net/ipv4/netfilter/nf_nat_l3proto_ipv4.c

   1 /*
   2  * (C) 1999-2001 Paul `Rusty' Russell
   3  * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org>
   4  * (C) 2011 Patrick McHardy <kaber@trash.net>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10
  11 #include <linux/types.h>
  12 #include <linux/module.h>
  13 #include <linux/skbuff.h>
  14 #include <linux/ip.h>
  15 #include <linux/icmp.h>
  16 #include <linux/netfilter.h>
  17 #include <linux/netfilter_ipv4.h>
  18 #include <net/secure_seq.h>
  19 #include <net/checksum.h>
  20 #include <net/route.h>
  21 #include <net/ip.h>
  22
  23 #include <net/netfilter/nf_conntrack_core.h>
  24 #include <net/netfilter/nf_conntrack.h>
  25 #include <net/netfilter/nf_nat_core.h>
  26 #include <net/netfilter/nf_nat_l3proto.h>
  27 #include <net/netfilter/nf_nat_l4proto.h>
  28
  29 static const struct nf_nat_l3proto nf_nat_l3proto_ipv4;
  30
  31 #ifdef CONFIG_XFRM
  32 static void nf_nat_ipv4_decode_session(struct sk_buff *skb,
  33                                        const struct nf_conn *ct,
  34                                        enum ip_conntrack_dir dir,
  35                                        unsigned long statusbit,
  36                                        struct flowi *fl)
  37 {
  38         const struct nf_conntrack_tuple *t = &ct->tuplehash[dir].tuple;
  39         struct flowi4 *fl4 = &fl->u.ip4;
  40
  41         if (ct->status & statusbit) {
  42                 fl4->daddr = t->dst.u3.ip;
  43                 if (t->dst.protonum == IPPROTO_TCP ||
  44                     t->dst.protonum == IPPROTO_UDP ||
  45                     t->dst.protonum == IPPROTO_UDPLITE ||
  46                     t->dst.protonum == IPPROTO_DCCP ||
  47                     t->dst.protonum == IPPROTO_SCTP)
  48                         fl4->fl4_dport = t->dst.u.all;
  49         }
  50
  51         statusbit ^= IPS_NAT_MASK;
  52
  53         if (ct->status & statusbit) {
  54                 fl4->saddr = t->src.u3.ip;
  55                 if (t->dst.protonum == IPPROTO_TCP ||
  56                     t->dst.protonum == IPPROTO_UDP ||
  57                     t->dst.protonum == IPPROTO_UDPLITE ||
  58                     t->dst.protonum == IPPROTO_DCCP ||
  59                     t->dst.protonum == IPPROTO_SCTP)
  60                         fl4->fl4_sport = t->src.u.all;
  61         }
  62 }
  63 #endif /* CONFIG_XFRM */
  64
  65 static bool nf_nat_ipv4_in_range(const struct nf_conntrack_tuple *t,
  66                                  const struct nf_nat_range *range)
  67 {
  68         return ntohl(t->src.u3.ip) >= ntohl(range->min_addr.ip) &&
  69                ntohl(t->src.u3.ip) <= ntohl(range->max_addr.ip);
  70 }
  71
  72 static u32 nf_nat_ipv4_secure_port(const struct nf_conntrack_tuple *t,
  73                                    __be16 dport)
  74 {
  75         return secure_ipv4_port_ephemeral(t->src.u3.ip, t->dst.u3.ip, dport);
  76 }
  77
  78 static bool nf_nat_ipv4_manip_pkt(struct sk_buff *skb,
  79                                   unsigned int iphdroff,
  80                                   const struct nf_nat_l4proto *l4proto,
  81                                   const struct nf_conntrack_tuple *target,
  82                                   enum nf_nat_manip_type maniptype)
  83 {
  84         struct iphdr *iph;
  85         unsigned int hdroff;
  86
  87         if (!skb_make_writable(skb, iphdroff + sizeof(*iph)))
  88                 return false;
  89
  90         iph = (void *)skb->data + iphdroff;
  91         hdroff = iphdroff + iph->ihl * 4;
  92
  93         if (!l4proto->manip_pkt(skb, &nf_nat_l3proto_ipv4, iphdroff, hdroff,
  94                                 target, maniptype))
  95                 return false;
  96         iph = (void *)skb->data + iphdroff;
  97
  98         if (maniptype == NF_NAT_MANIP_SRC) {
