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 <linux/ioam6_genl.h>
54 #include <net/ioam6.h>
55 #include <net/dst_metadata.h>
57 #include <linux/uaccess.h>
59 /*********************
61 *********************/
63 /* An unknown option is detected, decide what to do */
65 static bool ip6_tlvopt_unknown(struct sk_buff
*skb
, int optoff
,
66 bool disallow_unknowns
)
68 if (disallow_unknowns
) {
69 /* If unknown TLVs are disallowed by configuration
70 * then always silently drop packet. Note this also
71 * means no ICMP parameter problem is sent which
72 * could be a good property to mitigate a reflection DOS
79 switch ((skb_network_header(skb
)[optoff
] & 0xC0) >> 6) {
83 case 1: /* drop packet */
86 case 3: /* Send ICMP if not a multicast address and drop packet */
87 /* Actually, it is redundant check. icmp_send
88 will recheck in any case.
90 if (ipv6_addr_is_multicast(&ipv6_hdr(skb
)->daddr
))
93 case 2: /* send ICMP PARM PROB regardless and drop packet */
94 icmpv6_param_prob_reason(skb
, ICMPV6_UNK_OPTION
, optoff
,
95 SKB_DROP_REASON_UNHANDLED_PROTO
);
100 kfree_skb_reason(skb
, SKB_DROP_REASON_UNHANDLED_PROTO
);
104 static bool ipv6_hop_ra(struct sk_buff
*skb
, int optoff
);
105 static bool ipv6_hop_ioam(struct sk_buff
*skb
, int optoff
);
106 static bool ipv6_hop_jumbo(struct sk_buff
*skb
, int optoff
);
107 static bool ipv6_hop_calipso(struct sk_buff
*skb
, int optoff
);
108 #if IS_ENABLED(CONFIG_IPV6_MIP6)
109 static bool ipv6_dest_hao(struct sk_buff
*skb
, int optoff
);
112 /* Parse tlv encoded option header (hop-by-hop or destination) */
114 static bool ip6_parse_tlv(bool hopbyhop
,
118 int len
= (skb_transport_header(skb
)[1] + 1) << 3;
119 const unsigned char *nh
= skb_network_header(skb
);
120 int off
= skb_network_header_len(skb
);
121 bool disallow_unknowns
= false;
125 if (unlikely(max_count
< 0)) {
126 disallow_unknowns
= true;
127 max_count
= -max_count
;
136 if (nh
[off
] == IPV6_TLV_PAD1
) {
146 optlen
= nh
[off
+ 1] + 2;
150 if (nh
[off
] == IPV6_TLV_PADN
) {
151 /* RFC 2460 states that the purpose of PadN is
152 * to align the containing header to multiples
153 * of 8. 7 is therefore the highest valid value.
154 * See also RFC 4942, Section 2.1.9.5.
159 /* RFC 4942 recommends receiving hosts to
160 * actively check PadN payload to contain
163 for (i
= 2; i
< optlen
; i
++) {
164 if (nh
[off
+ i
] != 0)
169 if (tlv_count
> max_count
)
174 case IPV6_TLV_ROUTERALERT
:
175 if (!ipv6_hop_ra(skb
, off
))
179 if (!ipv6_hop_ioam(skb
, off
))
182 nh
= skb_network_header(skb
);
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
= min(READ_ONCE(net
->ipv6
.devconf_all
->seg6_enabled
),
383 READ_ONCE(idev
->cnf
.seg6_enabled
));
390 #ifdef CONFIG_IPV6_SEG6_HMAC
391 if (!seg6_hmac_validate_skb(skb
)) {
398 if (hdr
->segments_left
== 0) {
399 if (hdr
->nexthdr
== NEXTHDR_IPV6
|| hdr
->nexthdr
== NEXTHDR_IPV4
) {
400 int offset
= (hdr
->hdrlen
+ 1) << 3;
402 skb_postpull_rcsum(skb
, skb_network_header(skb
),
403 skb_network_header_len(skb
));
404 skb_pull(skb
, offset
);
405 skb_postpull_rcsum(skb
, skb_transport_header(skb
),
408 skb_reset_network_header(skb
);
409 skb_reset_transport_header(skb
);
410 skb
->encapsulation
= 0;
411 if (hdr
->nexthdr
== NEXTHDR_IPV4
)
412 skb
->protocol
= htons(ETH_P_IP
);
413 __skb_tunnel_rx(skb
, skb
->dev
, net
);
419 opt
->srcrt
= skb_network_header_len(skb
);
420 opt
->lastopt
= opt
->srcrt
;
421 skb
->transport_header
+= (hdr
->hdrlen
+ 1) << 3;
422 opt
->nhoff
= (&hdr
->nexthdr
) - skb_network_header(skb
