Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / ipv6 / exthdrs.c
1 /*
2  *      Extension Header handling for IPv6
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Andi Kleen              <ak@muc.de>
8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
9  *
10  *      This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  */
15
16 /* Changes:
17  *      yoshfuji                : ensure not to overrun while parsing
18  *                                tlv options.
19  *      Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
20  *      YOSHIFUJI Hideaki @USAGI  Register inbound extension header
21  *                                handlers as inet6_protocol{}.
22  */
23
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/netdevice.h>
30 #include <linux/in6.h>
31 #include <linux/icmpv6.h>
32 #include <linux/slab.h>
33 #include <linux/export.h>
34
35 #include <net/dst.h>
36 #include <net/sock.h>
37 #include <net/snmp.h>
38
39 #include <net/ipv6.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>
46 #if IS_ENABLED(CONFIG_IPV6_MIP6)
47 #include <net/xfrm.h>
48 #endif
49
50 #include <linux/uaccess.h>
51
52 /*
53  *      Parsing tlv encoded headers.
54  *
55  *      Parsing function "func" returns true, if parsing succeed
56  *      and false, if it failed.
57  *      It MUST NOT touch skb->h.
58  */
59
60 struct tlvtype_proc {
61         int     type;
62         bool    (*func)(struct sk_buff *skb, int offset);
63 };
64
65 /*********************
66   Generic functions
67  *********************/
68
69 /* An unknown option is detected, decide what to do */
70
71 static bool ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
72 {
73         switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
74         case 0: /* ignore */
75                 return true;
76
77         case 1: /* drop packet */
78                 break;
79
80         case 3: /* Send ICMP if not a multicast address and drop packet */
81                 /* Actually, it is redundant check. icmp_send
82                    will recheck in any case.
83                  */
84                 if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
85                         break;
86         case 2: /* send ICMP PARM PROB regardless and drop packet */
87                 icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
88                 return false;
89         }
90
91         kfree_skb(skb);
92         return false;
93 }
94
95 /* Parse tlv encoded option header (hop-by-hop or destination) */
96
97 static bool ip6_parse_tlv(const struct tlvtype_proc *procs, struct sk_buff *skb)
98 {
99         const struct tlvtype_proc *curr;
100         const unsigned char *nh = skb_network_header(skb);
101         int off = skb_network_header_len(skb);
102         int len = (skb_transport_header(skb)[1] + 1) << 3;
103         int padlen = 0;
104
105         if (skb_transport_offset(skb) + len > skb_headlen(skb))
106                 goto bad;
107
108         off += 2;
109         len -= 2;
110
111         while (len > 0) {
112                 int optlen = nh[off + 1] + 2;
113                 int i;
114
115                 switch (nh[off]) {
116                 case IPV6_TLV_PAD1:
117                         optlen = 1;
118                         padlen++;
119                         if (padlen > 7)
120                                 goto bad;
121                         break;
122
123                 case IPV6_TLV_PADN:
124                         /* RFC 2460 states that the purpose of PadN is
125                          * to align the containing header to multiples
126                          * of 8. 7 is therefore the highest valid value.
127                          * See also RFC 4942, Section 2.1.9.5.
128                          */
129                         padlen += optlen;
130                         if (padlen > 7)
131                                 goto bad;
132                         /* RFC 4942 recommends receiving hosts to
133                          * actively check PadN payload to contain
134                          * only zeroes.
135                          */
136                         for (i = 2; i < optlen; i++) {
137                                 if (nh[off + i] != 0)
138                                         goto bad;
139                         }
140                         break;
141
142                 default: /* Other TLV code so scan list */
143                         if (optlen > len)
144                                 goto bad;
145                         for (curr = procs; curr->type >= 0; curr++) {
146                                 if (curr->type == nh[off]) {
147                                         /* type specific length/alignment
148                                            checks will be performed in the
149                                            func(). */
150                                         if (curr->func(skb, off) == false)
151                                                 return false;
152                                         break;
153                                 }
154                         }
155                         if (curr->type < 0) {
156                                 if (ip6_tlvopt_unknown(skb, off) == 0)
157                                         return false;
158                         }
159                         padlen = 0;
160                         break;
161                 }
162                 off += optlen;
163                 len -= optlen;
164         }
165
166         if (len == 0)
167                 return true;
168 bad:
169         kfree_skb(skb);
170         return false;
