1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Extension Header handling for IPv6
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Andi Kleen <ak@muc.de>
9 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
13 * yoshfuji : ensure not to overrun while parsing
15 * Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
16 * YOSHIFUJI Hideaki @USAGI Register inbound extension header
17 * handlers as inet6_protocol{}.
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/netdevice.h>
26 #include <linux/in6.h>
27 #include <linux/icmpv6.h>
28 #include <linux/slab.h>
29 #include <linux/export.h>
36 #include <net/protocol.h>
37 #include <net/transp_v6.h>
38 #include <net/rawv6.h>
39 #include <net/ndisc.h>
40 #include <net/ip6_route.h>
41 #include <net/addrconf.h>
42 #include <net/calipso.h>
43 #if IS_ENABLED(CONFIG_IPV6_MIP6)
46 #include <linux/seg6.h>
48 #ifdef CONFIG_IPV6_SEG6_HMAC
49 #include <net/seg6_hmac.h>
53 #include <linux/uaccess.h>
56 * Parsing tlv encoded headers.
58 * Parsing function "func" returns true, if parsing succeed
59 * and false, if it failed.
60 * It MUST NOT touch skb->h.
65 bool (*func)(struct sk_buff *skb, int offset);
68 /*********************
70 *********************/
72 /* An unknown option is detected, decide what to do */
74 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff,
75 bool disallow_unknowns)
77 if (disallow_unknowns) {
78 /* If unknown TLVs are disallowed by configuration
79 * then always silently drop packet. Note this also
80 * means no ICMP parameter problem is sent which
81 * could be a good property to mitigate a reflection DOS
88 switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
92 case 1: /* drop packet */
95 case 3: /* Send ICMP if not a multicast address and drop packet */
96 /* Actually, it is redundant check. icmp_send
97 will recheck in any case.
99 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
102 case 2: /* send ICMP PARM PROB regardless and drop packet */
103 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
112 /* Parse tlv encoded option header (hop-by-hop or destination) */
114 static bool ip6_parse_tlv(const struct tlvtype_proc *procs,
118 int len = (skb_transport_header(skb)[1] + 1) << 3;
119 const unsigned char *nh = skb_network_header(skb);
120 int off = skb_network_header_len(skb);
121 const struct tlvtype_proc *curr;
122 bool disallow_unknowns = false;
126 if (unlikely(max_count < 0)) {
127 disallow_unknowns = true;
128 max_count = -max_count;
131 if (skb_transport_offset(skb) + len > skb_headlen(skb))
138 int optlen = nh[off + 1] + 2;
150 /* RFC 2460 states that the purpose of PadN is
151 * to align the containing header to multiples
152 * of 8. 7 is therefore the highest valid value.
153 * See also RFC 4942, Section 2.1.9.5.
158 /* RFC 4942 recommends receiving hosts to
159 * actively check PadN payload to contain
162 for (i = 2; i < optlen; i++) {
163 if (nh[off + i] != 0)
168 default: /* Other TLV code so scan list */
173 if (tlv_count > max_count)
176 for (curr = procs; curr->type >= 0; curr++) {
177 if (curr->type == nh[off]) {
178 /* type specific length/alignment
179 checks will be performed in the
181 if (curr->func(skb, off) == false)
186 if (curr->type < 0 &&
187 !ip6_tlvopt_unknown(skb, off, disallow_unknowns))
