| 1 | /* |
| 2 | * Copyright (C)2004 USAGI/WIDE Project |
| 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 | * Author: |
| 9 | * Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp> |
| 10 | */ |
| 11 | |
| 12 | #include <linux/types.h> |
| 13 | #include <linux/ipv6.h> |
| 14 | #include <linux/in6.h> |
| 15 | #include <linux/netfilter.h> |
| 16 | #include <linux/module.h> |
| 17 | #include <linux/skbuff.h> |
| 18 | #include <linux/icmp.h> |
| 19 | #include <linux/sysctl.h> |
| 20 | #include <net/ipv6.h> |
| 21 | #include <net/inet_frag.h> |
| 22 | |
| 23 | #include <linux/netfilter_ipv6.h> |
| 24 | #include <net/netfilter/nf_conntrack.h> |
| 25 | #include <net/netfilter/nf_conntrack_helper.h> |
| 26 | #include <net/netfilter/nf_conntrack_l4proto.h> |
| 27 | #include <net/netfilter/nf_conntrack_l3proto.h> |
| 28 | #include <net/netfilter/nf_conntrack_core.h> |
| 29 | #include <net/netfilter/ipv6/nf_conntrack_ipv6.h> |
| 30 | #include <net/netfilter/nf_log.h> |
| 31 | |
| 32 | static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff, |
| 33 | struct nf_conntrack_tuple *tuple) |
| 34 | { |
| 35 | const u_int32_t *ap; |
| 36 | u_int32_t _addrs[8]; |
| 37 | |
| 38 | ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr), |
| 39 | sizeof(_addrs), _addrs); |
| 40 | if (ap == NULL) |
| 41 | return false; |
| 42 | |
| 43 | memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6)); |
| 44 | memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6)); |
| 45 | |
| 46 | return true; |
| 47 | } |
| 48 | |
| 49 | static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple, |
| 50 | const struct nf_conntrack_tuple *orig) |
| 51 | { |
| 52 | memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6)); |
| 53 | memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6)); |
| 54 | |
| 55 | return true; |
| 56 | } |
| 57 | |
| 58 | static int ipv6_print_tuple(struct seq_file *s, |
| 59 | const struct nf_conntrack_tuple *tuple) |
| 60 | { |
| 61 | return seq_printf(s, "src=%pI6 dst=%pI6 ", |
| 62 | tuple->src.u3.ip6, tuple->dst.u3.ip6); |
| 63 | } |
| 64 | |
| 65 | /* |
| 66 | * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c |
| 67 | * |
| 68 | * This function parses (probably truncated) exthdr set "hdr" |
| 69 | * of length "len". "nexthdrp" initially points to some place, |
| 70 | * where type of the first header can be found. |
| 71 | * |
| 72 | * It skips all well-known exthdrs, and returns pointer to the start |
| 73 | * of unparsable area i.e. the first header with unknown type. |
| 74 | * if success, *nexthdr is updated by type/protocol of this header. |
| 75 | * |
| 76 | * NOTES: - it may return pointer pointing beyond end of packet, |
| 77 | * if the last recognized header is truncated in the middle. |
| 78 | * - if packet is truncated, so that all parsed headers are skipped, |
| 79 | * it returns -1. |
| 80 | * - if packet is fragmented, return pointer of the fragment header. |
| 81 | * - ESP is unparsable for now and considered like |
| 82 | * normal payload protocol. |
| 83 | * - Note also special handling of AUTH header. Thanks to IPsec wizards. |
| 84 | */ |
| 85 | |
