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_manip_pkt(struct sk_buff *skb,
  65                                   unsigned int iphdroff,
  66                                   const struct nf_conntrack_tuple *target,
  67                                   enum nf_nat_manip_type maniptype)
  68 {
  69         struct ipv6hdr *ipv6h;
  70         __be16 frag_off;
  71         int hdroff;
  72         u8 nexthdr;
  73
  74         if (!skb_make_writable(skb, iphdroff + sizeof(*ipv6h)))
  75                 return false;
  76
  77         ipv6h = (void *)skb->data + iphdroff;
  78         nexthdr = ipv6h->nexthdr;
  79         hdroff = ipv6_skip_exthdr(skb, iphdroff + sizeof(*ipv6h),
  80                                   &nexthdr, &frag_off);
  81         if (hdroff < 0)
  82                 goto manip_addr;
  83
  84         if ((frag_off & htons(~0x7)) == 0 &&
  85             !nf_nat_l4proto_manip_pkt(skb, &nf_nat_l3proto_ipv6, iphdroff, hdroff,
  86                                       target, maniptype))
  87                 return false;
  88
  89         /* must reload, offset might have changed */
  90         ipv6h = (void *)skb->data + iphdroff;
  91
  92 manip_addr:
  93         if (maniptype == NF_NAT_MANIP_SRC)
  94                 ipv6h->saddr = target->src.u3.in6;
  95         else
  96                 ipv6h->daddr = target->dst.u3.in6;
  97
  98         return true;
  99 }
 100
 101 static void nf_nat_ipv6_csum_update(struct sk_buff *skb,
 102                                     unsigned int iphdroff, __sum16 *check,
 103                                     const struct nf_conntrack_tuple *t,
 104                                     enum nf_nat_manip_type maniptype)
 105 {
 106         const struct ipv6hdr *ipv6h = (struct ipv6hdr *)(skb->data + iphdroff);
 107         const struct in6_addr *oldip, *newip;
 108
 109         if (maniptype == NF_NAT_MANIP_SRC) {
 110                 oldip = &ipv6h->saddr;
 111                 newip = &t->src.u3.in6;
 112         } else {
 113                 oldip = &ipv6h->daddr;
 114                 newip = &t->dst.u3.in6;
 115         }
 116         inet_proto_csum_replace16(check, skb, oldip->s6_addr32,
 117                                   newip->s6_addr32, true);
 118 }
 119
 120 static void nf_nat_ipv6_csum_recalc(struct sk_buff *skb,
 121                                     u8 proto, void *data, __sum16 *check,
 122                                     int datalen, int oldlen)
 123 {
 124         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 125                 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 126
 127                 skb->ip_summed = CHECKSUM_PARTIAL;
 128                 skb->csum_start = skb_headroom(skb) + skb_network_offset(skb) +
 129                         (data - (void *)skb->data);
 130                 skb->csum_offset = (void *)check - data;
 131                 *check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
 132                                           datalen, proto, 0);
 133         } else
 134                 inet_proto_csum_replace2(check, skb,
 135                                          htons(oldlen), htons(datalen), true);
 136 }
 137
 138 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 139 static int nf_nat_ipv6_nlattr_to_range(struct nlattr *tb[],
 140                                        struct nf_nat_range2 *range)
 141 {
 142         if (tb[CTA_NAT_V6_MINIP]) {
 143                 nla_memcpy(&range->min_addr.ip6, tb[CTA_NAT_V6_MINIP],
 144                            sizeof(struct in6_addr));
 145                 range->flags |= NF_NAT_RANGE_MAP_IPS;
 146         }
 147
 148         if (tb[CTA_NAT_V6_MAXIP])
 149                 nla_memcpy(&range->max_addr.ip6, tb[CTA_NAT_V6_MAXIP],
 150                            sizeof(struct in6_addr));
 151         else
 152                 range->max_addr = range->min_addr;
 153
 154         return 0;
 155 }
 156 #endif
 157
 158 static const struct nf_nat_l3proto nf_nat_l3proto_ipv6 = {
 159         .l3proto                = NFPROTO_IPV6,
 160         .manip_pkt              = nf_nat_ipv6_manip_pkt,
 161         .csum_update            = nf_nat_ipv6_csum_update,
 162         .csum_recalc            = nf_nat_ipv6_csum_recalc,
 163 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 164         .nlattr_to_range        = nf_nat_ipv6_nlattr_to_range,
 165 #endif
 166 #ifdef CONFIG_XFRM
 167         .decode_session = nf_nat_ipv6_decode_session,
 168 #endif
 169 };
 170
 171 int nf_nat_icmpv6_reply_translation(struct sk_buff *skb,
 172                                     struct nf_conn *ct,
 173                                     enum ip_conntrack_info ctinfo,
 174                                     unsigned int hooknum,
 175                                     unsigned int hdrlen)