204 /*****************************
205 Destination options header.
206 *****************************/
208 #if IS_ENABLED(CONFIG_IPV6_MIP6)
209 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
211 struct ipv6_destopt_hao *hao;
212 struct inet6_skb_parm *opt = IP6CB(skb);
213 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
217 net_dbg_ratelimited("hao duplicated\n");
220 opt->dsthao = opt->dst1;
223 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
225 if (hao->length != 16) {
226 net_dbg_ratelimited("hao invalid option length = %d\n",
231 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
232 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
237 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
238 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
239 if (unlikely(ret < 0))
242 if (skb_cloned(skb)) {
243 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
246 /* update all variable using below by copied skbuff */
247 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
249 ipv6h = ipv6_hdr(skb);
252 if (skb->ip_summed == CHECKSUM_COMPLETE)
253 skb->ip_summed = CHECKSUM_NONE;
255 swap(ipv6h->saddr, hao->addr);
257 if (skb->tstamp == 0)
258 __net_timestamp(skb);
268 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
269 #if IS_ENABLED(CONFIG_IPV6_MIP6)
271 .type = IPV6_TLV_HAO,
272 .func = ipv6_dest_hao,
278 static int ipv6_destopt_rcv(struct sk_buff *skb)
280 struct inet6_dev *idev = __in6_dev_get(skb->dev);
281 struct inet6_skb_parm *opt = IP6CB(skb);
282 #if IS_ENABLED(CONFIG_IPV6_MIP6)
285 struct dst_entry *dst = skb_dst(skb);
286 struct net *net = dev_net(skb->dev);
289 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
290 !pskb_may_pull(skb, (skb_transport_offset(skb) +
291 ((skb_transport_header(skb)[1] + 1) << 3)))) {
292 __IP6_INC_STATS(dev_net(dst->dev), idev,
293 IPSTATS_MIB_INHDRERRORS);
299 extlen = (skb_transport_header(skb)[1] + 1) << 3;
300 if (extlen > net->ipv6.sysctl.max_dst_opts_len)
303 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
304 #if IS_ENABLED(CONFIG_IPV6_MIP6)
308 if (ip6_parse_tlv(tlvprocdestopt_lst, skb,
309 init_net.ipv6.sysctl.max_dst_opts_cnt)) {
310 skb->transport_header += extlen;
312 #if IS_ENABLED(CONFIG_IPV6_MIP6)
315 opt->nhoff = opt->dst1;
320 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
324 static void seg6_update_csum(struct sk_buff *skb)
326 struct ipv6_sr_hdr *hdr;
327 struct in6_addr *addr;
330 /* srh is at transport offset and seg_left is already decremented
331 * but daddr is not yet updated with next segment
334 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
335 addr = hdr->segments + hdr->segments_left;
337 hdr->segments_left++;
338 from = *(__be32 *)hdr;
340 hdr->segments_left--;
343 /* update skb csum with diff resulting from seg_left decrement */
345 update_csum_diff4(skb, from, to);
347 /* compute csum diff between current and next segment and update */
349 update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
353 static int ipv6_srh_rcv(struct sk_buff *skb)
355 struct inet6_skb_parm *opt = IP6CB(skb);
356 struct net *net = dev_net(skb->dev);
357 struct ipv6_sr_hdr *hdr;
358 struct inet6_dev *idev;
359 struct in6_addr *addr;
362 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
364 idev = __in6_dev_get(skb->dev);
366 accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
367 if (accept_seg6 > idev->cnf.seg6_enabled)
368 accept_seg6 = idev->cnf.seg6_enabled;
375 #ifdef CONFIG_IPV6_SEG6_HMAC
376 if (!seg6_hmac_validate_skb(skb)) {
383 if (hdr->segments_left == 0) {
384 if (hdr->nexthdr == NEXTHDR_IPV6 || hdr->nexthdr == NEXTHDR_IPV4) {
385 int offset = (hdr->hdrlen + 1) << 3;
387 skb_postpull_rcsum(skb, skb_network_header(skb),
388 skb_network_header_len(skb));
390 if (!pskb_pull(skb, offset)) {
394 skb_postpull_rcsum(skb, skb_transport_header(skb),
397 skb_reset_network_header(skb);
398 skb_reset_transport_header(skb);
399 skb->encapsulation = 0;
400 if (hdr->nexthdr == NEXTHDR_IPV4)
401 skb->protocol = htons(ETH_P_IP);
402 __skb_tunnel_rx(skb, skb->dev, net);
408 opt->srcrt = skb_network_header_len(skb);
409 opt->lastopt = opt->srcrt;
410 skb->transport_header += (hdr->hdrlen + 1) << 3;
411 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
416 if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
417 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
418 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
419 ((&hdr->segments_left) -
420 skb_network_header(skb)));
424 if (skb_cloned(skb)) {
425 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
426 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
427 IPSTATS_MIB_OUTDISCARDS);
433 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
435 hdr->segments_left--;
436 addr = hdr->segments + hdr->segments_left;
438 skb_push(skb, sizeof(struct ipv6hdr));
440 if (skb->ip_summed == CHECKSUM_COMPLETE)
441 seg6_update_csum(skb);
443 ipv6_hdr(skb)->daddr = *addr;
