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 <net/compat.h>
65 #include <net/ioam6.h>
66 #include <net/rawv6.h>
69 #include <linux/uaccess.h>
70 #include <linux/mroute6.h>
72 #include "ip6_offload.h"
74 MODULE_AUTHOR("Cast of dozens");
75 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
76 MODULE_LICENSE("GPL");
78 /* The inetsw6 table contains everything that inet6_create needs to
81 static struct list_head inetsw6
[SOCK_MAX
];
82 static DEFINE_SPINLOCK(inetsw6_lock
);
84 struct ipv6_params ipv6_defaults
= {
89 static int disable_ipv6_mod
;
91 module_param_named(disable
, disable_ipv6_mod
, int, 0444);
92 MODULE_PARM_DESC(disable
, "Disable IPv6 module such that it is non-functional");
94 module_param_named(disable_ipv6
, ipv6_defaults
.disable_ipv6
, int, 0444);
95 MODULE_PARM_DESC(disable_ipv6
, "Disable IPv6 on all interfaces");
97 module_param_named(autoconf
, ipv6_defaults
.autoconf
, int, 0444);
98 MODULE_PARM_DESC(autoconf
, "Enable IPv6 address autoconfiguration on all interfaces");
100 bool ipv6_mod_enabled(void)
102 return disable_ipv6_mod
== 0;
104 EXPORT_SYMBOL_GPL(ipv6_mod_enabled
);
106 static struct ipv6_pinfo
*inet6_sk_generic(struct sock
*sk
)
108 const int offset
= sk
->sk_prot
->ipv6_pinfo_offset
;
110 return (struct ipv6_pinfo
*)(((u8
*)sk
) + offset
);
113 void inet6_sock_destruct(struct sock
*sk
)
115 inet6_cleanup_sock(sk
);
116 inet_sock_destruct(sk
);
118 EXPORT_SYMBOL_GPL(inet6_sock_destruct
);
120 static int inet6_create(struct net
*net
, struct socket
*sock
, int protocol
,
123 struct inet_sock
*inet
;
124 struct ipv6_pinfo
*np
;
126 struct inet_protosw
*answer
;
127 struct proto
*answer_prot
;
128 unsigned char answer_flags
;
129 int try_loading_module
= 0;
132 if (protocol
< 0 || protocol
>= IPPROTO_MAX
)
135 /* Look for the requested type/protocol pair. */
137 err
= -ESOCKTNOSUPPORT
;
139 list_for_each_entry_rcu(answer
, &inetsw6
[sock
->type
], list
) {
142 /* Check the non-wild match. */
143 if (protocol
== answer
->protocol
) {
144 if (protocol
!= IPPROTO_IP
)
147 /* Check for the two wild cases. */
148 if (IPPROTO_IP
== protocol
) {
149 protocol
= answer
->protocol
;
152 if (IPPROTO_IP
== answer
->protocol
)
155 err
= -EPROTONOSUPPORT
;
159 if (try_loading_module
< 2) {
162 * Be more specific, e.g. net-pf-10-proto-132-type-1
163 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
165 if (++try_loading_module
== 1)
166 request_module("net-pf-%d-proto-%d-type-%d",
167 PF_INET6
, protocol
, sock
->type
);
169 * Fall back to generic, e.g. net-pf-10-proto-132
170 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
173 request_module("net-pf-%d-proto-%d",
175 goto lookup_protocol
;
181 if (sock
->type
== SOCK_RAW
&& !kern
&&
182 !ns_capable(net
->user_ns
, CAP_NET_RAW
))
185 sock
->ops
= answer
->ops
;
186 answer_prot
= answer
->prot
;
187 answer_flags
= answer
->flags
;
190 WARN_ON(!answer_prot
->slab
);
193 sk
= sk_alloc(net
, PF_INET6
, GFP_KERNEL
, answer_prot
, kern
);
197 sock_init_data(sock
, sk
);
200 if (INET_PROTOSW_REUSE
& answer_flags
)
201 sk
->sk_reuse
= SK_CAN_REUSE
;
203 if (INET_PROTOSW_ICSK
& answer_flags
)
204 inet_init_csk_locks(sk
);
207 inet_assign_bit(IS_ICSK
, sk
, INET_PROTOSW_ICSK
& answer_flags
);
209 if (SOCK_RAW
== sock
->type
) {
210 inet
->inet_num
= protocol
;
211 if (IPPROTO_RAW
== protocol
)
212 inet_set_bit(HDRINCL
, sk
);
215 sk
->sk_destruct
= inet6_sock_destruct
;
216 sk
->sk_family
= PF_INET6
;
217 sk
->sk_protocol
= protocol
;
219 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
221 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
223 np
->mcast_hops
= IPV6_DEFAULT_MCASTHOPS
;
224 inet6_set_bit(MC6_LOOP
, sk
);
225 inet6_set_bit(MC6_ALL
, sk
);
226 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
227 inet6_assign_bit(REPFLOW
, sk
, net
->ipv6
.sysctl
.flowlabel_reflect
&
228 FLOWLABEL_REFLECT_ESTABLISHED
);
229 sk
->sk_ipv6only
= net
->ipv6
.sysctl
.bindv6only
;
230 sk
->sk_txrehash
= READ_ONCE(net
->core
.sysctl_txrehash
);
232 /* Init the ipv4 part of the socket since we can have sockets
233 * using v6 API for ipv4.
