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>
52 #include <linux/ioam6.h>
53 #include <net/ioam6.h>
54 #include <net/dst_metadata.h>
56 #include <linux/uaccess.h>
58 /*********************
60 *********************/
62 /* An unknown option is detected, decide what to do */
64 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff,
65 bool disallow_unknowns)
67 if (disallow_unknowns) {
68 /* If unknown TLVs are disallowed by configuration
69 * then always silently drop packet. Note this also
70 * means no ICMP parameter problem is sent which
71 * could be a good property to mitigate a reflection DOS
78 switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
82 case 1: /* drop packet */
85 case 3: /* Send ICMP if not a multicast address and drop packet */
86 /* Actually, it is redundant check. icmp_send
87 will recheck in any case.
89 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
92 case 2: /* send ICMP PARM PROB regardless and drop packet */
93 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
102 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff);
103 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff);
104 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff);
105 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff);
106 #if IS_ENABLED(CONFIG_IPV6_MIP6)
107 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff);
110 /* Parse tlv encoded option header (hop-by-hop or destination) */
112 static bool ip6_parse_tlv(bool hopbyhop,
116 int len = (skb_transport_header(skb)[1] + 1) << 3;
117 const unsigned char *nh = skb_network_header(skb);
118 int off = skb_network_header_len(skb);
119 bool disallow_unknowns = false;
123 if (unlikely(max_count < 0)) {
124 disallow_unknowns = true;
125 max_count = -max_count;
128 if (skb_transport_offset(skb) + len > skb_headlen(skb))
137 if (nh[off] == IPV6_TLV_PAD1) {
147 optlen = nh[off + 1] + 2;
151 if (nh[off] == IPV6_TLV_PADN) {
152 /* RFC 2460 states that the purpose of PadN is
153 * to align the containing header to multiples
154 * of 8. 7 is therefore the highest valid value.
155 * See also RFC 4942, Section 2.1.9.5.
160 /* RFC 4942 recommends receiving hosts to
161 * actively check PadN payload to contain
164 for (i = 2; i < optlen; i++) {
165 if (nh[off + i] != 0)
170 if (tlv_count > max_count)
175 case IPV6_TLV_ROUTERALERT:
176 if (!ipv6_hop_ra(skb, off))
180 if (!ipv6_hop_ioam(skb, off))
184 if (!ipv6_hop_jumbo(skb, off))
187 case IPV6_TLV_CALIPSO:
188 if (!ipv6_hop_calipso(skb, off))
192 if (!ip6_tlvopt_unknown(skb, off,
199 #if IS_ENABLED(CONFIG_IPV6_MIP6)
201 if (!ipv6_dest_hao(skb, off))
206 if (!ip6_tlvopt_unknown(skb, off,
225 /*****************************
226 Destination options header.
227 *****************************/
229 #if IS_ENABLED(CONFIG_IPV6_MIP6)
230 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
232 struct ipv6_destopt_hao *hao;
233 struct inet6_skb_parm *opt = IP6CB(skb);
234 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
238 net_dbg_ratelimited("hao duplicated\n");
241 opt->dsthao = opt->dst1;
244 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
246 if (hao->length != 16) {
247 net_dbg_ratelimited("hao invalid option length = %d\n",
252 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
253 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
258 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
259 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
260 if (unlikely(ret < 0))
263 if (skb_cloned(skb)) {
264 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
267 /* update all variable using below by copied skbuff */
268 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
270 ipv6h = ipv6_hdr(skb);
273 if (skb->ip_summed == CHECKSUM_COMPLETE)
274 skb->ip_summed = CHECKSUM_NONE;
276 swap(ipv6h->saddr, hao->addr);
278 if (skb->tstamp == 0)
279 __net_timestamp(skb);
289 static int ipv6_destopt_rcv(struct sk_buff *skb)
291 struct inet6_dev *idev = __in6_dev_get(skb->dev);
292 struct inet6_skb_parm *opt = IP6CB(skb);
293 #if IS_ENABLED(CONFIG_IPV6_MIP6)
296 struct dst_entry *dst = skb_dst(skb);
297 struct net *net = dev_net(skb->dev);
300 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
301 !pskb_may_pull(skb, (skb_transport_offset(skb) +
302 ((skb_transport_header(skb)[1] + 1) << 3)))) {
303 __IP6_INC_STATS(dev_net(dst->dev), idev,
304 IPSTATS_MIB_INHDRERRORS);
310 extlen = (skb_transport_header(skb)[1] + 1) << 3;
311 if (extlen > net->ipv6.sysctl.max_dst_opts_len)
314 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
315 #if IS_ENABLED(CONFIG_IPV6_MIP6)