447 ip6_route_input(skb);
449 if (skb_dst(skb)->error) {
454 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
455 if (ipv6_hdr(skb)->hop_limit <= 1) {
456 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
457 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
458 ICMPV6_EXC_HOPLIMIT, 0);
462 ipv6_hdr(skb)->hop_limit--;
464 skb_pull(skb, sizeof(struct ipv6hdr));
473 static int ipv6_rpl_srh_rcv(struct sk_buff *skb)
475 struct ipv6_rpl_sr_hdr *hdr, *ohdr, *chdr;
476 struct inet6_skb_parm *opt = IP6CB(skb);
477 struct net *net = dev_net(skb->dev);
478 struct inet6_dev *idev;
479 struct ipv6hdr *oldhdr;
480 struct in6_addr addr;
487 idev = __in6_dev_get(skb->dev);
489 accept_rpl_seg = net->ipv6.devconf_all->rpl_seg_enabled;
490 if (accept_rpl_seg > idev->cnf.rpl_seg_enabled)
491 accept_rpl_seg = idev->cnf.rpl_seg_enabled;
493 if (!accept_rpl_seg) {
499 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
501 if (hdr->segments_left == 0) {
502 if (hdr->nexthdr == NEXTHDR_IPV6) {
503 int offset = (hdr->hdrlen + 1) << 3;
505 skb_postpull_rcsum(skb, skb_network_header(skb),
506 skb_network_header_len(skb));
508 if (!pskb_pull(skb, offset)) {
512 skb_postpull_rcsum(skb, skb_transport_header(skb),
515 skb_reset_network_header(skb);
516 skb_reset_transport_header(skb);
517 skb->encapsulation = 0;
519 __skb_tunnel_rx(skb, skb->dev, net);
525 opt->srcrt = skb_network_header_len(skb);
526 opt->lastopt = opt->srcrt;
527 skb->transport_header += (hdr->hdrlen + 1) << 3;
528 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
533 if (!pskb_may_pull(skb, sizeof(*hdr))) {
538 n = (hdr->hdrlen << 3) - hdr->pad - (16 - hdr->cmpre);
539 r = do_div(n, (16 - hdr->cmpri));
540 /* checks if calculation was without remainder and n fits into
541 * unsigned char which is segments_left field. Should not be
544 if (r || (n + 1) > 255) {
549 if (hdr->segments_left > n + 1) {
550 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
551 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
552 ((&hdr->segments_left) -
553 skb_network_header(skb)));
557 if (skb_cloned(skb)) {
558 if (pskb_expand_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE, 0,
560 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
561 IPSTATS_MIB_OUTDISCARDS);
566 err = skb_cow_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE);
573 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
575 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
581 hdr->segments_left--;
582 i = n - hdr->segments_left;
584 buf = kcalloc(struct_size(hdr, segments.addr, n + 2), 2, GFP_ATOMIC);
585 if (unlikely(!buf)) {
590 ohdr = (struct ipv6_rpl_sr_hdr *)buf;
591 ipv6_rpl_srh_decompress(ohdr, hdr, &ipv6_hdr(skb)->daddr, n);
592 chdr = (struct ipv6_rpl_sr_hdr *)(buf + ((ohdr->hdrlen + 1) << 3));
594 if ((ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST) ||
595 (ipv6_addr_type(&ohdr->rpl_segaddr[i]) & IPV6_ADDR_MULTICAST)) {
601 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
603 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
609 addr = ipv6_hdr(skb)->daddr;
610 ipv6_hdr(skb)->daddr = ohdr->rpl_segaddr[i];
611 ohdr->rpl_segaddr[i] = addr;
613 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
615 oldhdr = ipv6_hdr(skb);
617 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
618 skb_postpull_rcsum(skb, oldhdr,
619 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
620 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
621 skb_reset_network_header(skb);
622 skb_mac_header_rebuild(skb);
623 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
625 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
626 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
628 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
629 skb_postpush_rcsum(skb, ipv6_hdr(skb),
630 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
636 ip6_route_input(skb);
638 if (skb_dst(skb)->error) {
643 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
644 if (ipv6_hdr(skb)->hop_limit <= 1) {
645 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
646 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
647 ICMPV6_EXC_HOPLIMIT, 0);
651 ipv6_hdr(skb)->hop_limit--;
653 skb_pull(skb, sizeof(struct ipv6hdr));
662 /********************************
664 ********************************/
666 /* called with rcu_read_lock() */
667 static int ipv6_rthdr_rcv(struct sk_buff *skb)
669 struct inet6_dev *idev = __in6_dev_get(skb->dev);
670 struct inet6_skb_parm *opt = IP6CB(skb);
671 struct in6_addr *addr = NULL;
672 struct in6_addr daddr;
674 struct ipv6_rt_hdr *hdr;
675 struct rt0_hdr *rthdr;
676 struct net *net = dev_net(skb->dev);
677 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
679 idev = __in6_dev_get(skb->dev);
680 if (idev && accept_source_route > idev->cnf.accept_source_route)
681 accept_source_route = idev->cnf.accept_source_route;
683 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
