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->suppress_ifgroup != -1 && 420 r->suppress_ifgroup != rule->suppress_ifgroup) 421 continue; 422 423 if (rule->suppress_prefixlen != -1 && 424 r->suppress_prefixlen != rule->suppress_prefixlen) 425 continue; 426 427 if (rule->mark_mask && r->mark_mask != rule->mark_mask) 428 continue; 429 430 if (rule->tun_id && r->tun_id != rule->tun_id) 431 continue; 432 433 if (r->fr_net != rule->fr_net) 434 continue; 435 436 if (rule->l3mdev && r->l3mdev != rule->l3mdev) 437 continue; 438 439 if (uid_range_set(&rule->uid_range) && 440 (!uid_eq(r->uid_range.start, rule->uid_range.start) || 441 !uid_eq(r->uid_range.end, rule->uid_range.end))) 442 continue; 443 444 if (rule->ip_proto && r->ip_proto != rule->ip_proto) 445 continue; 446 447 if (rule->proto && r->proto != rule->proto) 448 continue; 449 450 if (fib_rule_port_range_set(&rule->sport_range) && 451 !fib_rule_port_range_compare(&r->sport_range, 452 &rule->sport_range)) 453 continue; 454 455 if (fib_rule_port_range_set(&rule->dport_range) && 456 !fib_rule_port_range_compare(&r->dport_range, 457 &rule->dport_range)) 458 continue; 459 460 if (!ops->compare(r, frh, tb)) 461 continue; 462 return r; 463 } 464 465 return NULL; 466 } 467 468 #ifdef CONFIG_NET_L3_MASTER_DEV 469 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule, 470 struct netlink_ext_ack *extack) 471 { 472 nlrule->l3mdev = nla_get_u8(nla); 473 if (nlrule->l3mdev != 1) { 474 NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute"); 475 return -1; 476 } 477 478 return 0; 479 } 480 #else 481 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule, 482 struct netlink_ext_ack *extack) 483 { 484 NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel"); 485 return -1; 486 } 487 #endif 488 489 static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh, 490 struct netlink_ext_ack *extack, 491 struct fib_rules_ops *ops, 492 struct nlattr *tb[], 493 struct fib_rule **rule, 494 bool *user_priority) 495 { 496 struct net *net = sock_net(skb->sk); 497 struct fib_rule_hdr *frh = nlmsg_data(nlh); 498 struct fib_rule *nlrule = NULL; 499 int err = -EINVAL; 500 501 if (frh->src_len) 502 if (!tb[FRA_SRC] || 503 frh->src_len > (ops->addr_size * 8) || 504 nla_len(tb[FRA_SRC]) != ops->addr_size) { 505 NL_SET_ERR_MSG(extack, "Invalid source address"); 506 goto errout; 507 } 508 509 if (frh->dst_len) 510 if (!tb[FRA_DST] || 511 frh->dst_len > (ops->addr_size * 8) || 512 nla_len(tb[FRA_DST]) != ops->addr_size) { 513 NL_SET_ERR_MSG(extack, "Invalid dst address"); 514 goto errout; 515 } 516 517 nlrule = kzalloc(ops->rule_size, GFP_KERNEL); 518 if (!nlrule) { 519 err = -ENOMEM; 520 goto errout; 521 } 522 refcount_set(&nlrule->refcnt, 1); 523 nlrule->fr_net = net; 524 525 if (tb[FRA_PRIORITY]) { 526 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]); 527 *user_priority = true; 528 } else { 529 nlrule->pref = fib_default_rule_pref(ops); 530 } 531 532 nlrule->proto = tb[FRA_PROTOCOL] ? 533 nla_get_u8(tb[FRA_PROTOCOL]) : RTPROT_UNSPEC; 534 535 if (tb[FRA_IIFNAME]) { 536 struct net_device *dev; 537 538 nlrule->iifindex = -1; 539 nla_strlcpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ); 540 dev = __dev_get_by_name(net, nlrule->iifname); 541 if (dev) 542 nlrule->iifindex = dev->ifindex; 543 } 544 545 if (tb[FRA_OIFNAME]) { 546 struct net_device *dev; 547 548 nlrule->oifindex = -1; 549 nla_strlcpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ); 550 dev = __dev_get_by_name(net, nlrule->oifname); 551 if (dev) 552 nlrule->oifindex = dev->ifindex; 553 } 554 555 if (tb[FRA_FWMARK]) { 556 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]); 557 if (nlrule->mark) 558 /* compatibility: if the mark value is non-zero all bits 559 * are compared unless a mask is explicitly specified. 