| 1 | |
| 2 | /* (C) 1999-2001 Paul `Rusty' Russell |
| 3 | * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org> |
| 4 | * (C) 2006-2012 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/ip.h> |
| 13 | #include <linux/netfilter.h> |
| 14 | #include <linux/module.h> |
| 15 | #include <linux/skbuff.h> |
| 16 | #include <linux/icmp.h> |
| 17 | #include <linux/sysctl.h> |
| 18 | #include <net/route.h> |
| 19 | #include <net/ip.h> |
| 20 | |
| 21 | #include <linux/netfilter_ipv4.h> |
| 22 | #include <net/netfilter/nf_conntrack.h> |
| 23 | #include <net/netfilter/nf_conntrack_helper.h> |
| 24 | #include <net/netfilter/nf_conntrack_l4proto.h> |
| 25 | #include <net/netfilter/nf_conntrack_l3proto.h> |
| 26 | #include <net/netfilter/nf_conntrack_zones.h> |
| 27 | #include <net/netfilter/nf_conntrack_core.h> |
| 28 | #include <net/netfilter/nf_conntrack_seqadj.h> |
| 29 | #include <net/netfilter/ipv4/nf_conntrack_ipv4.h> |
| 30 | #include <net/netfilter/nf_nat_helper.h> |
| 31 | #include <net/netfilter/ipv4/nf_defrag_ipv4.h> |
| 32 | #include <net/netfilter/nf_log.h> |
| 33 | |
| 34 | static int conntrack4_net_id __read_mostly; |
| 35 | static DEFINE_MUTEX(register_ipv4_hooks); |
| 36 | |
| 37 | struct conntrack4_net { |
| 38 | unsigned int users; |
| 39 | }; |
| 40 | |
| 41 | static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff, |
| 42 | struct nf_conntrack_tuple *tuple) |
| 43 | { |
| 44 | const __be32 *ap; |
| 45 | __be32 _addrs[2]; |
| 46 | ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr), |
| 47 | sizeof(u_int32_t) * 2, _addrs); |
| 48 | if (ap == NULL) |
| 49 | return false; |
| 50 | |
| 51 | tuple->src.u3.ip = ap[0]; |
| 52 | tuple->dst.u3.ip = ap[1]; |
| 53 | |
| 54 | return true; |
| 55 | } |
| 56 | |
| 57 | static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple, |
| 58 | const struct nf_conntrack_tuple *orig) |
| 59 | { |
| 60 | tuple->src.u3.ip = orig->dst.u3.ip; |
| 61 | tuple->dst.u3.ip = orig->src.u3.ip; |
| 62 | |
| 63 | return true; |
| 64 | } |
| 65 | |
| 66 | static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff, |
| 67 | unsigned int *dataoff, u_int8_t *protonum) |
| 68 | { |
| 69 | const struct iphdr *iph; |
| 70 | struct iphdr _iph; |
| 71 | |
| 72 | iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph); |
| 73 | if (iph == NULL) |
| 74 | return -NF_ACCEPT; |
| 75 | |
| 76 | /* Conntrack defragments packets, we might still see fragments |
| 77 | * inside ICMP packets though. */ |
| 78 | if (iph->frag_off & htons(IP_OFFSET)) |
| 79 | return -NF_ACCEPT; |
| 80 | |
| 81 | *dataoff = nhoff + (iph->ihl << 2); |
| 82 | *protonum = iph->protocol; |
| 83 | |
| 84 | /* Check bogus IP headers */ |
| 85 | if (*dataoff > skb->len) { |
| 86 | pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: " |
| 87 | "nhoff %u, ihl %u, skblen %u\n", |
| 88 | nhoff, iph->ihl << 2, skb->len); |
| 89 | return -NF_ACCEPT; |
| 90 | } |
| 91 | |
| 92 | return NF_ACCEPT; |
| 93 | } |
| 94 | |
| 95 | static unsigned int ipv4_helper(void *priv, |
| 96 | struct sk_buff *skb, |
| 97 | const struct nf_hook_state *state) |
| 98 | { |
| 99 | struct nf_conn *ct; |
| 100 | enum ip_conntrack_info ctinfo; |
| 101 | const struct nf_conn_help *help; |
| 102 | const struct nf_conntrack_helper *helper; |
