net/ipv6/fib6_rules.c

   1 /*
   2  * net/ipv6/fib6_rules.c        IPv6 Routing Policy Rules
   3  *
   4  * Copyright (C)2003-2006 Helsinki University of Technology
   5  * Copyright (C)2003-2006 USAGI/WIDE Project
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License as
   9  *      published by the Free Software Foundation, version 2.
  10  *
  11  * Authors
  12  *      Thomas Graf             <tgraf@suug.ch>
  13  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
  14  */
  15
  16 #include <linux/netdevice.h>
  17 #include <linux/notifier.h>
  18 #include <linux/export.h>
  19
  20 #include <net/fib_rules.h>
  21 #include <net/ipv6.h>
  22 #include <net/addrconf.h>
  23 #include <net/ip6_route.h>
  24 #include <net/netlink.h>
  25
  26 struct fib6_rule {
  27         struct fib_rule         common;
  28         struct rt6key           src;
  29         struct rt6key           dst;
  30         u8                      tclass;
  31 };
  32
  33 static bool fib6_rule_matchall(const struct fib_rule *rule)
  34 {
  35         struct fib6_rule *r = container_of(rule, struct fib6_rule, common);
  36
  37         if (r->dst.plen || r->src.plen || r->tclass)
  38                 return false;
  39         return fib_rule_matchall(rule);
  40 }
  41
  42 bool fib6_rule_default(const struct fib_rule *rule)
  43 {
  44         if (!fib6_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
  45             rule->l3mdev)
  46                 return false;
  47         if (rule->table != RT6_TABLE_LOCAL && rule->table != RT6_TABLE_MAIN)
  48                 return false;
  49         return true;
  50 }
  51 EXPORT_SYMBOL_GPL(fib6_rule_default);
  52
  53 int fib6_rules_dump(struct net *net, struct notifier_block *nb)
  54 {
  55         return fib_rules_dump(net, nb, AF_INET6);
  56 }
  57
  58 unsigned int fib6_rules_seq_read(struct net *net)
  59 {
  60         return fib_rules_seq_read(net, AF_INET6);
  61 }
  62
  63 /* called with rcu lock held; no reference taken on fib6_info */
  64 struct fib6_info *fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
  65                               int flags)
  66 {
  67         struct fib6_info *f6i;
  68         int err;
  69
  70         if (net->ipv6.fib6_has_custom_rules) {
  71                 struct fib_lookup_arg arg = {
  72                         .lookup_ptr = fib6_table_lookup,
  73                         .lookup_data = &oif,
  74                         .flags = FIB_LOOKUP_NOREF,
  75                 };
  76
  77                 l3mdev_update_flow(net, flowi6_to_flowi(fl6));
  78
  79                 err = fib_rules_lookup(net->ipv6.fib6_rules_ops,
  80                                        flowi6_to_flowi(fl6), flags, &arg);
  81                 if (err)
  82                         return ERR_PTR(err);
  83
  84                 f6i = arg.result ? : net->ipv6.fib6_null_entry;
  85         } else {
  86                 f6i = fib6_table_lookup(net, net->ipv6.fib6_local_tbl,
  87                                         oif, fl6, flags);
  88                 if (!f6i || f6i == net->ipv6.fib6_null_entry)
  89                         f6i = fib6_table_lookup(net, net->ipv6.fib6_main_tbl,
  90                                                 oif, fl6, flags);
  91         }
  92
  93         return f6i;
  94 }
  95
  96 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
  97                                    const struct sk_buff *skb,
  98                                    int flags, pol_lookup_t lookup)
  99 {
 100         if (net->ipv6.fib6_has_custom_rules) {
 101                 struct fib_lookup_arg arg = {
 102                         .lookup_ptr = lookup,
 103                         .lookup_data = skb,
 104                         .flags = FIB_LOOKUP_NOREF,
 105                 };
 106
 107                 /* update flow if oif or iif point to device enslaved to l3mdev */
 108                 l3mdev_update_flow(net, flowi6_to_flowi(fl6));
 109
 110                 fib_rules_lookup(net->ipv6.fib6_rules_ops,
 111                                  flowi6_to_flowi(fl6), flags, &arg);
 112
 113                 if (arg.result)
 114                         return arg.result;
 115         } else {
 116                 struct rt6_info *rt;
 117
 118                 rt = lookup(net, net->ipv6.fib6_local_tbl, fl6, skb, flags);
 119                 if (rt != net->ipv6.ip6_null_entry && rt->dst.error != -EAGAIN)
 120                         return &rt->dst;
 121                 ip6_rt_put(rt);
 122                 rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
 123                 if (rt->dst.error != -EAGAIN)
 124                         return &rt->dst;
 125                 ip6_rt_put(rt);
 126         }
 127
 128         dst_hold(&net->ipv6.ip6_null_entry->dst);
 129         return &net->ipv6.ip6_null_entry->dst;
 130 }
 131
 132 static int fib6_rule_saddr(struct net *net, struct fib_rule *rule, int flags,
 133                            struct flowi6 *flp6, const struct net_device *dev)
 134 {
 135         struct fib6_rule *r = (struct fib6_rule *)rule;
 136
 137         /* If we need to find a source address for this traffic,
 138          * we check the result if it meets requirement of the rule.
