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_reason(skb, ICMPV6_UNK_OPTION, optoff,
94 SKB_DROP_REASON_UNHANDLED_PROTO);
99 kfree_skb_reason(skb, SKB_DROP_REASON_UNHANDLED_PROTO);
103 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff);
104 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff);
105 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff);
106 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff);
107 #if IS_ENABLED(CONFIG_IPV6_MIP6)
108 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff);
111 /* Parse tlv encoded option header (hop-by-hop or destination) */
113 static bool ip6_parse_tlv(bool hopbyhop,
117 int len = (skb_transport_header(skb)[1] + 1) << 3;
118 const unsigned char *nh = skb_network_header(skb);
119 int off = skb_network_header_len(skb);
120 bool disallow_unknowns = false;
124 if (unlikely(max_count < 0)) {
125 disallow_unknowns = true;
126 max_count = -max_count;
129 if (skb_transport_offset(skb) + len > skb_headlen(skb))
138 if (nh[off] == IPV6_TLV_PAD1) {
148 optlen = nh[off + 1] + 2;
152 if (nh[off] == IPV6_TLV_PADN) {
153 /* RFC 2460 states that the purpose of PadN is
154 * to align the containing header to multiples
155 * of 8. 7 is therefore the highest valid value.
156 * See also RFC 4942, Section 2.1.9.5.
161 /* RFC 4942 recommends receiving hosts to
162 * actively check PadN payload to contain
165 for (i = 2; i < optlen; i++) {
166 if (nh[off + i] != 0)
171 if (tlv_count > max_count)
176 case IPV6_TLV_ROUTERALERT:
177 if (!ipv6_hop_ra(skb, off))
181 if (!ipv6_hop_ioam(skb, off))
185 if (!ipv6_hop_jumbo(skb, off))
188 case IPV6_TLV_CALIPSO:
189 if (!ipv6_hop_calipso(skb, off))
193 if (!ip6_tlvopt_unknown(skb, off,
200 #if IS_ENABLED(CONFIG_IPV6_MIP6)
202 if (!ipv6_dest_hao(skb, off))
207 if (!ip6_tlvopt_unknown(skb, off,
222 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
226 /*****************************
227 Destination options header.
228 *****************************/
230 #if IS_ENABLED(CONFIG_IPV6_MIP6)
231 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
233 struct ipv6_destopt_hao *hao;
234 struct inet6_skb_parm *opt = IP6CB(skb);
235 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
240 net_dbg_ratelimited("hao duplicated\n");
243 opt->dsthao = opt->dst1;
246 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
248 if (hao->length != 16) {
249 net_dbg_ratelimited("hao invalid option length = %d\n",
251 SKB_DR_SET(reason, IP_INHDR);
255 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
256 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
258 SKB_DR_SET(reason, INVALID_PROTO);
262 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
263 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
264 if (unlikely(ret < 0)) {
265 SKB_DR_SET(reason, XFRM_POLICY);
269 if (skb_cloned(skb)) {
270 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
273 /* update all variable using below by copied skbuff */
274 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
276 ipv6h = ipv6_hdr(skb);
279 if (skb->ip_summed == CHECKSUM_COMPLETE)
280 skb->ip_summed = CHECKSUM_NONE;
282 swap(ipv6h->saddr, hao->addr);
284 if (skb->tstamp == 0)
285 __net_timestamp(skb);
290 kfree_skb_reason(skb, reason);
295 static int ipv6_destopt_rcv(struct sk_buff *skb)
297 struct inet6_dev *idev = __in6_dev_get(skb->dev);
298 struct inet6_skb_parm *opt = IP6CB(skb);
299 #if IS_ENABLED(CONFIG_IPV6_MIP6)
302 struct dst_entry *dst = skb_dst(skb);
303 struct net *net = dev_net(skb->dev);
306 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
307 !pskb_may_pull(skb, (skb_transport_offset(skb) +
308 ((skb_transport_header(skb)[1] + 1) << 3)))) {
309 __IP6_INC_STATS(dev_net(dst->dev), idev,
310 IPSTATS_MIB_INHDRERRORS);
316 extlen = (skb_transport_header(skb)[1] + 1) << 3;
317 if (extlen > net->ipv6.sysctl.max_dst_opts_len)
320 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
321 #if IS_ENABLED(CONFIG_IPV6_MIP6)
325 if (ip6_parse_tlv(false, skb, net->ipv6.sysctl.max_dst_opts_cnt)) {
326 skb->transport_header += extlen;
328 #if IS_ENABLED(CONFIG_IPV6_MIP6)
331 opt->nhoff = opt->dst1;
336 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
340 static void seg6_update_csum(struct sk_buff *skb)
342 struct ipv6_sr_hdr *hdr;
343 struct in6_addr *addr;
346 /* srh is at transport offset and seg_left is already decremented
