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);
167 * Sock is moving from IPv6 to IPv4 (sk_prot), so
168 * remove it from the refcnt debug socks count in the
171 sk_refcnt_debug_dec(sk);
173 if (sk->sk_protocol == IPPROTO_TCP) {
174 struct tcp_sock *tp = tcp_sk(sk);
177 sock_prot_dec_use(sk->sk_prot);
178 sock_prot_inc_use(&tcp_prot);
180 sk->sk_prot = &tcp_prot;
181 tp->af_specific = &ipv4_specific;
182 sk->sk_socket->ops = &inet_stream_ops;
183 sk->sk_family = PF_INET;
184 tcp_sync_mss(sk, tp->pmtu_cookie);
187 sock_prot_dec_use(sk->sk_prot);
188 sock_prot_inc_use(&udp_prot);
190 sk->sk_prot = &udp_prot;
191 sk->sk_socket->ops = &inet_dgram_ops;
192 sk->sk_family = PF_INET;
194 opt = xchg(&np->opt, NULL);
196 sock_kfree_s(sk, opt, opt->tot_len);
197 pktopt = xchg(&np->pktoptions, NULL);
201 sk->sk_destruct = inet_sock_destruct;
203 * ... and add it to the refcnt debug socks count
204 * in the new family. -acme
206 sk_refcnt_debug_inc(sk);
207 module_put(THIS_MODULE);
214 if (inet_sk(sk)->num)
216 np->ipv6only = valbool;
221 np->rxopt.bits.rxinfo = valbool;
226 np->rxopt.bits.rxhlim = valbool;
231 if (val < 0 || val > 2)
233 np->rxopt.bits.srcrt = val;
238 np->rxopt.bits.hopopts = valbool;
243 np->rxopt.bits.dstopts = valbool;
248 np->rxopt.bits.rxflow = valbool;
252 case IPV6_PKTOPTIONS:
254 struct ipv6_txoptions *opt = NULL;
259 fl.fl6_flowlabel = 0;
260 fl.oif = sk->sk_bound_dev_if;
265 /* 1K is probably excessive
266 * 1K is surely not enough, 2K per standard header is 16K.
269 if (optlen > 64*1024)
272 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
277 memset(opt, 0, sizeof(*opt));
278 opt->tot_len = sizeof(*opt) + optlen;
280 if (copy_from_user(opt+1, optval, optlen))
283 msg.msg_controllen = optlen;
284 msg.msg_control = (void*)(opt+1);
286 retv = datagram_send_ctl(&msg, &fl, opt, &junk);
291 if (sk->sk_type == SOCK_STREAM) {
293 struct tcp_sock *tp = tcp_sk(sk);
294 if (!((1 << sk->sk_state) &
295 (TCPF_LISTEN | TCPF_CLOSE))
296 && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
297 tp->ext_header_len = opt->opt_flen + opt->opt_nflen;
298 tcp_sync_mss(sk, tp->pmtu_cookie);
301 opt = xchg(&np->opt, opt);
304 write_lock(&sk->sk_dst_lock);
305 opt = xchg(&np->opt, opt);
306 write_unlock(&sk->sk_dst_lock);
312 sock_kfree_s(sk, opt, opt->tot_len);
315 case IPV6_UNICAST_HOPS:
316 if (val > 255 || val < -1)
322 case IPV6_MULTICAST_HOPS:
323 if (sk->sk_type == SOCK_STREAM)
325 if (val > 255 || val < -1)
327 np->mcast_hops = val;
331 case IPV6_MULTICAST_LOOP:
332 np->mc_loop = valbool;
336 case IPV6_MULTICAST_IF:
337 if (sk->sk_type == SOCK_STREAM)
339 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
342 if (__dev_get_by_index(val) == NULL) {
349 case IPV6_ADD_MEMBERSHIP:
350 case IPV6_DROP_MEMBERSHIP:
352 struct ipv6_mreq mreq;
355 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
358 if (optname == IPV6_ADD_MEMBERSHIP)
359 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
361 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
364 case IPV6_JOIN_ANYCAST:
365 case IPV6_LEAVE_ANYCAST:
367 struct ipv6_mreq mreq;
369 if (optlen != sizeof(struct ipv6_mreq))
373 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
376 if (optname == IPV6_JOIN_ANYCAST)
377 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
379 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
382 case MCAST_JOIN_GROUP:
383 case MCAST_LEAVE_GROUP:
385 struct group_req greq;
386 struct sockaddr_in6 *psin6;
389 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
391 if (greq.gr_group.ss_family != AF_INET6) {
392 retv = -EADDRNOTAVAIL;
395 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
396 if (optname == MCAST_JOIN_GROUP)
397 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
400 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
404 case MCAST_JOIN_SOURCE_GROUP:
405 case MCAST_LEAVE_SOURCE_GROUP:
406 case MCAST_BLOCK_SOURCE:
407 case MCAST_UNBLOCK_SOURCE:
409 struct group_source_req greqs;