);
427 if (hdr
->segments_left
>= (hdr
->hdrlen
>> 1)) {
428 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
429 icmpv6_param_prob(skb
, ICMPV6_HDR_FIELD
,
430 ((&hdr
->segments_left
) -
431 skb_network_header(skb
)));
435 if (skb_cloned(skb
)) {
436 if (pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
)) {
437 __IP6_INC_STATS(net
, ip6_dst_idev(skb_dst(skb
)),
438 IPSTATS_MIB_OUTDISCARDS
);
443 hdr
= (struct ipv6_sr_hdr
*)skb_transport_header(skb
);
446 hdr
->segments_left
--;
447 addr
= hdr
->segments
+ hdr
->segments_left
;
449 skb_push(skb
, sizeof(struct ipv6hdr
));
451 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
452 seg6_update_csum(skb
);
454 ipv6_hdr(skb
)->daddr
= *addr
;
456 ip6_route_input(skb
);
458 if (skb_dst(skb
)->error
) {
463 if (skb_dst(skb
)->dev
->flags
& IFF_LOOPBACK
) {
464 if (ipv6_hdr(skb
)->hop_limit
<= 1) {
465 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
466 icmpv6_send(skb
, ICMPV6_TIME_EXCEED
,
467 ICMPV6_EXC_HOPLIMIT
, 0);
471 ipv6_hdr(skb
)->hop_limit
--;
473 skb_pull(skb
, sizeof(struct ipv6hdr
));
482 static int ipv6_rpl_srh_rcv(struct sk_buff
*skb
)
484 struct ipv6_rpl_sr_hdr
*hdr
, *ohdr
, *chdr
;
485 struct inet6_skb_parm
*opt
= IP6CB(skb
);
486 struct net
*net
= dev_net(skb
->dev
);
487 struct inet6_dev
*idev
;
488 struct ipv6hdr
*oldhdr
;
495 idev
= __in6_dev_get(skb
->dev
);
497 accept_rpl_seg
= net
->ipv6
.devconf_all
->rpl_seg_enabled
;
498 if (accept_rpl_seg
> idev
->cnf
.rpl_seg_enabled
)
499 accept_rpl_seg
= idev
->cnf
.rpl_seg_enabled
;
501 if (!accept_rpl_seg
) {
507 hdr
= (struct ipv6_rpl_sr_hdr
*)skb_transport_header(skb
);
509 if (hdr
->segments_left
== 0) {
510 if (hdr
->nexthdr
== NEXTHDR_IPV6
) {
511 int offset
= (hdr
->hdrlen
+ 1) << 3;
513 skb_postpull_rcsum(skb
, skb_network_header(skb
),
514 skb_network_header_len(skb
));
515 skb_pull(skb
, offset
);
516 skb_postpull_rcsum(skb
, skb_transport_header(skb
),
519 skb_reset_network_header(skb
);
520 skb_reset_transport_header(skb
);
521 skb
->encapsulation
= 0;
523 __skb_tunnel_rx(skb
, skb
->dev
, net
);
529 opt
->srcrt
= skb_network_header_len(skb
);
530 opt
->lastopt
= opt
->srcrt
;
531 skb
->transport_header
+= (hdr
->hdrlen
+ 1) << 3;
532 opt
->nhoff
= (&hdr
->nexthdr
) - skb_network_header(skb
);
537 n
= (hdr
->hdrlen
<< 3) - hdr
->pad
- (16 - hdr
->cmpre
);
538 r
= do_div(n
, (16 - hdr
->cmpri
));
539 /* checks if calculation was without remainder and n fits into
540 * unsigned char which is segments_left field. Should not be
543 if (r
|| (n
+ 1) > 255) {
548 if (hdr
->segments_left
> n
+ 1) {
549 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
550 icmpv6_param_prob(skb
, ICMPV6_HDR_FIELD
,
551 ((&hdr
->segments_left
) -
552 skb_network_header(skb
)));
556 hdr
->segments_left
--;
557 i
= n
- hdr
->segments_left
;
559 buf
= kcalloc(struct_size(hdr
, segments
.addr
, n
+ 2), 2, GFP_ATOMIC
);
560 if (unlikely(!buf
)) {
565 ohdr
= (struct ipv6_rpl_sr_hdr
*)buf
;
566 ipv6_rpl_srh_decompress(ohdr
, hdr
, &ipv6_hdr(skb
)->daddr
, n
);
567 chdr
= (struct ipv6_rpl_sr_hdr
*)(buf
+ ((ohdr
->hdrlen
+ 1) << 3));
569 if (ipv6_addr_is_multicast(&ohdr
->rpl_segaddr
[i
])) {
575 err
= ipv6_chk_rpl_srh_loop(net
, ohdr
->rpl_segaddr
, n
+ 1);
577 icmpv6_send(skb
, ICMPV6_PARAMPROB
, 0, 0);
583 swap(ipv6_hdr(skb
)->daddr
, ohdr
->rpl_segaddr
[i
]);
585 ipv6_rpl_srh_compress(chdr
, ohdr
, &ipv6_hdr(skb
)->daddr
, n
);
587 oldhdr
= ipv6_hdr(skb
);
589 skb_pull(skb
, ((hdr
->hdrlen
+ 1) << 3));
590 skb_postpull_rcsum(skb
