net/ipv6/netfilter/nf_conntrack_l3proto_ipv6.c

   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 <net/ipv6.h>
  20 #include <net/inet_frag.h>
  21
  22 #include <linux/netfilter_bridge.h>
  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/nf_conntrack_zones.h>
  30 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
  31 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  32 #include <net/netfilter/nf_log.h>
  33
  34 static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  35                               struct nf_conntrack_tuple *tuple)
  36 {
  37         const u_int32_t *ap;
  38         u_int32_t _addrs[8];
  39
  40         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
  41                                 sizeof(_addrs), _addrs);
  42         if (ap == NULL)
  43                 return false;
  44
  45         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
  46         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
  47
  48         return true;
  49 }
  50
  51 static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
  52                               const struct nf_conntrack_tuple *orig)
  53 {
  54         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
  55         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
  56
  57         return true;
  58 }
  59
  60 static int ipv6_print_tuple(struct seq_file *s,
  61                             const struct nf_conntrack_tuple *tuple)
  62 {
  63         return seq_printf(s, "src=%pI6 dst=%pI6 ",
  64                           tuple->src.u3.ip6, tuple->dst.u3.ip6);
  65 }
  66
  67 /*
  68  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
  69  *
  70  * This function parses (probably truncated) exthdr set "hdr"
  71  * of length "len". "nexthdrp" initially points to some place,
  72  * where type of the first header can be found.
  73  *
  74  * It skips all well-known exthdrs, and returns pointer to the start
  75  * of unparsable area i.e. the first header with unknown type.
  76  * if success, *nexthdr is updated by type/protocol of this header.
  77  *
  78  * NOTES: - it may return pointer pointing beyond end of packet,
  79  *          if the last recognized header is truncated in the middle.
  80  *        - if packet is truncated, so that all parsed headers are skipped,
  81  *          it returns -1.
  82  *        - if packet is fragmented, return pointer of the fragment header.
  83  *        - ESP is unparsable for now and considered like
  84  *          normal payload protocol.
  85  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
  86  */
  87
  88 static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
  89                                   u8 *nexthdrp, int len)
  90 {
  91         u8 nexthdr = *nexthdrp;
  92
  93         while (ipv6_ext_hdr(nexthdr)) {
  94                 struct ipv6_opt_hdr hdr;
  95                 int hdrlen;
  96
  97                 if (len < (int)sizeof(struct ipv6_opt_hdr))
  98                         return -1;
  99                 if (nexthdr == NEXTHDR_NONE)
 100                         break;
 101                 if (nexthdr == NEXTHDR_FRAGMENT)
 102                         break;
 103                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
 104                         BUG();
 105                 if (nexthdr == NEXTHDR_AUTH)
 106                         hdrlen = (hdr.hdrlen+2)<<2;
 107                 else
 108                         hdrlen = ipv6_optlen(&hdr);
 109
 110                 nexthdr = hdr.nexthdr;
 111                 len -= hdrlen;
 112                 start += hdrlen;
 113         }
 114
 115         *nexthdrp = nexthdr;
 116         return start;
 117 }
 118
 119 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
 120                             unsigned int *dataoff, u_int8_t *protonum)
 121 {
 122         unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
 123         unsigned char pnum;
 124         int protoff;
 125
 126         if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
 127                           &pnum, sizeof(pnum)) != 0) {
 128                 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
 129                 return -NF_ACCEPT;
 130         }
 131         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
 132         /*
 133          * (protoff == skb->len) mean that the packet doesn't have no data
 134          * except of IPv6 & ext headers. but it's tracked anyway. - YK
 135          */
 136         if ((protoff < 0) || (protoff > skb->len)) {
 137                 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
 138                 return -NF_ACCEPT;
 139         }
 140
 141         *dataoff = protoff;
 142         *protonum = pnum;
 143         return NF_ACCEPT;
 144 }
 145
 146 static unsigned int ipv6_confirm(unsigned int hooknum,
 147                                  struct sk_buff *skb,
 148                                  const struct net_device *in,
