Merge tag 'iommu-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
[linux/fpc-iii.git] / net / ipv6 / exthdrs.c
blob6126f8bf94b3992f75648adbcf37c7cdbd2425be
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Extension Header handling for IPv6
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Andi Kleen <ak@muc.de>
9 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
12 /* Changes:
13 * yoshfuji : ensure not to overrun while parsing
14 * tlv options.
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>
31 #include <net/dst.h>
32 #include <net/sock.h>
33 #include <net/snmp.h>
35 #include <net/ipv6.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)
44 #include <net/xfrm.h>
45 #endif
46 #include <linux/seg6.h>
47 #include <net/seg6.h>
48 #ifdef CONFIG_IPV6_SEG6_HMAC
49 #include <net/seg6_hmac.h>
50 #endif
51 #include <net/rpl.h>
53 #include <linux/uaccess.h>
56 * Parsing tlv encoded headers.
58 * Parsing function "func" returns true, if parsing succeed
59 * and false, if it failed.
60 * It MUST NOT touch skb->h.
63 struct tlvtype_proc {
64 int type;
65 bool (*func)(struct sk_buff *skb, int offset);
68 /*********************
69 Generic functions
70 *********************/
72 /* An unknown option is detected, decide what to do */
74 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff,
75 bool disallow_unknowns)
77 if (disallow_unknowns) {
78 /* If unknown TLVs are disallowed by configuration
79 * then always silently drop packet. Note this also
80 * means no ICMP parameter problem is sent which
81 * could be a good property to mitigate a reflection DOS
82 * attack.
85 goto drop;
88 switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
89 case 0: /* ignore */
90 return true;
92 case 1: /* drop packet */
93 break;
95 case 3: /* Send ICMP if not a multicast address and drop packet */
96 /* Actually, it is redundant check. icmp_send
97 will recheck in any case.
99 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
100 break;
101 fallthrough;
102 case 2: /* send ICMP PARM PROB regardless and drop packet */
103 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
104 return false;
107 drop:
108 kfree_skb(skb);
109 return false;
112 /* Parse tlv encoded option header (hop-by-hop or destination) */
114 static bool ip6_parse_tlv(const struct tlvtype_proc *procs,
115 struct sk_buff *skb,
116 int max_count)
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 const struct tlvtype_proc *curr;
122 bool disallow_unknowns = false;
123 int tlv_count = 0;
124 int padlen = 0;
126 if (unlikely(max_count < 0)) {
127 disallow_unknowns = true;
128 max_count = -max_count;
131 if (skb_transport_offset(skb) + len > skb_headlen(skb))
132 goto bad;
134 off += 2;
135 len -= 2;
137 while (len > 0) {
138 int optlen = nh[off + 1] + 2;
139 int i;
141 switch (nh[off]) {
142 case IPV6_TLV_PAD1:
143 optlen = 1;
144 padlen++;
145 if (padlen > 7)
146 goto bad;
147 break;
149 case IPV6_TLV_PADN:
150 /* RFC 2460 states that the purpose of PadN is
151 * to align the containing header to multiples
152 * of 8. 7 is therefore the highest valid value.
153 * See also RFC 4942, Section 2.1.9.5.
155 padlen += optlen;
156 if (padlen > 7)
157 goto bad;
158 /* RFC 4942 recommends receiving hosts to
159 * actively check PadN payload to contain
160 * only zeroes.
162 for (i = 2; i < optlen; i++) {
163 if (nh[off + i] != 0)
164 goto bad;
166 break;
168 default: /* Other TLV code so scan list */
169 if (optlen > len)
170 goto bad;
172 tlv_count++;
173 if (tlv_count > max_count)
174 goto bad;
176 for (curr = procs; curr->type >= 0; curr++) {
177 if (curr->type == nh[off]) {
178 /* type specific length/alignment
179 checks will be performed in the
180 func(). */
181 if (curr->func(skb, off) == false)
182 return false;
183 break;
186 if (curr->type < 0 &&
187 !ip6_tlvopt_unknown(skb, off, disallow_unknowns))
188 return false;
190 padlen = 0;
191 break;
193 off += optlen;
194 len -= optlen;
197 if (len == 0)
198 return true;
199 bad:
200 kfree_skb(skb);
201 return false;
