1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Linux INET6 implementation 4 * Forwarding Information Database 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Changes: 10 * Yuji SEKIYA @USAGI: Support default route on router node; 11 * remove ip6_null_entry from the top of 12 * routing table. 13 * Ville Nuorvala: Fixed routing subtrees. 14 */ 15 16 #define pr_fmt(fmt) "IPv6: " fmt 17 18 #include <linux/errno.h> 19 #include <linux/types.h> 20 #include <linux/net.h> 21 #include <linux/route.h> 22 #include <linux/netdevice.h> 23 #include <linux/in6.h> 24 #include <linux/init.h> 25 #include <linux/list.h> 26 #include <linux/slab.h> 27 28 #include <net/ip.h> 29 #include <net/ipv6.h> 30 #include <net/ndisc.h> 31 #include <net/addrconf.h> 32 #include <net/lwtunnel.h> 33 #include <net/fib_notifier.h> 34 35 #include <net/ip6_fib.h> 36 #include <net/ip6_route.h> 37 38 static struct kmem_cache *fib6_node_kmem __read_mostly; 39 40 struct fib6_cleaner { 41 struct fib6_walker w; 42 struct net *net; 43 int (*func)(struct fib6_info *, void *arg); 44 int sernum; 45 void *arg; 46 bool skip_notify; 47 }; 48 49 #ifdef CONFIG_IPV6_SUBTREES 50 #define FWS_INIT FWS_S 51 #else 52 #define FWS_INIT FWS_L 53 #endif 54 55 static struct fib6_info *fib6_find_prefix(struct net *net, 56 struct fib6_table *table, 57 struct fib6_node *fn); 58 static struct fib6_node *fib6_repair_tree(struct net *net, 59 struct fib6_table *table, 60 struct fib6_node *fn); 61 static int fib6_walk(struct net *net, struct fib6_walker *w); 62 static int fib6_walk_continue(struct fib6_walker *w); 63 64 /* 65 * A routing update causes an increase of the serial number on the 66 * affected subtree. This allows for cached routes to be asynchronously 67 * tested when modifications are made to the destination cache as a 68 * result of redirects, path MTU changes, etc. 69 */ 70 71 static void fib6_gc_timer_cb(struct timer_list *t); 72 73 #define FOR_WALKERS(net, w) \ 74 list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh) 75 76 static void fib6_walker_link(struct net *net, struct fib6_walker *w) 77 { 78 write_lock_bh(&net->ipv6.fib6_walker_lock); 79 list_add(&w->lh, &net->ipv6.fib6_walkers); 80 write_unlock_bh(&net->ipv6.fib6_walker_lock); 81 } 82 83 static void fib6_walker_unlink(struct net *net, struct fib6_walker *w) 84 { 85 write_lock_bh(&net->ipv6.fib6_walker_lock); 86 list_del(&w->lh); 87 write_unlock_bh(&net->ipv6.fib6_walker_lock); 88 } 89 90 static int fib6_new_sernum(struct net *net) 91 { 92 int new, old; 93 94 do { 95 old = atomic_read(&net->ipv6.fib6_sernum); 96 new = old < INT_MAX ? old + 1 : 1; 97 } while (atomic_cmpxchg(&net->ipv6.fib6_sernum, 98 old, new) != old); 99 return new; 100 } 101 102 enum { 103 FIB6_NO_SERNUM_CHANGE = 0, 104 }; 105 106 void fib6_update_sernum(struct net *net, struct fib6_info *f6i) 107 { 108 struct fib6_node *fn; 109 110 fn = rcu_dereference_protected(f6i->fib6_node, 111 lockdep_is_held(&f6i->fib6_table->tb6_lock)); 112 if (fn) 113 fn->fn_sernum = fib6_new_sernum(net); 114 } 115 116 /* 117 * Auxiliary address test functions for the radix tree. 118 * 119 * These assume a 32bit processor (although it will work on 120 * 64bit processors) 121 */ 122 123 /* 124 * test bit 125 */ 126 #if