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))
184 nh = skb_network_header(skb);
187 if (!ipv6_hop_jumbo(skb, off))
190 case IPV6_TLV_CALIPSO:
191 if (!ipv6_hop_calipso(skb, off))
195 if (!ip6_tlvopt_unknown(skb, off,
202 #if IS_ENABLED(CONFIG_IPV6_MIP6)
204 if (!ipv6_dest_hao(skb, off))
209 if (!ip6_tlvopt_unknown(skb, off,
224 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
228 /*****************************
229 Destination options header.
230 *****************************/
232 #if IS_ENABLED(CONFIG_IPV6_MIP6)
233 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
235 struct ipv6_destopt_hao *hao;
236 struct inet6_skb_parm *opt = IP6CB(skb);
237 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
242 net_dbg_ratelimited("hao duplicated\n");
245 opt->dsthao = opt->dst1;
248 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
250 if (hao->length != 16) {
251 net_dbg_ratelimited("hao invalid option length = %d\n",
253 SKB_DR_SET(reason, IP_INHDR);
257 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
258 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
260 SKB_DR_SET(reason, INVALID_PROTO);
264 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
265 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
266 if (unlikely(ret < 0)) {
267 SKB_DR_SET(reason, XFRM_POLICY);
271 if (skb_cloned(skb)) {
272 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
275 /* update all variable using below by copied skbuff */
276 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
278 ipv6h = ipv6_hdr(skb);
281 if (skb->ip_summed == CHECKSUM_COMPLETE)
282 skb->ip_summed = CHECKSUM_NONE;
284 swap(ipv6h->saddr, hao->addr);
286 if (skb->tstamp == 0)
287 __net_timestamp(skb);
292 kfree_skb_reason(skb, reason);
297 static int ipv6_destopt_rcv(struct sk_buff *skb)
299 struct inet6_dev *idev = __in6_dev_get(skb->dev);
300 struct inet6_skb_parm *opt = IP6CB(skb);
301 #if IS_ENABLED(CONFIG_IPV6_MIP6)
304 struct dst_entry *dst = skb_dst(skb);
305 struct net *net = dev_net(skb->dev);
308 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
309 !pskb_may_pull(skb, (skb_transport_offset(skb) +
310 ((skb_transport_header(skb)[1] + 1) << 3)))) {
311 __IP6_INC_STATS(dev_net(dst->dev), idev,
312 IPSTATS_MIB_INHDRERRORS);
318 extlen = (skb_transport_header(skb)[1] + 1) << 3;
319 if (extlen > net->ipv6.sysctl.max_dst_opts_len)
322 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
323 #if IS_ENABLED(CONFIG_IPV6_MIP6)
327 if (ip6_parse_tlv(false, skb, net->ipv6.sysctl.max_dst_opts_cnt)) {
328 skb->transport_header += extlen;
330 #if IS_ENABLED(CONFIG_IPV6_MIP6)
333 opt->nhoff = opt->dst1;
338 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
342 static void seg6_update_csum(struct sk_buff *skb)
344 struct ipv6_sr_hdr *hdr;
345 struct in6_addr *addr;
348 /* srh is at transport offset and seg_left is already decremented
349 * but daddr is not yet updated with next segment
352 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
353 addr = hdr->segments + hdr->segments_left;
355 hdr->segments_left++;
356 from = *(__be32 *)hdr;
358 hdr->segments_left--;
361 /* update skb csum with diff resulting from seg_left decrement */
363 update_csum_diff4(skb, from, to);
365 /* compute csum diff between current and next segment and update */
367 update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
371 static int ipv6_srh_rcv(struct sk_buff *skb)
373 struct inet6_skb_parm *opt = IP6CB(skb);
374 struct net *net = dev_net(skb->dev);
375 struct ipv6_sr_hdr *hdr;
376 struct inet6_dev *idev;
377 struct in6_addr *addr;
380 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
382 idev = __in6_dev_get(skb->dev);
384 accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
385 if (accept_seg6 > idev->cnf.seg6_enabled)
386 accept_seg6 = idev->cnf.seg6_enabled;
393 #ifdef CONFIG_IPV6_SEG6_HMAC
