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
;
97 /* Look for the requested type/protocol pair. */
100 list_for_each_rcu(p
, &inetsw6
[sock
->type
]) {
101 answer
= list_entry(p
, struct inet_protosw
, list
);
103 /* Check the non-wild match. */
104 if (protocol
== answer
->protocol
) {
105 if (protocol
!= IPPROTO_IP
)
108 /* Check for the two wild cases. */
109 if (IPPROTO_IP
== protocol
) {
110 protocol
= answer
->protocol
;
113 if (IPPROTO_IP
== answer
->protocol
)
119 rc
= -ESOCKTNOSUPPORT
;
123 if (answer
->capability
> 0 && !capable(answer
->capability
))
125 rc
= -EPROTONOSUPPORT
;
129 sock
->ops
= answer
->ops
;
131 answer_prot
= answer
->prot
;
132 answer_no_check
= answer
->no_check
;
133 answer_flags
= answer
->flags
;
136 BUG_TRAP(answer_prot
->slab
!= NULL
);
139 sk
= sk_alloc(PF_INET6
, GFP_KERNEL
, answer_prot
, 1);
143 sock_init_data(sock
, sk
);
146 sk
->sk_no_check
= answer_no_check
;
147 if (INET_PROTOSW_REUSE
& answer_flags
)
152 if (SOCK_RAW
== sock
->type
) {
153 inet
->num
= protocol
;
154 if (IPPROTO_RAW
== protocol
)
158 sk
->sk_destruct
= inet_sock_destruct
;
159 sk
->sk_family
= PF_INET6
;
160 sk
->sk_protocol
= protocol
;
162 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
164 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
168 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
169 np
->ipv6only
= sysctl_ipv6_bindv6only
;
171 /* Init the ipv4 part of the socket since we can have sockets
172 * using v6 API for ipv4.
179 inet
->mc_list
= NULL
;
181 if (ipv4_config
.no_pmtu_disc
)
182 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
184 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
186 * Increment only the relevant sk_prot->socks debug field, this changes
187 * the previous behaviour of incrementing both the equivalent to
188 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
190 * This allows better debug granularity as we'll know exactly how many
191 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
192 * transport protocol socks. -acme
194 sk_refcnt_debug_inc(sk
);
197 /* It assumes that any protocol which allows
198 * the user to assign a number at socket
199 * creation time automatically shares.
201 inet
->sport
= ntohs(inet
->num
);
202 sk
->sk_prot
->hash(sk
);
204 if (sk
->sk_prot
->init
) {
205 rc
= sk
->sk_prot
->init(sk
);
207 sk_common_release(sk
);
219 /* bind for INET6 API */
220 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
222 struct sockaddr_in6
*addr
=(struct sockaddr_in6
*)uaddr
;
223 struct sock
*sk
= sock
->sk
;
224 struct inet_sock
*inet
= inet_sk(sk
);
225 struct ipv6_pinfo
*np
= inet6_sk(sk
);
231 /* If the socket has its own bind function then use it. */
232 if (sk
->sk_prot
->bind
)
233 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
235 if (addr_len
< SIN6_LEN_RFC2133
)
237 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
238 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sock
->type
== SOCK_STREAM
)
241 snum
= ntohs(addr
->sin6_port
);
242 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
247 /* Check these errors (active socket, double bind). */
248 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->num
) {
253 /* Check if the address belongs to the host. */
254 if (addr_type
== IPV6_ADDR_MAPPED
) {
255 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
256 if (inet_addr_type(v4addr
) != RTN_LOCAL
) {
257 err
= -EADDRNOTAVAIL
;
261 if (addr_type
!= IPV6_ADDR_ANY
) {
262 struct net_device
*dev
= NULL
;
264 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
265 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
266 addr
->sin6_scope_id
) {
267 /* Override any existing binding, if another one
268 * is supplied by user.
270 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
273 /* Binding to link-local address requires an interface */
274 if (!sk
->sk_bound_dev_if
) {
278 dev
= dev_get_by_index(sk
->sk_bound_dev_if
);
285 /* ipv4 addr of the socket is invalid. Only the
286 * unspecified and mapped address have a v4 equivalent.
