1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Adapted from linux/net/ipv4/af_inet.c
12 * piggy, Karl Knutson : Socket protocol table
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * Arnaldo Melo : check proc_net_create return, cleanups
17 #define pr_fmt(fmt) "IPv6: " fmt
19 #include <linux/module.h>
20 #include <linux/capability.h>
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
25 #include <linux/kernel.h>
26 #include <linux/timer.h>
27 #include <linux/string.h>
28 #include <linux/sockios.h>
29 #include <linux/net.h>
30 #include <linux/fcntl.h>
32 #include <linux/interrupt.h>
33 #include <linux/proc_fs.h>
34 #include <linux/stat.h>
35 #include <linux/init.h>
36 #include <linux/slab.h>
38 #include <linux/inet.h>
39 #include <linux/netdevice.h>
40 #include <linux/icmpv6.h>
41 #include <linux/netfilter_ipv6.h>
46 #include <net/udplite.h>
49 #include <net/protocol.h>
50 #include <net/inet_common.h>
51 #include <net/route.h>
52 #include <net/transp_v6.h>
53 #include <net/ip6_route.h>
54 #include <net/addrconf.h>
55 #include <net/ipv6_stubs.h>
56 #include <net/ndisc.h>
57 #ifdef CONFIG_IPV6_TUNNEL
58 #include <net/ip6_tunnel.h>
60 #include <net/calipso.h>
63 #include <linux/uaccess.h>
64 #include <linux/mroute6.h>
66 #include "ip6_offload.h"
68 MODULE_AUTHOR("Cast of dozens");
69 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
70 MODULE_LICENSE("GPL");
72 /* The inetsw6 table contains everything that inet6_create needs to
75 static struct list_head inetsw6
[SOCK_MAX
];
76 static DEFINE_SPINLOCK(inetsw6_lock
);
78 struct ipv6_params ipv6_defaults
= {
83 static int disable_ipv6_mod
;
85 module_param_named(disable
, disable_ipv6_mod
, int, 0444);
86 MODULE_PARM_DESC(disable
, "Disable IPv6 module such that it is non-functional");
88 module_param_named(disable_ipv6
, ipv6_defaults
.disable_ipv6
, int, 0444);
89 MODULE_PARM_DESC(disable_ipv6
, "Disable IPv6 on all interfaces");
91 module_param_named(autoconf
, ipv6_defaults
.autoconf
, int, 0444);
92 MODULE_PARM_DESC(autoconf
, "Enable IPv6 address autoconfiguration on all interfaces");
94 bool ipv6_mod_enabled(void)
96 return disable_ipv6_mod
== 0;
98 EXPORT_SYMBOL_GPL(ipv6_mod_enabled
);
100 static __inline__
struct ipv6_pinfo
*inet6_sk_generic(struct sock
*sk
)
102 const int offset
= sk
->sk_prot
->obj_size
- sizeof(struct ipv6_pinfo
);
104 return (struct ipv6_pinfo
*)(((u8
*)sk
) + offset
);
107 static int inet6_create(struct net
*net
, struct socket
*sock
, int protocol
,
110 struct inet_sock
*inet
;
111 struct ipv6_pinfo
*np
;
113 struct inet_protosw
*answer
;
114 struct proto
*answer_prot
;
115 unsigned char answer_flags
;
116 int try_loading_module
= 0;
119 if (protocol
< 0 || protocol
>= IPPROTO_MAX
)
122 /* Look for the requested type/protocol pair. */
124 err
= -ESOCKTNOSUPPORT
;
126 list_for_each_entry_rcu(answer
, &inetsw6
[sock
->type
], list
) {
129 /* Check the non-wild match. */
130 if (protocol
== answer
->protocol
) {
131 if (protocol
!= IPPROTO_IP
)
134 /* Check for the two wild cases. */
135 if (IPPROTO_IP
== protocol
) {
136 protocol
= answer
->protocol
;
139 if (IPPROTO_IP
== answer
->protocol
)
142 err
= -EPROTONOSUPPORT
;
146 if (try_loading_module
< 2) {
149 * Be more specific, e.g. net-pf-10-proto-132-type-1
150 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
152 if (++try_loading_module
== 1)
153 request_module("net-pf-%d-proto-%d-type-%d",
154 PF_INET6
, protocol
, sock
->type
);
156 * Fall back to generic, e.g. net-pf-10-proto-132
157 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
160 request_module("net-pf-%d-proto-%d",
162 goto lookup_protocol
;
168 if (sock
->type
== SOCK_RAW
&& !kern
&&
169 !ns_capable(net
->user_ns
, CAP_NET_RAW
))
172 sock
->ops
= answer
->ops
;
173 answer_prot
= answer
->prot
;
174 answer_flags
