net/ipv6/netfilter/nf_nat_l3proto_ipv6.c

   1 /*
   2  * Copyright (c) 2011 Patrick McHardy <kaber@trash.net>
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License version 2 as
   6  * published by the Free Software Foundation.
   7  *
   8  * Development of IPv6 NAT funded by Astaro.
   9  */
  10 #include <linux/types.h>
  11 #include <linux/module.h>
  12 #include <linux/skbuff.h>
  13 #include <linux/ipv6.h>
  14 #include <linux/netfilter.h>
  15 #include <linux/netfilter_ipv6.h>
  16 #include <net/secure_seq.h>
  17 #include <net/checksum.h>
  18 #include <net/ip6_checksum.h>
  19 #include <net/ip6_route.h>
  20 #include <net/ipv6.h>
  21
  22 #include <net/netfilter/nf_conntrack_core.h>
  23 #include <net/netfilter/nf_conntrack.h>
  24 #include <net/netfilter/nf_nat_core.h>
  25 #include <net/netfilter/nf_nat_l3proto.h>
  26 #include <net/netfilter/nf_nat_l4proto.h>
  27
  28 static const struct nf_nat_l3proto nf_nat_l3proto_ipv6;
  29
  30 #ifdef CONFIG_XFRM
  31 static void nf_nat_ipv6_decode_session(struct sk_buff *skb,
  32                                        const struct nf_conn *ct,
  33                                        enum ip_conntrack_dir dir,
  34                                        unsigned long statusbit,
  35                                        struct flowi *fl)
  36 {
  37         const struct nf_conntrack_tuple *t = &ct->tuplehash[dir].tuple;
  38         struct flowi6 *fl6 = &fl->u.ip6;
  39
  40         if (ct->status & statusbit) {
  41                 fl6->daddr = t->dst.u3.in6;
  42                 if (t->dst.protonum == IPPROTO_TCP ||
  43                     t->dst.protonum == IPPROTO_UDP ||
  44                     t->dst.protonum == IPPROTO_UDPLITE ||
  45                     t->dst.protonum == IPPROTO_DCCP ||
  46                     t->dst.protonum == IPPROTO_SCTP)
  47                         fl6->fl6_dport = t->dst.u.all;
  48         }
  49
  50         statusbit ^= IPS_NAT_MASK;
  51
  52         if (ct->status & statusbit) {
  53                 fl6->saddr = t->src.u3.in6;
  54                 if (t->dst.protonum == IPPROTO_TCP ||
  55                     t->dst.protonum == IPPROTO_UDP ||
  56                     t->dst.protonum == IPPROTO_UDPLITE ||
  57                     t->dst.protonum == IPPROTO_DCCP ||
  58                     t->dst.protonum == IPPROTO_SCTP)
  59                         fl6->fl6_sport = t->src.u.all;
  60         }
  61 }
  62 #endif
  63
  64 static bool nf_nat_ipv6_in_range(const struct nf_conntrack_tuple *t,
  65                                  const struct nf_nat_range *range)
  66 {
  67         return ipv6_addr_cmp(&t->src.u3.in6, &range->min_addr.in6) >= 0 &&
  68                ipv6_addr_cmp(&t->src.u3.in6, &range->max_addr.in6) <= 0;
  69 }
  70
  71 static u32 nf_nat_ipv6_secure_port(const struct nf_conntrack_tuple *t,
  72                                    __be16 dport)
  73 {
  74         return secure_ipv6_port_ephemeral(t->src.u3.ip6, t->dst.u3.ip6, dport);
  75 }
  76
  77 static bool nf_nat_ipv6_manip_pkt(struct sk_buff *skb,
  78                                   unsigned int iphdroff,
  79                                   const struct nf_nat_l4proto *l4proto,
  80                                   const struct nf_conntrack_tuple *target,
  81                                   enum nf_nat_manip_type maniptype)
  82 {
  83         struct ipv6hdr *ipv6h;
  84         __be16 frag_off;
  85         int hdroff;
  86         u8 nexthdr;
  87
  88         if (!skb_make_writable(skb, iphdroff + sizeof(*ipv6h)))
  89                 return false;
  90
  91         ipv6h = (void *)skb->data + iphdroff;
  92         nexthdr = ipv6h->nexthdr;
  93         hdroff = ipv6_skip_exthdr(skb, iphdroff + sizeof(*ipv6h),
  94                                   &nexthdr, &frag_off);
  95         if (hdroff < 0)
  96                 goto manip_addr;
  97
  98         if ((frag_off & htons(~0x7)) == 0 &&
  99             !l4proto->manip_pkt(skb, &nf_nat_l3proto_ipv6, iphdroff, hdroff,
 100                                 target, maniptype))
 101                 return false;
 102 manip_addr:
 103         if (maniptype == NF_NAT_MANIP_SRC)
 104                 ipv6h->saddr = target->src.u3.in6;
 105         else
 106                 ipv6h->daddr = target->dst.u3.in6;
 107