 149                                  const struct net_device *out,
 150                                  int (*okfn)(struct sk_buff *))
 151 {
 152         struct nf_conn *ct;
 153         const struct nf_conn_help *help;
 154         const struct nf_conntrack_helper *helper;
 155         enum ip_conntrack_info ctinfo;
 156         unsigned int ret, protoff;
 157         unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
 158         unsigned char pnum = ipv6_hdr(skb)->nexthdr;
 159
 160
 161         /* This is where we call the helper: as the packet goes out. */
 162         ct = nf_ct_get(skb, &ctinfo);
 163         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 164                 goto out;
 165
 166         help = nfct_help(ct);
 167         if (!help)
 168                 goto out;
 169         /* rcu_read_lock()ed by nf_hook_slow */
 170         helper = rcu_dereference(help->helper);
 171         if (!helper)
 172                 goto out;
 173
 174         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum,
 175                                          skb->len - extoff);
 176         if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) {
 177                 pr_debug("proto header not found\n");
 178                 return NF_ACCEPT;
 179         }
 180
 181         ret = helper->help(skb, protoff, ct, ctinfo);
 182         if (ret != NF_ACCEPT) {
 183                 nf_log_packet(NFPROTO_IPV6, hooknum, skb, in, out, NULL,
 184                               "nf_ct_%s: dropping packet", helper->name);
 185                 return ret;
 186         }
 187 out:
 188         /* We've seen it coming out the other side: confirm it */
 189         return nf_conntrack_confirm(skb);
 190 }
 191
 192 static unsigned int __ipv6_conntrack_in(struct net *net,
 193                                         unsigned int hooknum,
 194                                         struct sk_buff *skb,
 195                                         int (*okfn)(struct sk_buff *))
 196 {
 197         struct sk_buff *reasm = skb->nfct_reasm;
 198
 199         /* This packet is fragmented and has reassembled packet. */
 200         if (reasm) {
 201                 /* Reassembled packet isn't parsed yet ? */
 202                 if (!reasm->nfct) {
 203                         unsigned int ret;
 204
 205                         ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
 206                         if (ret != NF_ACCEPT)
 207                                 return ret;
 208                 }
 209                 nf_conntrack_get(reasm->nfct);
 210                 skb->nfct = reasm->nfct;
 211                 skb->nfctinfo = reasm->nfctinfo;
 212                 return NF_ACCEPT;
 213         }
 214
 215         return nf_conntrack_in(net, PF_INET6, hooknum, skb);
 216 }
 217
 218 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
 219                                       struct sk_buff *skb,
 220                                       const struct net_device *in,
 221                                       const struct net_device *out,
 222                                       int (*okfn)(struct sk_buff *))
 223 {
 224         return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn);
 225 }
 226
 227 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
 228                                          struct sk_buff *skb,
 229                                          const struct net_device *in,
 230                                          const struct net_device *out,
 231                                          int (*okfn)(struct sk_buff *))
 232 {
 233         /* root is playing with raw sockets. */
 234         if (skb->len < sizeof(struct ipv6hdr)) {
 235                 if (net_ratelimit())
 236                         pr_notice("ipv6_conntrack_local: packet too short\n");
 237                 return NF_ACCEPT;
 238         }
 239         return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn);
 240 }
 241
 242 static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
 243         {
 244                 .hook           = ipv6_conntrack_in,
 245                 .owner          = THIS_MODULE,
 246                 .pf             = NFPROTO_IPV6,
 247                 .hooknum        = NF_INET_PRE_ROUTING,
 248                 .priority       = NF_IP6_PRI_CONNTRACK,
 249         },
 250         {
 251                 .hook           = ipv6_conntrack_local,
 252                 .owner          = THIS_MODULE,
 253                 .pf             = NFPROTO_IPV6,
 254                 .hooknum        = NF_INET_LOCAL_OUT,
 255                 .priority       = NF_IP6_PRI_CONNTRACK,
 256         },
 257         {
 258                 .hook           = ipv6_confirm,
 259                 .owner          = THIS_MODULE,
 260                 .pf             = NFPROTO_IPV6,
 261                 .hooknum        = NF_INET_POST_ROUTING,
 262                 .priority       = NF_IP6_PRI_LAST,
 263         },
 264         {
 265                 .hook           = ipv6_confirm,
