Patrick McHardy | 9651851 | 2013-10-14 11:00:02 +0200 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (c) 2008 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 this code funded by Astaro AG (http://www.astaro.com/) |
| 9 | */ |
| 10 | |
| 11 | #include <linux/module.h> |
| 12 | #include <linux/init.h> |
| 13 | #include <linux/list.h> |
| 14 | #include <linux/skbuff.h> |
| 15 | #include <linux/ip.h> |
| 16 | #include <linux/netlink.h> |
| 17 | #include <linux/netfilter.h> |
| 18 | #include <linux/netfilter_ipv4.h> |
| 19 | #include <linux/netfilter/nfnetlink.h> |
| 20 | #include <linux/netfilter/nf_tables.h> |
| 21 | #include <net/netfilter/nf_conntrack.h> |
| 22 | #include <net/netfilter/nf_nat.h> |
| 23 | #include <net/netfilter/nf_nat_core.h> |
| 24 | #include <net/netfilter/nf_tables.h> |
| 25 | #include <net/netfilter/nf_nat_l3proto.h> |
| 26 | #include <net/ip.h> |
| 27 | |
| 28 | struct nft_nat { |
| 29 | enum nft_registers sreg_addr_min:8; |
| 30 | enum nft_registers sreg_addr_max:8; |
| 31 | enum nft_registers sreg_proto_min:8; |
| 32 | enum nft_registers sreg_proto_max:8; |
| 33 | enum nf_nat_manip_type type; |
| 34 | }; |
| 35 | |
| 36 | static void nft_nat_eval(const struct nft_expr *expr, |
| 37 | struct nft_data data[NFT_REG_MAX + 1], |
| 38 | const struct nft_pktinfo *pkt) |
| 39 | { |
| 40 | const struct nft_nat *priv = nft_expr_priv(expr); |
| 41 | enum ip_conntrack_info ctinfo; |
| 42 | struct nf_conn *ct = nf_ct_get(pkt->skb, &ctinfo); |
| 43 | struct nf_nat_range range; |
| 44 | |
| 45 | memset(&range, 0, sizeof(range)); |
| 46 | if (priv->sreg_addr_min) { |
| 47 | range.min_addr.ip = data[priv->sreg_addr_min].data[0]; |
| 48 | range.max_addr.ip = data[priv->sreg_addr_max].data[0]; |
| 49 | range.flags |= NF_NAT_RANGE_MAP_IPS; |
| 50 | } |
| 51 | |
| 52 | if (priv->sreg_proto_min) { |
| 53 | range.min_proto.all = data[priv->sreg_proto_min].data[0]; |
| 54 | range.max_proto.all = data[priv->sreg_proto_max].data[0]; |
| 55 | range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; |
| 56 | } |
| 57 | |
| 58 | data[NFT_REG_VERDICT].verdict = |
| 59 | nf_nat_setup_info(ct, &range, priv->type); |
| 60 | } |
| 61 | |
| 62 | static const struct nla_policy nft_nat_policy[NFTA_NAT_MAX + 1] = { |
| 63 | [NFTA_NAT_ADDR_MIN] = { .type = NLA_U32 }, |
| 64 | [NFTA_NAT_ADDR_MAX] = { .type = NLA_U32 }, |
| 65 | [NFTA_NAT_PROTO_MIN] = { .type = NLA_U32 }, |
| 66 | [NFTA_NAT_PROTO_MAX] = { .type = NLA_U32 }, |
| 67 | [NFTA_NAT_TYPE] = { .type = NLA_U32 }, |
| 68 | }; |
| 69 | |
| 70 | static int nft_nat_init(const struct nft_ctx *ctx, const struct nft_expr *expr, |
| 71 | const struct nlattr * const tb[]) |
| 72 | { |
| 73 | struct nft_nat *priv = nft_expr_priv(expr); |
| 74 | int err; |
| 75 | |
| 76 | if (tb[NFTA_NAT_TYPE] == NULL) |