204 /*****************************
205 Destination options header.
206 *****************************/
208 #if IS_ENABLED(CONFIG_IPV6_MIP6)
209 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
211 struct ipv6_destopt_hao *hao;
212 struct inet6_skb_parm *opt = IP6CB(skb);
213 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
214 int ret;
216 if (opt->dsthao) {
217 net_dbg_ratelimited("hao duplicated\n");
218 goto discard;
220 opt->dsthao = opt->dst1;
221 opt->dst1 = 0;
223 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
225 if (hao->length != 16) {
226 net_dbg_ratelimited("hao invalid option length = %d\n",
227 hao->length);
228 goto discard;
231 if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
232 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
233 &hao->addr);
234 goto discard;
237 ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
238 (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
239 if (unlikely(ret < 0))
240 goto discard;
242 if (skb_cloned(skb)) {
243 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
244 goto discard;
246 /* update all variable using below by copied skbuff */
247 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
248 optoff);
249 ipv6h = ipv6_hdr(skb);
252 if (skb->ip_summed == CHECKSUM_COMPLETE)
253 skb->ip_summed = CHECKSUM_NONE;
255 swap(ipv6h->saddr, hao->addr);
257 if (skb->tstamp == 0)
258 __net_timestamp(skb);
260 return true;
262 discard:
263 kfree_skb(skb);
264 return false;
266 #endif
268 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
269 #if IS_ENABLED(CONFIG_IPV6_MIP6)
271 .type = IPV6_TLV_HAO,
272 .func = ipv6_dest_hao,
274 #endif
275 {-1, NULL}
278 static int ipv6_destopt_rcv(struct sk_buff *skb)
280 struct inet6_dev *idev = __in6_dev_get(skb->dev);
281 struct inet6_skb_parm *opt = IP6CB(skb);
282 #if IS_ENABLED(CONFIG_IPV6_MIP6)
283 __u16 dstbuf;
284 #endif
285 struct dst_entry *dst = skb_dst(skb);
286 struct net *net = dev_net(skb->dev);
287 int extlen;
289 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
290 !pskb_may_pull(skb, (skb_transport_offset(skb) +
291 ((skb_transport_header(skb)[1] + 1) << 3)))) {
292 __IP6_INC_STATS(dev_net(dst->dev), idev,
293 IPSTATS_MIB_INHDRERRORS);
294 fail_and_free:
295 kfree_skb(skb);
296 return -1;
299 extlen = (skb_transport_header(skb)[1] + 1) << 3;
300 if (extlen > net->ipv6.sysctl.max_dst_opts_len)
301 goto fail_and_free;
303 opt->lastopt = opt->dst1 = skb_network_header_len(skb);
304 #if IS_ENABLED(CONFIG_IPV6_MIP6)
305 dstbuf = opt->dst1;
306 #endif
308 if (ip6_parse_tlv(tlvprocdestopt_lst, skb,
309 init_net.ipv6.sysctl.max_dst_opts_cnt)) {
310 skb->transport_header += extlen;
311 opt = IP6CB(skb);
312 #if IS_ENABLED(CONFIG_IPV6_MIP6)
313 opt->nhoff = dstbuf;
314 #else
315 opt->nhoff = opt->dst1;
316 #endif
317 return 1;
320 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
321 return -1;
324 static void seg6_update_csum(struct sk_buff *skb)
326 struct ipv6_sr_hdr *hdr;
327 struct in6_addr *addr;
328 __be32 from, to;
330 /* srh is at transport offset and seg_left is already decremented
331 * but daddr is not yet updated with next segment
334 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
335 addr = hdr->segments + hdr->segments_left;
337 hdr->segments_left++;
338 from = *(__be32 *)hdr;
340 hdr->segments_left--;
341 to = *(__be32 *)hdr;
343 /* update skb csum with diff resulting from seg_left decrement */
345 update_csum_diff4(skb, from, to);
347 /* compute csum diff between current and next segment and update */
349 update_csum_diff16(skb, (__be32 *)(&ipv6_hdr(skb)->daddr),
350 (__be32 *)addr);
353 static int ipv6_srh_rcv(struct sk_buff *skb)
355 struct inet6_skb_parm *opt = IP6CB(skb);
356 struct net *net = dev_net(skb->dev);
357 struct ipv6_sr_hdr *hdr;
358 struct inet6_dev *idev;
359 struct in6_addr *addr;
360 int accept_seg6;
362 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
