2 * Extension Header handling for IPv6
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Andi Kleen <ak@muc.de>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
10 * $Id: exthdrs.c,v 1.13 2001/06/19 15:58:56 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.
19 * yoshfuji : ensure not to overrun while parsing
21 * Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
22 * YOSHIFUJI Hideaki @USAGI Register inbound extension header
23 * handlers as inet6_protocol{}.
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
29 #include <linux/sockios.h>
30 #include <linux/net.h>
31 #include <linux/netdevice.h>
32 #include <linux/in6.h>
33 #include <linux/icmpv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/rawv6.h>
43 #include <net/ndisc.h>
44 #include <net/ip6_route.h>
45 #include <net/addrconf.h>
46 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
50 #include <asm/uaccess.h>
52 int ipv6_find_tlv(struct sk_buff *skb, int offset, int type)
54 const unsigned char *nh = skb_network_header(skb);
55 int packet_len = skb->tail - skb->network_header;
56 struct ipv6_opt_hdr *hdr;
59 if (offset + 2 > packet_len)
61 hdr = (struct ipv6_opt_hdr *)(nh + offset);
62 len = ((hdr->hdrlen + 1) << 3);
64 if (offset + len > packet_len)
71 int opttype = nh[offset];
82 optlen = nh[offset + 1] + 2;
94 EXPORT_SYMBOL_GPL(ipv6_find_tlv);
97 * Parsing tlv encoded headers.
99 * Parsing function "func" returns 1, if parsing succeed
100 * and 0, if it failed.
101 * It MUST NOT touch skb->h.
104 struct tlvtype_proc {
106 int (*func)(struct sk_buff *skb, int offset);
109 /*********************
111 *********************/
113 /* An unknown option is detected, decide what to do */
115 static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
117 switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
121 case 1: /* drop packet */
124 case 3: /* Send ICMP if not a multicast address and drop packet */
125 /* Actually, it is redundant check. icmp_send
126 will recheck in any case.
128 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
130 case 2: /* send ICMP PARM PROB regardless and drop packet */
131 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
139 /* Parse tlv encoded option header (hop-by-hop or destination) */
141 static int ip6_parse_tlv(struct tlvtype_proc *procs, struct sk_buff *skb)
143 struct tlvtype_proc *curr;
144 const unsigned char *nh = skb_network_header(skb);
145 int off = skb_network_header_len(skb);
146 int len = (skb_transport_header(skb)[1] + 1) << 3;
148 if (skb_transport_offset(skb) + len > skb_headlen(skb))
155 int optlen = nh[off + 1] + 2;
165 default: /* Other TLV code so scan list */
168 for (curr=procs; curr->type >= 0; curr++) {
169 if (curr->type == nh[off]) {
170 /* type specific length/alignment
171 checks will be performed in the
173 if (curr->func(skb, off) == 0)
178 if (curr->type < 0) {
179 if (ip6_tlvopt_unknown(skb, off) == 0)