347 * but daddr is not yet updated with next segment
350 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
351 addr = hdr->segments + hdr->segments_left;
353 hdr->segments_left++;
354 from = *(__be32 *)hdr;
356 hdr->segments_left--;
359 /* update skb csum with diff resulting from seg_left decrement */
361 update_csum_diff4(skb, from, to);
363 /* compute csum diff between current and next segment and update */
365 update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
369 static int ipv6_srh_rcv(struct sk_buff *skb)
371 struct inet6_skb_parm *opt = IP6CB(skb);
372 struct net *net = dev_net(skb->dev);
373 struct ipv6_sr_hdr *hdr;
374 struct inet6_dev *idev;
375 struct in6_addr *addr;
378 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
380 idev = __in6_dev_get(skb->dev);
382 accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
383 if (accept_seg6 > idev->cnf.seg6_enabled)
384 accept_seg6 = idev->cnf.seg6_enabled;
391 #ifdef CONFIG_IPV6_SEG6_HMAC
392 if (!seg6_hmac_validate_skb(skb)) {
399 if (hdr->segments_left == 0) {
400 if (hdr->nexthdr == NEXTHDR_IPV6 || hdr->nexthdr == NEXTHDR_IPV4) {
401 int offset = (hdr->hdrlen + 1) << 3;
403 skb_postpull_rcsum(skb, skb_network_header(skb),
404 skb_network_header_len(skb));
406 if (!pskb_pull(skb, offset)) {
410 skb_postpull_rcsum(skb, skb_transport_header(skb),
413 skb_reset_network_header(skb);
414 skb_reset_transport_header(skb);
415 skb->encapsulation = 0;
416 if (hdr->nexthdr == NEXTHDR_IPV4)
417 skb->protocol = htons(ETH_P_IP);
418 __skb_tunnel_rx(skb, skb->dev, net);
424 opt->srcrt = skb_network_header_len(skb);
425 opt->lastopt = opt->srcrt;
426 skb->transport_header += (hdr->hdrlen + 1) << 3;
427 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
432 if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
433 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
434 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
435 ((&hdr->segments_left) -
436 skb_network_header(skb)));
440 if (skb_cloned(skb)) {
441 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
442 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
443 IPSTATS_MIB_OUTDISCARDS);
449 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
451 hdr->segments_left--;
452 addr = hdr->segments + hdr->segments_left;
454 skb_push(skb, sizeof(struct ipv6hdr));
456 if (skb->ip_summed == CHECKSUM_COMPLETE)
457 seg6_update_csum(skb);
459 ipv6_hdr(skb)->daddr = *addr;
463 ip6_route_input(skb);
465 if (skb_dst(skb)->error) {
470 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
471 if (ipv6_hdr(skb)->hop_limit <= 1) {
472 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
473 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
474 ICMPV6_EXC_HOPLIMIT, 0);
478 ipv6_hdr(skb)->hop_limit--;
480 skb_pull(skb, sizeof(struct ipv6hdr));
489 static int ipv6_rpl_srh_rcv(struct sk_buff *skb)
491 struct ipv6_rpl_sr_hdr *hdr, *ohdr, *chdr;
492 struct inet6_skb_parm *opt = IP6CB(skb);
493 struct net *net = dev_net(skb->dev);
494 struct inet6_dev *idev;
495 struct ipv6hdr *oldhdr;
502 idev = __in6_dev_get(skb->dev);
504 accept_rpl_seg = net->ipv6.devconf_all->rpl_seg_enabled;
505 if (accept_rpl_seg > idev->cnf.rpl_seg_enabled)
506 accept_rpl_seg = idev->cnf.rpl_seg_enabled;
508 if (!accept_rpl_seg) {
514 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
516 if (hdr->segments_left == 0) {
517 if (hdr->nexthdr == NEXTHDR_IPV6) {
518 int offset = (hdr->hdrlen + 1) << 3;
520 skb_postpull_rcsum(skb, skb_network_header(skb),
521 skb_network_header_len(skb));
523 if (!pskb_pull(skb, offset)) {
527 skb_postpull_rcsum(skb, skb_transport_header(skb),
530 skb_reset_network_header(skb);
531 skb_reset_transport_header(skb);
532 skb->encapsulation = 0;
534 __skb_tunnel_rx(skb, skb->dev, net);
540 opt->srcrt = skb_network_header_len(skb);
541 opt->lastopt = opt->srcrt;
542 skb->transport_header += (hdr->hdrlen + 1) << 3;
543 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
548 if (!pskb_may_pull(skb, sizeof(*hdr))) {
553 n = (hdr->hdrlen << 3) - hdr->pad - (16 - hdr->cmpre);
554 r = do_div(n, (16 - hdr->cmpri));
555 /* checks if calculation was without remainder and n fits into
556 * unsigned char which is segments_left field. Should not be
559 if (r || (n + 1) > 255) {
564 if (hdr->segments_left > n + 1) {
565 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
