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>
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/icmpv6.h>
43 #include <linux/netfilter_ipv6.h>
48 #include <net/udplite.h>
51 #include <net/protocol.h>
52 #include <net/inet_common.h>
53 #include <net/route.h>
54 #include <net/transp_v6.h>
55 #include <net/ip6_route.h>
56 #include <net/addrconf.h>
57 #ifdef CONFIG_IPV6_TUNNEL
58 #include <net/ip6_tunnel.h>
61 #include <asm/uaccess.h>
62 #include <asm/system.h>
63 #include <linux/mroute6.h>
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
69 /* The inetsw6 table contains everything that inet6_create needs to
72 static struct list_head inetsw6
[SOCK_MAX
];
73 static DEFINE_SPINLOCK(inetsw6_lock
);
75 struct ipv6_params ipv6_defaults
= {
80 static int disable_ipv6_mod
= 0;
82 module_param_named(disable
, disable_ipv6_mod
, int, 0444);
83 MODULE_PARM_DESC(disable
, "Disable IPv6 module such that it is non-functional");
85 module_param_named(disable_ipv6
, ipv6_defaults
.disable_ipv6
, int, 0444);
86 MODULE_PARM_DESC(disable_ipv6
, "Disable IPv6 on all interfaces");
88 module_param_named(autoconf
, ipv6_defaults
.autoconf
, int, 0444);
89 MODULE_PARM_DESC(autoconf
, "Enable IPv6 address autoconfiguration on all interfaces");
91 static __inline__
struct ipv6_pinfo
*inet6_sk_generic(struct sock
*sk
)
93 const int offset
= sk
->sk_prot
->obj_size
- sizeof(struct ipv6_pinfo
);
95 return (struct ipv6_pinfo
*)(((u8
*)sk
) + offset
);
98 static int inet6_create(struct net
*net
, struct socket
*sock
, int protocol
)
100 struct inet_sock
*inet
;
101 struct ipv6_pinfo
*np
;
103 struct inet_protosw
*answer
;
104 struct proto
*answer_prot
;
105 unsigned char answer_flags
;
106 char answer_no_check
;
107 int try_loading_module
= 0;
110 if (sock
->type
!= SOCK_RAW
&&
111 sock
->type
!= SOCK_DGRAM
&&
113 build_ehash_secret();
115 /* Look for the requested type/protocol pair. */
117 err
= -ESOCKTNOSUPPORT
;
119 list_for_each_entry_rcu(answer
, &inetsw6
[sock
->type
], list
) {
122 /* Check the non-wild match. */
123 if (protocol
== answer
->protocol
) {
124 if (protocol
!= IPPROTO_IP
)
127 /* Check for the two wild cases. */
128 if (IPPROTO_IP
== protocol
) {
129 protocol
= answer
->protocol
;
132 if (IPPROTO_IP
== answer
->protocol
)
135 err
= -EPROTONOSUPPORT
;
139 if (try_loading_module
< 2) {
142 * Be more specific, e.g. net-pf-10-proto-132-type-1
143 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
145 if (++try_loading_module
== 1)
146 request_module("net-pf-%d-proto-%d-type-%d",
147 PF_INET6
, protocol
, sock
->type
);
149 * Fall back to generic, e.g. net-pf-10-proto-132
150 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 request_module("net-pf-%d-proto-%d",
155 goto lookup_protocol
;
161 if (answer
->capability
> 0 && !capable(answer
->capability
))
164 sock
->ops
= answer
->ops
;
165 answer_prot
= answer
->prot
;
166 answer_no_check
= answer
->no_check
;
167 answer_flags
= answer
->flags
;
170 WARN_ON(answer_prot
->slab
== NULL
);
173 sk
= sk_alloc(net
, PF_INET6
, GFP_KERNEL
, answer_prot
);
177 sock_init_data(sock
, sk
);
180 sk
->sk_no_check
= answer_no_check
;
181 if (INET_PROTOSW_REUSE
& answer_flags
)
185 inet
->is_icsk
= (INET_PROTOSW_ICSK
& answer_flags
) != 0;
187 if (SOCK_RAW
== sock
->type
) {
188 inet
->num
= protocol
;
189 if (IPPROTO_RAW
== protocol
)
193 sk
->sk_destruct
= inet_sock_destruct
;
194 sk
->sk_family
= PF_INET6
;
195 sk
->sk_protocol
= protocol
;
197 sk
->sk_backlog_rcv
= answer
->prot
->backlog_rcv
;
199 inet_sk(sk
)->pinet6
= np
= inet6_sk_generic(sk
);
203 np
->pmtudisc
= IPV6_PMTUDISC_WANT
;
204 np
->ipv6only
= net
->ipv6
.sysctl
.bindv6only
;
206 /* Init the ipv4 part of the socket since we can have sockets
207 * using v6 API for ipv4.