171 }
172
173 /*****************************
174   Destination options header.
175  *****************************/
176
177 #if IS_ENABLED(CONFIG_IPV6_MIP6)
178 static bool ipv6_dest_hao(struct sk_buff *skb, int optoff)
179 {
180         struct ipv6_destopt_hao *hao;
181         struct inet6_skb_parm *opt = IP6CB(skb);
182         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
183         struct in6_addr tmp_addr;
184         int ret;
185
186         if (opt->dsthao) {
187                 net_dbg_ratelimited("hao duplicated\n");
188                 goto discard;
189         }
190         opt->dsthao = opt->dst1;
191         opt->dst1 = 0;
192
193         hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
194
195         if (hao->length != 16) {
196                 net_dbg_ratelimited("hao invalid option length = %d\n",
197                                     hao->length);
198                 goto discard;
199         }
200
201         if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
202                 net_dbg_ratelimited("hao is not an unicast addr: %pI6\n",
203                                     &hao->addr);
204                 goto discard;
205         }
206
207         ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
208                                (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
209         if (unlikely(ret < 0))
210                 goto discard;
211
212         if (skb_cloned(skb)) {
213                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
214                         goto discard;
215
216                 /* update all variable using below by copied skbuff */
217                 hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
218                                                   optoff);
219                 ipv6h = ipv6_hdr(skb);
220         }
221
222         if (skb->ip_summed == CHECKSUM_COMPLETE)
223                 skb->ip_summed = CHECKSUM_NONE;
224
225         tmp_addr = ipv6h->saddr;
226         ipv6h->saddr = hao->addr;
227         hao->addr = tmp_addr;
228
229         if (skb->tstamp.tv64 == 0)
230                 __net_timestamp(skb);
231
232         return true;
233
234  discard:
235         kfree_skb(skb);
236         return false;
237 }
238 #endif
239
240 static const struct tlvtype_proc tlvprocdestopt_lst[] = {
241 #if IS_ENABLED(CONFIG_IPV6_MIP6)
242         {
243                 .type   = IPV6_TLV_HAO,
244                 .func   = ipv6_dest_hao,
245         },
246 #endif
247         {-1,                    NULL}
248 };
249
250 static int ipv6_destopt_rcv(struct sk_buff *skb)
251 {
252         struct inet6_skb_parm *opt = IP6CB(skb);
253 #if IS_ENABLED(CONFIG_IPV6_MIP6)
254         __u16 dstbuf;
255 #endif
256         struct dst_entry *dst = skb_dst(skb);
257
258         if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
259             !pskb_may_pull(skb, (skb_transport_offset(skb) +
260                                  ((skb_transport_header(skb)[1] + 1) << 3)))) {
261                 IP6_INC_STATS_BH(dev_net(dst->dev), ip6_dst_idev(dst),
262                                  IPSTATS_MIB_INHDRERRORS);
263                 kfree_skb(skb);
264                 return -1;
265         }
266
267         opt->lastopt = opt->dst1 = skb_network_header_len(skb);
268 #if IS_ENABLED(CONFIG_IPV6_MIP6)
269         dstbuf = opt->dst1;
270 #endif
271
272         if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
273                 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
274                 opt = IP6CB(skb);
275 #if IS_ENABLED(CONFIG_IPV6_MIP6)
276                 opt->nhoff = dstbuf;
277 #else
278                 opt->nhoff = opt->dst1;
279 #endif
280                 return 1;
281         }
282
283         IP6_INC_STATS_BH(dev_net(dst->dev),
284                          ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
285         return -1;