237 inet_set_bit(MC_LOOP
, sk
);
240 RCU_INIT_POINTER(inet
->mc_list
, NULL
);
243 if (READ_ONCE(net
->ipv4
.sysctl_ip_no_pmtu_disc
))
244 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
246 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
248 if (inet
->inet_num
) {
249 /* It assumes that any protocol which allows
250 * the user to assign a number at socket
251 * creation time automatically shares.
253 inet
->inet_sport
= htons(inet
->inet_num
);
254 err
= sk
->sk_prot
->hash(sk
);
258 if (sk
->sk_prot
->init
) {
259 err
= sk
->sk_prot
->init(sk
);
265 err
= BPF_CGROUP_RUN_PROG_INET_SOCK(sk
);
275 sk_common_release(sk
);
280 static int __inet6_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
,
283 struct sockaddr_in6
*addr
= (struct sockaddr_in6
*)uaddr
;
284 struct inet_sock
*inet
= inet_sk(sk
);
285 struct ipv6_pinfo
*np
= inet6_sk(sk
);
286 struct net
*net
= sock_net(sk
);
293 if (addr
->sin6_family
!= AF_INET6
)
294 return -EAFNOSUPPORT
;
296 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
297 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sk
->sk_type
== SOCK_STREAM
)
300 snum
= ntohs(addr
->sin6_port
);
301 if (!(flags
& BIND_NO_CAP_NET_BIND_SERVICE
) &&
302 snum
&& inet_port_requires_bind_service(net
, snum
) &&
303 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
306 if (flags
& BIND_WITH_LOCK
)
309 /* Check these errors (active socket, double bind). */
310 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->inet_num
) {
315 /* Check if the address belongs to the host. */
316 if (addr_type
== IPV6_ADDR_MAPPED
) {
317 struct net_device
*dev
= NULL
;
320 /* Binding to v4-mapped address on a v6-only socket
323 if (ipv6_only_sock(sk
)) {
329 if (sk
->sk_bound_dev_if
) {
330 dev
= dev_get_by_index_rcu(net
, sk
->sk_bound_dev_if
);
337 /* Reproduce AF_INET checks to make the bindings consistent */
338 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
339 chk_addr_ret
= inet_addr_type_dev_table(net
, dev
, v4addr
);
342 if (!inet_addr_valid_or_nonlocal(net
, inet
, v4addr
,
344 err
= -EADDRNOTAVAIL
;
348 if (addr_type
!= IPV6_ADDR_ANY
) {
349 struct net_device
*dev
= NULL
;
352 if (__ipv6_addr_needs_scope_id(addr_type
)) {
353 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
354 addr
->sin6_scope_id
) {
355 /* Override any existing binding, if another one
356 * is supplied by user.
358 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
361 /* Binding to link-local address requires an interface */
362 if (!sk
->sk_bound_dev_if
) {
368 if (sk
->sk_bound_dev_if
) {
369 dev
= dev_get_by_index_rcu(net
, sk
->sk_bound_dev_if
);
376 /* ipv4 addr of the socket is invalid. Only the
377 * unspecified and mapped address have a v4 equivalent.
