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
11 * piggy, Karl Knutson : Socket protocol table
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * Arnaldo Melo : check proc_net_create return, cleanups
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
28 #include <linux/kernel.h>
29 #include <linux/timer.h>
30 #include <linux/string.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/fcntl.h>
35 #include <linux/interrupt.h>
36 #include <linux/proc_fs.h>
37 #include <linux/stat.h>
38 #include <linux/init.h>
39 #include <linux/slab.h>
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/icmpv6.h>
44 #include <linux/netfilter_ipv6.h>
49 #include <net/udplite.h>
52 #include <net/protocol.h>
53 #include <net/inet_common.h>
54 #include <net/route.h>
55 #include <net/transp_v6.h>
56 #include <net/ip6_route.h>
57 #include <net/addrconf.h>
58 #ifdef CONFIG_IPV6_TUNNEL
59 #include <net/ip6_tunnel.h>
62 #include <asm/uaccess.h>
63 #include <asm/system.h>
64 #include <linux/mroute6.h>
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
70 /* The inetsw6 table contains everything that inet6_create needs to
73 static struct list_head inetsw6
[SOCK_MAX
];
74 static DEFINE_SPINLOCK(inetsw6_lock
);
76 struct ipv6_params ipv6_defaults
= {
81 static int disable_ipv6_mod
= 0;
83 module_param_named(disable
, disable_ipv6_mod
, int, 0444);
84 MODULE_PARM_DESC(disable
, "Disable IPv6 module such that it is non-functional");
86 module_param_named(disable_ipv6
, ipv6_defaults
.disable_ipv6
, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6
, "Disable IPv6 on all interfaces");
89 module_param_named(autoconf
, ipv6_defaults
.autoconf
, int, 0444);
90 MODULE_PARM_DESC(autoconf
, "Enable IPv6 address autoconfiguration on all interfaces");
92 static __inline__
struct ipv6_pinfo
*inet6_sk_generic(struct sock
*sk
)
94 const int offset
= sk
->sk_prot
->obj_size
- sizeof(struct ipv6_pinfo
);
96 return (struct ipv6_pinfo
*)(((u8
*)sk
) + offset
);
99 static int inet6_create(struct net
*net
, struct socket
*sock
, int protocol
,
102 struct inet_sock
*inet
;
103 struct ipv6_pinfo
*np
;
105 struct inet_protosw
*answer
;
106 struct proto
*answer_prot
;
107 unsigned char answer_flags
;
108 char answer_no_check
;
109 int try_loading_module
= 0;
112 if (sock
->type
!= SOCK_RAW
&&
113 sock
->type
!= SOCK_DGRAM
&&
115 build_ehash_secret();
117 /* Look for the requested type/protocol pair. */
119 err
= -ESOCKTNOSUPPORT
;
121 list_for_each_entry_rcu(answer
, &inetsw6
[sock
->type
], list
) {
124 /* Check the non-wild match. */
125 if (protocol
== answer
->protocol
) {
126 if (protocol
!= IPPROTO_IP
)
129 /* Check for the two wild cases. */
130 if (IPPROTO_IP
== protocol
) {
131 protocol
= answer
->protocol
;
134 if (IPPROTO_IP
== answer
->protocol
)
137 err
= -EPROTONOSUPPORT
;
141 if (try_loading_module
< 2) {
144 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
147 if (++try_loading_module
== 1)
148 request_module("net-pf-%d-proto-%d-type-%d",
149 PF_INET6
, protocol
, sock
->type
);
151 * Fall back to generic, e.g. net-pf-10-proto-132
152 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
155 request_module("net-pf-%d-proto-%d",
157 goto lookup_protocol
;
163 if (sock
->type
== SOCK_RAW
&& !kern
&& !capable(CAP_NET_RAW
))
166 sock
->ops
= answer
->ops
;
167 answer_prot
= answer
->prot
;
168 answer_no_check
= answer
->no_check
;
169 answer_flags
= answer
->flags
;
172 WARN_ON(answer_prot
->slab
== NULL
);
175 sk
= sk_alloc(net
, PF_INET6
, GFP_KERNEL
, answer_prot
);
179 sock_init_data(sock
, sk
);
182 sk
->sk_no_check
= answer_no_check
;
183 if (INET_PROTOSW_REUSE
& answer_flags
)
187 inet
->is_icsk
= (INET_PROTOSW_ICSK
& answer_flags
) != 0;
189 if (SOCK_RAW
== sock
->type
) {
190 inet
->inet_num
= protocol
;
191 if (IPPROTO_RAW
== protocol
)
195 sk
->sk_destruct
= inet_sock_destruct
;
196 sk
->sk_family
= PF_INET6
;
197 sk
->sk_protocol
= protocol
;
199 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
201 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
203 np
->mcast_hops
= IPV6_DEFAULT_MCASTHOPS
;
205 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
206 np
->ipv6only
= net
->ipv6
.sysctl
.bindv6only
;
208 /* Init the ipv4 part of the socket since we can have sockets
209 * using v6 API for ipv4.