, oldhdr
,
591 sizeof(struct ipv6hdr
) + ((hdr
->hdrlen
+ 1) << 3));
592 if (unlikely(!hdr
->segments_left
)) {
593 if (pskb_expand_head(skb
, sizeof(struct ipv6hdr
) + ((chdr
->hdrlen
+ 1) << 3), 0,
595 __IP6_INC_STATS(net
, ip6_dst_idev(skb_dst(skb
)), IPSTATS_MIB_OUTDISCARDS
);
601 oldhdr
= ipv6_hdr(skb
);
603 skb_push(skb
, ((chdr
->hdrlen
+ 1) << 3) + sizeof(struct ipv6hdr
));
604 skb_reset_network_header(skb
);
605 skb_mac_header_rebuild(skb
);
606 skb_set_transport_header(skb
, sizeof(struct ipv6hdr
));
608 memmove(ipv6_hdr(skb
), oldhdr
, sizeof(struct ipv6hdr
));
609 memcpy(skb_transport_header(skb
), chdr
, (chdr
->hdrlen
+ 1) << 3);
611 ipv6_hdr(skb
)->payload_len
= htons(skb
->len
- sizeof(struct ipv6hdr
));
612 skb_postpush_rcsum(skb
, ipv6_hdr(skb
),
613 sizeof(struct ipv6hdr
) + ((chdr
->hdrlen
+ 1) << 3));
617 ip6_route_input(skb
);
619 if (skb_dst(skb
)->error
) {
624 if (skb_dst(skb
)->dev
->flags
& IFF_LOOPBACK
) {
625 if (ipv6_hdr(skb
)->hop_limit
<= 1) {
626 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
627 icmpv6_send(skb
, ICMPV6_TIME_EXCEED
,
628 ICMPV6_EXC_HOPLIMIT
, 0);
632 ipv6_hdr(skb
)->hop_limit
--;
634 skb_pull(skb
, sizeof(struct ipv6hdr
));
643 /********************************
645 ********************************/
647 /* called with rcu_read_lock() */
648 static int ipv6_rthdr_rcv(struct sk_buff
*skb
)
650 struct inet6_dev
*idev
= __in6_dev_get(skb
->dev
);
651 struct inet6_skb_parm
*opt
= IP6CB(skb
);
652 struct in6_addr
*addr
= NULL
;
654 struct ipv6_rt_hdr
*hdr
;
655 struct rt0_hdr
*rthdr
;
656 struct net
*net
= dev_net(skb
->dev
);
657 int accept_source_route
;
659 accept_source_route
= READ_ONCE(net
->ipv6
.devconf_all
->accept_source_route
);
662 accept_source_route
= min(accept_source_route
,
663 READ_ONCE(idev
->cnf
.accept_source_route
));
665 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) + 8) ||
666 !pskb_may_pull(skb
, (skb_transport_offset(skb
) +
667 ((skb_transport_header(skb
)[1] + 1) << 3)))) {
668 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
673 hdr
= (struct ipv6_rt_hdr
*)skb_transport_header(skb
);
675 if (ipv6_addr_is_multicast(&ipv6_hdr(skb
)->daddr
) ||
676 skb
->pkt_type
!= PACKET_HOST
) {
677 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INADDRERRORS
);
683 case IPV6_SRCRT_TYPE_4
:
684 /* segment routing */
685 return ipv6_srh_rcv(skb
);
686 case IPV6_SRCRT_TYPE_3
:
687 /* rpl segment routing */
688 return ipv6_rpl_srh_rcv(skb
);
694 if (hdr
->segments_left
== 0) {
696 #if IS_ENABLED(CONFIG_IPV6_MIP6)
697 case IPV6_SRCRT_TYPE_2
:
698 /* Silently discard type 2 header unless it was
702 __IP6_INC_STATS(net
, idev
,
703 IPSTATS_MIB_INADDRERRORS
);
713 opt
->lastopt
= opt
->srcrt
= skb_network_header_len(skb
);
714 skb
->transport_header
+= (hdr
->hdrlen
+ 1) << 3;
715 opt
->dst0
= opt
->dst1
;
717 opt
->nhoff
= (&hdr
->nexthdr
) - skb_network_header(skb
);
722 #if IS_ENABLED(CONFIG_IPV6_MIP6)
723 case IPV6_SRCRT_TYPE_2
:
724 if (accept_source_route
< 0)
726 /* Silently discard invalid RTH type 2 */
727 if (hdr
->hdrlen
!= 2 || hdr
->segments_left
!= 1) {
728 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
739 * This is the routing header forwarding algorithm from
743 n
= hdr
->hdrlen
>> 1;
745 if (hdr
->segments_left
> n
) {
746 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
747 icmpv6_param_prob(skb
, ICMPV6_HDR_FIELD
,
748 ((&hdr
->segments_left
) -
749 skb_network_header(skb
)));
753 /* We are about to mangle packet header. Be careful!