288 v4addr
= LOOPBACK4_IPV6
;
289 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
290 if (!ipv6_chk_addr(&addr
->sin6_addr
, dev
, 0)) {
293 err
= -EADDRNOTAVAIL
;
302 inet
->rcv_saddr
= v4addr
;
303 inet
->saddr
= v4addr
;
305 ipv6_addr_copy(&np
->rcv_saddr
, &addr
->sin6_addr
);
307 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
308 ipv6_addr_copy(&np
->saddr
, &addr
->sin6_addr
);
310 /* Make sure we are allowed to bind here. */
311 if (sk
->sk_prot
->get_port(sk
, snum
)) {
312 inet_reset_saddr(sk
);
317 if (addr_type
!= IPV6_ADDR_ANY
)
318 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
320 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
321 inet
->sport
= ntohs(inet
->num
);
329 int inet6_release(struct socket
*sock
)
331 struct sock
*sk
= sock
->sk
;
337 ipv6_sock_mc_close(sk
);
340 ipv6_sock_ac_close(sk
);
342 return inet_release(sock
);
345 int inet6_destroy_sock(struct sock
*sk
)
347 struct ipv6_pinfo
*np
= inet6_sk(sk
);
349 struct ipv6_txoptions
*opt
;
352 * Release destination entry
357 /* Release rx options */
359 if ((skb
= xchg(&np
->pktoptions
, NULL
)) != NULL
)
362 /* Free flowlabels */
363 fl6_free_socklist(sk
);
365 /* Free tx options */
367 if ((opt
= xchg(&np
->opt
, NULL
)) != NULL
)
368 sock_kfree_s(sk
, opt
, opt
->tot_len
);
374 * This does both peername and sockname.
377 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
378 int *uaddr_len
, int peer
)
380 struct sockaddr_in6
*sin
=(struct sockaddr_in6
*)uaddr
;
381 struct sock
*sk
= sock
->sk
;
382 struct inet_sock
*inet
= inet_sk(sk
);
383 struct ipv6_pinfo
*np
= inet6_sk(sk
);
385 sin
->sin6_family
= AF_INET6
;
386 sin
->sin6_flowinfo
= 0;
387 sin
->sin6_scope_id
= 0;
391 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
394 sin
->sin6_port
= inet
->dport
;
395 ipv6_addr_copy(&sin
->sin6_addr
, &np
->daddr
);
397 sin
->sin6_flowinfo
= np
->flow_label
;
399 if (ipv6_addr_any(&np
->rcv_saddr
))
400 ipv6_addr_copy(&sin
->sin6_addr
, &np
->saddr
);
402 ipv6_addr_copy(&sin
->sin6_addr
, &np
->rcv_saddr
);
404 sin
->sin6_port
= inet
->sport
;
406 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
407 sin
->sin6_scope_id
= sk
->sk_bound_dev_if
;
408 *uaddr_len
= sizeof(*sin
);
412 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
414 struct sock
*sk
= sock
->sk
;
420 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
425 return(ipv6_route_ioctl(cmd
,(void __user
*)arg
));
428 return addrconf_add_ifaddr((void __user
*) arg
);
430 return addrconf_del_ifaddr((void __user
*) arg
);
432 return addrconf_set_dstaddr((void __user
*) arg
);
434 if (!sk
->sk_prot
->ioctl
||
435 (err
= sk
->sk_prot
->ioctl(sk
, cmd
, arg
)) == -ENOIOCTLCMD
)
436 return(dev_ioctl(cmd
,(void __user
*) arg
));
443 struct proto_ops inet6_stream_ops
= {
445 .owner
= THIS_MODULE
,
446 .release
= inet6_release
,
448 .connect
= inet_stream_connect
, /* ok */
449 .socketpair
= sock_no_socketpair
, /* a do nothing */
450 .accept