379 v4addr
= LOOPBACK4_IPV6
;
380 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
381 if (!ipv6_can_nonlocal_bind(net
, inet
) &&
382 !ipv6_chk_addr(net
, &addr
->sin6_addr
,
384 err
= -EADDRNOTAVAIL
;
392 inet
->inet_rcv_saddr
= v4addr
;
393 inet
->inet_saddr
= v4addr
;
395 sk
->sk_v6_rcv_saddr
= addr
->sin6_addr
;
397 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
398 np
->saddr
= addr
->sin6_addr
;
400 saved_ipv6only
= sk
->sk_ipv6only
;
401 if (addr_type
!= IPV6_ADDR_ANY
&& addr_type
!= IPV6_ADDR_MAPPED
)
404 /* Make sure we are allowed to bind here. */
405 if (snum
|| !(inet_test_bit(BIND_ADDRESS_NO_PORT
, sk
) ||
406 (flags
& BIND_FORCE_ADDRESS_NO_PORT
))) {
407 err
= sk
->sk_prot
->get_port(sk
, snum
);
409 sk
->sk_ipv6only
= saved_ipv6only
;
410 inet_reset_saddr(sk
);
413 if (!(flags
& BIND_FROM_BPF
)) {
414 err
= BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk
);
416 sk
->sk_ipv6only
= saved_ipv6only
;
417 inet_reset_saddr(sk
);
418 if (sk
->sk_prot
->put_port
)
419 sk
->sk_prot
->put_port(sk
);
425 if (addr_type
!= IPV6_ADDR_ANY
)
426 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
428 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
429 inet
->inet_sport
= htons(inet
->inet_num
);
430 inet
->inet_dport
= 0;
431 inet
->inet_daddr
= 0;
433 if (flags
& BIND_WITH_LOCK
)
441 int inet6_bind_sk(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
443 u32 flags
= BIND_WITH_LOCK
;
444 const struct proto
*prot
;
447 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
448 prot
= READ_ONCE(sk
->sk_prot
);
449 /* If the socket has its own bind function then use it. */
451 return prot
->bind(sk
, uaddr
, addr_len
);
453 if (addr_len
< SIN6_LEN_RFC2133
)
456 /* BPF prog is run before any checks are done so that if the prog
457 * changes context in a wrong way it will be caught.
459 err
= BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk
, uaddr
, &addr_len
,
460 CGROUP_INET6_BIND
, &flags
);
464 return __inet6_bind(sk
, uaddr
, addr_len
, flags
);
467 /* bind for INET6 API */
468 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
470 return inet6_bind_sk(sock
->sk
, uaddr
, addr_len
);
472 EXPORT_SYMBOL(inet6_bind
);
474 int inet6_release(struct socket
*sock
)
476 struct sock
*sk
= sock
->sk
;
482 ipv6_sock_mc_close(sk
);
485 ipv6_sock_ac_close(sk
);
487 return inet_release(sock
);
489 EXPORT_SYMBOL(inet6_release
);
491 void inet6_cleanup_sock(struct sock
*sk
)
493 struct ipv6_pinfo
*np
= inet6_sk(sk
);
495 struct ipv6_txoptions
*opt
;
497 /* Release rx options */
499 skb
= xchg(&np
->pktoptions
, NULL
);
502 skb
= xchg(&np
->rxpmtu
, NULL
);
505 /* Free flowlabels */
506 fl6_free_socklist(sk
);
508 /* Free tx options */
510 opt
= unrcu_pointer(xchg(&np
->opt
, NULL
));
512 atomic_sub(opt
->tot_len
, &sk
->sk_omem_alloc
);
516 EXPORT_SYMBOL_GPL(inet6_cleanup_sock
);
519 * This does both peername and sockname.
521 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
524 struct sockaddr_in6
*sin
= (struct sockaddr_in6
*)uaddr
;
525 int sin_addr_len
= sizeof(*sin
);
526 struct sock
*sk
= sock
->sk
;
527 struct inet_sock
*inet
= inet_sk(sk
);
528 struct ipv6_pinfo
*np
= inet6_sk(sk
);
530 sin
->sin6_family
= AF_INET6
;
531 sin
->sin6_flowinfo
= 0;
532 sin
->sin6_scope_id
= 0;
535 if (!inet
->inet_dport
||
536 (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
541 sin
->sin6_port
= inet
->inet_dport
;
542 sin
->sin6_addr
= sk
->sk_v6_daddr
;
543 if (inet6_test_bit(SNDFLOW
, sk
))
544 sin
->sin6_flowinfo