 176 {
 177         struct {
 178                 struct icmp6hdr icmp6;
 179                 struct ipv6hdr  ip6;
 180         } *inside;
 181         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 182         enum nf_nat_manip_type manip = HOOK2MANIP(hooknum);
 183         struct nf_conntrack_tuple target;
 184         unsigned long statusbit;
 185
 186         WARN_ON(ctinfo != IP_CT_RELATED && ctinfo != IP_CT_RELATED_REPLY);
 187
 188         if (!skb_make_writable(skb, hdrlen + sizeof(*inside)))
 189                 return 0;
 190         if (nf_ip6_checksum(skb, hooknum, hdrlen, IPPROTO_ICMPV6))
 191                 return 0;
 192
 193         inside = (void *)skb->data + hdrlen;
 194         if (inside->icmp6.icmp6_type == NDISC_REDIRECT) {
 195                 if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK)
 196                         return 0;
 197                 if (ct->status & IPS_NAT_MASK)
 198                         return 0;
 199         }
 200
 201         if (manip == NF_NAT_MANIP_SRC)
 202                 statusbit = IPS_SRC_NAT;
 203         else
 204                 statusbit = IPS_DST_NAT;
 205
 206         /* Invert if this is reply direction */
 207         if (dir == IP_CT_DIR_REPLY)
 208                 statusbit ^= IPS_NAT_MASK;
 209
 210         if (!(ct->status & statusbit))
 211                 return 1;
 212
 213         if (!nf_nat_ipv6_manip_pkt(skb, hdrlen + sizeof(inside->icmp6),
 214                                    &ct->tuplehash[!dir].tuple, !manip))
 215                 return 0;
 216
 217         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 218                 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 219                 inside = (void *)skb->data + hdrlen;
 220                 inside->icmp6.icmp6_cksum = 0;
 221                 inside->icmp6.icmp6_cksum =
 222                         csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
 223                                         skb->len - hdrlen, IPPROTO_ICMPV6,
 224                                         skb_checksum(skb, hdrlen,
 225                                                      skb->len - hdrlen, 0));
 226         }
 227
 228         nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple);
 229         target.dst.protonum = IPPROTO_ICMPV6;
 230         if (!nf_nat_ipv6_manip_pkt(skb, 0, &target, manip))
 231                 return 0;
 232
 233         return 1;
 234 }
 235 EXPORT_SYMBOL_GPL(nf_nat_icmpv6_reply_translation);
 236
 237 static unsigned int
 238 nf_nat_ipv6_fn(void *priv, struct sk_buff *skb,
 239                const struct nf_hook_state *state)
 240 {
 241         struct nf_conn *ct;
 242         enum ip_conntrack_info ctinfo;
 243         __be16 frag_off;
 244         int hdrlen;
 245         u8 nexthdr;
 246
 247         ct = nf_ct_get(skb, &ctinfo);
 248         /* Can't track?  It's not due to stress, or conntrack would
 249          * have dropped it.  Hence it's the user's responsibilty to
 250          * packet filter it out, or implement conntrack/NAT for that
 251          * protocol. 8) --RR
 252          */
 253         if (!ct)
 254                 return NF_ACCEPT;
 255
 256         if (ctinfo == IP_CT_RELATED || ctinfo == IP_CT_RELATED_REPLY) {
 257                 nexthdr = ipv6_hdr(skb)->nexthdr;
 258                 hdrlen = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
 259                                           &nexthdr, &frag_off);
 260
 261                 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
 262                         if (!nf_nat_icmpv6_reply_translation(skb, ct, ctinfo,
 263                                                              state->hook,
 264                                                              hdrlen))
 265                                 return NF_DROP;
 266                         else
 267                                 return NF_ACCEPT;
 268                 }
 269         }
 270
 271         return nf_nat_inet_fn(priv, skb, state);
 272 }
 273
 274 static unsigned int
 275 nf_nat_ipv6_in(void *priv, struct sk_buff *skb,
 276                const struct nf_hook_state *state)
 277 {
 278         unsigned int ret;
 279         struct in6_addr daddr = ipv6_hdr(skb)->daddr;
 280
 281         ret = nf_nat_ipv6_fn(priv, skb, state);
 282         if (ret != NF_DROP && ret != NF_STOLEN &&
 283             ipv6_addr_cmp(&daddr, &ipv6_hdr(skb)->daddr))
 284                 skb_dst_drop(skb);
 285
 286         return ret;
 287 }
 288
 289 static unsigned int
 290 nf_nat_ipv6_out(void *priv, struct sk_buff *skb,
 291                 const struct nf_hook_state *state)
 292 {
 293 #ifdef CONFIG_XFRM
 294         const struct nf_conn *ct;
 295         enum ip_conntrack_info ctinfo;
 296         int err;
 297 #endif
 298         unsigned int ret;