| 77 | return -EINVAL; |
| 78 | |
| 79 | switch (ntohl(nla_get_be32(tb[NFTA_NAT_TYPE]))) { |
| 80 | case NFT_NAT_SNAT: |
| 81 | priv->type = NF_NAT_MANIP_SRC; |
| 82 | break; |
| 83 | case NFT_NAT_DNAT: |
| 84 | priv->type = NF_NAT_MANIP_DST; |
| 85 | break; |
| 86 | default: |
| 87 | return -EINVAL; |
| 88 | } |
| 89 | |
| 90 | if (tb[NFTA_NAT_ADDR_MIN]) { |
| 91 | priv->sreg_addr_min = ntohl(nla_get_be32(tb[NFTA_NAT_ADDR_MIN])); |
| 92 | err = nft_validate_input_register(priv->sreg_addr_min); |
| 93 | if (err < 0) |
| 94 | return err; |
| 95 | } |
| 96 | |
| 97 | if (tb[NFTA_NAT_ADDR_MAX]) { |
| 98 | priv->sreg_addr_max = ntohl(nla_get_be32(tb[NFTA_NAT_ADDR_MAX])); |
| 99 | err = nft_validate_input_register(priv->sreg_addr_max); |
| 100 | if (err < 0) |
| 101 | return err; |
| 102 | } else |
| 103 | priv->sreg_addr_max = priv->sreg_addr_min; |
| 104 | |
| 105 | if (tb[NFTA_NAT_PROTO_MIN]) { |
| 106 | priv->sreg_proto_min = ntohl(nla_get_be32(tb[NFTA_NAT_PROTO_MIN])); |
| 107 | err = nft_validate_input_register(priv->sreg_proto_min); |
| 108 | if (err < 0) |
| 109 | return err; |
| 110 | } |
| 111 | |
| 112 | if (tb[NFTA_NAT_PROTO_MAX]) { |
| 113 | priv->sreg_proto_max = ntohl(nla_get_be32(tb[NFTA_NAT_PROTO_MAX])); |
| 114 | err = nft_validate_input_register(priv->sreg_proto_max); |
| 115 | if (err < 0) |
| 116 | return err; |
| 117 | } else |
| 118 | priv->sreg_proto_max = priv->sreg_proto_min; |
| 119 | |
| 120 | return 0; |
| 121 | } |
| 122 | |
| 123 | static int nft_nat_dump(struct sk_buff *skb, const struct nft_expr *expr) |
| 124 | { |
| 125 | const struct nft_nat *priv = nft_expr_priv(expr); |
| 126 | |
| 127 | switch (priv->type) { |
| 128 | case NF_NAT_MANIP_SRC: |
| 129 | if (nla_put_be32(skb, NFTA_NAT_TYPE, htonl(NFT_NAT_SNAT))) |
| 130 | goto nla_put_failure; |
| 131 | break; |
| 132 | case NF_NAT_MANIP_DST: |
| 133 | if (nla_put_be32(skb, NFTA_NAT_TYPE, htonl(NFT_NAT_DNAT))) |
| 134 | goto nla_put_failure; |
| 135 | break; |
| 136 | } |
| 137 | |
| 138 | if (nla_put_be32(skb, NFTA_NAT_ADDR_MIN, htonl(priv->sreg_addr_min))) |
| 139 | goto nla_put_failure; |
| 140 | if (nla_put_be32(skb, NFTA_NAT_ADDR_MAX, htonl(priv->sreg_addr_max))) |
| 141 | goto nla_put_failure; |
| 142 | if (nla_put_be32(skb, NFTA_NAT_PROTO_MIN, htonl(priv->sreg_proto_min))) |
| 143 | goto nla_put_failure; |
| 144 | if (nla_put_be32(skb, NFTA_NAT_PROTO_MAX, htonl(priv->sreg_proto_max))) |
| 145 | goto nla_put_failure; |
| 146 | return 0; |
| 147 | |
| 148 | nla_put_failure: |
| 149 | return -1; |
| 150 | } |
| 151 | |
| 152 | static struct nft_expr_ops nft_nat_ops __read_mostly = { |
| 153 | .name = "nat", |
| 154 | .size = NFT_EXPR_SIZE(sizeof(struct nft_nat)), |
| 155 | .owner = THIS_MODULE, |
| 156 | .eval = nft_nat_eval, |
| 157 | .init = nft_nat_init, |