  99                 csum_replace4(&iph->check, iph->saddr, target->src.u3.ip);
 100                 iph->saddr = target->src.u3.ip;
 101         } else {
 102                 csum_replace4(&iph->check, iph->daddr, target->dst.u3.ip);
 103                 iph->daddr = target->dst.u3.ip;
 104         }
 105         return true;
 106 }
 107
 108 static void nf_nat_ipv4_csum_update(struct sk_buff *skb,
 109                                     unsigned int iphdroff, __sum16 *check,
 110                                     const struct nf_conntrack_tuple *t,
 111                                     enum nf_nat_manip_type maniptype)
 112 {
 113         struct iphdr *iph = (struct iphdr *)(skb->data + iphdroff);
 114         __be32 oldip, newip;
 115
 116         if (maniptype == NF_NAT_MANIP_SRC) {
 117                 oldip = iph->saddr;
 118                 newip = t->src.u3.ip;
 119         } else {
 120                 oldip = iph->daddr;
 121                 newip = t->dst.u3.ip;
 122         }
 123         inet_proto_csum_replace4(check, skb, oldip, newip, true);
 124 }
 125
 126 static void nf_nat_ipv4_csum_recalc(struct sk_buff *skb,
 127                                     u8 proto, void *data, __sum16 *check,
 128                                     int datalen, int oldlen)
 129 {
 130         const struct iphdr *iph = ip_hdr(skb);
 131         struct rtable *rt = skb_rtable(skb);
 132
 133         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 134                 if (!(rt->rt_flags & RTCF_LOCAL) &&
 135                     (!skb->dev || skb->dev->features & NETIF_F_V4_CSUM)) {
 136                         skb->ip_summed = CHECKSUM_PARTIAL;
 137                         skb->csum_start = skb_headroom(skb) +
 138                                           skb_network_offset(skb) +
 139                                           ip_hdrlen(skb);
 140                         skb->csum_offset = (void *)check - data;
 141                         *check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
 142                                                     datalen, proto, 0);
 143                 } else {
 144                         *check = 0;
 145                         *check = csum_tcpudp_magic(iph->saddr, iph->daddr,
 146                                                    datalen, proto,
 147                                                    csum_partial(data, datalen,
 148                                                                 0));
 149                         if (proto == IPPROTO_UDP && !*check)
 150                                 *check = CSUM_MANGLED_0;
 151                 }
 152         } else
 153                 inet_proto_csum_replace2(check, skb,
 154                                          htons(oldlen), htons(datalen), true);
 155 }
 156
 157 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 158 static int nf_nat_ipv4_nlattr_to_range(struct nlattr *tb[],
 159                                        struct nf_nat_range *range)
 160 {
 161         if (tb[CTA_NAT_V4_MINIP]) {
 162                 range->min_addr.ip = nla_get_be32(tb[CTA_NAT_V4_MINIP]);
 163                 range->flags |= NF_NAT_RANGE_MAP_IPS;
 164         }
 165
 166         if (tb[CTA_NAT_V4_MAXIP])
 167                 range->max_addr.ip = nla_get_be32(tb[CTA_NAT_V4_MAXIP]);
 168         else
 169                 range->max_addr.ip = range->min_addr.ip;
 170
 171         return 0;
 172 }
 173 #endif
 174
 175 static const struct nf_nat_l3proto nf_nat_l3proto_ipv4 = {
 176         .l3proto                = NFPROTO_IPV4,
 177         .in_range               = nf_nat_ipv4_in_range,
 178         .secure_port            = nf_nat_ipv4_secure_port,
 179         .manip_pkt              = nf_nat_ipv4_manip_pkt,
 180         .csum_update            = nf_nat_ipv4_csum_update,
 181         .csum_recalc            = nf_nat_ipv4_csum_recalc,
