net/ipv6/fib6_rules.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * net/ipv6/fib6_rules.c        IPv6 Routing Policy Rules
   4  *
   5  * Copyright (C)2003-2006 Helsinki University of Technology
   6  * Copyright (C)2003-2006 USAGI/WIDE Project
   7  *
   8  * Authors
   9  *      Thomas Graf             <tgraf@suug.ch>
  10  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
  11  */
  12
  13 #include <linux/netdevice.h>
  14 #include <linux/notifier.h>
  15 #include <linux/export.h>
  16 #include <linux/indirect_call_wrapper.h>
  17
  18 #include <net/fib_rules.h>
  19 #include <net/inet_dscp.h>
  20 #include <net/ipv6.h>
  21 #include <net/addrconf.h>
  22 #include <net/ip6_route.h>
  23 #include <net/netlink.h>
  24
  25 struct fib6_rule {
  26         struct fib_rule         common;
  27         struct rt6key           src;
  28         struct rt6key           dst;
  29         dscp_t                  dscp;
  30 };
  31
  32 static bool fib6_rule_matchall(const struct fib_rule *rule)
  33 {
  34         struct fib6_rule *r = container_of(rule, struct fib6_rule, common);
  35
  36         if (r->dst.plen || r->src.plen || r->dscp)
  37                 return false;
  38         return fib_rule_matchall(rule);
  39 }
  40
  41 bool fib6_rule_default(const struct fib_rule *rule)
  42 {
  43         if (!fib6_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
  44             rule->l3mdev)
  45                 return false;
  46         if (rule->table != RT6_TABLE_LOCAL && rule->table != RT6_TABLE_MAIN)
  47                 return false;
  48         return true;
  49 }
  50 EXPORT_SYMBOL_GPL(fib6_rule_default);
  51
  52 int fib6_rules_dump(struct net *net, struct notifier_block *nb,
  53                     struct netlink_ext_ack *extack)
  54 {
  55         return fib_rules_dump(net, nb, AF_INET6, extack);
  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 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
  65                 struct fib6_result *res, int flags)
  66 {
  67         int err;
  68
  69         if (net->ipv6.fib6_has_custom_rules) {
  70                 struct fib_lookup_arg arg = {
  71                         .lookup_ptr = fib6_table_lookup,
  72                         .lookup_data = &oif,
  73                         .result = res,
  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         } else {
  82                 err = fib6_table_lookup(net, net->ipv6.fib6_local_tbl, oif,
  83                                         fl6, res, flags);
  84                 if (err || res->f6i == net->ipv6.fib6_null_entry)
  85                         err = fib6_table_lookup(net, net->ipv6.fib6_main_tbl,
  86                                                 oif, fl6, res, flags);
  87         }
  88
  89         return err;
  90 }
  91
  92 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
  93                                    const struct sk_buff *skb,
  94                                    int flags, pol_lookup_t lookup)
  95 {
  96         if (net->ipv6.fib6_has_custom_rules) {
  97                 struct fib6_result res = {};
  98                 struct fib_lookup_arg arg = {
  99                         .lookup_ptr = lookup,
 100                         .lookup_data = skb,
 101                         .result = &res,
 102                         .flags = FIB_LOOKUP_NOREF,
 103                 };
 104
 105                 /* update flow if oif or iif point to device enslaved to l3mdev */
 106                 l3mdev_update_flow(net, flowi6_to_flowi(fl6));
 107
 108                 fib_rules_lookup(net->ipv6.fib6_rules_ops,
 109                                  flowi6_to_flowi(fl6), flags, &arg);
 110
 111                 if (res.rt6)
 112                         return &res.rt6->dst;
 113         } else {
 114                 struct rt6_info *rt;
 115
 116                 rt = pol_lookup_func(lookup,
 117                              net, net->ipv6.fib6_local_tbl, fl6, skb, flags);
 118                 if (rt != net->ipv6.ip6_null_entry && rt->dst.error != -EAGAIN)
 119                         return &rt->dst;
 120                 ip6_rt_put_flags(rt, flags);
 121                 rt = pol_lookup_func(lookup,
 122                              net, net->ipv6.fib6_main_tbl, fl6, skb, flags);
 123                 if (rt->dst.error != -EAGAIN)
 124                         return &rt->dst;
 125                 ip6_rt_put_flags(rt, flags);
 126         }
 127
 128         if (!(flags & RT6_LOOKUP_F_DST_NOREF))
 129                 dst_hold(&net->ipv6.ip6_null_entry->dst);
 130         return &net->ipv6.ip6_null_entry->dst;
 131 }
 132
 133 static int fib6_rule_saddr(struct net *net, struct fib_rule *rule, int flags,
 134                            struct flowi6 *flp6, const struct net_device *dev)
 135 {
 136         struct fib6_rule *r = (struct fib6_rule *)rule;
 137
 138         /* If we need to find a source address for this traffic,
 139          * we check the result if it meets requirement of the rule.