| 103 | |
| 104 | /* This is where we call the helper: as the packet goes out. */ |
| 105 | ct = nf_ct_get(skb, &ctinfo); |
| 106 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) |
| 107 | return NF_ACCEPT; |
| 108 | |
| 109 | help = nfct_help(ct); |
| 110 | if (!help) |
| 111 | return NF_ACCEPT; |
| 112 | |
| 113 | /* rcu_read_lock()ed by nf_hook_thresh */ |
| 114 | helper = rcu_dereference(help->helper); |
| 115 | if (!helper) |
| 116 | return NF_ACCEPT; |
| 117 | |
| 118 | return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb), |
| 119 | ct, ctinfo); |
| 120 | } |
| 121 | |
| 122 | static unsigned int ipv4_confirm(void *priv, |
| 123 | struct sk_buff *skb, |
| 124 | const struct nf_hook_state *state) |
| 125 | { |
| 126 | struct nf_conn *ct; |
| 127 | enum ip_conntrack_info ctinfo; |
| 128 | |
| 129 | ct = nf_ct_get(skb, &ctinfo); |
| 130 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) |
| 131 | goto out; |
| 132 | |
| 133 | /* adjust seqs for loopback traffic only in outgoing direction */ |
| 134 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) && |
| 135 | !nf_is_loopback_packet(skb)) { |
| 136 | if (!nf_ct_seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) { |
| 137 | NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop); |
| 138 | return NF_DROP; |
| 139 | } |
| 140 | } |
| 141 | out: |
| 142 | /* We've seen it coming out the other side: confirm it */ |
| 143 | return nf_conntrack_confirm(skb); |
| 144 | } |
| 145 | |
| 146 | static unsigned int ipv4_conntrack_in(void *priv, |
| 147 | struct sk_buff *skb, |
| 148 | const struct nf_hook_state *state) |
| 149 | { |
| 150 | return nf_conntrack_in(state->net, PF_INET, state->hook, skb); |
| 151 | } |
| 152 | |
| 153 | static unsigned int ipv4_conntrack_local(void *priv, |
| 154 | struct sk_buff *skb, |
| 155 | const struct nf_hook_state *state) |
| 156 | { |
| 157 | /* root is playing with raw sockets. */ |
| 158 | if (skb->len < sizeof(struct iphdr) || |
| 159 | ip_hdrlen(skb) < sizeof(struct iphdr)) |
| 160 | return NF_ACCEPT; |
| 161 | |
| 162 | if (ip_is_fragment(ip_hdr(skb))) { /* IP_NODEFRAG setsockopt set */ |
| 163 | enum ip_conntrack_info ctinfo; |
| 164 | struct nf_conn *tmpl; |
| 165 | |
| 166 | tmpl = nf_ct_get(skb, &ctinfo); |
| 167 | if (tmpl && nf_ct_is_template(tmpl)) { |
| 168 | /* when skipping ct, clear templates to avoid fooling |
| 169 | * later targets/matches |
| 170 | */ |
| 171 | skb->_nfct = 0; |
| 172 | nf_ct_put(tmpl); |
| 173 | } |
| 174 | return NF_ACCEPT; |
| 175 | } |
| 176 | |
| 177 | return nf_conntrack_in(state->net, PF_INET, state->hook, skb); |
| 178 | } |
| 179 | |
| 180 | /* Connection tracking may drop packets, but never alters them, so |
| 181 | make it the first hook. */ |
| 182 | static const struct nf_hook_ops ipv4_conntrack_ops[] = { |
| 183 | { |
| 184 | .hook = ipv4_conntrack_in, |
| 185 | .pf = NFPROTO_IPV4, |
| 186 | .hooknum = NF_INET_PRE_ROUTING, |
| 187 | .priority = NF_IP_PRI_CONNTRACK, |
| 188 | }, |
| 189 | { |
| 190 | .hook = ipv4_conntrack_local, |
| 191 | .pf = NFPROTO_IPV4, |
| 192 | .hooknum = NF_INET_LOCAL_OUT, |
| 193 | .priority = NF_IP_PRI_CONNTRACK, |
| 194 | }, |
| 195 | { |
| 196 | .hook = ipv4_helper, |
| 197 | .pf = NFPROTO_IPV4, |
