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
)
171 if (SOCK_RAW
== sock
->type
) {
172 inet
->num
= protocol
;
173 if (IPPROTO_RAW
== protocol
)
177 sk
->sk_destruct
= inet_sock_destruct
;
178 sk
->sk_family
= PF_INET6
;
179 sk
->sk_protocol
= protocol
;
181 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
183 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
187 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
188 np
->ipv6only
= sysctl_ipv6_bindv6only
;
190 /* Init the ipv4 part of the socket since we can have sockets
191 * using v6 API for ipv4.
198 inet
->mc_list
= NULL
;
200 if (ipv4_config
.no_pmtu_disc
)
201 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
203 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
205 * Increment only the relevant sk_prot->socks debug field, this changes
206 * the previous behaviour of incrementing both the equivalent to
207 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
209 * This allows better debug granularity as we'll know exactly how many
210 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
211 * transport protocol socks. -acme
213 sk_refcnt_debug_inc(sk
);
216 /* It assumes that any protocol which allows
217 * the user to assign a number at socket
218 * creation time automatically shares.
220 inet
->sport
= ntohs(inet
->num
);
221 sk
->sk_prot
->hash(sk
);
223 if (sk
->sk_prot
->init
) {
224 err
= sk
->sk_prot
->init(sk
);
226 sk_common_release(sk
);
238 /* bind for INET6 API */
239 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
241 struct sockaddr_in6
*addr
=(struct sockaddr_in6
*)uaddr
;
242 struct sock
*sk
= sock
->sk
;
243 struct inet_sock
*inet
= inet_sk(sk
);
244 struct ipv6_pinfo
*np
= inet6_sk(sk
);
250 /* If the socket has its own bind function then use it. */
251 if (sk
->sk_prot
->bind
)
252 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
254 if (addr_len
< SIN6_LEN_RFC2133
)
256 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
257 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sock
->type
== SOCK_STREAM
)
260 snum
= ntohs(addr
->sin6_port
);
261 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
266 /* Check these errors (active socket, double bind). */
267 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->num
) {
272 /* Check if the address belongs to the host. */
273 if (addr_type
== IPV6_ADDR_MAPPED
) {
274 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
275 if (inet_addr_type(v4addr
) != RTN_LOCAL
) {
276 err
= -EADDRNOTAVAIL
;
280 if (addr_type
!= IPV6_ADDR_ANY
) {
281 struct net_device
*dev
= NULL
;
283 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
284 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
285 addr
->sin6_scope_id
) {
286 /* Override any existing binding, if another one
287 * is supplied by user.
289 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
292 /* Binding to link-local address requires an interface */
293 if (!sk
->sk_bound_dev_if
) {
297 dev
= dev_get_by_index(sk
->sk_bound_dev_if
);
304 /* ipv4 addr of the socket is invalid. Only the
305 * unspecified and mapped address have a v4 equivalent.
307 v4addr
= LOOPBACK4_IPV6
;
308 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
309 if (!ipv6_chk_addr(&addr
->sin6_addr
, dev
, 0)) {
312 err
= -EADDRNOTAVAIL
;
321 inet
->rcv_saddr
= v4addr
;
322 inet
->saddr
= v4addr
;
324 ipv6_addr_copy(&np
->rcv_saddr
, &addr
->sin6_addr
);
326 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
327 ipv6_addr_copy(&np
->saddr
, &addr
->sin6_addr
);
329 /* Make sure we are allowed to bind here. */
330 if (sk
->sk_prot
->get_port(sk
, snum
)) {
331 inet_reset_saddr(sk
);
336 if (addr_type
!= IPV6_ADDR_ANY
)
337 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
339 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
340 inet
->sport
= ntohs(inet
->num
);
348 int inet6_release(struct socket
*sock
)
350 struct sock
*sk
= sock
->sk
;
356 ipv6_sock_mc_close(sk
);
359 ipv6_sock_ac_close(sk
);
361 return inet_release(sock
);
364 int inet6_destroy_sock(struct sock
*sk
)
366 struct ipv6_pinfo
*np
= inet6_sk(sk
);
368 struct ipv6_txoptions
*opt
;
371 * Release destination entry
376 /* Release rx options */
378 if ((skb
= xchg(&np
->pktoptions
, NULL
)) != NULL
)
381 /* Free flowlabels */
382 fl6_free_socklist(sk
);
384 /* Free tx options */
386 if ((opt
= xchg(&np
->opt
, NULL
)) != NULL
)
387 sock_kfree_s(sk
, opt
, opt
->tot_len
);
393 * This does both peername and sockname.