 182 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 183         .nlattr_to_range        = nf_nat_ipv4_nlattr_to_range,
 184 #endif
 185 #ifdef CONFIG_XFRM
 186         .decode_session         = nf_nat_ipv4_decode_session,
 187 #endif
 188 };
 189
 190 int nf_nat_icmp_reply_translation(struct sk_buff *skb,
 191                                   struct nf_conn *ct,
 192                                   enum ip_conntrack_info ctinfo,
 193                                   unsigned int hooknum)
 194 {
 195         struct {
 196                 struct icmphdr  icmp;
 197                 struct iphdr    ip;
 198         } *inside;
 199         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 200         enum nf_nat_manip_type manip = HOOK2MANIP(hooknum);
 201         unsigned int hdrlen = ip_hdrlen(skb);
 202         const struct nf_nat_l4proto *l4proto;
 203         struct nf_conntrack_tuple target;
 204         unsigned long statusbit;
 205
 206         NF_CT_ASSERT(ctinfo == IP_CT_RELATED || ctinfo == IP_CT_RELATED_REPLY);
 207
 208         if (!skb_make_writable(skb, hdrlen + sizeof(*inside)))
 209                 return 0;
 210         if (nf_ip_checksum(skb, hooknum, hdrlen, 0))
 211                 return 0;
 212
 213         inside = (void *)skb->data + hdrlen;
 214         if (inside->icmp.type == ICMP_REDIRECT) {
 215                 if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK)
 216                         return 0;
 217                 if (ct->status & IPS_NAT_MASK)
 218                         return 0;
 219         }
 220
 221         if (manip == NF_NAT_MANIP_SRC)
 222                 statusbit = IPS_SRC_NAT;
 223         else
 224                 statusbit = IPS_DST_NAT;
 225
 226         /* Invert if this is reply direction */
 227         if (dir == IP_CT_DIR_REPLY)
 228                 statusbit ^= IPS_NAT_MASK;
 229
 230         if (!(ct->status & statusbit))
 231                 return 1;
 232
 233         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV4, inside->ip.protocol);
 234         if (!nf_nat_ipv4_manip_pkt(skb, hdrlen + sizeof(inside->icmp),
 235                                    l4proto, &ct->tuplehash[!dir].tuple, !manip))
 236                 return 0;
 237
 238         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 239                 /* Reloading "inside" here since manip_pkt may reallocate */
 240                 inside = (void *)skb->data + hdrlen;
 241                 inside->icmp.checksum = 0;
 242                 inside->icmp.checksum =
 243                         csum_fold(skb_checksum(skb, hdrlen,
 244                                                skb->len - hdrlen, 0));
 245         }
 246
 247         /* Change outer to look like the reply to an incoming packet */
 248         nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple);
 249         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV4, 0);
 250         if (!nf_nat_ipv4_manip_pkt(skb, 0, l4proto, &target, manip))
 251                 return 0;
 252
 253         return 1;
 254 }
 255 EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation);
 256
 257 unsigned int
 258 nf_nat_ipv4_fn(void *priv, struct sk_buff *skb,
 259                const struct nf_hook_state *state,
 260                unsigned int (*do_chain)(void *priv,
 261                                         struct sk_buff *skb,
 262                                         const struct nf_hook_state *state,
 263                                         struct nf_conn *ct))
 264 {
 265         struct nf_conn *ct;
 266         enum ip_conntrack_info ctinfo;
 267         struct nf_conn_nat *nat;
 268         /* maniptype == SRC for postrouting. */
 269         enum nf_nat_manip_type maniptype = HOOK2MANIP(state->hook);
 270
 271         /* We never see fragments: conntrack defrags on pre-routing
 272          * and local-out, and nf_nat_out protects post-routing.