 139          */
 140         if ((rule->flags & FIB_RULE_FIND_SADDR) &&
 141             r->src.plen && !(flags & RT6_LOOKUP_F_HAS_SADDR)) {
 142                 struct in6_addr saddr;
 143
 144                 if (ipv6_dev_get_saddr(net, dev, &flp6->daddr,
 145                                        rt6_flags2srcprefs(flags), &saddr))
 146                         return -EAGAIN;
 147
 148                 if (!ipv6_prefix_equal(&saddr, &r->src.addr, r->src.plen))
 149                         return -EAGAIN;
 150
 151                 flp6->saddr = saddr;
 152         }
 153
 154         return 0;
 155 }
 156
 157 static int fib6_rule_action_alt(struct fib_rule *rule, struct flowi *flp,
 158                                 int flags, struct fib_lookup_arg *arg)
 159 {
 160         struct flowi6 *flp6 = &flp->u.ip6;
 161         struct net *net = rule->fr_net;
 162         struct fib6_table *table;
 163         struct fib6_info *f6i;
 164         int err = -EAGAIN, *oif;
 165         u32 tb_id;
 166
 167         switch (rule->action) {
 168         case FR_ACT_TO_TBL:
 169                 break;
 170         case FR_ACT_UNREACHABLE:
 171                 return -ENETUNREACH;
 172         case FR_ACT_PROHIBIT:
 173                 return -EACCES;
 174         case FR_ACT_BLACKHOLE:
 175         default:
 176                 return -EINVAL;
 177         }
 178
 179         tb_id = fib_rule_get_table(rule, arg);
 180         table = fib6_get_table(net, tb_id);
 181         if (!table)
 182                 return -EAGAIN;
 183
 184         oif = (int *)arg->lookup_data;
 185         f6i = fib6_table_lookup(net, table, *oif, flp6, flags);
 186         if (f6i != net->ipv6.fib6_null_entry) {
 187                 err = fib6_rule_saddr(net, rule, flags, flp6,
 188                                       fib6_info_nh_dev(f6i));
 189
 190                 if (likely(!err))
 191                         arg->result = f6i;
 192         }
 193
 194         return err;
 195 }
 196
 197 static int __fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
 198                               int flags, struct fib_lookup_arg *arg)
 199 {
 200         struct flowi6 *flp6 = &flp->u.ip6;
 201         struct rt6_info *rt = NULL;
 202         struct fib6_table *table;
 203         struct net *net = rule->fr_net;
 204         pol_lookup_t lookup = arg->lookup_ptr;
 205         int err = 0;
 206         u32 tb_id;
 207
 208         switch (rule->action) {
 209         case FR_ACT_TO_TBL:
 210                 break;
 211         case FR_ACT_UNREACHABLE:
 212                 err = -ENETUNREACH;
 213                 rt = net->ipv6.ip6_null_entry;
 214                 goto discard_pkt;
 215         default:
 216         case FR_ACT_BLACKHOLE:
 217                 err = -EINVAL;
 218                 rt = net->ipv6.ip6_blk_hole_entry;
 219                 goto discard_pkt;
 220         case FR_ACT_PROHIBIT:
 221                 err = -EACCES;
 222                 rt = net->ipv6.ip6_prohibit_entry;
 223                 goto discard_pkt;
 224         }
 225
 226         tb_id = fib_rule_get_table(rule, arg);
 227         table = fib6_get_table(net, tb_id);
 228         if (!table) {
 229                 err = -EAGAIN;
 230                 goto out;
 231         }
 232
 233         rt = lookup(net, table, flp6, arg->lookup_data, flags);
 234         if (rt != net->ipv6.ip6_null_entry) {
 235                 struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
 236
 237                 if (!idev)
 238                         goto again;
 239                 err = fib6_rule_saddr(net, rule, flags, flp6,
 240                                       idev->dev);
 241
 242                 if (err == -EAGAIN)
 243                         goto again;
 244
 245                 err = rt->dst.error;
 246                 if (err != -EAGAIN)
 247                         goto out;
 248         }
 249 again:
 250         ip6_rt_put(rt);
 251         err = -EAGAIN;
 252         rt = NULL;
 253         goto out;
 254
 255 discard_pkt:
 256         dst_hold(&rt->dst);
 257 out:
 258         arg->result = rt;
 259         return err;
 260 }
 261
 262 static int fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
 263                             int flags, struct fib_lookup_arg *arg)
 264 {
 265         if (arg->lookup_ptr == fib6_table_lookup)