684 !pskb_may_pull(skb, (skb_transport_offset(skb) +
685 ((skb_transport_header(skb)[1] + 1) << 3)))) {
686 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
691 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
693 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
694 skb->pkt_type != PACKET_HOST) {
695 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
701 case IPV6_SRCRT_TYPE_4:
702 /* segment routing */
703 return ipv6_srh_rcv(skb);
704 case IPV6_SRCRT_TYPE_3:
705 /* rpl segment routing */
706 return ipv6_rpl_srh_rcv(skb);
712 if (hdr->segments_left == 0) {
714 #if IS_ENABLED(CONFIG_IPV6_MIP6)
715 case IPV6_SRCRT_TYPE_2:
716 /* Silently discard type 2 header unless it was
720 __IP6_INC_STATS(net, idev,
721 IPSTATS_MIB_INADDRERRORS);
731 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
732 skb->transport_header += (hdr->hdrlen + 1) << 3;
733 opt->dst0 = opt->dst1;
735 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
740 #if IS_ENABLED(CONFIG_IPV6_MIP6)
741 case IPV6_SRCRT_TYPE_2:
742 if (accept_source_route < 0)
744 /* Silently discard invalid RTH type 2 */
745 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
746 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
757 * This is the routing header forwarding algorithm from
761 n = hdr->hdrlen >> 1;
763 if (hdr->segments_left > n) {
764 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
765 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
766 ((&hdr->segments_left) -
767 skb_network_header(skb)));
771 /* We are about to mangle packet header. Be careful!
772 Do not damage packets queued somewhere.
774 if (skb_cloned(skb)) {
775 /* the copy is a forwarded packet */
776 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
777 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
778 IPSTATS_MIB_OUTDISCARDS);
782 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
785 if (skb->ip_summed == CHECKSUM_COMPLETE)
786 skb->ip_summed = CHECKSUM_NONE;
788 i = n - --hdr->segments_left;
790 rthdr = (struct rt0_hdr *) hdr;
795 #if IS_ENABLED(CONFIG_IPV6_MIP6)
796 case IPV6_SRCRT_TYPE_2:
797 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
798 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
799 IPPROTO_ROUTING) < 0) {
800 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
804 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
805 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
815 if (ipv6_addr_is_multicast(addr)) {
816 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
822 *addr = ipv6_hdr(skb)->daddr;
823 ipv6_hdr(skb)->daddr = daddr;
826 ip6_route_input(skb);
827 if (skb_dst(skb)->error) {
828 skb_push(skb, skb->data - skb_network_header(skb));
833 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
834 if (ipv6_hdr(skb)->hop_limit <= 1) {
835 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
836 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
841 ipv6_hdr(skb)->hop_limit--;
845 skb_push(skb, skb->data - skb_network_header(skb));
850 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
851 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
852 (&hdr->type) - skb_network_header(skb));
856 static const struct inet6_protocol rthdr_protocol = {
857 .handler = ipv6_rthdr_rcv,
858 .flags = INET6_PROTO_NOPOLICY,
861 static const struct inet6_protocol destopt_protocol = {
862 .handler = ipv6_destopt_rcv,
863 .flags = INET6_PROTO_NOPOLICY,
866 static const struct inet6_protocol nodata_protocol = {
867 .handler = dst_discard,
868 .flags = INET6_PROTO_NOPOLICY,
871 int __init ipv6_exthdrs_init(void)
875 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
879 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
883 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
890 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
892 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
896 void ipv6_exthdrs_exit(void)
898 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
899 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
900 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
903 /**********************************
905 **********************************/
908 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
910 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
912 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
915 /* Router Alert as of RFC 2711 */
917 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
919 const unsigned char *nh = skb_network_header(skb);
921 if (nh[optoff + 1] == 2) {
922 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
923 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
926 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
934 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
936 const unsigned char *nh = skb_network_header(skb);
937 struct inet6_dev *idev = __in6_dev_get_safely(skb->dev);
938 struct net *net = ipv6_skb_net(skb);