394 if (!seg6_hmac_validate_skb(skb)) {
401 if (hdr->segments_left == 0) {
402 if (hdr->nexthdr == NEXTHDR_IPV6 || hdr->nexthdr == NEXTHDR_IPV4) {
403 int offset = (hdr->hdrlen + 1) << 3;
405 skb_postpull_rcsum(skb, skb_network_header(skb),
406 skb_network_header_len(skb));
408 if (!pskb_pull(skb, offset)) {
412 skb_postpull_rcsum(skb, skb_transport_header(skb),
415 skb_reset_network_header(skb);
416 skb_reset_transport_header(skb);
417 skb->encapsulation = 0;
418 if (hdr->nexthdr == NEXTHDR_IPV4)
419 skb->protocol = htons(ETH_P_IP);
420 __skb_tunnel_rx(skb, skb->dev, net);
426 opt->srcrt = skb_network_header_len(skb);
427 opt->lastopt = opt->srcrt;
428 skb->transport_header += (hdr->hdrlen + 1) << 3;
429 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
434 if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
435 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
436 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
437 ((&hdr->segments_left) -
438 skb_network_header(skb)));
442 if (skb_cloned(skb)) {
443 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
444 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
445 IPSTATS_MIB_OUTDISCARDS);
451 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
453 hdr->segments_left--;
454 addr = hdr->segments + hdr->segments_left;
456 skb_push(skb, sizeof(struct ipv6hdr));
458 if (skb->ip_summed == CHECKSUM_COMPLETE)
459 seg6_update_csum(skb);
461 ipv6_hdr(skb)->daddr = *addr;
465 ip6_route_input(skb);
467 if (skb_dst(skb)->error) {
472 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
473 if (ipv6_hdr(skb)->hop_limit <= 1) {
474 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
475 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
476 ICMPV6_EXC_HOPLIMIT, 0);
480 ipv6_hdr(skb)->hop_limit--;
482 skb_pull(skb, sizeof(struct ipv6hdr));
491 static int ipv6_rpl_srh_rcv(struct sk_buff *skb)
493 struct ipv6_rpl_sr_hdr *hdr, *ohdr, *chdr;
494 struct inet6_skb_parm *opt = IP6CB(skb);
495 struct net *net = dev_net(skb->dev);
496 struct inet6_dev *idev;
497 struct ipv6hdr *oldhdr;
504 idev = __in6_dev_get(skb->dev);
506 accept_rpl_seg = net->ipv6.devconf_all->rpl_seg_enabled;
507 if (accept_rpl_seg > idev->cnf.rpl_seg_enabled)
508 accept_rpl_seg = idev->cnf.rpl_seg_enabled;
510 if (!accept_rpl_seg) {
516 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
518 if (hdr->segments_left == 0) {
519 if (hdr->nexthdr == NEXTHDR_IPV6) {
520 int offset = (hdr->hdrlen + 1) << 3;
522 skb_postpull_rcsum(skb, skb_network_header(skb),
523 skb_network_header_len(skb));
525 if (!pskb_pull(skb, offset)) {
529 skb_postpull_rcsum(skb, skb_transport_header(skb),
532 skb_reset_network_header(skb);
533 skb_reset_transport_header(skb);
534 skb->encapsulation = 0;
536 __skb_tunnel_rx(skb, skb->dev, net);
542 opt->srcrt = skb_network_header_len(skb);
543 opt->lastopt = opt->srcrt;
544 skb->transport_header += (hdr->hdrlen + 1) << 3;
545 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
550 if (!pskb_may_pull(skb, sizeof(*hdr))) {
555 n = (hdr->hdrlen << 3) - hdr->pad - (16 - hdr->cmpre);
556 r = do_div(n, (16 - hdr->cmpri));
557 /* checks if calculation was without remainder and n fits into
558 * unsigned char which is segments_left field. Should not be
561 if (r || (n + 1) > 255) {
566 if (hdr->segments_left > n + 1) {
567 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
568 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
569 ((&hdr->segments_left) -
570 skb_network_header(skb)));
574 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
580 hdr->segments_left--;
581 i = n - hdr->segments_left;
583 buf = kcalloc(struct_size(hdr, segments.addr, n + 2), 2, GFP_ATOMIC);
584 if (unlikely(!buf)) {
589 ohdr = (struct ipv6_rpl_sr_hdr *)buf;
590 ipv6_rpl_srh_decompress(ohdr, hdr, &ipv6_hdr(skb)->daddr, n);
591 chdr = (struct ipv6_rpl_sr_hdr *)(buf + ((ohdr->hdrlen + 1) << 3));
