2 * IPv6 BSD socket options interface
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on linux/net/ipv4/ip_sockglue.c
10 * $Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
17 * FIXME: Make the setsockopt code POSIX compliant: That is
19 * o Return -EINVAL for setsockopt of short lengths
20 * o Truncate getsockopt returns
21 * o Return an optlen of the truncated length if need be
24 * David L Stevens <dlstevens@us.ibm.com>:
25 * - added multicast source filtering API for MLDv2
28 #include <linux/module.h>
29 #include <linux/config.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/sched.h>
35 #include <linux/net.h>
36 #include <linux/in6.h>
37 #include <linux/netdevice.h>
38 #include <linux/if_arp.h>
39 #include <linux/init.h>
40 #include <linux/sysctl.h>
41 #include <linux/netfilter.h>
46 #include <net/ndisc.h>
47 #include <net/protocol.h>
48 #include <net/transp_v6.h>
49 #include <net/ip6_route.h>
50 #include <net/addrconf.h>
51 #include <net/inet_common.h>
56 #include <asm/uaccess.h>
58 DEFINE_SNMP_STAT(struct ipstats_mib
, ipv6_statistics
) __read_mostly
;
60 static struct packet_type ipv6_packet_type
= {
61 .type
= __constant_htons(ETH_P_IPV6
),
65 struct ip6_ra_chain
*ip6_ra_chain
;
66 DEFINE_RWLOCK(ip6_ra_lock
);
68 int ip6_ra_control(struct sock
*sk
, int sel
, void (*destructor
)(struct sock
*))
70 struct ip6_ra_chain
*ra
, *new_ra
, **rap
;
72 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
73 if (sk
->sk_type
!= SOCK_RAW
|| inet_sk(sk
)->num
!= IPPROTO_RAW
)
76 new_ra
= (sel
>=0) ? kmalloc(sizeof(*new_ra
), GFP_KERNEL
) : NULL
;
78 write_lock_bh(&ip6_ra_lock
);
79 for (rap
= &ip6_ra_chain
; (ra
=*rap
) != NULL
; rap
= &ra
->next
) {
82 write_unlock_bh(&ip6_ra_lock
);
88 write_unlock_bh(&ip6_ra_lock
);
98 write_unlock_bh(&ip6_ra_lock
);
103 new_ra
->destructor
= destructor
;
107 write_unlock_bh(&ip6_ra_lock
);
111 int ipv6_setsockopt(struct sock
*sk
, int level
, int optname
,
112 char __user
*optval
, int optlen
)
114 struct ipv6_pinfo
*np
= inet6_sk(sk
);
116 int retv
= -ENOPROTOOPT
;
118 if (level
== SOL_IP
&& sk
->sk_type
!= SOCK_RAW
)
119 return udp_prot
.setsockopt(sk
, level
, optname
, optval
, optlen
);
126 else if (get_user(val
, (int __user
*) optval
))
136 if (val
== PF_INET
) {
137 struct ipv6_txoptions
*opt
;
138 struct sk_buff
*pktopt
;
140 if (sk
->sk_protocol
!= IPPROTO_UDP
&&
141 sk
->sk_protocol
!= IPPROTO_TCP
)
144 if (sk
->sk_state
!= TCP_ESTABLISHED
) {
149 if (ipv6_only_sock(sk
) ||
150 !(ipv6_addr_type(&np
->daddr
) & IPV6_ADDR_MAPPED
)) {
151 retv
= -EADDRNOTAVAIL
;
155 fl6_free_socklist(sk
);
156 ipv6_sock_mc_close(sk
);
159 * Sock is moving from IPv6 to IPv4 (sk_prot), so
160 * remove it from the refcnt debug socks count in the
163 sk_refcnt_debug_dec(sk
);
165 if (sk
->sk_protocol
== IPPROTO_TCP
) {
166 struct inet_connection_sock
*icsk
= inet_csk(sk
);
169 sock_prot_dec_use(sk
->sk_prot
);