| 86 | static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start, |
| 87 | u8 *nexthdrp, int len) |
| 88 | { |
| 89 | u8 nexthdr = *nexthdrp; |
| 90 | |
| 91 | while (ipv6_ext_hdr(nexthdr)) { |
| 92 | struct ipv6_opt_hdr hdr; |
| 93 | int hdrlen; |
| 94 | |
| 95 | if (len < (int)sizeof(struct ipv6_opt_hdr)) |
| 96 | return -1; |
| 97 | if (nexthdr == NEXTHDR_NONE) |
| 98 | break; |
| 99 | if (nexthdr == NEXTHDR_FRAGMENT) |
| 100 | break; |
| 101 | if (skb_copy_bits(skb, start, &hdr, sizeof(hdr))) |
| 102 | BUG(); |
| 103 | if (nexthdr == NEXTHDR_AUTH) |
| 104 | hdrlen = (hdr.hdrlen+2)<<2; |
| 105 | else |
| 106 | hdrlen = ipv6_optlen(&hdr); |
| 107 | |
| 108 | nexthdr = hdr.nexthdr; |
| 109 | len -= hdrlen; |
| 110 | start += hdrlen; |
| 111 | } |
| 112 | |
| 113 | *nexthdrp = nexthdr; |
| 114 | return start; |
| 115 | } |
| 116 | |
| 117 | static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff, |
| 118 | unsigned int *dataoff, u_int8_t *protonum) |
| 119 | { |
| 120 | unsigned int extoff = nhoff + sizeof(struct ipv6hdr); |
| 121 | unsigned char pnum; |
| 122 | int protoff; |
| 123 | |
| 124 | if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr), |
| 125 | &pnum, sizeof(pnum)) != 0) { |
| 126 | pr_debug("ip6_conntrack_core: can't get nexthdr\n"); |
| 127 | return -NF_ACCEPT; |
| 128 | } |
| 129 | protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff); |
| 130 | /* |
| 131 | * (protoff == skb->len) mean that the packet doesn't have no data |
| 132 | * except of IPv6 & ext headers. but it's tracked anyway. - YK |
| 133 | */ |
| 134 | if ((protoff < 0) || (protoff > skb->len)) { |
| 135 | pr_debug("ip6_conntrack_core: can't find proto in pkt\n"); |
| 136 | return -NF_ACCEPT; |
| 137 | } |
| 138 | |
| 139 | *dataoff = protoff; |
| 140 | *protonum = pnum; |
| 141 | return NF_ACCEPT; |
| 142 | } |
| 143 | |
| 144 | static unsigned int ipv6_confirm(unsigned int hooknum, |
| 145 | struct sk_buff *skb, |
| 146 | const struct net_device *in, |
| 147 | const struct net_device *out, |
| 148 | int (*okfn)(struct sk_buff *)) |
| 149 | { |
| 150 | struct nf_conn *ct; |
| 151 | const struct nf_conn_help *help; |
| 152 | const struct nf_conntrack_helper *helper; |
| 153 | enum ip_conntrack_info ctinfo; |
| 154 | unsigned int ret, protoff; |
| 155 | unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; |
| 156 | unsigned char pnum = ipv6_hdr(skb)->nexthdr; |
| 157 | |
| 158 | |
| 159 | /* This is where we call the helper: as the packet goes out. */ |
| 160 | ct = nf_ct_get(skb, &ctinfo); |
| 161 | if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY) |
| 162 | goto out; |
| 163 | |
| 164 | help = nfct_help(ct); |
| 165 | if (!help) |
| 166 | goto out; |
| 167 | /* rcu_read_lock()ed by nf_hook_slow */ |
| 168 | helper = rcu_dereference(help->helper); |
| 169 | if (!helper) |
| 170 | goto out; |
| 171 | |
| 172 | protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, |
| 173 | skb->len - extoff); |
| 174 | if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) { |
| 175 | pr_debug("proto header not found\n"); |
| 176 | return NF_ACCEPT; |
| 177 | } |
| 178 | |
| 179 | ret = helper->help(skb, protoff, ct, ctinfo); |