593 if ((ipv6_addr_type(&ipv6_hdr(skb)->daddr) & IPV6_ADDR_MULTICAST) ||
594 (ipv6_addr_type(&ohdr->rpl_segaddr[i]) & IPV6_ADDR_MULTICAST)) {
600 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
602 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
608 swap(ipv6_hdr(skb)->daddr, ohdr->rpl_segaddr[i]);
610 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
612 oldhdr = ipv6_hdr(skb);
614 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
615 skb_postpull_rcsum(skb, oldhdr,
616 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
617 if (unlikely(!hdr->segments_left)) {
618 if (pskb_expand_head(skb, sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3), 0,
620 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_OUTDISCARDS);
626 oldhdr = ipv6_hdr(skb);
628 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
629 skb_reset_network_header(skb);
630 skb_mac_header_rebuild(skb);
631 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
633 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
634 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
636 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
637 skb_postpush_rcsum(skb, ipv6_hdr(skb),
638 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
644 ip6_route_input(skb);
646 if (skb_dst(skb)->error) {
651 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
652 if (ipv6_hdr(skb)->hop_limit <= 1) {
653 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
654 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
655 ICMPV6_EXC_HOPLIMIT, 0);
659 ipv6_hdr(skb)->hop_limit--;
661 skb_pull(skb, sizeof(struct ipv6hdr));
670 /********************************
672 ********************************/
674 /* called with rcu_read_lock() */
675 static int ipv6_rthdr_rcv(struct sk_buff *skb)
677 struct inet6_dev *idev = __in6_dev_get(skb->dev);
678 struct inet6_skb_parm *opt = IP6CB(skb);
679 struct in6_addr *addr = NULL;
680 struct in6_addr daddr;
682 struct ipv6_rt_hdr *hdr;
683 struct rt0_hdr *rthdr;
684 struct net *net = dev_net(skb->dev);
685 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
687 if (idev && accept_source_route > idev->cnf.accept_source_route)
688 accept_source_route = idev->cnf.accept_source_route;
690 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
691 !pskb_may_pull(skb, (skb_transport_offset(skb) +
692 ((skb_transport_header(skb)[1] + 1) << 3)))) {
693 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
698 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
700 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
701 skb->pkt_type != PACKET_HOST) {
702 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
708 case IPV6_SRCRT_TYPE_4:
709 /* segment routing */
710 return ipv6_srh_rcv(skb);
711 case IPV6_SRCRT_TYPE_3:
712 /* rpl segment routing */
713 return ipv6_rpl_srh_rcv(skb);
719 if (hdr->segments_left == 0) {
721 #if IS_ENABLED(CONFIG_IPV6_MIP6)
722 case IPV6_SRCRT_TYPE_2:
723 /* Silently discard type 2 header unless it was
727 __IP6_INC_STATS(net, idev,
728 IPSTATS_MIB_INADDRERRORS);
738 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
739 skb->transport_header += (hdr->hdrlen + 1) << 3;
740 opt->dst0 = opt->dst1;
742 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
747 #if IS_ENABLED(CONFIG_IPV6_MIP6)
748 case IPV6_SRCRT_TYPE_2:
749 if (accept_source_route < 0)
751 /* Silently discard invalid RTH type 2 */
752 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
753 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
764 * This is the routing header forwarding algorithm from
768 n = hdr->hdrlen >> 1;
770 if (hdr->segments_left > n) {
771 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
772 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
773 ((&hdr->segments_left) -
774 skb_network_header(skb)));
778 /* We are about to mangle packet header. Be careful!
