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/major.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>
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 #ifdef INET_REFCNT_DEBUG
90 atomic_t inet6_sock_nr
;
93 /* The inetsw table contains everything that inet_create needs to
96 static struct list_head inetsw6
[SOCK_MAX
];
97 static spinlock_t inetsw6_lock
= SPIN_LOCK_UNLOCKED
;
99 static void inet6_sock_destruct(struct sock
*sk
)
101 inet_sock_destruct(sk
);
103 #ifdef INET_REFCNT_DEBUG
104 atomic_dec(&inet6_sock_nr
);
108 static __inline__
struct ipv6_pinfo
*inet6_sk_generic(struct sock
*sk
)
110 const int offset
= sk
->sk_prot
->slab_obj_size
- sizeof(struct ipv6_pinfo
);
112 return (struct ipv6_pinfo
*)(((u8
*)sk
) + offset
);
115 static int inet6_create(struct socket
*sock
, int protocol
)
117 struct inet_opt
*inet
;
118 struct ipv6_pinfo
*np
;
120 struct tcp6_sock
* tcp6sk
;
122 struct inet_protosw
*answer
;
123 struct proto
*answer_prot
;
124 unsigned char answer_flags
;
125 char answer_no_check
;
128 /* Look for the requested type/protocol pair. */
131 list_for_each_rcu(p
, &inetsw6
[sock
->type
]) {
132 answer
= list_entry(p
, struct inet_protosw
, list
);
134 /* Check the non-wild match. */
135 if (protocol
== answer
->protocol
) {
136 if (protocol
!= IPPROTO_IP
)
139 /* Check for the two wild cases. */
140 if (IPPROTO_IP
== protocol
) {
141 protocol
= answer
->protocol
;
144 if (IPPROTO_IP
== answer
->protocol
)
150 rc
= -ESOCKTNOSUPPORT
;
154 if (answer
->capability
> 0 && !capable(answer
->capability
))
156 rc
= -EPROTONOSUPPORT
;
160 sock
->ops
= answer
->ops
;
162 answer_prot
= answer
->prot
;
163 answer_no_check
= answer
->no_check
;
164 answer_flags
= answer
->flags
;
167 BUG_TRAP(answer_prot
->slab
!= NULL
);
170 sk
= sk_alloc(PF_INET6
, GFP_KERNEL
,
171 answer_prot
->slab_obj_size
,
176 sock_init_data(sock
, sk
);
177 sk_set_owner(sk
, THIS_MODULE
);
180 sk
->sk_prot
= answer_prot
;
181 sk
->sk_no_check
= answer_no_check
;
182 if (INET_PROTOSW_REUSE
& answer_flags
)
187 if (SOCK_RAW
== sock
->type
) {
188 inet
->num
= protocol
;
189 if (IPPROTO_RAW
== protocol
)
193 sk
->sk_destruct
= inet6_sock_destruct
;
194 sk
->sk_family
= PF_INET6
;
195 sk
->sk_protocol
= protocol
;
197 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
199 tcp6sk
= (struct tcp6_sock
*)sk
;
200 tcp6sk
->pinet6
= np
= inet6_sk_generic(sk
);
204 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
205 np
->ipv6only
= sysctl_ipv6_bindv6only
;
207 /* Init the ipv4 part of the socket since we can have sockets
208 * using v6 API for ipv4.
215 inet
->mc_list
= NULL
;
217 if (ipv4_config
.no_pmtu_disc
)
218 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
220 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
223 #ifdef INET_REFCNT_DEBUG
224 atomic_inc(&inet6_sock_nr
);
225 atomic_inc(&inet_sock_nr
);
228 /* It assumes that any protocol which allows
229 * the user to assign a number at socket
230 * creation time automatically shares.
