1 /* 2 * net/core/fib_rules.c Generic Routing Rules 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public License as 6 * published by the Free Software Foundation, version 2. 7 * 8 * Authors: Thomas Graf <tgraf@suug.ch> 9 */ 10 11 #include <linux/types.h> 12 #include <linux/kernel.h> 13 #include <linux/slab.h> 14 #include <linux/list.h> 15 #include <linux/module.h> 16 #include <net/net_namespace.h> 17 #include <net/sock.h> 18 #include <net/fib_rules.h> 19 #include <net/ip_tunnels.h> 20 21 static const struct fib_kuid_range fib_kuid_range_unset = { 22 KUIDT_INIT(0), 23 KUIDT_INIT(~0), 24 }; 25 26 bool fib_rule_matchall(const struct fib_rule *rule) 27 { 28 if (rule->iifindex || rule->oifindex || rule->mark || rule->tun_id || 29 rule->flags) 30 return false; 31 if (rule->suppress_ifgroup != -1 || rule->suppress_prefixlen != -1) 32 return false; 33 if (!uid_eq(rule->uid_range.start, fib_kuid_range_unset.start) || 34 !uid_eq(rule->uid_range.end, fib_kuid_range_unset.end)) 35 return false; 36 if (fib_rule_port_range_set(&rule->sport_range)) 37 return false; 38 if (fib_rule_port_range_set(&rule->dport_range)) 39 return false; 40 return true; 41 } 42 EXPORT_SYMBOL_GPL(fib_rule_matchall); 43 44 int fib_default_rule_add(struct fib_rules_ops *ops, 45 u32 pref, u32 table, u32 flags) 46 { 47 struct fib_rule *r; 48 49 r = kzalloc(ops->rule_size, GFP_KERNEL); 50 if (r == NULL) 51 return -ENOMEM; 52 53 refcount_set(&r->refcnt, 1); 54 r->action = FR_ACT_TO_TBL; 55 r->pref = pref; 56 r->table = table; 57 r->flags = flags; 58 r->proto = RTPROT_KERNEL; 59 r->fr_net = ops->fro_net; 60 r->uid_range = fib_kuid_range_unset; 61 62 r->suppress_prefixlen = -1; 63 r->suppress_ifgroup = -1; 64 65 /* The lock is not required here, the list in unreacheable 66 * at the moment this function is called */ 67 list_add_tail(&r->list, &ops->rules_list); 68 return 0; 69 } 70 EXPORT_SYMBOL(fib_default_rule_add); 71 72 static u32 fib_default_rule_pref(struct fib_rules_ops *ops) 73 { 74 struct list_head *pos; 75 struct fib_rule *rule; 76 77 if (!list_empty(&ops->rules_list)) { 78 pos = ops->rules_list.next; 79 if (pos->next != &ops->rules_list) { 80 rule = list_entry(pos->next, struct fib_rule, list); 81 if (rule->pref) 82 return rule->pref - 1; 83 } 84 } 85 86 return 0; 87 } 88 89 static void notify_rule_change(int event, struct fib_rule *rule, 90 struct fib_rules_ops *ops, struct nlmsghdr *nlh, 91 u32 pid); 92 93 static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family) 94 { 95 struct fib_rules_ops *ops; 96 97 rcu_read_lock(); 98 list_for_each_entry_rcu(ops, &net->rules_ops, list) { 99 if (ops->family == family) { 100 if (!try_module_get(ops->owner)) 101 ops = NULL; 102 rcu_read_unlock(); 103 return ops; 104 } 105 } 106 rcu_read_unlock(); 107 108 return NULL; 109 } 110 111 static void rules_ops_put(struct fib_rules_ops *ops) 112 { 113 if (ops) 114 module_put(ops->owner); 115 } 116 117 static void flush_route_cache(struct fib_rules_ops *ops) 118 { 119 if (ops->flush_cache) 120 ops->flush_cache(ops); 121 } 122 123 static int __fib_rules_register(struct fib_rules_ops *ops) 124 { 125 int err = -EEXIST; 126 struct fib_rules_ops *o; 127 struct net *net; 128 129 net = ops->fro_net; 130 131 if (ops->rule_size < sizeof(struct fib_rule)) 132 return -EINVAL; 133 134 if (ops->match == NULL || ops->configure == NULL || 135 ops->compare == NULL || ops->fill == NULL || 136 ops->action == NULL) 137 return -EINVAL; 138 139 spin_lock(&net->rules_mod_lock); 140 list_for_each_entry(o, &net->rules_ops, list) 141 if (ops->family == o->family) 142 goto errout; 143 144 list_add_tail_rcu(&ops->list, &net->rules_ops); 145 err = 0; 146 errout: 147 spin_unlock(&net->rules_mod_lock); 148 149 return