566 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
567 ((&hdr->segments_left) -
568 skb_network_header(skb)));
572 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
578 hdr->segments_left--;
579 i = n - hdr->segments_left;
581 buf = kcalloc(struct_size(hdr, segments.addr, n + 2), 2, GFP_ATOMIC);
582 if (unlikely(!buf)) {
587 ohdr = (struct ipv6_rpl_sr_hdr *)buf;
588 ipv6_rpl_srh_decompress(ohdr, hdr, &ipv6_hdr(skb)->daddr, n);
589 chdr = (struct ipv6_rpl_sr_hdr *)(buf + ((ohdr->hdrlen + 1) << 3));
591 if ((ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST) ||
592 (ipv6_addr_type(&ohdr->rpl_segaddr[i]) & IPV6_ADDR_MULTICAST)) {
598 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
600 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
606 swap(ipv6_hdr(skb)->daddr, ohdr->rpl_segaddr[i]);
608 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
610 oldhdr = ipv6_hdr(skb);
612 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
613 skb_postpull_rcsum(skb, oldhdr,
614 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
615 if (unlikely(!hdr->segments_left)) {
616 if (pskb_expand_head(skb, sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3), 0,
618 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_OUTDISCARDS);
624 oldhdr = ipv6_hdr(skb);
626 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
627 skb_reset_network_header(skb);
628 skb_mac_header_rebuild(skb);
629 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
631 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
632 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
634 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
635 skb_postpush_rcsum(skb, ipv6_hdr(skb),
636 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
642 ip6_route_input(skb);
644 if (skb_dst(skb)->error) {
649 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
650 if (ipv6_hdr(skb)->hop_limit <= 1) {
651 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
652 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
653 ICMPV6_EXC_HOPLIMIT, 0);
657 ipv6_hdr(skb)->hop_limit--;
659 skb_pull(skb, sizeof(struct ipv6hdr));
668 /********************************
670 ********************************/
672 /* called with rcu_read_lock() */
673 static int ipv6_rthdr_rcv(struct sk_buff *skb)
675 struct inet6_dev *idev = __in6_dev_get(skb->dev);
676 struct inet6_skb_parm *opt = IP6CB(skb);
677 struct in6_addr *addr = NULL;
678 struct in6_addr daddr;
680 struct ipv6_rt_hdr *hdr;
681 struct rt0_hdr *rthdr;
682 struct net *net = dev_net(skb->dev);
683 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
685 if (idev && accept_source_route > idev->cnf.accept_source_route)
686 accept_source_route = idev->cnf.accept_source_route;
688 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
689 !pskb_may_pull(skb, (skb_transport_offset(skb) +
690 ((skb_transport_header(skb)[1] + 1) << 3)))) {
691 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
696 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
698 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
699 skb->pkt_type != PACKET_HOST) {
700 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
706 case IPV6_SRCRT_TYPE_4:
707 /* segment routing */
708 return ipv6_srh_rcv(skb);
709 case IPV6_SRCRT_TYPE_3:
710 /* rpl segment routing */
711 return ipv6_rpl_srh_rcv(skb);
717 if (hdr->segments_left == 0) {
719 #if IS_ENABLED(CONFIG_IPV6_MIP6)
720 case IPV6_SRCRT_TYPE_2:
721 /* Silently discard type 2 header unless it was
725 __IP6_INC_STATS(net, idev,
726 IPSTATS_MIB_INADDRERRORS);
736 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
737 skb->transport_header += (hdr->hdrlen + 1) << 3;
738 opt->dst0 = opt->dst1;
740 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
745 #if IS_ENABLED(CONFIG_IPV6_MIP6)
746 case IPV6_SRCRT_TYPE_2:
747 if (accept_source_route < 0)
749 /* Silently discard invalid RTH type 2 */
750 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
751 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
762 * This is the routing header forwarding algorithm from
766 n = hdr->hdrlen >> 1;
768 if (hdr->segments_left > n) {
769 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
770 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
771 ((&hdr->segments_left) -
772 skb_network_header(skb)));
776 /* We are about to mangle packet header. Be careful!