194 /*****************************
195 Destination options header.
196 *****************************/
198 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
199 static int ipv6_dest_hao(struct sk_buff *skb, int optoff)
201 struct ipv6_destopt_hao *hao;
202 struct inet6_skb_parm *opt = IP6CB(skb);
203 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
204 struct in6_addr tmp_addr;
208 LIMIT_NETDEBUG(KERN_DEBUG "hao duplicated\n");
211 opt->dsthao = opt->dst1;
214 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
216 if (hao->length != 16) {
218 KERN_DEBUG "hao invalid option length = %d\n", hao->length);
222 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
224 KERN_DEBUG "hao is not an unicast addr: " NIP6_FMT "\n", NIP6(hao->addr));
228 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
229 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
230 if (unlikely(ret < 0))
233 if (skb_cloned(skb)) {
234 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
237 /* update all variable using below by copied skbuff */
238 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
240 ipv6h = ipv6_hdr(skb);
243 if (skb->ip_summed == CHECKSUM_COMPLETE)
244 skb->ip_summed = CHECKSUM_NONE;
246 ipv6_addr_copy(&tmp_addr, &ipv6h->saddr);
247 ipv6_addr_copy(&ipv6h->saddr, &hao->addr);
248 ipv6_addr_copy(&hao->addr, &tmp_addr);
250 if (skb->tstamp.tv64 == 0)
251 __net_timestamp(skb);
261 static struct tlvtype_proc tlvprocdestopt_lst[] = {
262 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
264 .type = IPV6_TLV_HAO,
265 .func = ipv6_dest_hao,
271 static int ipv6_destopt_rcv(struct sk_buff *skb)
273 struct inet6_skb_parm *opt = IP6CB(skb);
274 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
277 struct dst_entry *dst;
279 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
280 !pskb_may_pull(skb, (skb_transport_offset(skb) +
281 ((skb_transport_header(skb)[1] + 1) << 3)))) {
282 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
283 IPSTATS_MIB_INHDRERRORS);
288 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
289 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
293 dst = dst_clone(skb->dst);
294 if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
296 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
298 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
301 opt->nhoff = opt->dst1;
306 IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
311 /********************************
313 ********************************/
315 static int ipv6_rthdr_rcv(struct sk_buff *skb)
317 struct inet6_skb_parm *opt = IP6CB(skb);
318 struct in6_addr *addr = NULL;
319 struct in6_addr daddr;
320 struct inet6_dev *idev;
322 struct ipv6_rt_hdr *hdr;
323 struct rt0_hdr *rthdr;
324 int accept_source_route = ipv6_devconf.accept_source_route;
326 idev = in6_dev_get(skb->dev);
328 if (accept_source_route > idev->cnf.accept_source_route)
329 accept_source_route = idev->cnf.accept_source_route;
333 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
334 !pskb_may_pull(skb, (skb_transport_offset(skb) +
335 ((skb_transport_header(skb)[1] + 1) << 3)))) {
336 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
337 IPSTATS_MIB_INHDRERRORS);
342 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
344 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
345 skb->pkt_type != PACKET_HOST) {
346 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
347 IPSTATS_MIB_INADDRERRORS);
353 if (hdr->segments_left == 0) {
355 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
356 case IPV6_SRCRT_TYPE_2:
357 /* Silently discard type 2 header unless it was
361 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
362 IPSTATS_MIB_INADDRERRORS);
372 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
373 skb->transport_header += (hdr->hdrlen + 1) << 3;
374 opt->dst0 = opt->dst1;
376 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
381 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
382 case IPV6_SRCRT_TYPE_2:
383 if (accept_source_route < 0)
385 /* Silently discard invalid RTH type 2 */
386 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
387 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
388 IPSTATS_MIB_INHDRERRORS);
399 * This is the routing header forwarding algorithm from
403 n = hdr->hdrlen >> 1;
405 if (hdr->segments_left > n) {
406 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
407 IPSTATS_MIB_INHDRERRORS);
408 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
409 ((&hdr->segments_left) -
410 skb_network_header(skb)));
414 /* We are about to mangle packet header. Be careful!