286 }
287
288 /********************************
289   Routing header.
290  ********************************/
291
292 /* called with rcu_read_lock() */
293 static int ipv6_rthdr_rcv(struct sk_buff *skb)
294 {
295         struct inet6_skb_parm *opt = IP6CB(skb);
296         struct in6_addr *addr = NULL;
297         struct in6_addr daddr;
298         struct inet6_dev *idev;
299         int n, i;
300         struct ipv6_rt_hdr *hdr;
301         struct rt0_hdr *rthdr;
302         struct net *net = dev_net(skb->dev);
303         int accept_source_route = net->ipv6.devconf_all->accept_source_route;
304
305         idev = __in6_dev_get(skb->dev);
306         if (idev && accept_source_route > idev->cnf.accept_source_route)
307                 accept_source_route = idev->cnf.accept_source_route;
308
309         if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
310             !pskb_may_pull(skb, (skb_transport_offset(skb) +
311                                  ((skb_transport_header(skb)[1] + 1) << 3)))) {
312                 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
313                                  IPSTATS_MIB_INHDRERRORS);
314                 kfree_skb(skb);
315                 return -1;
316         }
317
318         hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
319
320         if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
321             skb->pkt_type != PACKET_HOST) {
322                 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
323                                  IPSTATS_MIB_INADDRERRORS);
324                 kfree_skb(skb);
325                 return -1;
326         }
327
328 looped_back:
329         if (hdr->segments_left == 0) {
330                 switch (hdr->type) {
331 #if IS_ENABLED(CONFIG_IPV6_MIP6)
332                 case IPV6_SRCRT_TYPE_2:
333                         /* Silently discard type 2 header unless it was
334                          * processed by own
335                          */
336                         if (!addr) {
337                                 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
338                                                  IPSTATS_MIB_INADDRERRORS);
339                                 kfree_skb(skb);
340                                 return -1;
341                         }
342                         break;
343 #endif
344                 default:
345                         break;
346                 }
347
348                 opt->lastopt = opt->srcrt = skb_network_header_len(skb);
349                 skb->transport_header += (hdr->hdrlen + 1) << 3;
350                 opt->dst0 = opt->dst1;
351                 opt->dst1 = 0;
352                 opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
353                 return 1;
354         }
355
356         switch (hdr->type) {
357 #if IS_ENABLED(CONFIG_IPV6_MIP6)
358         case IPV6_SRCRT_TYPE_2:
359                 if (accept_source_route < 0)
360                         goto unknown_rh;
361                 /* Silently discard invalid RTH type 2 */
362                 if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
363                         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
364                                          IPSTATS_MIB_INHDRERRORS);
365                         kfree_skb(skb);
366                         return -1;
367                 }
368                 break;
369 #endif
370         default:
371                 goto unknown_rh;
372         }
373
374         /*
375          *      This is the routing header forwarding algorithm from
376          *      RFC 2460, page 16.
377          */
378
379         n = hdr->hdrlen >> 1;
380
381         if (hdr->segments_left > n) {
382                 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
383                                  IPSTATS_MIB_INHDRERRORS);
384                 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
385                                   ((&hdr->segments_left) -
386                                    skb_network_header(skb)));
387                 return -1;
388         }
389
390         /* We are about to mangle packet header. Be careful!
391            Do not damage packets queued somewhere.
392          */
393         if (skb_cloned(skb)) {
394                 /* the copy is a forwarded packet */
395                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
396                         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
397                                          IPSTATS_MIB_OUTDISCARDS);
398                         kfree_skb(skb);
399                         return -1;
400                 }
401                 hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
402         }
403
404         if (skb->ip_summed == CHECKSUM_COMPLETE)
405                 skb->ip_summed = CHECKSUM_NONE;
406
407         i = n - --hdr->segments_left;
408
409         rthdr = (struct rt0_hdr *) hdr;
410         addr = rthdr->addr;
411         addr += i - 1;
412
413         switch (hdr->type) {
414 #if IS_ENABLED(CONFIG_IPV6_MIP6)
415         case IPV6_SRCRT_TYPE_2:
416                 if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
417                                      (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
418                                      IPPROTO_ROUTING) < 0) {
419                         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
420                                          IPSTATS_MIB_INADDRERRORS);
421                         kfree_skb(skb);
422                         return -1;
423                 }
424                 if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
425                         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
426                                          IPSTATS_MIB_INADDRERRORS);
427                         kfree_skb(skb);
428                         return -1;
429                 }
430                 break;
431 #endif
432         default:
433                 break;
434         }
435
436         if (ipv6_addr_is_multicast(addr)) {
437                 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
438                                  IPSTATS_MIB_INADDRERRORS);
439                 kfree_skb(skb);
440                 return -1;
441         }
442
443         daddr = *addr;