| 198 | .hooknum = NF_INET_POST_ROUTING, |
| 199 | .priority = NF_IP_PRI_CONNTRACK_HELPER, |
| 200 | }, |
| 201 | { |
| 202 | .hook = ipv4_confirm, |
| 203 | .pf = NFPROTO_IPV4, |
| 204 | .hooknum = NF_INET_POST_ROUTING, |
| 205 | .priority = NF_IP_PRI_CONNTRACK_CONFIRM, |
| 206 | }, |
| 207 | { |
| 208 | .hook = ipv4_helper, |
| 209 | .pf = NFPROTO_IPV4, |
| 210 | .hooknum = NF_INET_LOCAL_IN, |
| 211 | .priority = NF_IP_PRI_CONNTRACK_HELPER, |
| 212 | }, |
| 213 | { |
| 214 | .hook = ipv4_confirm, |
| 215 | .pf = NFPROTO_IPV4, |
| 216 | .hooknum = NF_INET_LOCAL_IN, |
| 217 | .priority = NF_IP_PRI_CONNTRACK_CONFIRM, |
| 218 | }, |
| 219 | }; |
| 220 | |
| 221 | /* Fast function for those who don't want to parse /proc (and I don't |
| 222 | blame them). */ |
| 223 | /* Reversing the socket's dst/src point of view gives us the reply |
| 224 | mapping. */ |
| 225 | static int |
| 226 | getorigdst(struct sock *sk, int optval, void __user *user, int *len) |
| 227 | { |
| 228 | const struct inet_sock *inet = inet_sk(sk); |
| 229 | const struct nf_conntrack_tuple_hash *h; |
| 230 | struct nf_conntrack_tuple tuple; |
| 231 | |
| 232 | memset(&tuple, 0, sizeof(tuple)); |
| 233 | |
| 234 | lock_sock(sk); |
| 235 | tuple.src.u3.ip = inet->inet_rcv_saddr; |
| 236 | tuple.src.u.tcp.port = inet->inet_sport; |
| 237 | tuple.dst.u3.ip = inet->inet_daddr; |
| 238 | tuple.dst.u.tcp.port = inet->inet_dport; |
| 239 | tuple.src.l3num = PF_INET; |
| 240 | tuple.dst.protonum = sk->sk_protocol; |
| 241 | release_sock(sk); |
| 242 | |
| 243 | /* We only do TCP and SCTP at the moment: is there a better way? */ |
| 244 | if (tuple.dst.protonum != IPPROTO_TCP && |
| 245 | tuple.dst.protonum != IPPROTO_SCTP) { |
| 246 | pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n"); |
| 247 | return -ENOPROTOOPT; |
| 248 | } |
| 249 | |
| 250 | if ((unsigned int) *len < sizeof(struct sockaddr_in)) { |
| 251 | pr_debug("SO_ORIGINAL_DST: len %d not %zu\n", |
| 252 | *len, sizeof(struct sockaddr_in)); |
| 253 | return -EINVAL; |
| 254 | } |
| 255 | |
| 256 | h = nf_conntrack_find_get(sock_net(sk), &nf_ct_zone_dflt, &tuple); |
| 257 | if (h) { |
| 258 | struct sockaddr_in sin; |
| 259 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); |
| 260 | |
| 261 | sin.sin_family = AF_INET; |
| 262 | sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL] |
| 263 | .tuple.dst.u.tcp.port; |
| 264 | sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL] |
| 265 | .tuple.dst.u3.ip; |
| 266 | memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); |
| 267 | |
| 268 | pr_debug("SO_ORIGINAL_DST: %pI4 %u\n", |
| 269 | &sin.sin_addr.s_addr, ntohs(sin.sin_port)); |
| 270 | nf_ct_put(ct); |
| 271 | if (copy_to_user(user, &sin, sizeof(sin)) != 0) |
| 272 | return -EFAULT; |
| 273 | else |
| 274 | return 0; |
| 275 | } |
| 276 | pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n", |
| 277 | &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port), |
| 278 | &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port)); |
| 279 | return -ENOENT; |
| 280 | } |
| 281 | |
| 282 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
| 283 | |
| 284 | #include <linux/netfilter/nfnetlink.h> |
| 285 | #include <linux/netfilter/nfnetlink_conntrack.h> |
| 286 | |
