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         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 131                 const struct iphdr *iph = ip_hdr(skb);
 132
 133                 skb->ip_summed = CHECKSUM_PARTIAL;
 134                 skb->csum_start = skb_headroom(skb) + skb_network_offset(skb) +
 135                         ip_hdrlen(skb);
 136                 skb->csum_offset = (void *)check - data;
 137                 *check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, datalen,
 138                                             proto, 0);
 139         } else
 140                 inet_proto_csum_replace2(check, skb,
 141                                          htons(oldlen), htons(datalen), true);
 142 }
 143
 144 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 145 static int nf_nat_ipv4_nlattr_to_range(struct nlattr *tb[],
 146                                        struct nf_nat_range *range)
 147 {
 148         if (tb[CTA_NAT_V4_MINIP]) {
 149                 range->min_addr.ip = nla_get_be32(tb[CTA_NAT_V4_MINIP]);
 150                 range->flags |= NF_NAT_RANGE_MAP_IPS;
 151         }
 152
 153         if (tb[CTA_NAT_V4_MAXIP])
 154                 range->max_addr.ip = nla_get_be32(tb[CTA_NAT_V4_MAXIP]);
 155         else
 156                 range->max_addr.ip = range->min_addr.ip;
 157
 158         return 0;
 159 }
 160 #endif
 161
 162 static const struct nf_nat_l3proto nf_nat_l3proto_ipv4 = {
 163         .l3proto                = NFPROTO_IPV4,
 164         .in_range               = nf_nat_ipv4_in_range,
 165         .secure_port            = nf_nat_ipv4_secure_port,
 166         .manip_pkt              = nf_nat_ipv4_manip_pkt,
 167         .csum_update            = nf_nat_ipv4_csum_update,
 168         .csum_recalc            = nf_nat_ipv4_csum_recalc,
 169 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
 170         .nlattr_to_range        = nf_nat_ipv4_nlattr_to_range,
 171 #endif
 172 #ifdef CONFIG_XFRM
 173         .decode_session         = nf_nat_ipv4_decode_session,
 174 #endif
 175 };
 176
 177 int nf_nat_icmp_reply_translation(struct sk_buff *skb,
 178                                   struct nf_conn *ct,
 179                                   enum ip_conntrack_info ctinfo,
 180                                   unsigned int hooknum)
 181 {
 182         struct {
 183                 struct icmphdr  icmp;
 184                 struct iphdr    ip;
 185         } *inside;
 186         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 187         enum nf_nat_manip_type manip = HOOK2MANIP(hooknum);
 188         unsigned int hdrlen = ip_hdrlen(skb);
 189         const struct nf_nat_l4proto *l4proto;
 190         struct nf_conntrack_tuple target;
 191         unsigned long statusbit;
 192
 193         NF_CT_ASSERT(ctinfo == IP_CT_RELATED || ctinfo == IP_CT_RELATED_REPLY);
 194
 195         if (!skb_make_writable(skb, hdrlen + sizeof(*inside)))
 196                 return 0;
 197         if (nf_ip_checksum(skb, hooknum, hdrlen, 0))
 198                 return 0;
 199
 200         inside = (void *)skb->data + hdrlen;
 201         if (inside->icmp.type == ICMP_REDIRECT) {
 202                 if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK)
 203                         return 0;
 204                 if (ct->status & IPS_NAT_MASK)
 205                         return 0;
 206         }
 207
 208         if (manip == NF_NAT_MANIP_SRC)
 209                 statusbit = IPS_SRC_NAT;
 210         else
 211                 statusbit = IPS_DST_NAT;
 212
 213         /* Invert if this is reply direction */
 214         if (dir == IP_CT_DIR_REPLY)
 215                 statusbit ^= IPS_NAT_MASK;
 216
 217         if (!(ct->status & statusbit))
 218                 return 1;
 219
 220         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV4, inside->ip.protocol);
 221         if (!nf_nat_ipv4_manip_pkt(skb, hdrlen + sizeof(inside->icmp),
 222                                    l4proto, &ct->tuplehash[!dir].tuple, !manip))
 223                 return 0;
 224
 225         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 226                 /* Reloading "inside" here since manip_pkt may reallocate */
 227                 inside = (void *)skb->data + hdrlen;
 228                 inside->icmp.checksum = 0;
 229                 inside->icmp.checksum =
 230                         csum_fold(skb_checksum(skb, hdrlen,
 231                                                skb->len - hdrlen, 0));
 232         }
 233
 234         /* Change outer to look like the reply to an incoming packet */
 235         nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple);
 236         l4proto = __nf_nat_l4proto_find(NFPROTO_IPV4, 0);
 237         if (!nf_nat_ipv4_manip_pkt(skb, 0, l4proto, &target, manip))
 238                 return 0;
 239
 240         return 1;
 241 }
 242 EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation);
 243
 244 unsigned int
 245 nf_nat_ipv4_fn(void *priv, struct sk_buff *skb,
 246                const struct nf_hook_state *state,
 247                unsigned int (*do_chain)(void *priv,
 248                                         struct sk_buff *skb,
 249                                         const struct nf_hook_state *state,
 250                                         struct nf_conn *ct))
 251 {
 252         struct nf_conn *ct;
 253         enum ip_conntrack_info ctinfo;
 254         struct nf_conn_nat *nat;
 255         /* maniptype == SRC for postrouting. */
 256         enum nf_nat_manip_type maniptype = HOOK2MANIP(state->hook);
 257
 258         /* We never see fragments: conntrack defrags on pre-routing
 259          * and local-out, and nf_nat_out protects post-routing.