 108         return true;
 109 }
 110
 111 static void nf_nat_ipv6_csum_update(struct sk_buff *skb,
 112                                     unsigned int iphdroff, __sum16 *check,
 113                                     const struct nf_conntrack_tuple *t,
 114                                     enum nf_nat_manip_type maniptype)
 115 {
 116         const struct ipv6hdr *ipv6h = (struct ipv6hdr *)(skb->data + iphdroff);
 117         const struct in6_addr *oldip, *newip;
 118
 119         if (maniptype == NF_NAT_MANIP_SRC) {
 120                 oldip = &ipv6h->saddr;
 121                 newip = &t->src.u3.in6;
 122         } else {
 123                 oldip = &ipv6h->daddr;
 124                 newip = &t->dst.u3.in6;
 125         }
 126         inet_proto_csum_replace16(check, skb, oldip->s6_addr32,
 127                                   newip->s6_addr32, true);
 128 }
 129
 130 static void nf_nat_ipv6_csum_recalc(struct sk_buff *skb,
 131                                     u8 proto, void *data, __sum16 *check,
 132                                     int datalen, int oldlen)
 133 {
 134         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 135                 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 136
 137                 skb->ip_summed = CHECKSUM_PARTIAL;
 138                 skb->csum_start = skb_headroom(skb) + skb_network_offset(skb) +
 139                         (data - (void *)skb->data);
 140                 skb->csum_offset = (void *)check - data;
 141                 *check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
 142                                           datalen, proto, 0);
 143         } else
 144                 inet_proto_csum_replace2(check, skb,
 145                                          htons(oldlen), htons(datalen), true);
 146 }
 147
 148 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 149 static int nf_nat_ipv6_nlattr_to_range(struct nlattr *tb[],
 150                                        struct nf_nat_range *range)
 151 {
 152         if (tb[CTA_NAT_V6_MINIP]) {
 153                 nla_memcpy(&range->min_addr.ip6, tb[CTA_NAT_V6_MINIP],
 154                            sizeof(struct in6_addr));
 155                 range->flags |= NF_NAT_RANGE_MAP_IPS;
 156         }
 157
 158         if (tb[CTA_NAT_V6_MAXIP])
 159                 nla_memcpy(&range->max_addr.ip6, tb[CTA_NAT_V6_MAXIP],
 160                            sizeof(struct in6_addr));
 161         else
 162                 range->max_addr = range->min_addr;
 163
 164         return 0;
 165 }
 166 #endif
 167
 168 static const struct nf_nat_l3proto nf_nat_l3proto_ipv6 = {
 169         .l3proto                = NFPROTO_IPV6,
 170         .secure_port            = nf_nat_ipv6_secure_port,
 171         .in_range               = nf_nat_ipv6_in_range,
 172         .manip_pkt              = nf_nat_ipv6_manip_pkt,
 173         .csum_update            = nf_nat_ipv6_csum_update,
 174         .csum_recalc            = nf_nat_ipv6_csum_recalc,
 175 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 176         .nlattr_to_range        = nf_nat_ipv6_nlattr_to_range,
 177 #endif
 178 #ifdef CONFIG_XFRM
 179         .decode_session = nf_nat_ipv6_decode_session,
 180 #endif
 181 };
 182
 183 int nf_nat_icmpv6_reply_translation(struct sk_buff *skb,
 184                                     struct nf_conn *ct,
 185                                     enum ip_conntrack_info ctinfo,
 186                                     unsigned int hooknum,
 187                                     unsigned int hdrlen)
 188 {
 189         struct {
 190                 struct icmp6hdr icmp6;
 191                 struct ipv6hdr  ip6;
 192         } *inside;
 193         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 194         enum nf_nat_manip_type manip = HOOK2MANIP(hooknum);
 195         const struct nf_nat_l4proto *l4proto;
 196         struct nf_conntrack_tuple target;
 197         unsigned long statusbit;
 198
 199         NF_CT_ASSERT(ctinfo == IP_CT_RELATED || ctinfo == IP_CT_RELATED_REPLY);
 200
 201         if (!skb_make_writable(skb, hdrlen + sizeof(*inside)))
 202                 return 0;
 203         if (nf_ip6_checksum(skb, hooknum, hdrlen, IPPROTO_ICMPV6))
 204                 return 0;
 205
 206         inside = (void *)skb->data + hdrlen;