779 Do not damage packets queued somewhere.
781 if (skb_cloned(skb)) {
782 /* the copy is a forwarded packet */
783 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
784 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
785 IPSTATS_MIB_OUTDISCARDS);
789 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
792 if (skb->ip_summed == CHECKSUM_COMPLETE)
793 skb->ip_summed = CHECKSUM_NONE;
795 i = n - --hdr->segments_left;
797 rthdr = (struct rt0_hdr *) hdr;
802 #if IS_ENABLED(CONFIG_IPV6_MIP6)
803 case IPV6_SRCRT_TYPE_2:
804 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
805 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
806 IPPROTO_ROUTING) < 0) {
807 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
811 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
812 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
822 if (ipv6_addr_is_multicast(addr)) {
823 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
829 *addr = ipv6_hdr(skb)->daddr;
830 ipv6_hdr(skb)->daddr = daddr;
833 ip6_route_input(skb);
834 if (skb_dst(skb)->error) {
835 skb_push(skb, skb->data - skb_network_header(skb));
840 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
841 if (ipv6_hdr(skb)->hop_limit <= 1) {
842 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
843 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
848 ipv6_hdr(skb)->hop_limit--;
852 skb_push(skb, skb->data - skb_network_header(skb));
857 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
858 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
859 (&hdr->type) - skb_network_header(skb));
863 static const struct inet6_protocol rthdr_protocol = {
864 .handler = ipv6_rthdr_rcv,
865 .flags = INET6_PROTO_NOPOLICY,
868 static const struct inet6_protocol destopt_protocol = {
869 .handler = ipv6_destopt_rcv,
870 .flags = INET6_PROTO_NOPOLICY,
873 static const struct inet6_protocol nodata_protocol = {
874 .handler = dst_discard,
875 .flags = INET6_PROTO_NOPOLICY,
878 int __init ipv6_exthdrs_init(void)
882 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
886 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
890 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
897 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
899 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
903 void ipv6_exthdrs_exit(void)
905 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
906 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
907 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
910 /**********************************
912 **********************************/
915 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
917 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
919 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
922 /* Router Alert as of RFC 2711 */
924 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
926 const unsigned char *nh = skb_network_header(skb);
928 if (nh[optoff + 1] == 2) {
929 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
930 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
933 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
935 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
941 static bool ipv6_hop_ioam(struct sk_buff *skb, int optoff)
943 struct ioam6_trace_hdr *trace;
944 struct ioam6_namespace *ns;
945 struct ioam6_hdr *hdr;
947 /* Bad alignment (must be 4n-aligned) */
951 /* Ignore if IOAM is not enabled on ingress */
952 if (!__in6_dev_get(skb->dev)->cnf.ioam6_enabled)
955 /* Truncated Option header */
956 hdr = (struct ioam6_hdr *)(skb_network_header(skb) + optoff);
957 if (hdr->opt_len < 2)
961 case IOAM6_TYPE_PREALLOC:
962 /* Truncated Pre-allocated Trace header */
963 if (hdr->opt_len < 2 + sizeof(*trace))
966 /* Malformed Pre-allocated Trace header */
967 trace = (struct ioam6_trace_hdr *)((u8 *)hdr + sizeof(*hdr));
968 if (hdr->opt_len < 2 + sizeof(*trace) + trace->remlen * 4)
971 /* Ignore if the IOAM namespace is unknown */
972 ns = ioam6_namespace(ipv6_skb_net(skb), trace->namespace_id);
976 if (!skb_valid_dst(skb))
977 ip6_route_input(skb);
979 /* About to mangle packet header */
980 if (skb_ensure_writable(skb, optoff + 2 + hdr->opt_len))
983 /* Trace pointer may have changed */
984 trace = (struct ioam6_trace_hdr *)(skb_network_header(skb)
985 + optoff + sizeof(*hdr));
987 ioam6_fill_trace_data(skb, ns, trace, true);
997 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
1003 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
1005 const unsigned char *nh = skb_network_header(skb);
1009 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
1010 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
1012 SKB_DR_SET(reason, IP_INHDR);
1016 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
1017 if (pkt_len <= IPV6_MAXPLEN) {
1018 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff + 2,
1019 SKB_DROP_REASON_IP_INHDR);
1022 if (ipv6_hdr(skb)->payload_len) {
1023 icmpv6_param_prob_reason(skb, ICMPV6_HDR_FIELD, optoff,
1024 SKB_DROP_REASON_IP_INHDR);
1028 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
1029 SKB_DR_SET(reason, PKT_TOO_SMALL);
1033 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
1036 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
1040 kfree_skb_reason(skb, reason);
1044 /* CALIPSO RFC 5570 */
1046 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
1048 const unsigned char *nh = skb_network_header(skb);
1050 if (nh[optoff + 1] < 8)
1053 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
1056 if (!calipso_validate(skb, nh + optoff))
1062 kfree_skb_reason(skb, SKB_DROP_REASON_IP_INHDR);
1066 int ipv6_parse_hopopts(struct sk_buff *skb)
1068 struct inet6_skb_parm *opt = IP6CB(skb);
1069 struct net *net = dev_net(skb->dev);
1073 * skb_network_header(skb) is equal to skb->data, and
1074 * skb_network_header_len(skb) is always equal to
1075 * sizeof(struct ipv6hdr) by definition of
1076 * hop-by-hop options.