216 inet
->mc_list
= NULL
;
219 if (ipv4_config
.no_pmtu_disc
)
220 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
222 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
224 * Increment only the relevant sk_prot->socks debug field, this changes
225 * the previous behaviour of incrementing both the equivalent to
226 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
228 * This allows better debug granularity as we'll know exactly how many
229 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
230 * transport protocol socks. -acme
232 sk_refcnt_debug_inc(sk
);
234 if (inet
->inet_num
) {
235 /* It assumes that any protocol which allows
236 * the user to assign a number at socket
237 * creation time automatically shares.
239 inet
->inet_sport
= htons(inet
->inet_num
);
240 sk
->sk_prot
->hash(sk
);
242 if (sk
->sk_prot
->init
) {
243 err
= sk
->sk_prot
->init(sk
);
245 sk_common_release(sk
);
257 /* bind for INET6 API */
258 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
260 struct sockaddr_in6
*addr
=(struct sockaddr_in6
*)uaddr
;
261 struct sock
*sk
= sock
->sk
;
262 struct inet_sock
*inet
= inet_sk(sk
);
263 struct ipv6_pinfo
*np
= inet6_sk(sk
);
264 struct net
*net
= sock_net(sk
);
270 /* If the socket has its own bind function then use it. */
271 if (sk
->sk_prot
->bind
)
272 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
274 if (addr_len
< SIN6_LEN_RFC2133
)
277 if (addr
->sin6_family
!= AF_INET6
)
278 return -EAFNOSUPPORT
;
280 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
281 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sock
->type
== SOCK_STREAM
)
284 snum
= ntohs(addr
->sin6_port
);
285 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
290 /* Check these errors (active socket, double bind). */
291 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->inet_num
) {
296 /* Check if the address belongs to the host. */
297 if (addr_type
== IPV6_ADDR_MAPPED
) {
300 /* Binding to v4-mapped address on a v6-only socket
308 /* Reproduce AF_INET checks to make the bindings consistent */
309 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
310 chk_addr_ret
= inet_addr_type(net
, v4addr
);
311 if (!sysctl_ip_nonlocal_bind
&&
312 !(inet
->freebind
|| inet
->transparent
) &&
313 v4addr
!= htonl(INADDR_ANY
) &&
314 chk_addr_ret
!= RTN_LOCAL
&&
315 chk_addr_ret
!= RTN_MULTICAST
&&
316 chk_addr_ret
!= RTN_BROADCAST
) {
317 err
= -EADDRNOTAVAIL
;
321 if (addr_type
!= IPV6_ADDR_ANY
) {
322 struct net_device
*dev
= NULL
;
325 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
326 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
327 addr
->sin6_scope_id
) {
328 /* Override any existing binding, if another one
329 * is supplied by user.
331 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
334 /* Binding to link-local address requires an interface */
335 if (!sk
->sk_bound_dev_if
) {
339 dev
= dev_get_by_index_rcu(net
, sk
->sk_bound_dev_if
);
346 /* ipv4 addr of the socket is invalid. Only the
347 * unspecified and mapped address have a v4 equivalent.