396 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
397 int *uaddr_len
, int peer
)
399 struct sockaddr_in6
*sin
=(struct sockaddr_in6
*)uaddr
;
400 struct sock
*sk
= sock
->sk
;
401 struct inet_sock
*inet
= inet_sk(sk
);
402 struct ipv6_pinfo
*np
= inet6_sk(sk
);
404 sin
->sin6_family
= AF_INET6
;
405 sin
->sin6_flowinfo
= 0;
406 sin
->sin6_scope_id
= 0;
410 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
413 sin
->sin6_port
= inet
->dport
;
414 ipv6_addr_copy(&sin
->sin6_addr
, &np
->daddr
);
416 sin
->sin6_flowinfo
= np
->flow_label
;
418 if (ipv6_addr_any(&np
->rcv_saddr
))
419 ipv6_addr_copy(&sin
->sin6_addr
, &np
->saddr
);
421 ipv6_addr_copy(&sin
->sin6_addr
, &np
->rcv_saddr
);
423 sin
->sin6_port
= inet
->sport
;
425 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
426 sin
->sin6_scope_id
= sk
->sk_bound_dev_if
;
427 *uaddr_len
= sizeof(*sin
);
431 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
433 struct sock
*sk
= sock
->sk
;
439 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
444 return(ipv6_route_ioctl(cmd
,(void __user
*)arg
));
447 return addrconf_add_ifaddr((void __user
*) arg
);
449 return addrconf_del_ifaddr((void __user
*) arg
);
451 return addrconf_set_dstaddr((void __user
*) arg
);
453 if (!sk
->sk_prot
->ioctl
||
454 (err
= sk
->sk_prot
->ioctl(sk
, cmd
, arg
)) == -ENOIOCTLCMD
)
455 return(dev_ioctl(cmd
,(void __user
*) arg
));
462 struct proto_ops inet6_stream_ops
= {
464 .owner
= THIS_MODULE
,
465 .release
= inet6_release
,
467 .connect
= inet_stream_connect
, /* ok */
468 .socketpair
= sock_no_socketpair
, /* a do nothing */
469 .accept
= inet_accept
, /* ok */
470 .getname
= inet6_getname
,
471 .poll
= tcp_poll
, /* ok */
472 .ioctl
= inet6_ioctl
, /* must change */
473 .listen
= inet_listen
, /* ok */
474 .shutdown
= inet_shutdown
, /* ok */
475 .setsockopt
= sock_common_setsockopt
, /* ok */
476 .getsockopt
= sock_common_getsockopt
, /* ok */
477 .sendmsg
= inet_sendmsg
, /* ok */
478 .recvmsg
= sock_common_recvmsg
, /* ok */
479 .mmap
= sock_no_mmap
,
480 .sendpage
= tcp_sendpage
483 struct proto_ops inet6_dgram_ops
= {
485 .owner
= THIS_MODULE
,
486 .release
= inet6_release
,
488 .connect
= inet_dgram_connect
, /* ok */
489 .socketpair
= sock_no_socketpair
, /* a do nothing */
490 .accept
= sock_no_accept
, /* a do nothing */
491 .getname
= inet6_getname
,
492 .poll
= udp_poll
, /* ok */
493 .ioctl
= inet6_ioctl
, /* must change */
494 .listen
= sock_no_listen
, /* ok */
495 .shutdown
= inet_shutdown
, /* ok */
496 .setsockopt
= sock_common_setsockopt
, /* ok */
497 .getsockopt
= sock_common_getsockopt
, /* ok */
498 .sendmsg
= inet_sendmsg
, /* ok */
499 .recvmsg
= sock_common_recvmsg
, /* ok */
500 .mmap
= sock_no_mmap
,
501 .sendpage
= sock_no_sendpage
,
504 static struct net_proto_family inet6_family_ops
= {
506 .create
= inet6_create
,
507 .owner
= THIS_MODULE
,
510 /* Same as inet6_dgram_ops, sans udp_poll. */
511 static struct proto_ops inet6_sockraw_ops
= {
513 .owner