754 Do not damage packets queued somewhere.
756 if (skb_cloned(skb
)) {
757 /* the copy is a forwarded packet */
758 if (pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
)) {
759 __IP6_INC_STATS(net
, ip6_dst_idev(skb_dst(skb
)),
760 IPSTATS_MIB_OUTDISCARDS
);
764 hdr
= (struct ipv6_rt_hdr
*)skb_transport_header(skb
);
767 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
768 skb
->ip_summed
= CHECKSUM_NONE
;
770 i
= n
- --hdr
->segments_left
;
772 rthdr
= (struct rt0_hdr
*) hdr
;
777 #if IS_ENABLED(CONFIG_IPV6_MIP6)
778 case IPV6_SRCRT_TYPE_2
:
779 if (xfrm6_input_addr(skb
, (xfrm_address_t
*)addr
,
780 (xfrm_address_t
*)&ipv6_hdr(skb
)->saddr
,
781 IPPROTO_ROUTING
) < 0) {
782 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INADDRERRORS
);
786 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb
)->dev
), addr
)) {
787 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INADDRERRORS
);
797 if (ipv6_addr_is_multicast(addr
)) {
798 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INADDRERRORS
);
803 swap(*addr
, ipv6_hdr(skb
)->daddr
);
805 ip6_route_input(skb
);
806 if (skb_dst(skb
)->error
) {
807 skb_push(skb
, -skb_network_offset(skb
));
812 if (skb_dst(skb
)->dev
->flags
&IFF_LOOPBACK
) {
813 if (ipv6_hdr(skb
)->hop_limit
<= 1) {
814 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
815 icmpv6_send(skb
, ICMPV6_TIME_EXCEED
, ICMPV6_EXC_HOPLIMIT
,
820 ipv6_hdr(skb
)->hop_limit
--;
824 skb_push(skb
, -skb_network_offset(skb
));
829 __IP6_INC_STATS(net
, idev
, IPSTATS_MIB_INHDRERRORS
);
830 icmpv6_param_prob(skb
, ICMPV6_HDR_FIELD
,
831 (&hdr
->type
) - skb_network_header(skb
));
835 static const struct inet6_protocol rthdr_protocol
= {
836 .handler
= ipv6_rthdr_rcv
,
837 .flags
= INET6_PROTO_NOPOLICY
,
840 static const struct inet6_protocol destopt_protocol
= {
841 .handler
= ipv6_destopt_rcv
,
842 .flags
= INET6_PROTO_NOPOLICY
,
845 static const struct inet6_protocol nodata_protocol
= {
846 .handler
= dst_discard
,
847 .flags
= INET6_PROTO_NOPOLICY
,
850 int __init
ipv6_exthdrs_init(void)
854 ret
= inet6_add_protocol(&rthdr_protocol
, IPPROTO_ROUTING
);
858 ret
= inet6_add_protocol(&destopt_protocol
, IPPROTO_DSTOPTS
);
862 ret
= inet6_add_protocol(&nodata_protocol
, IPPROTO_NONE
);
869 inet6_del_protocol(&destopt_protocol
, IPPROTO_DSTOPTS
);
871 inet6_del_protocol(&rthdr_protocol
, IPPROTO_ROUTING
);
875 void ipv6_exthdrs_exit(void)
877 inet6_del_protocol(&nodata_protocol
, IPPROTO_NONE
);
878 inet6_del_protocol(&destopt_protocol
, IPPROTO_DSTOPTS
);
879 inet6_del_protocol(&rthdr_protocol
, IPPROTO_ROUTING
);
882 /**********************************
884 **********************************/
886 /* Router Alert as of RFC 2711 */
888 static bool ipv6_hop_ra(struct sk_buff
*skb
, int optoff
)
890 const unsigned char *nh
= skb_network_header(skb
);
892 if (nh
[optoff
+ 1] == 2) {
893 IP6CB(skb
)->flags
|= IP6SKB_ROUTERALERT
;
894 memcpy(&IP6CB(skb
)->ra
, nh
+ optoff
+ 2, sizeof(IP6CB(skb
)->ra
));
