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 /* IPv6 procfs goodies... */
73 extern int raw6_proc_init(void);
74 extern void raw6_proc_exit(void);
75 extern int tcp6_proc_init(void);
76 extern void tcp6_proc_exit(void);
77 extern int udp6_proc_init(void);
78 extern void udp6_proc_exit(void);
79 extern int ipv6_misc_proc_init(void);
80 extern void ipv6_misc_proc_exit(void);
81 extern int ac6_proc_init(void);
82 extern void ac6_proc_exit(void);
83 extern int if6_proc_init(void);
84 extern void if6_proc_exit(void);
87 int sysctl_ipv6_bindv6only;
89 /* The inetsw table contains everything that inet_create needs to
92 static struct list_head inetsw6[SOCK_MAX];
93 static DEFINE_SPINLOCK(inetsw6_lock);
95 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
97 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
99 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
102 static int inet6_create(struct socket *sock, int protocol)
104 struct inet_sock *inet;
105 struct ipv6_pinfo *np;
108 struct inet_protosw *answer;
109 struct proto *answer_prot;
110 unsigned char answer_flags;
111 char answer_no_check;
114 /* Look for the requested type/protocol pair. */
117 list_for_each_rcu(p, &inetsw6[sock->type]) {
118 answer = list_entry(p, struct inet_protosw, list);
120 /* Check the non-wild match. */
121 if (protocol == answer->protocol) {
122 if (protocol != IPPROTO_IP)
125 /* Check for the two wild cases. */
126 if (IPPROTO_IP == protocol) {
127 protocol = answer->protocol;
130 if (IPPROTO_IP == answer->protocol)
136 rc = -ESOCKTNOSUPPORT;
140 if (answer->capability > 0 && !capable(answer->capability))
142 rc = -EPROTONOSUPPORT;
146 sock->ops = answer->ops;
148 answer_prot = answer->prot;
149 answer_no_check = answer->no_check;
150 answer_flags = answer->flags;
153 BUG_TRAP(answer_prot->slab != NULL);
156 sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
160 sock_init_data(sock, sk);
163 sk->sk_no_check = answer_no_check;
164 if (INET_PROTOSW_REUSE & answer_flags)
169 if (SOCK_RAW == sock->type) {
170 inet->num = protocol;
171 if (IPPROTO_RAW == protocol)
175 sk->sk_destruct = inet_sock_destruct;
176 sk->sk_family = PF_INET6;
177 sk->sk_protocol = protocol;
179 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
181 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
185 np->pmtudisc = IPV6_PMTUDISC_WANT;
186 np->ipv6only = sysctl_ipv6_bindv6only;
188 /* Init the ipv4 part of the socket since we can have sockets
189 * using v6 API for ipv4.
196 inet->mc_list = NULL;
198 if (ipv4_config.no_pmtu_disc)
199 inet->pmtudisc = IP_PMTUDISC_DONT;
201 inet->pmtudisc = IP_PMTUDISC_WANT;
203 * Increment only the relevant sk_prot->socks debug field, this changes
204 * the previous behaviour of incrementing both the equivalent to
205 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
207 * This allows better debug granularity as we'll know exactly how many
208 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
209 * transport protocol socks. -acme
211 sk_refcnt_debug_inc(sk);
214 /* It assumes that any protocol which allows
215 * the user to assign a number at socket
216 * creation time automatically shares.
