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 (skb_cloned(skb)) {
573 if (pskb_expand_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE, 0,
575 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
576 IPSTATS_MIB_OUTDISCARDS);
581 err = skb_cow_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE);
588 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
590 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
596 hdr->segments_left--;
597 i = n - hdr->segments_left;
599 buf = kcalloc(struct_size(hdr, segments.addr, n + 2), 2, GFP_ATOMIC);
600 if (unlikely(!buf)) {
605 ohdr = (struct ipv6_rpl_sr_hdr *)buf;
606 ipv6_rpl_srh_decompress(ohdr, hdr, &ipv6_hdr(skb)->daddr, n);
607 chdr = (struct ipv6_rpl_sr_hdr *)(buf + ((ohdr->hdrlen + 1) << 3));
609 if ((ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST) ||
610 (ipv6_addr_type(&ohdr->rpl_segaddr[i]) & IPV6_ADDR_MULTICAST)) {
616 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
618 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
624 swap(ipv6_hdr(skb)->daddr, ohdr->rpl_segaddr[i]);
626 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
628 oldhdr = ipv6_hdr(skb);
630 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
631 skb_postpull_rcsum(skb, oldhdr,
632 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
633 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
634 skb_reset_network_header(skb);
635 skb_mac_header_rebuild(skb);
636 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
638 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
639 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
641 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
642 skb_postpush_rcsum(skb, ipv6_hdr(skb),
643 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
649 ip6_route_input(skb);
651 if (skb_dst(skb)->error) {
656 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
657 if (ipv6_hdr(skb)->hop_limit <= 1) {
658 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
659 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
660 ICMPV6_EXC_HOPLIMIT, 0);
664 ipv6_hdr(skb)->hop_limit--;
666 skb_pull(skb, sizeof(struct ipv6hdr));
675 /********************************
677 ********************************/
679 /* called with rcu_read_lock() */
680 static int ipv6_rthdr_rcv(struct sk_buff *skb)
682 struct inet6_dev *idev = __in6_dev_get(skb->dev);
683 struct inet6_skb_parm *opt = IP6CB(skb);
684 struct in6_addr *addr = NULL;
685 struct in6_addr daddr;
687 struct ipv6_rt_hdr *hdr;
688 struct rt0_hdr *rthdr;
689 struct net *net = dev_net(skb->dev);
690 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
692 if (idev && accept_source_route > idev->cnf.accept_source_route)
693 accept_source_route = idev->cnf.accept_source_route;
695 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
696 !pskb_may_pull(skb, (skb_transport_offset(skb) +
697 ((skb_transport_header(skb)[1] + 1) << 3)))) {
698 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
703 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
705 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
706 skb->pkt_type != PACKET_HOST) {
707 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
713 case IPV6_SRCRT_TYPE_4:
714 /* segment routing */
715 return ipv6_srh_rcv(skb);
716 case IPV6_SRCRT_TYPE_3:
717 /* rpl segment routing */
718 return ipv6_rpl_srh_rcv(skb);
724 if (hdr->segments_left == 0) {
726 #if IS_ENABLED(CONFIG_IPV6_MIP6)
727 case IPV6_SRCRT_TYPE_2:
728 /* Silently discard type 2 header unless it was
732 __IP6_INC_STATS(net, idev,
733 IPSTATS_MIB_INADDRERRORS);
743 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
744 skb->transport_header += (hdr->hdrlen + 1) << 3;
745 opt->dst0 = opt->dst1;
747 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
752 #if IS_ENABLED(CONFIG_IPV6_MIP6)
753 case IPV6_SRCRT_TYPE_2:
754 if (accept_source_route < 0)
756 /* Silently discard invalid RTH type 2 */
757 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
758 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
769 * This is the routing header forwarding algorithm from
773 n = hdr->hdrlen >> 1;
775 if (hdr->segments_left > n) {
776 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
777 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
778 ((&hdr->segments_left) -
779 skb_network_header(skb)));
783 /* We are about to mangle packet header. Be careful!