232 inet
->sport
= ntohs(inet
->num
);
233 sk
->sk_prot
->hash(sk
);
235 if (sk
->sk_prot
->init
) {
236 rc
= sk
->sk_prot
->init(sk
);
238 sk_common_release(sk
);
250 /* bind for INET6 API */
251 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
253 struct sockaddr_in6
*addr
=(struct sockaddr_in6
*)uaddr
;
254 struct sock
*sk
= sock
->sk
;
255 struct inet_opt
*inet
= inet_sk(sk
);
256 struct ipv6_pinfo
*np
= inet6_sk(sk
);
262 /* If the socket has its own bind function then use it. */
263 if (sk
->sk_prot
->bind
)
264 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
266 if (addr_len
< SIN6_LEN_RFC2133
)
268 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
269 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sock
->type
== SOCK_STREAM
)
272 snum
= ntohs(addr
->sin6_port
);
273 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
278 /* Check these errors (active socket, double bind). */
279 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->num
) {
284 /* Check if the address belongs to the host. */
285 if (addr_type
== IPV6_ADDR_MAPPED
) {
286 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
287 if (inet_addr_type(v4addr
) != RTN_LOCAL
) {
288 err
= -EADDRNOTAVAIL
;
292 if (addr_type
!= IPV6_ADDR_ANY
) {
293 struct net_device
*dev
= NULL
;
295 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
296 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
297 addr
->sin6_scope_id
) {
298 /* Override any existing binding, if another one
299 * is supplied by user.
301 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
304 /* Binding to link-local address requires an interface */
305 if (!sk
->sk_bound_dev_if
) {
309 dev
= dev_get_by_index(sk
->sk_bound_dev_if
);
316 /* ipv4 addr of the socket is invalid. Only the
317 * unspecified and mapped address have a v4 equivalent.
319 v4addr
= LOOPBACK4_IPV6
;
320 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
321 if (!ipv6_chk_addr(&addr
->sin6_addr
, dev
, 0)) {
324 err
= -EADDRNOTAVAIL
;
333 inet
->rcv_saddr
= v4addr
;
334 inet
->saddr
= v4addr
;
336 ipv6_addr_copy(&np
->rcv_saddr
, &addr
->sin6_addr
);
338 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
339 ipv6_addr_copy(&np
->saddr
, &addr
->sin6_addr
);
341 /* Make sure we are allowed to bind here. */
342 if (sk
->sk_prot
->get_port(sk
, snum
)) {
343 inet_reset_saddr(sk
);
348 if (addr_type
!= IPV6_ADDR_ANY
)
349 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
351 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
352 inet
->sport
= ntohs(inet
->num
);
360 int inet6_release(struct socket
*sock
)
362 struct sock
*sk
= sock
->sk
;
368 ipv6_sock_mc_close(sk
);
371 ipv6_sock_ac_close(sk
);
373 return inet_release(sock
);
376 int inet6_destroy_sock(struct sock
*sk
)
378 struct ipv6_pinfo
*np
= inet6_sk(sk
);
380 struct ipv6_txoptions
*opt
;
383 * Release destination entry
388 /* Release rx options */
390 if ((skb
= xchg(&np
->pktoptions
, NULL
)) != NULL
)
393 /* Free flowlabels */
394 fl6_free_socklist(sk
);
396 /* Free tx options */
398 if ((opt
= xchg(&np
->opt
, NULL
)) != NULL
)
399 sock_kfree_s(sk
, opt
, opt
->tot_len
);
405 * This does both peername and sockname.
408 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
409 int *uaddr_len
, int peer
)
411 struct sockaddr_in6
*sin
=(struct sockaddr_in6
*)uaddr
;
412 struct sock
*sk
= sock
->sk
;
413 struct inet_opt
*inet
= inet_sk(sk
);
414 struct ipv6_pinfo
*np
= inet6_sk(sk
);
416 sin
->sin6_family
= AF_INET6
;
417 sin
->sin6_flowinfo
= 0;
418 sin
->sin6_scope_id
= 0;
422 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
425 sin
->sin6_port
= inet
->dport
;
426 ipv6_addr_copy(&sin