= THIS_MODULE
,
514 .release
= inet6_release
,
516 .connect
= inet_dgram_connect
, /* ok */
517 .socketpair
= sock_no_socketpair
, /* a do nothing */
518 .accept
= sock_no_accept
, /* a do nothing */
519 .getname
= inet6_getname
,
520 .poll
= datagram_poll
, /* ok */
521 .ioctl
= inet6_ioctl
, /* must change */
522 .listen
= sock_no_listen
, /* ok */
523 .shutdown
= inet_shutdown
, /* ok */
524 .setsockopt
= sock_common_setsockopt
, /* ok */
525 .getsockopt
= sock_common_getsockopt
, /* ok */
526 .sendmsg
= inet_sendmsg
, /* ok */
527 .recvmsg
= sock_common_recvmsg
, /* ok */
528 .mmap
= sock_no_mmap
,
529 .sendpage
= sock_no_sendpage
,
532 static struct inet_protosw rawv6_protosw
= {
534 .protocol
= IPPROTO_IP
, /* wild card */
536 .ops
= &inet6_sockraw_ops
,
537 .capability
= CAP_NET_RAW
,
538 .no_check
= UDP_CSUM_DEFAULT
,
539 .flags
= INET_PROTOSW_REUSE
,
543 inet6_register_protosw(struct inet_protosw
*p
)
545 struct list_head
*lh
;
546 struct inet_protosw
*answer
;
547 int protocol
= p
->protocol
;
548 struct list_head
*last_perm
;
550 spin_lock_bh(&inetsw6_lock
);
552 if (p
->type
>= SOCK_MAX
)
555 /* If we are trying to override a permanent protocol, bail. */
557 last_perm
= &inetsw6
[p
->type
];
558 list_for_each(lh
, &inetsw6
[p
->type
]) {
559 answer
= list_entry(lh
, struct inet_protosw
, list
);
561 /* Check only the non-wild match. */
562 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
563 if (protocol
== answer
->protocol
)
573 /* Add the new entry after the last permanent entry if any, so that
574 * the new entry does not override a permanent entry when matched with
575 * a wild-card protocol. But it is allowed to override any existing
576 * non-permanent entry. This means that when we remove this entry, the
577 * system automatically returns to the old behavior.
579 list_add_rcu(&p
->list
, last_perm
);
581 spin_unlock_bh(&inetsw6_lock
);
585 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
591 "Ignoring attempt to register invalid socket type %d.\n",
597 inet6_unregister_protosw(struct inet_protosw
*p
)
599 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
601 "Attempt to unregister permanent protocol %d.\n",
604 spin_lock_bh(&inetsw6_lock
);
605 list_del_rcu(&p
->list
);
606 spin_unlock_bh(&inetsw6_lock
);
613 snmp6_mib_init(void *ptr
[2], size_t mibsize
, size_t mibalign
)
618 ptr
[0] = __alloc_percpu(mibsize
, mibalign
);
622 ptr
[1] = __alloc_percpu(mibsize
, mibalign
);
636 snmp6_mib_free(void *ptr
[2])
644 ptr
[0] = ptr
[1] = NULL
;
647 static int __init
init_ipv6_mibs(void)
649 if (snmp6_mib_init((void **)ipv6_statistics
, sizeof (struct ipstats_mib
),
650 __alignof__(struct ipstats_mib
)) < 0)
652 if (snmp6_mib_init((void **)icmpv6_statistics
, sizeof (struct icmpv6_mib
),
653 __alignof__(struct icmpv6_mib
)) < 0)
655 if (snmp6_mib_init((void **)udp_stats_in6
, sizeof (struct udp_mib
),
656 __alignof__(struct udp_mib
)) < 0)
661 snmp6_mib_free((void **)icmpv6_statistics
);