560 */ 561 nlrule->mark_mask = 0xFFFFFFFF; 562 } 563 564 if (tb[FRA_FWMASK]) 565 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]); 566 567 if (tb[FRA_TUN_ID]) 568 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]); 569 570 err = -EINVAL; 571 if (tb[FRA_L3MDEV] && 572 fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0) 573 goto errout_free; 574 575 nlrule->action = frh->action; 576 nlrule->flags = frh->flags; 577 nlrule->table = frh_get_table(frh, tb); 578 if (tb[FRA_SUPPRESS_PREFIXLEN]) 579 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]); 580 else 581 nlrule->suppress_prefixlen = -1; 582 583 if (tb[FRA_SUPPRESS_IFGROUP]) 584 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]); 585 else 586 nlrule->suppress_ifgroup = -1; 587 588 if (tb[FRA_GOTO]) { 589 if (nlrule->action != FR_ACT_GOTO) { 590 NL_SET_ERR_MSG(extack, "Unexpected goto"); 591 goto errout_free; 592 } 593 594 nlrule->target = nla_get_u32(tb[FRA_GOTO]); 595 /* Backward jumps are prohibited to avoid endless loops */ 596 if (nlrule->target <= nlrule->pref) { 597 NL_SET_ERR_MSG(extack, "Backward goto not supported"); 598 goto errout_free; 599 } 600 } else if (nlrule->action == FR_ACT_GOTO) { 601 NL_SET_ERR_MSG(extack, "Missing goto target for action goto"); 602 goto errout_free; 603 } 604 605 if (nlrule->l3mdev && nlrule->table) { 606 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive"); 607 goto errout_free; 608 } 609 610 if (tb[FRA_UID_RANGE]) { 611 if (current_user_ns() != net->user_ns) { 612 err = -EPERM; 613 NL_SET_ERR_MSG(extack, "No permission to set uid"); 614 goto errout_free; 615 } 616 617 nlrule->uid_range = nla_get_kuid_range(tb); 618 619 if (!uid_range_set(&nlrule->uid_range) || 620 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) { 621 NL_SET_ERR_MSG(extack, "Invalid uid range"); 622 goto errout_free; 623 } 624 } else { 625 nlrule->uid_range = fib_kuid_range_unset; 626 } 627 628 if (tb[FRA_IP_PROTO]) 629 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]); 630 631 if (tb[FRA_SPORT_RANGE]) { 632 err = nla_get_port_range(tb[FRA_SPORT_RANGE], 633 &nlrule->sport_range); 634 if (err) { 635 NL_SET_ERR_MSG(extack, "Invalid sport range"); 636 goto errout_free; 637 } 638 } 639 640 if (tb[FRA_DPORT_RANGE]) { 641 err = nla_get_port_range(tb[FRA_DPORT_RANGE], 642 &nlrule->dport_range); 643 if (err) { 644 NL_SET_ERR_MSG(extack, "Invalid dport range"); 645 goto errout_free; 646 } 647 } 648 649 *rule = nlrule; 650 651 return 0; 652 653 errout_free: 654 kfree(nlrule); 655 errout: 656 return err; 657 } 658 659 static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh, 660 struct nlattr **tb, struct fib_rule *rule) 661 { 662 struct fib_rule *r; 663 664 list_for_each_entry(r, &ops->rules_list, list) { 665 if (r->action != rule->action) 666 continue; 667 668 if (r->table != rule->table) 669 continue; 670 671 if (r->pref != rule->pref) 672 continue; 673 674 if (memcmp(r->iifname, rule->iifname, IFNAMSIZ)) 675 continue; 676 677 if (memcmp(r->oifname, rule->oifname, IFNAMSIZ)) 678 continue; 679 680 if (r->mark != rule->mark) 681 continue; 682 683 if (r->suppress_ifgroup != rule->suppress_ifgroup) 684 continue; 685 686 if (r->suppress_prefixlen != rule->suppress_prefixlen) 687 continue; 688 689 if (r->mark_mask != rule->mark_mask) 690 continue; 691 692 if (r->tun_id != rule->tun_id) 693 continue; 694 695 if (r->fr_net != rule->fr_net) 696 continue; 697 698 if (r->l3mdev != rule->l3mdev) 699 continue; 700 701 if (!uid_eq(r->uid_range.start, rule->uid_range.start) || 702 !uid_eq(r->uid_range.end, rule->uid_range.end)) 703 continue; 704 705 if (r->ip_proto != rule->ip_proto) 706 continue; 707 708 if (r->proto != rule->proto) 709 continue; 710 711 if (!fib_rule_port_range_compare(&r->sport_range, 712 &rule->sport_range)) 713 continue; 714 715 if (!fib_rule_port_range_compare(&r->dport_range, 716 &rule->dport_range)) 717 continue; 718 719 if (!ops->compare(r, frh, tb)) 720 continue; 721 return 1; 722 } 723 return 0; 724 } 725 726 int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh, 727 struct netlink_ext_ack *extack) 728 { 729 struct net *net = sock_net(skb->sk); 730 struct fib_rule_hdr *frh = nlmsg_data(nlh); 731 struct fib_rules_ops *ops = NULL; 732 struct fib_rule *rule = NULL, *r, *last = NULL; 733 struct nlattr *tb[FRA_MAX + 1]; 734 int err = -EINVAL, unresolved = 0; 735 bool user_priority = false; 736 737 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) { 738 NL_SET_ERR_MSG(extack, "Invalid msg length"); 739 goto errout; 740 } 741 742 ops = lookup_rules_ops(net, frh->family); 743 if (!ops) { 744 err = -EAFNOSUPPORT; 745 NL_SET_ERR_MSG(extack, "Rule family not supported"); 746 goto errout; 747 } 748 749 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy, extack); 750 if (err < 0) { 751 NL_SET_ERR_MSG(extack, "Error parsing msg"); 752 goto errout; 753 } 754 755 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority); 756 if (err) 757 goto errout; 758 759 if ((nlh->nlmsg_flags & NLM_F_EXCL) && 760 rule_exists(ops, frh, tb, rule)) { 761 err = -EEXIST; 762 goto errout_free; 763 } 764 765 err = ops->configure(rule, skb, frh, tb, extack); 766 if (err < 0) 767 goto errout_free; 768 769 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops, 770 extack); 771 if (err < 0) 772 goto errout_free; 773 774 list_for_each_entry(r, &ops->rules_list, list) { 775 if (r->pref == rule->target) { 776 RCU_INIT_POINTER(rule->ctarget, r); 777 break; 778 } 779 } 780 781 if (rcu_dereference_protected(rule->ctarget, 1) == NULL) 782 unresolved = 1; 783 784 list_for_each_entry(r, &ops->rules_list, list) { 785 if (r->pref > rule->pref) 786 break; 787 last = r; 788 } 789 790 if (last) 791 list_add_rcu(&rule->list, &last->list); 792 else 793 list_add_rcu(&rule->list, &ops->rules_list); 794 795 if (ops->unresolved_rules) { 796 /* 797 * There are unresolved goto rules in the list, check if 798 * any of them are pointing to this new rule. 799 */ 800 list_for_each_entry(r, &ops->rules_list, list) { 801 if (r->action == FR_ACT_GOTO && 802 r->target == rule->pref && 803 rtnl_dereference(r->ctarget) == NULL) { 804 rcu_assign_pointer(r->ctarget, rule); 805 if (--ops->unresolved_rules == 0) 806 break; 807 } 808 } 809 } 810 811 if (rule->action == FR_ACT_GOTO) 812 ops->nr_goto_rules++; 813 814 if (unresolved) 815 ops->unresolved_rules++; 816 817 if (rule->tun_id) 818 ip_tunnel_need_metadata(); 819 820 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid); 821 flush_route_cache(ops); 822 rules_ops_put(ops); 823 return 0; 824 825 errout_free: 826 kfree(rule); 827 errout: 828 rules_ops_put(ops); 829 return err; 830 } 831 EXPORT_SYMBOL_GPL(fib_nl_newrule); 832 833 int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh, 834 struct netlink_ext_ack *extack) 835 { 836 struct net *net = sock_net(skb->sk); 837 struct fib_rule_hdr *frh = nlmsg_data(nlh); 838 struct fib_rules_ops *ops = NULL; 839 struct fib_rule *rule = NULL, *r, *nlrule = NULL; 840 struct nlattr *tb[FRA_MAX+1]; 841 int err = -EINVAL; 842 bool user_priority = false; 