897 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
899 kfree_skb_reason(skb
, SKB_DROP_REASON_IP_INHDR
);
905 static bool ipv6_hop_ioam(struct sk_buff
*skb
, int optoff
)
907 struct ioam6_trace_hdr
*trace
;
908 struct ioam6_namespace
*ns
;
909 struct ioam6_hdr
*hdr
;
911 /* Bad alignment (must be 4n-aligned) */
915 /* Ignore if IOAM is not enabled on ingress */
916 if (!READ_ONCE(__in6_dev_get(skb
->dev
)->cnf
.ioam6_enabled
))
919 /* Truncated Option header */
920 hdr
= (struct ioam6_hdr
*)(skb_network_header(skb
) + optoff
);
921 if (hdr
->opt_len
< 2)
925 case IOAM6_TYPE_PREALLOC
:
926 /* Truncated Pre-allocated Trace header */
927 if (hdr
->opt_len
< 2 + sizeof(*trace
))
930 /* Malformed Pre-allocated Trace header */
931 trace
= (struct ioam6_trace_hdr
*)((u8
*)hdr
+ sizeof(*hdr
));
932 if (hdr
->opt_len
< 2 + sizeof(*trace
) + trace
->remlen
* 4)
935 /* Ignore if the IOAM namespace is unknown */
936 ns
= ioam6_namespace(dev_net(skb
->dev
), trace
->namespace_id
);
940 if (!skb_valid_dst(skb
))
941 ip6_route_input(skb
);
943 /* About to mangle packet header */
944 if (skb_ensure_writable(skb
, optoff
+ 2 + hdr
->opt_len
))
947 /* Trace pointer may have changed */
948 trace
= (struct ioam6_trace_hdr
*)(skb_network_header(skb
)
949 + optoff
+ sizeof(*hdr
));
951 ioam6_fill_trace_data(skb
, ns
, trace
, true);
953 ioam6_event(IOAM6_EVENT_TRACE
, dev_net(skb
->dev
),
954 GFP_ATOMIC
, (void *)trace
, hdr
->opt_len
- 2);
964 kfree_skb_reason(skb
, SKB_DROP_REASON_IP_INHDR
);
970 static bool ipv6_hop_jumbo(struct sk_buff
*skb
, int optoff
)
972 const unsigned char *nh
= skb_network_header(skb
);
976 if (nh
[optoff
+ 1] != 4 || (optoff
& 3) != 2) {
977 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
979 SKB_DR_SET(reason
, IP_INHDR
);
983 pkt_len
= ntohl(*(__be32
*)(nh
+ optoff
+ 2));
984 if (pkt_len
<= IPV6_MAXPLEN
) {
985 icmpv6_param_prob_reason(skb
, ICMPV6_HDR_FIELD
, optoff
+ 2,
986 SKB_DROP_REASON_IP_INHDR
);
989 if (ipv6_hdr(skb
)->payload_len
) {
990 icmpv6_param_prob_reason(skb
, ICMPV6_HDR_FIELD
, optoff
,
991 SKB_DROP_REASON_IP_INHDR
);
995 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
)) {
996 SKB_DR_SET(reason
, PKT_TOO_SMALL
);
1000 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
1003 IP6CB(skb
)->flags
|= IP6SKB_JUMBOGRAM
;
1007 kfree_skb_reason(skb
, reason
);
1011 /* CALIPSO RFC 5570 */
1013 static bool ipv6_hop_calipso(struct sk_buff
*skb
, int optoff
)
1015 const unsigned char *nh
= skb_network_header(skb
);
1017 if (nh
[optoff
+ 1] < 8)
1020 if (nh
[optoff
+ 6] * 4 + 8 > nh
[optoff
+ 1])
1023 if (!calipso_validate(skb
, nh
+ optoff
))
1029 kfree_skb_reason(skb
, SKB_DROP_REASON_IP_INHDR
);
1033 int ipv6_parse_hopopts(struct sk_buff
*skb
)
1035 struct inet6_skb_parm
*opt
= IP6CB(skb
);
1036 struct net
*net
= dev_net(skb
->dev
);
1040 * skb_network_header(skb) is equal to skb->data, and
1041 * skb_network_header_len(skb) is always equal to
1042 * sizeof(struct ipv6hdr) by definition of
1043 * hop-by-hop options.