319 if (ip6_parse_tlv(false, skb, net->ipv6.sysctl.max_dst_opts_cnt)) {
320 skb->transport_header += extlen;
322 #if IS_ENABLED(CONFIG_IPV6_MIP6)
325 opt->nhoff = opt->dst1;
330 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
334 static void seg6_update_csum(struct sk_buff *skb)
336 struct ipv6_sr_hdr *hdr;
337 struct in6_addr *addr;
340 /* srh is at transport offset and seg_left is already decremented
341 * but daddr is not yet updated with next segment
344 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
345 addr = hdr->segments + hdr->segments_left;
347 hdr->segments_left++;
348 from = *(__be32 *)hdr;
350 hdr->segments_left--;
353 /* update skb csum with diff resulting from seg_left decrement */
355 update_csum_diff4(skb, from, to);
357 /* compute csum diff between current and next segment and update */
359 update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
363 static int ipv6_srh_rcv(struct sk_buff *skb)
365 struct inet6_skb_parm *opt = IP6CB(skb);
366 struct net *net = dev_net(skb->dev);
367 struct ipv6_sr_hdr *hdr;
368 struct inet6_dev *idev;
369 struct in6_addr *addr;
372 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
374 idev = __in6_dev_get(skb->dev);
376 accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
377 if (accept_seg6 > idev->cnf.seg6_enabled)
378 accept_seg6 = idev->cnf.seg6_enabled;
385 #ifdef CONFIG_IPV6_SEG6_HMAC
386 if (!seg6_hmac_validate_skb(skb)) {
393 if (hdr->segments_left == 0) {
394 if (hdr->nexthdr == NEXTHDR_IPV6 || hdr->nexthdr == NEXTHDR_IPV4) {
395 int offset = (hdr->hdrlen + 1) << 3;
397 skb_postpull_rcsum(skb, skb_network_header(skb),
398 skb_network_header_len(skb));
400 if (!pskb_pull(skb, offset)) {
404 skb_postpull_rcsum(skb, skb_transport_header(skb),
407 skb_reset_network_header(skb);
408 skb_reset_transport_header(skb);
409 skb->encapsulation = 0;
410 if (hdr->nexthdr == NEXTHDR_IPV4)
411 skb->protocol = htons(ETH_P_IP);
412 __skb_tunnel_rx(skb, skb->dev, net);
418 opt->srcrt = skb_network_header_len(skb);
419 opt->lastopt = opt->srcrt;
420 skb->transport_header += (hdr->hdrlen + 1) << 3;
421 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
426 if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
427 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
428 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
429 ((&hdr->segments_left) -
430 skb_network_header(skb)));
434 if (skb_cloned(skb)) {
435 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
436 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
437 IPSTATS_MIB_OUTDISCARDS);
443 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
445 hdr->segments_left--;
446 addr = hdr->segments + hdr->segments_left;
448 skb_push(skb, sizeof(struct ipv6hdr));
450 if (skb->ip_summed == CHECKSUM_COMPLETE)
451 seg6_update_csum(skb);
453 ipv6_hdr(skb)->daddr = *addr;
457 ip6_route_input(skb);
459 if (skb_dst(skb)->error) {
464 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
465 if (ipv6_hdr(skb)->hop_limit <= 1) {
466 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
467 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
468 ICMPV6_EXC_HOPLIMIT, 0);
472 ipv6_hdr(skb)->hop_limit--;
474 skb_pull(skb, sizeof(struct ipv6hdr));
483 static int ipv6_rpl_srh_rcv(struct sk_buff *skb)
485 struct ipv6_rpl_sr_hdr *hdr, *ohdr, *chdr;
486 struct inet6_skb_parm *opt = IP6CB(skb);
487 struct net *net = dev_net(skb->dev);
488 struct inet6_dev *idev;
489 struct ipv6hdr *oldhdr;
490 struct in6_addr addr;
497 idev = __in6_dev_get(skb->dev);
499 accept_rpl_seg = net->ipv6.devconf_all->rpl_seg_enabled;
500 if (accept_rpl_seg > idev->cnf.rpl_seg_enabled)
501 accept_rpl_seg = idev->cnf.rpl_seg_enabled;
503 if (!accept_rpl_seg) {
509 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
511 if (hdr->segments_left == 0) {
512 if (hdr->nexthdr == NEXTHDR_IPV6) {
513 int offset = (hdr->hdrlen + 1) << 3;
515 skb_postpull_rcsum(skb, skb_network_header(skb),
516 skb_network_header_len(skb));
518 if (!pskb_pull(skb, offset)) {
522 skb_postpull_rcsum(skb, skb_transport_header(skb),
525 skb_reset_network_header(skb);
526 skb_reset_transport_header(skb);
527 skb->encapsulation = 0;
529 __skb_tunnel_rx(skb, skb->dev, net);
535 opt->srcrt = skb_network_header_len(skb);
536 opt->lastopt = opt->srcrt;
537 skb->transport_header += (hdr->hdrlen + 1) << 3;
538 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
543 if (!pskb_may_pull(skb, sizeof(*hdr))) {
548 n = (hdr->hdrlen << 3) - hdr->pad - (16 - hdr->cmpre);
549 r = do_div(n, (16 - hdr->cmpri));
550 /* checks if calculation was without remainder and n fits into
551 * unsigned char which is segments_left field. Should not be