 299
 300         ret = nf_nat_ipv6_fn(priv, skb, state);
 301 #ifdef CONFIG_XFRM
 302         if (ret != NF_DROP && ret != NF_STOLEN &&
 303             !(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 304             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 305                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 306
 307                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.src.u3,
 308                                       &ct->tuplehash[!dir].tuple.dst.u3) ||
 309                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 310                      ct->tuplehash[dir].tuple.src.u.all !=
 311                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 312                         err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
 313                         if (err < 0)
 314                                 ret = NF_DROP_ERR(err);
 315                 }
 316         }
 317 #endif
 318         return ret;
 319 }
 320
 321 static unsigned int
 322 nf_nat_ipv6_local_fn(void *priv, struct sk_buff *skb,
 323                      const struct nf_hook_state *state)
 324 {
 325         const struct nf_conn *ct;
 326         enum ip_conntrack_info ctinfo;
 327         unsigned int ret;
 328         int err;
 329
 330         ret = nf_nat_ipv6_fn(priv, skb, state);
 331         if (ret != NF_DROP && ret != NF_STOLEN &&
 332             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 333                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 334
 335                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.dst.u3,
 336                                       &ct->tuplehash[!dir].tuple.src.u3)) {
 337                         err = ip6_route_me_harder(state->net, skb);
 338                         if (err < 0)
 339                                 ret = NF_DROP_ERR(err);
 340                 }
 341 #ifdef CONFIG_XFRM
 342                 else if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 343                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 344                          ct->tuplehash[dir].tuple.dst.u.all !=
 345                          ct->tuplehash[!dir].tuple.src.u.all) {
 346                         err = nf_xfrm_me_harder(state->net, skb, AF_INET6);
 347                         if (err < 0)
 348                                 ret = NF_DROP_ERR(err);
 349                 }
 350 #endif
 351         }
 352         return ret;
 353 }
 354
 355 static const struct nf_hook_ops nf_nat_ipv6_ops[] = {
 356         /* Before packet filtering, change destination */
 357         {
 358                 .hook           = nf_nat_ipv6_in,
 359                 .pf             = NFPROTO_IPV6,
 360                 .hooknum        = NF_INET_PRE_ROUTING,
 361                 .priority       = NF_IP6_PRI_NAT_DST,
 362         },
 363         /* After packet filtering, change source */
 364         {
 365                 .hook           = nf_nat_ipv6_out,
 366                 .pf             = NFPROTO_IPV6,
 367                 .hooknum        = NF_INET_POST_ROUTING,
 368                 .priority       = NF_IP6_PRI_NAT_SRC,
 369         },
 370         /* Before packet filtering, change destination */
 371         {
 372                 .hook           = nf_nat_ipv6_local_fn,
 373                 .pf             = NFPROTO_IPV6,
 374                 .hooknum        = NF_INET_LOCAL_OUT,
 375                 .priority       = NF_IP6_PRI_NAT_DST,
 376         },
 377         /* After packet filtering, change source */
 378         {
 379                 .hook           = nf_nat_ipv6_fn,
 380                 .pf             = NFPROTO_IPV6,
 381                 .hooknum        = NF_INET_LOCAL_IN,
 382                 .priority       = NF_IP6_PRI_NAT_SRC,
 383         },
 384 };
 385
 386 int nf_nat_l3proto_ipv6_register_fn(struct net *net, const struct nf_hook_ops *ops)
 387 {
 388         return nf_nat_register_fn(net, ops, nf_nat_ipv6_ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 389 }
 390 EXPORT_SYMBOL_GPL(nf_nat_l3proto_ipv6_register_fn);
 391
 392 void nf_nat_l3proto_ipv6_unregister_fn(struct net *net, const struct nf_hook_ops *ops)
 393 {
 394         nf_nat_unregister_fn(net, ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 395 }
 396 EXPORT_SYMBOL_GPL(nf_nat_l3proto_ipv6_unregister_fn);
 397
 398 static int __init nf_nat_l3proto_ipv6_init(void)
 399 {
 400         return nf_nat_l3proto_register(&nf_nat_l3proto_ipv6);
 401 }
 402
 403 static void __exit nf_nat_l3proto_ipv6_exit(void)
 404 {
 405         nf_nat_l3proto_unregister(&nf_nat_l3proto_ipv6);
 406 }
 407
 408 MODULE_LICENSE("GPL");
 409 MODULE_ALIAS("nf-nat-" __stringify(AF_INET6));
 410
 411 module_init(nf_nat_l3proto_ipv6_init);
 412 module_exit(nf_nat_l3proto_ipv6_exit);