 266                 return fib6_rule_action_alt(rule, flp, flags, arg);
 267
 268         return __fib6_rule_action(rule, flp, flags, arg);
 269 }
 270
 271 static bool fib6_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
 272 {
 273         struct rt6_info *rt = (struct rt6_info *) arg->result;
 274         struct net_device *dev = NULL;
 275
 276         if (rt->rt6i_idev)
 277                 dev = rt->rt6i_idev->dev;
 278
 279         /* do not accept result if the route does
 280          * not meet the required prefix length
 281          */
 282         if (rt->rt6i_dst.plen <= rule->suppress_prefixlen)
 283                 goto suppress_route;
 284
 285         /* do not accept result if the route uses a device
 286          * belonging to a forbidden interface group
 287          */
 288         if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
 289                 goto suppress_route;
 290
 291         return false;
 292
 293 suppress_route:
 294         ip6_rt_put(rt);
 295         return true;
 296 }
 297
 298 static int fib6_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 299 {
 300         struct fib6_rule *r = (struct fib6_rule *) rule;
 301         struct flowi6 *fl6 = &fl->u.ip6;
 302
 303         if (r->dst.plen &&
 304             !ipv6_prefix_equal(&fl6->daddr, &r->dst.addr, r->dst.plen))
 305                 return 0;
 306
 307         /*
 308          * If FIB_RULE_FIND_SADDR is set and we do not have a
 309          * source address for the traffic, we defer check for
 310          * source address.
 311          */
 312         if (r->src.plen) {
 313                 if (flags & RT6_LOOKUP_F_HAS_SADDR) {
 314                         if (!ipv6_prefix_equal(&fl6->saddr, &r->src.addr,
 315                                                r->src.plen))
 316                                 return 0;
 317                 } else if (!(r->common.flags & FIB_RULE_FIND_SADDR))
 318                         return 0;
 319         }
 320
 321         if (r->tclass && r->tclass != ip6_tclass(fl6->flowlabel))
 322                 return 0;
 323
 324         if (rule->ip_proto && (rule->ip_proto != fl6->flowi6_proto))
 325                 return 0;
 326
 327         if (fib_rule_port_range_set(&rule->sport_range) &&
 328             !fib_rule_port_inrange(&rule->sport_range, fl6->fl6_sport))
 329                 return 0;
 330
 331         if (fib_rule_port_range_set(&rule->dport_range) &&
 332             !fib_rule_port_inrange(&rule->dport_range, fl6->fl6_dport))
 333                 return 0;
 334
 335         return 1;
 336 }
 337
 338 static const struct nla_policy fib6_rule_policy[FRA_MAX+1] = {
 339         FRA_GENERIC_POLICY,
 340 };
 341
 342 static int fib6_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 343                                struct fib_rule_hdr *frh,
 344                                struct nlattr **tb,
 345                                struct netlink_ext_ack *extack)
 346 {
 347         int err = -EINVAL;
 348         struct net *net = sock_net(skb->sk);
 349         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 350
 351         if (rule->action == FR_ACT_TO_TBL && !rule->l3mdev) {
 352                 if (rule->table == RT6_TABLE_UNSPEC) {
 353                         NL_SET_ERR_MSG(extack, "Invalid table");
 354                         goto errout;
 355                 }
 356
 357                 if (fib6_new_table(net, rule->table) == NULL) {
 358                         err = -ENOBUFS;
 359                         goto errout;
 360                 }
 361         }
 362
 363         if (frh->src_len)
 364                 rule6->src.addr = nla_get_in6_addr(tb[FRA_SRC]);
 365
 366         if (frh->dst_len)
 367                 rule6->dst.addr = nla_get_in6_addr(tb[FRA_DST]);
 368
 369         rule6->src.plen = frh->src_len;
 370         rule6->dst.plen = frh->dst_len;
 371         rule6->tclass = frh->tos;
 372
 373         if (fib_rule_requires_fldissect(rule))
 374                 net->ipv6.fib6_rules_require_fldissect++;
 375
 376         net->ipv6.fib6_has_custom_rules = true;
 377         err = 0;
 378 errout:
 379         return err;
 380 }
 381
 382 static int fib6_rule_delete(struct fib_rule *rule)
 383 {
 384         struct net *net = rule->fr_net;
 385
 386         if (net->ipv6.fib6_rules_require_fldissect &&
 387             fib_rule_requires_fldissect(rule))