554 if (r || (n + 1) > 255) {
559 if (hdr->segments_left > n + 1) {
560 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
561 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
562 ((&hdr->segments_left) -
563 skb_network_header(skb)));
567 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
573 hdr->segments_left--;
574 i = n - hdr->segments_left;
576 buf = kcalloc(struct_size(hdr, segments.addr, n + 2), 2, GFP_ATOMIC);
577 if (unlikely(!buf)) {
582 ohdr = (struct ipv6_rpl_sr_hdr *)buf;
583 ipv6_rpl_srh_decompress(ohdr, hdr, &ipv6_hdr(skb)->daddr, n);
584 chdr = (struct ipv6_rpl_sr_hdr *)(buf + ((ohdr->hdrlen + 1) << 3));
586 if ((ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST) ||
587 (ipv6_addr_type(&ohdr->rpl_segaddr[i]) & IPV6_ADDR_MULTICAST)) {
593 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
595 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
601 addr = ipv6_hdr(skb)->daddr;
602 ipv6_hdr(skb)->daddr = ohdr->rpl_segaddr[i];
603 ohdr->rpl_segaddr[i] = addr;
605 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
607 oldhdr = ipv6_hdr(skb);
609 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
610 skb_postpull_rcsum(skb, oldhdr,
611 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
612 if (unlikely(!hdr->segments_left)) {
613 if (pskb_expand_head(skb, sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3), 0,
615 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_OUTDISCARDS);
621 oldhdr = ipv6_hdr(skb);
623 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
624 skb_reset_network_header(skb);
625 skb_mac_header_rebuild(skb);
626 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
628 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
629 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
631 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
632 skb_postpush_rcsum(skb, ipv6_hdr(skb),
633 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
639 ip6_route_input(skb);
641 if (skb_dst(skb)->error) {
646 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
647 if (ipv6_hdr(skb)->hop_limit <= 1) {
648 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
649 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
650 ICMPV6_EXC_HOPLIMIT, 0);
654 ipv6_hdr(skb)->hop_limit--;
656 skb_pull(skb, sizeof(struct ipv6hdr));
665 /********************************
667 ********************************/
669 /* called with rcu_read_lock() */
670 static int ipv6_rthdr_rcv(struct sk_buff *skb)
672 struct inet6_dev *idev = __in6_dev_get(skb->dev);
673 struct inet6_skb_parm *opt = IP6CB(skb);
674 struct in6_addr *addr = NULL;
675 struct in6_addr daddr;
677 struct ipv6_rt_hdr *hdr;
678 struct rt0_hdr *rthdr;
679 struct net *net = dev_net(skb->dev);
680 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
682 idev = __in6_dev_get(skb->dev);
683 if (idev && accept_source_route > idev->cnf.accept_source_route)
684 accept_source_route = idev->cnf.accept_source_route;
686 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
687 !pskb_may_pull(skb, (skb_transport_offset(skb) +
688 ((skb_transport_header(skb)[1] + 1) << 3)))) {
689 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
694 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
696 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
697 skb->pkt_type != PACKET_HOST) {
698 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
704 case IPV6_SRCRT_TYPE_4:
705 /* segment routing */
706 return ipv6_srh_rcv(skb);
707 case IPV6_SRCRT_TYPE_3:
708 /* rpl segment routing */
709 return ipv6_rpl_srh_rcv(skb);
715 if (hdr->segments_left == 0) {
717 #if IS_ENABLED(CONFIG_IPV6_MIP6)
718 case IPV6_SRCRT_TYPE_2:
719 /* Silently discard type 2 header unless it was
723 __IP6_INC_STATS(net, idev,
724 IPSTATS_MIB_INADDRERRORS);
734 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
735 skb->transport_header += (hdr->hdrlen + 1) << 3;
736 opt->dst0 = opt->dst1;
738 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
743 #if IS_ENABLED(CONFIG_IPV6_MIP6)
744 case IPV6_SRCRT_TYPE_2:
745 if (accept_source_route < 0)
747 /* Silently discard invalid RTH type 2 */
748 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
749 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
760 * This is the routing header forwarding algorithm from
764 n = hdr->hdrlen >> 1;
766 if (hdr->segments_left > n) {
767 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
768 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
769 ((&hdr->segments_left) -
770 skb_network_header(skb)));
774 /* We are about to mangle packet header. Be careful!