1078 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1079 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1080 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1086 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1087 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1090 opt->flags |= IP6SKB_HOPBYHOP;
1091 if (ip6_parse_tlv(true, skb, net->ipv6.sysctl.max_hbh_opts_cnt)) {
1092 skb->transport_header += extlen;
1094 opt->nhoff = sizeof(struct ipv6hdr);
1101 * Creating outbound headers.
1103 * "build" functions work when skb is filled from head to tail (datagram)
1104 * "push" functions work when headers are added from tail to head (tcp)
1106 * In both cases we assume, that caller reserved enough room
1110 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1111 struct ipv6_rt_hdr *opt,
1112 struct in6_addr **addr_p, struct in6_addr *saddr)
1114 struct rt0_hdr *phdr, *ihdr;
1117 ihdr = (struct rt0_hdr *) opt;
1119 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1120 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1122 hops = ihdr->rt_hdr.hdrlen >> 1;
1125 memcpy(phdr->addr, ihdr->addr + 1,
1126 (hops - 1) * sizeof(struct in6_addr));
1128 phdr->addr[hops - 1] = **addr_p;
1129 *addr_p = ihdr->addr;
1131 phdr->rt_hdr.nexthdr = *proto;
1132 *proto = NEXTHDR_ROUTING;
1135 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1136 struct ipv6_rt_hdr *opt,
1137 struct in6_addr **addr_p, struct in6_addr *saddr)
1139 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1142 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1143 plen = (sr_ihdr->hdrlen + 1) << 3;
1145 sr_phdr = skb_push(skb, plen);
1146 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1148 hops = sr_ihdr->first_segment + 1;
1149 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1150 (hops - 1) * sizeof(struct in6_addr));
1152 sr_phdr->segments[0] = **addr_p;
1153 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1155 if (sr_ihdr->hdrlen > hops * 2) {
1156 int tlvs_offset, tlvs_length;
1158 tlvs_offset = (1 + hops * 2) << 3;
1159 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1160 memcpy((char *)sr_phdr + tlvs_offset,
1161 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1164 #ifdef CONFIG_IPV6_SEG6_HMAC
1165 if (sr_has_hmac(sr_phdr)) {
1166 struct net *net = NULL;
1169 net = dev_net(skb->dev);
1171 net = sock_net(skb->sk);
1176 seg6_push_hmac(net, saddr, sr_phdr);
1180 sr_phdr->nexthdr = *proto;
1181 *proto = NEXTHDR_ROUTING;
1184 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1185 struct ipv6_rt_hdr *opt,
1186 struct in6_addr **addr_p, struct in6_addr *saddr)
1188 switch (opt->type) {
1189 case IPV6_SRCRT_TYPE_0:
1190 case IPV6_SRCRT_STRICT:
1191 case IPV6_SRCRT_TYPE_2:
1192 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1194 case IPV6_SRCRT_TYPE_4:
1195 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1202 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1204 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1206 memcpy(h, opt, ipv6_optlen(opt));
1207 h->nexthdr = *proto;
1211 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1213 struct in6_addr **daddr, struct in6_addr *saddr)
1216 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1218 * IPV6_RTHDRDSTOPTS is ignored
1219 * unless IPV6_RTHDR is set (RFC3542).