777 Do not damage packets queued somewhere.
779 if (skb_cloned(skb)) {
780 /* the copy is a forwarded packet */
781 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
782 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
783 IPSTATS_MIB_OUTDISCARDS);
787 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
790 if (skb->ip_summed == CHECKSUM_COMPLETE)
791 skb->ip_summed = CHECKSUM_NONE;
793 i = n - --hdr->segments_left;
795 rthdr = (struct rt0_hdr *) hdr;
800 #if IS_ENABLED(CONFIG_IPV6_MIP6)
801 case IPV6_SRCRT_TYPE_2:
802 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
803 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
804 IPPROTO_ROUTING) < 0) {
805 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
809 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
810 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
820 if (ipv6_addr_is_multicast(addr)) {
821 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
827 *addr = ipv6_hdr(skb)->daddr;
828 ipv6_hdr(skb)->daddr = daddr;
831 ip6_route_input(skb);
832 if (skb_dst(skb)->error) {
833 skb_push(skb, skb->data - skb_network_header(skb));
838 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
839 if (ipv6_hdr(skb)->hop_limit <= 1) {
840 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
841 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
846 ipv6_hdr(skb)->hop_limit--;
850 skb_push(skb, skb->data - skb_network_header(skb));
855 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
856 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
857 (&hdr->type) - skb_network_header(skb));
861 static const struct inet6_protocol rthdr_protocol = {
862 .handler = ipv6_rthdr_rcv,
863 .flags = INET6_PROTO_NOPOLICY,
866 static const struct inet6_protocol destopt_protocol = {
867 .handler = ipv6_destopt_rcv,
868 .flags = INET6_PROTO_NOPOLICY,
871 static const struct inet6_protocol nodata_protocol = {
872 .handler = dst_discard,
873 .flags = INET6_PROTO_NOPOLICY,
876 int __init ipv6_exthdrs_init(void)
880 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
884 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
888 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
895 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
897 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
901 void ipv6_exthdrs_exit(void)
903 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
904 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
905 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
908 /**********************************
910 **********************************/
913 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
915 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
917 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
920 /* Router Alert as of RFC 2711 */
922 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
924 const unsigned char *nh = skb_network_header(skb);
926 if (nh[optoff + 1] == 2) {
927 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
928 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
931 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
933 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
939 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff)
941 struct ioam6_trace_hdr *trace;
942 struct ioam6_namespace *ns;
943 struct ioam6_hdr *hdr;
945 /* Bad alignment (must be 4n-aligned) */
949 /* Ignore if IOAM is not enabled on ingress */
950 if (!__in6_dev_get(skb->dev)->cnf.ioam6_enabled)
953 /* Truncated Option header */
954 hdr = (struct ioam6_hdr *)(skb_network_header(skb) + optoff);
955 if (hdr->opt_len < 2)
959 case IOAM6_TYPE_PREALLOC:
960 /* Truncated Pre-allocated Trace header */
961 if (hdr->opt_len < 2 + sizeof(*trace))
964 /* Malformed Pre-allocated Trace header */
965 trace = (struct ioam6_trace_hdr *)((u8 *)hdr + sizeof(*hdr));
966 if (hdr->opt_len < 2 + sizeof(*trace) + trace->remlen * 4)
969 /* Ignore if the IOAM namespace is unknown */
970 ns = ioam6_namespace(ipv6_skb_net(skb), trace->namespace_id);
974 if (!skb_valid_dst(skb))
975 ip6_route_input(skb);
977 ioam6_fill_trace_data(skb, ns, trace, true);
987 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
993 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
995 const unsigned char *nh = skb_network_header(skb);
999 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
1000 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
1002 SKB_DR_SET(reason, IP_INHDR);
1006 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
1007 if (pkt_len <= IPV6_MAXPLEN) {
1008 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff + 2,
1009 SKB_DROP_REASON_IP_INHDR);
1012 if (ipv6_hdr(skb)->payload_len) {
1013 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff,
1014 SKB_DROP_REASON_IP_INHDR);
1018 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
1019 SKB_DR_SET(reason, PKT_TOO_SMALL);
1023 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
1026 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
1030 kfree_skb_reason(skb, reason);
1034 /* CALIPSO RFC 5570 */
1036 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
1038 const unsigned char *nh = skb_network_header(skb);
1040 if (nh[optoff + 1] < 8)
1043 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
1046 if (!calipso_validate(skb, nh + optoff))
1052 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
1056 int ipv6_parse_hopopts(struct sk_buff *skb)
1058 struct inet6_skb_parm *opt = IP6CB(skb);
1059 struct net *net = dev_net(skb->dev);
1063 * skb_network_header(skb) is equal to skb->data, and
1064 * skb_network_header_len(skb) is always equal to
1065 * sizeof(struct ipv6hdr) by definition of
1066 * hop-by-hop options.