412 if (optlen != sizeof(struct group_source_req))
414 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
418 if (greqs.gsr_group.ss_family != AF_INET6 ||
419 greqs.gsr_source.ss_family != AF_INET6) {
420 retv = -EADDRNOTAVAIL;
423 if (optname == MCAST_BLOCK_SOURCE) {
424 omode = MCAST_EXCLUDE;
426 } else if (optname == MCAST_UNBLOCK_SOURCE) {
427 omode = MCAST_EXCLUDE;
429 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
430 struct sockaddr_in6 *psin6;
432 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
433 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
435 /* prior join w/ different source is ok */
436 if (retv && retv != -EADDRINUSE)
438 omode = MCAST_INCLUDE;
440 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
441 omode = MCAST_INCLUDE;
444 retv = ip6_mc_source(add, omode, sk, &greqs);
449 extern int sysctl_optmem_max;
450 extern int sysctl_mld_max_msf;
451 struct group_filter *gsf;
453 if (optlen < GROUP_FILTER_SIZE(0))
455 if (optlen > sysctl_optmem_max) {
459 gsf = (struct group_filter *)kmalloc(optlen,GFP_KERNEL);
465 if (copy_from_user(gsf, optval, optlen)) {
469 /* numsrc >= (4G-140)/128 overflow in 32 bits */
470 if (gsf->gf_numsrc >= 0x1ffffffU ||
471 gsf->gf_numsrc > sysctl_mld_max_msf) {
476 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
481 retv = ip6_mc_msfilter(sk, gsf);
486 case IPV6_ROUTER_ALERT:
487 retv = ip6_ra_control(sk, val, NULL);
489 case IPV6_MTU_DISCOVER:
496 if (val && val < IPV6_MIN_MTU)
502 np->recverr = valbool;
504 skb_queue_purge(&sk->sk_error_queue);
507 case IPV6_FLOWINFO_SEND:
508 np->sndflow = valbool;
511 case IPV6_FLOWLABEL_MGR:
512 retv = ipv6_flowlabel_opt(sk, optval, optlen);
514 case IPV6_IPSEC_POLICY:
515 case IPV6_XFRM_POLICY:
517 if (!capable(CAP_NET_ADMIN))
519 retv = xfrm_user_policy(sk, optname, optval, optlen);
522 #ifdef CONFIG_NETFILTER
524 retv = nf_setsockopt(sk, PF_INET6, optname, optval,
540 int ipv6_getsockopt(struct sock *sk, int level, int optname,
541 char __user *optval, int __user *optlen)
543 struct ipv6_pinfo *np = inet6_sk(sk);
547 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
548 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
551 if (get_user(len, optlen))
555 if (sk->sk_protocol != IPPROTO_UDP &&
556 sk->sk_protocol != IPPROTO_TCP)
558 if (sk->sk_state != TCP_ESTABLISHED)
564 struct group_filter gsf;
567 if (len < GROUP_FILTER_SIZE(0))
569 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
572 err = ip6_mc_msfget(sk, &gsf,
573 (struct group_filter __user *)optval, optlen);
578 case IPV6_PKTOPTIONS:
583 if (sk->sk_type != SOCK_STREAM)
586 msg.msg_control = optval;
587 msg.msg_controllen = len;
591 skb = np->pktoptions;
593 atomic_inc(&skb->users);
597 int err = datagram_recv_ctl(sk, &msg, skb);
602 if (np->rxopt.bits.rxinfo) {
603 struct in6_pktinfo src_info;
604 src_info.ipi6_ifindex = np->mcast_oif;
605 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
606 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
608 if (np->rxopt.bits.rxhlim) {
609 int hlim = np->mcast_hops;
610 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
613 len -= msg.msg_controllen;
614 return put_user(len, optlen);
618 struct dst_entry *dst;
621 dst = sk_dst_get(sk);
637 val = np->rxopt.bits.rxinfo;
641 val = np->rxopt.bits.rxhlim;
645 val = np->rxopt.bits.srcrt;
649 val = np->rxopt.bits.hopopts;
653 val = np->rxopt.bits.dstopts;
657 val = np->rxopt.bits.rxflow;
660 case IPV6_UNICAST_HOPS:
664 case IPV6_MULTICAST_HOPS:
665 val = np->mcast_hops;
668 case IPV6_MULTICAST_LOOP:
672 case IPV6_MULTICAST_IF:
676 case IPV6_MTU_DISCOVER:
684 case IPV6_FLOWINFO_SEND:
689 #ifdef CONFIG_NETFILTER
691 val = nf_getsockopt(sk, PF_INET6, optname, optval,
695 val = put_user(len, optlen);
701 len = min_t(unsigned int, sizeof(int), len);
702 if(put_user(len, optlen))
704 if(copy_to_user(optval,&val,len))
709 void __init ipv6_packet_init(void)
711 dev_add_pack(&ipv6_packet_type);
714 void ipv6_packet_cleanup(void)
716 dev_remove_pack(&ipv6_packet_type);