net/ipv4/fib_rules.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              IPv4 Forwarding Information Base: policy rules.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *              Thomas Graf <tgraf@suug.ch>
  11  *
  12  * Fixes:
  13  *              Rani Assaf      :       local_rule cannot be deleted
  14  *              Marc Boucher    :       routing by fwmark
  15  */
  16
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netlink.h>
  21 #include <linux/inetdevice.h>
  22 #include <linux/init.h>
  23 #include <linux/list.h>
  24 #include <linux/rcupdate.h>
  25 #include <linux/export.h>
  26 #include <net/ip.h>
  27 #include <net/route.h>
  28 #include <net/tcp.h>
  29 #include <net/ip_fib.h>
  30 #include <net/nexthop.h>
  31 #include <net/fib_rules.h>
  32
  33 struct fib4_rule {
  34         struct fib_rule         common;
  35         u8                      dst_len;
  36         u8                      src_len;
  37         u8                      tos;
  38         __be32                  src;
  39         __be32                  srcmask;
  40         __be32                  dst;
  41         __be32                  dstmask;
  42 #ifdef CONFIG_IP_ROUTE_CLASSID
  43         u32                     tclassid;
  44 #endif
  45 };
  46
  47 static bool fib4_rule_matchall(const struct fib_rule *rule)
  48 {
  49         struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
  50
  51         if (r->dst_len || r->src_len || r->tos)
  52                 return false;
  53         return fib_rule_matchall(rule);
  54 }
  55
  56 bool fib4_rule_default(const struct fib_rule *rule)
  57 {
  58         if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
  59             rule->l3mdev)
  60                 return false;
  61         if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
  62             rule->table != RT_TABLE_DEFAULT)
  63                 return false;
  64         return true;
  65 }
  66 EXPORT_SYMBOL_GPL(fib4_rule_default);
  67
  68 int fib4_rules_dump(struct net *net, struct notifier_block *nb,
  69                     struct netlink_ext_ack *extack)
  70 {
  71         return fib_rules_dump(net, nb, AF_INET, extack);
  72 }
  73
  74 unsigned int fib4_rules_seq_read(struct net *net)
  75 {
  76         return fib_rules_seq_read(net, AF_INET);
  77 }
  78
  79 int __fib_lookup(struct net *net, struct flowi4 *flp,
  80                  struct fib_result *res, unsigned int flags)
  81 {
  82         struct fib_lookup_arg arg = {
  83                 .result = res,
  84                 .flags = flags,
  85         };
  86         int err;
  87
  88         /* update flow if oif or iif point to device enslaved to l3mdev */
  89         l3mdev_update_flow(net, flowi4_to_flowi(flp));
  90
  91         err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
  92 #ifdef CONFIG_IP_ROUTE_CLASSID
  93         if (arg.rule)
  94                 res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
  95         else
  96                 res->tclassid = 0;
  97 #endif
  98
  99         if (err == -ESRCH)
 100                 err = -ENETUNREACH;
 101
 102         return err;
 103 }
 104 EXPORT_SYMBOL_GPL(__fib_lookup);
 105
 106 static int fib4_rule_action(struct fib_rule *rule, struct flowi *flp,
 107                             int flags, struct fib_lookup_arg *arg)
 108 {
 109         int err = -EAGAIN;
 110         struct fib_table *tbl;
 111         u32 tb_id;
 112
 113         switch (rule->action) {
 114         case FR_ACT_TO_TBL:
 115                 break;
 116
 117         case FR_ACT_UNREACHABLE:
 118                 return -ENETUNREACH;
 119
 120         case FR_ACT_PROHIBIT:
 121                 return -EACCES;
 122
 123         case FR_ACT_BLACKHOLE:
 124         default:
 125                 return -EINVAL;
 126         }
 127
 128         rcu_read_lock();
 129
 130         tb_id = fib_rule_get_table(rule, arg);
 131         tbl = fib_get_table(rule->fr_net, tb_id);
 132         if (tbl)
 133                 err = fib_table_lookup(tbl, &flp->u.ip4,
 134                                        (struct fib_result *)arg->result,
 135                                        arg->flags);
 136
 137         rcu_read_unlock();
 138         return err;
 139 }
 140
 141 static bool fib4_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg)
 142 {
 143         struct fib_result *result = (struct fib_result *) arg->result;
 144         struct net_device *dev = NULL;
 145
 146         if (result->fi) {
 147                 struct fib_nh_common *nhc = fib_info_nhc(result->fi, 0);
 148
 149                 dev = nhc->nhc_dev;
 150         }
 151
 152         /* do not accept result if the route does