= inet_accept
, /* ok */
451 .getname
= inet6_getname
,
452 .poll
= tcp_poll
, /* ok */
453 .ioctl
= inet6_ioctl
, /* must change */
454 .listen
= inet_listen
, /* ok */
455 .shutdown
= inet_shutdown
, /* ok */
456 .setsockopt
= sock_common_setsockopt
, /* ok */
457 .getsockopt
= sock_common_getsockopt
, /* ok */
458 .sendmsg
= inet_sendmsg
, /* ok */
459 .recvmsg
= sock_common_recvmsg
, /* ok */
460 .mmap
= sock_no_mmap
,
461 .sendpage
= tcp_sendpage
464 struct proto_ops inet6_dgram_ops
= {
466 .owner
= THIS_MODULE
,
467 .release
= inet6_release
,
469 .connect
= inet_dgram_connect
, /* ok */
470 .socketpair
= sock_no_socketpair
, /* a do nothing */
471 .accept
= sock_no_accept
, /* a do nothing */
472 .getname
= inet6_getname
,
473 .poll
= udp_poll
, /* ok */
474 .ioctl
= inet6_ioctl
, /* must change */
475 .listen
= sock_no_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
= sock_no_sendpage
,
485 static struct net_proto_family inet6_family_ops
= {
487 .create
= inet6_create
,
488 .owner
= THIS_MODULE
,
491 /* Same as inet6_dgram_ops, sans udp_poll. */
492 static struct proto_ops inet6_sockraw_ops
= {
494 .owner
= THIS_MODULE
,
495 .release
= inet6_release
,
497 .connect
= inet_dgram_connect
, /* ok */
498 .socketpair
= sock_no_socketpair
, /* a do nothing */
499 .accept
= sock_no_accept
, /* a do nothing */
500 .getname
= inet6_getname
,
501 .poll
= datagram_poll
, /* ok */
502 .ioctl
= inet6_ioctl
, /* must change */
503 .listen
= sock_no_listen
, /* ok */
504 .shutdown
= inet_shutdown
, /* ok */
505 .setsockopt
= sock_common_setsockopt
, /* ok */
506 .getsockopt
= sock_common_getsockopt
, /* ok */
507 .sendmsg
= inet_sendmsg
, /* ok */
508 .recvmsg
= sock_common_recvmsg
, /* ok */
509 .mmap
= sock_no_mmap
,
510 .sendpage
= sock_no_sendpage
,
513 static struct inet_protosw rawv6_protosw
= {
515 .protocol
= IPPROTO_IP
, /* wild card */
517 .ops
= &inet6_sockraw_ops
,
518 .capability
= CAP_NET_RAW
,
519 .no_check
= UDP_CSUM_DEFAULT
,
520 .flags
= INET_PROTOSW_REUSE
,
524 inet6_register_protosw(struct inet_protosw
*p
)
526 struct list_head
*lh
;
527 struct inet_protosw
*answer
;
528 int protocol
= p
->protocol
;
529 struct list_head
*last_perm
;
531 spin_lock_bh(&inetsw6_lock
);
533 if (p
->type
>= SOCK_MAX
)
536 /* If we are trying to override a permanent protocol, bail. */
538 last_perm
= &inetsw6
[p
->type
];
539 list_for_each(lh
, &inetsw6
[p
->type
]) {
540 answer
= list_entry(lh
, struct inet_protosw
, list
);
542 /* Check only the non-wild match. */
543 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
544 if (protocol
== answer
->protocol
)
554 /* Add the new entry after the last permanent entry if any, so that
555 * the new entry does not override a permanent entry when matched with
556 * a wild-card protocol. But it is allowed to override any existing
557 * non-permanent entry. This means that when we remove this entry, the
558 * system automatically returns to the old behavior.