775 Do not damage packets queued somewhere.
777 if (skb_cloned(skb)) {
778 /* the copy is a forwarded packet */
779 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
780 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
781 IPSTATS_MIB_OUTDISCARDS);
785 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
788 if (skb->ip_summed == CHECKSUM_COMPLETE)
789 skb->ip_summed = CHECKSUM_NONE;
791 i = n - --hdr->segments_left;
793 rthdr = (struct rt0_hdr *) hdr;
798 #if IS_ENABLED(CONFIG_IPV6_MIP6)
799 case IPV6_SRCRT_TYPE_2:
800 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
801 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
802 IPPROTO_ROUTING) < 0) {
803 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
807 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
808 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
818 if (ipv6_addr_is_multicast(addr)) {
819 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
825 *addr = ipv6_hdr(skb)->daddr;
826 ipv6_hdr(skb)->daddr = daddr;
829 ip6_route_input(skb);
830 if (skb_dst(skb)->error) {
831 skb_push(skb, skb->data - skb_network_header(skb));
836 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
837 if (ipv6_hdr(skb)->hop_limit <= 1) {
838 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
839 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
844 ipv6_hdr(skb)->hop_limit--;
848 skb_push(skb, skb->data - skb_network_header(skb));
853 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
854 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
855 (&hdr->type) - skb_network_header(skb));
859 static const struct inet6_protocol rthdr_protocol = {
860 .handler = ipv6_rthdr_rcv,
861 .flags = INET6_PROTO_NOPOLICY,
864 static const struct inet6_protocol destopt_protocol = {
865 .handler = ipv6_destopt_rcv,
866 .flags = INET6_PROTO_NOPOLICY,
869 static const struct inet6_protocol nodata_protocol = {
870 .handler = dst_discard,
871 .flags = INET6_PROTO_NOPOLICY,
874 int __init ipv6_exthdrs_init(void)
878 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
882 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
886 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
893 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
895 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
899 void ipv6_exthdrs_exit(void)
901 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
902 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
903 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
906 /**********************************
908 **********************************/
911 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
913 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
915 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
918 /* Router Alert as of RFC 2711 */
920 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
922 const unsigned char *nh = skb_network_header(skb);
924 if (nh[optoff + 1] == 2) {
925 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
926 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
929 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
937 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff)
939 struct ioam6_trace_hdr *trace;
940 struct ioam6_namespace *ns;
941 struct ioam6_hdr *hdr;
943 /* Bad alignment (must be 4n-aligned) */
947 /* Ignore if IOAM is not enabled on ingress */
948 if (!__in6_dev_get(skb->dev)->cnf.ioam6_enabled)
951 /* Truncated Option header */
952 hdr = (struct ioam6_hdr *)(skb_network_header(skb) + optoff);
953 if (hdr->opt_len < 2)
957 case IOAM6_TYPE_PREALLOC:
958 /* Truncated Pre-allocated Trace header */
959 if (hdr->opt_len < 2 + sizeof(*trace))
962 /* Malformed Pre-allocated Trace header */
963 trace = (struct ioam6_trace_hdr *)((u8 *)hdr + sizeof(*hdr));
964 if (hdr->opt_len < 2 + sizeof(*trace) + trace->remlen * 4)
967 /* Ignore if the IOAM namespace is unknown */
968 ns = ioam6_namespace(ipv6_skb_net(skb), trace->namespace_id);
972 if (!skb_valid_dst(skb))
973 ip6_route_input(skb);
975 ioam6_fill_trace_data(skb, ns, trace);
991 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
993 const unsigned char *nh = skb_network_header(skb);
994 struct inet6_dev *idev = __in6_dev_get_safely(skb->dev);
995 struct net *net = ipv6_skb_net(skb);
998 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
999 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
1001 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
1005 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