941 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
942 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
944 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
948 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
949 if (pkt_len <= IPV6_MAXPLEN) {
950 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
951 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
954 if (ipv6_hdr(skb)->payload_len) {
955 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
956 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
960 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
961 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INTRUNCATEDPKTS);
965 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
968 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
976 /* CALIPSO RFC 5570 */
978 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
980 const unsigned char *nh = skb_network_header(skb);
982 if (nh[optoff + 1] < 8)
985 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
988 if (!calipso_validate(skb, nh + optoff))
998 static const struct tlvtype_proc tlvprochopopt_lst[] = {
1000 .type = IPV6_TLV_ROUTERALERT,
1001 .func = ipv6_hop_ra,
1004 .type = IPV6_TLV_JUMBO,
1005 .func = ipv6_hop_jumbo,
1008 .type = IPV6_TLV_CALIPSO,
1009 .func = ipv6_hop_calipso,
1014 int ipv6_parse_hopopts(struct sk_buff *skb)
1016 struct inet6_skb_parm *opt = IP6CB(skb);
1017 struct net *net = dev_net(skb->dev);
1021 * skb_network_header(skb) is equal to skb->data, and
1022 * skb_network_header_len(skb) is always equal to
1023 * sizeof(struct ipv6hdr) by definition of
1024 * hop-by-hop options.
1026 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1027 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1028 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1034 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1035 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1038 opt->flags |= IP6SKB_HOPBYHOP;
1039 if (ip6_parse_tlv(tlvprochopopt_lst, skb,
1040 init_net.ipv6.sysctl.max_hbh_opts_cnt)) {
1041 skb->transport_header += extlen;
1043 opt->nhoff = sizeof(struct ipv6hdr);
1050 * Creating outbound headers.
1052 * "build" functions work when skb is filled from head to tail (datagram)
1053 * "push" functions work when headers are added from tail to head (tcp)
1055 * In both cases we assume, that caller reserved enough room
1059 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1060 struct ipv6_rt_hdr *opt,
1061 struct in6_addr **addr_p, struct in6_addr *saddr)
1063 struct rt0_hdr *phdr, *ihdr;
1066 ihdr = (struct rt0_hdr *) opt;
1068 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1069 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1071 hops = ihdr->rt_hdr.hdrlen >> 1;
1074 memcpy(phdr->addr, ihdr->addr + 1,
1075 (hops - 1) * sizeof(struct in6_addr));
1077 phdr->addr[hops - 1] = **addr_p;
1078 *addr_p = ihdr->addr;
1080 phdr->rt_hdr.nexthdr = *proto;
1081 *proto = NEXTHDR_ROUTING;
1084 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1085 struct ipv6_rt_hdr *opt,
1086 struct in6_addr **addr_p, struct in6_addr *saddr)
1088 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1091 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1092 plen = (sr_ihdr->hdrlen + 1) << 3;
1094 sr_phdr = skb_push(skb, plen);
1095 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1097 hops = sr_ihdr->first_segment + 1;
1098 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1099 (hops - 1) * sizeof(struct in6_addr));
1101 sr_phdr->segments[0] = **addr_p;
1102 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1104 if (sr_ihdr->hdrlen > hops * 2) {
1105 int tlvs_offset, tlvs_length;
1107 tlvs_offset = (1 + hops * 2) << 3;
1108 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1109 memcpy((char *)sr_phdr + tlvs_offset,
1110 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1113 #ifdef CONFIG_IPV6_SEG6_HMAC
1114 if (sr_has_hmac(sr_phdr)) {
1115 struct net *net = NULL;
1118 net = dev_net(skb->dev);
1120 net = sock_net(skb->sk);
1125 seg6_push_hmac(net, saddr, sr_phdr);
1129 sr_phdr->nexthdr = *proto;
1130 *proto = NEXTHDR_ROUTING;
1133 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1134 struct ipv6_rt_hdr *opt,
1135 struct in6_addr **addr_p, struct in6_addr *saddr)
1137 switch (opt->type) {
1138 case IPV6_SRCRT_TYPE_0:
1139 case IPV6_SRCRT_STRICT:
1140 case IPV6_SRCRT_TYPE_2:
1141 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1143 case IPV6_SRCRT_TYPE_4:
1144 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1151 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1153 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1155 memcpy(h, opt, ipv6_optlen(opt));
1156 h->nexthdr = *proto;
1160 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1162 struct in6_addr **daddr, struct in6_addr *saddr)
1165 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1167 * IPV6_RTHDRDSTOPTS is ignored
1168 * unless IPV6_RTHDR is set (RFC3542).