663 snmp6_mib_free((void **)ipv6_statistics
);
669 static void cleanup_ipv6_mibs(void)
671 snmp6_mib_free((void **)ipv6_statistics
);
672 snmp6_mib_free((void **)icmpv6_statistics
);
673 snmp6_mib_free((void **)udp_stats_in6
);
676 static int __init
inet6_init(void)
678 struct sk_buff
*dummy_skb
;
683 #if 0 /* FIXME --RR */
684 if (!mod_member_present(&__this_module
, can_unload
))
687 __this_module
.can_unload
= &ipv6_unload
;
691 if (sizeof(struct inet6_skb_parm
) > sizeof(dummy_skb
->cb
)) {
692 printk(KERN_CRIT
"inet6_proto_init: size fault\n");
696 err
= proto_register(&tcpv6_prot
, 1);
700 err
= proto_register(&udpv6_prot
, 1);
702 goto out_unregister_tcp_proto
;
704 err
= proto_register(&rawv6_prot
, 1);
706 goto out_unregister_udp_proto
;
709 /* Register the socket-side information for inet6_create. */
710 for(r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
713 /* We MUST register RAW sockets before we create the ICMP6,
714 * IGMP6, or NDISC control sockets.
716 inet6_register_protosw(&rawv6_protosw
);
718 /* Register the family here so that the init calls below will
719 * be able to create sockets. (?? is this dangerous ??)
721 err
= sock_register(&inet6_family_ops
);
723 goto out_unregister_raw_proto
;
725 /* Initialise ipv6 mibs */
726 err
= init_ipv6_mibs();
728 goto out_unregister_sock
;
731 * ipngwg API draft makes clear that the correct semantics
732 * for TCP and UDP is to consider one TCP and UDP instance
733 * in a host availiable by both INET and INET6 APIs and
734 * able to communicate via both network protocols.
738 ipv6_sysctl_register();
740 err
= icmpv6_init(&inet6_family_ops
);
743 err
= ndisc_init(&inet6_family_ops
);
746 err
= igmp6_init(&inet6_family_ops
);
749 err
= ipv6_netfilter_init();
752 /* Create /proc/foo6 entries. */
753 #ifdef CONFIG_PROC_FS
755 if (raw6_proc_init())
757 if (tcp6_proc_init())
759 if (udp6_proc_init())
761 if (ipv6_misc_proc_init())
762 goto proc_misc6_fail
;
765 goto proc_anycast6_fail
;
770 ip6_flowlabel_init();
771 err
= addrconf_init();
776 /* Init v6 extension headers. */
782 /* Init v6 transport protocols. */
792 ip6_flowlabel_cleanup();
794 #ifdef CONFIG_PROC_FS
799 ipv6_misc_proc_exit();
808 ipv6_netfilter_fini();
817 ipv6_sysctl_unregister();
821 sock_unregister(PF_INET6
);
822 out_unregister_raw_proto
:
823 proto_unregister(&rawv6_prot
);
824 out_unregister_udp_proto
:
825 proto_unregister(&udpv6_prot
);
826 out_unregister_tcp_proto
:
827 proto_unregister(&tcpv6_prot
);
830 module_init(inet6_init
);
832 static void __exit
inet6_exit(void)
834 /* First of all disallow new sockets creation. */
835 sock_unregister(PF_INET6
);
836 #ifdef CONFIG_PROC_FS
839 ipv6_misc_proc_exit();
844 /* Cleanup code parts. */
846 ip6_flowlabel_cleanup();
849 ipv6_packet_cleanup();
851 ipv6_netfilter_fini();
855 ipv6_sysctl_unregister();
858 proto_unregister(&rawv6_prot
);
859 proto_unregister(&udpv6_prot
);
860 proto_unregister(&tcpv6_prot
);
862 module_exit(inet6_exit
);
864 MODULE_ALIAS_NETPROTO(PF_INET6
);