 153          * not meet the required prefix length
 154          */
 155         if (result->prefixlen <= rule->suppress_prefixlen)
 156                 goto suppress_route;
 157
 158         /* do not accept result if the route uses a device
 159          * belonging to a forbidden interface group
 160          */
 161         if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
 162                 goto suppress_route;
 163
 164         return false;
 165
 166 suppress_route:
 167         if (!(arg->flags & FIB_LOOKUP_NOREF))
 168                 fib_info_put(result->fi);
 169         return true;
 170 }
 171
 172 static int fib4_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 173 {
 174         struct fib4_rule *r = (struct fib4_rule *) rule;
 175         struct flowi4 *fl4 = &fl->u.ip4;
 176         __be32 daddr = fl4->daddr;
 177         __be32 saddr = fl4->saddr;
 178
 179         if (((saddr ^ r->src) & r->srcmask) ||
 180             ((daddr ^ r->dst) & r->dstmask))
 181                 return 0;
 182
 183         if (r->tos && (r->tos != fl4->flowi4_tos))
 184                 return 0;
 185
 186         if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
 187                 return 0;
 188
 189         if (fib_rule_port_range_set(&rule->sport_range) &&
 190             !fib_rule_port_inrange(&rule->sport_range, fl4->fl4_sport))
 191                 return 0;
 192
 193         if (fib_rule_port_range_set(&rule->dport_range) &&
 194             !fib_rule_port_inrange(&rule->dport_range, fl4->fl4_dport))
 195                 return 0;
 196
 197         return 1;
 198 }
 199
 200 static struct fib_table *fib_empty_table(struct net *net)
 201 {
 202         u32 id = 1;
 203
 204         while (1) {
 205                 if (!fib_get_table(net, id))
 206                         return fib_new_table(net, id);
 207
 208                 if (id++ == RT_TABLE_MAX)
 209                         break;
 210         }
 211         return NULL;
 212 }
 213
 214 static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = {
 215         FRA_GENERIC_POLICY,
 216         [FRA_FLOW]      = { .type = NLA_U32 },
 217 };
 218
 219 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 220                                struct fib_rule_hdr *frh,
 221                                struct nlattr **tb,
 222                                struct netlink_ext_ack *extack)
 223 {
 224         struct net *net = sock_net(skb->sk);
 225         int err = -EINVAL;
 226         struct fib4_rule *rule4 = (struct fib4_rule *) rule;
 227
 228         if (frh->tos & ~IPTOS_TOS_MASK) {
 229                 NL_SET_ERR_MSG(extack, "Invalid tos");
 230                 goto errout;
 231         }
 232
 233         /* split local/main if they are not already split */
 234         err = fib_unmerge(net);
 235         if (err)
 236                 goto errout;
 237
 238         if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
 239                 if (rule->action == FR_ACT_TO_TBL) {
 240                         struct fib_table *table;
 241
 242                         table = fib_empty_table(net);
 243                         if (!table) {
 244                                 err = -ENOBUFS;
 245                                 goto errout;
 246                         }
 247
 248                         rule->table = table->tb_id;
 249                 }
 250         }
 251
 252         if (frh->src_len)
 253                 rule4->src = nla_get_in_addr(tb[FRA_SRC]);
 254
 255         if (frh->dst_len)
 256                 rule4->dst = nla_get_in_addr(tb[FRA_DST]);
 257
 258 #ifdef CONFIG_IP_ROUTE_CLASSID
 259         if (tb[FRA_FLOW]) {
 260                 rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
 261                 if (rule4->tclassid)
 262                         net->ipv4.fib_num_tclassid_users++;
 263         }
 264 #endif
 265
 266         if (fib_rule_requires_fldissect(rule))
 267                 net->ipv4.fib_rules_require_fldissect++;
 268
 269         rule4->src_len = frh->src_len;
 270         rule4->srcmask = inet_make_mask(rule4->src_len);
 271         rule4->dst_len = frh->dst_len;
 272         rule4->dstmask = inet_make_mask(rule4->dst_len);
 273         rule4->tos = frh->tos;
 274
 275         net->ipv4.fib_has_custom_rules = true;
 276
 277         err = 0;
 278 errout:
 279         return err;
 280 }
 281
 282 static int fib4_rule_delete(struct fib_rule *rule)
 283 {
 284         struct net *net = rule->fr_net;
 285         int err;
 286
 287         /* split local/main if they are not already split */
 288         err = fib_unmerge(net);
 289         if (err)
 290                 goto errout;