= answer
->flags
;
177 WARN_ON(!answer_prot
->slab
);
180 sk
= sk_alloc(net
, PF_INET6
, GFP_KERNEL
, answer_prot
, kern
);
184 sock_init_data(sock
, sk
);
187 if (INET_PROTOSW_REUSE
& answer_flags
)
188 sk
->sk_reuse
= SK_CAN_REUSE
;
191 inet
->is_icsk
= (INET_PROTOSW_ICSK
& answer_flags
) != 0;
193 if (SOCK_RAW
== sock
->type
) {
194 inet
->inet_num
= protocol
;
195 if (IPPROTO_RAW
== protocol
)
199 sk
->sk_destruct
= inet_sock_destruct
;
200 sk
->sk_family
= PF_INET6
;
201 sk
->sk_protocol
= protocol
;
203 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
205 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
207 np
->mcast_hops
= IPV6_DEFAULT_MCASTHOPS
;
210 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
211 np
->repflow
= net
->ipv6
.sysctl
.flowlabel_reflect
;
212 sk
->sk_ipv6only
= net
->ipv6
.sysctl
.bindv6only
;
214 /* Init the ipv4 part of the socket since we can have sockets
215 * using v6 API for ipv4.
222 inet
->mc_list
= NULL
;
225 if (net
->ipv4
.sysctl_ip_no_pmtu_disc
)
226 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
228 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
230 * Increment only the relevant sk_prot->socks debug field, this changes
231 * the previous behaviour of incrementing both the equivalent to
232 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
234 * This allows better debug granularity as we'll know exactly how many
235 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
236 * transport protocol socks. -acme
238 sk_refcnt_debug_inc(sk
);
240 if (inet
->inet_num
) {
241 /* It assumes that any protocol which allows
242 * the user to assign a number at socket
243 * creation time automatically shares.
245 inet
->inet_sport
= htons(inet
->inet_num
);
246 err
= sk
->sk_prot
->hash(sk
);
248 sk_common_release(sk
);
252 if (sk
->sk_prot
->init
) {
253 err
= sk
->sk_prot
->init(sk
);
255 sk_common_release(sk
);
261 err
= BPF_CGROUP_RUN_PROG_INET_SOCK(sk
);
263 sk_common_release(sk
);
274 static int __inet6_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
,
275 bool force_bind_address_no_port
, bool with_lock
)
277 struct sockaddr_in6
*addr
= (struct sockaddr_in6
*)uaddr
;
278 struct inet_sock
*inet
= inet_sk(sk
);
279 struct ipv6_pinfo
*np
= inet6_sk(sk
);
280 struct net
*net
= sock_net(sk
);
287 if (addr
->sin6_family
!= AF_INET6
)
288 return -EAFNOSUPPORT
;
290 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
291 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sk
->sk_type
== SOCK_STREAM
)
294 snum
= ntohs(addr
->sin6_port
);
295 if (snum
&& snum
< inet_prot_sock(net
) &&
296 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
302 /* Check these errors (active socket, double bind). */
303 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->inet_num
) {
308 /* Check if the address belongs to the host. */
309 if (addr_type
== IPV6_ADDR_MAPPED
) {
310 struct net_device
*dev
= NULL
;
313 /* Binding to v4-mapped address on a v6-only socket
316 if (sk
->sk_ipv6only
) {
322 if (sk
->sk_bound_dev_if
) {
323 dev
= dev_get_by_index_rcu(net
, sk
->sk_bound_dev_if
);
330 /* Reproduce AF_INET checks to make the bindings consistent */
331 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
332 chk_addr_ret
= inet_addr_type_dev_table(net
, dev
, v4addr
);
335 if (!inet_can_nonlocal_bind(net
, inet
) &&
336 v4addr
!= htonl(INADDR_ANY
) &&
337 chk_addr_ret
!= RTN_LOCAL
&&
338 chk_addr_ret
!= RTN_MULTICAST
&&
339 chk_addr_ret
!= RTN_BROADCAST
) {
340 err
= -EADDRNOTAVAIL
;
344 if (addr_type
!= IPV6_ADDR_ANY
) {
345 struct net_device
*dev
= NULL
;
348 if (__ipv6_addr_needs_scope_id(addr_type
)) {
349 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
350 addr
->sin6_scope_id
) {
351 /* Override any existing binding, if another one
352 * is supplied by user.