err; 150 } 151 152 struct fib_rules_ops * 153 fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net) 154 { 155 struct fib_rules_ops *ops; 156 int err; 157 158 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL); 159 if (ops == NULL) 160 return ERR_PTR(-ENOMEM); 161 162 INIT_LIST_HEAD(&ops->rules_list); 163 ops->fro_net = net; 164 165 err = __fib_rules_register(ops); 166 if (err) { 167 kfree(ops); 168 ops = ERR_PTR(err); 169 } 170 171 return ops; 172 } 173 EXPORT_SYMBOL_GPL(fib_rules_register); 174 175 static void fib_rules_cleanup_ops(struct fib_rules_ops *ops) 176 { 177 struct fib_rule *rule, *tmp; 178 179 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) { 180 list_del_rcu(&rule->list); 181 if (ops->delete) 182 ops->delete(rule); 183 fib_rule_put(rule); 184 } 185 } 186 187 void fib_rules_unregister(struct fib_rules_ops *ops) 188 { 189 struct net *net = ops->fro_net; 190 191 spin_lock(&net->rules_mod_lock); 192 list_del_rcu(&ops->list); 193 spin_unlock(&net->rules_mod_lock); 194 195 fib_rules_cleanup_ops(ops); 196 kfree_rcu(ops, rcu); 197 } 198 EXPORT_SYMBOL_GPL(fib_rules_unregister); 199 200 static int uid_range_set(struct fib_kuid_range *range) 201 { 202 return uid_valid(range->start) && uid_valid(range->end); 203 } 204 205 static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb) 206 { 207 struct fib_rule_uid_range *in; 208 struct fib_kuid_range out; 209 210 in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]); 211 212 out.start = make_kuid(current_user_ns(), in->start); 213 out.end = make_kuid(current_user_ns(), in->end); 214 215 return out; 216 } 217 218 static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range) 219 { 220 struct fib_rule_uid_range out = { 221 from_kuid_munged(current_user_ns(), range->start), 222 from_kuid_munged(current_user_ns(), range->end) 223 }; 224 225 return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out); 226 } 227 228 static int nla_get_port_range(struct nlattr *pattr, 229 struct fib_rule_port_range *port_range) 230 { 231 const struct fib_rule_port_range *pr = nla_data(pattr); 232 233 if (!fib_rule_port_range_valid(pr)) 234 return -EINVAL; 235 236 port_range->start = pr->start; 237 port_range->end = pr->end; 238 239 return 0; 240 } 241 242 static int nla_put_port_range(struct sk_buff *skb, int attrtype, 243 struct fib_rule_port_range *range) 244 { 245 return nla_put(skb, attrtype, sizeof(*range), range); 246 } 247 248 static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops, 249 struct flowi *fl, int flags, 250 struct fib_lookup_arg *arg) 251 { 252 int ret = 0; 253 254 if (rule->iifindex && (rule->iifindex != fl->flowi_iif)) 255 goto out; 256 257 if (rule->oifindex && (rule->oifindex != fl->flowi_oif)) 258 goto out; 259 260 if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask) 261 goto out; 262 263 if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id)) 264 goto out; 265 266 if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg)) 267 goto out; 268 269 if (uid_lt(fl->flowi_uid, rule->uid_range.start) || 270 uid_gt(fl->flowi_uid, rule->uid_range.end)) 271 goto out; 272 273 ret = ops->match(rule, fl, flags); 274 out: 275 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret; 276 } 277 278 int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl, 279 int flags, struct fib_lookup_arg *arg) 280 { 281 struct fib_rule *rule; 282 int err; 283 284 rcu_read_lock(); 285 286 list_for_each_entry_rcu(rule, &ops->rules_list, list) { 287 jumped: 288 if (!fib_rule_match(rule, ops, fl, flags, arg)) 289 continue; 290 291 if (rule->action == FR_ACT_GOTO) { 292 struct fib_rule *target; 293 294 target = rcu_dereference(rule->ctarget); 295 if (target == NULL) { 296 continue; 