444         *addr = ipv6_hdr(skb)->daddr;
445         ipv6_hdr(skb)->daddr = daddr;
446
447         skb_dst_drop(skb);
448         ip6_route_input(skb);
449         if (skb_dst(skb)->error) {
450                 skb_push(skb, skb->data - skb_network_header(skb));
451                 dst_input(skb);
452                 return -1;
453         }
454
455         if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
456                 if (ipv6_hdr(skb)->hop_limit <= 1) {
457                         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
458                                          IPSTATS_MIB_INHDRERRORS);
459                         icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
460                                     0);
461                         kfree_skb(skb);
462                         return -1;
463                 }
464                 ipv6_hdr(skb)->hop_limit--;
465                 goto looped_back;
466         }
467
468         skb_push(skb, skb->data - skb_network_header(skb));
469         dst_input(skb);
470         return -1;
471
472 unknown_rh:
473         IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
474         icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
475                           (&hdr->type) - skb_network_header(skb));
476         return -1;
477 }
478
479 static const struct inet6_protocol rthdr_protocol = {
480         .handler        =       ipv6_rthdr_rcv,
481         .flags          =       INET6_PROTO_NOPOLICY,
482 };
483
484 static const struct inet6_protocol destopt_protocol = {
485         .handler        =       ipv6_destopt_rcv,
486         .flags          =       INET6_PROTO_NOPOLICY,
487 };
488
489 static const struct inet6_protocol nodata_protocol = {
490         .handler        =       dst_discard,
491         .flags          =       INET6_PROTO_NOPOLICY,
492 };
493
494 int __init ipv6_exthdrs_init(void)
495 {
496         int ret;
497
498         ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
499         if (ret)
500                 goto out;
501
502         ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
503         if (ret)
504                 goto out_rthdr;
505
506         ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
507         if (ret)
508                 goto out_destopt;
509
510 out:
511         return ret;
512 out_destopt:
513         inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
514 out_rthdr:
515         inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
516         goto out;
517 };
518
519 void ipv6_exthdrs_exit(void)
520 {
521         inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
522         inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
523         inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
524 }
525
526 /**********************************
527   Hop-by-hop options.
528  **********************************/
529
530 /*
531  * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
532  */
533 static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
534 {
535         return skb_dst(skb) ? ip6_dst_idev(skb_dst(skb)) : __in6_dev_get(skb->dev);
536 }
537
538 static inline struct net *ipv6_skb_net(struct sk_buff *skb)
539 {
540         return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
541 }
542
543 /* Router Alert as of RFC 2711 */
544
545 static bool ipv6_hop_ra(struct sk_buff *skb, int optoff)
546 {
547         const unsigned char *nh = skb_network_header(skb);
548
549         if (nh[optoff + 1] == 2) {
550                 IP6CB(skb)->flags |= IP6SKB_ROUTERALERT;
551                 memcpy(&IP6CB(skb)->ra, nh + optoff + 2, sizeof(IP6CB(skb)->ra));
552                 return true;
553         }
554         net_dbg_ratelimited("ipv6_hop_ra: wrong RA length %d\n",
555                             nh[optoff + 1]);
556         kfree_skb(skb);
557         return false;
558 }
559
560 /* Jumbo payload */
561
562 static bool ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
563 {
564         const unsigned char *nh = skb_network_header(skb);
565         struct net *net = ipv6_skb_net(skb);
566         u32 pkt_len;
567
568         if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
569                 net_dbg_ratelimited("ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
570                                     nh[optoff+1]);
571                 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
572                                  IPSTATS_MIB_INHDRERRORS);
573                 goto drop;
574         }
575
576         pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
577         if (pkt_len <= IPV6_MAXPLEN) {
578                 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
579                                  IPSTATS_MIB_INHDRERRORS);
580                 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
581                 return false;
582         }
583         if (ipv6_hdr(skb)->payload_len) {
584                 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
585                                  IPSTATS_MIB_INHDRERRORS);
586                 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
587                 return false;
588         }
589
590         if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
591                 IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
592                                  IPSTATS_MIB_INTRUNCATEDPKTS);
593                 goto drop;
594         }
595
596         if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
597                 goto drop;
598
599         return true;
600
601 drop:
602         kfree_skb(skb);
603         return false;
604 }
605
606 static const struct tlvtype_proc tlvprochopopt_lst[] = {
607         {
608                 .type   = IPV6_TLV_ROUTERALERT,
609                 .func   = ipv6_hop_ra,
610         },
611         {
612                 .type   = IPV6_TLV_JUMBO,
613                 .func   = ipv6_hop_jumbo,
614         },
615         { -1, }
616 };
617
618 int ipv6_parse_hopopts(struct sk_buff *skb)
619 {
620         struct inet6_skb_parm *opt = IP6CB(skb);
621
622         /*
623          * skb_network_header(skb) is equal to skb->data, and
624          * skb_network_header_len(skb) is always equal to
625          * sizeof(struct ipv6hdr) by definition of
626          * hop-by-hop options.