364 idev = __in6_dev_get(skb->dev);
366 accept_seg6 = net->ipv6.devconf_all->seg6_enabled;
367 if (accept_seg6 > idev->cnf.seg6_enabled)
368 accept_seg6 = idev->cnf.seg6_enabled;
370 if (!accept_seg6) {
371 kfree_skb(skb);
372 return -1;
375 #ifdef CONFIG_IPV6_SEG6_HMAC
376 if (!seg6_hmac_validate_skb(skb)) {
377 kfree_skb(skb);
378 return -1;
380 #endif
382 looped_back:
383 if (hdr->segments_left == 0) {
384 if (hdr->nexthdr == NEXTHDR_IPV6) {
385 int offset = (hdr->hdrlen + 1) << 3;
387 skb_postpull_rcsum(skb, skb_network_header(skb),
388 skb_network_header_len(skb));
390 if (!pskb_pull(skb, offset)) {
391 kfree_skb(skb);
392 return -1;
394 skb_postpull_rcsum(skb, skb_transport_header(skb),
395 offset);
397 skb_reset_network_header(skb);
398 skb_reset_transport_header(skb);
399 skb->encapsulation = 0;
401 __skb_tunnel_rx(skb, skb->dev, net);
403 netif_rx(skb);
404 return -1;
407 opt->srcrt = skb_network_header_len(skb);
408 opt->lastopt = opt->srcrt;
409 skb->transport_header += (hdr->hdrlen + 1) << 3;
410 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
412 return 1;
415 if (hdr->segments_left >= (hdr->hdrlen >> 1)) {
416 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
417 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
418 ((&hdr->segments_left) -
419 skb_network_header(skb)));
420 return -1;
423 if (skb_cloned(skb)) {
424 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
425 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
426 IPSTATS_MIB_OUTDISCARDS);
427 kfree_skb(skb);
428 return -1;
432 hdr = (struct ipv6_sr_hdr *)skb_transport_header(skb);
434 hdr->segments_left--;
435 addr = hdr->segments + hdr->segments_left;
437 skb_push(skb, sizeof(struct ipv6hdr));
439 if (skb->ip_summed == CHECKSUM_COMPLETE)
440 seg6_update_csum(skb);
442 ipv6_hdr(skb)->daddr = *addr;
444 skb_dst_drop(skb);
446 ip6_route_input(skb);
448 if (skb_dst(skb)->error) {
449 dst_input(skb);
450 return -1;
453 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
454 if (ipv6_hdr(skb)->hop_limit <= 1) {
455 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
456 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
457 ICMPV6_EXC_HOPLIMIT, 0);
458 kfree_skb(skb);
459 return -1;
461 ipv6_hdr(skb)->hop_limit--;
463 skb_pull(skb, sizeof(struct ipv6hdr));
464 goto looped_back;
467 dst_input(skb);
469 return -1;
472 static int ipv6_rpl_srh_rcv(struct sk_buff *skb)
474 struct ipv6_rpl_sr_hdr *hdr, *ohdr, *chdr;
475 struct inet6_skb_parm *opt = IP6CB(skb);
476 struct net *net = dev_net(skb->dev);
477 struct inet6_dev *idev;
478 struct ipv6hdr *oldhdr;
479 struct in6_addr addr;
480 unsigned char *buf;
481 int accept_rpl_seg;
482 int i, err;
483 u64 n = 0;
484 u32 r;
486 idev = __in6_dev_get(skb->dev);
488 accept_rpl_seg = net->ipv6.devconf_all->rpl_seg_enabled;
489 if (accept_rpl_seg > idev->cnf.rpl_seg_enabled)
490 accept_rpl_seg = idev->cnf.rpl_seg_enabled;
492 if (!accept_rpl_seg) {
493 kfree_skb(skb);
494 return -1;
497 looped_back:
498 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
500 if (hdr->segments_left == 0) {
501 if (hdr->nexthdr == NEXTHDR_IPV6) {
502 int offset = (hdr->hdrlen + 1) << 3;
504 skb_postpull_rcsum(skb, skb_network_header(skb),
505 skb_network_header_len(skb));
507 if (!pskb_pull(skb, offset)) {
508 kfree_skb(skb);
509 return -1;
511 skb_postpull_rcsum(skb, skb_transport_header(skb),
512 offset);
514 skb_reset_network_header(skb);
515 skb_reset_transport_header(skb);
516 skb->encapsulation = 0;
518 __skb_tunnel_rx(skb, skb->dev, net);
520 netif_rx(skb);
521 return -1;
524 opt->srcrt = skb_network_header_len(skb);
525 opt->lastopt = opt->srcrt;
526 skb->transport_header += (hdr->hdrlen + 1) << 3;
527 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
529 return 1;
532 if (!pskb_may_pull(skb, sizeof(*hdr))) {
533 kfree_skb(skb);
534 return -1;
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
541 * higher than that.
