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
);
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
);
89 write_unlock_bh(&ip6_ra_lock
);
99 write_unlock_bh(&ip6_ra_lock
);
104 new_ra
->destructor
= destructor
;
108 write_unlock_bh(&ip6_ra_lock
);
112 extern int ip6_mc_source(int add
, int omode
, struct sock
*sk
,
113 struct group_source_req
*pgsr
);
114 extern int ip6_mc_msfilter(struct sock
*sk
, struct group_filter
*gsf
);
115 extern int ip6_mc_msfget(struct sock
*sk
, struct group_filter
*gsf
,
116 struct group_filter __user
*optval
, int __user
*optlen
);
119 int ipv6_setsockopt(struct sock
*sk
, int level
, int optname
,
120 char __user
*optval
, int optlen
)
122 struct ipv6_pinfo
*np
= inet6_sk(sk
);
124 int retv
= -ENOPROTOOPT
;
126 if (level
== SOL_IP
&& sk
->sk_type
!= SOCK_RAW
)
127 return udp_prot
.setsockopt(sk
, level
, optname
, optval
, optlen
);
134 else if (get_user(val
, (int __user
*) optval
))
144 if (val
== PF_INET
) {
145 struct ipv6_txoptions
*opt
;
146 struct sk_buff
*pktopt
;
148 if (sk
->sk_protocol
!= IPPROTO_UDP
&&
149 sk
->sk_protocol
!= IPPROTO_TCP
)
152 if (sk
->sk_state
!= TCP_ESTABLISHED
) {
157 if (ipv6_only_sock(sk
) ||
158 !(ipv6_addr_type(&np
->daddr
) & IPV6_ADDR_MAPPED
)) {
159 retv
= -EADDRNOTAVAIL
;
163 fl6_free_socklist(sk
);
164 ipv6_sock_mc_close(sk
);
166 if (sk
->sk_protocol
== IPPROTO_TCP
) {
167 struct tcp_sock
*tp
= tcp_sk(sk
);
170 sock_prot_dec_use(sk
->sk_prot
);
171 sock_prot_inc_use(&tcp_prot
);
173 sk
->sk_prot
= &tcp_prot
;
174 tp
->af_specific
= &ipv4_specific
;
175 sk
->sk_socket
->ops
= &inet_stream_ops
;
176 sk
->sk_family
= PF_INET
;
177 tcp_sync_mss(sk
, tp
->pmtu_cookie
);
180 sock_prot_dec_use(sk
->sk_prot
);
181 sock_prot_inc_use(&udp_prot
);
183 sk
->sk_prot
= &udp_prot
;
184 sk
->sk_socket
->ops
= &inet_dgram_ops
;
185 sk
->sk_family
= PF_INET
;
187 opt
= xchg(&np
->opt
, NULL
);
189 sock_kfree_s(sk
, opt
, opt
->tot_len
);
190 pktopt
= xchg(&np
->pktoptions
, NULL
);
194 sk
->sk_destruct
= inet_sock_destruct
;
195 #ifdef INET_REFCNT_DEBUG
196 atomic_dec(&inet6_sock_nr
);
198 module_put(THIS_MODULE
);
205 if (inet_sk(sk
)->num
)
207 np
->ipv6only
= valbool
;
212 np
->rxopt
.bits
.rxinfo
= valbool
;
217 np
->rxopt
.bits
.rxhlim
= valbool
;
222 if (val
< 0 || val
> 2)
224 np
->rxopt
.bits
.srcrt
= val
;
229 np
->rxopt
.bits
.hopopts
= valbool
;
234 np
->rxopt
.bits
.dstopts
= valbool
;
239 np
->rxopt
.bits
.rxflow
= valbool
;
243 case IPV6_PKTOPTIONS
:
245 struct ipv6_txoptions
*opt
= NULL
;
250 fl
.fl6_flowlabel
= 0;
251 fl
.oif
= sk
->sk_bound_dev_if
;
256 /* 1K is probably excessive
257 * 1K is surely not enough, 2K per standard header is 16K.