 266                 .owner          = THIS_MODULE,
 267                 .pf             = NFPROTO_IPV6,
 268                 .hooknum        = NF_INET_LOCAL_IN,
 269                 .priority       = NF_IP6_PRI_LAST-1,
 270         },
 271 };
 272
 273 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 274
 275 #include <linux/netfilter/nfnetlink.h>
 276 #include <linux/netfilter/nfnetlink_conntrack.h>
 277
 278 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
 279                                 const struct nf_conntrack_tuple *tuple)
 280 {
 281         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
 282                 &tuple->src.u3.ip6);
 283         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
 284                 &tuple->dst.u3.ip6);
 285         return 0;
 286
 287 nla_put_failure:
 288         return -1;
 289 }
 290
 291 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
 292         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
 293         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
 294 };
 295
 296 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
 297                                 struct nf_conntrack_tuple *t)
 298 {
 299         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
 300                 return -EINVAL;
 301
 302         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
 303                sizeof(u_int32_t) * 4);
 304         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
 305                sizeof(u_int32_t) * 4);
 306
 307         return 0;
 308 }
 309
 310 static int ipv6_nlattr_tuple_size(void)
 311 {
 312         return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1);
 313 }
 314 #endif
 315
 316 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
 317         .l3proto                = PF_INET6,
 318         .name                   = "ipv6",
 319         .pkt_to_tuple           = ipv6_pkt_to_tuple,
 320         .invert_tuple           = ipv6_invert_tuple,
 321         .print_tuple            = ipv6_print_tuple,
 322         .get_l4proto            = ipv6_get_l4proto,
 323 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 324         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
 325         .nlattr_tuple_size      = ipv6_nlattr_tuple_size,
 326         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
 327         .nla_policy             = ipv6_nla_policy,
 328 #endif
 329         .me                     = THIS_MODULE,
 330 };
 331
 332 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
 333 MODULE_LICENSE("GPL");
 334 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
 335
 336 static int __init nf_conntrack_l3proto_ipv6_init(void)
 337 {
 338         int ret = 0;
 339
 340         need_conntrack();
 341         nf_defrag_ipv6_enable();
 342
 343         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
 344         if (ret < 0) {
 345                 pr_err("nf_conntrack_ipv6: can't register tcp.\n");
 346                 return ret;
 347         }
 348
 349         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
 350         if (ret < 0) {
 351                 pr_err("nf_conntrack_ipv6: can't register udp.\n");
 352                 goto cleanup_tcp;
 353         }
 354
 355         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
 356         if (ret < 0) {
 357                 pr_err("nf_conntrack_ipv6: can't register icmpv6.\n");
 358                 goto cleanup_udp;
 359         }
 360
 361         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
 362         if (ret < 0) {
 363                 pr_err("nf_conntrack_ipv6: can't register ipv6\n");
 364                 goto cleanup_icmpv6;
 365         }
 366
 367         ret = nf_register_hooks(ipv6_conntrack_ops,
 368                                 ARRAY_SIZE(ipv6_conntrack_ops));
 369         if (ret < 0) {
 370                 pr_err("nf_conntrack_ipv6: can't register pre-routing defrag "
 371                        "hook.\n");
 372                 goto cleanup_ipv6;
 373         }
 374         return ret;
 375
 376  cleanup_ipv6:
 377         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
 378  cleanup_icmpv6:
 379         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
 380  cleanup_udp:
 381         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
 382  cleanup_tcp:
 383         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
 384         return ret;
 385 }
 386
 387 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
 388 {
 389         synchronize_net();
 390         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
 391         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
 392         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
 393         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
 394         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
 395 }
 396
 397 module_init(nf_conntrack_l3proto_ipv6_init);
 398 module_exit(nf_conntrack_l3proto_ipv6_fini);