 273          */
 274         NF_CT_ASSERT(!ip_is_fragment(ip_hdr(skb)));
 275
 276         ct = nf_ct_get(skb, &ctinfo);
 277         /* Can't track?  It's not due to stress, or conntrack would
 278          * have dropped it.  Hence it's the user's responsibilty to
 279          * packet filter it out, or implement conntrack/NAT for that
 280          * protocol. 8) --RR
 281          */
 282         if (!ct)
 283                 return NF_ACCEPT;
 284
 285         /* Don't try to NAT if this packet is not conntracked */
 286         if (nf_ct_is_untracked(ct))
 287                 return NF_ACCEPT;
 288
 289         nat = nf_ct_nat_ext_add(ct);
 290         if (nat == NULL)
 291                 return NF_ACCEPT;
 292
 293         switch (ctinfo) {
 294         case IP_CT_RELATED:
 295         case IP_CT_RELATED_REPLY:
 296                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 297                         if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
 298                                                            state->hook))
 299                                 return NF_DROP;
 300                         else
 301                                 return NF_ACCEPT;
 302                 }
 303                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 304         case IP_CT_NEW:
 305                 /* Seen it before?  This can happen for loopback, retrans,
 306                  * or local packets.
 307                  */
 308                 if (!nf_nat_initialized(ct, maniptype)) {
 309                         unsigned int ret;
 310
 311                         ret = do_chain(priv, skb, state, ct);
 312                         if (ret != NF_ACCEPT)
 313                                 return ret;
 314
 315                         if (nf_nat_initialized(ct, HOOK2MANIP(state->hook)))
 316                                 break;
 317
 318                         ret = nf_nat_alloc_null_binding(ct, state->hook);
 319                         if (ret != NF_ACCEPT)
 320                                 return ret;
 321                 } else {
 322                         pr_debug("Already setup manip %s for ct %p\n",
 323                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 324                                  ct);
 325                         if (nf_nat_oif_changed(state->hook, ctinfo, nat,
 326                                                state->out))
 327                                 goto oif_changed;
 328                 }
 329                 break;
 330
 331         default:
 332                 /* ESTABLISHED */
 333                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 334                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 335                 if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
 336                         goto oif_changed;
 337         }
 338
 339         return nf_nat_packet(ct, ctinfo, state->hook, skb);
 340
 341 oif_changed:
 342         nf_ct_kill_acct(ct, ctinfo, skb);
 343         return NF_DROP;
 344 }
 345 EXPORT_SYMBOL_GPL(nf_nat_ipv4_fn);
 346
 347 unsigned int
 348 nf_nat_ipv4_in(void *priv, struct sk_buff *skb,
 349                const struct nf_hook_state *state,
 350                unsigned int (*do_chain)(void *priv,
 351                                          struct sk_buff *skb,
 352                                          const struct nf_hook_state *state,
 353                                          struct nf_conn *ct))
 354 {
 355         unsigned int ret;
 356         __be32 daddr = ip_hdr(skb)->daddr;
 357
 358         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 359         if (ret != NF_DROP && ret != NF_STOLEN &&
 360             daddr != ip_hdr(skb)->daddr)
 361                 skb_dst_drop(skb);
 362
 363         return ret;
 364 }
 365 EXPORT_SYMBOL_GPL(nf_nat_ipv4_in);
 366
 367 unsigned int
 368 nf_nat_ipv4_out(void *priv, struct sk_buff *skb,
 369                 const struct nf_hook_state *state,
 370                 unsigned int (*do_chain)(void *priv,
 371                                           struct sk_buff *skb,
 372                                           const struct nf_hook_state *state,
 373                                           struct nf_conn *ct))
 374 {
 375 #ifdef CONFIG_XFRM