| 287 | static int ipv4_tuple_to_nlattr(struct sk_buff *skb, |
| 288 | const struct nf_conntrack_tuple *tuple) |
| 289 | { |
| 290 | if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) || |
| 291 | nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip)) |
| 292 | goto nla_put_failure; |
| 293 | return 0; |
| 294 | |
| 295 | nla_put_failure: |
| 296 | return -1; |
| 297 | } |
| 298 | |
| 299 | static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = { |
| 300 | [CTA_IP_V4_SRC] = { .type = NLA_U32 }, |
| 301 | [CTA_IP_V4_DST] = { .type = NLA_U32 }, |
| 302 | }; |
| 303 | |
| 304 | static int ipv4_nlattr_to_tuple(struct nlattr *tb[], |
| 305 | struct nf_conntrack_tuple *t) |
| 306 | { |
| 307 | if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST]) |
| 308 | return -EINVAL; |
| 309 | |
| 310 | t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]); |
| 311 | t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]); |
| 312 | |
| 313 | return 0; |
| 314 | } |
| 315 | #endif |
| 316 | |
| 317 | static struct nf_sockopt_ops so_getorigdst = { |
| 318 | .pf = PF_INET, |
| 319 | .get_optmin = SO_ORIGINAL_DST, |
| 320 | .get_optmax = SO_ORIGINAL_DST+1, |
| 321 | .get = getorigdst, |
| 322 | .owner = THIS_MODULE, |
| 323 | }; |
| 324 | |
| 325 | static int ipv4_hooks_register(struct net *net) |
| 326 | { |
| 327 | struct conntrack4_net *cnet = net_generic(net, conntrack4_net_id); |
| 328 | int err = 0; |
| 329 | |
| 330 | mutex_lock(®ister_ipv4_hooks); |
| 331 | |
| 332 | cnet->users++; |
| 333 | if (cnet->users > 1) |
| 334 | goto out_unlock; |
| 335 | |
| 336 | err = nf_defrag_ipv4_enable(net); |
| 337 | if (err) { |
| 338 | cnet->users = 0; |
| 339 | goto out_unlock; |
| 340 | } |
| 341 | |
| 342 | err = nf_register_net_hooks(net, ipv4_conntrack_ops, |
| 343 | ARRAY_SIZE(ipv4_conntrack_ops)); |
| 344 | |
| 345 | if (err) |
| 346 | cnet->users = 0; |
| 347 | out_unlock: |
| 348 | mutex_unlock(®ister_ipv4_hooks); |
| 349 | return err; |
| 350 | } |
| 351 | |
| 352 | static void ipv4_hooks_unregister(struct net *net) |
| 353 | { |
| 354 | struct conntrack4_net *cnet = net_generic(net, conntrack4_net_id); |
| 355 | |
| 356 | mutex_lock(®ister_ipv4_hooks); |
| 357 | if (cnet->users && (--cnet->users == 0)) |
| 358 | nf_unregister_net_hooks(net, ipv4_conntrack_ops, |
| 359 | ARRAY_SIZE(ipv4_conntrack_ops)); |
| 360 | mutex_unlock(®ister_ipv4_hooks); |
| 361 | } |
| 362 | |
| 363 | struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = { |
| 364 | .l3proto = PF_INET, |
| 365 | .pkt_to_tuple = ipv4_pkt_to_tuple, |
| 366 | .invert_tuple = ipv4_invert_tuple, |
| 367 | .get_l4proto = ipv4_get_l4proto, |
| 368 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
| 369 | .tuple_to_nlattr = ipv4_tuple_to_nlattr, |
| 370 | .nlattr_to_tuple = ipv4_nlattr_to_tuple, |
| 371 | .nla_policy = ipv4_nla_policy, |
| 372 | .nla_size = NLA_ALIGN(NLA_HDRLEN + sizeof(u32)) + /* CTA_IP_V4_SRC */ |
| 373 | NLA_ALIGN(NLA_HDRLEN + sizeof(u32)), /* CTA_IP_V4_DST */ |
| 374 | #endif |
| 375 | .net_ns_get = ipv4_hooks_register, |
| 376 | .net_ns_put = ipv4_hooks_unregister, |
| 377 | .me = THIS_MODULE, |
| 378 | }; |
| 379 | |
| 380 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
| 381 | &nf_conntrack_htable_size, 0600); |
| 382 | |
