *) arg);
456 if (!sk->sk_prot->ioctl)
458 return sk->sk_prot->ioctl(sk, cmd, arg);
464 const struct proto_ops inet6_stream_ops = {
466 .owner = THIS_MODULE,
467 .release = inet6_release,
469 .connect = inet_stream_connect, /* ok */
470 .socketpair = sock_no_socketpair, /* a do nothing */
471 .accept = inet_accept, /* ok */
472 .getname = inet6_getname,
473 .poll = tcp_poll, /* ok */
474 .ioctl = inet6_ioctl, /* must change */
475 .listen = inet_listen, /* ok */
476 .shutdown = inet_shutdown, /* ok */
477 .setsockopt = sock_common_setsockopt, /* ok */
478 .getsockopt = sock_common_getsockopt, /* ok */
479 .sendmsg = inet_sendmsg, /* ok */
480 .recvmsg = sock_common_recvmsg, /* ok */
481 .mmap = sock_no_mmap,
482 .sendpage = tcp_sendpage
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,
506 static struct net_proto_family inet6_family_ops = {
508 .create = inet6_create,
509 .owner = THIS_MODULE,
512 /* Same as inet6_dgram_ops, sans udp_poll. */
513 static const struct proto_ops inet6_sockraw_ops = {
515 .owner = THIS_MODULE,
516 .release = inet6_release,
518 .connect = inet_dgram_connect, /* ok */
519 .socketpair = sock_no_socketpair, /* a do nothing */
520 .accept = sock_no_accept, /* a do nothing */
521 .getname = inet6_getname,
522 .poll = datagram_poll, /* ok */
523 .ioctl = inet6_ioctl, /* must change */
524 .listen = sock_no_listen, /* ok */
525 .shutdown = inet_shutdown, /* ok */
526 .setsockopt = sock_common_setsockopt, /* ok */
527 .getsockopt = sock_common_getsockopt, /* ok */
528 .sendmsg = inet_sendmsg, /* ok */
529 .recvmsg = sock_common_recvmsg, /* ok */
530 .mmap = sock_no_mmap,
531 .sendpage = sock_no_sendpage,
534 static struct inet_protosw rawv6_protosw = {
536 .protocol = IPPROTO_IP, /* wild card */
538 .ops = &inet6_sockraw_ops,
539 .capability = CAP_NET_RAW,
540 .no_check = UDP_CSUM_DEFAULT,
541 .flags = INET_PROTOSW_REUSE,
545 inet6_register_protosw(struct inet_protosw *p)
547 struct list_head *lh;
548 struct inet_protosw *answer;
549 int protocol = p->protocol;
550 struct list_head *last_perm;
552 spin_lock_bh(&inetsw6_lock);
554 if (p->type >= SOCK_MAX)
557 /* If we are trying to override a permanent protocol, bail. */
559 last_perm = &inetsw6[p->type];
560 list_for_each(lh, &inetsw6[p->type]) {
561 answer = list_entry(lh, struct inet_protosw, list);
563 /* Check only the non-wild match. */
564 if (INET_PROTOSW_PERMANENT & answer->flags) {
565 if (protocol == answer->protocol)
575 /* Add the new entry after the last permanent entry if any, so that
576 * the new entry does not override a permanent entry when matched with
577 * a wild-card protocol. But it is allowed to override any existing
578 * non-permanent entry. This means that when we remove this entry, the
579 * system automatically returns to the old behavior.