415 Do not damage packets queued somewhere.
417 if (skb_cloned(skb)) {
418 /* the copy is a forwarded packet */
419 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
420 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
421 IPSTATS_MIB_OUTDISCARDS);
425 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
428 if (skb->ip_summed == CHECKSUM_COMPLETE)
429 skb->ip_summed = CHECKSUM_NONE;
431 i = n - --hdr->segments_left;
433 rthdr = (struct rt0_hdr *) hdr;
438 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
439 case IPV6_SRCRT_TYPE_2:
440 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
441 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
442 IPPROTO_ROUTING) < 0) {
443 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
444 IPSTATS_MIB_INADDRERRORS);
448 if (!ipv6_chk_home_addr(&init_net, addr)) {
449 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
450 IPSTATS_MIB_INADDRERRORS);
460 if (ipv6_addr_is_multicast(addr)) {
461 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
462 IPSTATS_MIB_INADDRERRORS);
467 ipv6_addr_copy(&daddr, addr);
468 ipv6_addr_copy(addr, &ipv6_hdr(skb)->daddr);
469 ipv6_addr_copy(&ipv6_hdr(skb)->daddr, &daddr);
471 dst_release(xchg(&skb->dst, NULL));
472 ip6_route_input(skb);
473 if (skb->dst->error) {
474 skb_push(skb, skb->data - skb_network_header(skb));
479 if (skb->dst->dev->flags&IFF_LOOPBACK) {
480 if (ipv6_hdr(skb)->hop_limit <= 1) {
481 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
482 IPSTATS_MIB_INHDRERRORS);
483 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
488 ipv6_hdr(skb)->hop_limit--;
492 skb_push(skb, skb->data - skb_network_header(skb));
497 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
498 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
499 (&hdr->type) - skb_network_header(skb));
503 static struct inet6_protocol rthdr_protocol = {
504 .handler = ipv6_rthdr_rcv,
505 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_GSO_EXTHDR,
508 static struct inet6_protocol destopt_protocol = {
509 .handler = ipv6_destopt_rcv,
510 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_GSO_EXTHDR,
513 static struct inet6_protocol nodata_protocol = {
514 .handler = dst_discard,
515 .flags = INET6_PROTO_NOPOLICY,
518 int __init ipv6_exthdrs_init(void)
522 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
526 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
530 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
537 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
539 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
543 void ipv6_exthdrs_exit(void)
545 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
546 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
547 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
550 /**********************************
552 **********************************/
555 * Note: we cannot rely on skb->dst before we assign it in ip6_route_input().
557 static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
559 return skb->dst ? ip6_dst_idev(skb->dst) : __in6_dev_get(skb->dev);
562 /* Router Alert as of RFC 2711 */
564 static int ipv6_hop_ra(struct sk_buff *skb, int optoff)
566 const unsigned char *nh = skb_network_header(skb);
568 if (nh[optoff + 1] == 2) {
569 IP6CB(skb)->ra = optoff;
572 LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_ra: wrong RA length %d\n",
580 static int ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
582 const unsigned char *nh = skb_network_header(skb);
585 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
586 LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
588 IP6_INC_STATS_BH(ipv6_skb_idev(skb),
589 IPSTATS_MIB_INHDRERRORS);
593 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
594 if (pkt_len <= IPV6_MAXPLEN) {
595 IP6_INC_STATS_BH(ipv6_skb_idev(skb), IPSTATS_MIB_INHDRERRORS);
596 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
599 if (ipv6_hdr(skb)->payload_len) {
600 IP6_INC_STATS_BH(ipv6_skb_idev(skb), IPSTATS_MIB_INHDRERRORS);
601 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
605 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
606 IP6_INC_STATS_BH(ipv6_skb_idev(skb), IPSTATS_MIB_INTRUNCATEDPKTS);
610 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
620 static struct tlvtype_proc tlvprochopopt_lst[] = {
622 .type = IPV6_TLV_ROUTERALERT,
626 .type = IPV6_TLV_JUMBO,
627 .func = ipv6_hop_jumbo,
632 int ipv6_parse_hopopts(struct sk_buff *skb)
634 struct inet6_skb_parm *opt = IP6CB(skb);
637 * skb_network_header(skb) is equal to skb->data, and
638 * skb_network_header_len(skb) is always equal to
639 * sizeof(struct ipv6hdr) by definition of
640 * hop-by-hop options.
642 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
643 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
644 ((skb_transport_header(skb)[1] + 1) << 3)))) {
649 opt->hop = sizeof(struct ipv6hdr);
650 if (ip6_parse_tlv(tlvprochopopt_lst, skb)) {
651 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
653 opt->nhoff = sizeof(struct ipv6hdr);
660 * Creating outbound headers.