170 sock_prot_inc_use(&tcp_prot
);
172 sk
->sk_prot
= &tcp_prot
;
173 icsk
->icsk_af_ops
= &ipv4_specific
;
174 sk
->sk_socket
->ops
= &inet_stream_ops
;
175 sk
->sk_family
= PF_INET
;
176 tcp_sync_mss(sk
, icsk
->icsk_pmtu_cookie
);
179 sock_prot_dec_use(sk
->sk_prot
);
180 sock_prot_inc_use(&udp_prot
);
182 sk
->sk_prot
= &udp_prot
;
183 sk
->sk_socket
->ops
= &inet_dgram_ops
;
184 sk
->sk_family
= PF_INET
;
186 opt
= xchg(&np
->opt
, NULL
);
188 sock_kfree_s(sk
, opt
, opt
->tot_len
);
189 pktopt
= xchg(&np
->pktoptions
, NULL
);
193 sk
->sk_destruct
= inet_sock_destruct
;
195 * ... and add it to the refcnt debug socks count
196 * in the new family. -acme
198 sk_refcnt_debug_inc(sk
);
199 module_put(THIS_MODULE
);
206 if (inet_sk(sk
)->num
)
208 np
->ipv6only
= valbool
;
212 case IPV6_RECVPKTINFO
:
213 np
->rxopt
.bits
.rxinfo
= valbool
;
217 case IPV6_2292PKTINFO
:
218 np
->rxopt
.bits
.rxoinfo
= valbool
;
222 case IPV6_RECVHOPLIMIT
:
223 np
->rxopt
.bits
.rxhlim
= valbool
;
227 case IPV6_2292HOPLIMIT
:
228 np
->rxopt
.bits
.rxohlim
= valbool
;
233 if (val
< 0 || val
> 2)
235 np
->rxopt
.bits
.srcrt
= val
;
240 if (val
< 0 || val
> 2)
242 np
->rxopt
.bits
.osrcrt
= val
;
246 case IPV6_RECVHOPOPTS
:
247 np
->rxopt
.bits
.hopopts
= valbool
;
251 case IPV6_2292HOPOPTS
:
252 np
->rxopt
.bits
.ohopopts
= valbool
;
256 case IPV6_RECVDSTOPTS
:
257 np
->rxopt
.bits
.dstopts
= valbool
;
261 case IPV6_2292DSTOPTS
:
262 np
->rxopt
.bits
.odstopts
= valbool
;
267 if (val
< 0 || val
> 0xff)
273 case IPV6_RECVTCLASS
:
274 np
->rxopt
.bits
.rxtclass
= valbool
;
279 np
->rxopt
.bits
.rxflow
= valbool
;
284 case IPV6_RTHDRDSTOPTS
:
288 struct ipv6_txoptions
*opt
;
292 /* hop-by-hop / destination options are privileged option */
294 if (optname
!= IPV6_RTHDR
&& !capable(CAP_NET_RAW
))
298 if (optlen
& 0x7 || optlen
> 8 * 255)
301 opt
= ipv6_renew_options(sk
, np
->opt
, optname
,
302 (struct ipv6_opt_hdr __user
*)optval
,
309 /* routing header option needs extra check */
310 if (optname
== IPV6_RTHDR
&& opt
->srcrt
) {
311 struct ipv6_rt_hdr
*rthdr
= opt
->srcrt
;
314 if ((rthdr
->hdrlen
& 1) ||
315 (rthdr
->hdrlen
>> 1) != rthdr
->segments_left
)
320 if (inet_sk(sk
)->is_icsk
) {
322 struct inet_connection_sock
*icsk
= inet_csk(sk
);
323 if (!((1 << sk
->sk_state
) &
324 (TCPF_LISTEN
| TCPF_CLOSE
))
325 && inet_sk(sk
)->daddr
!= LOOPBACK4_IPV6
) {
326 icsk
->icsk_ext_hdr_len
=
327 opt
->opt_flen
+ opt
->opt_nflen
;
328 icsk
->icsk_sync_mss(sk
, icsk
->icsk_pmtu_cookie
);
331 opt
= xchg(&np
->opt
, opt
);
334 write_lock(&sk
->sk_dst_lock
);
335 opt
= xchg(&np
->opt
, opt
);
336 write_unlock(&sk
->sk_dst_lock
);
341 sock_kfree_s(sk
, opt
, opt
->tot_len
);
345 case IPV6_2292PKTOPTIONS
:
347 struct ipv6_txoptions
*opt
= NULL
;
352 fl
.fl6_flowlabel
= 0;
353 fl
.oif
= sk
->sk_bound_dev_if
;
358 /* 1K is probably excessive
359 * 1K is surely not enough, 2K per standard header is 16K.