| 158 | .dump = nft_nat_dump, |
| 159 | .policy = nft_nat_policy, |
| 160 | .maxattr = NFTA_NAT_MAX, |
| 161 | }; |
| 162 | |
| 163 | /* |
| 164 | * NAT table |
| 165 | */ |
| 166 | |
| 167 | static unsigned int nf_nat_fn(const struct nf_hook_ops *ops, |
| 168 | struct sk_buff *skb, |
| 169 | const struct net_device *in, |
| 170 | const struct net_device *out, |
| 171 | int (*okfn)(struct sk_buff *)) |
| 172 | { |
| 173 | enum ip_conntrack_info ctinfo; |
| 174 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
| 175 | struct nf_conn_nat *nat; |
| 176 | enum nf_nat_manip_type maniptype = HOOK2MANIP(ops->hooknum); |
| 177 | unsigned int ret; |
| 178 | |
| 179 | if (ct == NULL || nf_ct_is_untracked(ct)) |
| 180 | return NF_ACCEPT; |
| 181 | |
| 182 | NF_CT_ASSERT(!(ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET))); |
| 183 | |
| 184 | nat = nfct_nat(ct); |
| 185 | if (nat == NULL) { |
| 186 | /* Conntrack module was loaded late, can't add extension. */ |
| 187 | if (nf_ct_is_confirmed(ct)) |
| 188 | return NF_ACCEPT; |
| 189 | nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC); |
| 190 | if (nat == NULL) |
| 191 | return NF_ACCEPT; |
| 192 | } |
| 193 | |
| 194 | switch (ctinfo) { |
| 195 | case IP_CT_RELATED: |
| 196 | case IP_CT_RELATED + IP_CT_IS_REPLY: |
| 197 | if (ip_hdr(skb)->protocol == IPPROTO_ICMP) { |
| 198 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, |
| 199 | ops->hooknum)) |
| 200 | return NF_DROP; |
| 201 | else |
| 202 | return NF_ACCEPT; |
| 203 | } |
| 204 | /* Fall through */ |
| 205 | case IP_CT_NEW: |
| 206 | if (nf_nat_initialized(ct, maniptype)) |
| 207 | break; |
| 208 | |
| 209 | ret = nft_do_chain(ops, skb, in, out, okfn); |
| 210 | if (ret != NF_ACCEPT) |
| 211 | return ret; |
| 212 | if (!nf_nat_initialized(ct, maniptype)) { |
| 213 | ret = nf_nat_alloc_null_binding(ct, ops->hooknum); |
| 214 | if (ret != NF_ACCEPT) |
| 215 | return ret; |
| 216 | } |
| 217 | default: |
| 218 | break; |
| 219 | } |
| 220 | |
| 221 | return nf_nat_packet(ct, ctinfo, ops->hooknum, skb); |
| 222 | } |
| 223 | |
| 224 | static unsigned int nf_nat_prerouting(const struct nf_hook_ops *ops, |
| 225 | struct sk_buff *skb, |
| 226 | const struct net_device *in, |
| 227 | const struct net_device *out, |
| 228 | int (*okfn)(struct sk_buff *)) |
| 229 | { |
| 230 | __be32 daddr = ip_hdr(skb)->daddr; |
| 231 | unsigned int ret; |
| 232 | |
| 233 | ret = nf_nat_fn(ops, skb, in, out, okfn); |
| 234 | if (ret != NF_DROP && ret != NF_STOLEN && |
| 235 | ip_hdr(skb)->daddr != daddr) { |
| 236 | skb_dst_drop(skb); |
| 237 | } |
| 238 | return ret; |
| 239 | } |
| 240 | |
| 241 | static unsigned int nf_nat_postrouting(const struct nf_hook_ops *ops, |
| 242 | struct sk_buff *skb, |
| 243 | const struct net_device *in, |
| 244 | const struct net_device *out, |
