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 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
17 * yoshfuji : ensure not to overrun while parsing
19 * Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
20 * YOSHIFUJI Hideaki @USAGI Register inbound extension header
21 * handlers as inet6_protocol{}.
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/netdevice.h>
30 #include <linux/in6.h>
31 #include <linux/icmpv6.h>
38 #include <net/protocol.h>
39 #include <net/transp_v6.h>
40 #include <net/rawv6.h>
41 #include <net/ndisc.h>
42 #include <net/ip6_route.h>
43 #include <net/addrconf.h>
44 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
48 #include <asm/uaccess.h>
50 int ipv6_find_tlv(struct sk_buff *skb, int offset, int type)
52 const unsigned char *nh = skb_network_header(skb);
53 int packet_len = skb->tail - skb->network_header;
54 struct ipv6_opt_hdr *hdr;
57 if (offset + 2 > packet_len)
59 hdr = (struct ipv6_opt_hdr *)(nh + offset);
60 len = ((hdr->hdrlen + 1) << 3);
62 if (offset + len > packet_len)
69 int opttype = nh[offset];
80 optlen = nh[offset + 1] + 2;
92 EXPORT_SYMBOL_GPL(ipv6_find_tlv);
95 * Parsing tlv encoded headers.
97 * Parsing function "func" returns 1, if parsing succeed
98 * and 0, if it failed.
99 * It MUST NOT touch skb->h.
102 struct tlvtype_proc {
104 int (*func)(struct sk_buff *skb, int offset);
107 /*********************
109 *********************/
111 /* An unknown option is detected, decide what to do */
113 static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
115 switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
119 case 1: /* drop packet */
122 case 3: /* Send ICMP if not a multicast address and drop packet */
123 /* Actually, it is redundant check. icmp_send
124 will recheck in any case.
126 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
128 case 2: /* send ICMP PARM PROB regardless and drop packet */
129 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
137 /* Parse tlv encoded option header (hop-by-hop or destination) */
139 static int ip6_parse_tlv(struct tlvtype_proc *procs, struct sk_buff *skb)
141 struct tlvtype_proc *curr;
142 const unsigned char *nh = skb_network_header(skb);
143 int off = skb_network_header_len(skb);
144 int len = (skb_transport_header(skb)[1] + 1) << 3;
146 if (skb_transport_offset(skb) + len > skb_headlen(skb))
153 int optlen = nh[off + 1] + 2;
163 default: /* Other TLV code so scan list */
166 for (curr=procs; curr->type >= 0; curr++) {
167 if (curr->type == nh[off]) {
168 /* type specific length/alignment
169 checks will be performed in the
171 if (curr->func(skb, off) == 0)
176 if (curr->type < 0) {
177 if (ip6_tlvopt_unknown(skb, off) == 0)