297 } else { 298 rule = target; 299 goto jumped; 300 } 301 } else if (rule->action == FR_ACT_NOP) 302 continue; 303 else 304 err = ops->action(rule, fl, flags, arg); 305 306 if (!err && ops->suppress && ops->suppress(rule, arg)) 307 continue; 308 309 if (err != -EAGAIN) { 310 if ((arg->flags & FIB_LOOKUP_NOREF) || 311 likely(refcount_inc_not_zero(&rule->refcnt))) { 312 arg->rule = rule; 313 goto out; 314 } 315 break; 316 } 317 } 318 319 err = -ESRCH; 320 out: 321 rcu_read_unlock(); 322 323 return err; 324 } 325 EXPORT_SYMBOL_GPL(fib_rules_lookup); 326 327 static int call_fib_rule_notifier(struct notifier_block *nb, struct net *net, 328 enum fib_event_type event_type, 329 struct fib_rule *rule, int family) 330 { 331 struct fib_rule_notifier_info info = { 332 .info.family = family, 333 .rule = rule, 334 }; 335 336 return call_fib_notifier(nb, net, event_type, &info.info); 337 } 338 339 static int call_fib_rule_notifiers(struct net *net, 340 enum fib_event_type event_type, 341 struct fib_rule *rule, 342 struct fib_rules_ops *ops, 343 struct netlink_ext_ack *extack) 344 { 345 struct fib_rule_notifier_info info = { 346 .info.family = ops->family, 347 .info.extack = extack, 348 .rule = rule, 349 }; 350 351 ops->fib_rules_seq++; 352 return call_fib_notifiers(net, event_type, &info.info); 353 } 354 355 /* Called with rcu_read_lock() */ 356 int fib_rules_dump(struct net *net, struct notifier_block *nb, int family) 357 { 358 struct fib_rules_ops *ops; 359 struct fib_rule *rule; 360 361 ops = lookup_rules_ops(net, family); 362 if (!ops) 363 return -EAFNOSUPPORT; 364 list_for_each_entry_rcu(rule, &ops->rules_list, list) 365 call_fib_rule_notifier(nb, net, FIB_EVENT_RULE_ADD, rule, 366 family); 367 rules_ops_put(ops); 368 369 return 0; 370 } 371 EXPORT_SYMBOL_GPL(fib_rules_dump); 372 373 unsigned int fib_rules_seq_read(struct net *net, int family) 374 { 375 unsigned int fib_rules_seq; 376 struct fib_rules_ops *ops; 377 378 ASSERT_RTNL(); 379 380 ops = lookup_rules_ops(net, family); 381 if (!ops) 382 return 0; 383 fib_rules_seq = ops->fib_rules_seq; 384 rules_ops_put(ops); 385 386 return fib_rules_seq; 387 } 388 EXPORT_SYMBOL_GPL(fib_rules_seq_read); 389 390 static struct fib_rule *rule_find(struct fib_rules_ops *ops, 391 struct fib_rule_hdr *frh, 392 struct nlattr **tb, 393 struct fib_rule *rule, 394 bool user_priority) 395 { 396 struct fib_rule *r; 397 398 list_for_each_entry(r, &ops->rules_list, list) { 399 if (rule->action && r->action != rule->action) 400 continue; 401 402 if (rule->table && r->table != rule->table) 403 continue; 404 405 if (user_priority && r->pref != rule->pref) 406 continue; 407 408 if (rule->iifname[0] && 409 memcmp(r->iifname, rule->iifname, IFNAMSIZ)) 410 continue; 411 412 if (rule->oifname[0] && 413 memcmp(r->oifname, rule->oifname, IFNAMSIZ)) 414 continue; 415 416 if (rule->mark && r->mark != rule->mark) 417 continue; 418 419 if (rule->mark_mask && r->mark_mask != rule->mark_mask) 420 continue; 421 422 if (rule->tun_id && r->tun_id != rule->tun_id) 423 continue; 424 425 if (r->fr_net != rule->fr_net) 426 continue; 427 428 if (rule->l3mdev && r->l3mdev != rule->l3mdev) 429 continue; 430 431 if (uid_range_set(&rule->uid_range) && 432 (!uid_eq(r->uid_range.start, rule->uid_range.start) || 433 !uid_eq(r->uid_range.end, rule->uid_range.end))) 434 continue; 435 436 if (rule->ip_proto && r->ip_proto != rule->ip_proto) 437 continue; 438 439 if (fib_rule_port_range_set(&rule->sport_range) && 440 !fib_rule_port_range_compare(&r->sport_range, 441 &rule->sport_range)) 442 continue; 443 444 if (fib_rule_port_range_set(&rule->dport_range) && 445 !fib_rule_port_range_compare(&r->dport_range, 446 &rule->dport_range)) 447 continue; 448 449 if (!ops->compare(r, frh, tb)) 450 continue; 451 return r; 452 } 453 454 return NULL; 455 } 456 457 static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh, 458 struct netlink_ext_ack *extack, 459 struct fib_rules_ops *ops, 460 struct nlattr *tb[], 461 struct fib_rule **rule, 462 bool *user_priority) 463 { 464 struct net *net = sock_net(skb->sk); 465 struct fib_rule_hdr *frh = nlmsg_data(nlh); 466 struct fib_rule *nlrule = NULL; 467 int err = -EINVAL; 468 469 if (frh->src_len) 470 if (!tb[FRA_SRC] || 471 frh->src_len > (ops->addr_size * 8) || 472 nla_len(tb[FRA_SRC]) != ops->addr_size) { 473 NL_SET_ERR_MSG(extack, "Invalid source address"); 474 goto errout; 475 } 476 477 if (frh->dst_len) 478 if (!tb[FRA_DST] || 479 frh->dst_len > (ops->addr_size * 8) || 480 nla_len(tb[FRA_DST]) != ops->addr_size) { 481 NL_SET_ERR_MSG(extack, "Invalid dst address"); 482 goto errout; 483 } 484 485 nlrule = kzalloc(ops->rule_size, GFP_KERNEL); 486 if (!nlrule) { 487 err = -ENOMEM; 488 goto errout; 489 } 490 refcount_set(&nlrule->refcnt, 1); 491 nlrule->fr_net = net; 492 493 if (tb[FRA_PRIORITY]) { 494 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]); 495 *user_priority = true; 496 } else { 497 nlrule->pref = fib_default_rule_pref(ops); 498 } 499 500 nlrule->proto = tb[FRA_PROTOCOL] ? 501 nla_get_u8(tb[FRA_PROTOCOL]) : RTPROT_UNSPEC; 502 503 if (tb[FRA_IIFNAME]) { 504 struct net_device *dev; 505 506 nlrule->iifindex = -1; 507 nla_strlcpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ); 508 dev = __dev_get_by_name(net, nlrule->iifname); 509 if (dev) 510 nlrule->iifindex = dev->ifindex; 511 } 512 513 if (tb[FRA_OIFNAME]) { 514 struct net_device *dev; 515 516 nlrule->oifindex = -1; 517 nla_strlcpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ); 518 dev = __dev_get_by_name(net, nlrule->oifname); 519 if (dev) 520 nlrule->oifindex = dev->ifindex; 521 } 522 523 if (tb[FRA_FWMARK]) { 524 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]); 525 if (nlrule->mark) 526 /* compatibility: if the mark value is non-zero all bits 527 * are compared unless a mask is explicitly specified. 528 */ 529 nlrule->mark_mask = 0xFFFFFFFF; 530 } 531 532 if (tb[FRA_FWMASK]) 533 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]); 534 535 if (tb[FRA_TUN_ID]) 536 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]); 537 538 err = -EINVAL; 539 if (tb[FRA_L3MDEV]) { 540 #ifdef CONFIG_NET_L3_MASTER_DEV 541 nlrule->l3mdev = nla_get_u8(tb[FRA_L3MDEV]); 542 if (nlrule->l3mdev != 1) 543 #endif 544 NL_SET_ERR_MSG(extack, "Invalid l3mdev"); 545 goto errout_free; 546 } 547 548 nlrule->action = frh->action; 549 nlrule->flags = frh->flags; 550 nlrule->table = frh_get_table(frh, tb); 551 if (tb[FRA_SUPPRESS_PREFIXLEN]) 552 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]); 553 else 554 nlrule->suppress_prefixlen = -1; 555 556 if (tb[FRA_SUPPRESS_IFGROUP]) 557 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]); 558 else 559 nlrule->suppress_ifgroup = -1; 560 561 if (tb[FRA_GOTO]) { 562 if (nlrule->action != FR_ACT_GOTO) { 563 NL_SET_ERR_MSG(extack, "Unexpected goto"); 564 goto errout_free; 565 } 566 567 nlrule->target = nla_get_u32(tb[FRA_GOTO]); 568 /* Backward jumps are prohibited to avoid endless loops */ 569 if (nlrule->target <= nlrule->pref) { 570 NL_SET_ERR_MSG(extack, "Backward goto not supported"); 571 goto errout_free; 572 } 573 } else if (nlrule->action == FR_ACT_GOTO) { 574 NL_SET_ERR_MSG(extack, "Missing goto target for action goto"); 575 goto errout_free; 576 } 577 578 if (nlrule->l3mdev && nlrule->table) { 579 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive"); 580 goto errout_free; 581 } 582 583 if (tb[FRA_UID_RANGE]) { 584 if (current_user_ns() != net->user_ns) { 585 err = -EPERM; 586 NL_SET_ERR_MSG(extack, "No permission to set uid"); 587 goto errout_free; 588 } 589 590 nlrule->uid_range = nla_get_kuid_range(tb); 591 592 if (!uid_range_set(&nlrule->uid_range) || 593 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) { 594 NL_SET_ERR_MSG(extack, "Invalid uid range"); 595 goto errout_free; 596 } 597 } else { 598 nlrule->uid_range = fib_kuid_range_unset; 599 } 600 601 if (tb[FRA_IP_PROTO]) 602 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]); 603 604 if (tb[FRA_SPORT_RANGE]) { 605 err = nla_get_port_range(tb[FRA_SPORT_RANGE], 606 &nlrule->sport_range); 607 if (err) { 608 NL_SET_ERR_MSG(extack, "Invalid sport range"); 609 goto errout_free; 610 } 611 } 612 613 if (tb[FRA_DPORT_RANGE]) { 614 err = nla_get_port_range(tb[FRA_DPORT_RANGE], 615 &nlrule->dport_range); 616 if (err) { 617 NL_SET_ERR_MSG(extack, "Invalid dport range"); 618 goto errout_free; 619 } 620 } 621 622 *rule = nlrule; 623 624 return 0; 625 626 errout_free: 627 kfree(nlrule); 628 errout: 629 return err; 630 } 631 632 int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh, 633 struct netlink_ext_ack *extack) 634 { 635 struct net *net = sock_net(skb->sk); 636 struct fib_rule_hdr *frh = nlmsg_data(nlh); 637 struct fib_rules_ops *ops = NULL; 638 struct fib_rule *rule = NULL, *r, *last = NULL; 639 struct nlattr *tb[FRA_MAX + 1]; 640 int err = -EINVAL, unresolved = 0; 641 bool user_priority = false; 642 643 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) { 644 NL_SET_ERR_MSG(extack, "Invalid msg length"); 645 goto errout; 646 } 647 648 ops = lookup_rules_ops(net, frh->family); 649 if (!ops) { 650 err = -EAFNOSUPPORT; 651 NL_SET_ERR_MSG(extack, "Rule family not supported"); 652 goto errout; 653 } 654 655 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy, extack); 656 if (err < 0) { 657 NL_SET_ERR_MSG(extack, "Error parsing msg"); 658 goto errout; 659 } 660 661 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority); 662 if (err) 663 goto errout; 664 665 if ((nlh->nlmsg_flags & NLM_F_EXCL) && 666 rule_find(ops, frh, tb, rule, user_priority)) { 667 err = -EEXIST; 668 goto errout_free; 669 } 670 671 err = ops->configure(rule, skb, frh, tb, extack); 672 if (err < 0) 673 goto errout_free; 674 675 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops, 676 extack); 677 if (err < 0) 678 goto errout_free; 679 680 list_for_each_entry(r, &ops->rules_list, list) { 681 if (r->pref == rule->target) { 682 RCU_INIT_POINTER(rule->ctarget, r); 683 break; 684 } 685 } 686 687 if (rcu_dereference_protected(rule->ctarget, 1) == NULL) 688 unresolved = 1; 689 690 list_for_each_entry(r, &ops->rules_list, list) { 691 if (r->pref > rule->pref) 692 break; 693 last = r; 694 } 695 696 if (last) 697 list_add_rcu(&rule->list, &last->list); 698 else 699 list_add_rcu(&rule->list, &ops->rules_list); 700 701 if (ops->unresolved_rules) { 702 /* 703 * There are unresolved goto rules in the list, check if 704 * any of them are pointing to this new rule. 705 */ 706 list_for_each_entry(r, &ops->rules_list, list) { 707 if (r->action == FR_ACT_GOTO && 708 r->target == rule->pref && 709 rtnl_dereference(r->ctarget) == NULL) { 710 rcu_assign_pointer(r->ctarget, rule); 711 if (--ops->unresolved_rules == 0) 712 break; 713 } 714 } 715 } 716 717 if (rule->action == FR_ACT_GOTO) 718 ops->nr_goto_rules++; 719 720 if (unresolved) 721 ops->unresolved_rules++; 722 723 if (rule->tun_id) 724 ip_tunnel_need_metadata(); 725 726 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid); 727 flush_route_cache(ops); 728 rules_ops_put(ops); 729 return 0; 730 731 errout_free: 732 kfree(rule); 733 errout: 734 rules_ops_put(ops); 735 return err; 736 } 737 