->sin6_addr
, &np
->daddr
);
428 sin
->sin6_flowinfo
= np
->flow_label
;
430 if (ipv6_addr_any(&np
->rcv_saddr
))
431 ipv6_addr_copy(&sin
->sin6_addr
, &np
->saddr
);
433 ipv6_addr_copy(&sin
->sin6_addr
, &np
->rcv_saddr
);
435 sin
->sin6_port
= inet
->sport
;
437 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
438 sin
->sin6_scope_id
= sk
->sk_bound_dev_if
;
439 *uaddr_len
= sizeof(*sin
);
443 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
445 struct sock
*sk
= sock
->sk
;
451 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
456 return(ipv6_route_ioctl(cmd
,(void __user
*)arg
));
459 return addrconf_add_ifaddr((void __user
*) arg
);
461 return addrconf_del_ifaddr((void __user
*) arg
);
463 return addrconf_set_dstaddr((void __user
*) arg
);
465 if (!sk
->sk_prot
->ioctl
||
466 (err
= sk
->sk_prot
->ioctl(sk
, cmd
, arg
)) == -ENOIOCTLCMD
)
467 return(dev_ioctl(cmd
,(void __user
*) arg
));
474 struct proto_ops inet6_stream_ops
= {
476 .owner
= THIS_MODULE
,
477 .release
= inet6_release
,
479 .connect
= inet_stream_connect
, /* ok */
480 .socketpair
= sock_no_socketpair
, /* a do nothing */
481 .accept
= inet_accept
, /* ok */
482 .getname
= inet6_getname
,
483 .poll
= tcp_poll
, /* ok */
484 .ioctl
= inet6_ioctl
, /* must change */
485 .listen
= inet_listen
, /* ok */
486 .shutdown
= inet_shutdown
, /* ok */
487 .setsockopt
= sock_common_setsockopt
, /* ok */
488 .getsockopt
= sock_common_getsockopt
, /* ok */
489 .sendmsg
= inet_sendmsg
, /* ok */
490 .recvmsg
= sock_common_recvmsg
, /* ok */
491 .mmap
= sock_no_mmap
,
492 .sendpage
= tcp_sendpage
495 struct proto_ops inet6_dgram_ops
= {
497 .owner
= THIS_MODULE
,
498 .release
= inet6_release
,
500 .connect
= inet_dgram_connect
, /* ok */
501 .socketpair
= sock_no_socketpair
, /* a do nothing */
502 .accept
= sock_no_accept
, /* a do nothing */
503 .getname
= inet6_getname
,
504 .poll
= datagram_poll
, /* ok */
505 .ioctl
= inet6_ioctl
, /* must change */
506 .listen
= sock_no_listen
, /* ok */
507 .shutdown
= inet_shutdown
, /* ok */
508 .setsockopt
= sock_common_setsockopt
, /* ok */
509 .getsockopt
= sock_common_getsockopt
, /* ok */
510 .sendmsg
= inet_sendmsg
, /* ok */
511 .recvmsg
= sock_common_recvmsg
, /* ok */
512 .mmap
= sock_no_mmap
,
513 .sendpage
= sock_no_sendpage
,
516 static struct net_proto_family inet6_family_ops
= {
518 .create
= inet6_create
,
519 .owner
= THIS_MODULE
,
523 extern void ipv6_sysctl_register(void);
524 extern void ipv6_sysctl_unregister(void);
527 static struct inet_protosw rawv6_protosw
= {
529 .protocol
= IPPROTO_IP
, /* wild card */
531 .ops
= &inet6_dgram_ops
,
532 .capability
= CAP_NET_RAW
,
533 .no_check
= UDP_CSUM_DEFAULT
,
534 .flags
= INET_PROTOSW_REUSE
,
538 inet6_register_protosw(struct inet_protosw
*p
)
540 struct list_head
*lh
;
541 struct inet_protosw
*answer
;
542 int protocol
= p
->protocol
;
543 struct list_head
*last_perm
;
545 spin_lock_bh(&inetsw6_lock
);
547 if (p
->type
>= SOCK_MAX
)
550 /* If we are trying to override a permanent protocol, bail. */
552 last_perm
= &inetsw6
[p
->type
];
553 list_for_each(lh
, &inetsw6
[p
->type
]) {
554 answer
= list_entry(lh
, struct inet_protosw
, list
);
556 /* Check only the non-wild match. */
557 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
558 if (protocol
== answer
->protocol
)
568 /* Add the new entry after the last permanent entry if any, so that
569 * the new entry does not override a permanent entry when matched with
570 * a wild-card protocol. But it is allowed to override any existing
571 * non-permanent entry. This means that when we remove this entry, the
572 * system automatically returns to the old behavior.