627          */
628         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
629             !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
630                                  ((skb_transport_header(skb)[1] + 1) << 3)))) {
631                 kfree_skb(skb);
632                 return -1;
633         }
634
635         opt->hop = sizeof(struct ipv6hdr);
636         if (ip6_parse_tlv(tlvprochopopt_lst, skb)) {
637                 skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
638                 opt = IP6CB(skb);
639                 opt->nhoff = sizeof(struct ipv6hdr);
640                 return 1;
641         }
642         return -1;
643 }
644
645 /*
646  *      Creating outbound headers.
647  *
648  *      "build" functions work when skb is filled from head to tail (datagram)
649  *      "push"  functions work when headers are added from tail to head (tcp)
650  *
651  *      In both cases we assume, that caller reserved enough room
652  *      for headers.
653  */
654
655 static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
656                             struct ipv6_rt_hdr *opt,
657                             struct in6_addr **addr_p)
658 {
659         struct rt0_hdr *phdr, *ihdr;
660         int hops;
661
662         ihdr = (struct rt0_hdr *) opt;
663
664         phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
665         memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
666
667         hops = ihdr->rt_hdr.hdrlen >> 1;
668
669         if (hops > 1)
670                 memcpy(phdr->addr, ihdr->addr + 1,
671                        (hops - 1) * sizeof(struct in6_addr));
672
673         phdr->addr[hops - 1] = **addr_p;
674         *addr_p = ihdr->addr;
675
676         phdr->rt_hdr.nexthdr = *proto;
677         *proto = NEXTHDR_ROUTING;
678 }
679
680 static void ipv6_push_exthdr(struct sk_buff *skb, u8 *proto, u8 type, struct ipv6_opt_hdr *opt)
681 {
682         struct ipv6_opt_hdr *h = (struct ipv6_opt_hdr *)skb_push(skb, ipv6_optlen(opt));
683
684         memcpy(h, opt, ipv6_optlen(opt));
685         h->nexthdr = *proto;
686         *proto = type;
687 }
688
689 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
690                           u8 *proto,
691                           struct in6_addr **daddr)
692 {
693         if (opt->srcrt) {
694                 ipv6_push_rthdr(skb, proto, opt->srcrt, daddr);
695                 /*
696                  * IPV6_RTHDRDSTOPTS is ignored
697                  * unless IPV6_RTHDR is set (RFC3542).
698                  */
699                 if (opt->dst0opt)
700                         ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst0opt);
701         }
702         if (opt->hopopt)
703                 ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
704 }
705 EXPORT_SYMBOL(ipv6_push_nfrag_opts);
706
707 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
708 {
709         if (opt->dst1opt)
710                 ipv6_push_exthdr(skb, proto, NEXTHDR_DEST, opt->dst1opt);
711 }
712
713 struct ipv6_txoptions *
714 ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
715 {
716         struct ipv6_txoptions *opt2;
717
718         opt2 = sock_kmalloc(sk, opt->tot_len, GFP_ATOMIC);
719         if (opt2) {
720                 long dif = (char *)opt2 - (char *)opt;
721                 memcpy(opt2, opt, opt->tot_len);
722                 if (opt2->hopopt)
723                         *((char **)&opt2->hopopt) += dif;
724                 if (opt2->dst0opt)
725                         *((char **)&opt2->dst0opt) += dif;
726                 if (opt2->dst1opt)
727                         *((char **)&opt2->dst1opt) += dif;
728                 if (opt2->srcrt)
729                         *((char **)&opt2->srcrt) += dif;
730         }
731         return opt2;
732 }
733 EXPORT_SYMBOL_GPL(ipv6_dup_options);
734
735 static int ipv6_renew_option(void *ohdr,
736                              struct ipv6_opt_hdr __user *newopt, int newoptlen,
737                              int inherit,
738                              struct ipv6_opt_hdr **hdr,
739                              char **p)
740 {
741         if (inherit) {
742                 if (ohdr) {
743                         memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
744                         *hdr = (struct ipv6_opt_hdr *)*p;
745                         *p += CMSG_ALIGN(ipv6_optlen(*hdr));
746                 }
747         } else {
748                 if (newopt) {
749                         if (copy_from_user(*p, newopt, newoptlen))
750                                 return -EFAULT;
751                         *hdr = (struct ipv6_opt_hdr *)*p;