| 180 | if (ret != NF_ACCEPT) { |
| 181 | nf_log_packet(NFPROTO_IPV6, hooknum, skb, in, out, NULL, |
| 182 | "nf_ct_%s: dropping packet", helper->name); |
| 183 | return ret; |
| 184 | } |
| 185 | out: |
| 186 | /* We've seen it coming out the other side: confirm it */ |
| 187 | return nf_conntrack_confirm(skb); |
| 188 | } |
| 189 | |
| 190 | static enum ip6_defrag_users nf_ct6_defrag_user(unsigned int hooknum, |
| 191 | struct sk_buff *skb) |
| 192 | { |
| 193 | if (hooknum == NF_INET_PRE_ROUTING) |
| 194 | return IP6_DEFRAG_CONNTRACK_IN; |
| 195 | else |
| 196 | return IP6_DEFRAG_CONNTRACK_OUT; |
| 197 | |
| 198 | } |
| 199 | |
| 200 | static unsigned int ipv6_defrag(unsigned int hooknum, |
| 201 | struct sk_buff *skb, |
| 202 | const struct net_device *in, |
| 203 | const struct net_device *out, |
| 204 | int (*okfn)(struct sk_buff *)) |
| 205 | { |
| 206 | struct sk_buff *reasm; |
| 207 | |
| 208 | /* Previously seen (loopback)? */ |
| 209 | if (skb->nfct) |
| 210 | return NF_ACCEPT; |
| 211 | |
| 212 | reasm = nf_ct_frag6_gather(skb, nf_ct6_defrag_user(hooknum, skb)); |
| 213 | /* queued */ |
| 214 | if (reasm == NULL) |
| 215 | return NF_STOLEN; |
| 216 | |
| 217 | /* error occured or not fragmented */ |
| 218 | if (reasm == skb) |
| 219 | return NF_ACCEPT; |
| 220 | |
| 221 | nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in, |
| 222 | (struct net_device *)out, okfn); |
| 223 | |
| 224 | return NF_STOLEN; |
| 225 | } |
| 226 | |
| 227 | static unsigned int __ipv6_conntrack_in(struct net *net, |
| 228 | unsigned int hooknum, |
| 229 | struct sk_buff *skb, |
| 230 | int (*okfn)(struct sk_buff *)) |
| 231 | { |
| 232 | struct sk_buff *reasm = skb->nfct_reasm; |
| 233 | |
| 234 | /* This packet is fragmented and has reassembled packet. */ |
| 235 | if (reasm) { |
| 236 | /* Reassembled packet isn't parsed yet ? */ |
| 237 | if (!reasm->nfct) { |
| 238 | unsigned int ret; |
| 239 | |
| 240 | ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm); |
| 241 | if (ret != NF_ACCEPT) |
| 242 | return ret; |
| 243 | } |
| 244 | nf_conntrack_get(reasm->nfct); |
| 245 | skb->nfct = reasm->nfct; |
| 246 | skb->nfctinfo = reasm->nfctinfo; |
| 247 | return NF_ACCEPT; |
| 248 | } |
| 249 | |
| 250 | return nf_conntrack_in(net, PF_INET6, hooknum, skb); |
| 251 | } |
| 252 | |
| 253 | static unsigned int ipv6_conntrack_in(unsigned int hooknum, |
| 254 | struct sk_buff *skb, |
| 255 | const struct net_device *in, |
| 256 | const struct net_device *out, |
| 257 | int (*okfn)(struct sk_buff *)) |
| 258 | { |
| 259 | return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn); |
| 260 | } |
| 261 | |
| 262 | static unsigned int ipv6_conntrack_local(unsigned int hooknum, |
| 263 | struct sk_buff *skb, |
| 264 | const struct net_device *in, |
| 265 | const struct net_device *out, |
| 266 | int (*okfn)(struct sk_buff *)) |
| 267 | { |
| 268 | /* root is playing with raw sockets. */ |
| 269 | if (skb->len < sizeof(struct ipv6hdr)) { |
| 270 | if (net_ratelimit()) |
| 271 | printk("ipv6_conntrack_local: packet too short\n"); |
| 272 | return NF_ACCEPT; |
| 273 | } |
| 274 | return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn); |