 291
 292 #ifdef CONFIG_IP_ROUTE_CLASSID
 293         if (((struct fib4_rule *)rule)->tclassid)
 294                 net->ipv4.fib_num_tclassid_users--;
 295 #endif
 296         net->ipv4.fib_has_custom_rules = true;
 297
 298         if (net->ipv4.fib_rules_require_fldissect &&
 299             fib_rule_requires_fldissect(rule))
 300                 net->ipv4.fib_rules_require_fldissect--;
 301 errout:
 302         return err;
 303 }
 304
 305 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 306                              struct nlattr **tb)
 307 {
 308         struct fib4_rule *rule4 = (struct fib4_rule *) rule;
 309
 310         if (frh->src_len && (rule4->src_len != frh->src_len))
 311                 return 0;
 312
 313         if (frh->dst_len && (rule4->dst_len != frh->dst_len))
 314                 return 0;
 315
 316         if (frh->tos && (rule4->tos != frh->tos))
 317                 return 0;
 318
 319 #ifdef CONFIG_IP_ROUTE_CLASSID
 320         if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
 321                 return 0;
 322 #endif
 323
 324         if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
 325                 return 0;
 326
 327         if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
 328                 return 0;
 329
 330         return 1;
 331 }
 332
 333 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 334                           struct fib_rule_hdr *frh)
 335 {
 336         struct fib4_rule *rule4 = (struct fib4_rule *) rule;
 337
 338         frh->dst_len = rule4->dst_len;
 339         frh->src_len = rule4->src_len;
 340         frh->tos = rule4->tos;
 341
 342         if ((rule4->dst_len &&
 343              nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
 344             (rule4->src_len &&
 345              nla_put_in_addr(skb, FRA_SRC, rule4->src)))
 346                 goto nla_put_failure;
 347 #ifdef CONFIG_IP_ROUTE_CLASSID
 348         if (rule4->tclassid &&
 349             nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
 350                 goto nla_put_failure;
 351 #endif
 352         return 0;
 353
 354 nla_put_failure:
 355         return -ENOBUFS;
 356 }
 357
 358 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
 359 {
 360         return nla_total_size(4) /* dst */
 361                + nla_total_size(4) /* src */
 362                + nla_total_size(4); /* flow */
 363 }
 364
 365 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
 366 {
 367         rt_cache_flush(ops->fro_net);
 368 }
 369
 370 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
 371         .family         = AF_INET,
 372         .rule_size      = sizeof(struct fib4_rule),
 373         .addr_size      = sizeof(u32),
 374         .action         = fib4_rule_action,
 375         .suppress       = fib4_rule_suppress,
 376         .match          = fib4_rule_match,
 377         .configure      = fib4_rule_configure,
 378         .delete         = fib4_rule_delete,
 379         .compare        = fib4_rule_compare,
 380         .fill           = fib4_rule_fill,
 381         .nlmsg_payload  = fib4_rule_nlmsg_payload,
 382         .flush_cache    = fib4_rule_flush_cache,
 383         .nlgroup        = RTNLGRP_IPV4_RULE,
 384         .policy         = fib4_rule_policy,
 385         .owner          = THIS_MODULE,
 386 };
 387
 388 static int fib_default_rules_init(struct fib_rules_ops *ops)
 389 {
 390         int err;
 391
 392         err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0);
 393         if (err < 0)
 394                 return err;
 395         err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0);
 396         if (err < 0)
 397                 return err;
 398         err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0);
 399         if (err < 0)
 400                 return err;
 401         return 0;
 402 }
 403
 404 int __net_init fib4_rules_init(struct net *net)
 405 {
 406         int err;
 407         struct fib_rules_ops *ops;
 408
 409         ops = fib_rules_register(&fib4_rules_ops_template, net);
 410         if (IS_ERR(ops))
 411                 return PTR_ERR(ops);
 412
 413         err = fib_default_rules_init(ops);
 414         if (err < 0)
 415                 goto fail;
 416         net->ipv4.rules_ops = ops;
 417         net->ipv4.fib_has_custom_rules = false;
 418         net->ipv4.fib_rules_require_fldissect = 0;
 419         return 0;
 420
 421 fail:
 422         /* also cleans all rules already added */
 423         fib_rules_unregister(ops);
 424         return err;
 425 }
 426
 427 void __net_exit fib4_rules_exit(struct net *net)
 428 {
 429         fib_rules_unregister(net->ipv4.rules_ops);
 430 }