1045 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
) + 8) ||
1046 !pskb_may_pull(skb
, (sizeof(struct ipv6hdr
) +
1047 ((skb_transport_header(skb
)[1] + 1) << 3)))) {
1053 extlen
= (skb_transport_header(skb
)[1] + 1) << 3;
1054 if (extlen
> net
->ipv6
.sysctl
.max_hbh_opts_len
)
1057 opt
->flags
|= IP6SKB_HOPBYHOP
;
1058 if (ip6_parse_tlv(true, skb
, net
->ipv6
.sysctl
.max_hbh_opts_cnt
)) {
1059 skb
->transport_header
+= extlen
;
1061 opt
->nhoff
= sizeof(struct ipv6hdr
);
1068 * Creating outbound headers.
1070 * "build" functions work when skb is filled from head to tail (datagram)
1071 * "push" functions work when headers are added from tail to head (tcp)
1073 * In both cases we assume, that caller reserved enough room
1077 static void ipv6_push_rthdr0(struct sk_buff
*skb
, u8
*proto
,
1078 struct ipv6_rt_hdr
*opt
,
1079 struct in6_addr
**addr_p
, struct in6_addr
*saddr
)
1081 struct rt0_hdr
*phdr
, *ihdr
;
1084 ihdr
= (struct rt0_hdr
*) opt
;
1086 phdr
= skb_push(skb
, (ihdr
->rt_hdr
.hdrlen
+ 1) << 3);
1087 memcpy(phdr
, ihdr
, sizeof(struct rt0_hdr
));
1089 hops
= ihdr
->rt_hdr
.hdrlen
>> 1;
1092 memcpy(phdr
->addr
, ihdr
->addr
+ 1,
1093 (hops
- 1) * sizeof(struct in6_addr
));
1095 phdr
->addr
[hops
- 1] = **addr_p
;
1096 *addr_p
= ihdr
->addr
;
1098 phdr
->rt_hdr
.nexthdr
= *proto
;
1099 *proto
= NEXTHDR_ROUTING
;
1102 static void ipv6_push_rthdr4(struct sk_buff
*skb
, u8
*proto
,
1103 struct ipv6_rt_hdr
*opt
,
1104 struct in6_addr
**addr_p
, struct in6_addr
*saddr
)
1106 struct ipv6_sr_hdr
*sr_phdr
, *sr_ihdr
;
1109 sr_ihdr
= (struct ipv6_sr_hdr
*)opt
;
1110 plen
= (sr_ihdr
->hdrlen
+ 1) << 3;
1112 sr_phdr
= skb_push(skb
, plen
);
1113 memcpy(sr_phdr
, sr_ihdr
, sizeof(struct ipv6_sr_hdr
));
1115 hops
= sr_ihdr
->first_segment
+ 1;
1116 memcpy(sr_phdr
->segments
+ 1, sr_ihdr
->segments
+ 1,
1117 (hops
- 1) * sizeof(struct in6_addr
));
1119 sr_phdr
->segments
[0] = **addr_p
;
1120 *addr_p
= &sr_ihdr
->segments
[sr_ihdr
->segments_left
];
1122 if (sr_ihdr
->hdrlen
> hops
* 2) {
1123 int tlvs_offset
, tlvs_length
;
1125 tlvs_offset
= (1 + hops
* 2) << 3;
1126 tlvs_length
= (sr_ihdr
->hdrlen
- hops
* 2) << 3;
1127 memcpy((char *)sr_phdr
+ tlvs_offset
,
1128 (char *)sr_ihdr
+ tlvs_offset
, tlvs_length
);
1131 #ifdef CONFIG_IPV6_SEG6_HMAC
1132 if (sr_has_hmac(sr_phdr
)) {
1133 struct net
*net
= NULL
;
1136 net
= dev_net(skb
->dev
);
1138 net
= sock_net(skb
->sk
);
1143 seg6_push_hmac(net
, saddr
, sr_phdr
);
1147 sr_phdr
->nexthdr
= *proto
;
1148 *proto
= NEXTHDR_ROUTING
;
1151 static void ipv6_push_rthdr(struct sk_buff
*skb
, u8
*proto
,
1152 struct ipv6_rt_hdr
*opt
,
1153 struct in6_addr
**addr_p
, struct in6_addr