349 v4addr
= LOOPBACK4_IPV6
;
350 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
351 if (!(inet
->freebind
|| inet
->transparent
) &&
352 !ipv6_chk_addr(net
, &addr
->sin6_addr
,
354 err
= -EADDRNOTAVAIL
;
362 inet
->inet_rcv_saddr
= v4addr
;
363 inet
->inet_saddr
= v4addr
;
365 np
->rcv_saddr
= addr
->sin6_addr
;
367 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
368 np
->saddr
= addr
->sin6_addr
;
370 /* Make sure we are allowed to bind here. */
371 if (sk
->sk_prot
->get_port(sk
, snum
)) {
372 inet_reset_saddr(sk
);
377 if (addr_type
!= IPV6_ADDR_ANY
) {
378 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
379 if (addr_type
!= IPV6_ADDR_MAPPED
)
383 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
384 inet
->inet_sport
= htons(inet
->inet_num
);
385 inet
->inet_dport
= 0;
386 inet
->inet_daddr
= 0;
395 EXPORT_SYMBOL(inet6_bind
);
397 int inet6_release(struct socket
*sock
)
399 struct sock
*sk
= sock
->sk
;
405 ipv6_sock_mc_close(sk
);
408 ipv6_sock_ac_close(sk
);
410 return inet_release(sock
);
413 EXPORT_SYMBOL(inet6_release
);
415 void inet6_destroy_sock(struct sock
*sk
)
417 struct ipv6_pinfo
*np
= inet6_sk(sk
);
419 struct ipv6_txoptions
*opt
;
421 /* Release rx options */
423 if ((skb
= xchg(&np
->pktoptions
, NULL
)) != NULL
)
426 if ((skb
= xchg(&np
->rxpmtu
, NULL
)) != NULL
)
429 /* Free flowlabels */
430 fl6_free_socklist(sk
);
432 /* Free tx options */
434 if ((opt
= xchg(&np
->opt
, NULL
)) != NULL
)
435 sock_kfree_s(sk
, opt
, opt
->tot_len
);
438 EXPORT_SYMBOL_GPL(inet6_destroy_sock
);
441 * This does both peername and sockname.
444 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
445 int *uaddr_len
, int peer
)
447 struct sockaddr_in6
*sin
=(struct sockaddr_in6
*)uaddr
;
448 struct sock
*sk
= sock
->sk
;
449 struct inet_sock
*inet
= inet_sk(sk
);
450 struct ipv6_pinfo
*np
= inet6_sk(sk
);
452 sin
->sin6_family
= AF_INET6
;
453 sin
->sin6_flowinfo
= 0;
454 sin
->sin6_scope_id
= 0;
456 if (!inet
->inet_dport
)
458 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
461 sin
->sin6_port
= inet
->inet_dport
;
462 sin
->sin6_addr
= np
->daddr
;
464 sin
->sin6_flowinfo
= np
->flow_label
;
466 if (ipv6_addr_any(&np
->rcv_saddr
))
467 sin
->sin6_addr
= np
->saddr
;
469 sin
->sin6_addr
= np
->rcv_saddr
;
471 sin
->sin6_port
= inet
->inet_sport
;
473 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
474 sin
->sin6_scope_id
= sk
->sk_bound_dev_if
;
475 *uaddr_len
= sizeof(*sin
);
479 EXPORT_SYMBOL(inet6_getname
);
481 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
483 struct sock
*sk
= sock
->sk
;
484 struct net
*net
= sock_net(sk
);
489 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
492 return sock_get_timestampns(sk
, (struct timespec __user
*)arg
);
497 return ipv6_route_ioctl(net
, cmd
, (void __user
*)arg
);
500 return addrconf_add_ifaddr(net
, (void __user
*) arg
);
502 return addrconf_del_ifaddr(net
, (void __user
*) arg
);
504 return addrconf_set_dstaddr(net
, (void __user
*) arg
);
506 if (!sk
->sk_prot
->ioctl
)
508 return sk
->sk_prot
->ioctl(sk
, cmd
, arg
);
514 EXPORT_SYMBOL(inet6_ioctl
);
516 const struct proto_ops inet6_stream_ops
= {
518 .owner