EXPORT_SYMBOL_GPL(fib_nl_newrule); 738 739 int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh, 740 struct netlink_ext_ack *extack) 741 { 742 struct net *net = sock_net(skb->sk); 743 struct fib_rule_hdr *frh = nlmsg_data(nlh); 744 struct fib_rules_ops *ops = NULL; 745 struct fib_rule *rule = NULL, *r, *nlrule = NULL; 746 struct nlattr *tb[FRA_MAX+1]; 747 int err = -EINVAL; 748 bool user_priority = false; 749 750 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) { 751 NL_SET_ERR_MSG(extack, "Invalid msg length"); 752 goto errout; 753 } 754 755 ops = lookup_rules_ops(net, frh->family); 756 if (ops == NULL) { 757 err = -EAFNOSUPPORT; 758 NL_SET_ERR_MSG(extack, "Rule family not supported"); 759 goto errout; 760 } 761 762 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy, extack); 763 if (err < 0) { 764 NL_SET_ERR_MSG(extack, "Error parsing msg"); 765 goto errout; 766 } 767 768 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority); 769 if (err) 770 goto errout; 771 772 rule = rule_find(ops, frh, tb, nlrule, user_priority); 773 if (!rule) { 774 err = -ENOENT; 775 goto errout; 776 } 777 778 if (rule->flags & FIB_RULE_PERMANENT) { 779 err = -EPERM; 780 goto errout; 781 } 782 783 if (ops->delete) { 784 err = ops->delete(rule); 785 if (err) 786 goto errout; 787 } 788 789 if (rule->tun_id) 790 ip_tunnel_unneed_metadata(); 791 792 list_del_rcu(&rule->list); 793 794 if (rule->action == FR_ACT_GOTO) { 795 ops->nr_goto_rules--; 796 if (rtnl_dereference(rule->ctarget) == NULL) 797 ops->unresolved_rules--; 798 } 799 800 /* 801 * Check if this rule is a target to any of them. If so, 802 * adjust to the next one with the same preference or 803 * disable them. As this operation is eventually very 804 * expensive, it is only performed if goto rules, except 805 * current if it is goto rule, have actually been added. 806 */ 807 if (ops->nr_goto_rules > 0) { 808 struct fib_rule *n; 809 810 n = list_next_entry(rule, list); 811 if (&n->list == &ops->rules_list || n->pref != rule->pref) 812 n = NULL; 813 list_for_each_entry(r, &ops->rules_list, list) { 814 if (rtnl_dereference(r->ctarget) != rule) 815 continue; 816 rcu_assign_pointer(r->ctarget, n); 817 if (!n) 818 ops->unresolved_rules++; 819 } 820 } 821 822 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops, 823 NULL); 824 notify_rule_change(RTM_DELRULE, rule, ops, nlh, 825 NETLINK_CB(skb).portid); 826 fib_rule_put(rule); 827 flush_route_cache(ops); 828 rules_ops_put(ops); 829 kfree(nlrule); 830 return 0; 831 832 errout: 833 if (nlrule) 834 kfree(nlrule); 835 rules_ops_put(ops); 836 return err; 837 } 838 EXPORT_SYMBOL_GPL(fib_nl_delrule); 839 840 static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops, 841 struct fib_rule *rule) 842 { 843 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr)) 844 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */ 845 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */ 846 + nla_total_size(4) /* FRA_PRIORITY */ 847 + nla_total_size(4) /* FRA_TABLE */ 848 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */ 849 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */ 850 + nla_total_size(4) /* FRA_FWMARK */ 851 + nla_total_size(4) /* FRA_FWMASK */ 852 + nla_total_size_64bit(8) /* FRA_TUN_ID */ 853 + nla_total_size(sizeof(struct fib_kuid_range)) 854 + nla_total_size(1) /* FRA_PROTOCOL */ 855 + nla_total_size(1) /* FRA_IP_PROTO */ 856 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */ 857 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */ 858 859 if (ops->nlmsg_payload) 860 payload += ops->nlmsg_payload(rule); 861 862 return payload; 863 } 864 865 static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule, 866 u32 pid, u32 seq, int