*saddr
)
1155 switch (opt
->type
) {
1156 case IPV6_SRCRT_TYPE_0
:
1157 case IPV6_SRCRT_STRICT
:
1158 case IPV6_SRCRT_TYPE_2
:
1159 ipv6_push_rthdr0(skb
, proto
, opt
, addr_p
, saddr
);
1161 case IPV6_SRCRT_TYPE_4
:
1162 ipv6_push_rthdr4(skb
, proto
, opt
, addr_p
, saddr
);
1169 static void ipv6_push_exthdr(struct sk_buff
*skb
, u8
*proto
, u8 type
, struct ipv6_opt_hdr
*opt
)
1171 struct ipv6_opt_hdr
*h
= skb_push(skb
, ipv6_optlen(opt
));
1173 memcpy(h
, opt
, ipv6_optlen(opt
));
1174 h
->nexthdr
= *proto
;
1178 void ipv6_push_nfrag_opts(struct sk_buff
*skb
, struct ipv6_txoptions
*opt
,
1180 struct in6_addr
**daddr
, struct in6_addr
*saddr
)
1183 ipv6_push_rthdr(skb
, proto
, opt
->srcrt
, daddr
, saddr
);
1185 * IPV6_RTHDRDSTOPTS is ignored
1186 * unless IPV6_RTHDR is set (RFC3542).
1189 ipv6_push_exthdr(skb
, proto
, NEXTHDR_DEST
, opt
->dst0opt
);
1192 ipv6_push_exthdr(skb
, proto
, NEXTHDR_HOP
, opt
->hopopt
);
1195 void ipv6_push_frag_opts(struct sk_buff
*skb
, struct ipv6_txoptions
*opt
, u8
*proto
)
1198 ipv6_push_exthdr(skb
, proto
, NEXTHDR_DEST
, opt
->dst1opt
);
1200 EXPORT_SYMBOL(ipv6_push_frag_opts
);
1202 struct ipv6_txoptions
*
1203 ipv6_dup_options(struct sock
*sk
, struct ipv6_txoptions
*opt
)
1205 struct ipv6_txoptions
*opt2
;
1207 opt2
= sock_kmalloc(sk
, opt
->tot_len
, GFP_ATOMIC
);
1209 long dif
= (char *)opt2
- (char *)opt
;
1210 memcpy(opt2
, opt
, opt
->tot_len
);
1212 *((char **)&opt2
->hopopt
) += dif
;
1214 *((char **)&opt2
->dst0opt
) += dif
;
1216 *((char **)&opt2
->dst1opt
) += dif
;
1218 *((char **)&opt2
->srcrt
) += dif
;
1219 refcount_set(&opt2
->refcnt
, 1);
1223 EXPORT_SYMBOL_GPL(ipv6_dup_options
);
1225 static void ipv6_renew_option(int renewtype
,
1226 struct ipv6_opt_hdr
**dest
,
1227 struct ipv6_opt_hdr
*old
,
1228 struct ipv6_opt_hdr
*new,
1229 int newtype
, char **p
)
1231 struct ipv6_opt_hdr
*src
;
1233 src
= (renewtype
== newtype
? new : old
);
1237 memcpy(*p
, src
, ipv6_optlen(src
));
1238 *dest
= (struct ipv6_opt_hdr
*)*p
;
1239 *p
+= CMSG_ALIGN(ipv6_optlen(*dest
));
1243 * ipv6_renew_options - replace a specific ext hdr with a new one.
1245 * @sk: sock from which to allocate memory
1246 * @opt: original options
1247 * @newtype: option type to replace in @opt
1248 * @newopt: new option of type @newtype to replace (user-mem)
1250 * Returns a new set of options which is a copy of @opt with the
1251 * option type @newtype replaced with @newopt.
1253 * @opt may be NULL, in which case a new set of options is returned
1254 * containing just @newopt.
1256 * @newopt may be NULL, in which case the specified option type is
1257 * not copied into the new set of options.