| 245 | int (*okfn)(struct sk_buff *)) |
| 246 | { |
| 247 | enum ip_conntrack_info ctinfo __maybe_unused; |
| 248 | const struct nf_conn *ct __maybe_unused; |
| 249 | unsigned int ret; |
| 250 | |
| 251 | ret = nf_nat_fn(ops, skb, in, out, okfn); |
| 252 | #ifdef CONFIG_XFRM |
| 253 | if (ret != NF_DROP && ret != NF_STOLEN && |
| 254 | (ct = nf_ct_get(skb, &ctinfo)) != NULL) { |
| 255 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
| 256 | |
| 257 | if (ct->tuplehash[dir].tuple.src.u3.ip != |
| 258 | ct->tuplehash[!dir].tuple.dst.u3.ip || |
| 259 | ct->tuplehash[dir].tuple.src.u.all != |
| 260 | ct->tuplehash[!dir].tuple.dst.u.all) |
| 261 | return nf_xfrm_me_harder(skb, AF_INET) == 0 ? |
| 262 | ret : NF_DROP; |
| 263 | } |
| 264 | #endif |
| 265 | return ret; |
| 266 | } |
| 267 | |
| 268 | static unsigned int nf_nat_output(const struct nf_hook_ops *ops, |
| 269 | struct sk_buff *skb, |
| 270 | const struct net_device *in, |
| 271 | const struct net_device *out, |
| 272 | int (*okfn)(struct sk_buff *)) |
| 273 | { |
| 274 | enum ip_conntrack_info ctinfo; |
| 275 | const struct nf_conn *ct; |
| 276 | unsigned int ret; |
| 277 | |
| 278 | ret = nf_nat_fn(ops, skb, in, out, okfn); |
| 279 | if (ret != NF_DROP && ret != NF_STOLEN && |
| 280 | (ct = nf_ct_get(skb, &ctinfo)) != NULL) { |
| 281 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
| 282 | |
| 283 | if (ct->tuplehash[dir].tuple.dst.u3.ip != |
| 284 | ct->tuplehash[!dir].tuple.src.u3.ip) { |
| 285 | if (ip_route_me_harder(skb, RTN_UNSPEC)) |
| 286 | ret = NF_DROP; |
| 287 | } |
| 288 | #ifdef CONFIG_XFRM |
| 289 | else if (ct->tuplehash[dir].tuple.dst.u.all != |
| 290 | ct->tuplehash[!dir].tuple.src.u.all) |
| 291 | if (nf_xfrm_me_harder(skb, AF_INET)) |
| 292 | ret = NF_DROP; |
| 293 | #endif |
| 294 | } |
| 295 | return ret; |
| 296 | } |
| 297 | |
| 298 | static struct nft_base_chain nf_chain_nat_prerouting __read_mostly = { |
| 299 | .chain = { |
| 300 | .name = "PREROUTING", |
| 301 | .rules = LIST_HEAD_INIT(nf_chain_nat_prerouting.chain.rules), |
| 302 | .flags = NFT_BASE_CHAIN | NFT_CHAIN_BUILTIN, |
| 303 | }, |
| 304 | .ops = { |
| 305 | .hook = nf_nat_prerouting, |
| 306 | .owner = THIS_MODULE, |
| 307 | .pf = NFPROTO_IPV4, |
| 308 | .hooknum = NF_INET_PRE_ROUTING, |
| 309 | .priority = NF_IP_PRI_NAT_DST, |
| 310 | .priv = &nf_chain_nat_prerouting.chain, |
| 311 | }, |
| 312 | }; |
| 313 | |
| 314 | static struct nft_base_chain nf_chain_nat_postrouting __read_mostly = { |
| 315 | .chain = { |
| 316 | .name = "POSTROUTING", |
| 317 | .rules = LIST_HEAD_INIT(nf_chain_nat_postrouting.chain.rules), |
| 318 | .flags = NFT_BASE_CHAIN | NFT_CHAIN_BUILTIN, |
| 319 | }, |
| 320 | .ops = { |
| 321 | .hook = nf_nat_postrouting, |
| 322 | .owner = THIS_MODULE, |
| 323 | .pf = NFPROTO_IPV4, |