843 844 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) { 845 NL_SET_ERR_MSG(extack, "Invalid msg length"); 846 goto errout; 847 } 848 849 ops = lookup_rules_ops(net, frh->family); 850 if (ops == NULL) { 851 err = -EAFNOSUPPORT; 852 NL_SET_ERR_MSG(extack, "Rule family not supported"); 853 goto errout; 854 } 855 856 err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy, extack); 857 if (err < 0) { 858 NL_SET_ERR_MSG(extack, "Error parsing msg"); 859 goto errout; 860 } 861 862 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority); 863 if (err) 864 goto errout; 865 866 rule = rule_find(ops, frh, tb, nlrule, user_priority); 867 if (!rule) { 868 err = -ENOENT; 869 goto errout; 870 } 871 872 if (rule->flags & FIB_RULE_PERMANENT) { 873 err = -EPERM; 874 goto errout; 875 } 876 877 if (ops->delete) { 878 err = ops->delete(rule); 879 if (err) 880 goto errout; 881 } 882 883 if (rule->tun_id) 884 ip_tunnel_unneed_metadata(); 885 886 list_del_rcu(&rule->list); 887 888 if (rule->action == FR_ACT_GOTO) { 889 ops->nr_goto_rules--; 890 if (rtnl_dereference(rule->ctarget) == NULL) 891 ops->unresolved_rules--; 892 } 893 894 /* 895 * Check if this rule is a target to any of them. If so, 896 * adjust to the next one with the same preference or 897 * disable them. As this operation is eventually very 898 * expensive, it is only performed if goto rules, except 899 * current if it is goto rule, have actually been added. 900 */ 901 if (ops->nr_goto_rules > 0) { 902 struct fib_rule *n; 903 904 n = list_next_entry(rule, list); 905 if (&n->list == &ops->rules_list || n->pref != rule->pref) 906 n = NULL; 907 list_for_each_entry(r, &ops->rules_list, list) { 908 if (rtnl_dereference(r->ctarget) != rule) 909 continue; 910 rcu_assign_pointer(r->ctarget, n); 911 if (!n) 912 ops->unresolved_rules++; 913 } 914 } 915 916 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops, 917 NULL); 918 notify_rule_change(RTM_DELRULE, rule, ops, nlh, 919 NETLINK_CB(skb).portid); 920 fib_rule_put(rule); 921 flush_route_cache(ops); 922 rules_ops_put(ops); 923 kfree(nlrule); 924 return 0; 925 926 errout: 927 if (nlrule) 928 kfree(nlrule); 929 rules_ops_put(ops); 930 return err; 931 } 932 EXPORT_SYMBOL_GPL(fib_nl_delrule); 933 934 static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops, 935 struct fib_rule *rule) 936 { 937 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr)) 938 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */ 939 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */ 940 + nla_total_size(4) /* FRA_PRIORITY */ 941 + nla_total_size(4) /* FRA_TABLE */ 942 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */ 943 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */ 944 + nla_total_size(4) /* FRA_FWMARK */ 945 + nla_total_size(4) /* FRA_FWMASK */ 946 + nla_total_size_64bit(8) /* FRA_TUN_ID */ 947 + nla_total_size(sizeof(struct fib_kuid_range)) 948 + nla_total_size(1) /* FRA_PROTOCOL */ 949 + nla_total_size(1) /* FRA_IP_PROTO */ 950 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */ 951 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */ 952 953 if (ops->nlmsg_payload) 954 payload += ops->nlmsg_payload(rule); 955 956 return payload; 957 } 958 959 static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule, 960 u32 pid, u32 seq, int type, int flags, 961 struct fib_rules_ops *ops) 962 { 963 struct nlmsghdr *nlh; 964 struct fib_rule_hdr *frh; 965 966 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags); 967 if (nlh == NULL) 968 return -EMSGSIZE; 969 970 frh = nlmsg_data(nlh); 971 frh->family = ops->family; 972 frh->table = rule->table; 973 if (nla_put_u32(skb, FRA_TABLE, rule->table)) 974 goto nla_put_failure; 975 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen)) 976 goto nla_put_failure; 977 frh->res1 = 0; 978 frh->res2 = 0; 979 frh->action = rule->action; 980 frh->flags = rule->flags; 981 982 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto)) 983 goto nla_put_failure; 984 985 if (rule->action == FR_ACT_GOTO && 986 rcu_access_pointer(rule->ctarget) == NULL) 987 frh->flags |= FIB_RULE_UNRESOLVED; 988 989 if (rule->iifname[0]) { 990 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname)) 991 goto nla_put_failure; 992 if (rule->iifindex == -1) 993 frh->flags |= FIB_RULE_IIF_DETACHED; 994 } 995 996 if (rule->oifname[0]) { 997 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname)) 998 goto nla_put_failure; 999 if (rule->oifindex == -1) 1000 frh->flags |= FIB_RULE_OIF_DETACHED; 1001 } 1002 1003 if ((rule->pref && 1004 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) || 1005 (rule->mark && 1006 nla_put_u32(skb, FRA_FWMARK, rule->mark)) || 1007 ((rule->mark_mask || rule->mark) && 1008 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) || 1009 (rule->target && 1010 nla_put_u32(skb, FRA_GOTO, rule->target)) || 1011 (rule->tun_id && 1012 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) || 1013 (rule->l3mdev && 1014 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) || 1015 (uid_range_set(&rule->uid_range) && 1016 nla_put_uid_range(skb, &rule->uid_range)) || 1017 (fib_rule_port_range_set(&rule->sport_range) && 1018 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) || 1019 (fib_rule_port_range_set(&rule->dport_range) && 1020 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) || 1021 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto))) 1022 goto nla_put_failure; 1023 1024 if (rule->suppress_ifgroup != -1) { 1025 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup)) 1026 goto nla_put_failure; 1027 } 1028 1029 if (ops->fill(rule, skb, frh) < 0) 1030 goto nla_put_failure; 1031 1032 nlmsg_end(skb, nlh); 1033 return 0; 1034 1035 nla_put_failure: 1036 nlmsg_cancel(skb, nlh); 1037 return -EMSGSIZE; 1038 } 1039 1040 static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb, 1041 struct fib_rules_ops *ops) 1042 { 1043 int idx = 0; 1044 struct fib_rule *rule; 1045 int err = 0; 1046 1047 rcu_read_lock(); 1048 list_for_each_entry_rcu(rule, &ops->rules_list, list) { 1049 if (idx < cb->args[1]) 1050 goto skip; 1051 1052 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid, 1053 cb->nlh->nlmsg_seq, RTM_NEWRULE, 1054 NLM_F_MULTI, ops); 1055 if (err) 1056 break; 1057 skip: 1058 idx++; 1059 } 1060 rcu_read_unlock(); 1061 cb->args[1] = idx; 1062 rules_ops_put(ops); 1063 1064 return err; 1065 } 1066 1067 static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb) 1068 { 1069 struct net *net = sock_net(skb->sk); 1070 struct fib_rules_ops *ops; 1071 int idx = 0, family; 1072 1073 family = rtnl_msg_family(cb->nlh); 1074 if (family != AF_UNSPEC) { 1075 /* Protocol specific dump request */ 1076 ops = lookup_rules_ops(net, family); 1077 if (ops == NULL) 1078 return -EAFNOSUPPORT; 1079 1080 dump_rules(skb, cb, ops); 1081 1082 return skb->len; 1083 } 1084 1085 rcu_read_lock(); 1086 list_for_each_entry_rcu(ops, &net->rules_ops, list) { 1087 if (idx < cb->args[0] || !try_module_get(ops->owner)) 1088 goto skip; 1089 1090 if (dump_rules(skb, cb, ops) < 0) 1091 break; 1092 1093 cb->args[1] = 0; 1094 skip: 1095 idx++; 1096 } 1097 rcu_read_unlock(); 1098 cb->args[0] = idx; 1099 1100 return skb->len; 1101 } 1102 1103 static void