214 inet
->mc_list
= NULL
;
216 if (ipv4_config
.no_pmtu_disc
)
217 inet
->pmtudisc
= IP_PMTUDISC_DONT
;
219 inet
->pmtudisc
= IP_PMTUDISC_WANT
;
221 * Increment only the relevant sk_prot->socks debug field, this changes
222 * the previous behaviour of incrementing both the equivalent to
223 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
225 * This allows better debug granularity as we'll know exactly how many
226 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
227 * transport protocol socks. -acme
229 sk_refcnt_debug_inc(sk
);
232 /* It assumes that any protocol which allows
233 * the user to assign a number at socket
234 * creation time automatically shares.
236 inet
->sport
= htons(inet
->num
);
237 sk
->sk_prot
->hash(sk
);
239 if (sk
->sk_prot
->init
) {
240 err
= sk
->sk_prot
->init(sk
);
242 sk_common_release(sk
);
254 /* bind for INET6 API */
255 int inet6_bind(struct socket
*sock
, struct sockaddr
*uaddr
, int addr_len
)
257 struct sockaddr_in6
*addr
=(struct sockaddr_in6
*)uaddr
;
258 struct sock
*sk
= sock
->sk
;
259 struct inet_sock
*inet
= inet_sk(sk
);
260 struct ipv6_pinfo
*np
= inet6_sk(sk
);
261 struct net
*net
= sock_net(sk
);
267 /* If the socket has its own bind function then use it. */
268 if (sk
->sk_prot
->bind
)
269 return sk
->sk_prot
->bind(sk
, uaddr
, addr_len
);
271 if (addr_len
< SIN6_LEN_RFC2133
)
273 addr_type
= ipv6_addr_type(&addr
->sin6_addr
);
274 if ((addr_type
& IPV6_ADDR_MULTICAST
) && sock
->type
== SOCK_STREAM
)
277 snum
= ntohs(addr
->sin6_port
);
278 if (snum
&& snum
< PROT_SOCK
&& !capable(CAP_NET_BIND_SERVICE
))
283 /* Check these errors (active socket, double bind). */
284 if (sk
->sk_state
!= TCP_CLOSE
|| inet
->num
) {
289 /* Check if the address belongs to the host. */
290 if (addr_type
== IPV6_ADDR_MAPPED
) {
293 /* Binding to v4-mapped address on a v6-only socket
301 /* Reproduce AF_INET checks to make the bindings consitant */
302 v4addr
= addr
->sin6_addr
.s6_addr32
[3];
303 chk_addr_ret
= inet_addr_type(net
, v4addr
);
304 if (!sysctl_ip_nonlocal_bind
&&
305 !(inet
->freebind
|| inet
->transparent
) &&
306 v4addr
!= htonl(INADDR_ANY
) &&
307 chk_addr_ret
!= RTN_LOCAL
&&
308 chk_addr_ret
!= RTN_MULTICAST
&&
309 chk_addr_ret
!= RTN_BROADCAST
) {
310 err
= -EADDRNOTAVAIL
;
314 if (addr_type
!= IPV6_ADDR_ANY
) {
315 struct net_device
*dev
= NULL
;
317 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
318 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
319 addr
->sin6_scope_id
) {
320 /* Override any existing binding, if another one
321 * is supplied by user.
323 sk
->sk_bound_dev_if
= addr
->sin6_scope_id
;
326 /* Binding to link-local address requires an interface */
327 if (!sk
->sk_bound_dev_if
) {
331 dev
= dev_get_by_index(net
, sk
->sk_bound_dev_if
);
338 /* ipv4 addr of the socket is invalid. Only the
339 * unspecified and mapped address have a v4 equivalent.