type, int flags, 867 struct fib_rules_ops *ops) 868 { 869 struct nlmsghdr *nlh; 870 struct fib_rule_hdr *frh; 871 872 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags); 873 if (nlh == NULL) 874 return -EMSGSIZE; 875 876 frh = nlmsg_data(nlh); 877 frh->family = ops->family; 878 frh->table = rule->table; 879 if (nla_put_u32(skb, FRA_TABLE, rule->table)) 880 goto nla_put_failure; 881 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen)) 882 goto nla_put_failure; 883 frh->res1 = 0; 884 frh->res2 = 0; 885 frh->action = rule->action; 886 frh->flags = rule->flags; 887 888 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto)) 889 goto nla_put_failure; 890 891 if (rule->action == FR_ACT_GOTO && 892 rcu_access_pointer(rule->ctarget) == NULL) 893 frh->flags |= FIB_RULE_UNRESOLVED; 894 895 if (rule->iifname[0]) { 896 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname)) 897 goto nla_put_failure; 898 if (rule->iifindex == -1) 899 frh->flags |= FIB_RULE_IIF_DETACHED; 900 } 901 902 if (rule->oifname[0]) { 903 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname)) 904 goto nla_put_failure; 905 if (rule->oifindex == -1) 906 frh->flags |= FIB_RULE_OIF_DETACHED; 907 } 908 909 if ((rule->pref && 910 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) || 911 (rule->mark && 912 nla_put_u32(skb, FRA_FWMARK, rule->mark)) || 913 ((rule->mark_mask || rule->mark) && 914 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) || 915 (rule->target && 916 nla_put_u32(skb, FRA_GOTO, rule->target)) || 917 (rule->tun_id && 918 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) || 919 (rule->l3mdev && 920 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) || 921 (uid_range_set(&rule->uid_range) && 922 nla_put_uid_range(skb, &rule->uid_range)) || 923 (fib_rule_port_range_set(&rule->sport_range) && 924 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) || 925 (fib_rule_port_range_set(&rule->dport_range) && 926 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) || 927 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto))) 928 goto nla_put_failure; 929 930 if (rule->suppress_ifgroup != -1) { 931 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup)) 932 goto nla_put_failure; 933 } 934 935 if (ops->fill(rule, skb, frh) < 0) 936 goto nla_put_failure; 937 938 nlmsg_end(skb, nlh); 939 return 0; 940 941 nla_put_failure: 942 nlmsg_cancel(skb, nlh); 943 return -EMSGSIZE; 944 } 945 946 static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb, 947 struct fib_rules_ops *ops) 948 { 949 int idx = 0; 950 struct fib_rule *rule; 951 int err = 0; 952 953 rcu_read_lock(); 954 list_for_each_entry_rcu(rule, &ops->rules_list, list) { 955 if (idx < cb->args[1]) 956 goto skip; 957 958 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid, 959 cb->nlh->nlmsg_seq, RTM_NEWRULE, 960 NLM_F_MULTI, ops); 961 if (err) 962 break; 963 skip: 964 idx++; 965 } 966 rcu_read_unlock(); 967 cb->args[1] = idx; 968 rules_ops_put(ops); 969 970 return err; 971 } 972 973 static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb) 974 { 975 struct net *net = sock_net(skb->sk); 976 struct fib_rules_ops *ops; 977 int idx = 0, family; 978 979 family = rtnl_msg_family(cb->nlh); 980 if (family != AF_UNSPEC) { 981 /* Protocol specific dump request */ 982 ops = lookup_rules_ops(net, family); 983 if (ops == NULL) 984 return -EAFNOSUPPORT; 985 986 dump_rules(skb, cb, ops); 987 988 return skb->len; 989 } 990 991 rcu_read_lock(); 992 list_for_each_entry_rcu(ops, &net->rules_ops, list) { 993 if (idx < cb->args[0] || !try_module_get(ops->owner)) 994 goto skip; 995 996 if (dump_rules(skb, cb, ops) < 0) 997 break; 998 999 cb->args[1] = 0; 1000 skip: 1001 idx++; 1002 } 1003 