| 324 | .hooknum = NF_INET_POST_ROUTING, |
| 325 | .priority = NF_IP_PRI_NAT_SRC, |
| 326 | .priv = &nf_chain_nat_postrouting.chain, |
| 327 | }, |
| 328 | }; |
| 329 | |
| 330 | static struct nft_base_chain nf_chain_nat_output __read_mostly = { |
| 331 | .chain = { |
| 332 | .name = "OUTPUT", |
| 333 | .rules = LIST_HEAD_INIT(nf_chain_nat_output.chain.rules), |
| 334 | .flags = NFT_BASE_CHAIN | NFT_CHAIN_BUILTIN, |
| 335 | }, |
| 336 | .ops = { |
| 337 | .hook = nf_nat_output, |
| 338 | .owner = THIS_MODULE, |
| 339 | .pf = NFPROTO_IPV4, |
| 340 | .hooknum = NF_INET_LOCAL_OUT, |
| 341 | .priority = NF_IP_PRI_NAT_DST, |
| 342 | .priv = &nf_chain_nat_output.chain, |
| 343 | }, |
| 344 | }; |
| 345 | |
| 346 | static struct nft_base_chain nf_chain_nat_input __read_mostly = { |
| 347 | .chain = { |
| 348 | .name = "INPUT", |
| 349 | .rules = LIST_HEAD_INIT(nf_chain_nat_input.chain.rules), |
| 350 | .flags = NFT_BASE_CHAIN | NFT_CHAIN_BUILTIN, |
| 351 | }, |
| 352 | .ops = { |
| 353 | .hook = nf_nat_fn, |
| 354 | .owner = THIS_MODULE, |
| 355 | .pf = NFPROTO_IPV4, |
| 356 | .hooknum = NF_INET_LOCAL_IN, |
| 357 | .priority = NF_IP_PRI_NAT_SRC, |
| 358 | .priv = &nf_chain_nat_input.chain, |
| 359 | }, |
| 360 | }; |
| 361 | |
| 362 | |
| 363 | static struct nft_table nf_table_nat_ipv4 __read_mostly = { |
| 364 | .name = "nat", |
| 365 | .chains = LIST_HEAD_INIT(nf_table_nat_ipv4.chains), |
| 366 | }; |
| 367 | |
| 368 | static int __init nf_table_nat_init(void) |
| 369 | { |
| 370 | int err; |
| 371 | |
| 372 | list_add_tail(&nf_chain_nat_prerouting.chain.list, |
| 373 | &nf_table_nat_ipv4.chains); |
| 374 | list_add_tail(&nf_chain_nat_postrouting.chain.list, |
| 375 | &nf_table_nat_ipv4.chains); |
| 376 | list_add_tail(&nf_chain_nat_output.chain.list, |
| 377 | &nf_table_nat_ipv4.chains); |
| 378 | list_add_tail(&nf_chain_nat_input.chain.list, |
| 379 | &nf_table_nat_ipv4.chains); |
| 380 | |
| 381 | err = nft_register_table(&nf_table_nat_ipv4, NFPROTO_IPV4); |
| 382 | if (err < 0) |
| 383 | goto err1; |
| 384 | |
| 385 | err = nft_register_expr(&nft_nat_ops); |
| 386 | if (err < 0) |
| 387 | goto err2; |
| 388 | |
| 389 | return 0; |
| 390 | |
| 391 | err2: |
| 392 | nft_unregister_table(&nf_table_nat_ipv4, NFPROTO_IPV4); |
| 393 | err1: |
| 394 | return err; |
| 395 | } |
| 396 | |
| 397 | static void __exit nf_table_nat_exit(void) |
| 398 | { |
| 399 | nft_unregister_expr(&nft_nat_ops); |
| 400 | nft_unregister_table(&nf_table_nat_ipv4, AF_INET); |
| 401 | } |
| 402 | |
| 403 | module_init(nf_table_nat_init); |
| 404 | module_exit(nf_table_nat_exit); |
| 405 | |
| 406 | MODULE_LICENSE("GPL"); |
| 407 | MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); |
| 408 | MODULE_ALIAS_NFT_TABLE(AF_INET, "nat"); |
| 409 | MODULE_ALIAS_NFT_EXPR("nat"); |