543 if (r || (n + 1) > 255) {
544 kfree_skb(skb);
545 return -1;
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)));
553 return -1;
556 if (skb_cloned(skb)) {
557 if (pskb_expand_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE, 0,
558 GFP_ATOMIC)) {
559 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
560 IPSTATS_MIB_OUTDISCARDS);
561 kfree_skb(skb);
562 return -1;
564 } else {
565 err = skb_cow_head(skb, IPV6_RPL_SRH_WORST_SWAP_SIZE);
566 if (unlikely(err)) {
567 kfree_skb(skb);
568 return -1;
572 hdr = (struct ipv6_rpl_sr_hdr *)skb_transport_header(skb);
574 if (!pskb_may_pull(skb, ipv6_rpl_srh_size(n, hdr->cmpri,
575 hdr->cmpre))) {
576 kfree_skb(skb);
577 return -1;
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)) {
585 kfree_skb(skb);
586 return -1;
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)) {
595 kfree_skb(skb);
596 kfree(buf);
597 return -1;
600 err = ipv6_chk_rpl_srh_loop(net, ohdr->rpl_segaddr, n + 1);
601 if (err) {
602 icmpv6_send(skb, ICMPV6_PARAMPROB, 0, 0);
603 kfree_skb(skb);
604 kfree(buf);
605 return -1;
608 addr = ipv6_hdr(skb)->daddr;
609 ipv6_hdr(skb)->daddr = ohdr->rpl_segaddr[i];
610 ohdr->rpl_segaddr[i] = addr;
612 ipv6_rpl_srh_compress(chdr, ohdr, &ipv6_hdr(skb)->daddr, n);
614 oldhdr = ipv6_hdr(skb);
616 skb_pull(skb, ((hdr->hdrlen + 1) << 3));
617 skb_postpull_rcsum(skb, oldhdr,
618 sizeof(struct ipv6hdr) + ((hdr->hdrlen + 1) << 3));
619 skb_push(skb, ((chdr->hdrlen + 1) << 3) + sizeof(struct ipv6hdr));
620 skb_reset_network_header(skb);
621 skb_mac_header_rebuild(skb);
622 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
624 memmove(ipv6_hdr(skb), oldhdr, sizeof(struct ipv6hdr));
625 memcpy(skb_transport_header(skb), chdr, (chdr->hdrlen + 1) << 3);
627 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
628 skb_postpush_rcsum(skb, ipv6_hdr(skb),
629 sizeof(struct ipv6hdr) + ((chdr->hdrlen + 1) << 3));
631 kfree(buf);
633 skb_dst_drop(skb);
635 ip6_route_input(skb);
637 if (skb_dst(skb)->error) {
638 dst_input(skb);
639 return -1;
642 if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
643 if (ipv6_hdr(skb)->hop_limit <= 1) {
644 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
645 icmpv6_send(skb, ICMPV6_TIME_EXCEED,
646 ICMPV6_EXC_HOPLIMIT, 0);
647 kfree_skb(skb);
648 return -1;
650 ipv6_hdr(skb)->hop_limit--;
652 skb_pull(skb, sizeof(struct ipv6hdr));
653 goto looped_back;
656 dst_input(skb);
658 return -1;