662 * "build" functions work when skb is filled from head to tail (datagram)
663 * "push" functions work when headers are added from tail to head (tcp)
665 * In both cases we assume, that caller reserved enough room
669 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
670 struct ipv6_rt_hdr *opt,
671 struct in6_addr **addr_p)
673 struct rt0_hdr *phdr, *ihdr;
676 ihdr = (struct rt0_hdr *) opt;
678 phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
679 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
681 hops = ihdr->rt_hdr.hdrlen >> 1;
684 memcpy(phdr->addr, ihdr->addr + 1,
685 (hops - 1) * sizeof(struct in6_addr));
687 ipv6_addr_copy(phdr->addr + (hops - 1), *addr_p);
688 *addr_p = ihdr->addr;
690 phdr->rt_hdr.nexthdr = *proto;
691 *proto = NEXTHDR_ROUTING;
694 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
696 struct ipv6_opt_hdr *h = (struct ipv6_opt_hdr *)skb_push(skb, ipv6_optlen(opt));
698 memcpy(h, opt, ipv6_optlen(opt));
703 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
705 struct in6_addr **daddr)
708 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr);
710 * IPV6_RTHDRDSTOPTS is ignored
711 * unless IPV6_RTHDR is set (RFC3542).
714 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
717 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
720 EXPORT_SYMBOL(ipv6_push_nfrag_opts);
722 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
725 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
728 struct ipv6_txoptions *
729 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
731 struct ipv6_txoptions *opt2;
733 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
735 long dif = (char*)opt2 - (char*)opt;
736 memcpy(opt2, opt, opt->tot_len);
738 *((char**)&opt2->hopopt) += dif;
740 *((char**)&opt2->dst0opt) += dif;
742 *((char**)&opt2->dst1opt) += dif;
744 *((char**)&opt2->srcrt) += dif;
749 EXPORT_SYMBOL_GPL(ipv6_dup_options);
751 static int ipv6_renew_option(void *ohdr,
752 struct ipv6_opt_hdr __user *newopt, int newoptlen,
754 struct ipv6_opt_hdr **hdr,
759 memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
760 *hdr = (struct ipv6_opt_hdr *)*p;
761 *p += CMSG_ALIGN(ipv6_optlen(*(struct ipv6_opt_hdr **)hdr));
765 if (copy_from_user(*p, newopt, newoptlen))
767 *hdr = (struct ipv6_opt_hdr *)*p;
768 if (ipv6_optlen(*(struct ipv6_opt_hdr **)hdr) > newoptlen)
770 *p += CMSG_ALIGN(newoptlen);
776 struct ipv6_txoptions *
777 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
779 struct ipv6_opt_hdr __user *newopt, int newoptlen)
783 struct ipv6_txoptions *opt2;
787 if (newtype != IPV6_HOPOPTS && opt->hopopt)
788 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
789 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
790 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
791 if (newtype != IPV6_RTHDR && opt->srcrt)
792 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
793 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
794 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
797 if (newopt && newoptlen)
798 tot_len += CMSG_ALIGN(newoptlen);
803 tot_len += sizeof(*opt2);
804 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
806 return ERR_PTR(-ENOBUFS);
808 memset(opt2, 0, tot_len);
810 opt2->tot_len = tot_len;
811 p = (char *)(opt2 + 1);
813 err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
814 newtype != IPV6_HOPOPTS,
819 err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
820 newtype != IPV6_RTHDRDSTOPTS,
825 err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
826 newtype != IPV6_RTHDR,
827 (struct ipv6_opt_hdr **)&opt2->srcrt, &p);
831 err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
832 newtype != IPV6_DSTOPTS,
837 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
838 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
839 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
840 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
844 sock_kfree_s(sk, opt2, opt2->tot_len);
848 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
849 struct ipv6_txoptions *opt)
852 * ignore the dest before srcrt unless srcrt is being included.
855 if (opt && opt->dst0opt && !opt->srcrt) {
856 if (opt_space != opt) {
857 memcpy(opt_space, opt, sizeof(*opt_space));
860 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);