784 Do not damage packets queued somewhere.
786 if (skb_cloned(skb)) {
787 /* the copy is a forwarded packet */
788 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
789 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
790 IPSTATS_MIB_OUTDISCARDS);
794 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
797 if (skb->ip_summed == CHECKSUM_COMPLETE)
798 skb->ip_summed = CHECKSUM_NONE;
800 i = n - --hdr->segments_left;
802 rthdr = (struct rt0_hdr *) hdr;
807 #if IS_ENABLED(CONFIG_IPV6_MIP6)
808 case IPV6_SRCRT_TYPE_2:
809 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
810 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
811 IPPROTO_ROUTING) < 0) {
812 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
816 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
817 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
827 if (ipv6_addr_is_multicast(addr)) {
828 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
834 *addr = ipv6_hdr(skb)->daddr;
835 ipv6_hdr(skb)->daddr = daddr;
838 ip6_route_input(skb);
839 if (skb_dst(skb)->error) {
840 skb_push(skb, skb->data - skb_network_header(skb));
845 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
846 if (ipv6_hdr(skb)->hop_limit <= 1) {
847 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
848 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
853 ipv6_hdr(skb)->hop_limit--;
857 skb_push(skb, skb->data - skb_network_header(skb));
862 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
863 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
864 (&hdr->type) - skb_network_header(skb));
868 static const struct inet6_protocol rthdr_protocol = {
869 .handler = ipv6_rthdr_rcv,
870 .flags = INET6_PROTO_NOPOLICY,
873 static const struct inet6_protocol destopt_protocol = {
874 .handler = ipv6_destopt_rcv,
875 .flags = INET6_PROTO_NOPOLICY,
878 static const struct inet6_protocol nodata_protocol = {
879 .handler = dst_discard,
880 .flags = INET6_PROTO_NOPOLICY,
883 int __init ipv6_exthdrs_init(void)
887 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
891 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
895 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
902 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
904 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
908 void ipv6_exthdrs_exit(void)
910 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
911 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
912 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
915 /**********************************
917 **********************************/
920 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
922 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
924 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
927 /* Router Alert as of RFC 2711 */
929 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
931 const unsigned char *nh = skb_network_header(skb);
933 if (nh[optoff + 1] == 2) {
934 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
935 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
938 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
940 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
946 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff)
948 struct ioam6_trace_hdr *trace;
949 struct ioam6_namespace *ns;
950 struct ioam6_hdr *hdr;
952 /* Bad alignment (must be 4n-aligned) */
956 /* Ignore if IOAM is not enabled on ingress */
957 if (!__in6_dev_get(skb->dev)->cnf.ioam6_enabled)
960 /* Truncated Option header */
961 hdr = (struct ioam6_hdr *)(skb_network_header(skb) + optoff);
962 if (hdr->opt_len < 2)
966 case IOAM6_TYPE_PREALLOC:
967 /* Truncated Pre-allocated Trace header */
968 if (hdr->opt_len < 2 + sizeof(*trace))
971 /* Malformed Pre-allocated Trace header */
972 trace = (struct ioam6_trace_hdr *)((u8 *)hdr + sizeof(*hdr));
973 if (hdr->opt_len < 2 + sizeof(*trace) + trace->remlen * 4)
976 /* Ignore if the IOAM namespace is unknown */
977 ns = ioam6_namespace(ipv6_skb_net(skb), trace->namespace_id);
981 if (!skb_valid_dst(skb))
982 ip6_route_input(skb);
984 ioam6_fill_trace_data(skb, ns, trace, true);
994 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
1000 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
1002 const unsigned char *nh = skb_network_header(skb);
1006 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
1007 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
1009 SKB_DR_SET(reason, IP_INHDR);
1013 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
1014 if (pkt_len <= IPV6_MAXPLEN) {
1015 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff + 2,
1016 SKB_DROP_REASON_IP_INHDR);
1019 if (ipv6_hdr(skb)->payload_len) {
1020 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff,
1021 SKB_DROP_REASON_IP_INHDR);
1025 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
1026 SKB_DR_SET(reason, PKT_TOO_SMALL);
1030 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
1033 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
1037 kfree_skb_reason(skb, reason);
1041 /* CALIPSO RFC 5570 */
1043 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
1045 const unsigned char *nh = skb_network_header(skb);
1047 if (nh[optoff + 1] < 8)
1050 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
1053 if (!calipso_validate(skb, nh + optoff))
1059 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
1063 int ipv6_parse_hopopts(struct sk_buff *skb)
1065 struct inet6_skb_parm *opt = IP6CB(skb);
1066 struct net *net = dev_net(skb->dev);
1070 * skb_network_header(skb) is equal to skb->data, and
1071 * skb_network_header_len(skb) is always equal to
1072 * sizeof(struct ipv6hdr) by definition of
1073 * hop-by-hop options.