661 /********************************
662 Routing header.
663 ********************************/
665 /* called with rcu_read_lock() */
666 static int ipv6_rthdr_rcv(struct sk_buff *skb)
668 struct inet6_dev *idev = __in6_dev_get(skb->dev);
669 struct inet6_skb_parm *opt = IP6CB(skb);
670 struct in6_addr *addr = NULL;
671 struct in6_addr daddr;
672 int n, i;
673 struct ipv6_rt_hdr *hdr;
674 struct rt0_hdr *rthdr;
675 struct net *net = dev_net(skb->dev);
676 int accept_source_route = net->ipv6.devconf_all->accept_source_route;
678 idev = __in6_dev_get(skb->dev);
679 if (idev && accept_source_route > idev->cnf.accept_source_route)
680 accept_source_route = idev->cnf.accept_source_route;
682 if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
683 !pskb_may_pull(skb, (skb_transport_offset(skb) +
684 ((skb_transport_header(skb)[1] + 1) << 3)))) {
685 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
686 kfree_skb(skb);
687 return -1;
690 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
692 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
693 skb->pkt_type != PACKET_HOST) {
694 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
695 kfree_skb(skb);
696 return -1;
699 switch (hdr->type) {
700 case IPV6_SRCRT_TYPE_4:
701 /* segment routing */
702 return ipv6_srh_rcv(skb);
703 case IPV6_SRCRT_TYPE_3:
704 /* rpl segment routing */
705 return ipv6_rpl_srh_rcv(skb);
706 default:
707 break;
710 looped_back:
711 if (hdr->segments_left == 0) {
712 switch (hdr->type) {
713 #if IS_ENABLED(CONFIG_IPV6_MIP6)
714 case IPV6_SRCRT_TYPE_2:
715 /* Silently discard type 2 header unless it was
716 * processed by own
718 if (!addr) {
719 __IP6_INC_STATS(net, idev,
720 IPSTATS_MIB_INADDRERRORS);
721 kfree_skb(skb);
722 return -1;
724 break;
725 #endif
726 default:
727 break;
730 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
731 skb->transport_header += (hdr->hdrlen + 1) << 3;
732 opt->dst0 = opt->dst1;
733 opt->dst1 = 0;
734 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
735 return 1;
738 switch (hdr->type) {
739 #if IS_ENABLED(CONFIG_IPV6_MIP6)
740 case IPV6_SRCRT_TYPE_2:
741 if (accept_source_route < 0)
742 goto unknown_rh;
743 /* Silently discard invalid RTH type 2 */
744 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
745 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
746 kfree_skb(skb);
747 return -1;
749 break;
750 #endif
751 default:
752 goto unknown_rh;
756 * This is the routing header forwarding algorithm from
757 * RFC 2460, page 16.
760 n = hdr->hdrlen >> 1;
762 if (hdr->segments_left > n) {
763 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
764 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
765 ((&hdr->segments_left) -
766 skb_network_header(skb)));
767 return -1;
770 /* We are about to mangle packet header. Be careful!
771 Do not damage packets queued somewhere.
773 if (skb_cloned(skb)) {
774 /* the copy is a forwarded packet */
775 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
776 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
777 IPSTATS_MIB_OUTDISCARDS);
778 kfree_skb(skb);
779 return -1;
781 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
784 if (skb->ip_summed == CHECKSUM_COMPLETE)
785 skb->ip_summed = CHECKSUM_NONE;
787 i = n - --hdr->segments_left;
789 rthdr = (struct rt0_hdr *) hdr;
790 addr = rthdr->addr;
791 addr += i - 1;
793 switch (hdr->type) {
794 #if IS_ENABLED(CONFIG_IPV6_MIP6)
795 case IPV6_SRCRT_TYPE_2:
796 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
797 (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
798 IPPROTO_ROUTING) < 0) {
799 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
800 kfree_skb(skb);
801 return -1;
803 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
804 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
805 kfree_skb(skb);
806 return -1;
808 break;
809 #endif
810 default:
811 break;
814 if (ipv6_addr_is_multicast(addr)) {
815 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
816 kfree_skb(skb);
817 return -1;
820 daddr = *addr;
821 *addr = ipv6_hdr(skb)->daddr;
822 ipv6_hdr(skb)->daddr = daddr;
824 skb_dst_drop(skb);
825 ip6_route_input(skb);
826 if (skb_dst(skb)->error) {
827 skb_push(skb, skb->data - skb_network_header(skb));
828 dst_input(skb);
829 return -1;
832 if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
833 if (ipv6_hdr(skb)->hop_limit <= 1) {
834 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
835 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
837 kfree_skb(skb);
838 return -1;
840 ipv6_hdr(skb)->hop_limit--;
841 goto looped_back;
844 skb_push(skb, skb->data - skb_network_header(skb));
845 dst_input(skb);
846 return -1;
848 unknown_rh:
849 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
850 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
851 (&hdr->type) - skb_network_header(skb));
852 return -1;
855 static const struct inet6_protocol rthdr_protocol = {
856 .handler = ipv6_rthdr_rcv,
857 .flags = INET6_PROTO_NOPOLICY,
860 static const struct inet6_protocol destopt_protocol = {
861 .handler = ipv6_destopt_rcv,
862 .flags = INET6_PROTO_NOPOLICY,
865 static const struct inet6_protocol nodata_protocol = {
866 .handler = dst_discard,
867 .flags = INET6_PROTO_NOPOLICY,
870 int __init ipv6_exthdrs_init(void)
872 int ret;
874 ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
875 if (ret)
876 goto out;
878 ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
879 if (ret)
880 goto out_rthdr;
882 ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
883 if (ret)
884 goto out_destopt;
886 out:
887 return ret;
888 out_destopt:
889 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
890 out_rthdr:
891 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
892 goto out;
895 void ipv6_exthdrs_exit(void)