1006 if (pkt_len <= IPV6_MAXPLEN) {
1007 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
1008 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
1011 if (ipv6_hdr(skb)->payload_len) {
1012 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
1013 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
1017 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
1018 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INTRUNCATEDPKTS);
1022 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
1025 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
1033 /* CALIPSO RFC 5570 */
1035 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
1037 const unsigned char *nh = skb_network_header(skb);
1039 if (nh[optoff + 1] < 8)
1042 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
1045 if (!calipso_validate(skb, nh + optoff))
1055 int ipv6_parse_hopopts(struct sk_buff *skb)
1057 struct inet6_skb_parm *opt = IP6CB(skb);
1058 struct net *net = dev_net(skb->dev);
1062 * skb_network_header(skb) is equal to skb->data, and
1063 * skb_network_header_len(skb) is always equal to
1064 * sizeof(struct ipv6hdr) by definition of
1065 * hop-by-hop options.
1067 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1068 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1069 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1075 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1076 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1079 opt->flags |= IP6SKB_HOPBYHOP;
1080 if (ip6_parse_tlv(true, skb, net->ipv6.sysctl.max_hbh_opts_cnt)) {
1081 skb->transport_header += extlen;
1083 opt->nhoff = sizeof(struct ipv6hdr);
1090 * Creating outbound headers.
1092 * "build" functions work when skb is filled from head to tail (datagram)
1093 * "push" functions work when headers are added from tail to head (tcp)
1095 * In both cases we assume, that caller reserved enough room
1099 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1100 struct ipv6_rt_hdr *opt,
1101 struct in6_addr **addr_p, struct in6_addr *saddr)
1103 struct rt0_hdr *phdr, *ihdr;
1106 ihdr = (struct rt0_hdr *) opt;
1108 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1109 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1111 hops = ihdr->rt_hdr.hdrlen >> 1;
1114 memcpy(phdr->addr, ihdr->addr + 1,
1115 (hops - 1) * sizeof(struct in6_addr));
1117 phdr->addr[hops - 1] = **addr_p;
1118 *addr_p = ihdr->addr;
1120 phdr->rt_hdr.nexthdr = *proto;
1121 *proto = NEXTHDR_ROUTING;
1124 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1125 struct ipv6_rt_hdr *opt,
1126 struct in6_addr **addr_p, struct in6_addr *saddr)
1128 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1131 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1132 plen = (sr_ihdr->hdrlen + 1) << 3;
1134 sr_phdr = skb_push(skb, plen);
1135 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1137 hops = sr_ihdr->first_segment + 1;
1138 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1139 (hops - 1) * sizeof(struct in6_addr));
1141 sr_phdr->segments[0] = **addr_p;
1142 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1144 if (sr_ihdr->hdrlen > hops * 2) {
1145 int tlvs_offset, tlvs_length;
1147 tlvs_offset = (1 + hops * 2) << 3;
1148 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1149 memcpy((char *)sr_phdr + tlvs_offset,
1150 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1153 #ifdef CONFIG_IPV6_SEG6_HMAC
1154 if (sr_has_hmac(sr_phdr)) {
1155 struct net *net = NULL;
1158 net = dev_net(skb->dev);
1160 net = sock_net(skb->sk);
1165 seg6_push_hmac(net, saddr, sr_phdr);
1169 sr_phdr->nexthdr = *proto;
1170 *proto = NEXTHDR_ROUTING;
1173 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1174 struct ipv6_rt_hdr *opt,
1175 struct in6_addr **addr_p, struct in6_addr *saddr)
1177 switch (opt->type) {
1178 case IPV6_SRCRT_TYPE_0:
1179 case IPV6_SRCRT_STRICT:
1180 case IPV6_SRCRT_TYPE_2:
1181 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1183 case IPV6_SRCRT_TYPE_4:
1184 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1191 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1193 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1195 memcpy(h, opt, ipv6_optlen(opt));
1196 h->nexthdr = *proto;
1200 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1202 struct in6_addr **daddr, struct in6_addr *saddr)
1205 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1207 * IPV6_RTHDRDSTOPTS is ignored
1208 * unless IPV6_RTHDR is set (RFC3542).