notify_rule_change(int event, struct fib_rule *rule, 1104 struct fib_rules_ops *ops, struct nlmsghdr *nlh, 1105 u32 pid) 1106 { 1107 struct net *net; 1108 struct sk_buff *skb; 1109 int err = -ENOBUFS; 1110 1111 net = ops->fro_net; 1112 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL); 1113 if (skb == NULL) 1114 goto errout; 1115 1116 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops); 1117 if (err < 0) { 1118 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */ 1119 WARN_ON(err == -EMSGSIZE); 1120 kfree_skb(skb); 1121 goto errout; 1122 } 1123 1124 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL); 1125 return; 1126 errout: 1127 if (err < 0) 1128 rtnl_set_sk_err(net, ops->nlgroup, err); 1129 } 1130 1131 static void attach_rules(struct list_head *rules, struct net_device *dev) 1132 { 1133 struct fib_rule *rule; 1134 1135 list_for_each_entry(rule, rules, list) { 1136 if (rule->iifindex == -1 && 1137 strcmp(dev->name, rule->iifname) == 0) 1138 rule->iifindex = dev->ifindex; 1139 if (rule->oifindex == -1 && 1140 strcmp(dev->name, rule->oifname) == 0) 1141 rule->oifindex = dev->ifindex; 1142 } 1143 } 1144 1145 static void detach_rules(struct list_head *rules, struct net_device *dev) 1146 { 1147 struct fib_rule *rule; 1148 1149 list_for_each_entry(rule, rules, list) { 1150 if (rule->iifindex == dev->ifindex) 1151 rule->iifindex = -1; 1152 if (rule->oifindex == dev->ifindex) 1153 rule->oifindex = -1; 1154 } 1155 } 1156 1157 1158 static int fib_rules_event(struct notifier_block *this, unsigned long event, 1159 void *ptr) 1160 { 1161 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1162 struct net *net = dev_net(dev); 1163 struct fib_rules_ops *ops; 1164 1165 ASSERT_RTNL(); 1166 1167 switch (event) { 1168 case NETDEV_REGISTER: 1169 list_for_each_entry(ops, &net->rules_ops, list) 1170 attach_rules(&ops->rules_list, dev); 1171 break; 1172 1173 case NETDEV_CHANGENAME: 1174 list_for_each_entry(ops, &net->rules_ops, list) { 1175 detach_rules(&ops->rules_list, dev); 1176 attach_rules(&ops->rules_list, dev); 1177 } 1178 break; 1179 1180 case NETDEV_UNREGISTER: 1181 list_for_each_entry(ops, &net->rules_ops, list) 1182 detach_rules(&ops->rules_list, dev); 1183 break; 1184 } 1185 1186 return NOTIFY_DONE; 1187 } 1188 1189 static struct notifier_block fib_rules_notifier = { 1190 .notifier_call = fib_rules_event, 1191 }; 1192 1193 static int __net_init fib_rules_net_init(struct net *net) 1194 { 1195 INIT_LIST_HEAD(&net->rules_ops); 1196 spin_lock_init(&net->rules_mod_lock); 1197 return 0; 1198 } 1199 1200 static void __net_exit fib_rules_net_exit(struct net *net) 1201 { 1202 WARN_ON_ONCE(!list_empty(&net->rules_ops)); 1203 } 1204 1205 static struct pernet_operations fib_rules_net_ops = { 1206 .init = fib_rules_net_init, 1207 .exit = fib_rules_net_exit, 1208 }; 1209 1210 static int __init fib_rules_init(void) 1211 { 1212 int err; 1213 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, 0); 1214 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, 0); 1215 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule, 0); 1216 1217 err = register_pernet_subsys(&fib_rules_net_ops); 1218 if (err < 0) 1219 goto fail; 1220 1221 err = register_netdevice_notifier(&fib_rules_notifier); 1222 if (err < 0) 1223 goto fail_unregister; 1224 1225 return 0; 1226 1227 fail_unregister: 1228 unregister_pernet_subsys(&fib_rules_net_ops); 1229 fail: 1230 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE); 1231 rtnl_unregister(PF_UNSPEC, RTM_DELRULE); 1232 rtnl_unregister(PF_UNSPEC, RTM_GETRULE); 1233 return err; 1234 } 1235 1236 subsys_initcall(fib_rules_init); 1237