897 inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
898 inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
899 inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
902 /**********************************
903 Hop-by-hop options.
904 **********************************/
907 * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
909 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
911 return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
914 /* Router Alert as of RFC 2711 */
916 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
918 const unsigned char *nh = skb_network_header(skb);
920 if (nh[optoff + 1] == 2) {
921 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
922 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
923 return true;
925 net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
926 nh[optoff + 1]);
927 kfree_skb(skb);
928 return false;
931 /* Jumbo payload */
933 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
935 const unsigned char *nh = skb_network_header(skb);
936 struct inet6_dev *idev = __in6_dev_get_safely(skb->dev);
937 struct net *net = ipv6_skb_net(skb);
938 u32 pkt_len;
940 if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
941 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
942 nh[optoff+1]);
943 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
944 goto drop;
947 pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
948 if (pkt_len <= IPV6_MAXPLEN) {
949 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
950 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
951 return false;
953 if (ipv6_hdr(skb)->payload_len) {
954 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
955 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
956 return false;
959 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
960 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INTRUNCATEDPKTS);
961 goto drop;
964 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
965 goto drop;
967 IP6CB(skb)->flags |= IP6SKB_JUMBOGRAM;
968 return true;
970 drop:
971 kfree_skb(skb);
972 return false;
975 /* CALIPSO RFC 5570 */
977 static bool ipv6_hop_calipso(struct sk_buff *skb, int optoff)
979 const unsigned char *nh = skb_network_header(skb);
981 if (nh[optoff + 1] < 8)
982 goto drop;
984 if (nh[optoff + 6] * 4 + 8 > nh[optoff + 1])
985 goto drop;
987 if (!calipso_validate(skb, nh + optoff))
988 goto drop;
990 return true;
992 drop:
993 kfree_skb(skb);
994 return false;
997 static const struct tlvtype_proc tlvprochopopt_lst[] = {
999 .type = IPV6_TLV_ROUTERALERT,
1000 .func = ipv6_hop_ra,
1003 .type = IPV6_TLV_JUMBO,
1004 .func = ipv6_hop_jumbo,
1007 .type = IPV6_TLV_CALIPSO,
1008 .func = ipv6_hop_calipso,
1010 { -1, }
1013 int ipv6_parse_hopopts(struct sk_buff *skb)
1015 struct inet6_skb_parm *opt = IP6CB(skb);
1016 struct net *net = dev_net(skb->dev);
1017 int extlen;
1020 * skb_network_header(skb) is equal to skb->data, and
1021 * skb_network_header_len(skb) is always equal to
1022 * sizeof(struct ipv6hdr) by definition of
1023 * hop-by-hop options.
1025 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
1026 !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
1027 ((skb_transport_header(skb)[1] + 1) << 3)))) {
1028 fail_and_free:
1029 kfree_skb(skb);
1030 return -1;
1033 extlen = (skb_transport_header(skb)[1] + 1) << 3;
1034 if (extlen > net->ipv6.sysctl.max_hbh_opts_len)
1035 goto fail_and_free;
1037 opt->flags |= IP6SKB_HOPBYHOP;
1038 if (ip6_parse_tlv(tlvprochopopt_lst, skb,
1039 init_net.ipv6.sysctl.max_hbh_opts_cnt)) {
1040 skb->transport_header += extlen;
1041 opt = IP6CB(skb);
1042 opt->nhoff = sizeof(struct ipv6hdr);
1043 return 1;
1045 return -1;
1049 * Creating outbound headers.
1051 * "build" functions work when skb is filled from head to tail (datagram)
1052 * "push" functions work when headers are added from tail to head (tcp)
1054 * In both cases we assume, that caller reserved enough room
1055 * for headers.