1211 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1214 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1217 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1220 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1222 EXPORT_SYMBOL(ipv6_push_frag_opts);
1224 struct ipv6_txoptions *
1225 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1227 struct ipv6_txoptions *opt2;
1229 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1231 long dif = (char *)opt2 - (char *)opt;
1232 memcpy(opt2, opt, opt->tot_len);
1234 *((char **)&opt2->hopopt) += dif;
1236 *((char **)&opt2->dst0opt) += dif;
1238 *((char **)&opt2->dst1opt) += dif;
1240 *((char **)&opt2->srcrt) += dif;
1241 refcount_set(&opt2->refcnt, 1);
1245 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1247 static void ipv6_renew_option(int renewtype,
1248 struct ipv6_opt_hdr **dest,
1249 struct ipv6_opt_hdr *old,
1250 struct ipv6_opt_hdr *new,
1251 int newtype, char **p)
1253 struct ipv6_opt_hdr *src;
1255 src = (renewtype == newtype ? new : old);
1259 memcpy(*p, src, ipv6_optlen(src));
1260 *dest = (struct ipv6_opt_hdr *)*p;
1261 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1265 * ipv6_renew_options - replace a specific ext hdr with a new one.
1267 * @sk: sock from which to allocate memory
1268 * @opt: original options
1269 * @newtype: option type to replace in @opt
1270 * @newopt: new option of type @newtype to replace (user-mem)
1272 * Returns a new set of options which is a copy of @opt with the
1273 * option type @newtype replaced with @newopt.
1275 * @opt may be NULL, in which case a new set of options is returned
1276 * containing just @newopt.
1278 * @newopt may be NULL, in which case the specified option type is
1279 * not copied into the new set of options.
1281 * The new set of options is allocated from the socket option memory
1284 struct ipv6_txoptions *
1285 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1286 int newtype, struct ipv6_opt_hdr *newopt)
1290 struct ipv6_txoptions *opt2;
1293 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1294 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1295 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1296 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1297 if (newtype != IPV6_RTHDR && opt->srcrt)
1298 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1299 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1300 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1304 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1309 tot_len += sizeof(*opt2);
1310 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1312 return ERR_PTR(-ENOBUFS);
1314 memset(opt2, 0, tot_len);
1315 refcount_set(&opt2->refcnt, 1);
1316 opt2->tot_len = tot_len;
1317 p = (char *)(opt2 + 1);
1319 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1320 (opt ? opt->hopopt : NULL),
1321 newopt, newtype, &p);
1322 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1323 (opt ? opt->dst0opt : NULL),
1324 newopt, newtype, &p);
1325 ipv6_renew_option(IPV6_RTHDR,
1326 (struct ipv6_opt_hdr **)&opt2->srcrt,
1327 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1328 newopt, newtype, &p);
1329 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1330 (opt ? opt->dst1opt : NULL),
1331 newopt, newtype, &p);
1333 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1334 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1335 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1336 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1341 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1342 struct ipv6_txoptions *opt)
1345 * ignore the dest before srcrt unless srcrt is being included.
1348 if (opt && opt->dst0opt && !opt->srcrt) {
1349 if (opt_space != opt) {
1350 memcpy(opt_space, opt, sizeof(*opt_space));
1353 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1354 opt->dst0opt = NULL;
1359 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
1362 * fl6_update_dst - update flowi destination address with info given
1363 * by srcrt option, if any.
1365 * @fl6: flowi6 for which daddr is to be updated
1366 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1367 * @orig: copy of original daddr address if modified
1369 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1370 * and initial value of fl6->daddr set in orig
1372 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1373 const struct ipv6_txoptions *opt,
1374 struct in6_addr *orig)
1376 if (!opt || !opt->srcrt)
1381 switch (opt->srcrt->type) {
1382 case IPV6_SRCRT_TYPE_0:
1383 case IPV6_SRCRT_STRICT:
1384 case IPV6_SRCRT_TYPE_2:
1385 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1387 case IPV6_SRCRT_TYPE_4:
1389 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1391 fl6->daddr = srh->segments[srh->segments_left];
1400 EXPORT_SYMBOL_GPL(fl6_update_dst);