defined(__LITTLE_ENDIAN) 127 # define BITOP_BE32_SWIZZLE (0x1F & ~7) 128 #else 129 # define BITOP_BE32_SWIZZLE 0 130 #endif 131 132 static __be32 addr_bit_set(const void *token, int fn_bit) 133 { 134 const __be32 *addr = token; 135 /* 136 * Here, 137 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f) 138 * is optimized version of 139 * htonl(1 << ((~fn_bit)&0x1F)) 140 * See include/asm-generic/bitops/le.h. 141 */ 142 return (__force __be32)(1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)) & 143 addr[fn_bit >> 5]; 144 } 145 146 struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh) 147 { 148 struct fib6_info *f6i; 149 size_t sz = sizeof(*f6i); 150 151 if (with_fib6_nh) 152 sz += sizeof(struct fib6_nh); 153 154 f6i = kzalloc(sz, gfp_flags); 155 if (!f6i) 156 return NULL; 157 158 /* fib6_siblings is a union with nh_list, so this initializes both */ 159 INIT_LIST_HEAD(&f6i->fib6_siblings); 160 refcount_set(&f6i->fib6_ref, 1); 161 162 return f6i; 163 } 164 165 void fib6_info_destroy_rcu(struct rcu_head *head) 166 { 167 struct fib6_info *f6i = container_of(head, struct fib6_info, rcu); 168 169 WARN_ON(f6i->fib6_node); 170 171 if (f6i->nh) 172 nexthop_put(f6i->nh); 173 else 174 fib6_nh_release(f6i->fib6_nh); 175 176 ip_fib_metrics_put(f6i->fib6_metrics); 177 kfree(f6i); 178 } 179 EXPORT_SYMBOL_GPL(fib6_info_destroy_rcu); 180 181 static struct fib6_node *node_alloc(struct net *net) 182 { 183 struct fib6_node *fn; 184 185 fn = kmem_cache_zalloc(fib6_node_kmem, GFP_ATOMIC); 186 if (fn) 187 net->ipv6.rt6_stats->fib_nodes++; 188 189 return fn; 190 } 191 192 static void node_free_immediate(struct net *net, struct fib6_node *fn) 193 { 194 kmem_cache_free(fib6_node_kmem, fn); 195 net->ipv6.rt6_stats->fib_nodes--; 196 } 197 198 static void node_free_rcu(struct rcu_head *head) 199 { 200 struct fib6_node *fn = container_of(head, struct fib6_node, rcu); 201 202 kmem_cache_free(fib6_node_kmem, fn); 203 } 204 205 static void node_free(struct net *net, struct fib6_node *fn) 206 { 207 call_rcu(&fn->rcu, node_free_rcu); 208 net->ipv6.rt6_stats->fib_nodes--; 209 } 210 211 static void fib6_free_table(struct fib6_table *table) 212 { 213 inetpeer_invalidate_tree(&table->tb6_peers); 214 kfree(table); 215 } 216 217 static void fib6_link_table(struct net *net, struct fib6_table *tb) 218 { 219 unsigned int h; 220 221 /* 222 * Initialize table lock at a single place to give lockdep a key, 223 * tables aren't visible prior to being linked to the list. 224 */ 225 spin_lock_init(&tb->tb6_lock); 226 h = tb->tb6_id & (FIB6_TABLE_HASHSZ - 1); 227 228 /* 229 * No protection necessary, this is the only list mutatation 230 * operation, tables never disappear once they exist. 231 */ 232 hlist_add_head_rcu(&tb->tb6_hlist, &net->ipv6.fib_table_hash[h]); 233 } 234 235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 236 237 static struct fib6_table *fib6_alloc_table(struct net *net, u32 id) 238 { 239 struct fib6_table *table; 240 241 table = kzalloc(sizeof(*table), GFP_ATOMIC); 242 if (table) { 243 table->tb6_id = id; 244 rcu_assign_pointer(table->tb6_root.leaf, 245 net->ipv6.fib6_null_entry); 246 table->tb6_root.fn_flags = RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; 247 inet_peer_base_init(&table->tb6_peers); 