354 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
357 /* Binding to link-local address requires an interface */
358 if (!sk
->sk_bound_dev_if
) {
364 if (sk
->sk_bound_dev_if
) {
365 dev
= dev_get_by_index_rcu(net
, sk
->sk_bound_dev_if
);
372 /* ipv4 addr of the socket is invalid. Only the
373 * unspecified and mapped address have a v4 equivalent.
375 v4addr
= LOOPBACK4_IPV6
;
376 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
377 if (!ipv6_can_nonlocal_bind(net
, inet
) &&
378 !ipv6_chk_addr(net
, &addr
->sin6_addr
,
380 err
= -EADDRNOTAVAIL
;
388 inet
->inet_rcv_saddr
= v4addr
;
389 inet
->inet_saddr
= v4addr
;
391 sk
->sk_v6_rcv_saddr
= addr
->sin6_addr
;
393 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
394 np
->saddr
= addr
->sin6_addr
;
396 saved_ipv6only
= sk
->sk_ipv6only
;
397 if (addr_type
!= IPV6_ADDR_ANY
&& addr_type
!= IPV6_ADDR_MAPPED
)
400 /* Make sure we are allowed to bind here. */
401 if (snum
|| !(inet
->bind_address_no_port
||
402 force_bind_address_no_port
)) {
403 if (sk
->sk_prot
->get_port(sk
, snum
)) {
404 sk
->sk_ipv6only
= saved_ipv6only
;
405 inet_reset_saddr(sk
);
409 err
= BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk
);
411 sk
->sk_ipv6only
= saved_ipv6only
;
412 inet_reset_saddr(sk
);
417 if (addr_type
!= IPV6_ADDR_ANY
)
418 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
420 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
421 inet
->inet_sport
= htons(inet
->inet_num
);
422 inet
->inet_dport
= 0;
423 inet
->inet_daddr
= 0;
433 /* bind for INET6 API */
434 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
436 struct sock
*sk
= sock
->sk
;
439 /* If the socket has its own bind function then use it. */
440 if (sk
->sk_prot
->bind
)
441 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
443 if (addr_len
< SIN6_LEN_RFC2133
)
446 /* BPF prog is run before any checks are done so that if the prog
447 * changes context in a wrong way it will be caught.
449 err
= BPF_CGROUP_RUN_PROG_INET6_BIND(sk
, uaddr
);
453 return __inet6_bind(sk
, uaddr
, addr_len
, false, true);
455 EXPORT_SYMBOL(inet6_bind
);
457 int inet6_release(struct socket
*sock
)
459 struct sock
*sk
= sock
->sk
;
465 ipv6_sock_mc_close(sk
);
468 ipv6_sock_ac_close(sk
);
470 return inet_release(sock
);
472 EXPORT_SYMBOL(inet6_release
);
474 void inet6_destroy_sock(struct sock
*sk
)
476 struct ipv6_pinfo
*np
= inet6_sk(sk
);
478 struct ipv6_txoptions
*opt
;
480 /* Release rx options */
482 skb
= xchg(&np
->pktoptions
, NULL
);
485 skb
= xchg(&np
->rxpmtu
, NULL
);
488 /* Free flowlabels */
489 fl6_free_socklist(sk
);
491 /* Free tx options */
493 opt
= xchg((__force
struct ipv6_txoptions
**)&np
->opt
, NULL
);
495 atomic_sub(opt
->tot_len
, &sk
->sk_omem_alloc
);
499 EXPORT_SYMBOL_GPL(inet6_destroy_sock
);
502 * This does both peername and sockname.