= THIS_MODULE
,
519 .release
= inet6_release
,
521 .connect
= inet_stream_connect
, /* ok */
522 .socketpair
= sock_no_socketpair
, /* a do nothing */
523 .accept
= inet_accept
, /* ok */
524 .getname
= inet6_getname
,
525 .poll
= tcp_poll
, /* ok */
526 .ioctl
= inet6_ioctl
, /* must change */
527 .listen
= inet_listen
, /* ok */
528 .shutdown
= inet_shutdown
, /* ok */
529 .setsockopt
= sock_common_setsockopt
, /* ok */
530 .getsockopt
= sock_common_getsockopt
, /* ok */
531 .sendmsg
= inet_sendmsg
, /* ok */
532 .recvmsg
= inet_recvmsg
, /* ok */
533 .mmap
= sock_no_mmap
,
534 .sendpage
= inet_sendpage
,
535 .splice_read
= tcp_splice_read
,
537 .compat_setsockopt
= compat_sock_common_setsockopt
,
538 .compat_getsockopt
= compat_sock_common_getsockopt
,
542 const struct proto_ops inet6_dgram_ops
= {
544 .owner
= THIS_MODULE
,
545 .release
= inet6_release
,
547 .connect
= inet_dgram_connect
, /* ok */
548 .socketpair
= sock_no_socketpair
, /* a do nothing */
549 .accept
= sock_no_accept
, /* a do nothing */
550 .getname
= inet6_getname
,
551 .poll
= udp_poll
, /* ok */
552 .ioctl
= inet6_ioctl
, /* must change */
553 .listen
= sock_no_listen
, /* ok */
554 .shutdown
= inet_shutdown
, /* ok */
555 .setsockopt
= sock_common_setsockopt
, /* ok */
556 .getsockopt
= sock_common_getsockopt
, /* ok */
557 .sendmsg
= inet_sendmsg
, /* ok */
558 .recvmsg
= inet_recvmsg
, /* ok */
559 .mmap
= sock_no_mmap
,
560 .sendpage
= sock_no_sendpage
,
562 .compat_setsockopt
= compat_sock_common_setsockopt
,
563 .compat_getsockopt
= compat_sock_common_getsockopt
,
567 static const struct net_proto_family inet6_family_ops
= {
569 .create
= inet6_create
,
570 .owner
= THIS_MODULE
,
573 int inet6_register_protosw(struct inet_protosw
*p
)
575 struct list_head
*lh
;
576 struct inet_protosw
*answer
;
577 struct list_head
*last_perm
;
578 int protocol
= p
->protocol
;
581 spin_lock_bh(&inetsw6_lock
);
584 if (p
->type
>= SOCK_MAX
)
587 /* If we are trying to override a permanent protocol, bail. */
590 last_perm
= &inetsw6
[p
->type
];
591 list_for_each(lh
, &inetsw6
[p
->type
]) {
592 answer
= list_entry(lh
, struct inet_protosw
, list
);
594 /* Check only the non-wild match. */
595 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
596 if (protocol
== answer
->protocol
)
606 /* Add the new entry after the last permanent entry if any, so that
607 * the new entry does not override a permanent entry when matched with
608 * a wild-card protocol. But it is allowed to override any existing
609 * non-permanent entry. This means that when we remove this entry, the
610 * system automatically returns to the old behavior.
612 list_add_rcu(&p
->list
, last_perm
);
615 spin_unlock_bh(&inetsw6_lock
);
619 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
625 "Ignoring attempt to register invalid socket type %d.\n",
630 EXPORT_SYMBOL(inet6_register_protosw
);
633 inet6_unregister_protosw(struct inet_protosw
*p
)
635 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
637 "Attempt to unregister permanent protocol %d.\n",
640 spin_lock_bh(&inetsw6_lock
);
641 list_del_rcu(&p
->list
);
642 spin_unlock_bh(&inetsw6_lock
);
648 EXPORT_SYMBOL(inet6_unregister_protosw