362 if (optlen
> 64*1024)
365 opt
= sock_kmalloc(sk
, sizeof(*opt
) + optlen
, GFP_KERNEL
);
370 memset(opt
, 0, sizeof(*opt
));
371 opt
->tot_len
= sizeof(*opt
) + optlen
;
373 if (copy_from_user(opt
+1, optval
, optlen
))
376 msg
.msg_controllen
= optlen
;
377 msg
.msg_control
= (void*)(opt
+1);
379 retv
= datagram_send_ctl(&msg
, &fl
, opt
, &junk
, &junk
);
384 if (inet_sk(sk
)->is_icsk
) {
386 struct inet_connection_sock
*icsk
= inet_csk(sk
);
387 if (!((1 << sk
->sk_state
) &
388 (TCPF_LISTEN
| TCPF_CLOSE
))
389 && inet_sk(sk
)->daddr
!= LOOPBACK4_IPV6
) {
390 icsk
->icsk_ext_hdr_len
=
391 opt
->opt_flen
+ opt
->opt_nflen
;
392 icsk
->icsk_sync_mss(sk
, icsk
->icsk_pmtu_cookie
);
395 opt
= xchg(&np
->opt
, opt
);
398 write_lock(&sk
->sk_dst_lock
);
399 opt
= xchg(&np
->opt
, opt
);
400 write_unlock(&sk
->sk_dst_lock
);
406 sock_kfree_s(sk
, opt
, opt
->tot_len
);
409 case IPV6_UNICAST_HOPS
:
410 if (val
> 255 || val
< -1)
416 case IPV6_MULTICAST_HOPS
:
417 if (sk
->sk_type
== SOCK_STREAM
)
419 if (val
> 255 || val
< -1)
421 np
->mcast_hops
= val
;
425 case IPV6_MULTICAST_LOOP
:
426 np
->mc_loop
= valbool
;
430 case IPV6_MULTICAST_IF
:
431 if (sk
->sk_type
== SOCK_STREAM
)
433 if (sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= val
)
436 if (__dev_get_by_index(val
) == NULL
) {
443 case IPV6_ADD_MEMBERSHIP
:
444 case IPV6_DROP_MEMBERSHIP
:
446 struct ipv6_mreq mreq
;
449 if (copy_from_user(&mreq
, optval
, sizeof(struct ipv6_mreq
)))
452 if (optname
== IPV6_ADD_MEMBERSHIP
)
453 retv
= ipv6_sock_mc_join(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_multiaddr
);
455 retv
= ipv6_sock_mc_drop(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_multiaddr
);
458 case IPV6_JOIN_ANYCAST
:
459 case IPV6_LEAVE_ANYCAST
:
461 struct ipv6_mreq mreq
;
463 if (optlen
!= sizeof(struct ipv6_mreq
))
467 if (copy_from_user(&mreq
, optval
, sizeof(struct ipv6_mreq
)))
470 if (optname
== IPV6_JOIN_ANYCAST
)
471 retv
= ipv6_sock_ac_join(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_acaddr
);
473 retv
= ipv6_sock_ac_drop(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_acaddr
);
476 case MCAST_JOIN_GROUP
:
477 case MCAST_LEAVE_GROUP
:
479 struct group_req greq
;
480 struct sockaddr_in6
*psin6
;
483 if (copy_from_user(&greq
, optval
, sizeof(struct group_req
)))
485 if (greq
.gr_group
.ss_family
!= AF_INET6
) {
486 retv
= -EADDRNOTAVAIL
;
489 psin6
= (struct sockaddr_in6
*)&greq
.gr_group
;
490 if (optname
== MCAST_JOIN_GROUP
)
491 retv
= ipv6_sock_mc_join(sk
, greq
.gr_interface
,
494 retv
= ipv6_sock_mc_drop(sk
, greq
.gr_interface
,
498 case MCAST_JOIN_SOURCE_GROUP
:
499 case MCAST_LEAVE_SOURCE_GROUP
:
500 case MCAST_BLOCK_SOURCE
:
501 case MCAST_UNBLOCK_SOURCE
