2 * Extension Header handling for IPv6
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Andi Kleen <ak@muc.de>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
10 * $Id: exthdrs.c,v 1.13 2001/06/19 15:58:56 davem Exp $
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
19 * yoshfuji : ensure not to overrun while parsing
21 * Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
22 * YOSHIFUJI Hideaki @USAGI Register inbound extension header
23 * handlers as inet6_protocol{}.
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
29 #include <linux/sockios.h>
30 #include <linux/sched.h>
31 #include <linux/net.h>
32 #include <linux/netdevice.h>
33 #include <linux/in6.h>
34 #include <linux/icmpv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/rawv6.h>
43 #include <net/ndisc.h>
44 #include <net/ip6_route.h>
45 #include <net/addrconf.h>
47 #include <asm/uaccess.h>
50 * Parsing tlv encoded headers.
52 * Parsing function "func" returns 1, if parsing succeed
53 * and 0, if it failed.
54 * It MUST NOT touch skb->h.
59 int (*func)(struct sk_buff *skb, int offset);
62 /*********************
64 *********************/
66 /* An unknown option is detected, decide what to do */
68 static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
70 switch ((skb->nh.raw[optoff] & 0xC0) >> 6) {
74 case 1: /* drop packet */
77 case 3: /* Send ICMP if not a multicast address and drop packet */
78 /* Actually, it is redundant check. icmp_send
79 will recheck in any case.
81 if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr))
83 case 2: /* send ICMP PARM PROB regardless and drop packet */
84 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
92 /* Parse tlv encoded option header (hop-by-hop or destination) */
94 static int ip6_parse_tlv(struct tlvtype_proc *procs, struct sk_buff *skb)
96 struct tlvtype_proc *curr;
97 int off = skb->h.raw - skb->nh.raw;
98 int len = ((skb->h.raw[1]+1)<<3);
100 if ((skb->h.raw + len) - skb->data > skb_headlen(skb))
107 int optlen = skb->nh.raw[off+1]+2;
109 switch (skb->nh.raw[off]) {
117 default: /* Other TLV code so scan list */
120 for (curr=procs; curr->type >= 0; curr++) {
121 if (curr->type == skb->nh.raw[off]) {
122 /* type specific length/alignment
123 checks will be performed in the
125 if (curr->func(skb, off) == 0)
130 if (curr->type < 0) {
131 if (ip6_tlvopt_unknown(skb, off) == 0)
146 /*****************************
147 Destination options header.
148 *****************************/
150 static struct tlvtype_proc tlvprocdestopt_lst[] = {
151 /* No destination options are defined now */
155 static int ipv6_destopt_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
157 struct sk_buff *skb = *skbp;
158 struct inet6_skb_parm *opt = IP6CB(skb);
160 if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+8) ||
161 !pskb_may_pull(skb, (skb->h.raw-skb->data)+((skb->h.raw[1]+1)<<3))) {
162 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
167 opt->lastopt = skb->h.raw - skb->nh.raw;
168 opt->dst1 = skb->h.raw - skb->nh.raw;
170 if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
171 skb->h.raw += ((skb->h.raw[1]+1)<<3);
176 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
180 static struct inet6_protocol destopt_protocol = {
181 .handler = ipv6_destopt_rcv,
182 .flags = INET6_PROTO_NOPOLICY,
185 void __init ipv6_destopt_init(void)
187 if (inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS) < 0)
188 printk(KERN_ERR "ipv6_destopt_init: Could not register protocol\n");
191 /********************************
192 NONE header. No data in packet.
193 ********************************/
195 static int ipv6_nodata_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
197 struct sk_buff *skb = *skbp;
203 static struct inet6_protocol nodata_protocol = {
204 .handler = ipv6_nodata_rcv,
205 .flags = INET6_PROTO_NOPOLICY,
208 void __init ipv6_nodata_init(void)
210 if (inet6_add_protocol(&nodata_protocol, IPPROTO_NONE) < 0)
211 printk(KERN_ERR "ipv6_nodata_init: Could not register protocol\n");
214 /********************************
216 ********************************/
218 static int ipv6_rthdr_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
220 struct sk_buff *skb = *skbp;
221 struct inet6_skb_parm *opt = IP6CB(skb);
222 struct in6_addr *addr;
223 struct in6_addr daddr;
226 struct ipv6_rt_hdr *hdr;
227 struct rt0_hdr *rthdr;
229 if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+8) ||
230 !pskb_may_pull(skb, (skb->h.raw-skb->data)+((skb->h.raw[1]+1)<<3))) {
231 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
236 hdr = (struct ipv6_rt_hdr *) skb->h.raw;
238 if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr) ||
239 skb->pkt_type != PACKET_HOST) {
240 IP6_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
246 if (hdr->segments_left == 0) {
247 opt->lastopt = skb->h.raw - skb->nh.raw;
248 opt->srcrt = skb->h.raw - skb->nh.raw;
249 skb->h.raw += (hdr->hdrlen + 1) << 3;
250 opt->dst0 = opt->dst1;
252 *nhoffp = (&hdr->nexthdr) - skb->nh.raw;
256 if (hdr->type != IPV6_SRCRT_TYPE_0) {
257 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
258 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->type) - skb->nh.raw);
262 if (hdr->hdrlen & 0x01) {
263 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
264 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->hdrlen) - skb->nh.raw);
269 * This is the routing header forwarding algorithm from
273 n = hdr->hdrlen >> 1;
275 if (hdr->segments_left > n) {
276 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
277 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->segments_left) - skb->nh.raw);
281 /* We are about to mangle packet header. Be careful!