752                         if (ipv6_optlen(*hdr) > newoptlen)
753                                 return -EINVAL;
754                         *p += CMSG_ALIGN(newoptlen);
755                 }
756         }
757         return 0;
758 }
759
760 struct ipv6_txoptions *
761 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
762                    int newtype,
763                    struct ipv6_opt_hdr __user *newopt, int newoptlen)
764 {
765         int tot_len = 0;
766         char *p;
767         struct ipv6_txoptions *opt2;
768         int err;
769
770         if (opt) {
771                 if (newtype != IPV6_HOPOPTS && opt->hopopt)
772                         tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
773                 if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
774                         tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
775                 if (newtype != IPV6_RTHDR && opt->srcrt)
776                         tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
777                 if (newtype != IPV6_DSTOPTS && opt->dst1opt)
778                         tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
779         }
780
781         if (newopt && newoptlen)
782                 tot_len += CMSG_ALIGN(newoptlen);
783
784         if (!tot_len)
785                 return NULL;
786
787         tot_len += sizeof(*opt2);
788         opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
789         if (!opt2)
790                 return ERR_PTR(-ENOBUFS);
791
792         memset(opt2, 0, tot_len);
793
794         opt2->tot_len = tot_len;
795         p = (char *)(opt2 + 1);
796
797         err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
798                                 newtype != IPV6_HOPOPTS,
799                                 &opt2->hopopt, &p);
800         if (err)
801                 goto out;
802
803         err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
804                                 newtype != IPV6_RTHDRDSTOPTS,
805                                 &opt2->dst0opt, &p);
806         if (err)
807                 goto out;
808
809         err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
810                                 newtype != IPV6_RTHDR,
811                                 (struct ipv6_opt_hdr **)&opt2->srcrt, &p);
812         if (err)
813                 goto out;
814
815         err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
816                                 newtype != IPV6_DSTOPTS,
817                                 &opt2->dst1opt, &p);
818         if (err)
819                 goto out;
820
821         opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
822                           (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
823                           (opt2->srcrt ? ipv6_optlen(opt2->srcrt) : 0);
824         opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
825
826         return opt2;
827 out:
828         sock_kfree_s(sk, opt2, opt2->tot_len);
829         return ERR_PTR(err);
830 }
831
832 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
833                                           struct ipv6_txoptions *opt)
834 {
835         /*
836          * ignore the dest before srcrt unless srcrt is being included.
837          * --yoshfuji
838          */
839         if (opt && opt->dst0opt && !opt->srcrt) {
840                 if (opt_space != opt) {
841                         memcpy(opt_space, opt, sizeof(*opt_space));
842                         opt = opt_space;
843                 }
844                 opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
845                 opt->dst0opt = NULL;
846         }
847
848         return opt;
849 }
850 EXPORT_SYMBOL_GPL(ipv6_fixup_options);
851
852 /**
853  * fl6_update_dst - update flowi destination address with info given
854  *                  by srcrt option, if any.
855  *
856  * @fl6: flowi6 for which daddr is to be updated
857  * @opt: struct ipv6_txoptions in which to look for srcrt opt
858  * @orig: copy of original daddr address if modified
859  *
860  * Returns NULL if no txoptions or no srcrt, otherwise returns orig
861  * and initial value of fl6->daddr set in orig
862  */
863 struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
864                                 const struct ipv6_txoptions *opt,
865                                 struct in6_addr *orig)
866 {
867         if (!opt || !opt->srcrt)
868                 return NULL;
869
870         *orig = fl6->daddr;
871         fl6->daddr = *((struct rt0_hdr *)opt->srcrt)->addr;
872         return orig;
873 }
874 EXPORT_SYMBOL_GPL(fl6_update_dst);