 207         if (inside->icmp6.icmp6_type == NDISC_REDIRECT) {
 208                 if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK)
 209                         return 0;
 210                 if (ct->status & IPS_NAT_MASK)
 211                         return 0;
 212         }
 213
 214         if (manip == NF_NAT_MANIP_SRC)
 215                 statusbit = IPS_SRC_NAT;
 216         else
 217                 statusbit = IPS_DST_NAT;
 218
 219         /* Invert if this is reply direction */
 220         if (dir == IP_CT_DIR_REPLY)
 221                 statusbit ^= IPS_NAT_MASK;
 222
 223         if (!(ct->status & statusbit))
 224                 return 1;
 225
 226         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV6, inside->ip6.nexthdr);
 227         if (!nf_nat_ipv6_manip_pkt(skb, hdrlen + sizeof(inside->icmp6),
 228                                    l4proto, &ct->tuplehash[!dir].tuple, !manip))
 229                 return 0;
 230
 231         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 232                 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 233                 inside = (void *)skb->data + hdrlen;
 234                 inside->icmp6.icmp6_cksum = 0;
 235                 inside->icmp6.icmp6_cksum =
 236                         csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
 237                                         skb->len - hdrlen, IPPROTO_ICMPV6,
 238                                         csum_partial(&inside->icmp6,
 239                                                      skb->len - hdrlen, 0));
 240         }
 241
 242         nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple);
 243         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV6, IPPROTO_ICMPV6);
 244         if (!nf_nat_ipv6_manip_pkt(skb, 0, l4proto, &target, manip))
 245                 return 0;
 246
 247         return 1;
 248 }
 249 EXPORT_SYMBOL_GPL(nf_nat_icmpv6_reply_translation);
 250
 251 unsigned int
 252 nf_nat_ipv6_fn(void *priv, struct sk_buff *skb,
 253                const struct nf_hook_state *state,
 254                unsigned int (*do_chain)(void *priv,
 255                                         struct sk_buff *skb,
 256                                         const struct nf_hook_state *state,
 257                                         struct nf_conn *ct))
 258 {
 259         struct nf_conn *ct;
 260         enum ip_conntrack_info ctinfo;
 261         struct nf_conn_nat *nat;
 262         enum nf_nat_manip_type maniptype = HOOK2MANIP(state->hook);
 263         __be16 frag_off;
 264         int hdrlen;
 265         u8 nexthdr;
 266
 267         ct = nf_ct_get(skb, &ctinfo);
 268         /* Can't track?  It's not due to stress, or conntrack would
 269          * have dropped it.  Hence it's the user's responsibilty to
 270          * packet filter it out, or implement conntrack/NAT for that
 271          * protocol. 8) --RR
 272          */
 273         if (!ct)
 274                 return NF_ACCEPT;
 275
 276         /* Don't try to NAT if this packet is not conntracked */
 277         if (nf_ct_is_untracked(ct))
 278                 return NF_ACCEPT;
 279
 280         nat = nf_ct_nat_ext_add(ct);
 281         if (nat == NULL)
 282                 return NF_ACCEPT;
 283
 284         switch (ctinfo) {
 285         case IP_CT_RELATED:
 286         case IP_CT_RELATED_REPLY:
 287                 nexthdr = ipv6_hdr(skb)->nexthdr;
 288                 hdrlen = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
 289                                           &nexthdr, &frag_off);
 290
 291                 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
 292                         if (!nf_nat_icmpv6_reply_translation(skb, ct, ctinfo,
 293                                                              state->hook,
 294                                                              hdrlen))
 295                                 return NF_DROP;
 296                         else
 297                                 return NF_ACCEPT;
 298                 }
 299                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 300         case IP_CT_NEW:
 301                 /* Seen it before?  This can happen for loopback, retrans,
 302                  * or local packets.