= np
->flow_label
;
545 BPF_CGROUP_RUN_SA_PROG(sk
, (struct sockaddr
*)sin
, &sin_addr_len
,
546 CGROUP_INET6_GETPEERNAME
);
548 if (ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
549 sin
->sin6_addr
= np
->saddr
;
551 sin
->sin6_addr
= sk
->sk_v6_rcv_saddr
;
552 sin
->sin6_port
= inet
->inet_sport
;
553 BPF_CGROUP_RUN_SA_PROG(sk
, (struct sockaddr
*)sin
, &sin_addr_len
,
554 CGROUP_INET6_GETSOCKNAME
);
556 sin
->sin6_scope_id
= ipv6_iface_scope_id(&sin
->sin6_addr
,
557 sk
->sk_bound_dev_if
);
561 EXPORT_SYMBOL(inet6_getname
);
563 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
565 void __user
*argp
= (void __user
*)arg
;
566 struct sock
*sk
= sock
->sk
;
567 struct net
*net
= sock_net(sk
);
568 const struct proto
*prot
;
573 struct in6_rtmsg rtmsg
;
575 if (copy_from_user(&rtmsg
, argp
, sizeof(rtmsg
)))
577 return ipv6_route_ioctl(net
, cmd
, &rtmsg
);
580 return addrconf_add_ifaddr(net
, argp
);
582 return addrconf_del_ifaddr(net
, argp
);
584 return addrconf_set_dstaddr(net
, argp
);
586 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
587 prot
= READ_ONCE(sk
->sk_prot
);
590 return sk_ioctl(sk
, cmd
, (void __user
*)arg
);
595 EXPORT_SYMBOL(inet6_ioctl
);
598 struct compat_in6_rtmsg
{
599 struct in6_addr rtmsg_dst
;
600 struct in6_addr rtmsg_src
;
601 struct in6_addr rtmsg_gateway
;
611 static int inet6_compat_routing_ioctl(struct sock
*sk
, unsigned int cmd
,
612 struct compat_in6_rtmsg __user
*ur
)
616 if (copy_from_user(&rt
.rtmsg_dst
, &ur
->rtmsg_dst
,
617 3 * sizeof(struct in6_addr
)) ||
618 get_user(rt
.rtmsg_type
, &ur
->rtmsg_type
) ||
619 get_user(rt
.rtmsg_dst_len
, &ur
->rtmsg_dst_len
) ||
620 get_user(rt
.rtmsg_src_len
, &ur
->rtmsg_src_len
) ||
621 get_user(rt
.rtmsg_metric
, &ur
->rtmsg_metric
) ||
622 get_user(rt
.rtmsg_info
, &ur
->rtmsg_info
) ||
623 get_user(rt
.rtmsg_flags
, &ur
->rtmsg_flags
) ||
624 get_user(rt
.rtmsg_ifindex
, &ur
->rtmsg_ifindex
))
628 return ipv6_route_ioctl(sock_net(sk
), cmd
, &rt
);
631 int inet6_compat_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
633 void __user
*argp
= compat_ptr(arg
);
634 struct sock
*sk
= sock
->sk
;
639 return inet6_compat_routing_ioctl(sk
, cmd
, argp
);
644 EXPORT_SYMBOL_GPL(inet6_compat_ioctl
);
645 #endif /* CONFIG_COMPAT */
647 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock
*, struct msghdr
*,
649 int inet6_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
651 struct sock
*sk
= sock
->sk
;
652 const struct proto
*prot
;
654 if (unlikely(inet_send_prepare(sk
)))
657 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
658 prot
= READ_ONCE(sk
->sk_prot
);
659 return INDIRECT_CALL_2(prot
->sendmsg
, tcp_sendmsg
, udpv6_sendmsg
,
663 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock
*, struct msghdr
*,
664 size_t, int, int *));
665 int inet6_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
668 struct sock
*sk
= sock
->sk
;
669 const struct proto
*prot
;
673 if (likely(!(flags
& MSG_ERRQUEUE
)))
674 sock_rps_record_flow(sk
);
676 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
677 prot
= READ_ONCE(sk
->sk_prot
);
678 err
= INDIRECT_CALL_2(prot
->recvmsg
, tcp_recvmsg
, udpv6_recvmsg
,
679 sk
, msg
, size
, flags
, &addr_len
);
681 msg
->msg_namelen
= addr_len
;
685 const struct proto_ops inet6_stream_ops
= {
687 .owner
= THIS_MODULE
,
688 .release
= inet6_release
,
690 .connect
= inet_stream_connect
, /* ok */
691 .socketpair
= sock_no_socketpair
, /* a do nothing */
692 .accept
= inet_accept
, /* ok */
693 .getname