560 list_add_rcu(&p
->list
, last_perm
);
562 spin_unlock_bh(&inetsw6_lock
);
566 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
572 "Ignoring attempt to register invalid socket type %d.\n",
578 inet6_unregister_protosw(struct inet_protosw
*p
)
580 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
582 "Attempt to unregister permanent protocol %d.\n",
585 spin_lock_bh(&inetsw6_lock
);
586 list_del_rcu(&p
->list
);
587 spin_unlock_bh(&inetsw6_lock
);
594 snmp6_mib_init(void *ptr
[2], size_t mibsize
, size_t mibalign
)
599 ptr
[0] = __alloc_percpu(mibsize
, mibalign
);
603 ptr
[1] = __alloc_percpu(mibsize
, mibalign
);
617 snmp6_mib_free(void *ptr
[2])
625 ptr
[0] = ptr
[1] = NULL
;
628 static int __init
init_ipv6_mibs(void)
630 if (snmp6_mib_init((void **)ipv6_statistics
, sizeof (struct ipstats_mib
),
631 __alignof__(struct ipstats_mib
)) < 0)
633 if (snmp6_mib_init((void **)icmpv6_statistics
, sizeof (struct icmpv6_mib
),
634 __alignof__(struct icmpv6_mib
)) < 0)
636 if (snmp6_mib_init((void **)udp_stats_in6
, sizeof (struct udp_mib
),
637 __alignof__(struct udp_mib
)) < 0)
642 snmp6_mib_free((void **)icmpv6_statistics
);
644 snmp6_mib_free((void **)ipv6_statistics
);
650 static void cleanup_ipv6_mibs(void)
652 snmp6_mib_free((void **)ipv6_statistics
);
653 snmp6_mib_free((void **)icmpv6_statistics
);
654 snmp6_mib_free((void **)udp_stats_in6
);
657 static int __init
inet6_init(void)
659 struct sk_buff
*dummy_skb
;
664 #if 0 /* FIXME --RR */
665 if (!mod_member_present(&__this_module
, can_unload
))
668 __this_module
.can_unload
= &ipv6_unload
;
672 if (sizeof(struct inet6_skb_parm
) > sizeof(dummy_skb
->cb
)) {
673 printk(KERN_CRIT
"inet6_proto_init: size fault\n");
677 err
= proto_register(&tcpv6_prot
, 1);
681 err
= proto_register(&udpv6_prot
, 1);
683 goto out_unregister_tcp_proto
;
685 err
= proto_register(&rawv6_prot
, 1);
687 goto out_unregister_udp_proto
;
690 /* Register the socket-side information for inet6_create. */
691 for(r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
694 /* We MUST register RAW sockets before we create the ICMP6,
695 * IGMP6, or NDISC control sockets.
697 inet6_register_protosw(&rawv6_protosw
);
699 /* Register the family here so that the init calls below will
700 * be able to create sockets. (?? is this dangerous ??)
702 (void) sock_register(&inet6_family_ops
);
704 /* Initialise ipv6 mibs */
705 err
= init_ipv6_mibs();
707 goto out_unregister_raw_proto
;
710 * ipngwg API draft makes clear that the correct semantics
711 * for TCP and UDP is to consider one TCP and UDP instance
712 * in a host availiable by both INET and INET6 APIs and
713 * able to communicate via both network protocols.
717 ipv6_sysctl_register();
719 err
= icmpv6_init(&inet6_family_ops
);
722 err
= ndisc_init(&inet6_family_ops
);
725 err
= igmp6_init(&inet6_family_ops
);
728 err
= ipv6_netfilter_init();
731 /* Create /proc/foo6 entries. */
732 #ifdef CONFIG_PROC_FS
734 if (raw6_proc_init())
736 if (tcp6_proc_init())
738 if (udp6_proc_init())
740 if (ipv6_misc_proc_init())
741 goto proc_misc6_fail
;
744 goto proc_anycast6_fail
;
749 ip6_flowlabel_init();
750 err
= addrconf_init();
755 /* Init v6 extension headers. */
761 /* Init v6 transport protocols. */
771 ip6_flowlabel_cleanup();
773 #ifdef CONFIG_PROC_FS
778 ipv6_misc_proc_exit();
787 ipv6_netfilter_fini();
796 ipv6_sysctl_unregister();
799 out_unregister_raw_proto
:
800 proto_unregister(&rawv6_prot
);
801 out_unregister_udp_proto
:
802 proto_unregister(&udpv6_prot
);
803 out_unregister_tcp_proto
:
804 proto_unregister(&tcpv6_prot
);
807 module_init(inet6_init
);
809 static void __exit
inet6_exit(void)
811 /* First of all disallow new sockets creation. */
812 sock_unregister(PF_INET6
);
813 #ifdef CONFIG_PROC_FS
816 ipv6_misc_proc_exit();
821 /* Cleanup code parts. */
823 ip6_flowlabel_cleanup();
826 ipv6_packet_cleanup();
828 ipv6_netfilter_fini();
832 ipv6_sysctl_unregister();
835 proto_unregister(&rawv6_prot
);
836 proto_unregister(&udpv6_prot
);
837 proto_unregister(&tcpv6_prot
);
839 module_exit(inet6_exit
);
841 MODULE_ALIAS_NETPROTO(PF_INET6
);