 260          */
 261         NF_CT_ASSERT(!ip_is_fragment(ip_hdr(skb)));
 262
 263         ct = nf_ct_get(skb, &ctinfo);
 264         /* Can't track?  It's not due to stress, or conntrack would
 265          * have dropped it.  Hence it's the user's responsibilty to
 266          * packet filter it out, or implement conntrack/NAT for that
 267          * protocol. 8) --RR
 268          */
 269         if (!ct)
 270                 return NF_ACCEPT;
 271
 272         /* Don't try to NAT if this packet is not conntracked */
 273         if (nf_ct_is_untracked(ct))
 274                 return NF_ACCEPT;
 275
 276         nat = nf_ct_nat_ext_add(ct);
 277         if (nat == NULL)
 278                 return NF_ACCEPT;
 279
 280         switch (ctinfo) {
 281         case IP_CT_RELATED:
 282         case IP_CT_RELATED_REPLY:
 283                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 284                         if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
 285                                                            state->hook))
 286                                 return NF_DROP;
 287                         else
 288                                 return NF_ACCEPT;
 289                 }
 290                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 291         case IP_CT_NEW:
 292                 /* Seen it before?  This can happen for loopback, retrans,
 293                  * or local packets.
 294                  */
 295                 if (!nf_nat_initialized(ct, maniptype)) {
 296                         unsigned int ret;
 297
 298                         ret = do_chain(priv, skb, state, ct);
 299                         if (ret != NF_ACCEPT)
 300                                 return ret;
 301
 302                         if (nf_nat_initialized(ct, HOOK2MANIP(state->hook)))
 303                                 break;
 304
 305                         ret = nf_nat_alloc_null_binding(ct, state->hook);
 306                         if (ret != NF_ACCEPT)
 307                                 return ret;
 308                 } else {
 309                         pr_debug("Already setup manip %s for ct %p\n",
 310                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 311                                  ct);
 312                         if (nf_nat_oif_changed(state->hook, ctinfo, nat,
 313                                                state->out))
 314                                 goto oif_changed;
 315                 }
 316                 break;
 317
 318         default:
 319                 /* ESTABLISHED */
 320                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 321                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 322                 if (nf_nat_oif_changed(state->hook, ctinfo, nat, state->out))
 323                         goto oif_changed;
 324         }
 325
 326         return nf_nat_packet(ct, ctinfo, state->hook, skb);
 327
 328 oif_changed:
 329         nf_ct_kill_acct(ct, ctinfo, skb);
 330         return NF_DROP;
 331 }
 332 EXPORT_SYMBOL_GPL(nf_nat_ipv4_fn);
 333
 334 unsigned int
 335 nf_nat_ipv4_in(void *priv, struct sk_buff *skb,
 336                const struct nf_hook_state *state,
 337                unsigned int (*do_chain)(void *priv,
 338                                          struct sk_buff *skb,
 339                                          const struct nf_hook_state *state,
 340                                          struct nf_conn *ct))
 341 {
 342         unsigned int ret;
 343         __be32 daddr = ip_hdr(skb)->daddr;
 344
 345         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 346         if (ret != NF_DROP && ret != NF_STOLEN &&
 347             daddr != ip_hdr(skb)->daddr)
 348                 skb_dst_drop(skb);
 349
 350         return ret;
 351 }
 352 EXPORT_SYMBOL_GPL(nf_nat_ipv4_in);
 353
 354 unsigned int
 355 nf_nat_ipv4_out(void *priv, struct sk_buff *skb,
 356                 const struct nf_hook_state *state,
 357                 unsigned int (*do_chain)(void *priv,
 358                                           struct sk_buff *skb,
 359                                           const struct nf_hook_state *state,
 360                                           struct nf_conn *ct))
 361 {
 362 #ifdef CONFIG_XFRM