= inet6_getname
,
694 .poll
= tcp_poll
, /* ok */
695 .ioctl
= inet6_ioctl
, /* must change */
696 .gettstamp
= sock_gettstamp
,
697 .listen
= inet_listen
, /* ok */
698 .shutdown
= inet_shutdown
, /* ok */
699 .setsockopt
= sock_common_setsockopt
, /* ok */
700 .getsockopt
= sock_common_getsockopt
, /* ok */
701 .sendmsg
= inet6_sendmsg
, /* retpoline's sake */
702 .recvmsg
= inet6_recvmsg
, /* retpoline's sake */
706 .splice_eof
= inet_splice_eof
,
707 .sendmsg_locked
= tcp_sendmsg_locked
,
708 .splice_read
= tcp_splice_read
,
709 .set_peek_off
= sk_set_peek_off
,
710 .read_sock
= tcp_read_sock
,
711 .read_skb
= tcp_read_skb
,
712 .peek_len
= tcp_peek_len
,
714 .compat_ioctl
= inet6_compat_ioctl
,
716 .set_rcvlowat
= tcp_set_rcvlowat
,
719 const struct proto_ops inet6_dgram_ops
= {
721 .owner
= THIS_MODULE
,
722 .release
= inet6_release
,
724 .connect
= inet_dgram_connect
, /* ok */
725 .socketpair
= sock_no_socketpair
, /* a do nothing */
726 .accept
= sock_no_accept
, /* a do nothing */
727 .getname
= inet6_getname
,
728 .poll
= udp_poll
, /* ok */
729 .ioctl
= inet6_ioctl
, /* must change */
730 .gettstamp
= sock_gettstamp
,
731 .listen
= sock_no_listen
, /* ok */
732 .shutdown
= inet_shutdown
, /* ok */
733 .setsockopt
= sock_common_setsockopt
, /* ok */
734 .getsockopt
= sock_common_getsockopt
, /* ok */
735 .sendmsg
= inet6_sendmsg
, /* retpoline's sake */
736 .recvmsg
= inet6_recvmsg
, /* retpoline's sake */
737 .read_skb
= udp_read_skb
,
738 .mmap
= sock_no_mmap
,
739 .set_peek_off
= udp_set_peek_off
,
741 .compat_ioctl
= inet6_compat_ioctl
,
745 static const struct net_proto_family inet6_family_ops
= {
747 .create
= inet6_create
,
748 .owner
= THIS_MODULE
,
751 int inet6_register_protosw(struct inet_protosw
*p
)
753 struct list_head
*lh
;
754 struct inet_protosw
*answer
;
755 struct list_head
*last_perm
;
756 int protocol
= p
->protocol
;
759 spin_lock_bh(&inetsw6_lock
);
762 if (p
->type
>= SOCK_MAX
)
765 /* If we are trying to override a permanent protocol, bail. */
768 last_perm
= &inetsw6
[p
->type
];
769 list_for_each(lh
, &inetsw6
[p
->type
]) {
770 answer
= list_entry(lh
, struct inet_protosw
, list
);
772 /* Check only the non-wild match. */
773 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
774 if (protocol
== answer
->protocol
)
784 /* Add the new entry after the last permanent entry if any, so that
785 * the new entry does not override a permanent entry when matched with
786 * a wild-card protocol. But it is allowed to override any existing
787 * non-permanent entry. This means that when we remove this entry, the
788 * system automatically returns to the old behavior.
790 list_add_rcu(&p
->list
, last_perm
);
793 spin_unlock_bh(&inetsw6_lock
);
797 pr_err("Attempt to override permanent protocol %d\n", protocol
);
801 pr_err("Ignoring attempt to register invalid socket type %d\n",
805 EXPORT_SYMBOL(inet6_register_protosw
);
808 inet6_unregister_protosw(struct inet_protosw
*p
)
810 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
811 pr_err("Attempt to unregister permanent protocol %d\n",
814 spin_lock_bh(&inetsw6_lock
);
815 list_del_rcu(&p
->list
);
816 spin_unlock_bh(&inetsw6_lock
);
821 EXPORT_SYMBOL(inet6_unregister_protosw
);
823 int inet6_sk_rebuild_header(struct sock
*sk
)
825 struct ipv6_pinfo
*np
= inet6_sk(sk
);
826 struct dst_entry
*dst
;
828 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
831 struct inet_sock
*inet
= inet_sk(sk
);
832 struct in6_addr
*final_p
, final
;
835 memset(&fl6