192 /*****************************
193 Destination options header.
194 *****************************/
196 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
197 static int ipv6_dest_hao(struct sk_buff *skb, int optoff)
199 struct ipv6_destopt_hao *hao;
200 struct inet6_skb_parm *opt = IP6CB(skb);
201 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
202 struct in6_addr tmp_addr;
206 LIMIT_NETDEBUG(KERN_DEBUG "hao duplicated\n");
209 opt->dsthao = opt->dst1;
212 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
214 if (hao->length != 16) {
216 KERN_DEBUG "hao invalid option length = %d\n", hao->length);
220 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
222 KERN_DEBUG "hao is not an unicast addr: %pI6\n", &hao->addr);
226 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
227 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
228 if (unlikely(ret < 0))
231 if (skb_cloned(skb)) {
232 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
235 /* update all variable using below by copied skbuff */
236 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
238 ipv6h = ipv6_hdr(skb);
241 if (skb->ip_summed == CHECKSUM_COMPLETE)
242 skb->ip_summed = CHECKSUM_NONE;
244 ipv6_addr_copy(&tmp_addr, &ipv6h->saddr);
245 ipv6_addr_copy(&ipv6h->saddr, &hao->addr);
246 ipv6_addr_copy(&hao->addr, &tmp_addr);
248 if (skb->tstamp.tv64 == 0)
249 __net_timestamp(skb);
259 static struct tlvtype_proc tlvprocdestopt_lst[] = {
260 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
262 .type = IPV6_TLV_HAO,
263 .func = ipv6_dest_hao,
269 static int ipv6_destopt_rcv(struct sk_buff *skb)
271 struct inet6_skb_parm *opt = IP6CB(skb);
272 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
275 struct dst_entry *dst;
277 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
278 !pskb_may_pull(skb, (skb_transport_offset(skb) +
279 ((skb_transport_header(skb)[1] + 1) << 3)))) {
280 IP6_INC_STATS_BH(dev_net(skb->dst->dev), ip6_dst_idev(skb->dst),
281 IPSTATS_MIB_INHDRERRORS);
286 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
287 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
291 dst = dst_clone(skb->dst);
292 if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
294 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
296 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
299 opt->nhoff = opt->dst1;
304 IP6_INC_STATS_BH(dev_net(dst->dev),
305 ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
310 /********************************
312 ********************************/
314 static int ipv6_rthdr_rcv(struct sk_buff *skb)
316 struct inet6_skb_parm *opt = IP6CB(skb);
317 struct in6_addr *addr = NULL;
318 struct in6_addr daddr;
319 struct inet6_dev *idev;
321 struct ipv6_rt_hdr *hdr;
322 struct rt0_hdr *rthdr;
323 struct net *net = dev_net(skb->dev);
324 int accept_source_route = net->ipv6.devconf_all->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(net, 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(net, 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(net, 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(net, 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(net, 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(net, 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(net, ip6_dst_idev(skb->dst),
444 IPSTATS_MIB_INADDRERRORS);
448 if (!ipv6_chk_home_addr(dev_net(skb->dst->dev), addr)) {
449 IP6_INC_STATS_BH(net, ip6_dst_idev(skb->dst),
450 IPSTATS_MIB_INADDRERRORS);
460 if (ipv6_addr_is_multicast(addr)) {
461 IP6_INC_STATS_BH(net, 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(net, 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(net, 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);
584 struct net *net = dev_net(skb->dst->dev);
586 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
587 LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
589 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
590 IPSTATS_MIB_INHDRERRORS);
594 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
595 if (pkt_len <= IPV6_MAXPLEN) {
596 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
597 IPSTATS_MIB_INHDRERRORS);
598 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
601 if (ipv6_hdr(skb)->payload_len) {
602 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
603 IPSTATS_MIB_INHDRERRORS);
604 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
608 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
609 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
610 IPSTATS_MIB_INTRUNCATEDPKTS);
614 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
624 static struct tlvtype_proc tlvprochopopt_lst[] = {
626 .type = IPV6_TLV_ROUTERALERT,
630 .type = IPV6_TLV_JUMBO,
631 .func = ipv6_hop_jumbo,
636 int ipv6_parse_hopopts(struct sk_buff *skb)
638 struct inet6_skb_parm *opt = IP6CB(skb);
641 * skb_network_header(skb) is equal to skb->data, and
642 * skb_network_header_len(skb) is always equal to
643 * sizeof(struct ipv6hdr) by definition of
644 * hop-by-hop options.
646 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
647 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
648 ((skb_transport_header(skb)[1] + 1) << 3)))) {
653 opt->hop = sizeof(struct ipv6hdr);
654 if (ip6_parse_tlv(tlvprochopopt_lst, skb)) {
655 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
657 opt->nhoff = sizeof(struct ipv6hdr);
664 * Creating outbound headers.