260 if (optlen
> 64*1024)
263 opt
= sock_kmalloc(sk
, sizeof(*opt
) + optlen
, GFP_KERNEL
);
268 memset(opt
, 0, sizeof(*opt
));
269 opt
->tot_len
= sizeof(*opt
) + optlen
;
271 if (copy_from_user(opt
+1, optval
, optlen
))
274 msg
.msg_controllen
= optlen
;
275 msg
.msg_control
= (void*)(opt
+1);
277 retv
= datagram_send_ctl(&msg
, &fl
, opt
, &junk
);
282 if (sk
->sk_type
== SOCK_STREAM
) {
284 struct tcp_sock
*tp
= tcp_sk(sk
);
285 if (!((1 << sk
->sk_state
) &
286 (TCPF_LISTEN
| TCPF_CLOSE
))
287 && inet_sk(sk
)->daddr
!= LOOPBACK4_IPV6
) {
288 tp
->ext_header_len
= opt
->opt_flen
+ opt
->opt_nflen
;
289 tcp_sync_mss(sk
, tp
->pmtu_cookie
);
292 opt
= xchg(&np
->opt
, opt
);
295 write_lock(&sk
->sk_dst_lock
);
296 opt
= xchg(&np
->opt
, opt
);
297 write_unlock(&sk
->sk_dst_lock
);
303 sock_kfree_s(sk
, opt
, opt
->tot_len
);
306 case IPV6_UNICAST_HOPS
:
307 if (val
> 255 || val
< -1)
313 case IPV6_MULTICAST_HOPS
:
314 if (sk
->sk_type
== SOCK_STREAM
)
316 if (val
> 255 || val
< -1)
318 np
->mcast_hops
= val
;
322 case IPV6_MULTICAST_LOOP
:
323 np
->mc_loop
= valbool
;
327 case IPV6_MULTICAST_IF
:
328 if (sk
->sk_type
== SOCK_STREAM
)
330 if (sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= val
)
333 if (__dev_get_by_index(val
) == NULL
) {
340 case IPV6_ADD_MEMBERSHIP
:
341 case IPV6_DROP_MEMBERSHIP
:
343 struct ipv6_mreq mreq
;
346 if (copy_from_user(&mreq
, optval
, sizeof(struct ipv6_mreq
)))
349 if (optname
== IPV6_ADD_MEMBERSHIP
)
350 retv
= ipv6_sock_mc_join(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_multiaddr
);
352 retv
= ipv6_sock_mc_drop(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_multiaddr
);
355 case IPV6_JOIN_ANYCAST
:
356 case IPV6_LEAVE_ANYCAST
:
358 struct ipv6_mreq mreq
;
360 if (optlen
!= sizeof(struct ipv6_mreq
))
364 if (copy_from_user(&mreq
, optval
, sizeof(struct ipv6_mreq
)))
367 if (optname
== IPV6_JOIN_ANYCAST
)
368 retv
= ipv6_sock_ac_join(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_acaddr
);
370 retv
= ipv6_sock_ac_drop(sk
, mreq
.ipv6mr_ifindex
, &mreq
.ipv6mr_acaddr
);
373 case MCAST_JOIN_GROUP
:
374 case MCAST_LEAVE_GROUP
:
376 struct group_req greq
;
377 struct sockaddr_in6
*psin6
;
380 if (copy_from_user(&greq
, optval
, sizeof(struct group_req
)))
382 if (greq
.gr_group
.ss_family
!= AF_INET6
) {
383 retv
= -EADDRNOTAVAIL
;
386 psin6
= (struct sockaddr_in6
*)&greq
.gr_group
;
387 if (optname
== MCAST_JOIN_GROUP
)
388 retv
= ipv6_sock_mc_join(sk
, greq
.gr_interface
,
391 retv
= ipv6_sock_mc_drop(sk
, greq
.gr_interface
,
395 case MCAST_JOIN_SOURCE_GROUP
:
396 case MCAST_LEAVE_SOURCE_GROUP
:
397 case MCAST_BLOCK_SOURCE
:
398 case MCAST_UNBLOCK_SOURCE
:
400 struct group_source_req greqs
;
403 if (optlen
!= sizeof(struct group_source_req
))
405 if (copy_from_user(&greqs
, optval
, sizeof(greqs
))) {
409 if (greqs
.gsr_group
.ss_family
!= AF_INET6
||
410 greqs
.gsr_source
.ss_family
!= AF_INET6
) {
411 retv
= -EADDRNOTAVAIL
;
414 if (optname
== MCAST_BLOCK_SOURCE
) {
415 omode
= MCAST_EXCLUDE
;
417 } else if (optname
== MCAST_UNBLOCK_SOURCE
) {
418 omode
= MCAST_EXCLUDE
;
420 } else if (optname
== MCAST_JOIN_SOURCE_GROUP
) {
421 struct sockaddr_in6
*psin6
;
423 psin6
= (struct sockaddr_in6
*)&greqs
.gsr_group
;
424 retv
= ipv6_sock_mc_join(sk
, greqs
.gsr_interface
,
426 /* prior join w/ different source is ok */
427 if (retv
&& retv
!= -EADDRINUSE
)
429 omode
= MCAST_INCLUDE
;
431 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