581 list_add_rcu(&p->list, last_perm);
583 spin_unlock_bh(&inetsw6_lock);
587 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
593 "Ignoring attempt to register invalid socket type %d.\n",
599 inet6_unregister_protosw(struct inet_protosw *p)
601 if (INET_PROTOSW_PERMANENT & p->flags) {
603 "Attempt to unregister permanent protocol %d.\n",
606 spin_lock_bh(&inetsw6_lock);
607 list_del_rcu(&p->list);
608 spin_unlock_bh(&inetsw6_lock);
614 int inet6_sk_rebuild_header(struct sock *sk)
617 struct dst_entry *dst;
618 struct ipv6_pinfo *np = inet6_sk(sk);
620 dst = __sk_dst_check(sk, np->dst_cookie);
623 struct inet_sock *inet = inet_sk(sk);
624 struct in6_addr *final_p = NULL, final;
627 memset(&fl, 0, sizeof(fl));
628 fl.proto = sk->sk_protocol;
629 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
630 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
631 fl.fl6_flowlabel = np->flow_label;
632 fl.oif = sk->sk_bound_dev_if;
633 fl.fl_ip_dport = inet->dport;
634 fl.fl_ip_sport = inet->sport;
636 if (np->opt && np->opt->srcrt) {
637 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
638 ipv6_addr_copy(&final, &fl.fl6_dst);
639 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
643 err = ip6_dst_lookup(sk, &dst, &fl);
645 sk->sk_route_caps = 0;
649 ipv6_addr_copy(&fl.fl6_dst, final_p);
651 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
652 sk->sk_err_soft = -err;
656 ip6_dst_store(sk, dst, NULL);
657 sk->sk_route_caps = dst->dev->features &
658 ~(NETIF_F_IP_CSUM | NETIF_F_TSO);
664 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
666 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
668 struct ipv6_pinfo *np = inet6_sk(sk);
669 struct inet6_skb_parm *opt = IP6CB(skb);
672 if ((opt->hop && (np->rxopt.bits.hopopts ||
673 np->rxopt.bits.ohopopts)) ||
674 ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
675 np->rxopt.bits.rxflow) ||
676 (opt->srcrt && (np->rxopt.bits.srcrt ||
677 np->rxopt.bits.osrcrt)) ||
678 ((opt->dst1 || opt->dst0) &&
679 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
685 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
688 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
693 ptr[0] = __alloc_percpu(mibsize);
697 ptr[1] = __alloc_percpu(mibsize);
711 snmp6_mib_free(void *ptr[2])
719 ptr[0] = ptr[1] = NULL;
722 static int __init init_ipv6_mibs(void)
724 if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
725 __alignof__(struct ipstats_mib)) < 0)
727 if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
728 __alignof__(struct icmpv6_mib)) < 0)
730 if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
731 __alignof__(struct udp_mib)) < 0)
736 snmp6_mib_free((void **)icmpv6_statistics);
738 snmp6_mib_free((void **)ipv6_statistics);
744 static void cleanup_ipv6_mibs(void)
746 snmp6_mib_free((void **)ipv6_statistics);
747 snmp6_mib_free((void **)icmpv6_statistics);
748 snmp6_mib_free((void **)udp_stats_in6);
751 static int __init inet6_init(void)
753 struct sk_buff *dummy_skb;
758 #if 0 /* FIXME --RR */
759 if (!mod_member_present(&__this_module, can_unload))
762 __this_module.can_unload = &ipv6_unload;
766 if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
767 printk(KERN_CRIT "inet6_proto_init: size fault\n");
771 err = proto_register(&tcpv6_prot, 1);
775 err = proto_register(&udpv6_prot, 1);
777 goto out_unregister_tcp_proto;
779 err = proto_register(&rawv6_prot, 1);
781 goto out_unregister_udp_proto;
784 /* Register the socket-side information for inet6_create. */
785 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
788 /* We MUST register RAW sockets before we create the ICMP6,
789 * IGMP6, or NDISC control sockets.
791 inet6_register_protosw(&rawv6_protosw);
793 /* Register the family here so that the init calls below will
794 * be able to create sockets. (?? is this dangerous ??)
796 err = sock_register(&inet6_family_ops);
798 goto out_unregister_raw_proto;
800 /* Initialise ipv6 mibs */
801 err = init_ipv6_mibs();
803 goto out_unregister_sock;
806 * ipngwg API draft makes clear that the correct semantics
807 * for TCP and UDP is to consider one TCP and UDP instance
808 * in a host availiable by both INET and INET6 APIs and
809 * able to communicate via both network protocols.
813 ipv6_sysctl_register();
815 err = icmpv6_init(&inet6_family_ops);
818 err = ndisc_init(&inet6_family_ops);
821 err = igmp6_init(&inet6_family_ops);
824 err = ipv6_netfilter_init();
827 /* Create /proc/foo6 entries. */
828 #ifdef CONFIG_PROC_FS
830 if (raw6_proc_init())
832 if (tcp6_proc_init())
834 if (udp6_proc_init())
836 if (ipv6_misc_proc_init())
837 goto proc_misc6_fail;
840 goto proc_anycast6_fail;
845 ip6_flowlabel_init();
846 err = addrconf_init();
851 /* Init v6 extension headers. */
857 /* Init v6 transport protocols. */
867 ip6_flowlabel_cleanup();
869 #ifdef CONFIG_PROC_FS
874 ipv6_misc_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_udp_proto:
900 proto_unregister(&udpv6_prot);
901 out_unregister_tcp_proto:
902 proto_unregister(&tcpv6_prot);
905 module_init(inet6_init);
907 static void __exit inet6_exit(void)
909 /* First of all disallow new sockets creation. */
910 sock_unregister(PF_INET6);
911 #ifdef CONFIG_PROC_FS
914 ipv6_misc_proc_exit();
919 /* Cleanup code parts. */
921 ip6_flowlabel_cleanup();
924 ipv6_packet_cleanup();
926 ipv6_netfilter_fini();
930 ipv6_sysctl_unregister();
933 proto_unregister(&rawv6_prot);
934 proto_unregister(&udpv6_prot);
935 proto_unregister(&tcpv6_prot);
937 module_exit(inet6_exit);
939 MODULE_ALIAS_NETPROTO(PF_INET6);