218 inet->sport = ntohs(inet->num);
219 sk->sk_prot->hash(sk);
221 if (sk->sk_prot->init) {
222 rc = sk->sk_prot->init(sk);
224 sk_common_release(sk);
236 /* bind for INET6 API */
237 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
239 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
240 struct sock *sk = sock->sk;
241 struct inet_sock *inet = inet_sk(sk);
242 struct ipv6_pinfo *np = inet6_sk(sk);
248 /* If the socket has its own bind function then use it. */
249 if (sk->sk_prot->bind)
250 return sk->sk_prot->bind(sk, uaddr, addr_len);
252 if (addr_len < SIN6_LEN_RFC2133)
254 addr_type = ipv6_addr_type(&addr->sin6_addr);
255 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
258 snum = ntohs(addr->sin6_port);
259 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
264 /* Check these errors (active socket, double bind). */
265 if (sk->sk_state != TCP_CLOSE || inet->num) {
270 /* Check if the address belongs to the host. */
271 if (addr_type == IPV6_ADDR_MAPPED) {
272 v4addr = addr->sin6_addr.s6_addr32[3];
273 if (inet_addr_type(v4addr) != RTN_LOCAL) {
274 err = -EADDRNOTAVAIL;
278 if (addr_type != IPV6_ADDR_ANY) {
279 struct net_device *dev = NULL;
281 if (addr_type & IPV6_ADDR_LINKLOCAL) {
282 if (addr_len >= sizeof(struct sockaddr_in6) &&
283 addr->sin6_scope_id) {
284 /* Override any existing binding, if another one
285 * is supplied by user.
287 sk->sk_bound_dev_if = addr->sin6_scope_id;
290 /* Binding to link-local address requires an interface */
291 if (!sk->sk_bound_dev_if) {
295 dev = dev_get_by_index(sk->sk_bound_dev_if);
302 /* ipv4 addr of the socket is invalid. Only the
303 * unspecified and mapped address have a v4 equivalent.
305 v4addr = LOOPBACK4_IPV6;
306 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
307 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
310 err = -EADDRNOTAVAIL;
319 inet->rcv_saddr = v4addr;
320 inet->saddr = v4addr;
322 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
324 if (!(addr_type & IPV6_ADDR_MULTICAST))
325 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
327 /* Make sure we are allowed to bind here. */
328 if (sk->sk_prot->get_port(sk, snum)) {
329 inet_reset_saddr(sk);
334 if (addr_type != IPV6_ADDR_ANY)
335 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
337 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
338 inet->sport = ntohs(inet->num);
346 int inet6_release(struct socket *sock)
348 struct sock *sk = sock->sk;
354 ipv6_sock_mc_close(sk);
357 ipv6_sock_ac_close(sk);
359 return inet_release(sock);
362 int inet6_destroy_sock(struct sock *sk)
364 struct ipv6_pinfo *np = inet6_sk(sk);
366 struct ipv6_txoptions *opt;
369 * Release destination entry
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);
391 * This does both peername and sockname.
394 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
395 int *uaddr_len, int peer)
397 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
398 struct sock *sk = sock->sk;
399 struct inet_sock *inet = inet_sk(sk);
400 struct ipv6_pinfo *np = inet6_sk(sk);
402 sin->sin6_family = AF_INET6;
403 sin->sin6_flowinfo = 0;
404 sin->sin6_scope_id = 0;
408 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
411 sin->sin6_port = inet->dport;
412 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
414 sin->sin6_flowinfo = np->flow_label;
416 if (ipv6_addr_any(&np->rcv_saddr))
417 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
419 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