 363         const struct nf_conn *ct;
 364         enum ip_conntrack_info ctinfo;
 365         int err;
 366 #endif
 367         unsigned int ret;
 368
 369         /* root is playing with raw sockets. */
 370         if (skb->len < sizeof(struct iphdr) ||
 371             ip_hdrlen(skb) < sizeof(struct iphdr))
 372                 return NF_ACCEPT;
 373
 374         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 375 #ifdef CONFIG_XFRM
 376         if (ret != NF_DROP && ret != NF_STOLEN &&
 377             !(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
 378             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 379                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 380
 381                 if ((ct->tuplehash[dir].tuple.src.u3.ip !=
 382                      ct->tuplehash[!dir].tuple.dst.u3.ip) ||
 383                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMP &&
 384                      ct->tuplehash[dir].tuple.src.u.all !=
 385                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 386                         err = nf_xfrm_me_harder(state->net, skb, AF_INET);
 387                         if (err < 0)
 388                                 ret = NF_DROP_ERR(err);
 389                 }
 390         }
 391 #endif
 392         return ret;
 393 }
 394 EXPORT_SYMBOL_GPL(nf_nat_ipv4_out);
 395
 396 unsigned int
 397 nf_nat_ipv4_local_fn(void *priv, struct sk_buff *skb,
 398                      const struct nf_hook_state *state,
 399                      unsigned int (*do_chain)(void *priv,
 400                                                struct sk_buff *skb,
 401                                                const struct nf_hook_state *state,
 402                                                struct nf_conn *ct))
 403 {
 404         const struct nf_conn *ct;
 405         enum ip_conntrack_info ctinfo;
 406         unsigned int ret;
 407         int err;
 408
 409         /* root is playing with raw sockets. */
 410         if (skb->len < sizeof(struct iphdr) ||
 411             ip_hdrlen(skb) < sizeof(struct iphdr))
 412                 return NF_ACCEPT;
 413
 414         ret = nf_nat_ipv4_fn(priv, skb, state, do_chain);
 415         if (ret != NF_DROP && ret != NF_STOLEN &&
 416             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 417                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 418
 419                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 420                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 421                         err = ip_route_me_harder(state->net, skb, RTN_UNSPEC);
 422                         if (err < 0)
 423                                 ret = NF_DROP_ERR(err);
 424                 }
 425 #ifdef CONFIG_XFRM
 426                 else if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
 427                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMP &&
 428                          ct->tuplehash[dir].tuple.dst.u.all !=
 429                          ct->tuplehash[!dir].tuple.src.u.all) {
 430                         err = nf_xfrm_me_harder(state->net, skb, AF_INET);
 431                         if (err < 0)
 432                                 ret = NF_DROP_ERR(err);
 433                 }
 434 #endif
 435         }
 436         return ret;
 437 }
 438 EXPORT_SYMBOL_GPL(nf_nat_ipv4_local_fn);
 439
 440 static int __init nf_nat_l3proto_ipv4_init(void)
 441 {
 442         int err;
 443
 444         err = nf_nat_l4proto_register(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 445         if (err < 0)
 446                 goto err1;
 447         err = nf_nat_l3proto_register(&nf_nat_l3proto_ipv4);
 448         if (err < 0)
 449                 goto err2;
 450         return err;
 451
 452 err2:
 453         nf_nat_l4proto_unregister(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 454 err1:
 455         return err;
 456 }
 457
 458 static void __exit nf_nat_l3proto_ipv4_exit(void)
 459 {
 460         nf_nat_l3proto_unregister(&nf_nat_l3proto_ipv4);
 461         nf_nat_l4proto_unregister(NFPROTO_IPV4, &nf_nat_l4proto_icmp);
 462 }
 463
 464 MODULE_LICENSE("GPL");
 465 MODULE_ALIAS("nf-nat-" __stringify(AF_INET));
 466
 467 module_init(nf_nat_l3proto_ipv4_init);
 468 module_exit(nf_nat_l3proto_ipv4_exit);