rcu_read_unlock(); 1004 cb->args[0] = idx; 1005 1006 return skb->len; 1007 } 1008 1009 static void notify_rule_change(int event, struct fib_rule *rule, 1010 struct fib_rules_ops *ops, struct nlmsghdr *nlh, 1011 u32 pid) 1012 { 1013 struct net *net; 1014 struct sk_buff *skb; 1015 int err = -ENOBUFS; 1016 1017 net = ops->fro_net; 1018 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL); 1019 if (skb == NULL) 1020 goto errout; 1021 1022 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops); 1023 if (err < 0) { 1024 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */ 1025 WARN_ON(err == -EMSGSIZE); 1026 kfree_skb(skb); 1027 goto errout; 1028 } 1029 1030 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL); 1031 return; 1032 errout: 1033 if (err < 0) 1034 rtnl_set_sk_err(net, ops->nlgroup, err); 1035 } 1036 1037 static void attach_rules(struct list_head *rules, struct net_device *dev) 1038 { 1039 struct fib_rule *rule; 1040 1041 list_for_each_entry(rule, rules, list) { 1042 if (rule->iifindex == -1 && 1043 strcmp(dev->name, rule->iifname) == 0) 1044 rule->iifindex = dev->ifindex; 1045 if (rule->oifindex == -1 && 1046 strcmp(dev->name, rule->oifname) == 0) 1047 rule->oifindex = dev->ifindex; 1048 } 1049 } 1050 1051 static void detach_rules(struct list_head *rules, struct net_device *dev) 1052 { 1053 struct fib_rule *rule; 1054 1055 list_for_each_entry(rule, rules, list) { 1056 if (rule->iifindex == dev->ifindex) 1057 rule->iifindex = -1; 1058 if (rule->oifindex == dev->ifindex) 1059 rule->oifindex = -1; 1060 } 1061 } 1062 1063 1064 static int fib_rules_event(struct notifier_block *this, unsigned long event, 1065 void *ptr) 1066 { 1067 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1068 struct net *net = dev_net(dev); 1069 struct fib_rules_ops *ops; 1070 1071 ASSERT_RTNL(); 1072 1073 switch (event) { 1074 case NETDEV_REGISTER: 1075 list_for_each_entry(ops, &net->rules_ops, list) 1076 attach_rules(&ops->rules_list, dev); 1077 break; 1078 1079 case NETDEV_CHANGENAME: 1080 list_for_each_entry(ops, &net->rules_ops, list) { 1081 detach_rules(&ops->rules_list, dev); 1082 attach_rules(&ops->rules_list, dev); 1083 } 1084 break; 1085 1086 case NETDEV_UNREGISTER: 1087 list_for_each_entry(ops, &net->rules_ops, list) 1088 detach_rules(&ops->rules_list, dev); 1089 break; 1090 } 1091 1092 return NOTIFY_DONE; 1093 } 1094 1095 static struct notifier_block fib_rules_notifier = { 1096 .notifier_call = fib_rules_event, 1097 }; 1098 1099 static int __net_init fib_rules_net_init(struct net *net) 1100 { 1101 INIT_LIST_HEAD(&net->rules_ops); 1102 spin_lock_init(&net->rules_mod_lock); 1103 return 0; 1104 } 1105 1106 static void __net_exit fib_rules_net_exit(struct net *net) 1107 { 1108 WARN_ON_ONCE(!list_empty(&net->rules_ops)); 1109 } 1110 1111 static struct pernet_operations fib_rules_net_ops = { 1112 .init = fib_rules_net_init, 1113 .exit = fib_rules_net_exit, 1114 }; 1115 1116 static int __init fib_rules_init(void) 1117 { 1118 int err; 1119 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, 0); 1120 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, 0); 1121 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule, 0); 1122 1123 err = register_pernet_subsys(&fib_rules_net_ops); 1124 if (err < 0) 1125 goto fail; 1126 1127 err = register_netdevice_notifier(&fib_rules_notifier); 1128 if (err < 0) 1129 goto fail_unregister; 1130 1131 return 0; 1132 1133 fail_unregister: 1134 unregister_pernet_subsys(&fib_rules_net_ops); 1135 fail: 1136 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE); 1137 rtnl_unregister(PF_UNSPEC, RTM_DELRULE); 1138 rtnl_unregister(PF_UNSPEC, RTM_GETRULE); 1139 return err; 1140 } 1141 1142 subsys_initcall(fib_rules_init); 1143