341 v4addr
= LOOPBACK4_IPV6
;
342 if (!(addr_type
& IPV6_ADDR_MULTICAST
)) {
343 if (!ipv6_chk_addr(net
, &addr
->sin6_addr
,
347 err
= -EADDRNOTAVAIL
;
356 inet
->rcv_saddr
= v4addr
;
357 inet
->saddr
= v4addr
;
359 ipv6_addr_copy(&np
->rcv_saddr
, &addr
->sin6_addr
);
361 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
362 ipv6_addr_copy(&np
->saddr
, &addr
->sin6_addr
);
364 /* Make sure we are allowed to bind here. */
365 if (sk
->sk_prot
->get_port(sk
, snum
)) {
366 inet_reset_saddr(sk
);
371 if (addr_type
!= IPV6_ADDR_ANY
) {
372 sk
->sk_userlocks
|= SOCK_BINDADDR_LOCK
;
373 if (addr_type
!= IPV6_ADDR_MAPPED
)
377 sk
->sk_userlocks
|= SOCK_BINDPORT_LOCK
;
378 inet
->sport
= htons(inet
->num
);
386 EXPORT_SYMBOL(inet6_bind
);
388 int inet6_release(struct socket
*sock
)
390 struct sock
*sk
= sock
->sk
;
396 ipv6_sock_mc_close(sk
);
399 ipv6_sock_ac_close(sk
);
401 return inet_release(sock
);
404 EXPORT_SYMBOL(inet6_release
);
406 void inet6_destroy_sock(struct sock
*sk
)
408 struct ipv6_pinfo
*np
= inet6_sk(sk
);
410 struct ipv6_txoptions
*opt
;
412 /* Release rx options */
414 if ((skb
= xchg(&np
->pktoptions
, NULL
)) != NULL
)
417 /* Free flowlabels */
418 fl6_free_socklist(sk
);
420 /* Free tx options */
422 if ((opt
= xchg(&np
->opt
, NULL
)) != NULL
)
423 sock_kfree_s(sk
, opt
, opt
->tot_len
);
426 EXPORT_SYMBOL_GPL(inet6_destroy_sock
);
429 * This does both peername and sockname.
432 int inet6_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
433 int *uaddr_len
, int peer
)
435 struct sockaddr_in6
*sin
=(struct sockaddr_in6
*)uaddr
;
436 struct sock
*sk
= sock
->sk
;
437 struct inet_sock
*inet
= inet_sk(sk
);
438 struct ipv6_pinfo
*np
= inet6_sk(sk
);
440 sin
->sin6_family
= AF_INET6
;
441 sin
->sin6_flowinfo
= 0;
442 sin
->sin6_scope_id
= 0;
446 if (((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_SYN_SENT
)) &&
449 sin
->sin6_port
= inet
->dport
;
450 ipv6_addr_copy(&sin
->sin6_addr
, &np
->daddr
);
452 sin
->sin6_flowinfo
= np
->flow_label
;
454 if (ipv6_addr_any(&np
->rcv_saddr
))
455 ipv6_addr_copy(&sin
->sin6_addr
, &np
->saddr
);
457 ipv6_addr_copy(&sin
->sin6_addr
, &np
->rcv_saddr
);
459 sin
->sin6_port
= inet
->sport
;
461 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
462 sin
->sin6_scope_id
= sk
->sk_bound_dev_if
;
463 *uaddr_len
= sizeof(*sin
);
467 EXPORT_SYMBOL(inet6_getname
);
469 int inet6_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
471 struct sock
*sk
= sock
->sk
;
472 struct net
*net
= sock_net(sk
);
477 return sock_get_timestamp(sk
, (struct timeval __user
*)arg
);
480 return sock_get_timestampns(sk
, (struct timespec __user
*)arg
);
485 return(ipv6_route_ioctl(net
, cmd
, (void __user
*)arg
));
488 return addrconf_add_ifaddr(net
, (void __user
*) arg
);
490 return addrconf_del_ifaddr(net
, (void __user
*) arg
);
492 return addrconf_set_dstaddr(net