| 275 | } |
| 276 | |
| 277 | static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = { |
| 278 | { |
| 279 | .hook = ipv6_defrag, |
| 280 | .owner = THIS_MODULE, |
| 281 | .pf = NFPROTO_IPV6, |
| 282 | .hooknum = NF_INET_PRE_ROUTING, |
| 283 | .priority = NF_IP6_PRI_CONNTRACK_DEFRAG, |
| 284 | }, |
| 285 | { |
| 286 | .hook = ipv6_conntrack_in, |
| 287 | .owner = THIS_MODULE, |
| 288 | .pf = NFPROTO_IPV6, |
| 289 | .hooknum = NF_INET_PRE_ROUTING, |
| 290 | .priority = NF_IP6_PRI_CONNTRACK, |
| 291 | }, |
| 292 | { |
| 293 | .hook = ipv6_conntrack_local, |
| 294 | .owner = THIS_MODULE, |
| 295 | .pf = NFPROTO_IPV6, |
| 296 | .hooknum = NF_INET_LOCAL_OUT, |
| 297 | .priority = NF_IP6_PRI_CONNTRACK, |
| 298 | }, |
| 299 | { |
| 300 | .hook = ipv6_defrag, |
| 301 | .owner = THIS_MODULE, |
| 302 | .pf = NFPROTO_IPV6, |
| 303 | .hooknum = NF_INET_LOCAL_OUT, |
| 304 | .priority = NF_IP6_PRI_CONNTRACK_DEFRAG, |
| 305 | }, |
| 306 | { |
| 307 | .hook = ipv6_confirm, |
| 308 | .owner = THIS_MODULE, |
| 309 | .pf = NFPROTO_IPV6, |
| 310 | .hooknum = NF_INET_POST_ROUTING, |
| 311 | .priority = NF_IP6_PRI_LAST, |
| 312 | }, |
| 313 | { |
| 314 | .hook = ipv6_confirm, |
| 315 | .owner = THIS_MODULE, |
| 316 | .pf = NFPROTO_IPV6, |
| 317 | .hooknum = NF_INET_LOCAL_IN, |
| 318 | .priority = NF_IP6_PRI_LAST-1, |
| 319 | }, |
| 320 | }; |
| 321 | |
| 322 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
| 323 | |
| 324 | #include <linux/netfilter/nfnetlink.h> |
| 325 | #include <linux/netfilter/nfnetlink_conntrack.h> |
| 326 | |
| 327 | static int ipv6_tuple_to_nlattr(struct sk_buff *skb, |
| 328 | const struct nf_conntrack_tuple *tuple) |
| 329 | { |
| 330 | NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4, |
| 331 | &tuple->src.u3.ip6); |
| 332 | NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4, |
| 333 | &tuple->dst.u3.ip6); |
| 334 | return 0; |
| 335 | |
| 336 | nla_put_failure: |
| 337 | return -1; |
| 338 | } |
| 339 | |
| 340 | static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = { |
| 341 | [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 }, |
| 342 | [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 }, |
| 343 | }; |
| 344 | |
| 345 | static int ipv6_nlattr_to_tuple(struct nlattr *tb[], |
| 346 | struct nf_conntrack_tuple *t) |
| 347 | { |
| 348 | if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST]) |
| 349 | return -EINVAL; |
| 350 | |
| 351 | memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]), |
| 352 | sizeof(u_int32_t) * 4); |
| 353 | memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]), |
| 354 | sizeof(u_int32_t) * 4); |
| 355 | |
| 356 | return 0; |
| 357 | } |
| 358 | |
| 359 | static int ipv6_nlattr_tuple_size(void) |
| 360 | { |
| 361 | return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1); |
| 362 | } |
| 363 | #endif |
| 364 | |
| 365 | struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = { |
| 366 | .l3proto = PF_INET6, |
| 367 | .name = "ipv6", |
| 368 | .pkt_to_tuple = ipv6_pkt_to_tuple, |
| 369 | .invert_tuple = ipv6_invert_tuple, |
| 370 | .print_tuple = ipv6_print_tuple, |
| 371 | .get_l4proto = ipv6_get_l4proto, |