1222 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1225 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1228 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1231 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1233 EXPORT_SYMBOL(ipv6_push_frag_opts);
1235 struct ipv6_txoptions *
1236 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1238 struct ipv6_txoptions *opt2;
1240 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1242 long dif = (char *)opt2 - (char *)opt;
1243 memcpy(opt2, opt, opt->tot_len);
1245 *((char **)&opt2->hopopt) += dif;
1247 *((char **)&opt2->dst0opt) += dif;
1249 *((char **)&opt2->dst1opt) += dif;
1251 *((char **)&opt2->srcrt) += dif;
1252 refcount_set(&opt2->refcnt, 1);
1256 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1258 static void ipv6_renew_option(int renewtype,
1259 struct ipv6_opt_hdr **dest,
1260 struct ipv6_opt_hdr *old,
1261 struct ipv6_opt_hdr *new,
1262 int newtype, char **p)
1264 struct ipv6_opt_hdr *src;
1266 src = (renewtype == newtype ? new : old);
1270 memcpy(*p, src, ipv6_optlen(src));
1271 *dest = (struct ipv6_opt_hdr *)*p;
1272 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1276 * ipv6_renew_options - replace a specific ext hdr with a new one.
1278 * @sk: sock from which to allocate memory
1279 * @opt: original options
1280 * @newtype: option type to replace in @opt
1281 * @newopt: new option of type @newtype to replace (user-mem)
1283 * Returns a new set of options which is a copy of @opt with the
1284 * option type @newtype replaced with @newopt.
1286 * @opt may be NULL, in which case a new set of options is returned
1287 * containing just @newopt.
1289 * @newopt may be NULL, in which case the specified option type is
1290 * not copied into the new set of options.
1292 * The new set of options is allocated from the socket option memory
1295 struct ipv6_txoptions *
1296 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1297 int newtype, struct ipv6_opt_hdr *newopt)
1301 struct ipv6_txoptions *opt2;
1304 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1305 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1306 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1307 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1308 if (newtype != IPV6_RTHDR && opt->srcrt)
1309 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1310 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1311 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1315 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1320 tot_len += sizeof(*opt2);
1321 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1323 return ERR_PTR(-ENOBUFS);
1325 memset(opt2, 0, tot_len);
1326 refcount_set(&opt2->refcnt, 1);
1327 opt2->tot_len = tot_len;
1328 p = (char *)(opt2 + 1);
1330 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1331 (opt ? opt->hopopt : NULL),
1332 newopt, newtype, &p);
1333 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1334 (opt ? opt->dst0opt : NULL),
1335 newopt, newtype, &p);
1336 ipv6_renew_option(IPV6_RTHDR,
1337 (struct ipv6_opt_hdr **)&opt2->srcrt,
1338 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1339 newopt, newtype, &p);
1340 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1341 (opt ? opt->dst1opt : NULL),
1342 newopt, newtype, &p);
1344 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1345 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1346 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1347 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1352 struct ipv6_txoptions *__ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1353 struct ipv6_txoptions *opt)
1356 * ignore the dest before srcrt unless srcrt is being included.
1359 if (opt->dst0opt && !opt->srcrt) {
1360 if (opt_space != opt) {
1361 memcpy(opt_space, opt, sizeof(*opt_space));
1364 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1365 opt->dst0opt = NULL;
1370 EXPORT_SYMBOL_GPL(__ipv6_fixup_options);
1373 * fl6_update_dst - update flowi destination address with info given
1374 * by srcrt option, if any.
1376 * @fl6: flowi6 for which daddr is to be updated
1377 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1378 * @orig: copy of original daddr address if modified
1380 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1381 * and initial value of fl6->daddr set in orig
1383 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1384 const struct ipv6_txoptions *opt,
1385 struct in6_addr *orig)
1387 if (!opt || !opt->srcrt)
1392 switch (opt->srcrt->type) {
1393 case IPV6_SRCRT_TYPE_0:
1394 case IPV6_SRCRT_STRICT:
1395 case IPV6_SRCRT_TYPE_2:
1396 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1398 case IPV6_SRCRT_TYPE_4:
1400 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1402 fl6->daddr = srh->segments[srh->segments_left];
1411 EXPORT_SYMBOL_GPL(fl6_update_dst);