);
650 int inet6_sk_rebuild_header(struct sock
*sk
)
652 struct ipv6_pinfo
*np
= inet6_sk(sk
);
653 struct dst_entry
*dst
;
655 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
658 struct inet_sock
*inet
= inet_sk(sk
);
659 struct in6_addr
*final_p
, final
;
662 memset(&fl6
, 0, sizeof(fl6
));
663 fl6
.flowi6_proto
= sk
->sk_protocol
;
664 fl6
.daddr
= np
->daddr
;
665 fl6
.saddr
= np
->saddr
;
666 fl6
.flowlabel
= np
->flow_label
;
667 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
668 fl6
.flowi6_mark
= sk
->sk_mark
;
669 fl6
.fl6_dport
= inet
->inet_dport
;
670 fl6
.fl6_sport
= inet
->inet_sport
;
671 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
673 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
675 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
, false);
677 sk
->sk_route_caps
= 0;
678 sk
->sk_err_soft
= -PTR_ERR(dst
);
682 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
688 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header
);
690 int ipv6_opt_accepted(struct sock
*sk
, struct sk_buff
*skb
)
692 struct ipv6_pinfo
*np
= inet6_sk(sk
);
693 struct inet6_skb_parm
*opt
= IP6CB(skb
);
696 if ((opt
->hop
&& (np
->rxopt
.bits
.hopopts
||
697 np
->rxopt
.bits
.ohopopts
)) ||
698 ((IPV6_FLOWINFO_MASK
&
699 *(__be32
*)skb_network_header(skb
)) &&
700 np
->rxopt
.bits
.rxflow
) ||
701 (opt
->srcrt
&& (np
->rxopt
.bits
.srcrt
||
702 np
->rxopt
.bits
.osrcrt
)) ||
703 ((opt
->dst1
|| opt
->dst0
) &&
704 (np
->rxopt
.bits
.dstopts
|| np
->rxopt
.bits
.odstopts
)))
710 EXPORT_SYMBOL_GPL(ipv6_opt_accepted
);
712 static int ipv6_gso_pull_exthdrs(struct sk_buff
*skb
, int proto
)
714 const struct inet6_protocol
*ops
= NULL
;
717 struct ipv6_opt_hdr
*opth
;
720 if (proto
!= NEXTHDR_HOP
) {
721 ops
= rcu_dereference(inet6_protos
[proto
]);
726 if (!(ops
->flags
& INET6_PROTO_GSO_EXTHDR
))
730 if (unlikely(!pskb_may_pull(skb
, 8)))
733 opth
= (void *)skb
->data
;
734 len
= ipv6_optlen(opth
);
736 if (unlikely(!pskb_may_pull(skb
, len
)))
739 proto
= opth
->nexthdr
;
740 __skb_pull(skb
, len
);
746 static int ipv6_gso_send_check(struct sk_buff
*skb
)
748 const struct ipv6hdr
*ipv6h
;
749 const struct inet6_protocol
*ops
;
752 if (unlikely(!pskb_may_pull(skb
, sizeof(*ipv6h
))))
755 ipv6h
= ipv6_hdr(skb
);
756 __skb_pull(skb
, sizeof(*ipv6h
));
757 err
= -EPROTONOSUPPORT
;
760 ops
= rcu_dereference(inet6_protos
[
761 ipv6_gso_pull_exthdrs(skb
, ipv6h
->nexthdr
)]);
763 if (likely(ops
&& ops
->gso_send_check
)) {
764 skb_reset_transport_header(skb
);
765 err
= ops
->gso_send_check(skb
);
773 static struct sk_buff
*ipv6_gso_segment(struct sk_buff
*skb
,
774 netdev_features_t features
)
776 struct sk_buff
*segs
= ERR_PTR(-EINVAL
);
777 struct ipv6hdr
*ipv6h
;
778 const struct inet6_protocol
*ops
;
780 struct frag_hdr
*fptr
;
781 unsigned int unfrag_ip6hlen
;
785 if (!(features
& NETIF_F_V6_CSUM
))
786 features
&= ~NETIF_F_SG
;
788 if (unlikely(skb_shinfo(skb
)->gso_type
&
796 if (unlikely(!pskb_may_pull(skb
, sizeof(*ipv6h
))))
799 ipv6h
= ipv6_hdr(skb
);
800 __skb_pull(skb
, sizeof(*ipv6h
));
801 segs
= ERR_PTR(-EPROTONOSUPPORT
);
803 proto
= ipv6_gso_pull_exthdrs(skb