421 sin->sin6_port = inet->sport;
423 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
424 sin->sin6_scope_id = sk->sk_bound_dev_if;
425 *uaddr_len = sizeof(*sin);
429 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
431 struct sock *sk = sock->sk;
437 return sock_get_timestamp(sk, (struct timeval __user *)arg);
442 return(ipv6_route_ioctl(cmd,(void __user *)arg));
445 return addrconf_add_ifaddr((void __user *) arg);
447 return addrconf_del_ifaddr((void __user *) arg);
449 return addrconf_set_dstaddr((void __user *) arg);
451 if (!sk->sk_prot->ioctl ||
452 (err = sk->sk_prot->ioctl(sk, cmd, arg)) == -ENOIOCTLCMD)
453 return(dev_ioctl(cmd,(void __user *) arg));
460 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
481 struct proto_ops inet6_dgram_ops = {
483 .owner = THIS_MODULE,
484 .release = inet6_release,
486 .connect = inet_dgram_connect, /* ok */
487 .socketpair = sock_no_socketpair, /* a do nothing */
488 .accept = sock_no_accept, /* a do nothing */
489 .getname = inet6_getname,
490 .poll = udp_poll, /* ok */
491 .ioctl = inet6_ioctl, /* must change */
492 .listen = sock_no_listen, /* ok */
493 .shutdown = inet_shutdown, /* ok */
494 .setsockopt = sock_common_setsockopt, /* ok */
495 .getsockopt = sock_common_getsockopt, /* ok */
496 .sendmsg = inet_sendmsg, /* ok */
497 .recvmsg = sock_common_recvmsg, /* ok */
498 .mmap = sock_no_mmap,
499 .sendpage = sock_no_sendpage,
502 static struct net_proto_family inet6_family_ops = {
504 .create = inet6_create,
505 .owner = THIS_MODULE,
509 extern void ipv6_sysctl_register(void);
510 extern void ipv6_sysctl_unregister(void);
513 /* Same as inet6_dgram_ops, sans udp_poll. */
514 static struct proto_ops inet6_sockraw_ops = {
516 .owner = THIS_MODULE,
517 .release = inet6_release,
519 .connect = inet_dgram_connect, /* ok */
520 .socketpair = sock_no_socketpair, /* a do nothing */
521 .accept = sock_no_accept, /* a do nothing */
522 .getname = inet6_getname,
523 .poll = datagram_poll, /* ok */
524 .ioctl = inet6_ioctl, /* must change */
525 .listen = sock_no_listen, /* ok */
526 .shutdown = inet_shutdown, /* ok */
527 .setsockopt = sock_common_setsockopt, /* ok */
528 .getsockopt = sock_common_getsockopt, /* ok */
529 .sendmsg = inet_sendmsg, /* ok */
530 .recvmsg = sock_common_recvmsg, /* ok */
531 .mmap = sock_no_mmap,
532 .sendpage = sock_no_sendpage,
535 static struct inet_protosw rawv6_protosw = {
537 .protocol = IPPROTO_IP, /* wild card */
539 .ops = &inet6_sockraw_ops,
540 .capability = CAP_NET_RAW,
541 .no_check = UDP_CSUM_DEFAULT,
542 .flags = INET_PROTOSW_REUSE,
546 inet6_register_protosw(struct inet_protosw *p)
548 struct list_head *lh;
549 struct inet_protosw *answer;
550 int protocol = p->protocol;
551 struct list_head *last_perm;
553 spin_lock_bh(&inetsw6_lock);
555 if (p->type >= SOCK_MAX)
558 /* If we are trying to override a permanent protocol, bail. */
560 last_perm = &inetsw6[p->type];
561 list_for_each(lh, &inetsw6[p->type]) {
562 answer = list_entry(lh, struct inet_protosw, list);
564 /* Check only the non-wild match. */
565 if (INET_PROTOSW_PERMANENT & answer->flags) {
566 if (protocol == answer->protocol)
576 /* Add the new entry after the last permanent entry if any, so that
577 * the new entry does not override a permanent entry when matched with
578 * a wild-card protocol. But it is allowed to override any existing
579 * non-permanent entry. This means that when we remove this entry, the
580 * system automatically returns to the old behavior.