, (void __user
*) arg
);
494 if (!sk
->sk_prot
->ioctl
)
496 return sk
->sk_prot
->ioctl(sk
, cmd
, arg
);
502 EXPORT_SYMBOL(inet6_ioctl
);
504 const struct proto_ops inet6_stream_ops
= {
506 .owner
= THIS_MODULE
,
507 .release
= inet6_release
,
509 .connect
= inet_stream_connect
, /* ok */
510 .socketpair
= sock_no_socketpair
, /* a do nothing */
511 .accept
= inet_accept
, /* ok */
512 .getname
= inet6_getname
,
513 .poll
= tcp_poll
, /* ok */
514 .ioctl
= inet6_ioctl
, /* must change */
515 .listen
= inet_listen
, /* ok */
516 .shutdown
= inet_shutdown
, /* ok */
517 .setsockopt
= sock_common_setsockopt
, /* ok */
518 .getsockopt
= sock_common_getsockopt
, /* ok */
519 .sendmsg
= tcp_sendmsg
, /* ok */
520 .recvmsg
= sock_common_recvmsg
, /* ok */
521 .mmap
= sock_no_mmap
,
522 .sendpage
= tcp_sendpage
,
523 .splice_read
= tcp_splice_read
,
525 .compat_setsockopt
= compat_sock_common_setsockopt
,
526 .compat_getsockopt
= compat_sock_common_getsockopt
,
530 const struct proto_ops inet6_dgram_ops
= {
532 .owner
= THIS_MODULE
,
533 .release
= inet6_release
,
535 .connect
= inet_dgram_connect
, /* ok */
536 .socketpair
= sock_no_socketpair
, /* a do nothing */
537 .accept
= sock_no_accept
, /* a do nothing */
538 .getname
= inet6_getname
,
539 .poll
= udp_poll
, /* ok */
540 .ioctl
= inet6_ioctl
, /* must change */
541 .listen
= sock_no_listen
, /* ok */
542 .shutdown
= inet_shutdown
, /* ok */
543 .setsockopt
= sock_common_setsockopt
, /* ok */
544 .getsockopt
= sock_common_getsockopt
, /* ok */
545 .sendmsg
= inet_sendmsg
, /* ok */
546 .recvmsg
= sock_common_recvmsg
, /* ok */
547 .mmap
= sock_no_mmap
,
548 .sendpage
= sock_no_sendpage
,
550 .compat_setsockopt
= compat_sock_common_setsockopt
,
551 .compat_getsockopt
= compat_sock_common_getsockopt
,
555 static struct net_proto_family inet6_family_ops
= {
557 .create
= inet6_create
,
558 .owner
= THIS_MODULE
,
561 int inet6_register_protosw(struct inet_protosw
*p
)
563 struct list_head
*lh
;
564 struct inet_protosw
*answer
;
565 struct list_head
*last_perm
;
566 int protocol
= p
->protocol
;
569 spin_lock_bh(&inetsw6_lock
);
572 if (p
->type
>= SOCK_MAX
)
575 /* If we are trying to override a permanent protocol, bail. */
578 last_perm
= &inetsw6
[p
->type
];
579 list_for_each(lh
, &inetsw6
[p
->type
]) {
580 answer
= list_entry(lh
, struct inet_protosw
, list
);
582 /* Check only the non-wild match. */
583 if (INET_PROTOSW_PERMANENT
& answer
->flags
) {
584 if (protocol
== answer
->protocol
)
594 /* Add the new entry after the last permanent entry if any, so that
595 * the new entry does not override a permanent entry when matched with
596 * a wild-card protocol. But it is allowed to override any existing
597 * non-permanent entry. This means that when we remove this entry, the
598 * system automatically returns to the old behavior.