, 0, sizeof(fl6
));
836 fl6
.flowi6_proto
= sk
->sk_protocol
;
837 fl6
.daddr
= sk
->sk_v6_daddr
;
838 fl6
.saddr
= np
->saddr
;
839 fl6
.flowlabel
= np
->flow_label
;
840 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
841 fl6
.flowi6_mark
= sk
->sk_mark
;
842 fl6
.fl6_dport
= inet
->inet_dport
;
843 fl6
.fl6_sport
= inet
->inet_sport
;
844 fl6
.flowi6_uid
= sk
->sk_uid
;
845 security_sk_classify_flow(sk
, flowi6_to_flowi_common(&fl6
));
848 final_p
= fl6_update_dst(&fl6
, rcu_dereference(np
->opt
),
852 dst
= ip6_dst_lookup_flow(sock_net(sk
), sk
, &fl6
, final_p
);
854 sk
->sk_route_caps
= 0;
855 WRITE_ONCE(sk
->sk_err_soft
, -PTR_ERR(dst
));
859 ip6_dst_store(sk
, dst
, NULL
, NULL
);
864 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header
);
866 bool ipv6_opt_accepted(const struct sock
*sk
, const struct sk_buff
*skb
,
867 const struct inet6_skb_parm
*opt
)
869 const struct ipv6_pinfo
*np
= inet6_sk(sk
);
872 if (((opt
->flags
& IP6SKB_HOPBYHOP
) &&
873 (np
->rxopt
.bits
.hopopts
|| np
->rxopt
.bits
.ohopopts
)) ||
874 (ip6_flowinfo((struct ipv6hdr
*) skb_network_header(skb
)) &&
875 np
->rxopt
.bits
.rxflow
) ||
876 (opt
->srcrt
&& (np
->rxopt
.bits
.srcrt
||
877 np
->rxopt
.bits
.osrcrt
)) ||
878 ((opt
->dst1
|| opt
->dst0
) &&
879 (np
->rxopt
.bits
.dstopts
|| np
->rxopt
.bits
.odstopts
)))
884 EXPORT_SYMBOL_GPL(ipv6_opt_accepted
);
886 static struct packet_type ipv6_packet_type __read_mostly
= {
887 .type
= cpu_to_be16(ETH_P_IPV6
),
889 .list_func
= ipv6_list_rcv
,
892 static int __init
ipv6_packet_init(void)
894 dev_add_pack(&ipv6_packet_type
);
898 static void ipv6_packet_cleanup(void)
900 dev_remove_pack(&ipv6_packet_type
);
903 static int __net_init
ipv6_init_mibs(struct net
*net
)
907 net
->mib
.udp_stats_in6
= alloc_percpu(struct udp_mib
);
908 if (!net
->mib
.udp_stats_in6
)
910 net
->mib
.udplite_stats_in6
= alloc_percpu(struct udp_mib
);
911 if (!net
->mib
.udplite_stats_in6
)
912 goto err_udplite_mib
;
913 net
->mib
.ipv6_statistics
= alloc_percpu(struct ipstats_mib
);
914 if (!net
->mib
.ipv6_statistics
)
917 for_each_possible_cpu(i
) {
918 struct ipstats_mib
*af_inet6_stats
;
919 af_inet6_stats
= per_cpu_ptr(net
->mib
.ipv6_statistics
, i
);
920 u64_stats_init(&af_inet6_stats
->syncp
);
924 net
->mib
.icmpv6_statistics
= alloc_percpu(struct icmpv6_mib
);
925 if (!net
->mib
.icmpv6_statistics
)
927 net
->mib
.icmpv6msg_statistics
= kzalloc(sizeof(struct icmpv6msg_mib
),
929 if (!net
->mib
.icmpv6msg_statistics
)
930 goto err_icmpmsg_mib
;
934 free_percpu(net
->mib
.icmpv6_statistics
);
936 free_percpu(net
->mib
.ipv6_statistics
);
938 free_percpu(net
->mib
.udplite_stats_in6
);
940 free_percpu(net
->mib
.udp_stats_in6
);
944 static void ipv6_cleanup_mibs(struct net
*net
)
946 free_percpu(net
->mib
.udp_stats_in6
);
947 free_percpu(net
->mib
.udplite_stats_in6
);
948 free_percpu(net
->mib
.ipv6_statistics
);
949 free_percpu(net
->mib
.icmpv6_statistics
);
950 kfree(net
->mib
.icmpv6msg_statistics
);
953 static int __net_init
inet6_net_init(struct net
*net
)
957 net
->ipv6
.sysctl
.bindv6only
= 0;
958 net
->ipv6
.sysctl
.icmpv6_time
= 1*HZ
;
959 net
->ipv6
.sysctl
.icmpv6_echo_ignore_all
= 0;
960 net
->ipv6
.sysctl
.icmpv6_echo_ignore_multicast
= 0;
961 net
->ipv6
.sysctl
.icmpv6_echo_ignore_anycast
= 0;
962 net
->ipv6
.sysctl
.icmpv6_error_anycast_as_unicast
= 0;
964 /* By default, rate limit error messages.
965 * Except for pmtu discovery, it would break it.
966 * proc_do_large_bitmap needs pointer to the bitmap.