, ipv6h
->nexthdr
);
805 ops
= rcu_dereference(inet6_protos
[proto
]);
806 if (likely(ops
&& ops
->gso_segment
)) {
807 skb_reset_transport_header(skb
);
808 segs
= ops
->gso_segment(skb
, features
);
815 for (skb
= segs
; skb
; skb
= skb
->next
) {
816 ipv6h
= ipv6_hdr(skb
);
817 ipv6h
->payload_len
= htons(skb
->len
- skb
->mac_len
-
819 if (proto
== IPPROTO_UDP
) {
820 unfrag_ip6hlen
= ip6_find_1stfragopt(skb
, &prevhdr
);
821 fptr
= (struct frag_hdr
*)(skb_network_header(skb
) +
823 fptr
->frag_off
= htons(offset
);
824 if (skb
->next
!= NULL
)
825 fptr
->frag_off
|= htons(IP6_MF
);
826 offset
+= (ntohs(ipv6h
->payload_len
) -
827 sizeof(struct frag_hdr
));
836 struct napi_gro_cb napi
;
840 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
842 static struct sk_buff
**ipv6_gro_receive(struct sk_buff
**head
,
845 const struct inet6_protocol
*ops
;
846 struct sk_buff
**pp
= NULL
;
856 off
= skb_gro_offset(skb
);
857 hlen
= off
+ sizeof(*iph
);
858 iph
= skb_gro_header_fast(skb
, off
);
859 if (skb_gro_header_hard(skb
, hlen
)) {
860 iph
= skb_gro_header_slow(skb
, hlen
, off
);
865 skb_gro_pull(skb
, sizeof(*iph
));
866 skb_set_transport_header(skb
, skb_gro_offset(skb
));
868 flush
+= ntohs(iph
->payload_len
) != skb_gro_len(skb
);
871 proto
= iph
->nexthdr
;
872 ops
= rcu_dereference(inet6_protos
[proto
]);
873 if (!ops
|| !ops
->gro_receive
) {
874 __pskb_pull(skb
, skb_gro_offset(skb
));
875 proto
= ipv6_gso_pull_exthdrs(skb
, proto
);
876 skb_gro_pull(skb
, -skb_transport_offset(skb
));
877 skb_reset_transport_header(skb
);
878 __skb_push(skb
, skb_gro_offset(skb
));
880 ops
= rcu_dereference(inet6_protos
[proto
]);
881 if (!ops
|| !ops
->gro_receive
)
887 IPV6_GRO_CB(skb
)->proto
= proto
;
890 nlen
= skb_network_header_len(skb
);
892 for (p
= *head
; p
; p
= p
->next
) {
893 struct ipv6hdr
*iph2
;
895 if (!NAPI_GRO_CB(p
)->same_flow
)
900 /* All fields must match except length. */
901 if (nlen
!= skb_network_header_len(p
) ||
902 memcmp(iph
, iph2
, offsetof(struct ipv6hdr
, payload_len
)) ||
903 memcmp(&iph
->nexthdr
, &iph2
->nexthdr
,
904 nlen
- offsetof(struct ipv6hdr
, nexthdr
))) {
905 NAPI_GRO_CB(p
)->same_flow
= 0;
909 NAPI_GRO_CB(p
)->flush
|= flush
;
912 NAPI_GRO_CB(skb
)->flush
|= flush
;
915 skb_postpull_rcsum(skb
, iph
, skb_network_header_len(skb
));
917 pp
= ops
->gro_receive(head
, skb
);
925 NAPI_GRO_CB(skb
)->flush
|= flush
;
930 static int ipv6_gro_complete(struct sk_buff
*skb
)
932 const struct inet6_protocol
*ops
;
933 struct ipv6hdr
*iph
= ipv6_hdr(skb
);
936 iph
->payload_len
= htons(skb
->len
- skb_network_offset(skb
) -
940 ops
= rcu_dereference(inet6_protos
[IPV6_GRO_CB(skb
)->proto
]);
941 if (WARN_ON(!ops
|| !ops
->gro_complete
))
944 err
= ops
->gro_complete(skb
);
952 static struct packet_type ipv6_packet_type __read_mostly
= {
953 .type
= cpu_to_be16(ETH_P_IPV6
),
955 .gso_send_check
= ipv6_gso_send_check
,
956 .gso_segment
= ipv6_gso_segment
,
957 .gro_receive
= ipv6_gro_receive
,
958 .gro_complete
= ipv6_gro_complete
,
961 static int __init
ipv6_packet_init(void)
963 dev_add_pack(&ipv6_packet_type
);
967 static void ipv6_packet_cleanup(void)
969 dev_remove_pack(&ipv6_packet_type