600 list_add_rcu(&p
->list
, last_perm
);
603 spin_unlock_bh(&inetsw6_lock
);
607 printk(KERN_ERR
"Attempt to override permanent protocol %d.\n",
613 "Ignoring attempt to register invalid socket type %d.\n",
618 EXPORT_SYMBOL(inet6_register_protosw
);
621 inet6_unregister_protosw(struct inet_protosw
*p
)
623 if (INET_PROTOSW_PERMANENT
& p
->flags
) {
625 "Attempt to unregister permanent protocol %d.\n",
628 spin_lock_bh(&inetsw6_lock
);
629 list_del_rcu(&p
->list
);
630 spin_unlock_bh(&inetsw6_lock
);
636 EXPORT_SYMBOL(inet6_unregister_protosw
);
638 int inet6_sk_rebuild_header(struct sock
*sk
)
641 struct dst_entry
*dst
;
642 struct ipv6_pinfo
*np
= inet6_sk(sk
);
644 dst
= __sk_dst_check(sk
, np
->dst_cookie
);
647 struct inet_sock
*inet
= inet_sk(sk
);
648 struct in6_addr
*final_p
= NULL
, final
;
651 memset(&fl
, 0, sizeof(fl
));
652 fl
.proto
= sk
->sk_protocol
;
653 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
654 ipv6_addr_copy(&fl
.fl6_src
, &np
->saddr
);
655 fl
.fl6_flowlabel
= np
->flow_label
;
656 fl
.oif
= sk
->sk_bound_dev_if
;
657 fl
.fl_ip_dport
= inet
->dport
;
658 fl
.fl_ip_sport
= inet
->sport
;
659 security_sk_classify_flow(sk
, &fl
);
661 if (np
->opt
&& np
->opt
->srcrt
) {
662 struct rt0_hdr
*rt0
= (struct rt0_hdr
*) np
->opt
->srcrt
;
663 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
664 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
668 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
670 sk
->sk_route_caps
= 0;
674 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
676 if ((err
= xfrm_lookup(sock_net(sk
), &dst
, &fl
, sk
, 0)) < 0) {
677 sk
->sk_err_soft
= -err
;
681 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
687 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header
);
689 int ipv6_opt_accepted(struct sock
*sk
, struct sk_buff
*skb
)
691 struct ipv6_pinfo
*np
= inet6_sk(sk
);
692 struct inet6_skb_parm
*opt
= IP6CB(skb
);
695 if ((opt
->hop
&& (np
->rxopt
.bits
.hopopts
||
696 np
->rxopt
.bits
.ohopopts
)) ||
697 ((IPV6_FLOWINFO_MASK
&
698 *(__be32
*)skb_network_header(skb
)) &&
699 np
->rxopt
.bits
.rxflow
) ||
700 (opt
->srcrt
&& (np
->rxopt
.bits
.srcrt
||
701 np
->rxopt
.bits
.osrcrt
)) ||
702 ((opt
->dst1
|| opt
->dst0
) &&
703 (np
->rxopt
.bits
.dstopts
|| np
->rxopt
.bits
.odstopts
)))
709 EXPORT_SYMBOL_GPL(ipv6_opt_accepted
);
711 static int ipv6_gso_pull_exthdrs(struct sk_buff
*skb
, int proto
)
713 const struct inet6_protocol
*ops
= NULL
;
716 struct ipv6_opt_hdr
*opth
;
719 if (proto
!= NEXTHDR_HOP
) {
720 ops
= rcu_dereference(inet6_protos
[proto
]);
725 if (!(ops
->flags
& INET6_PROTO_GSO_EXTHDR
))
729 if (unlikely(!pskb_may_pull(skb
, 8)))
732 opth
= (void *)skb
->data
;
733 len
= ipv6_optlen(opth
);
735 if (unlikely(!pskb_may_pull(skb
, len
)))
738 proto
= opth
->nexthdr
;
739 __skb_pull(skb
, len
);
745 static int ipv6_gso_send_check(struct sk_buff
*skb
)
747 struct ipv6hdr
*ipv6h
;
748 const struct inet6_protocol
*ops
;
751 if (unlikely(!pskb_may_pull(skb
, sizeof(*ipv6h
))))
754 ipv6h
= ipv6_hdr(skb
);
755 __skb_pull(skb
, sizeof(*ipv6h
));
756 err
= -EPROTONOSUPPORT
;
759 ops
= rcu_dereference(inet6_protos
[
760 ipv6_gso_pull_exthdrs(skb
, ipv6h
->nexthdr
)]);
762 if (likely(ops
&& ops
->gso_send_check
)) {
763 skb_reset_transport_header(skb
);
764 err
= ops
->gso_send_check(skb