968 bitmap_set(net
->ipv6
.sysctl
.icmpv6_ratemask
, 0, ICMPV6_ERRMSG_MAX
+ 1);
969 bitmap_clear(net
->ipv6
.sysctl
.icmpv6_ratemask
, ICMPV6_PKT_TOOBIG
, 1);
970 net
->ipv6
.sysctl
.icmpv6_ratemask_ptr
= net
->ipv6
.sysctl
.icmpv6_ratemask
;
972 net
->ipv6
.sysctl
.flowlabel_consistency
= 1;
973 net
->ipv6
.sysctl
.auto_flowlabels
= IP6_DEFAULT_AUTO_FLOW_LABELS
;
974 net
->ipv6
.sysctl
.idgen_retries
= 3;
975 net
->ipv6
.sysctl
.idgen_delay
= 1 * HZ
;
976 net
->ipv6
.sysctl
.flowlabel_state_ranges
= 0;
977 net
->ipv6
.sysctl
.max_dst_opts_cnt
= IP6_DEFAULT_MAX_DST_OPTS_CNT
;
978 net
->ipv6
.sysctl
.max_hbh_opts_cnt
= IP6_DEFAULT_MAX_HBH_OPTS_CNT
;
979 net
->ipv6
.sysctl
.max_dst_opts_len
= IP6_DEFAULT_MAX_DST_OPTS_LEN
;
980 net
->ipv6
.sysctl
.max_hbh_opts_len
= IP6_DEFAULT_MAX_HBH_OPTS_LEN
;
981 net
->ipv6
.sysctl
.fib_notify_on_flag_change
= 0;
982 atomic_set(&net
->ipv6
.fib6_sernum
, 1);
984 net
->ipv6
.sysctl
.ioam6_id
= IOAM6_DEFAULT_ID
;
985 net
->ipv6
.sysctl
.ioam6_id_wide
= IOAM6_DEFAULT_ID_WIDE
;
987 err
= ipv6_init_mibs(net
);
990 #ifdef CONFIG_PROC_FS
991 err
= udp6_proc_init(net
);
994 err
= tcp6_proc_init(net
);
997 err
= ac6_proc_init(net
);
1003 #ifdef CONFIG_PROC_FS
1005 tcp6_proc_exit(net
);
1007 udp6_proc_exit(net
);
1009 ipv6_cleanup_mibs(net
);
1014 static void __net_exit
inet6_net_exit(struct net
*net
)
1016 #ifdef CONFIG_PROC_FS
1017 udp6_proc_exit(net
);
1018 tcp6_proc_exit(net
);
1021 ipv6_cleanup_mibs(net
);
1024 static struct pernet_operations inet6_net_ops
= {
1025 .init
= inet6_net_init
,
1026 .exit
= inet6_net_exit
,
1029 static int ipv6_route_input(struct sk_buff
*skb
)
1031 ip6_route_input(skb
);
1032 return skb_dst(skb
)->error
;
1035 static const struct ipv6_stub ipv6_stub_impl
= {
1036 .ipv6_sock_mc_join
= ipv6_sock_mc_join
,
1037 .ipv6_sock_mc_drop
= ipv6_sock_mc_drop
,
1038 .ipv6_dst_lookup_flow
= ip6_dst_lookup_flow
,
1039 .ipv6_route_input
= ipv6_route_input
,
1040 .fib6_get_table
= fib6_get_table
,
1041 .fib6_table_lookup
= fib6_table_lookup
,
1042 .fib6_lookup
= fib6_lookup
,
1043 .fib6_select_path
= fib6_select_path
,
1044 .ip6_mtu_from_fib6
= ip6_mtu_from_fib6
,
1045 .fib6_nh_init
= fib6_nh_init
,
1046 .fib6_nh_release
= fib6_nh_release
,
1047 .fib6_nh_release_dsts
= fib6_nh_release_dsts
,
1048 .fib6_update_sernum
= fib6_update_sernum_stub
,
1049 .fib6_rt_update
= fib6_rt_update
,
1050 .ip6_del_rt
= ip6_del_rt
,
1051 .udpv6_encap_enable
= udpv6_encap_enable
,
1052 .ndisc_send_na
= ndisc_send_na
,
1053 #if IS_ENABLED(CONFIG_XFRM)
1054 .xfrm6_local_rxpmtu
= xfrm6_local_rxpmtu
,
1055 .xfrm6_udp_encap_rcv
= xfrm6_udp_encap_rcv
,
1056 .xfrm6_gro_udp_encap_rcv
= xfrm6_gro_udp_encap_rcv
,
1057 .xfrm6_rcv_encap
= xfrm6_rcv_encap
,
1060 .ipv6_fragment
= ip6_fragment
,
1061 .ipv6_dev_find
= ipv6_dev_find
,
1062 .ip6_xmit
= ip6_xmit
,
1065 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl
= {
1066 .inet6_bind
= __inet6_bind
,
1067 .udp6_lib_lookup
= __udp6_lib_lookup
,
1068 .ipv6_setsockopt
= do_ipv6_setsockopt
,
1069 .ipv6_getsockopt
= do_ipv6_getsockopt
,
1070 .ipv6_dev_get_saddr
= ipv6_dev_get_saddr
,
1073 static int __init
inet6_init(void)
1075 struct list_head
*r
;
1078 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm
));
1080 /* Register the socket-side information for inet6_create. */
1081 for (r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
1084 raw_hashinfo_init(&raw_v6_hashinfo
);
1086 if (disable_ipv6_mod
) {
1087 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1091 err
= proto_register(&tcpv6_prot
, 1);
1095 err
= proto_register(&udpv6_prot
, 1);
1097 goto out_unregister_tcp_proto
;
1099 err
= proto_register(&udplitev6_prot
, 1);
1101 goto out_unregister_udp_proto
;
1103 err
= proto_register(&rawv6_prot
, 1);
1105 goto out_unregister_udplite_proto
;
1107 err
= proto_register(&pingv6_prot
, 1);
1109 goto out_unregister_raw_proto
;
1111 /* We MUST register RAW sockets before we create the ICMP6,
1112 * IGMP6, or NDISC control sockets.