1058 static void ipv6_push_rthdr0(struct sk_buff *skb, u8 *proto,
1059 struct ipv6_rt_hdr *opt,
1060 struct in6_addr **addr_p, struct in6_addr *saddr)
1062 struct rt0_hdr *phdr, *ihdr;
1063 int hops;
1065 ihdr = (struct rt0_hdr *) opt;
1067 phdr = skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
1068 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
1070 hops = ihdr->rt_hdr.hdrlen >> 1;
1072 if (hops > 1)
1073 memcpy(phdr->addr, ihdr->addr + 1,
1074 (hops - 1) * sizeof(struct in6_addr));
1076 phdr->addr[hops - 1] = **addr_p;
1077 *addr_p = ihdr->addr;
1079 phdr->rt_hdr.nexthdr = *proto;
1080 *proto = NEXTHDR_ROUTING;
1083 static void ipv6_push_rthdr4(struct sk_buff *skb, u8 *proto,
1084 struct ipv6_rt_hdr *opt,
1085 struct in6_addr **addr_p, struct in6_addr *saddr)
1087 struct ipv6_sr_hdr *sr_phdr, *sr_ihdr;
1088 int plen, hops;
1090 sr_ihdr = (struct ipv6_sr_hdr *)opt;
1091 plen = (sr_ihdr->hdrlen + 1) << 3;
1093 sr_phdr = skb_push(skb, plen);
1094 memcpy(sr_phdr, sr_ihdr, sizeof(struct ipv6_sr_hdr));
1096 hops = sr_ihdr->first_segment + 1;
1097 memcpy(sr_phdr->segments + 1, sr_ihdr->segments + 1,
1098 (hops - 1) * sizeof(struct in6_addr));
1100 sr_phdr->segments[0] = **addr_p;
1101 *addr_p = &sr_ihdr->segments[sr_ihdr->segments_left];
1103 if (sr_ihdr->hdrlen > hops * 2) {
1104 int tlvs_offset, tlvs_length;
1106 tlvs_offset = (1 + hops * 2) << 3;
1107 tlvs_length = (sr_ihdr->hdrlen - hops * 2) << 3;
1108 memcpy((char *)sr_phdr + tlvs_offset,
1109 (char *)sr_ihdr + tlvs_offset, tlvs_length);
1112 #ifdef CONFIG_IPV6_SEG6_HMAC
1113 if (sr_has_hmac(sr_phdr)) {
1114 struct net *net = NULL;
1116 if (skb->dev)
1117 net = dev_net(skb->dev);
1118 else if (skb->sk)
1119 net = sock_net(skb->sk);
1121 WARN_ON(!net);
1123 if (net)
1124 seg6_push_hmac(net, saddr, sr_phdr);
1126 #endif
1128 sr_phdr->nexthdr = *proto;
1129 *proto = NEXTHDR_ROUTING;
1132 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
1133 struct ipv6_rt_hdr *opt,
1134 struct in6_addr **addr_p, struct in6_addr *saddr)
1136 switch (opt->type) {
1137 case IPV6_SRCRT_TYPE_0:
1138 case IPV6_SRCRT_STRICT:
1139 case IPV6_SRCRT_TYPE_2:
1140 ipv6_push_rthdr0(skb, proto, opt, addr_p, saddr);
1141 break;
1142 case IPV6_SRCRT_TYPE_4:
1143 ipv6_push_rthdr4(skb, proto, opt, addr_p, saddr);
1144 break;
1145 default:
1146 break;
1150 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
1152 struct ipv6_opt_hdr *h = skb_push(skb, ipv6_optlen(opt));
1154 memcpy(h, opt, ipv6_optlen(opt));
1155 h->nexthdr = *proto;
1156 *proto = type;
1159 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1160 u8 *proto,
1161 struct in6_addr **daddr, struct in6_addr *saddr)
1163 if (opt->srcrt) {
1164 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr, saddr);
1166 * IPV6_RTHDRDSTOPTS is ignored
1167 * unless IPV6_RTHDR is set (RFC3542).
1169 if (opt->dst0opt)
1170 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
1172 if (opt->hopopt)
1173 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
1176 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
1178 if (opt->dst1opt)
1179 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
1181 EXPORT_SYMBOL(ipv6_push_frag_opts);
1183 struct ipv6_txoptions *
1184 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
1186 struct ipv6_txoptions *opt2;
1188 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
1189 if (opt2) {
1190 long dif = (char *)opt2 - (char *)opt;
1191 memcpy(opt2, opt, opt->tot_len);
1192 if (opt2->hopopt)
1193 *((char **)&opt2->hopopt) += dif;
1194 if (opt2->dst0opt)
1195 *((char **)&opt2->dst0opt) += dif;
1196 if (opt2->dst1opt)
1197 *((char **)&opt2->dst1opt) += dif;
1198 if (opt2->srcrt)
1199 *((char **)&opt2->srcrt) += dif;
1200 refcount_set(&opt2->refcnt, 1);
1202 return opt2;
1204 EXPORT_SYMBOL_GPL(ipv6_dup_options);
1206 static void ipv6_renew_option(int renewtype,
1207 struct ipv6_opt_hdr **dest,
1208 struct ipv6_opt_hdr *old,
1209 struct ipv6_opt_hdr *new,
1210 int newtype, char **p)
1212 struct ipv6_opt_hdr *src;
1214 src = (renewtype == newtype ? new : old);
1215 if (!src)
1216 return;
1218 memcpy(*p, src, ipv6_optlen(src));
1219 *dest = (struct ipv6_opt_hdr *)*p;
1220 *p += CMSG_ALIGN(ipv6_optlen(*dest));
1224 * ipv6_renew_options - replace a specific ext hdr with a new one.