| 383 | MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET)); |
| 384 | MODULE_ALIAS("ip_conntrack"); |
| 385 | MODULE_LICENSE("GPL"); |
| 386 | |
| 387 | static struct nf_conntrack_l4proto *builtin_l4proto4[] = { |
| 388 | &nf_conntrack_l4proto_tcp4, |
| 389 | &nf_conntrack_l4proto_udp4, |
| 390 | &nf_conntrack_l4proto_icmp, |
| 391 | #ifdef CONFIG_NF_CT_PROTO_DCCP |
| 392 | &nf_conntrack_l4proto_dccp4, |
| 393 | #endif |
| 394 | #ifdef CONFIG_NF_CT_PROTO_SCTP |
| 395 | &nf_conntrack_l4proto_sctp4, |
| 396 | #endif |
| 397 | #ifdef CONFIG_NF_CT_PROTO_UDPLITE |
| 398 | &nf_conntrack_l4proto_udplite4, |
| 399 | #endif |
| 400 | }; |
| 401 | |
| 402 | static int ipv4_net_init(struct net *net) |
| 403 | { |
| 404 | return nf_ct_l4proto_pernet_register(net, builtin_l4proto4, |
| 405 | ARRAY_SIZE(builtin_l4proto4)); |
| 406 | } |
| 407 | |
| 408 | static void ipv4_net_exit(struct net *net) |
| 409 | { |
| 410 | nf_ct_l4proto_pernet_unregister(net, builtin_l4proto4, |
| 411 | ARRAY_SIZE(builtin_l4proto4)); |
| 412 | } |
| 413 | |
| 414 | static struct pernet_operations ipv4_net_ops = { |
| 415 | .init = ipv4_net_init, |
| 416 | .exit = ipv4_net_exit, |
| 417 | .id = &conntrack4_net_id, |
| 418 | .size = sizeof(struct conntrack4_net), |
| 419 | }; |
| 420 | |
| 421 | static int __init nf_conntrack_l3proto_ipv4_init(void) |
| 422 | { |
| 423 | int ret = 0; |
| 424 | |
| 425 | need_conntrack(); |
| 426 | |
| 427 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
| 428 | if (WARN_ON(nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1) != |
| 429 | nf_conntrack_l3proto_ipv4.nla_size)) |
| 430 | return -EINVAL; |
| 431 | #endif |
| 432 | ret = nf_register_sockopt(&so_getorigdst); |
| 433 | if (ret < 0) { |
| 434 | pr_err("Unable to register netfilter socket option\n"); |
| 435 | return ret; |
| 436 | } |
| 437 | |
| 438 | ret = register_pernet_subsys(&ipv4_net_ops); |
| 439 | if (ret < 0) { |
| 440 | pr_err("nf_conntrack_ipv4: can't register pernet ops\n"); |
| 441 | goto cleanup_sockopt; |
| 442 | } |
| 443 | |
| 444 | ret = nf_ct_l4proto_register(builtin_l4proto4, |
| 445 | ARRAY_SIZE(builtin_l4proto4)); |
| 446 | if (ret < 0) |
| 447 | goto cleanup_pernet; |
| 448 | |
| 449 | ret = nf_ct_l3proto_register(&nf_conntrack_l3proto_ipv4); |
| 450 | if (ret < 0) { |
| 451 | pr_err("nf_conntrack_ipv4: can't register ipv4 proto.\n"); |
| 452 | goto cleanup_l4proto; |
| 453 | } |
| 454 | |
| 455 | return ret; |
| 456 | cleanup_l4proto: |
| 457 | nf_ct_l4proto_unregister(builtin_l4proto4, |
| 458 | ARRAY_SIZE(builtin_l4proto4)); |
| 459 | cleanup_pernet: |
| 460 | unregister_pernet_subsys(&ipv4_net_ops); |
| 461 | cleanup_sockopt: |
| 462 | nf_unregister_sockopt(&so_getorigdst); |
| 463 | return ret; |
| 464 | } |
| 465 | |
| 466 | static void __exit nf_conntrack_l3proto_ipv4_fini(void) |
| 467 | { |
| 468 | synchronize_net(); |
| 469 | nf_ct_l3proto_unregister(&nf_conntrack_l3proto_ipv4); |
| 470 | nf_ct_l4proto_unregister(builtin_l4proto4, |
| 471 | ARRAY_SIZE(builtin_l4proto4)); |
| 472 | unregister_pernet_subsys(&ipv4_net_ops); |
| 473 | nf_unregister_sockopt(&so_getorigdst); |
| 474 | } |
| 475 | |
| 476 | module_init(nf_conntrack_l3proto_ipv4_init); |
| 477 | module_exit(nf_conntrack_l3proto_ipv4_fini); |