);
772 static struct sk_buff
*ipv6_gso_segment(struct sk_buff
*skb
, int features
)
774 struct sk_buff
*segs
= ERR_PTR(-EINVAL
);
775 struct ipv6hdr
*ipv6h
;
776 const struct inet6_protocol
*ops
;
778 struct frag_hdr
*fptr
;
779 unsigned int unfrag_ip6hlen
;
783 if (!(features
& NETIF_F_V6_CSUM
))
784 features
&= ~NETIF_F_SG
;
786 if (unlikely(skb_shinfo(skb
)->gso_type
&
794 if (unlikely(!pskb_may_pull(skb
, sizeof(*ipv6h
))))
797 ipv6h
= ipv6_hdr(skb
);
798 __skb_pull(skb
, sizeof(*ipv6h
));
799 segs
= ERR_PTR(-EPROTONOSUPPORT
);
801 proto
= ipv6_gso_pull_exthdrs(skb
, ipv6h
->nexthdr
);
803 ops
= rcu_dereference(inet6_protos
[proto
]);
804 if (likely(ops
&& ops
->gso_segment
)) {
805 skb_reset_transport_header(skb
);
806 segs
= ops
->gso_segment(skb
, features
);
810 if (unlikely(IS_ERR(segs
)))
813 for (skb
= segs
; skb
; skb
= skb
->next
) {
814 ipv6h
= ipv6_hdr(skb
);
815 ipv6h
->payload_len
= htons(skb
->len
- skb
->mac_len
-
817 if (proto
== IPPROTO_UDP
) {
818 unfrag_ip6hlen
= ip6_find_1stfragopt(skb
, &prevhdr
);
819 fptr
= (struct frag_hdr
*)(skb_network_header(skb
) +
821 fptr
->frag_off
= htons(offset
);
822 if (skb
->next
!= NULL
)
823 fptr
->frag_off
|= htons(IP6_MF
);
824 offset
+= (ntohs(ipv6h
->payload_len
) -
825 sizeof(struct frag_hdr
));
834 struct napi_gro_cb napi
;
838 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
840 static struct sk_buff
**ipv6_gro_receive(struct sk_buff
**head
,
843 const struct inet6_protocol
*ops
;
844 struct sk_buff
**pp
= NULL
;
854 off
= skb_gro_offset(skb
);
855 hlen
= off
+ sizeof(*iph
);
856 iph
= skb_gro_header_fast(skb
, off
);
857 if (skb_gro_header_hard(skb
, hlen
)) {
858 iph
= skb_gro_header_slow(skb
, hlen
, off
);
863 skb_gro_pull(skb
, sizeof(*iph
));
864 skb_set_transport_header(skb
, skb_gro_offset(skb
));
866 flush
+= ntohs(iph
->payload_len
) != skb_gro_len(skb
);
869 proto
= iph
->nexthdr
;
870 ops
= rcu_dereference(inet6_protos
[proto
]);
871 if (!ops
|| !ops
->gro_receive
) {
872 __pskb_pull(skb
, skb_gro_offset(skb
));
873 proto
= ipv6_gso_pull_exthdrs(skb
, proto
);
874 skb_gro_pull(skb
, -skb_transport_offset(skb
));
875 skb_reset_transport_header(skb
);
876 __skb_push(skb
, skb_gro_offset(skb
));
878 if (!ops
|| !ops
->gro_receive
)
884 IPV6_GRO_CB(skb
)->proto
= proto
;
887 nlen
= skb_network_header_len(skb
);
889 for (p
= *head
; p
; p
= p
->next
) {
890 struct ipv6hdr
*iph2
;
892 if (!NAPI_GRO_CB(p
)->same_flow
)
897 /* All fields must match except length. */
898 if (nlen
!= skb_network_header_len(p
) ||
899 memcmp(iph
, iph2
, offsetof(struct ipv6hdr
, payload_len
)) ||
900 memcmp(&iph
->nexthdr
, &iph2
->nexthdr
,
901 nlen
- offsetof(struct ipv6hdr
, nexthdr
))) {
902 NAPI_GRO_CB(p
)->same_flow
= 0;
906 NAPI_GRO_CB(p
)->flush
|= flush
;
909 NAPI_GRO_CB(skb
)->flush
|= flush
;
912 skb_postpull_rcsum(skb
, iph
, skb_network_header_len(skb
));
914 pp
= ops
->gro_receive(head
, skb
);
922 NAPI_GRO_CB(skb
)->flush
|= flush
;
927 static int ipv6_gro_complete(struct sk_buff
*skb
)
929 const struct inet6_protocol
*ops
;
930 struct ipv6hdr
*iph
= ipv6_hdr(skb
);
933 iph
->payload_len
= htons(skb
->len
- skb_network_offset(skb
) -
937 ops
= rcu_dereference(inet6_protos
[IPV6_GRO_CB(skb
)->proto
]);
938 if (WARN_ON(!ops
|| !ops
->gro_complete
))
941 err
= ops
->gro_complete(skb
);