574 list_add_rcu(&p
->list
, last_perm
);
576 spin_unlock_bh(&inetsw6_lock
);
580 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
586 "Ignoring attempt to register invalid socket type %d.\n",
592 inet6_unregister_protosw(struct inet_protosw
*p
)
594 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
596 "Attempt to unregister permanent protocol %d.\n",
599 spin_lock_bh(&inetsw6_lock
);
600 list_del_rcu(&p
->list
);
601 spin_unlock_bh(&inetsw6_lock
);
608 snmp6_mib_init(void *ptr
[2], size_t mibsize
, size_t mibalign
)
613 ptr
[0] = __alloc_percpu(mibsize
, mibalign
);
617 ptr
[1] = __alloc_percpu(mibsize
, mibalign
);
631 snmp6_mib_free(void *ptr
[2])
637 ptr
[0] = ptr
[1] = NULL
;
640 static int __init
init_ipv6_mibs(void)
642 if (snmp6_mib_init((void **)ipv6_statistics
, sizeof (struct ipstats_mib
),
643 __alignof__(struct ipstats_mib
)) < 0)
645 if (snmp6_mib_init((void **)icmpv6_statistics
, sizeof (struct icmpv6_mib
),
646 __alignof__(struct icmpv6_mib
)) < 0)
648 if (snmp6_mib_init((void **)udp_stats_in6
, sizeof (struct udp_mib
),
649 __alignof__(struct udp_mib
)) < 0)
654 snmp6_mib_free((void **)icmpv6_statistics
);
656 snmp6_mib_free((void **)ipv6_statistics
);
662 static void cleanup_ipv6_mibs(void)
664 snmp6_mib_free((void **)ipv6_statistics
);
665 snmp6_mib_free((void **)icmpv6_statistics
);
666 snmp6_mib_free((void **)udp_stats_in6
);
669 extern int ipv6_misc_proc_init(void);
671 static int __init
inet6_init(void)
673 struct sk_buff
*dummy_skb
;
678 #if 0 /* FIXME --RR */
679 if (!mod_member_present(&__this_module
, can_unload
))
682 __this_module
.can_unload
= &ipv6_unload
;
686 if (sizeof(struct inet6_skb_parm
) > sizeof(dummy_skb
->cb
)) {
687 printk(KERN_CRIT
"inet6_proto_init: size fault\n");
691 err
= sk_alloc_slab(&tcpv6_prot
, "tcpv6_sock");
693 sk_alloc_slab_error(&tcpv6_prot
);
696 err
= sk_alloc_slab(&udpv6_prot
, "udpv6_sock");
698 sk_alloc_slab_error(&udpv6_prot
);
699 goto out_tcp_free_slab
;
701 err
= sk_alloc_slab(&rawv6_prot
, "rawv6_sock");
703 sk_alloc_slab_error(&rawv6_prot
);
704 goto out_udp_free_slab
;
707 /* Register the socket-side information for inet6_create. */
708 for(r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
711 /* We MUST register RAW sockets before we create the ICMP6,
712 * IGMP6, or NDISC control sockets.
714 inet6_register_protosw(&rawv6_protosw
);
716 /* Register the family here so that the init calls below will
717 * be able to create sockets. (?? is this dangerous ??)
719 (void) sock_register(&inet6_family_ops
);
721 /* Initialise ipv6 mibs */
722 err
= init_ipv6_mibs();
724 goto out_raw_free_slab
;
727 * ipngwg API draft makes clear that the correct semantics
728 * for TCP and UDP is to consider one TCP and UDP instance
729 * in a host availiable by both INET and INET6 APIs and
730 * able to communicate via both network protocols.
734 ipv6_sysctl_register();
736 err
= icmpv6_init(&inet6_family_ops
);
739 err
= ndisc_init(&inet6_family_ops
);
742 err
= igmp6_init(&inet6_family_ops
);
745 /* Create /proc/foo6 entries. */
746 #ifdef CONFIG_PROC_FS
748 if (raw6_proc_init())
750 if (tcp6_proc_init())
752 if (udp6_proc_init())
754 if (ipv6_misc_proc_init())
755 goto proc_misc6_fail
;
758 goto proc_anycast6_fail
;
764 ip6_flowlabel_init();
768 /* Init v6 extension headers. */
774 /* Init v6 transport protocols. */
781 #ifdef CONFIG_PROC_FS
785 ipv6_misc_proc_exit();
801 ipv6_sysctl_unregister();
805 sk_free_slab(&rawv6_prot
);
807 sk_free_slab(&udpv6_prot
);
809 sk_free_slab(&tcpv6_prot
);
812 module_init(inet6_init
);
814 static void __exit
inet6_exit(void)
816 /* First of all disallow new sockets creation. */
817 sock_unregister(PF_INET6
);
818 #ifdef CONFIG_PROC_FS
821 ipv6_misc_proc_exit();
826 /* Cleanup code parts. */
828 ip6_flowlabel_cleanup();
831 ipv6_packet_cleanup();
836 ipv6_sysctl_unregister();
839 sk_free_slab(&rawv6_prot
);
840 sk_free_slab(&udpv6_prot
);
841 sk_free_slab(&tcpv6_prot
);
843 module_exit(inet6_exit
);
845 MODULE_ALIAS_NETPROTO(PF_INET6
);