666 * "build" functions work when skb is filled from head to tail (datagram)
667 * "push" functions work when headers are added from tail to head (tcp)
669 * In both cases we assume, that caller reserved enough room
673 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
674 struct ipv6_rt_hdr *opt,
675 struct in6_addr **addr_p)
677 struct rt0_hdr *phdr, *ihdr;
680 ihdr = (struct rt0_hdr *) opt;
682 phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
683 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
685 hops = ihdr->rt_hdr.hdrlen >> 1;
688 memcpy(phdr->addr, ihdr->addr + 1,
689 (hops - 1) * sizeof(struct in6_addr));
691 ipv6_addr_copy(phdr->addr + (hops - 1), *addr_p);
692 *addr_p = ihdr->addr;
694 phdr->rt_hdr.nexthdr = *proto;
695 *proto = NEXTHDR_ROUTING;
698 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
700 struct ipv6_opt_hdr *h = (struct ipv6_opt_hdr *)skb_push(skb, ipv6_optlen(opt));
702 memcpy(h, opt, ipv6_optlen(opt));
707 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
709 struct in6_addr **daddr)
712 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr);
714 * IPV6_RTHDRDSTOPTS is ignored
715 * unless IPV6_RTHDR is set (RFC3542).
718 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
721 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
724 EXPORT_SYMBOL(ipv6_push_nfrag_opts);
726 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
729 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
732 struct ipv6_txoptions *
733 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
735 struct ipv6_txoptions *opt2;
737 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
739 long dif = (char*)opt2 - (char*)opt;
740 memcpy(opt2, opt, opt->tot_len);
742 *((char**)&opt2->hopopt) += dif;
744 *((char**)&opt2->dst0opt) += dif;
746 *((char**)&opt2->dst1opt) += dif;
748 *((char**)&opt2->srcrt) += dif;
753 EXPORT_SYMBOL_GPL(ipv6_dup_options);
755 static int ipv6_renew_option(void *ohdr,
756 struct ipv6_opt_hdr __user *newopt, int newoptlen,
758 struct ipv6_opt_hdr **hdr,
763 memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
764 *hdr = (struct ipv6_opt_hdr *)*p;
765 *p += CMSG_ALIGN(ipv6_optlen(*(struct ipv6_opt_hdr **)hdr));
769 if (copy_from_user(*p, newopt, newoptlen))
771 *hdr = (struct ipv6_opt_hdr *)*p;
772 if (ipv6_optlen(*(struct ipv6_opt_hdr **)hdr) > newoptlen)
774 *p += CMSG_ALIGN(newoptlen);
780 struct ipv6_txoptions *
781 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
783 struct ipv6_opt_hdr __user *newopt, int newoptlen)
787 struct ipv6_txoptions *opt2;
791 if (newtype != IPV6_HOPOPTS && opt->hopopt)
792 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
793 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
794 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
795 if (newtype != IPV6_RTHDR && opt->srcrt)
796 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
797 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
798 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
801 if (newopt && newoptlen)
802 tot_len += CMSG_ALIGN(newoptlen);
807 tot_len += sizeof(*opt2);
808 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
810 return ERR_PTR(-ENOBUFS);
812 memset(opt2, 0, tot_len);
814 opt2->tot_len = tot_len;
815 p = (char *)(opt2 + 1);
817 err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
818 newtype != IPV6_HOPOPTS,
823 err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
824 newtype != IPV6_RTHDRDSTOPTS,
829 err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
830 newtype != IPV6_RTHDR,
831 (struct ipv6_opt_hdr **)&opt2->srcrt, &p);
835 err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
836 newtype != IPV6_DSTOPTS,
841 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
842 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
843 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
844 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
848 sock_kfree_s(sk, opt2, opt2->tot_len);
852 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
853 struct ipv6_txoptions *opt)
856 * ignore the dest before srcrt unless srcrt is being included.
859 if (opt && opt->dst0opt && !opt->srcrt) {
860 if (opt_space != opt) {
861 memcpy(opt_space, opt, sizeof(*opt_space));
864 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);