949 static struct packet_type ipv6_packet_type __read_mostly
= {
950 .type
= cpu_to_be16(ETH_P_IPV6
),
952 .gso_send_check
= ipv6_gso_send_check
,
953 .gso_segment
= ipv6_gso_segment
,
954 .gro_receive
= ipv6_gro_receive
,
955 .gro_complete
= ipv6_gro_complete
,
958 static int __init
ipv6_packet_init(void)
960 dev_add_pack(&ipv6_packet_type
);
964 static void ipv6_packet_cleanup(void)
966 dev_remove_pack(&ipv6_packet_type
);
969 static int __net_init
ipv6_init_mibs(struct net
*net
)
971 if (snmp_mib_init((void **)net
->mib
.udp_stats_in6
,
972 sizeof (struct udp_mib
)) < 0)
974 if (snmp_mib_init((void **)net
->mib
.udplite_stats_in6
,
975 sizeof (struct udp_mib
)) < 0)
976 goto err_udplite_mib
;
977 if (snmp_mib_init((void **)net
->mib
.ipv6_statistics
,
978 sizeof(struct ipstats_mib
)) < 0)
980 if (snmp_mib_init((void **)net
->mib
.icmpv6_statistics
,
981 sizeof(struct icmpv6_mib
)) < 0)
983 if (snmp_mib_init((void **)net
->mib
.icmpv6msg_statistics
,
984 sizeof(struct icmpv6msg_mib
)) < 0)
985 goto err_icmpmsg_mib
;
989 snmp_mib_free((void **)net
->mib
.icmpv6_statistics
);
991 snmp_mib_free((void **)net
->mib
.ipv6_statistics
);
993 snmp_mib_free((void **)net
->mib
.udplite_stats_in6
);
995 snmp_mib_free((void **)net
->mib
.udp_stats_in6
);
999 static void __net_exit
ipv6_cleanup_mibs(struct net
*net
)
1001 snmp_mib_free((void **)net
->mib
.udp_stats_in6
);
1002 snmp_mib_free((void **)net
->mib
.udplite_stats_in6
);
1003 snmp_mib_free((void **)net
->mib
.ipv6_statistics
);
1004 snmp_mib_free((void **)net
->mib
.icmpv6_statistics
);
1005 snmp_mib_free((void **)net
->mib
.icmpv6msg_statistics
);
1008 static int __net_init
inet6_net_init(struct net
*net
)
1012 net
->ipv6
.sysctl
.bindv6only
= 0;
1013 net
->ipv6
.sysctl
.icmpv6_time
= 1*HZ
;
1015 err
= ipv6_init_mibs(net
);
1018 #ifdef CONFIG_PROC_FS
1019 err
= udp6_proc_init(net
);
1022 err
= tcp6_proc_init(net
);
1024 goto proc_tcp6_fail
;
1025 err
= ac6_proc_init(net
);
1031 #ifdef CONFIG_PROC_FS
1033 tcp6_proc_exit(net
);
1035 udp6_proc_exit(net
);
1037 ipv6_cleanup_mibs(net
);
1042 static void inet6_net_exit(struct net
*net
)
1044 #ifdef CONFIG_PROC_FS
1045 udp6_proc_exit(net
);
1046 tcp6_proc_exit(net
);
1049 ipv6_cleanup_mibs(net
);
1052 static struct pernet_operations inet6_net_ops
= {
1053 .init
= inet6_net_init
,
1054 .exit
= inet6_net_exit
,
1057 static int __init
inet6_init(void)
1059 struct sk_buff
*dummy_skb
;
1060 struct list_head
*r
;
1063 BUILD_BUG_ON(sizeof(struct inet6_skb_parm
) > sizeof(dummy_skb
->cb
));
1065 /* Register the socket-side information for inet6_create. */
1066 for(r
= &inetsw6
[0]; r
< &inetsw6
[SOCK_MAX
]; ++r
)
1069 if (disable_ipv6_mod
) {
1071 "IPv6: Loaded, but administratively disabled, "
1072 "reboot required to enable\n");
1076 err
= proto_register(&tcpv6_prot
, 1);
1080 err
= proto_register(&udpv6_prot
, 1);
1082 goto out_unregister_tcp_proto
;
1084 err
= proto_register(&udplitev6_prot
, 1);
1086 goto out_unregister_udp_proto
;
1088 err
= proto_register(&rawv6_prot
, 1);
1090 goto out_unregister_udplite_proto
;
1093 /* We MUST register RAW sockets before we create the ICMP6,
1094 * IGMP6, or NDISC control sockets.
1098 goto out_unregister_raw_proto
;
1100 /* Register the family here so that the init calls below will
1101 * be able to create sockets. (?? is this dangerous ??)