432 omode
= MCAST_INCLUDE
;
435 retv
= ip6_mc_source(add
, omode
, sk
, &greqs
);
440 extern int sysctl_optmem_max
;
441 extern int sysctl_mld_max_msf
;
442 struct group_filter
*gsf
;
444 if (optlen
< GROUP_FILTER_SIZE(0))
446 if (optlen
> sysctl_optmem_max
) {
450 gsf
= (struct group_filter
*)kmalloc(optlen
,GFP_KERNEL
);
456 if (copy_from_user(gsf
, optval
, optlen
)) {
460 /* numsrc >= (4G-140)/128 overflow in 32 bits */
461 if (gsf
->gf_numsrc
>= 0x1ffffffU
||
462 gsf
->gf_numsrc
> sysctl_mld_max_msf
) {
467 if (GROUP_FILTER_SIZE(gsf
->gf_numsrc
) > optlen
) {
472 retv
= ip6_mc_msfilter(sk
, gsf
);
477 case IPV6_ROUTER_ALERT
:
478 retv
= ip6_ra_control(sk
, val
, NULL
);
480 case IPV6_MTU_DISCOVER
:
487 if (val
&& val
< IPV6_MIN_MTU
)
493 np
->recverr
= valbool
;
495 skb_queue_purge(&sk
->sk_error_queue
);
498 case IPV6_FLOWINFO_SEND
:
499 np
->sndflow
= valbool
;
502 case IPV6_FLOWLABEL_MGR
:
503 retv
= ipv6_flowlabel_opt(sk
, optval
, optlen
);
505 case IPV6_IPSEC_POLICY
:
506 case IPV6_XFRM_POLICY
:
508 if (!capable(CAP_NET_ADMIN
))
510 retv
= xfrm_user_policy(sk
, optname
, optval
, optlen
);
513 #ifdef CONFIG_NETFILTER
515 retv
= nf_setsockopt(sk
, PF_INET6
, optname
, optval
,
531 int ipv6_getsockopt(struct sock
*sk
, int level
, int optname
,
532 char __user
*optval
, int __user
*optlen
)
534 struct ipv6_pinfo
*np
= inet6_sk(sk
);
538 if (level
== SOL_IP
&& sk
->sk_type
!= SOCK_RAW
)
539 return udp_prot
.getsockopt(sk
, level
, optname
, optval
, optlen
);
542 if (get_user(len
, optlen
))
546 if (sk
->sk_protocol
!= IPPROTO_UDP
&&
547 sk
->sk_protocol
!= IPPROTO_TCP
)
549 if (sk
->sk_state
!= TCP_ESTABLISHED
)
555 struct group_filter gsf
;
558 if (len
< GROUP_FILTER_SIZE(0))
560 if (copy_from_user(&gsf
, optval
, GROUP_FILTER_SIZE(0)))
563 err
= ip6_mc_msfget(sk
, &gsf
,
564 (struct group_filter __user
*)optval
, optlen
);
569 case IPV6_PKTOPTIONS
:
574 if (sk
->sk_type
!= SOCK_STREAM
)
577 msg
.msg_control
= optval
;
578 msg
.msg_controllen
= len
;
582 skb
= np
->pktoptions
;
584 atomic_inc(&skb
->users
);
588 int err
= datagram_recv_ctl(sk
, &msg
, skb
);
593 if (np
->rxopt
.bits
.rxinfo
) {
594 struct in6_pktinfo src_info
;
595 src_info
.ipi6_ifindex
= np
->mcast_oif
;
596 ipv6_addr_copy(&src_info
.ipi6_addr
, &np
->daddr
);
597 put_cmsg(&msg
, SOL_IPV6
, IPV6_PKTINFO
, sizeof(src_info
), &src_info
);
599 if (np
->rxopt
.bits
.rxhlim
) {
600 int hlim
= np
->mcast_hops
;
601 put_cmsg(&msg
, SOL_IPV6
, IPV6_HOPLIMIT
, sizeof(hlim
), &hlim
);
604 len
-= msg
.msg_controllen
;
605 return put_user(len
, optlen
);
609 struct dst_entry
*dst
;
612 dst
= sk_dst_get(sk
);
628 val
= np
->rxopt
.bits
.rxinfo
;
632 val
= np
->rxopt
.bits
.rxhlim
;
636 val
= np
->rxopt
.bits
.srcrt
;
640 val
= np
->rxopt
.bits
.hopopts
;
644 val
= np
->rxopt
.bits
.dstopts
;
648 val
= np
->rxopt
.bits
.rxflow
;
651 case IPV6_UNICAST_HOPS
:
655 case IPV6_MULTICAST_HOPS
:
656 val
= np
->mcast_hops
;
659 case IPV6_MULTICAST_LOOP
:
663 case IPV6_MULTICAST_IF
:
667 case IPV6_MTU_DISCOVER
:
675 case IPV6_FLOWINFO_SEND
:
680 #ifdef CONFIG_NETFILTER
682 val
= nf_getsockopt(sk
, PF_INET6
, optname
, optval
,
686 val
= put_user(len
, optlen
);
692 len
= min_t(unsigned int, sizeof(int), len
);
693 if(put_user(len
, optlen
))
695 if(copy_to_user(optval
,&val
,len
))
700 void __init
ipv6_packet_init(void)
702 dev_add_pack(&ipv6_packet_type
);
705 void ipv6_packet_cleanup(void)
707 dev_remove_pack(&ipv6_packet_type
);