582 list_add_rcu(&p->list, last_perm);
584 spin_unlock_bh(&inetsw6_lock);
588 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
594 "Ignoring attempt to register invalid socket type %d.\n",
600 inet6_unregister_protosw(struct inet_protosw *p)
602 if (INET_PROTOSW_PERMANENT & p->flags) {
604 "Attempt to unregister permanent protocol %d.\n",
607 spin_lock_bh(&inetsw6_lock);
608 list_del_rcu(&p->list);
609 spin_unlock_bh(&inetsw6_lock);
616 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
621 ptr[0] = __alloc_percpu(mibsize, mibalign);
625 ptr[1] = __alloc_percpu(mibsize, mibalign);
639 snmp6_mib_free(void *ptr[2])
647 ptr[0] = ptr[1] = NULL;
650 static int __init init_ipv6_mibs(void)
652 if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
653 __alignof__(struct ipstats_mib)) < 0)
655 if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
656 __alignof__(struct icmpv6_mib)) < 0)
658 if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
659 __alignof__(struct udp_mib)) < 0)
664 snmp6_mib_free((void **)icmpv6_statistics);
666 snmp6_mib_free((void **)ipv6_statistics);
672 static void cleanup_ipv6_mibs(void)
674 snmp6_mib_free((void **)ipv6_statistics);
675 snmp6_mib_free((void **)icmpv6_statistics);
676 snmp6_mib_free((void **)udp_stats_in6);
679 extern int ipv6_misc_proc_init(void);
681 static int __init inet6_init(void)
683 struct sk_buff *dummy_skb;
688 #if 0 /* FIXME --RR */
689 if (!mod_member_present(&__this_module, can_unload))
692 __this_module.can_unload = &ipv6_unload;
696 if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
697 printk(KERN_CRIT "inet6_proto_init: size fault\n");
701 err = proto_register(&tcpv6_prot, 1);
705 err = proto_register(&udpv6_prot, 1);
707 goto out_unregister_tcp_proto;
709 err = proto_register(&rawv6_prot, 1);
711 goto out_unregister_udp_proto;
714 /* Register the socket-side information for inet6_create. */
715 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
718 /* We MUST register RAW sockets before we create the ICMP6,
719 * IGMP6, or NDISC control sockets.
721 inet6_register_protosw(&rawv6_protosw);
723 /* Register the family here so that the init calls below will
724 * be able to create sockets. (?? is this dangerous ??)
726 (void) sock_register(&inet6_family_ops);
728 /* Initialise ipv6 mibs */
729 err = init_ipv6_mibs();
731 goto out_unregister_raw_proto;
734 * ipngwg API draft makes clear that the correct semantics
735 * for TCP and UDP is to consider one TCP and UDP instance
736 * in a host availiable by both INET and INET6 APIs and
737 * able to communicate via both network protocols.
741 ipv6_sysctl_register();
743 err = icmpv6_init(&inet6_family_ops);
746 err = ndisc_init(&inet6_family_ops);
749 err = igmp6_init(&inet6_family_ops);
752 err = ipv6_netfilter_init();
755 /* Create /proc/foo6 entries. */
756 #ifdef CONFIG_PROC_FS
758 if (raw6_proc_init())
760 if (tcp6_proc_init())
762 if (udp6_proc_init())
764 if (ipv6_misc_proc_init())
765 goto proc_misc6_fail;
768 goto proc_anycast6_fail;
773 ip6_flowlabel_init();
774 err = addrconf_init();
779 /* Init v6 extension headers. */
785 /* Init v6 transport protocols. */
795 ip6_flowlabel_cleanup();
797 #ifdef CONFIG_PROC_FS
802 ipv6_misc_proc_exit();
811 ipv6_netfilter_fini();
820 ipv6_sysctl_unregister();
823 out_unregister_raw_proto:
824 proto_unregister(&rawv6_prot);
825 out_unregister_udp_proto:
826 proto_unregister(&udpv6_prot);
827 out_unregister_tcp_proto:
828 proto_unregister(&tcpv6_prot);
831 module_init(inet6_init);
833 static void __exit inet6_exit(void)
835 /* First of all disallow new sockets creation. */
836 sock_unregister(PF_INET6);
837 #ifdef CONFIG_PROC_FS
840 ipv6_misc_proc_exit();
845 /* Cleanup code parts. */
847 ip6_flowlabel_cleanup();
850 ipv6_packet_cleanup();
852 ipv6_netfilter_fini();
856 ipv6_sysctl_unregister();
859 proto_unregister(&rawv6_prot);
860 proto_unregister(&udpv6_prot);
861 proto_unregister(&tcpv6_prot);
863 module_exit(inet6_exit);
865 MODULE_ALIAS_NETPROTO(PF_INET6);