 140          */
 141         if ((rule->flags & FIB_RULE_FIND_SADDR) &&
 142             r->src.plen && !(flags & RT6_LOOKUP_F_HAS_SADDR)) {
 143                 struct in6_addr saddr;
 144
 145                 if (ipv6_dev_get_saddr(net, dev, &flp6->daddr,
 146                                        rt6_flags2srcprefs(flags), &saddr))
 147                         return -EAGAIN;
 148
 149                 if (!ipv6_prefix_equal(&saddr, &r->src.addr, r->src.plen))
 150                         return -EAGAIN;
 151
 152                 flp6->saddr = saddr;
 153         }
 154
 155         return 0;
 156 }
 157
 158 static int fib6_rule_action_alt(struct fib_rule *rule, struct flowi *flp,
 159                                 int flags, struct fib_lookup_arg *arg)
 160 {
 161         struct fib6_result *res = arg->result;
 162         struct flowi6 *flp6 = &flp->u.ip6;
 163         struct net *net = rule->fr_net;
 164         struct fib6_table *table;
 165         int err, *oif;
 166         u32 tb_id;
 167
 168         switch (rule->action) {
 169         case FR_ACT_TO_TBL:
 170                 break;
 171         case FR_ACT_UNREACHABLE:
 172                 return -ENETUNREACH;
 173         case FR_ACT_PROHIBIT:
 174                 return -EACCES;
 175         case FR_ACT_BLACKHOLE:
 176         default:
 177                 return -EINVAL;
 178         }
 179
 180         tb_id = fib_rule_get_table(rule, arg);
 181         table = fib6_get_table(net, tb_id);
 182         if (!table)
 183                 return -EAGAIN;
 184
 185         oif = (int *)arg->lookup_data;
 186         err = fib6_table_lookup(net, table, *oif, flp6, res, flags);
 187         if (!err && res->f6i != net->ipv6.fib6_null_entry)
 188                 err = fib6_rule_saddr(net, rule, flags, flp6,
 189                                       res->nh->fib_nh_dev);
 190         else
 191                 err = -EAGAIN;
 192
 193         return err;
 194 }
 195
 196 static int __fib6_rule_action(struct fib_rule *rule, struct flowi *flp,
 197                               int flags, struct fib_lookup_arg *arg)
 198 {
 199         struct fib6_result *res = arg->result;
 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 = pol_lookup_func(lookup,
 234                              net, table, flp6, arg->lookup_data, flags);
 235         if (rt != net->ipv6.ip6_null_entry) {
 236                 struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
 237
 238                 if (!idev)
 239                         goto again;
 240                 err = fib6_rule_saddr(net, rule, flags, flp6,
 241                                       idev->dev);
 242
 243                 if (err == -EAGAIN)
 244                         goto again;
 245
 246                 err = rt->dst.error;
 247                 if (err != -EAGAIN)
 248                         goto out;
 249         }
 250 again:
 251         ip6_rt_put_flags(rt, flags);
 252         err = -EAGAIN;
 253         rt = NULL;
 254         goto out;
 255
 256 discard_pkt:
 257         if (!(flags & RT6_LOOKUP_F_DST_NOREF))
 258                 dst_hold(&rt->dst);
 259 out:
 260         res->rt6 = rt;
 261         return err;
 262 }
 263
 264 INDIRECT_CALLABLE_SCOPE int fib6_rule_action(struct fib_rule *rule,
 265                                              struct flowi *flp, int flags,
 266                                              struct fib_lookup_arg *arg)
 267 {
 268         if (arg->lookup_ptr == fib6_table_lookup)
 269                 return fib6_rule_action_alt(rule, flp, flags, arg);
 270
 271         return __fib6_rule_action(rule, flp, flags, arg);
 272 }
 273
 274 INDIRECT_CALLABLE_SCOPE bool fib6_rule_suppress(struct fib_rule *rule,
 275                                                 int flags,
 276                                                 struct fib_lookup_arg *arg)
 277 {
 278         struct fib6_result *res = arg->result;
 279         struct rt6_info *rt = res->rt6;
 280         struct net_device *dev = NULL;
 281
 282         if (!rt)
 283                 return false;
 284
 285         if (rt->rt6i_idev)
 286                 dev = rt->rt6i_idev->dev;
 287
 288         /* do not accept result if the route does
 289          * not meet the required prefix length
 290          */
 291         if (rt->rt6i_dst.plen <= rule->suppress_prefixlen)
 292                 goto suppress_route;
 293
 294         /* do not accept result if the route uses a device
 295          * belonging to a forbidden interface group
 296          */
 297         if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
 298                 goto suppress_route;
 299
 300         return false;
 301
 302 suppress_route:
 303         ip6_rt_put_flags(rt, flags);
 304         return true;
 305 }
 306
 307 INDIRECT_CALLABLE_SCOPE int fib6_rule_match(struct fib_rule *rule,
 308                                             struct flowi *fl, int flags)
 309 {
 310         struct fib6_rule *r = (struct fib6_rule *) rule;
 311         struct flowi6 *fl6 = &fl->u.ip6;
 312
 313         if (r->dst.plen &&
 314             !ipv6_prefix_equal(&fl6->daddr, &r->dst.addr, r->dst.plen))
 315                 return 0;
 316
 317         /*
 318          * If FIB_RULE_FIND_SADDR is set and we do not have a
 319          * source address for the traffic, we defer check for
 320          * source address.