248 } 249 250 return table; 251 } 252 253 struct fib6_table *fib6_new_table(struct net *net, u32 id) 254 { 255 struct fib6_table *tb; 256 257 if (id == 0) 258 id = RT6_TABLE_MAIN; 259 tb = fib6_get_table(net, id); 260 if (tb) 261 return tb; 262 263 tb = fib6_alloc_table(net, id); 264 if (tb) 265 fib6_link_table(net, tb); 266 267 return tb; 268 } 269 EXPORT_SYMBOL_GPL(fib6_new_table); 270 271 struct fib6_table *fib6_get_table(struct net *net, u32 id) 272 { 273 struct fib6_table *tb; 274 struct hlist_head *head; 275 unsigned int h; 276 277 if (id == 0) 278 id = RT6_TABLE_MAIN; 279 h = id & (FIB6_TABLE_HASHSZ - 1); 280 rcu_read_lock(); 281 head = &net->ipv6.fib_table_hash[h]; 282 hlist_for_each_entry_rcu(tb, head, tb6_hlist) { 283 if (tb->tb6_id == id) { 284 rcu_read_unlock(); 285 return tb; 286 } 287 } 288 rcu_read_unlock(); 289 290 return NULL; 291 } 292 EXPORT_SYMBOL_GPL(fib6_get_table); 293 294 static void __net_init fib6_tables_init(struct net *net) 295 { 296 fib6_link_table(net, net->ipv6.fib6_main_tbl); 297 fib6_link_table(net, net->ipv6.fib6_local_tbl); 298 } 299 #else 300 301 struct fib6_table *fib6_new_table(struct net *net, u32 id) 302 { 303 return fib6_get_table(net, id); 304 } 305 306 struct fib6_table *fib6_get_table(struct net *net, u32 id) 307 { 308 return net->ipv6.fib6_main_tbl; 309 } 310 311 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6, 312 const struct sk_buff *skb, 313 int flags, pol_lookup_t lookup) 314 { 315 struct rt6_info *rt; 316 317 rt = lookup(net, net->ipv6.fib6_main_tbl, fl6, skb, flags); 318 if (rt->dst.error == -EAGAIN) { 319 ip6_rt_put_flags(rt, flags); 320 rt = net->ipv6.ip6_null_entry; 321 if (!(flags & RT6_LOOKUP_F_DST_NOREF)) 322 dst_hold(&rt->dst); 323 } 324 325 return &rt->dst; 326 } 327 328 /* called with rcu lock held; no reference taken on fib6_info */ 329 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6, 330 struct fib6_result *res, int flags) 331 { 332 return fib6_table_lookup(net, net->ipv6.fib6_main_tbl, oif, fl6, 333 res, flags); 334 } 335 336 static void __net_init fib6_tables_init(struct net *net) 337 { 338 fib6_link_table(net, net->ipv6.fib6_main_tbl); 339 } 340 341 #endif 342 343 unsigned int fib6_tables_seq_read(struct net *net) 344 { 345 unsigned int h, fib_seq = 0; 346 347 rcu_read_lock(); 348 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { 349 struct hlist_head *head = &net->ipv6.fib_table_hash[h]; 350 struct fib6_table *tb; 351 352 hlist_for_each_entry_rcu(tb, head, tb6_hlist) 353 fib_seq += tb->fib_seq; 354 } 355 rcu_read_unlock(); 356 357 return fib_seq; 358 } 359 360 static int call_fib6_entry_notifier(struct notifier_block *nb, 361 enum fib_event_type event_type, 362 struct fib6_info *rt, 363 struct netlink_ext_ack *extack) 364 { 365 struct fib6_entry_notifier_info info = { 366 .info.extack = extack, 367 .rt = rt, 368 }; 369 370 return call_fib6_notifier(nb, event_type, &info.info); 371 } 372 373 static int call_fib6_multipath_entry_notifier(struct notifier_block *nb, 374 enum fib_event_type event_type, 375 struct fib6_info *rt, 376 unsigned int nsiblings, 377 struct netlink_ext_ack *extack) 378 { 379 struct fib6_entry_notifier_info info = { 380 .info.extack = extack, 381 .rt = rt, 382 .nsiblings = nsiblings, 383 }; 384 385 return call_fib6_notifier(nb, event_type, &info.info); 386 } 387 388 int call_fib6_entry_notifiers(struct net *net, 389 enum fib_event_type event_type, 390 struct fib6_info *rt, 391 struct netlink_ext_ack *extack) 392 { 393 struct fib6_entry_notifier_info info = { 394 .info.extack = extack, 395 .rt = rt, 396 }; 397 398 rt->fib6_table->fib_seq++; 399 return call_fib6_notifiers(net, event_type, &info.info); 400 } 401 402 int call_fib6_multipath_entry_notifiers(struct net *net, 403 enum fib_event_type event_type, 404 struct fib6_info *rt, 405 unsigned int nsiblings, 406 struct netlink_ext_ack *extack) 407 { 408 struct fib6_entry_notifier_info info = { 409 .info.extack = extack, 410 .rt = rt, 411 .nsiblings = nsiblings, 412 }; 413 414 rt->fib6_table->fib_seq++; 415 return call_fib6_notifiers(net, event_type, &info.info); 416 } 417 418 int call_fib6_entry_notifiers_replace(struct net *net, struct fib6_info *rt) 419 { 420 struct fib6_entry_notifier_info info = { 421 .rt = rt, 422 .nsiblings = rt->fib6_nsiblings, 423 }; 424 425 rt->fib6_table->fib_seq++; 426 return call_fib6_notifiers(net, FIB_EVENT_ENTRY_REPLACE, &info.info); 427 } 428 429 struct fib6_dump_arg { 430 struct net *net; 431 struct notifier_block *nb; 432 struct netlink_ext_ack *extack; 433 }; 434 435 static int fib6_rt_dump(struct fib6_info *rt, struct fib6_dump_arg *arg) 436 { 437 enum fib_event_type fib_event = FIB_EVENT_ENTRY_REPLACE; 438 int err; 439 440 if (!rt || rt == arg->net->ipv6.fib6_null_entry) 441 return 0; 442 443 if (rt->fib6_nsiblings) 444 err = call_fib6_multipath_entry_notifier(arg->nb, fib_event, 445 rt, 446 rt->fib6_nsiblings, 447 arg->extack); 448 else 449 err = call_fib6_entry_notifier(arg->nb, fib_event, rt, 450 arg->extack); 451 452 return err; 453 } 454 455 static int fib6_node_dump(struct fib6_walker *w) 456 { 457 int err; 458 459 err = fib6_rt_dump(w->leaf, w->args); 460 w->leaf = NULL; 461 return err; 462 } 463 464 static int fib6_table_dump(struct net *net, struct fib6_table *tb, 465 struct fib6_walker *w) 466 { 467 int err; 468 469 w->root = &tb->tb6_root; 470 spin_lock_bh(&tb->tb6_lock); 471 err = fib6_walk(net, w); 472 spin_unlock_bh(&tb->tb6_lock); 473 return err; 474 } 475 476 /* Called with rcu_read_lock() */ 477 int fib6_tables_dump(struct net *net, struct notifier_block *nb, 478 struct netlink_ext_ack *extack) 479 { 480 struct fib6_dump_arg arg; 481 struct fib6_walker *w; 482 unsigned int h; 483 int err = 0; 484 485 w = kzalloc(sizeof(*w), GFP_ATOMIC); 486 if (!w) 487 return -ENOMEM; 488 489 w->func = fib6_node_dump; 490 arg.net = net; 491 arg.nb = nb; 492 arg.extack = extack; 493 w->args = &arg; 494 495 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { 496 struct hlist_head *head = &net->ipv6.fib_table_hash[h]; 497 struct fib6_table *tb; 498 499 hlist_for_each_entry_rcu(tb, head, tb6_hlist) { 500 err = fib6_table_dump(net, tb, w); 501 if (err < 0) 502 goto out; 503 } 504 } 505 506 out: 507 kfree(w); 508 509 return err; 510 } 511 512 static int fib6_dump_node(struct fib6_walker *w) 513 { 514 int res; 515 struct fib6_info *rt; 516 517 for_each_fib6_walker_rt(w) { 518 res = rt6_dump_route(rt, w->args, w->skip_in_node); 519 if (res >= 0) { 520 /* Frame is