1116 goto out_unregister_ping_proto
;
1118 /* Register the family here so that the init calls below will
1119 * be able to create sockets. (?? is this dangerous ??)
1121 err
= sock_register(&inet6_family_ops
);
1123 goto out_sock_register_fail
;
1126 * ipngwg API draft makes clear that the correct semantics
1127 * for TCP and UDP is to consider one TCP and UDP instance
1128 * in a host available by both INET and INET6 APIs and
1129 * able to communicate via both network protocols.
1132 err
= register_pernet_subsys(&inet6_net_ops
);
1134 goto register_pernet_fail
;
1135 err
= ip6_mr_init();
1138 err
= icmpv6_init();
1148 err
= ipv6_netfilter_init();
1150 goto netfilter_fail
;
1151 /* Create /proc/foo6 entries. */
1152 #ifdef CONFIG_PROC_FS
1154 if (raw6_proc_init())
1155 goto proc_raw6_fail
;
1156 if (udplite6_proc_init())
1157 goto proc_udplite6_fail
;
1158 if (ipv6_misc_proc_init())
1159 goto proc_misc6_fail
;
1160 if (if6_proc_init())
1163 err
= ip6_route_init();
1165 goto ip6_route_fail
;
1166 err
= ndisc_late_init();
1168 goto ndisc_late_fail
;
1169 err
= ip6_flowlabel_init();
1171 goto ip6_flowlabel_fail
;
1172 err
= ipv6_anycast_init();
1174 goto ipv6_anycast_fail
;
1175 err
= addrconf_init();
1179 /* Init v6 extension headers. */
1180 err
= ipv6_exthdrs_init();
1182 goto ipv6_exthdrs_fail
;
1184 err
= ipv6_frag_init();
1186 goto ipv6_frag_fail
;
1188 /* Init v6 transport protocols. */
1193 err
= udplitev6_init();
1195 goto udplitev6_fail
;
1197 err
= udpv6_offload_init();
1199 goto udpv6_offload_fail
;
1205 err
= ipv6_packet_init();
1207 goto ipv6_packet_fail
;
1209 err
= pingv6_init();
1213 err
= calipso_init();
1229 err
= igmp6_late_init();
1231 goto igmp6_late_err
;
1233 #ifdef CONFIG_SYSCTL
1234 err
= ipv6_sysctl_register();
1239 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1241 ipv6_stub
= &ipv6_stub_impl
;
1242 ipv6_bpf_stub
= &ipv6_bpf_stub_impl
;
1246 #ifdef CONFIG_SYSCTL
1248 igmp6_late_cleanup();
1261 ipv6_packet_cleanup();
1265 udpv6_offload_exit();
1273 ipv6_exthdrs_exit();
1277 ipv6_anycast_cleanup();
1279 ip6_flowlabel_cleanup();
1281 ndisc_late_cleanup();
1283 ip6_route_cleanup();
1285 #ifdef CONFIG_PROC_FS
1288 ipv6_misc_proc_exit();
1290 udplite6_proc_exit();
1295 ipv6_netfilter_fini();
1305 unregister_pernet_subsys(&inet6_net_ops
);
1306 register_pernet_fail
:
1307 sock_unregister(PF_INET6
);
1308 rtnl_unregister_all(PF_INET6
);
1309 out_sock_register_fail
:
1311 out_unregister_ping_proto
:
1312 proto_unregister(&pingv6_prot
);
1313 out_unregister_raw_proto
:
1314 proto_unregister(&rawv6_prot
);
1315 out_unregister_udplite_proto
:
1316 proto_unregister(&udplitev6_prot
);
1317 out_unregister_udp_proto
:
1318 proto_unregister(&udpv6_prot
);
1319 out_unregister_tcp_proto
:
1320 proto_unregister(&tcpv6_prot
);
1323 module_init(inet6_init
);
1325 MODULE_ALIAS_NETPROTO(PF_INET6
);