 321          */
 322         if (r->src.plen) {
 323                 if (flags & RT6_LOOKUP_F_HAS_SADDR) {
 324                         if (!ipv6_prefix_equal(&fl6->saddr, &r->src.addr,
 325                                                r->src.plen))
 326                                 return 0;
 327                 } else if (!(r->common.flags & FIB_RULE_FIND_SADDR))
 328                         return 0;
 329         }
 330
 331         if (r->dscp && r->dscp != ip6_dscp(fl6->flowlabel))
 332                 return 0;
 333
 334         if (rule->ip_proto && (rule->ip_proto != fl6->flowi6_proto))
 335                 return 0;
 336
 337         if (fib_rule_port_range_set(&rule->sport_range) &&
 338             !fib_rule_port_inrange(&rule->sport_range, fl6->fl6_sport))
 339                 return 0;
 340
 341         if (fib_rule_port_range_set(&rule->dport_range) &&
 342             !fib_rule_port_inrange(&rule->dport_range, fl6->fl6_dport))
 343                 return 0;
 344
 345         return 1;
 346 }
 347
 348 static int fib6_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 349                                struct fib_rule_hdr *frh,
 350                                struct nlattr **tb,
 351                                struct netlink_ext_ack *extack)
 352 {
 353         int err = -EINVAL;
 354         struct net *net = sock_net(skb->sk);
 355         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 356
 357         if (!inet_validate_dscp(frh->tos)) {
 358                 NL_SET_ERR_MSG(extack,
 359                                "Invalid dsfield (tos): ECN bits must be 0");
 360                 goto errout;
 361         }
 362         rule6->dscp = inet_dsfield_to_dscp(frh->tos);
 363
 364         if (rule->action == FR_ACT_TO_TBL && !rule->l3mdev) {
 365                 if (rule->table == RT6_TABLE_UNSPEC) {
 366                         NL_SET_ERR_MSG(extack, "Invalid table");
 367                         goto errout;
 368                 }
 369
 370                 if (fib6_new_table(net, rule->table) == NULL) {
 371                         err = -ENOBUFS;
 372                         goto errout;
 373                 }
 374         }
 375
 376         if (frh->src_len)
 377                 rule6->src.addr = nla_get_in6_addr(tb[FRA_SRC]);
 378
 379         if (frh->dst_len)
 380                 rule6->dst.addr = nla_get_in6_addr(tb[FRA_DST]);
 381
 382         rule6->src.plen = frh->src_len;
 383         rule6->dst.plen = frh->dst_len;
 384
 385         if (fib_rule_requires_fldissect(rule))
 386                 net->ipv6.fib6_rules_require_fldissect++;
 387
 388         net->ipv6.fib6_has_custom_rules = true;
 389         err = 0;
 390 errout:
 391         return err;
 392 }
 393
 394 static int fib6_rule_delete(struct fib_rule *rule)
 395 {
 396         struct net *net = rule->fr_net;
 397
 398         if (net->ipv6.fib6_rules_require_fldissect &&
 399             fib_rule_requires_fldissect(rule))
 400                 net->ipv6.fib6_rules_require_fldissect--;
 401
 402         return 0;
 403 }
 404
 405 static int fib6_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 406                              struct nlattr **tb)
 407 {
 408         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 409
 410         if (frh->src_len && (rule6->src.plen != frh->src_len))
 411                 return 0;
 412
 413         if (frh->dst_len && (rule6->dst.plen != frh->dst_len))
 414                 return 0;
 415
 416         if (frh->tos && inet_dscp_to_dsfield(rule6->dscp) != frh->tos)
 417                 return 0;
 418
 419         if (frh->src_len &&
 420             nla_memcmp(tb[FRA_SRC], &rule6->src.addr, sizeof(struct in6_addr)))
 421                 return 0;
 422