 388                 net->ipv6.fib6_rules_require_fldissect--;
 389
 390         return 0;
 391 }
 392
 393 static int fib6_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 394                              struct nlattr **tb)
 395 {
 396         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 397
 398         if (frh->src_len && (rule6->src.plen != frh->src_len))
 399                 return 0;
 400
 401         if (frh->dst_len && (rule6->dst.plen != frh->dst_len))
 402                 return 0;
 403
 404         if (frh->tos && (rule6->tclass != frh->tos))
 405                 return 0;
 406
 407         if (frh->src_len &&
 408             nla_memcmp(tb[FRA_SRC], &rule6->src.addr, sizeof(struct in6_addr)))
 409                 return 0;
 410
 411         if (frh->dst_len &&
 412             nla_memcmp(tb[FRA_DST], &rule6->dst.addr, sizeof(struct in6_addr)))
 413                 return 0;
 414
 415         return 1;
 416 }
 417
 418 static int fib6_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 419                           struct fib_rule_hdr *frh)
 420 {
 421         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 422
 423         frh->dst_len = rule6->dst.plen;
 424         frh->src_len = rule6->src.plen;
 425         frh->tos = rule6->tclass;
 426
 427         if ((rule6->dst.plen &&
 428              nla_put_in6_addr(skb, FRA_DST, &rule6->dst.addr)) ||
 429             (rule6->src.plen &&
 430              nla_put_in6_addr(skb, FRA_SRC, &rule6->src.addr)))
 431                 goto nla_put_failure;
 432         return 0;
 433
 434 nla_put_failure:
 435         return -ENOBUFS;
 436 }
 437
 438 static size_t fib6_rule_nlmsg_payload(struct fib_rule *rule)
 439 {
 440         return nla_total_size(16) /* dst */
 441                + nla_total_size(16); /* src */
 442 }
 443
 444 static void fib6_rule_flush_cache(struct fib_rules_ops *ops)
 445 {
 446         rt_genid_bump_ipv6(ops->fro_net);
 447 }
 448
 449 static const struct fib_rules_ops __net_initconst fib6_rules_ops_template = {
 450         .family                 = AF_INET6,
 451         .rule_size              = sizeof(struct fib6_rule),
 452         .addr_size              = sizeof(struct in6_addr),
 453         .action                 = fib6_rule_action,
 454         .match                  = fib6_rule_match,
 455         .suppress               = fib6_rule_suppress,
 456         .configure              = fib6_rule_configure,
 457         .delete                 = fib6_rule_delete,
 458         .compare                = fib6_rule_compare,
 459         .fill                   = fib6_rule_fill,
 460         .nlmsg_payload          = fib6_rule_nlmsg_payload,
 461         .flush_cache            = fib6_rule_flush_cache,
 462         .nlgroup                = RTNLGRP_IPV6_RULE,
 463         .policy                 = fib6_rule_policy,
 464         .owner                  = THIS_MODULE,
 465         .fro_net                = &init_net,
 466 };
 467
 468 static int __net_init fib6_rules_net_init(struct net *net)
 469 {
 470         struct fib_rules_ops *ops;
 471         int err = -ENOMEM;
 472
 473         ops = fib_rules_register(&fib6_rules_ops_template, net);
 474         if (IS_ERR(ops))
 475                 return PTR_ERR(ops);
 476
 477         err = fib_default_rule_add(ops, 0, RT6_TABLE_LOCAL, 0);
 478         if (err)
 479                 goto out_fib6_rules_ops;
 480
 481         err = fib_default_rule_add(ops, 0x7FFE, RT6_TABLE_MAIN, 0);
 482         if (err)
 483                 goto out_fib6_rules_ops;
 484
 485         net->ipv6.fib6_rules_ops = ops;
 486         net->ipv6.fib6_rules_require_fldissect = 0;
 487 out:
 488         return err;
 489
 490 out_fib6_rules_ops:
 491         fib_rules_unregister(ops);
 492         goto out;
 493 }
 494
 495 static void __net_exit fib6_rules_net_exit(struct net *net)
 496 {
 497         rtnl_lock();
 498         fib_rules_unregister(net->ipv6.fib6_rules_ops);
 499         rtnl_unlock();
 500 }
 501
 502 static struct pernet_operations fib6_rules_net_ops = {
 503         .init = fib6_rules_net_init,
 504         .exit = fib6_rules_net_exit,
 505 };
 506
 507 int __init fib6_rules_init(void)
 508 {
 509         return register_pernet_subsys(&fib6_rules_net_ops);
 510 }
 511
 512
 513 void fib6_rules_cleanup(void)
 514 {
 515         unregister_pernet_subsys(&fib6_rules_net_ops);
 516 }