1068 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1069 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1070 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1076 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1077 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1080 opt->flags |= IP6SKB_HOPBYHOP;
1081 if (ip6_parse_tlv(true, skb, net->ipv6.sysctl.max_hbh_opts_cnt)) {
1082 skb->transport_header += extlen;
1084 opt->nhoff = sizeof(struct ipv6hdr);
1091 * Creating outbound headers.
1093 * "build" functions work when skb is filled from head to tail (datagram)
1094 * "push" functions work when headers are added from tail to head (tcp)
1096 * In both cases we assume, that caller reserved enough room
1100 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1101 struct ipv6_rt_hdr *opt,
1102 struct in6_addr **addr_p, struct in6_addr *saddr)
1104 struct rt0_hdr *phdr, *ihdr;
1107 ihdr = (struct rt0_hdr *) opt;
1109 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1110 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1112 hops = ihdr->rt_hdr.hdrlen >> 1;
1115 memcpy(phdr->addr, ihdr->addr + 1,
1116 (hops - 1) * sizeof(struct in6_addr));
1118 phdr->addr[hops - 1] = **addr_p;
1119 *addr_p = ihdr->addr;
1121 phdr->rt_hdr.nexthdr = *proto;
1122 *proto = NEXTHDR_ROUTING;
1125 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1126 struct ipv6_rt_hdr *opt,
1127 struct in6_addr **addr_p, struct in6_addr *saddr)
1129 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1132 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1133 plen = (sr_ihdr->hdrlen + 1) << 3;
1135 sr_phdr = skb_push(skb, plen);
1136 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1138 hops = sr_ihdr->first_segment + 1;
1139 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1140 (hops - 1) * sizeof(struct in6_addr));
1142 sr_phdr->segments[0] = **addr_p;
1143 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1145 if (sr_ihdr->hdrlen > hops * 2) {
1146 int tlvs_offset, tlvs_length;
1148 tlvs_offset = (1 + hops * 2) << 3;
1149 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1150 memcpy((char *)sr_phdr + tlvs_offset,
1151 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1154 #ifdef CONFIG_IPV6_SEG6_HMAC
1155 if (sr_has_hmac(sr_phdr)) {
1156 struct net *net = NULL;
1159 net = dev_net(skb->dev);
1161 net = sock_net(skb->sk);
1166 seg6_push_hmac(net, saddr, sr_phdr);
1170 sr_phdr->nexthdr = *proto;
1171 *proto = NEXTHDR_ROUTING;
1174 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1175 struct ipv6_rt_hdr *opt,
1176 struct in6_addr **addr_p, struct in6_addr *saddr)
1178 switch (opt->type) {
1179 case IPV6_SRCRT_TYPE_0:
1180 case IPV6_SRCRT_STRICT:
1181 case IPV6_SRCRT_TYPE_2:
1182 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1184 case IPV6_SRCRT_TYPE_4:
1185 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1192 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1194 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1196 memcpy(h, opt, ipv6_optlen(opt));
1197 h->nexthdr = *proto;
1201 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1203 struct in6_addr **daddr, struct in6_addr *saddr)
1206 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1208 * IPV6_RTHDRDSTOPTS is ignored
1209 * unless IPV6_RTHDR is set (RFC3542).