 423         if (frh->dst_len &&
 424             nla_memcmp(tb[FRA_DST], &rule6->dst.addr, sizeof(struct in6_addr)))
 425                 return 0;
 426
 427         return 1;
 428 }
 429
 430 static int fib6_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 431                           struct fib_rule_hdr *frh)
 432 {
 433         struct fib6_rule *rule6 = (struct fib6_rule *) rule;
 434
 435         frh->dst_len = rule6->dst.plen;
 436         frh->src_len = rule6->src.plen;
 437         frh->tos = inet_dscp_to_dsfield(rule6->dscp);
 438
 439         if ((rule6->dst.plen &&
 440              nla_put_in6_addr(skb, FRA_DST, &rule6->dst.addr)) ||
 441             (rule6->src.plen &&
 442              nla_put_in6_addr(skb, FRA_SRC, &rule6->src.addr)))
 443                 goto nla_put_failure;
 444         return 0;
 445
 446 nla_put_failure:
 447         return -ENOBUFS;
 448 }
 449
 450 static size_t fib6_rule_nlmsg_payload(struct fib_rule *rule)
 451 {
 452         return nla_total_size(16) /* dst */
 453                + nla_total_size(16); /* src */
 454 }
 455
 456 static void fib6_rule_flush_cache(struct fib_rules_ops *ops)
 457 {
 458         rt_genid_bump_ipv6(ops->fro_net);
 459 }
 460
 461 static const struct fib_rules_ops __net_initconst fib6_rules_ops_template = {
 462         .family                 = AF_INET6,
 463         .rule_size              = sizeof(struct fib6_rule),
 464         .addr_size              = sizeof(struct in6_addr),
 465         .action                 = fib6_rule_action,
 466         .match                  = fib6_rule_match,
 467         .suppress               = fib6_rule_suppress,
 468         .configure              = fib6_rule_configure,
 469         .delete                 = fib6_rule_delete,
 470         .compare                = fib6_rule_compare,
 471         .fill                   = fib6_rule_fill,
 472         .nlmsg_payload          = fib6_rule_nlmsg_payload,
 473         .flush_cache            = fib6_rule_flush_cache,
 474         .nlgroup                = RTNLGRP_IPV6_RULE,
 475         .owner                  = THIS_MODULE,
 476         .fro_net                = &init_net,
 477 };
 478
 479 static int __net_init fib6_rules_net_init(struct net *net)
 480 {
 481         struct fib_rules_ops *ops;
 482         int err;
 483
 484         ops = fib_rules_register(&fib6_rules_ops_template, net);
 485         if (IS_ERR(ops))
 486                 return PTR_ERR(ops);
 487
 488         err = fib_default_rule_add(ops, 0, RT6_TABLE_LOCAL);
 489         if (err)
 490                 goto out_fib6_rules_ops;
 491
 492         err = fib_default_rule_add(ops, 0x7FFE, RT6_TABLE_MAIN);
 493         if (err)
 494                 goto out_fib6_rules_ops;
 495
 496         net->ipv6.fib6_rules_ops = ops;
 497         net->ipv6.fib6_rules_require_fldissect = 0;
 498 out:
 499         return err;
 500
 501 out_fib6_rules_ops:
 502         fib_rules_unregister(ops);
 503         goto out;
 504 }
 505
 506 static void __net_exit fib6_rules_net_exit_batch(struct list_head *net_list)
 507 {
 508         struct net *net;
 509
 510         rtnl_lock();
 511         list_for_each_entry(net, net_list, exit_list) {
 512                 fib_rules_unregister(net->ipv6.fib6_rules_ops);
 513                 cond_resched();
 514         }
 515         rtnl_unlock();
 516 }
 517
 518 static struct pernet_operations fib6_rules_net_ops = {
 519         .init = fib6_rules_net_init,
 520         .exit_batch = fib6_rules_net_exit_batch,
 521 };
 522
 523 int __init fib6_rules_init(void)
 524 {
 525         return register_pernet_subsys(&fib6_rules_net_ops);
 526 }
 527
 528
 529 void fib6_rules_cleanup(void)
 530 {
 531         unregister_pernet_subsys(&fib6_rules_net_ops);
 532 }