1259 * The new set of options is allocated from the socket option memory
1262 struct ipv6_txoptions
*
1263 ipv6_renew_options(struct sock
*sk
, struct ipv6_txoptions
*opt
,
1264 int newtype
, struct ipv6_opt_hdr
*newopt
)
1268 struct ipv6_txoptions
*opt2
;
1271 if (newtype
!= IPV6_HOPOPTS
&& opt
->hopopt
)
1272 tot_len
+= CMSG_ALIGN(ipv6_optlen(opt
->hopopt
));
1273 if (newtype
!= IPV6_RTHDRDSTOPTS
&& opt
->dst0opt
)
1274 tot_len
+= CMSG_ALIGN(ipv6_optlen(opt
->dst0opt
));
1275 if (newtype
!= IPV6_RTHDR
&& opt
->srcrt
)
1276 tot_len
+= CMSG_ALIGN(ipv6_optlen(opt
->srcrt
));
1277 if (newtype
!= IPV6_DSTOPTS
&& opt
->dst1opt
)
1278 tot_len
+= CMSG_ALIGN(ipv6_optlen(opt
->dst1opt
));
1282 tot_len
+= CMSG_ALIGN(ipv6_optlen(newopt
));
1287 tot_len
+= sizeof(*opt2
);
1288 opt2
= sock_kmalloc(sk
, tot_len
, GFP_ATOMIC
);
1290 return ERR_PTR(-ENOBUFS
);
1292 memset(opt2
, 0, tot_len
);
1293 refcount_set(&opt2
->refcnt
, 1);
1294 opt2
->tot_len
= tot_len
;
1295 p
= (char *)(opt2
+ 1);
1297 ipv6_renew_option(IPV6_HOPOPTS
, &opt2
->hopopt
,
1298 (opt
? opt
->hopopt
: NULL
),
1299 newopt
, newtype
, &p
);
1300 ipv6_renew_option(IPV6_RTHDRDSTOPTS
, &opt2
->dst0opt
,
1301 (opt
? opt
->dst0opt
: NULL
),
1302 newopt
, newtype
, &p
);
1303 ipv6_renew_option(IPV6_RTHDR
,
1304 (struct ipv6_opt_hdr
**)&opt2
->srcrt
,
1305 (opt
? (struct ipv6_opt_hdr
*)opt
->srcrt
: NULL
),
1306 newopt
, newtype
, &p
);
1307 ipv6_renew_option(IPV6_DSTOPTS
, &opt2
->dst1opt
,
1308 (opt
? opt
->dst1opt
: NULL
),
1309 newopt
, newtype
, &p
);
1311 opt2
->opt_nflen
= (opt2
->hopopt
? ipv6_optlen(opt2
->hopopt
) : 0) +
1312 (opt2
->dst0opt
? ipv6_optlen(opt2
->dst0opt
) : 0) +
1313 (opt2
->srcrt
? ipv6_optlen(opt2
->srcrt
) : 0);
1314 opt2
->opt_flen
= (opt2
->dst1opt
? ipv6_optlen(opt2
->dst1opt
) : 0);
1319 struct ipv6_txoptions
*__ipv6_fixup_options(struct ipv6_txoptions
*opt_space
,
1320 struct ipv6_txoptions
*opt
)
1323 * ignore the dest before srcrt unless srcrt is being included.
1326 if (opt
->dst0opt
&& !opt
->srcrt
) {
1327 if (opt_space
!= opt
) {
1328 memcpy(opt_space
, opt
, sizeof(*opt_space
));
1331 opt
->opt_nflen
-= ipv6_optlen(opt
->dst0opt
);
1332 opt
->dst0opt
= NULL
;
1337 EXPORT_SYMBOL_GPL(__ipv6_fixup_options
);
1340 * fl6_update_dst - update flowi destination address with info given
1341 * by srcrt option, if any.
1343 * @fl6: flowi6 for which daddr is to be updated
1344 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1345 * @orig: copy of original daddr address if modified
1347 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1348 * and initial value of fl6->daddr set in orig
1350 struct in6_addr
*fl6_update_dst(struct flowi6
*fl6
,
1351 const struct ipv6_txoptions
*opt
,
1352 struct in6_addr
*orig
)
1354 if (!opt
|| !opt
->srcrt
)
1359 switch (opt
->srcrt
->type
) {
1360 case IPV6_SRCRT_TYPE_0
:
1361 case IPV6_SRCRT_STRICT
:
1362 case IPV6_SRCRT_TYPE_2
:
1363 fl6
->daddr
= *((struct rt0_hdr
*)opt
->srcrt
)->addr
;
1365 case IPV6_SRCRT_TYPE_4
:
1367 struct ipv6_sr_hdr
*srh
= (struct ipv6_sr_hdr
*)opt
->srcrt
;
1369 fl6
->daddr
= srh
->segments
[srh
->segments_left
];
1378 EXPORT_SYMBOL_GPL(fl6_update_dst
);