282 Do not damage packets queued somewhere.
284 if (skb_cloned(skb)) {
285 struct sk_buff *skb2 = skb_copy(skb, GFP_ATOMIC);
287 /* the copy is a forwarded packet */
289 IP6_INC_STATS_BH(IPSTATS_MIB_OUTDISCARDS);
294 hdr = (struct ipv6_rt_hdr *) skb2->h.raw;
297 if (skb->ip_summed == CHECKSUM_HW)
298 skb->ip_summed = CHECKSUM_NONE;
300 i = n - --hdr->segments_left;
302 rthdr = (struct rt0_hdr *) hdr;
306 if (ipv6_addr_is_multicast(addr)) {
307 IP6_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
312 ipv6_addr_copy(&daddr, addr);
313 ipv6_addr_copy(addr, &skb->nh.ipv6h->daddr);
314 ipv6_addr_copy(&skb->nh.ipv6h->daddr, &daddr);
316 dst_release(xchg(&skb->dst, NULL));
317 ip6_route_input(skb);
318 if (skb->dst->error) {
319 skb_push(skb, skb->data - skb->nh.raw);
324 if (skb->dst->dev->flags&IFF_LOOPBACK) {
325 if (skb->nh.ipv6h->hop_limit <= 1) {
326 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
327 icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
332 skb->nh.ipv6h->hop_limit--;
336 skb_push(skb, skb->data - skb->nh.raw);
341 static struct inet6_protocol rthdr_protocol = {
342 .handler = ipv6_rthdr_rcv,
343 .flags = INET6_PROTO_NOPOLICY,
346 void __init ipv6_rthdr_init(void)
348 if (inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING) < 0)
349 printk(KERN_ERR "ipv6_rthdr_init: Could not register protocol\n");
353 This function inverts received rthdr.
354 NOTE: specs allow to make it automatically only if
355 packet authenticated.
357 I will not discuss it here (though, I am really pissed off at
358 this stupid requirement making rthdr idea useless)
360 Actually, it creates severe problems for us.
361 Embryonic requests has no associated sockets,
362 so that user have no control over it and
363 cannot not only to set reply options, but
364 even to know, that someone wants to connect
367 For now we need to test the engine, so that I created
368 temporary (or permanent) backdoor.
369 If listening socket set IPV6_RTHDR to 2, then we invert header.
373 struct ipv6_txoptions *
374 ipv6_invert_rthdr(struct sock *sk, struct ipv6_rt_hdr *hdr)
378 [ H1 -> H2 -> ... H_prev ] daddr=ME
381 [ H_prev -> ... -> H1 ] daddr =sender
383 Note, that IP output engine will rewrite this rthdr
384 by rotating it left by one addr.
388 struct rt0_hdr *rthdr = (struct rt0_hdr*)hdr;
389 struct rt0_hdr *irthdr;
390 struct ipv6_txoptions *opt;
391 int hdrlen = ipv6_optlen(hdr);
393 if (hdr->segments_left ||
394 hdr->type != IPV6_SRCRT_TYPE_0 ||
398 n = hdr->hdrlen >> 1;
399 opt = sock_kmalloc(sk, sizeof(*opt) + hdrlen, GFP_ATOMIC);
402 memset(opt, 0, sizeof(*opt));
403 opt->tot_len = sizeof(*opt) + hdrlen;
404 opt->srcrt = (void*)(opt+1);
405 opt->opt_nflen = hdrlen;
407 memcpy(opt->srcrt, hdr, sizeof(*hdr));
408 irthdr = (struct rt0_hdr*)opt->srcrt;
409 irthdr->reserved = 0;
410 opt->srcrt->segments_left = n;
412 memcpy(irthdr->addr+i, rthdr->addr+(n-1-i), 16);
416 /**********************************
418 **********************************/
420 /* Router Alert as of RFC 2711 */
422 static int ipv6_hop_ra(struct sk_buff *skb, int optoff)
424 if (skb->nh.raw[optoff+1] == 2) {
425 IP6CB(skb)->ra = optoff;
428 LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_ra: wrong RA length %d\n",
429 skb->nh.raw[optoff+1]);
436 static int ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
440 if (skb->nh.raw[optoff+1] != 4 || (optoff&3) != 2) {
441 LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
442 skb->nh.raw[optoff+1]);
443 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
447 pkt_len = ntohl(*(u32*)(skb->nh.raw+optoff+2));
448 if (pkt_len <= IPV6_MAXPLEN) {
449 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
450 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
453 if (skb->nh.ipv6h->payload_len) {
454 IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
455 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
459 if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
460 IP6_INC_STATS_BH(IPSTATS_MIB_INTRUNCATEDPKTS);
464 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
474 static struct tlvtype_proc tlvprochopopt_lst[] = {
476 .type = IPV6_TLV_ROUTERALERT,
480 .type = IPV6_TLV_JUMBO,
481 .func = ipv6_hop_jumbo,
486 int ipv6_parse_hopopts(struct sk_buff *skb, int nhoff)
488 IP6CB(skb)->hop = sizeof(struct ipv6hdr);
489 if (ip6_parse_tlv(tlvprochopopt_lst, skb))
490 return sizeof(struct ipv6hdr);
495 * Creating outbound headers.