1171 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1174 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1177 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1180 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1182 EXPORT_SYMBOL(ipv6_push_frag_opts);
1184 struct ipv6_txoptions *
1185 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1187 struct ipv6_txoptions *opt2;
1189 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1191 long dif = (char *)opt2 - (char *)opt;
1192 memcpy(opt2, opt, opt->tot_len);
1194 *((char **)&opt2->hopopt) += dif;
1196 *((char **)&opt2->dst0opt) += dif;
1198 *((char **)&opt2->dst1opt) += dif;
1200 *((char **)&opt2->srcrt) += dif;
1201 refcount_set(&opt2->refcnt, 1);
1205 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1207 static void ipv6_renew_option(int renewtype,
1208 struct ipv6_opt_hdr **dest,
1209 struct ipv6_opt_hdr *old,
1210 struct ipv6_opt_hdr *new,
1211 int newtype, char **p)
1213 struct ipv6_opt_hdr *src;
1215 src = (renewtype == newtype ? new : old);
1219 memcpy(*p, src, ipv6_optlen(src));
1220 *dest = (struct ipv6_opt_hdr *)*p;
1221 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1225 * ipv6_renew_options - replace a specific ext hdr with a new one.
1227 * @sk: sock from which to allocate memory
1228 * @opt: original options
1229 * @newtype: option type to replace in @opt
1230 * @newopt: new option of type @newtype to replace (user-mem)
1232 * Returns a new set of options which is a copy of @opt with the
1233 * option type @newtype replaced with @newopt.
1235 * @opt may be NULL, in which case a new set of options is returned
1236 * containing just @newopt.
1238 * @newopt may be NULL, in which case the specified option type is
1239 * not copied into the new set of options.
1241 * The new set of options is allocated from the socket option memory
1244 struct ipv6_txoptions *
1245 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1246 int newtype, struct ipv6_opt_hdr *newopt)
1250 struct ipv6_txoptions *opt2;
1253 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1254 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1255 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1256 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1257 if (newtype != IPV6_RTHDR && opt->srcrt)
1258 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1259 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1260 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1264 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1269 tot_len += sizeof(*opt2);
1270 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1272 return ERR_PTR(-ENOBUFS);
1274 memset(opt2, 0, tot_len);
1275 refcount_set(&opt2->refcnt, 1);
1276 opt2->tot_len = tot_len;
1277 p = (char *)(opt2 + 1);
1279 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1280 (opt ? opt->hopopt : NULL),
1281 newopt, newtype, &p);
1282 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1283 (opt ? opt->dst0opt : NULL),
1284 newopt, newtype, &p);
1285 ipv6_renew_option(IPV6_RTHDR,
1286 (struct ipv6_opt_hdr **)&opt2->srcrt,
1287 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1288 newopt, newtype, &p);
1289 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1290 (opt ? opt->dst1opt : NULL),
1291 newopt, newtype, &p);
1293 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1294 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1295 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1296 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1301 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1302 struct ipv6_txoptions *opt)
1305 * ignore the dest before srcrt unless srcrt is being included.
1308 if (opt && opt->dst0opt && !opt->srcrt) {
1309 if (opt_space != opt) {
1310 memcpy(opt_space, opt, sizeof(*opt_space));
1313 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1314 opt->dst0opt = NULL;
1319 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
1322 * fl6_update_dst - update flowi destination address with info given
1323 * by srcrt option, if any.
1325 * @fl6: flowi6 for which daddr is to be updated
1326 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1327 * @orig: copy of original daddr address if modified
1329 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1330 * and initial value of fl6->daddr set in orig
1332 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1333 const struct ipv6_txoptions *opt,
1334 struct in6_addr *orig)
1336 if (!opt || !opt->srcrt)
1341 switch (opt->srcrt->type) {
1342 case IPV6_SRCRT_TYPE_0:
1343 case IPV6_SRCRT_STRICT:
1344 case IPV6_SRCRT_TYPE_2:
1345 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1347 case IPV6_SRCRT_TYPE_4:
1349 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1351 fl6->daddr = srh->segments[srh->segments_left];
1360 EXPORT_SYMBOL_GPL(fl6_update_dst);