);
972 static int __net_init
ipv6_init_mibs(struct net
*net
)
974 if (snmp_mib_init((void __percpu
**)net
->mib
.udp_stats_in6
,
975 sizeof(struct udp_mib
),
976 __alignof__(struct udp_mib
)) < 0)
978 if (snmp_mib_init((void __percpu
**)net
->mib
.udplite_stats_in6
,
979 sizeof(struct udp_mib
),
980 __alignof__(struct udp_mib
)) < 0)
981 goto err_udplite_mib
;
982 if (snmp_mib_init((void __percpu
**)net
->mib
.ipv6_statistics
,
983 sizeof(struct ipstats_mib
),
984 __alignof__(struct ipstats_mib
)) < 0)
986 if (snmp_mib_init((void __percpu
**)net
->mib
.icmpv6_statistics
,
987 sizeof(struct icmpv6_mib
),
988 __alignof__(struct icmpv6_mib
)) < 0)
990 net
->mib
.icmpv6msg_statistics
= kzalloc(sizeof(struct icmpv6msg_mib
),
992 if (!net
->mib
.icmpv6msg_statistics
)
993 goto err_icmpmsg_mib
;
997 snmp_mib_free((void __percpu
**)net
->mib
.icmpv6_statistics
);
999 snmp_mib_free((void __percpu
**)net
->mib
.ipv6_statistics
);
1001 snmp_mib_free((void __percpu
**)net
->mib
.udplite_stats_in6
);
1003 snmp_mib_free((void __percpu
**)net
->mib
.udp_stats_in6
);
1007 static void ipv6_cleanup_mibs(struct net
*net
)
1009 snmp_mib_free((void __percpu
**)net
->mib
.udp_stats_in6
);
1010 snmp_mib_free((void __percpu
**)net
->mib
.udplite_stats_in6
);
1011 snmp_mib_free((void __percpu
**)net
->mib
.ipv6_statistics
);
1012 snmp_mib_free((void __percpu
**)net
->mib
.icmpv6_statistics
);
1013 kfree(net
->mib
.icmpv6msg_statistics
);
1016 static int __net_init
inet6_net_init(struct net
*net
)
1020 net
->ipv6
.sysctl
.bindv6only
= 0;
1021 net
->ipv6
.sysctl
.icmpv6_time
= 1*HZ
;
1023 err
= ipv6_init_mibs(net
);
1026 #ifdef CONFIG_PROC_FS
1027 err
= udp6_proc_init(net
);
1030 err
= tcp6_proc_init(net
);
1032 goto proc_tcp6_fail
;
1033 err
= ac6_proc_init(net
);
1039 #ifdef CONFIG_PROC_FS
1041 tcp6_proc_exit(net
);
1043 udp6_proc_exit(net
);
1045 ipv6_cleanup_mibs(net
);
1050 static void __net_exit
inet6_net_exit(struct net
*net
)
1052 #ifdef CONFIG_PROC_FS
1053 udp6_proc_exit(net
);
1054 tcp6_proc_exit(net
);
1057 ipv6_cleanup_mibs(net
);
1060 static struct pernet_operations inet6_net_ops
= {
1061 .init
= inet6_net_init
,
1062 .exit
= inet6_net_exit
,
1065 static int __init
inet6_init(void)
1067 struct sk_buff
*dummy_skb
;
1068 struct list_head
*r
;
1071 BUILD_BUG_ON(sizeof(struct inet6_skb_parm
) > sizeof(dummy_skb
->cb
));
1073 /* Register the socket-side information for inet6_create. */
1074 for(r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
1077 if (disable_ipv6_mod
) {
1079 "IPv6: Loaded, but administratively disabled, "
1080 "reboot required to enable\n");
1084 err
= proto_register(&tcpv6_prot
, 1);
1088 err
= proto_register(&udpv6_prot
, 1);
1090 goto out_unregister_tcp_proto
;
1092 err
= proto_register(&udplitev6_prot
, 1);
1094 goto out_unregister_udp_proto
;
1096 err
= proto_register(&rawv6_prot
, 1);
1098 goto out_unregister_udplite_proto
;
1101 /* We MUST register RAW sockets before we create the ICMP6,
1102 * IGMP6, or NDISC control sockets.
1106 goto out_unregister_raw_proto
;
1108 /* Register the family here so that the init calls below will
1109 * be able to create sockets. (?? is this dangerous ??)