1226 * @sk: sock from which to allocate memory
1227 * @opt: original options
1228 * @newtype: option type to replace in @opt
1229 * @newopt: new option of type @newtype to replace (user-mem)
1231 * Returns a new set of options which is a copy of @opt with the
1232 * option type @newtype replaced with @newopt.
1234 * @opt may be NULL, in which case a new set of options is returned
1235 * containing just @newopt.
1237 * @newopt may be NULL, in which case the specified option type is
1238 * not copied into the new set of options.
1240 * The new set of options is allocated from the socket option memory
1241 * buffer of @sk.
1243 struct ipv6_txoptions *
1244 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
1245 int newtype, struct ipv6_opt_hdr *newopt)
1247 int tot_len = 0;
1248 char *p;
1249 struct ipv6_txoptions *opt2;
1251 if (opt) {
1252 if (newtype != IPV6_HOPOPTS && opt->hopopt)
1253 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
1254 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
1255 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
1256 if (newtype != IPV6_RTHDR && opt->srcrt)
1257 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
1258 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
1259 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
1262 if (newopt)
1263 tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
1265 if (!tot_len)
1266 return NULL;
1268 tot_len += sizeof(*opt2);
1269 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
1270 if (!opt2)
1271 return ERR_PTR(-ENOBUFS);
1273 memset(opt2, 0, tot_len);
1274 refcount_set(&opt2->refcnt, 1);
1275 opt2->tot_len = tot_len;
1276 p = (char *)(opt2 + 1);
1278 ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
1279 (opt ? opt->hopopt : NULL),
1280 newopt, newtype, &p);
1281 ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
1282 (opt ? opt->dst0opt : NULL),
1283 newopt, newtype, &p);
1284 ipv6_renew_option(IPV6_RTHDR,
1285 (struct ipv6_opt_hdr **)&opt2->srcrt,
1286 (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
1287 newopt, newtype, &p);
1288 ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
1289 (opt ? opt->dst1opt : NULL),
1290 newopt, newtype, &p);
1292 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
1293 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
1294 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
1295 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
1297 return opt2;
1300 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
1301 struct ipv6_txoptions *opt)
1304 * ignore the dest before srcrt unless srcrt is being included.
1305 * --yoshfuji
1307 if (opt && opt->dst0opt && !opt->srcrt) {
1308 if (opt_space != opt) {
1309 memcpy(opt_space, opt, sizeof(*opt_space));
1310 opt = opt_space;
1312 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
1313 opt->dst0opt = NULL;
1316 return opt;
1318 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
1321 * fl6_update_dst - update flowi destination address with info given
1322 * by srcrt option, if any.
1324 * @fl6: flowi6 for which daddr is to be updated
1325 * @opt: struct ipv6_txoptions in which to look for srcrt opt
1326 * @orig: copy of original daddr address if modified
1328 * Returns NULL if no txoptions or no srcrt, otherwise returns orig
1329 * and initial value of fl6->daddr set in orig
1331 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1332 const struct ipv6_txoptions *opt,
1333 struct in6_addr *orig)
1335 if (!opt || !opt->srcrt)
1336 return NULL;
1338 *orig = fl6->daddr;
1340 switch (opt->srcrt->type) {
1341 case IPV6_SRCRT_TYPE_0:
1342 case IPV6_SRCRT_STRICT:
1343 case IPV6_SRCRT_TYPE_2:
1344 fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
1345 break;
1346 case IPV6_SRCRT_TYPE_4:
1348 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)opt->srcrt;
1350 fl6->daddr = srh->segments[srh->segments_left];
1351 break;
1353 default:
1354 return NULL;
1357 return orig;
1359 EXPORT_SYMBOL_GPL(fl6_update_dst);