1075 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1076 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1077 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1083 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1084 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1087 opt->flags |= IP6SKB_HOPBYHOP;
1088 if (ip6_parse_tlv(true, skb, net->ipv6.sysctl.max_hbh_opts_cnt)) {
1089 skb->transport_header += extlen;
1091 opt->nhoff = sizeof(struct ipv6hdr);
1098 * Creating outbound headers.
1100 * "build" functions work when skb is filled from head to tail (datagram)
1101 * "push" functions work when headers are added from tail to head (tcp)
1103 * In both cases we assume, that caller reserved enough room
1107 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1108 struct ipv6_rt_hdr *opt,
1109 struct in6_addr **addr_p, struct in6_addr *saddr)
1111 struct rt0_hdr *phdr, *ihdr;
1114 ihdr = (struct rt0_hdr *) opt;
1116 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1117 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1119 hops = ihdr->rt_hdr.hdrlen >> 1;
1122 memcpy(phdr->addr, ihdr->addr + 1,
1123 (hops - 1) * sizeof(struct in6_addr));
1125 phdr->addr[hops - 1] = **addr_p;
1126 *addr_p = ihdr->addr;
1128 phdr->rt_hdr.nexthdr = *proto;
1129 *proto = NEXTHDR_ROUTING;
1132 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1133 struct ipv6_rt_hdr *opt,
1134 struct in6_addr **addr_p, struct in6_addr *saddr)
1136 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1139 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1140 plen = (sr_ihdr->hdrlen + 1) << 3;
1142 sr_phdr = skb_push(skb, plen);
1143 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1145 hops = sr_ihdr->first_segment + 1;
1146 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1147 (hops - 1) * sizeof(struct in6_addr));
1149 sr_phdr->segments[0] = **addr_p;
1150 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1152 if (sr_ihdr->hdrlen > hops * 2) {
1153 int tlvs_offset, tlvs_length;
1155 tlvs_offset = (1 + hops * 2) << 3;
1156 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1157 memcpy((char *)sr_phdr + tlvs_offset,
1158 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1161 #ifdef CONFIG_IPV6_SEG6_HMAC
1162 if (sr_has_hmac(sr_phdr)) {
1163 struct net *net = NULL;
1166 net = dev_net(skb->dev);
1168 net = sock_net(skb->sk);
1173 seg6_push_hmac(net, saddr, sr_phdr);
1177 sr_phdr->nexthdr = *proto;
1178 *proto = NEXTHDR_ROUTING;
1181 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1182 struct ipv6_rt_hdr *opt,
1183 struct in6_addr **addr_p, struct in6_addr *saddr)
1185 switch (opt->type) {
1186 case IPV6_SRCRT_TYPE_0:
1187 case IPV6_SRCRT_STRICT:
1188 case IPV6_SRCRT_TYPE_2:
1189 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1191 case IPV6_SRCRT_TYPE_4:
1192 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1199 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1201 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1203 memcpy(h, opt, ipv6_optlen(opt));
1204 h->nexthdr = *proto;
1208 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1210 struct in6_addr **daddr, struct in6_addr *saddr)
1213 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1215 * IPV6_RTHDRDSTOPTS is ignored
1216 * unless IPV6_RTHDR is set (RFC3542).