505 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
508 struct sockaddr_in6
*sin
= (struct sockaddr_in6
*)uaddr
;
509 struct sock
*sk
= sock
->sk
;
510 struct inet_sock
*inet
= inet_sk(sk
);
511 struct ipv6_pinfo
*np
= inet6_sk(sk
);
513 sin
->sin6_family
= AF_INET6
;
514 sin
->sin6_flowinfo
= 0;
515 sin
->sin6_scope_id
= 0;
517 if (!inet
->inet_dport
)
519 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
522 sin
->sin6_port
= inet
->inet_dport
;
523 sin
->sin6_addr
= sk
->sk_v6_daddr
;
525 sin
->sin6_flowinfo
= np
->flow_label
;
527 if (ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
528 sin
->sin6_addr
= np
->saddr
;
530 sin
->sin6_addr
= sk
->sk_v6_rcv_saddr
;
532 sin
->sin6_port
= inet
->inet_sport
;
534 sin
->sin6_scope_id
= ipv6_iface_scope_id(&sin
->sin6_addr
,
535 sk
->sk_bound_dev_if
);
538 EXPORT_SYMBOL(inet6_getname
);
540 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
542 struct sock
*sk
= sock
->sk
;
543 struct net
*net
= sock_net(sk
);
549 return ipv6_route_ioctl(net
, cmd
, (void __user
*)arg
);
552 return addrconf_add_ifaddr(net
, (void __user
*) arg
);
554 return addrconf_del_ifaddr(net
, (void __user
*) arg
);
556 return addrconf_set_dstaddr(net
, (void __user
*) arg
);
558 if (!sk
->sk_prot
->ioctl
)
560 return sk
->sk_prot
->ioctl(sk
, cmd
, arg
);
565 EXPORT_SYMBOL(inet6_ioctl
);
567 const struct proto_ops inet6_stream_ops
= {
569 .owner
= THIS_MODULE
,
570 .release
= inet6_release
,
572 .connect
= inet_stream_connect
, /* ok */
573 .socketpair
= sock_no_socketpair
, /* a do nothing */
574 .accept
= inet_accept
, /* ok */
575 .getname
= inet6_getname
,
576 .poll
= tcp_poll
, /* ok */
577 .ioctl
= inet6_ioctl
, /* must change */
578 .gettstamp
= sock_gettstamp
,
579 .listen
= inet_listen
, /* ok */
580 .shutdown
= inet_shutdown
, /* ok */
581 .setsockopt
= sock_common_setsockopt
, /* ok */
582 .getsockopt
= sock_common_getsockopt
, /* ok */
583 .sendmsg
= inet_sendmsg
, /* ok */
584 .recvmsg
= inet_recvmsg
, /* ok */
588 .sendpage
= inet_sendpage
,
589 .sendmsg_locked
= tcp_sendmsg_locked
,
590 .sendpage_locked
= tcp_sendpage_locked
,
591 .splice_read
= tcp_splice_read
,
592 .read_sock
= tcp_read_sock
,
593 .peek_len
= tcp_peek_len
,
595 .compat_setsockopt
= compat_sock_common_setsockopt
,
596 .compat_getsockopt
= compat_sock_common_getsockopt
,
598 .set_rcvlowat
= tcp_set_rcvlowat
,
601 const struct proto_ops inet6_dgram_ops
= {
603 .owner
= THIS_MODULE
,
604 .release
= inet6_release
,
606 .connect
= inet_dgram_connect
, /* ok */
607 .socketpair
= sock_no_socketpair
, /* a do nothing */
608 .accept
= sock_no_accept
, /* a do nothing */
609 .getname
= inet6_getname
,
610 .poll
= udp_poll
, /* ok */
611 .ioctl
= inet6_ioctl
, /* must change */
612 .gettstamp
= sock_gettstamp
,
613 .listen
= sock_no_listen
, /* ok */
614 .shutdown
= inet_shutdown
, /* ok */
615 .setsockopt