 303                  */
 304                 if (!nf_nat_initialized(ct, maniptype)) {
 305                         unsigned int ret;
 306
 307                         ret = do_chain(priv, skb, state, ct);
 308                         if (ret != NF_ACCEPT)
 309                                 return ret;
 310
 311                         if (nf_nat_initialized(ct, HOOK2MANIP(state->hook)))
 312                                 break;
 313
 314                         ret = nf_nat_alloc_null_binding(ct, state->hook);
 315                         if (ret != NF_ACCEPT)
 316                                 return ret;
 317                 } else {
 318                         pr_debug("Already setup manip %s for ct %p\n",
 319                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 320                                  ct);
 321                         if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
 322                                 goto oif_changed;
 323                 }
 324                 break;
 325
 326         default:
 327                 /* ESTABLISHED */
 328                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 329                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 330                 if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
 331                         goto oif_changed;
 332         }
 333
 334         return nf_nat_packet(ct, ctinfo, state->hook, skb);
 335
 336 oif_changed:
 337         nf_ct_kill_acct(ct, ctinfo, skb);
 338         return NF_DROP;
 339 }
 340 EXPORT_SYMBOL_GPL(nf_nat_ipv6_fn);
 341
 342 unsigned int
 343 nf_nat_ipv6_in(void *priv, struct sk_buff *skb,
 344                const struct nf_hook_state *state,
 345                unsigned int (*do_chain)(void *priv,
 346                                         struct sk_buff *skb,
 347                                         const struct nf_hook_state *state,
 348                                         struct nf_conn *ct))
 349 {
 350         unsigned int ret;
 351         struct in6_addr daddr = ipv6_hdr(skb)->daddr;
 352
 353         ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
 354         if (ret != NF_DROP && ret != NF_STOLEN &&
 355             ipv6_addr_cmp(&daddr, &ipv6_hdr(skb)->daddr))
 356                 skb_dst_drop(skb);
 357
 358         return ret;
 359 }
 360 EXPORT_SYMBOL_GPL(nf_nat_ipv6_in);
 361
 362 unsigned int
 363 nf_nat_ipv6_out(void *priv, struct sk_buff *skb,
 364                 const struct nf_hook_state *state,
 365                 unsigned int (*do_chain)(void *priv,
 366                                          struct sk_buff *skb,
 367                                          const struct nf_hook_state *state,
 368                                          struct nf_conn *ct))
 369 {
 370 #ifdef CONFIG_XFRM
 371         const struct nf_conn *ct;
 372         enum ip_conntrack_info ctinfo;
 373         int err;
 374 #endif
 375         unsigned int ret;
 376
 377         /* root is playing with raw sockets. */
 378         if (skb->len < sizeof(struct ipv6hdr))
 379                 return NF_ACCEPT;
 380
 381         ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
 382 #ifdef CONFIG_XFRM
 383         if (ret != NF_DROP && ret != NF_STOLEN &&
 384             !(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 385             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 386                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 387
 388                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.src.u3,
 389                                       &ct->tuplehash[!dir].tuple.dst.u3) ||
 390                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 391                      ct->tuplehash[dir].tuple.src.u.all !=
 392                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 393                         err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
 394                         if (err < 0)
 395                                 ret = NF_DROP_ERR(err);
 396                 }
 397         }
 398 #endif
 399         return ret;
 400 }
 401 EXPORT_SYMBOL_GPL(nf_nat_ipv6_out);
 402
 403 unsigned int
 404 nf_nat_ipv6_local_fn(void *priv, struct sk_buff *skb,
 405                      const struct nf_hook_state *state,
 406                      unsigned int (*do_chain)(void *priv,
 407                                               struct sk_buff *skb,
 408                                               const struct nf_hook_state *state,
 409                                               struct nf_conn *ct))
 410 {
 411         const struct nf_conn *ct;
 412         enum ip_conntrack_info ctinfo;
 413         unsigned int ret;
 414         int err;
 415
 416         /* root is playing with raw sockets. */
 417         if (skb->len < sizeof(struct ipv6hdr))
 418                 return NF_ACCEPT;
 419
 420         ret = nf_nat_ipv6_fn(priv, skb, state, do_chain);
 421         if (ret != NF_DROP && ret != NF_STOLEN &&
 422             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 423                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 424
 425                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.dst.u3,
 426                                       &ct->tuplehash[!dir].tuple.src.u3)) {
 427                         err = ip6_route_me_harder(state->net, skb);
 428                         if (err < 0)
 429                                 ret = NF_DROP_ERR(err);
 430                 }
 431 #ifdef CONFIG_XFRM
 432                 else if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 433                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 434                          ct->tuplehash[dir].tuple.dst.u.all !=
 435                          ct->tuplehash[!dir].tuple.src.u.all) {
 436                         err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
 437                         if (err < 0)
 438                                 ret = NF_DROP_ERR(err);
 439                 }
 440 #endif
 441         }
 442         return ret;
 443 }
 444 EXPORT_SYMBOL_GPL(nf_nat_ipv6_local_fn);
 445
 446 static int __init nf_nat_l3proto_ipv6_init(void)
 447 {
 448         int err;
 449
 450         err = nf_nat_l4proto_register(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
 451         if (err < 0)
 452                 goto err1;
 453         err = nf_nat_l3proto_register(&nf_nat_l3proto_ipv6);
 454         if (err < 0)
 455                 goto err2;
 456         return err;
 457
 458 err2:
 459         nf_nat_l4proto_unregister(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
 460 err1:
 461         return err;
 462 }
 463
 464 static void __exit nf_nat_l3proto_ipv6_exit(void)
 465 {
 466         nf_nat_l3proto_unregister(&nf_nat_l3proto_ipv6);
 467         nf_nat_l4proto_unregister(NFPROTO_IPV6, &nf_nat_l4proto_icmpv6);
 468 }
 469
 470 MODULE_LICENSE("GPL");
 471 MODULE_ALIAS("nf-nat-" __stringify(AF_INET6));
 472
 473 module_init(nf_nat_l3proto_ipv6_init);
 474 module_exit(nf_nat_l3proto_ipv6_exit);