497 * "build" functions work when skb is filled from head to tail (datagram)
498 * "push" functions work when headers are added from tail to head (tcp)
500 * In both cases we assume, that caller reserved enough room
504 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
505 struct ipv6_rt_hdr *opt,
506 struct in6_addr **addr_p)
508 struct rt0_hdr *phdr, *ihdr;
511 ihdr = (struct rt0_hdr *) opt;
513 phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
514 memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
516 hops = ihdr->rt_hdr.hdrlen >> 1;
519 memcpy(phdr->addr, ihdr->addr + 1,
520 (hops - 1) * sizeof(struct in6_addr));
522 ipv6_addr_copy(phdr->addr + (hops - 1), *addr_p);
523 *addr_p = ihdr->addr;
525 phdr->rt_hdr.nexthdr = *proto;
526 *proto = NEXTHDR_ROUTING;
529 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
531 struct ipv6_opt_hdr *h = (struct ipv6_opt_hdr *)skb_push(skb, ipv6_optlen(opt));
533 memcpy(h, opt, ipv6_optlen(opt));
538 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
540 struct in6_addr **daddr)
543 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr);
545 * IPV6_RTHDRDSTOPTS is ignored
546 * unless IPV6_RTHDR is set (RFC3542).
549 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
552 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
555 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
558 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
561 struct ipv6_txoptions *
562 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
564 struct ipv6_txoptions *opt2;
566 opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
568 long dif = (char*)opt2 - (char*)opt;
569 memcpy(opt2, opt, opt->tot_len);
571 *((char**)&opt2->hopopt) += dif;
573 *((char**)&opt2->dst0opt) += dif;
575 *((char**)&opt2->dst1opt) += dif;
577 *((char**)&opt2->srcrt) += dif;
582 static int ipv6_renew_option(void *ohdr,
583 struct ipv6_opt_hdr __user *newopt, int newoptlen,
585 struct ipv6_opt_hdr **hdr,
590 memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
591 *hdr = (struct ipv6_opt_hdr *)*p;
592 *p += CMSG_ALIGN(ipv6_optlen(*(struct ipv6_opt_hdr **)hdr));
596 if (copy_from_user(*p, newopt, newoptlen))
598 *hdr = (struct ipv6_opt_hdr *)*p;
599 if (ipv6_optlen(*(struct ipv6_opt_hdr **)hdr) > newoptlen)
601 *p += CMSG_ALIGN(newoptlen);
607 struct ipv6_txoptions *
608 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
610 struct ipv6_opt_hdr __user *newopt, int newoptlen)
614 struct ipv6_txoptions *opt2;
617 if (newtype != IPV6_HOPOPTS && opt->hopopt)
618 tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
619 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
620 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
621 if (newtype != IPV6_RTHDR && opt->srcrt)
622 tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
623 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
624 tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
625 if (newopt && newoptlen)
626 tot_len += CMSG_ALIGN(newoptlen);
631 tot_len += sizeof(*opt2);
632 opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
634 return ERR_PTR(-ENOBUFS);
636 memset(opt2, 0, tot_len);
638 opt2->tot_len = tot_len;
639 p = (char *)(opt2 + 1);
641 err = ipv6_renew_option(opt->hopopt, newopt, newoptlen,
642 newtype != IPV6_HOPOPTS,
647 err = ipv6_renew_option(opt->dst0opt, newopt, newoptlen,
648 newtype != IPV6_RTHDRDSTOPTS,
653 err = ipv6_renew_option(opt->srcrt, newopt, newoptlen,
654 newtype != IPV6_RTHDR,
655 (struct ipv6_opt_hdr **)opt2->srcrt, &p);
659 err = ipv6_renew_option(opt->dst1opt, newopt, newoptlen,
660 newtype != IPV6_DSTOPTS,
665 opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
666 (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
667 (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
668 opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
672 sock_kfree_s(sk, opt2, opt2->tot_len);
676 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
677 struct ipv6_txoptions *opt)
680 * ignore the dest before srcrt unless srcrt is being included.
683 if (opt && opt->dst0opt && !opt->srcrt) {
684 if (opt_space != opt) {
685 memcpy(opt_space, opt, sizeof(*opt_space));
688 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);