= sock_common_setsockopt
, /* ok */
616 .getsockopt
= sock_common_getsockopt
, /* ok */
617 .sendmsg
= inet_sendmsg
, /* ok */
618 .recvmsg
= inet_recvmsg
, /* ok */
619 .mmap
= sock_no_mmap
,
620 .sendpage
= sock_no_sendpage
,
621 .set_peek_off
= sk_set_peek_off
,
623 .compat_setsockopt
= compat_sock_common_setsockopt
,
624 .compat_getsockopt
= compat_sock_common_getsockopt
,
628 static const struct net_proto_family inet6_family_ops
= {
630 .create
= inet6_create
,
631 .owner
= THIS_MODULE
,
634 int inet6_register_protosw(struct inet_protosw
*p
)
636 struct list_head
*lh
;
637 struct inet_protosw
*answer
;
638 struct list_head
*last_perm
;
639 int protocol
= p
->protocol
;
642 spin_lock_bh(&inetsw6_lock
);
645 if (p
->type
>= SOCK_MAX
)
648 /* If we are trying to override a permanent protocol, bail. */
651 last_perm
= &inetsw6
[p
->type
];
652 list_for_each(lh
, &inetsw6
[p
->type
]) {
653 answer
= list_entry(lh
, struct inet_protosw
, list
);
655 /* Check only the non-wild match. */
656 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
657 if (protocol
== answer
->protocol
)
667 /* Add the new entry after the last permanent entry if any, so that
668 * the new entry does not override a permanent entry when matched with
669 * a wild-card protocol. But it is allowed to override any existing
670 * non-permanent entry. This means that when we remove this entry, the
671 * system automatically returns to the old behavior.
673 list_add_rcu(&p
->list
, last_perm
);
676 spin_unlock_bh(&inetsw6_lock
);
680 pr_err("Attempt to override permanent protocol %d\n", protocol
);
684 pr_err("Ignoring attempt to register invalid socket type %d\n",
688 EXPORT_SYMBOL(inet6_register_protosw
);
691 inet6_unregister_protosw(struct inet_protosw
*p
)
693 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
694 pr_err("Attempt to unregister permanent protocol %d\n",
697 spin_lock_bh(&inetsw6_lock
);
698 list_del_rcu(&p
->list
);
699 spin_unlock_bh(&inetsw6_lock
);
704 EXPORT_SYMBOL(inet6_unregister_protosw
);
706 int inet6_sk_rebuild_header(struct sock
*sk
)
708 struct ipv6_pinfo
*np
= inet6_sk(sk
);
709 struct dst_entry
*dst
;
711 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
714 struct inet_sock
*inet
= inet_sk(sk
);
715 struct in6_addr
*final_p
, final
;
718 memset(&fl6
, 0, sizeof(fl6
));
719 fl6
.flowi6_proto
= sk
->sk_protocol
;
720 fl6
.daddr
= sk
->sk_v6_daddr
;
721 fl6
.saddr
= np
->saddr
;
722 fl6
.flowlabel
= np
->flow_label
;
723 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
724 fl6
.flowi6_mark
= sk
->sk_mark
;
725 fl6
.fl6_dport
= inet
->inet_dport
;
726 fl6
.fl6_sport
= inet
->inet_sport
;
727 fl6
.flowi6_uid
= sk
->sk_uid
;
728 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
731 final_p
= fl6_update_dst(&fl6
, rcu_dereference(np
->opt
),
735 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