full, suspend walking */ 521 w->leaf = rt; 522 523 /* We'll restart from this node, so if some routes were 524 * already dumped, skip them next time. 525 */ 526 w->skip_in_node += res; 527 528 return 1; 529 } 530 w->skip_in_node = 0; 531 532 /* Multipath routes are dumped in one route with the 533 * RTA_MULTIPATH attribute. Jump 'rt' to point to the 534 * last sibling of this route (no need to dump the 535 * sibling routes again) 536 */ 537 if (rt->fib6_nsiblings) 538 rt = list_last_entry(&rt->fib6_siblings, 539 struct fib6_info, 540 fib6_siblings); 541 } 542 w->leaf = NULL; 543 return 0; 544 } 545 546 static void fib6_dump_end(struct netlink_callback *cb) 547 { 548 struct net *net = sock_net(cb->skb->sk); 549 struct fib6_walker *w = (void *)cb->args[2]; 550 551 if (w) { 552 if (cb->args[4]) { 553 cb->args[4] = 0; 554 fib6_walker_unlink(net, w); 555 } 556 cb->args[2] = 0; 557 kfree(w); 558 } 559 cb->done = (void *)cb->args[3]; 560 cb->args[1] = 3; 561 } 562 563 static int fib6_dump_done(struct netlink_callback *cb) 564 { 565 fib6_dump_end(cb); 566 return cb->done ? cb->done(cb) : 0; 567 } 568 569 static int fib6_dump_table(struct fib6_table *table, struct sk_buff *skb, 570 struct netlink_callback *cb) 571 { 572 struct net *net = sock_net(skb->sk); 573 struct fib6_walker *w; 574 int res; 575 576 w = (void *)cb->args[2]; 577 w->root = &table->tb6_root; 578 579 if (cb->args[4] == 0) { 580 w->count = 0; 581 w->skip = 0; 582 w->skip_in_node = 0; 583 584 spin_lock_bh(&table->tb6_lock); 585 res = fib6_walk(net, w); 586 spin_unlock_bh(&table->tb6_lock); 587 if (res > 0) { 588 cb->args[4] = 1; 589 cb->args[5] = w->root->fn_sernum; 590 } 591 } else { 592 if (cb->args[5] != w->root->fn_sernum) { 593 /* Begin at the root if the tree changed */ 594 cb->args[5] = w->root->fn_sernum; 595 w->state = FWS_INIT; 596 w->node = w->root; 597 w->skip = w->count; 598 w->skip_in_node = 0; 599 } else 600 w->skip = 0; 601 602 spin_lock_bh(&table->tb6_lock); 603 res = fib6_walk_continue(w); 604 spin_unlock_bh(&table->tb6_lock); 605 if (res <= 0) { 606 fib6_walker_unlink(net, w); 607 cb->args[4] = 0; 608 } 609 } 610 611 return res; 612 } 613 614 static int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb) 615 { 616 struct rt6_rtnl_dump_arg arg = { .filter.dump_exceptions = true, 617 .filter.dump_routes = true }; 618 const struct nlmsghdr *nlh = cb->nlh; 619 struct net *net = sock_net(skb->sk); 620 unsigned int h, s_h; 621 unsigned int e = 0, s_e; 622 struct fib6_walker *w; 623 struct fib6_table *tb; 624 struct hlist_head *head; 625 int res = 0; 626 627 if (cb->strict_check) { 628 int err; 629 630 err = ip_valid_fib_dump_req(net, nlh, &arg.filter, cb); 631 if (err < 0) 632 return err; 633 } else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) { 634 struct rtmsg *rtm = nlmsg_data(nlh); 635 636 if (rtm->rtm_flags & RTM_F_PREFIX) 637 arg.filter.flags = RTM_F_PREFIX; 638 } 639 640 w = (void *)cb->args[2]; 641 if (!w) { 642 /* New dump: 643 * 644 * 1. hook callback destructor. 645 */ 646 cb->args[3] = (long)cb->done; 647 cb->done = fib6_dump_done; 648 649 /* 650 * 2. allocate and initialize walker. 