:
503 struct group_source_req greqs
;
506 if (optlen
!= sizeof(struct group_source_req
))
508 if (copy_from_user(&greqs
, optval
, sizeof(greqs
))) {
512 if (greqs
.gsr_group
.ss_family
!= AF_INET6
||
513 greqs
.gsr_source
.ss_family
!= AF_INET6
) {
514 retv
= -EADDRNOTAVAIL
;
517 if (optname
== MCAST_BLOCK_SOURCE
) {
518 omode
= MCAST_EXCLUDE
;
520 } else if (optname
== MCAST_UNBLOCK_SOURCE
) {
521 omode
= MCAST_EXCLUDE
;
523 } else if (optname
== MCAST_JOIN_SOURCE_GROUP
) {
524 struct sockaddr_in6
*psin6
;
526 psin6
= (struct sockaddr_in6
*)&greqs
.gsr_group
;
527 retv
= ipv6_sock_mc_join(sk
, greqs
.gsr_interface
,
529 /* prior join w/ different source is ok */
530 if (retv
&& retv
!= -EADDRINUSE
)
532 omode
= MCAST_INCLUDE
;
534 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
535 omode
= MCAST_INCLUDE
;
538 retv
= ip6_mc_source(add
, omode
, sk
, &greqs
);
543 extern int sysctl_mld_max_msf
;
544 struct group_filter
*gsf
;
546 if (optlen
< GROUP_FILTER_SIZE(0))
548 if (optlen
> sysctl_optmem_max
) {
552 gsf
= (struct group_filter
*)kmalloc(optlen
,GFP_KERNEL
);
558 if (copy_from_user(gsf
, optval
, optlen
)) {
562 /* numsrc >= (4G-140)/128 overflow in 32 bits */
563 if (gsf
->gf_numsrc
>= 0x1ffffffU
||
564 gsf
->gf_numsrc
> sysctl_mld_max_msf
) {
569 if (GROUP_FILTER_SIZE(gsf
->gf_numsrc
) > optlen
) {
574 retv
= ip6_mc_msfilter(sk
, gsf
);
579 case IPV6_ROUTER_ALERT
:
580 retv
= ip6_ra_control(sk
, val
, NULL
);
582 case IPV6_MTU_DISCOVER
:
589 if (val
&& val
< IPV6_MIN_MTU
)
595 np
->recverr
= valbool
;
597 skb_queue_purge(&sk
->sk_error_queue
);
600 case IPV6_FLOWINFO_SEND
:
601 np
->sndflow
= valbool
;
604 case IPV6_FLOWLABEL_MGR
:
605 retv
= ipv6_flowlabel_opt(sk
, optval
, optlen
);
607 case IPV6_IPSEC_POLICY
:
608 case IPV6_XFRM_POLICY
:
610 if (!capable(CAP_NET_ADMIN
))
612 retv
= xfrm_user_policy(sk
, optname
, optval
, optlen
);
615 #ifdef CONFIG_NETFILTER
617 retv
= nf_setsockopt(sk
, PF_INET6
, optname
, optval
,
633 static int ipv6_getsockopt_sticky(struct sock
*sk
, struct ipv6_opt_hdr
*hdr
,
634 char __user
*optval
, int len
)
638 len
= min_t(int, len
, ipv6_optlen(hdr
));
639 if (copy_to_user(optval
, hdr
, ipv6_optlen(hdr
)))
644 int ipv6_getsockopt(struct sock
*sk
, int level
, int optname
,
645 char __user
*optval
, int __user
*optlen
)
647 struct ipv6_pinfo
*np
= inet6_sk(sk
);
651 if (level
== SOL_IP
&& sk
->sk_type
!= SOCK_RAW
)
652 return udp_prot
.getsockopt(sk
, level
, optname
, optval
, optlen
);
655 if (get_user(len
, optlen
))
659 if (sk
->sk_protocol
!= IPPROTO_UDP
&&
660 sk
->sk_protocol
!= IPPROTO_TCP
)
662 if (sk
->sk_state
!= TCP_ESTABLISHED
)
668 struct group_filter gsf
;
671 if (len
< GROUP_FILTER_SIZE(0))
673 if (copy_from_user(&gsf