737 sk
->sk_route_caps
= 0;
738 sk
->sk_err_soft
= -PTR_ERR(dst
);
742 ip6_dst_store(sk
, dst
, NULL
, NULL
);
747 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header
);
749 bool ipv6_opt_accepted(const struct sock
*sk
, const struct sk_buff
*skb
,
750 const struct inet6_skb_parm
*opt
)
752 const struct ipv6_pinfo
*np
= inet6_sk(sk
);
755 if (((opt
->flags
& IP6SKB_HOPBYHOP
) &&
756 (np
->rxopt
.bits
.hopopts
|| np
->rxopt
.bits
.ohopopts
)) ||
757 (ip6_flowinfo((struct ipv6hdr
*) skb_network_header(skb
)) &&
758 np
->rxopt
.bits
.rxflow
) ||
759 (opt
->srcrt
&& (np
->rxopt
.bits
.srcrt
||
760 np
->rxopt
.bits
.osrcrt
)) ||
761 ((opt
->dst1
|| opt
->dst0
) &&
762 (np
->rxopt
.bits
.dstopts
|| np
->rxopt
.bits
.odstopts
)))
767 EXPORT_SYMBOL_GPL(ipv6_opt_accepted
);
769 static struct packet_type ipv6_packet_type __read_mostly
= {
770 .type
= cpu_to_be16(ETH_P_IPV6
),
772 .list_func
= ipv6_list_rcv
,
775 static int __init
ipv6_packet_init(void)
777 dev_add_pack(&ipv6_packet_type
);
781 static void ipv6_packet_cleanup(void)
783 dev_remove_pack(&ipv6_packet_type
);
786 static int __net_init
ipv6_init_mibs(struct net
*net
)
790 net
->mib
.udp_stats_in6
= alloc_percpu(struct udp_mib
);
791 if (!net
->mib
.udp_stats_in6
)
793 net
->mib
.udplite_stats_in6
= alloc_percpu(struct udp_mib
);
794 if (!net
->mib
.udplite_stats_in6
)
795 goto err_udplite_mib
;
796 net
->mib
.ipv6_statistics
= alloc_percpu(struct ipstats_mib
);
797 if (!net
->mib
.ipv6_statistics
)
800 for_each_possible_cpu(i
) {
801 struct ipstats_mib
*af_inet6_stats
;
802 af_inet6_stats
= per_cpu_ptr(net
->mib
.ipv6_statistics
, i
);
803 u64_stats_init(&af_inet6_stats
->syncp
);
807 net
->mib
.icmpv6_statistics
= alloc_percpu(struct icmpv6_mib
);
808 if (!net
->mib
.icmpv6_statistics
)
810 net
->mib
.icmpv6msg_statistics
= kzalloc(sizeof(struct icmpv6msg_mib
),
812 if (!net
->mib
.icmpv6msg_statistics
)
813 goto err_icmpmsg_mib
;
817 free_percpu(net
->mib
.icmpv6_statistics
);
819 free_percpu(net
->mib
.ipv6_statistics
);
821 free_percpu(net
->mib
.udplite_stats_in6
);
823 free_percpu(net
->mib
.udp_stats_in6
);
827 static void ipv6_cleanup_mibs(struct net
*net
)
829 free_percpu(net
->mib
.udp_stats_in6
);
830 free_percpu(net
->mib
.udplite_stats_in6
);
831 free_percpu(net
->mib
.ipv6_statistics
);
832 free_percpu(net
->mib
.icmpv6_statistics
);
833 kfree(net
->mib
.icmpv6msg_statistics
);
836 static int __net_init
inet6_net_init(struct net
*net
)
840 net
->ipv6
.sysctl
.bindv6only
= 0;
841 net
->ipv6
.sysctl
.icmpv6_time
= 1*HZ
;
842 net
->ipv6
.sysctl
.icmpv6_echo_ignore_all
= 0;
843 net
->ipv6
.sysctl
.icmpv6_echo_ignore_multicast
= 0;
844 net
->ipv6
.sysctl
.icmpv6_echo_ignore_anycast
= 0;
846 /* By default, rate limit error messages.
847 * Except for pmtu discovery, it would break it.
848 * proc_do_large_bitmap needs pointer to the bitmap.