1103 err
= sock_register(&inet6_family_ops
);
1105 goto out_sock_register_fail
;
1107 #ifdef CONFIG_SYSCTL
1108 err
= ipv6_static_sysctl_register();
1110 goto static_sysctl_fail
;
1113 * ipngwg API draft makes clear that the correct semantics
1114 * for TCP and UDP is to consider one TCP and UDP instance
1115 * in a host availiable by both INET and INET6 APIs and
1116 * able to communicate via both network protocols.
1119 err
= register_pernet_subsys(&inet6_net_ops
);
1121 goto register_pernet_fail
;
1122 err
= icmpv6_init();
1125 err
= ip6_mr_init();
1134 err
= ipv6_netfilter_init();
1136 goto netfilter_fail
;
1137 /* Create /proc/foo6 entries. */
1138 #ifdef CONFIG_PROC_FS
1140 if (raw6_proc_init())
1141 goto proc_raw6_fail
;
1142 if (udplite6_proc_init())
1143 goto proc_udplite6_fail
;
1144 if (ipv6_misc_proc_init())
1145 goto proc_misc6_fail
;
1146 if (if6_proc_init())
1149 err
= ip6_route_init();
1151 goto ip6_route_fail
;
1152 err
= ip6_flowlabel_init();
1154 goto ip6_flowlabel_fail
;
1155 err
= addrconf_init();
1159 /* Init v6 extension headers. */
1160 err
= ipv6_exthdrs_init();
1162 goto ipv6_exthdrs_fail
;
1164 err
= ipv6_frag_init();
1166 goto ipv6_frag_fail
;
1168 /* Init v6 transport protocols. */
1173 err
= udplitev6_init();
1175 goto udplitev6_fail
;
1181 err
= ipv6_packet_init();
1183 goto ipv6_packet_fail
;
1185 #ifdef CONFIG_SYSCTL
1186 err
= ipv6_sysctl_register();
1193 #ifdef CONFIG_SYSCTL
1195 ipv6_packet_cleanup();
1206 ipv6_exthdrs_exit();
1210 ip6_flowlabel_cleanup();
1212 ip6_route_cleanup();
1214 #ifdef CONFIG_PROC_FS
1217 ipv6_misc_proc_exit();
1219 udplite6_proc_exit();
1224 ipv6_netfilter_fini();
1234 unregister_pernet_subsys(&inet6_net_ops
);
1235 register_pernet_fail
:
1236 #ifdef CONFIG_SYSCTL
1237 ipv6_static_sysctl_unregister();
1240 sock_unregister(PF_INET6
);
1241 rtnl_unregister_all(PF_INET6
);
1242 out_sock_register_fail
:
1244 out_unregister_raw_proto
:
1245 proto_unregister(&rawv6_prot
);
1246 out_unregister_udplite_proto
:
1247 proto_unregister(&udplitev6_prot
);
1248 out_unregister_udp_proto
:
1249 proto_unregister(&udpv6_prot
);
1250 out_unregister_tcp_proto
:
1251 proto_unregister(&tcpv6_prot
);
1254 module_init(inet6_init
);
1256 static void __exit
inet6_exit(void)
1258 if (disable_ipv6_mod
)
1261 /* First of all disallow new sockets creation. */
1262 sock_unregister(PF_INET6
);
1263 /* Disallow any further netlink messages */
1264 rtnl_unregister_all(PF_INET6
);
1266 #ifdef CONFIG_SYSCTL
1267 ipv6_sysctl_unregister();
1273 /* Cleanup code parts. */
1274 ipv6_packet_cleanup();
1276 ipv6_exthdrs_exit();
1278 ip6_flowlabel_cleanup();
1279 ip6_route_cleanup();
1280 #ifdef CONFIG_PROC_FS
1282 /* Cleanup code parts. */
1284 ipv6_misc_proc_exit();
1285 udplite6_proc_exit();
1288 ipv6_netfilter_fini();
1295 unregister_pernet_subsys(&inet6_net_ops
);
1296 #ifdef CONFIG_SYSCTL
1297 ipv6_static_sysctl_unregister();
1299 proto_unregister(&rawv6_prot
);
1300 proto_unregister(&udplitev6_prot
);
1301 proto_unregister(&udpv6_prot
);
1302 proto_unregister(&tcpv6_prot
);
1304 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1306 module_exit(inet6_exit
);
1308 MODULE_ALIAS_NETPROTO(PF_INET6
);