, optval
, GROUP_FILTER_SIZE(0)))
676 err
= ip6_mc_msfget(sk
, &gsf
,
677 (struct group_filter __user
*)optval
, optlen
);
682 case IPV6_2292PKTOPTIONS
:
687 if (sk
->sk_type
!= SOCK_STREAM
)
690 msg
.msg_control
= optval
;
691 msg
.msg_controllen
= len
;
695 skb
= np
->pktoptions
;
697 atomic_inc(&skb
->users
);
701 int err
= datagram_recv_ctl(sk
, &msg
, skb
);
706 if (np
->rxopt
.bits
.rxinfo
) {
707 struct in6_pktinfo src_info
;
708 src_info
.ipi6_ifindex
= np
->mcast_oif
;
709 ipv6_addr_copy(&src_info
.ipi6_addr
, &np
->daddr
);
710 put_cmsg(&msg
, SOL_IPV6
, IPV6_PKTINFO
, sizeof(src_info
), &src_info
);
712 if (np
->rxopt
.bits
.rxhlim
) {
713 int hlim
= np
->mcast_hops
;
714 put_cmsg(&msg
, SOL_IPV6
, IPV6_HOPLIMIT
, sizeof(hlim
), &hlim
);
716 if (np
->rxopt
.bits
.rxoinfo
) {
717 struct in6_pktinfo src_info
;
718 src_info
.ipi6_ifindex
= np
->mcast_oif
;
719 ipv6_addr_copy(&src_info
.ipi6_addr
, &np
->daddr
);
720 put_cmsg(&msg
, SOL_IPV6
, IPV6_2292PKTINFO
, sizeof(src_info
), &src_info
);
722 if (np
->rxopt
.bits
.rxohlim
) {
723 int hlim
= np
->mcast_hops
;
724 put_cmsg(&msg
, SOL_IPV6
, IPV6_2292HOPLIMIT
, sizeof(hlim
), &hlim
);
727 len
-= msg
.msg_controllen
;
728 return put_user(len
, optlen
);
732 struct dst_entry
*dst
;
735 dst
= sk_dst_get(sk
);
750 case IPV6_RECVPKTINFO
:
751 val
= np
->rxopt
.bits
.rxinfo
;
754 case IPV6_2292PKTINFO
:
755 val
= np
->rxopt
.bits
.rxoinfo
;
758 case IPV6_RECVHOPLIMIT
:
759 val
= np
->rxopt
.bits
.rxhlim
;
762 case IPV6_2292HOPLIMIT
:
763 val
= np
->rxopt
.bits
.rxohlim
;
767 val
= np
->rxopt
.bits
.srcrt
;
771 val
= np
->rxopt
.bits
.osrcrt
;
775 case IPV6_RTHDRDSTOPTS
:
781 len
= ipv6_getsockopt_sticky(sk
, np
->opt
->hopopt
,
784 return put_user(len
, optlen
);
787 case IPV6_RECVHOPOPTS
:
788 val
= np
->rxopt
.bits
.hopopts
;
791 case IPV6_2292HOPOPTS
:
792 val
= np
->rxopt
.bits
.ohopopts
;
795 case IPV6_RECVDSTOPTS
:
796 val
= np
->rxopt
.bits
.dstopts
;
799 case IPV6_2292DSTOPTS
:
800 val
= np
->rxopt
.bits
.odstopts
;
807 case IPV6_RECVTCLASS
:
808 val
= np
->rxopt
.bits
.rxtclass
;
812 val
= np
->rxopt
.bits
.rxflow
;
815 case IPV6_UNICAST_HOPS
:
819 case IPV6_MULTICAST_HOPS
:
820 val
= np
->mcast_hops
;
823 case IPV6_MULTICAST_LOOP
:
827 case IPV6_MULTICAST_IF
:
831 case IPV6_MTU_DISCOVER
:
839 case IPV6_FLOWINFO_SEND
:
844 #ifdef CONFIG_NETFILTER
846 val
= nf_getsockopt(sk
, PF_INET6
, optname
, optval
,
850 val
= put_user(len
, optlen
);
856 len
= min_t(unsigned int, sizeof(int), len
);
857 if(put_user(len
, optlen
))
859 if(copy_to_user(optval
,&val
,len
))
864 void __init
ipv6_packet_init(void)
866 dev_add_pack(&ipv6_packet_type
);
869 void ipv6_packet_cleanup(void)
871 dev_remove_pack(&ipv6_packet_type
);