850 bitmap_set(net
->ipv6
.sysctl
.icmpv6_ratemask
, 0, ICMPV6_ERRMSG_MAX
+ 1);
851 bitmap_clear(net
->ipv6
.sysctl
.icmpv6_ratemask
, ICMPV6_PKT_TOOBIG
, 1);
852 net
->ipv6
.sysctl
.icmpv6_ratemask_ptr
= net
->ipv6
.sysctl
.icmpv6_ratemask
;
854 net
->ipv6
.sysctl
.flowlabel_consistency
= 1;
855 net
->ipv6
.sysctl
.auto_flowlabels
= IP6_DEFAULT_AUTO_FLOW_LABELS
;
856 net
->ipv6
.sysctl
.idgen_retries
= 3;
857 net
->ipv6
.sysctl
.idgen_delay
= 1 * HZ
;
858 net
->ipv6
.sysctl
.flowlabel_state_ranges
= 0;
859 net
->ipv6
.sysctl
.max_dst_opts_cnt
= IP6_DEFAULT_MAX_DST_OPTS_CNT
;
860 net
->ipv6
.sysctl
.max_hbh_opts_cnt
= IP6_DEFAULT_MAX_HBH_OPTS_CNT
;
861 net
->ipv6
.sysctl
.max_dst_opts_len
= IP6_DEFAULT_MAX_DST_OPTS_LEN
;
862 net
->ipv6
.sysctl
.max_hbh_opts_len
= IP6_DEFAULT_MAX_HBH_OPTS_LEN
;
863 atomic_set(&net
->ipv6
.fib6_sernum
, 1);
865 err
= ipv6_init_mibs(net
);
868 #ifdef CONFIG_PROC_FS
869 err
= udp6_proc_init(net
);
872 err
= tcp6_proc_init(net
);
875 err
= ac6_proc_init(net
);
881 #ifdef CONFIG_PROC_FS
887 ipv6_cleanup_mibs(net
);
892 static void __net_exit
inet6_net_exit(struct net
*net
)
894 #ifdef CONFIG_PROC_FS
899 ipv6_cleanup_mibs(net
);
902 static struct pernet_operations inet6_net_ops
= {
903 .init
= inet6_net_init
,
904 .exit
= inet6_net_exit
,
907 static int ipv6_route_input(struct sk_buff
*skb
)
909 ip6_route_input(skb
);
910 return skb_dst(skb
)->error
;
913 static const struct ipv6_stub ipv6_stub_impl
= {
914 .ipv6_sock_mc_join
= ipv6_sock_mc_join
,
915 .ipv6_sock_mc_drop
= ipv6_sock_mc_drop
,
916 .ipv6_dst_lookup
= ip6_dst_lookup
,
917 .ipv6_route_input
= ipv6_route_input
,
918 .fib6_get_table
= fib6_get_table
,
919 .fib6_table_lookup
= fib6_table_lookup
,
920 .fib6_lookup
= fib6_lookup
,
921 .fib6_select_path
= fib6_select_path
,
922 .ip6_mtu_from_fib6
= ip6_mtu_from_fib6
,
923 .fib6_nh_init
= fib6_nh_init
,
924 .fib6_nh_release
= fib6_nh_release
,
925 .udpv6_encap_enable
= udpv6_encap_enable
,
926 .ndisc_send_na
= ndisc_send_na
,
930 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl
= {
931 .inet6_bind
= __inet6_bind
,
932 .udp6_lib_lookup
= __udp6_lib_lookup
,
935 static int __init
inet6_init(void)
940 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm
));
942 /* Register the socket-side information for inet6_create. */
943 for (r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
946 if (disable_ipv6_mod
) {
947 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
951 err
= proto_register(&tcpv6_prot
, 1);
955 err
= proto_register(&udpv6_prot
, 1);
957 goto out_unregister_tcp_proto
;
959 err
= proto_register(&udplitev6_prot
, 1);
961 goto out_unregister_udp_proto
;
963 err
= proto_register(&rawv6_prot
, 1);
965 goto out_unregister_udplite_proto
;
967 err
= proto_register(&pingv6_prot
, 1);
969 goto out_unregister_raw_proto
;
971 /* We MUST register RAW sockets before we create the ICMP6,
972 * IGMP6, or NDISC control sockets.