1111 err
= sock_register(&inet6_family_ops
);
1113 goto out_sock_register_fail
;
1115 #ifdef CONFIG_SYSCTL
1116 err
= ipv6_static_sysctl_register();
1118 goto static_sysctl_fail
;
1120 tcpv6_prot
.sysctl_mem
= init_net
.ipv4
.sysctl_tcp_mem
;
1123 * ipngwg API draft makes clear that the correct semantics
1124 * for TCP and UDP is to consider one TCP and UDP instance
1125 * in a host available by both INET and INET6 APIs and
1126 * able to communicate via both network protocols.
1129 err
= register_pernet_subsys(&inet6_net_ops
);
1131 goto register_pernet_fail
;
1132 err
= icmpv6_init();
1135 err
= ip6_mr_init();
1144 err
= ipv6_netfilter_init();
1146 goto netfilter_fail
;
1147 /* Create /proc/foo6 entries. */
1148 #ifdef CONFIG_PROC_FS
1150 if (raw6_proc_init())
1151 goto proc_raw6_fail
;
1152 if (udplite6_proc_init())
1153 goto proc_udplite6_fail
;
1154 if (ipv6_misc_proc_init())
1155 goto proc_misc6_fail
;
1156 if (if6_proc_init())
1159 err
= ip6_route_init();
1161 goto ip6_route_fail
;
1162 err
= ip6_flowlabel_init();
1164 goto ip6_flowlabel_fail
;
1165 err
= addrconf_init();
1169 /* Init v6 extension headers. */
1170 err
= ipv6_exthdrs_init();
1172 goto ipv6_exthdrs_fail
;
1174 err
= ipv6_frag_init();
1176 goto ipv6_frag_fail
;
1178 /* Init v6 transport protocols. */
1183 err
= udplitev6_init();
1185 goto udplitev6_fail
;
1191 err
= ipv6_packet_init();
1193 goto ipv6_packet_fail
;
1195 #ifdef CONFIG_SYSCTL
1196 err
= ipv6_sysctl_register();
1203 #ifdef CONFIG_SYSCTL
1205 ipv6_packet_cleanup();
1216 ipv6_exthdrs_exit();
1220 ip6_flowlabel_cleanup();
1222 ip6_route_cleanup();
1224 #ifdef CONFIG_PROC_FS
1227 ipv6_misc_proc_exit();
1229 udplite6_proc_exit();
1234 ipv6_netfilter_fini();
1244 unregister_pernet_subsys(&inet6_net_ops
);
1245 register_pernet_fail
:
1246 #ifdef CONFIG_SYSCTL
1247 ipv6_static_sysctl_unregister();
1250 sock_unregister(PF_INET6
);
1251 rtnl_unregister_all(PF_INET6
);
1252 out_sock_register_fail
:
1254 out_unregister_raw_proto
:
1255 proto_unregister(&rawv6_prot
);
1256 out_unregister_udplite_proto
:
1257 proto_unregister(&udplitev6_prot
);
1258 out_unregister_udp_proto
:
1259 proto_unregister(&udpv6_prot
);
1260 out_unregister_tcp_proto
:
1261 proto_unregister(&tcpv6_prot
);
1264 module_init(inet6_init
);
1266 static void __exit
inet6_exit(void)
1268 if (disable_ipv6_mod
)
1271 /* First of all disallow new sockets creation. */
1272 sock_unregister(PF_INET6
);
1273 /* Disallow any further netlink messages */
1274 rtnl_unregister_all(PF_INET6
);
1276 #ifdef CONFIG_SYSCTL
1277 ipv6_sysctl_unregister();
1283 /* Cleanup code parts. */
1284 ipv6_packet_cleanup();
1286 ipv6_exthdrs_exit();
1288 ip6_flowlabel_cleanup();
1289 ip6_route_cleanup();
1290 #ifdef CONFIG_PROC_FS
1292 /* Cleanup code parts. */
1294 ipv6_misc_proc_exit();
1295 udplite6_proc_exit();
1298 ipv6_netfilter_fini();
1305 unregister_pernet_subsys(&inet6_net_ops
);
1306 #ifdef CONFIG_SYSCTL
1307 ipv6_static_sysctl_unregister();
1309 proto_unregister(&rawv6_prot
);
1310 proto_unregister(&udplitev6_prot
);
1311 proto_unregister(&udpv6_prot
);
1312 proto_unregister(&tcpv6_prot
);
1314 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1316 module_exit(inet6_exit
);
1318 MODULE_ALIAS_NETPROTO(PF_INET6
);