 376         const struct nf_conn *ct;
 377         enum ip_conntrack_info ctinfo;
 378         int err;
 379 #endif
 380         unsigned int ret;
 381
 382         /* root is playing with raw sockets. */
 383         if (skb->len < sizeof(struct iphdr) ||
 384             ip_hdrlen(skb) < sizeof(struct iphdr))
 385                 return NF_ACCEPT;
 386
 387         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 388 #ifdef CONFIG_XFRM
 389         if (ret != NF_DROP && ret != NF_STOLEN &&
 390             !(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
 391             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 392                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 393
 394                 if ((ct->tuplehash[dir].tuple.src.u3.ip !=
 395                      ct->tuplehash[!dir].tuple.dst.u3.ip) ||
 396                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMP &&
 397                      ct->tuplehash[dir].tuple.src.u.all !=
 398                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 399                         err = nf_xfrm_me_harder(state->net, skb, AF_INET);
 400                         if (err < 0)
 401                                 ret = NF_DROP_ERR(err);
 402                 }
 403         }
 404 #endif
 405         return ret;
 406 }
 407 EXPORT_SYMBOL_GPL(nf_nat_ipv4_out);
 408
 409 unsigned int
 410 nf_nat_ipv4_local_fn(void *priv, struct sk_buff *skb,
 411                      const struct nf_hook_state *state,
 412                      unsigned int (*do_chain)(void *priv,
 413                                                struct sk_buff *skb,
 414                                                const struct nf_hook_state *state,
 415                                                struct nf_conn *ct))
 416 {
 417         const struct nf_conn *ct;
 418         enum ip_conntrack_info ctinfo;
 419         unsigned int ret;
 420         int err;
 421
 422         /* root is playing with raw sockets. */
 423         if (skb->len < sizeof(struct iphdr) ||
 424             ip_hdrlen(skb) < sizeof(struct iphdr))
 425                 return NF_ACCEPT;
 426
 427         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 428         if (ret != NF_DROP && ret != NF_STOLEN &&
 429             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 430                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 431
 432                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 433                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 434                         err = ip_route_me_harder(skb, RTN_UNSPEC);
 435                         if (err < 0)
 436                                 ret = NF_DROP_ERR(err);
 437                 }
 438 #ifdef CONFIG_XFRM
 439                 else if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
 440                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMP &&
 441                          ct->tuplehash[dir].tuple.dst.u.all !=
 442                          ct->tuplehash[!dir].tuple.src.u.all) {
 443                         err = nf_xfrm_me_harder(state->net, skb, AF_INET);
 444                         if (err < 0)
 445                                 ret = NF_DROP_ERR(err);
 446                 }
 447 #endif
 448         }
 449         return ret;
 450 }
 451 EXPORT_SYMBOL_GPL(nf_nat_ipv4_local_fn);
 452
 453 static int __init nf_nat_l3proto_ipv4_init(void)
 454 {
 455         int err;
 456
 457         err = nf_nat_l4proto_register(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 458         if (err < 0)
 459                 goto err1;
 460         err = nf_nat_l3proto_register(&nf_nat_l3proto_ipv4);
 461         if (err < 0)
 462                 goto err2;
 463         return err;
 464
 465 err2:
 466         nf_nat_l4proto_unregister(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 467 err1:
 468         return err;
 469 }
 470
 471 static void __exit nf_nat_l3proto_ipv4_exit(void)
 472 {
 473         nf_nat_l3proto_unregister(&nf_nat_l3proto_ipv4);
 474         nf_nat_l4proto_unregister(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 475 }
 476
 477 MODULE_LICENSE("GPL");
 478 MODULE_ALIAS("nf-nat-" __stringify(AF_INET));
 479
 480 module_init(nf_nat_l3proto_ipv4_init);
 481 module_exit(nf_nat_l3proto_ipv4_exit);