1219 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1222 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1225 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1228 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1230 EXPORT_SYMBOL(ipv6_push_frag_opts);
1232 struct ipv6_txoptions *
1233 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1235 struct ipv6_txoptions *opt2;
1237 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1239 long dif = (char *)opt2 - (char *)opt;
1240 memcpy(opt2, opt, opt->tot_len);
1242 *((char **)&opt2->hopopt) += dif;
1244 *((char **)&opt2->dst0opt) += dif;
1246 *((char **)&opt2->dst1opt) += dif;
1248 *((char **)&opt2->srcrt) += dif;
1249 refcount_set(&opt2->refcnt, 1);
1253 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1255 static void ipv6_renew_option(int renewtype,
1256 struct ipv6_opt_hdr **dest,
1257 struct ipv6_opt_hdr *old,
1258 struct ipv6_opt_hdr *new,
1259 int newtype, char **p)
1261 struct ipv6_opt_hdr *src;
1263 src = (renewtype == newtype ? new : old);
1267 memcpy(*p, src, ipv6_optlen(src));
1268 *dest = (struct ipv6_opt_hdr *)*p;
1269 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1273 * ipv6_renew_options - replace a specific ext hdr with a new one.
1275 * @sk: sock from which to allocate memory
1276 * @opt: original options
1277 * @newtype: option type to replace in @opt
1278 * @newopt: new option of type @newtype to replace (user-mem)
1280 * Returns a new set of options which is a copy of @opt with the
1281 * option type @newtype replaced with @newopt.
1283 * @opt may be NULL, in which case a new set of options is returned
1284 * containing just @newopt.
1286 * @newopt may be NULL, in which case the specified option type is
1287 * not copied into the new set of options.
1289 * The new set of options is allocated from the socket option memory
1292 struct ipv6_txoptions *
1293 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1294 int newtype, struct ipv6_opt_hdr *newopt)
1298 struct ipv6_txoptions *opt2;
1301 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1302 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1303 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1304 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1305 if (newtype != IPV6_RTHDR && opt->srcrt)
1306 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1307 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1308 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1312 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1317 tot_len += sizeof(*opt2);
1318 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1320 return ERR_PTR(-ENOBUFS);
1322 memset(opt2, 0, tot_len);
1323 refcount_set(&opt2->refcnt, 1);
1324 opt2->tot_len = tot_len;
1325 p = (char *)(opt2 + 1);
1327 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1328 (opt ? opt->hopopt : NULL),
1329 newopt, newtype, &p);
1330 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1331 (opt ? opt->dst0opt : NULL),
1332 newopt, newtype, &p);
1333 ipv6_renew_option(IPV6_RTHDR,
1334 (struct ipv6_opt_hdr **)&opt2->srcrt,
1335 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1336 newopt, newtype, &p);
1337 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1338 (opt ? opt->dst1opt : NULL),
1339 newopt, newtype, &p);
1341 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1342 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1343 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1344 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1349 struct ipv6_txoptions *__ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1350 struct ipv6_txoptions *opt)
1353 * ignore the dest before srcrt unless srcrt is being included.
1356 if (opt->dst0opt && !opt->srcrt) {
1357 if (opt_space != opt) {
1358 memcpy(opt_space, opt, sizeof(*opt_space));
1361 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1362 opt->dst0opt = NULL;
1367 EXPORT_SYMBOL_GPL(__ipv6_fixup_options);
1370 * fl6_update_dst - update flowi destination address with info given
1371 * by srcrt option, if any.
1373 * @fl6: flowi6 for which daddr is to be updated
1374 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1375 * @orig: copy of original daddr address if modified
1377 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1378 * and initial value of fl6->daddr set in orig
1380 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1381 const struct ipv6_txoptions *opt,
1382 struct in6_addr *orig)
1384 if (!opt || !opt->srcrt)
1389 switch (opt->srcrt->type) {
1390 case IPV6_SRCRT_TYPE_0:
1391 case IPV6_SRCRT_STRICT:
1392 case IPV6_SRCRT_TYPE_2:
1393 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1395 case IPV6_SRCRT_TYPE_4:
1397 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1399 fl6->daddr = srh->segments[srh->segments_left];
1408 EXPORT_SYMBOL_GPL(fl6_update_dst);