| 372 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
| 373 | .tuple_to_nlattr = ipv6_tuple_to_nlattr, |
| 374 | .nlattr_tuple_size = ipv6_nlattr_tuple_size, |
| 375 | .nlattr_to_tuple = ipv6_nlattr_to_tuple, |
| 376 | .nla_policy = ipv6_nla_policy, |
| 377 | #endif |
| 378 | #ifdef CONFIG_SYSCTL |
| 379 | .ctl_table_path = nf_net_netfilter_sysctl_path, |
| 380 | .ctl_table = nf_ct_ipv6_sysctl_table, |
| 381 | #endif |
| 382 | .me = THIS_MODULE, |
| 383 | }; |
| 384 | |
| 385 | MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6)); |
| 386 | MODULE_LICENSE("GPL"); |
| 387 | MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>"); |
| 388 | |
| 389 | static int __init nf_conntrack_l3proto_ipv6_init(void) |
| 390 | { |
| 391 | int ret = 0; |
| 392 | |
| 393 | need_conntrack(); |
| 394 | |
| 395 | ret = nf_ct_frag6_init(); |
| 396 | if (ret < 0) { |
| 397 | printk("nf_conntrack_ipv6: can't initialize frag6.\n"); |
| 398 | return ret; |
| 399 | } |
| 400 | ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6); |
| 401 | if (ret < 0) { |
| 402 | printk("nf_conntrack_ipv6: can't register tcp.\n"); |
| 403 | goto cleanup_frag6; |
| 404 | } |
| 405 | |
| 406 | ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6); |
| 407 | if (ret < 0) { |
| 408 | printk("nf_conntrack_ipv6: can't register udp.\n"); |
| 409 | goto cleanup_tcp; |
| 410 | } |
| 411 | |
| 412 | ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6); |
| 413 | if (ret < 0) { |
| 414 | printk("nf_conntrack_ipv6: can't register icmpv6.\n"); |
| 415 | goto cleanup_udp; |
| 416 | } |
| 417 | |
| 418 | ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6); |
| 419 | if (ret < 0) { |
| 420 | printk("nf_conntrack_ipv6: can't register ipv6\n"); |
| 421 | goto cleanup_icmpv6; |
| 422 | } |
| 423 | |
| 424 | ret = nf_register_hooks(ipv6_conntrack_ops, |
| 425 | ARRAY_SIZE(ipv6_conntrack_ops)); |
| 426 | if (ret < 0) { |
| 427 | printk("nf_conntrack_ipv6: can't register pre-routing defrag " |
| 428 | "hook.\n"); |
| 429 | goto cleanup_ipv6; |
| 430 | } |
| 431 | return ret; |
| 432 | |
| 433 | cleanup_ipv6: |
| 434 | nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6); |
| 435 | cleanup_icmpv6: |
| 436 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6); |
| 437 | cleanup_udp: |
| 438 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6); |
| 439 | cleanup_tcp: |
| 440 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6); |
| 441 | cleanup_frag6: |
| 442 | nf_ct_frag6_cleanup(); |
| 443 | return ret; |
| 444 | } |
| 445 | |
| 446 | static void __exit nf_conntrack_l3proto_ipv6_fini(void) |
| 447 | { |
| 448 | synchronize_net(); |
| 449 | nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops)); |
| 450 | nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6); |
| 451 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6); |
| 452 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6); |
| 453 | nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6); |
| 454 | nf_ct_frag6_cleanup(); |
| 455 | } |
| 456 | |
| 457 | module_init(nf_conntrack_l3proto_ipv6_init); |
| 458 | module_exit(nf_conntrack_l3proto_ipv6_fini); |