1212 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1215 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1218 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1221 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1223 EXPORT_SYMBOL(ipv6_push_frag_opts);
1225 struct ipv6_txoptions *
1226 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1228 struct ipv6_txoptions *opt2;
1230 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1232 long dif = (char *)opt2 - (char *)opt;
1233 memcpy(opt2, opt, opt->tot_len);
1235 *((char **)&opt2->hopopt) += dif;
1237 *((char **)&opt2->dst0opt) += dif;
1239 *((char **)&opt2->dst1opt) += dif;
1241 *((char **)&opt2->srcrt) += dif;
1242 refcount_set(&opt2->refcnt, 1);
1246 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1248 static void ipv6_renew_option(int renewtype,
1249 struct ipv6_opt_hdr **dest,
1250 struct ipv6_opt_hdr *old,
1251 struct ipv6_opt_hdr *new,
1252 int newtype, char **p)
1254 struct ipv6_opt_hdr *src;
1256 src = (renewtype == newtype ? new : old);
1260 memcpy(*p, src, ipv6_optlen(src));
1261 *dest = (struct ipv6_opt_hdr *)*p;
1262 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1266 * ipv6_renew_options - replace a specific ext hdr with a new one.
1268 * @sk: sock from which to allocate memory
1269 * @opt: original options
1270 * @newtype: option type to replace in @opt
1271 * @newopt: new option of type @newtype to replace (user-mem)
1273 * Returns a new set of options which is a copy of @opt with the
1274 * option type @newtype replaced with @newopt.
1276 * @opt may be NULL, in which case a new set of options is returned
1277 * containing just @newopt.
1279 * @newopt may be NULL, in which case the specified option type is
1280 * not copied into the new set of options.
1282 * The new set of options is allocated from the socket option memory
1285 struct ipv6_txoptions *
1286 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1287 int newtype, struct ipv6_opt_hdr *newopt)
1291 struct ipv6_txoptions *opt2;
1294 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1295 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1296 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1297 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1298 if (newtype != IPV6_RTHDR && opt->srcrt)
1299 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1300 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1301 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1305 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1310 tot_len += sizeof(*opt2);
1311 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1313 return ERR_PTR(-ENOBUFS);
1315 memset(opt2, 0, tot_len);
1316 refcount_set(&opt2->refcnt, 1);
1317 opt2->tot_len = tot_len;
1318 p = (char *)(opt2 + 1);
1320 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1321 (opt ? opt->hopopt : NULL),
1322 newopt, newtype, &p);
1323 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1324 (opt ? opt->dst0opt : NULL),
1325 newopt, newtype, &p);
1326 ipv6_renew_option(IPV6_RTHDR,
1327 (struct ipv6_opt_hdr **)&opt2->srcrt,
1328 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1329 newopt, newtype, &p);
1330 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1331 (opt ? opt->dst1opt : NULL),
1332 newopt, newtype, &p);
1334 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1335 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1336 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1337 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1342 struct ipv6_txoptions *__ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1343 struct ipv6_txoptions *opt)
1346 * ignore the dest before srcrt unless srcrt is being included.
1349 if (opt->dst0opt && !opt->srcrt) {
1350 if (opt_space != opt) {
1351 memcpy(opt_space, opt, sizeof(*opt_space));
1354 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1355 opt->dst0opt = NULL;
1360 EXPORT_SYMBOL_GPL(__ipv6_fixup_options);
1363 * fl6_update_dst - update flowi destination address with info given
1364 * by srcrt option, if any.
1366 * @fl6: flowi6 for which daddr is to be updated
1367 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1368 * @orig: copy of original daddr address if modified
1370 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1371 * and initial value of fl6->daddr set in orig
1373 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1374 const struct ipv6_txoptions *opt,
1375 struct in6_addr *orig)
1377 if (!opt || !opt->srcrt)
1382 switch (opt->srcrt->type) {
1383 case IPV6_SRCRT_TYPE_0:
1384 case IPV6_SRCRT_STRICT:
1385 case IPV6_SRCRT_TYPE_2:
1386 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1388 case IPV6_SRCRT_TYPE_4:
1390 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1392 fl6->daddr = srh->segments[srh->segments_left];
1401 EXPORT_SYMBOL_GPL(fl6_update_dst);