976 goto out_unregister_ping_proto
;
978 /* Register the family here so that the init calls below will
979 * be able to create sockets. (?? is this dangerous ??)
981 err
= sock_register(&inet6_family_ops
);
983 goto out_sock_register_fail
;
986 * ipngwg API draft makes clear that the correct semantics
987 * for TCP and UDP is to consider one TCP and UDP instance
988 * in a host available by both INET and INET6 APIs and
989 * able to communicate via both network protocols.
992 err
= register_pernet_subsys(&inet6_net_ops
);
994 goto register_pernet_fail
;
1008 err
= ipv6_netfilter_init();
1010 goto netfilter_fail
;
1011 /* Create /proc/foo6 entries. */
1012 #ifdef CONFIG_PROC_FS
1014 if (raw6_proc_init())
1015 goto proc_raw6_fail
;
1016 if (udplite6_proc_init())
1017 goto proc_udplite6_fail
;
1018 if (ipv6_misc_proc_init())
1019 goto proc_misc6_fail
;
1020 if (if6_proc_init())
1023 err
= ip6_route_init();
1025 goto ip6_route_fail
;
1026 err
= ndisc_late_init();
1028 goto ndisc_late_fail
;
1029 err
= ip6_flowlabel_init();
1031 goto ip6_flowlabel_fail
;
1032 err
= ipv6_anycast_init();
1034 goto ipv6_anycast_fail
;
1035 err
= addrconf_init();
1039 /* Init v6 extension headers. */
1040 err
= ipv6_exthdrs_init();
1042 goto ipv6_exthdrs_fail
;
1044 err
= ipv6_frag_init();
1046 goto ipv6_frag_fail
;
1048 /* Init v6 transport protocols. */
1053 err
= udplitev6_init();
1055 goto udplitev6_fail
;
1057 err
= udpv6_offload_init();
1059 goto udpv6_offload_fail
;
1065 err
= ipv6_packet_init();
1067 goto ipv6_packet_fail
;
1069 err
= pingv6_init();
1073 err
= calipso_init();
1081 err
= igmp6_late_init();
1083 goto igmp6_late_err
;
1085 #ifdef CONFIG_SYSCTL
1086 err
= ipv6_sysctl_register();
1091 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1093 ipv6_stub
= &ipv6_stub_impl
;
1094 ipv6_bpf_stub
= &ipv6_bpf_stub_impl
;
1098 #ifdef CONFIG_SYSCTL
1100 igmp6_late_cleanup();
1109 ipv6_packet_cleanup();
1113 udpv6_offload_exit();
1121 ipv6_exthdrs_exit();
1125 ipv6_anycast_cleanup();
1127 ip6_flowlabel_cleanup();
1129 ndisc_late_cleanup();
1131 ip6_route_cleanup();
1133 #ifdef CONFIG_PROC_FS
1136 ipv6_misc_proc_exit();
1138 udplite6_proc_exit();
1143 ipv6_netfilter_fini();
1153 unregister_pernet_subsys(&inet6_net_ops
);
1154 register_pernet_fail
:
1155 sock_unregister(PF_INET6
);
1156 rtnl_unregister_all(PF_INET6
);
1157 out_sock_register_fail
:
1159 out_unregister_ping_proto
:
1160 proto_unregister(&pingv6_prot
);
1161 out_unregister_raw_proto
:
1162 proto_unregister(&rawv6_prot
);
1163 out_unregister_udplite_proto
:
1164 proto_unregister(&udplitev6_prot
);
1165 out_unregister_udp_proto
:
1166 proto_unregister(&udpv6_prot
);
1167 out_unregister_tcp_proto
:
1168 proto_unregister(&tcpv6_prot
);
1171 module_init(inet6_init
);
1173 MODULE_ALIAS_NETPROTO(PF_INET6
);