651 */ 652 w = kzalloc(sizeof(*w), GFP_ATOMIC); 653 if (!w) 654 return -ENOMEM; 655 w->func = fib6_dump_node; 656 cb->args[2] = (long)w; 657 } 658 659 arg.skb = skb; 660 arg.cb = cb; 661 arg.net = net; 662 w->args = &arg; 663 664 if (arg.filter.table_id) { 665 tb = fib6_get_table(net, arg.filter.table_id); 666 if (!tb) { 667 if (arg.filter.dump_all_families) 668 goto out; 669 670 NL_SET_ERR_MSG_MOD(cb->extack, "FIB table does not exist"); 671 return -ENOENT; 672 } 673 674 if (!cb->args[0]) { 675 res = fib6_dump_table(tb, skb, cb); 676 if (!res) 677 cb->args[0] = 1; 678 } 679 goto out; 680 } 681 682 s_h = cb->args[0]; 683 s_e = cb->args[1]; 684 685 rcu_read_lock(); 686 for (h = s_h; h < FIB6_TABLE_HASHSZ; h++, s_e = 0) { 687 e = 0; 688 head = &net->ipv6.fib_table_hash[h]; 689 hlist_for_each_entry_rcu(tb, head, tb6_hlist) { 690 if (e < s_e) 691 goto next; 692 res = fib6_dump_table(tb, skb, cb); 693 if (res != 0) 694 goto out_unlock; 695 next: 696 e++; 697 } 698 } 699 out_unlock: 700 rcu_read_unlock(); 701 cb->args[1] = e; 702 cb->args[0] = h; 703 out: 704 res = res < 0 ? res : skb->len; 705 if (res <= 0) 706 fib6_dump_end(cb); 707 return res; 708 } 709 710 void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val) 711 { 712 if (!f6i) 713 return; 714 715 if (f6i->fib6_metrics == &dst_default_metrics) { 716 struct dst_metrics *p = kzalloc(sizeof(*p), GFP_ATOMIC); 717 718 if (!p) 719 return; 720 721 refcount_set(&p->refcnt, 1); 722 f6i->fib6_metrics = p; 723 } 724 725 f6i->fib6_metrics->metrics[metric - 1] = val; 726 } 727 728 /* 729 * Routing Table 730 * 731 * return the appropriate node for a routing tree "add" operation 732 * by either creating and inserting or by returning an existing 733 * node. 734 */ 735 736 static struct fib6_node *fib6_add_1(struct net *net, 737 struct fib6_table *table, 738 struct fib6_node *root, 739 struct in6_addr *addr, int plen, 740 int offset, int allow_create, 741 int replace_required, 742 struct netlink_ext_ack *extack) 743 { 744 struct fib6_node *fn, *in, *ln; 745 struct fib6_node *pn = NULL; 746 struct rt6key *key; 747 int bit; 748 __be32 dir = 0; 749 750 RT6_TRACE("fib6_add_1\n"); 751 752 /* insert node in tree */ 753 754 fn = root; 755 756 do { 757 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, 758 lockdep_is_held(&table->tb6_lock)); 759 key = (struct rt6key *)((u8 *)leaf + offset); 760 761 /* 762 * Prefix match 763 */ 764 if (plen < fn->fn_bit || 765 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) { 766 if (!allow_create) { 767 if (replace_required) { 768 NL_SET_ERR_MSG(extack, 769 "Can not replace route - no match found"); 770 pr_warn("Can't replace route, no match found\n"); 771 return ERR_PTR(-ENOENT); 772 } 773 pr_warn("NLM_F_CREATE should be set when creating new route\n"); 774 } 775 goto insert_above; 776 } 777 778 /* 779 * Exact match ? 780 */ 781 782 if (plen == fn->fn_bit) { 783 /* clean up an intermediate node */ 784 if (!(fn->fn_flags & RTN_RTINFO)) { 785 RCU_INIT_POINTER(fn->leaf, NULL); 786 fib6_info_release(leaf); 787 /* remove null_entry in the root node */ 788 } else if (fn->fn_flags & RTN_TL_ROOT && 789 rcu_access_pointer(fn->leaf) == 790 net->ipv6.fib6_null_entry) { 791 RCU_INIT_POINTER(fn->leaf, NULL); 792 } 793 794 return fn; 795 } 796 797 /* 798 * We have more bits to go 799 */ 800 801 /* Try to walk down on tree. */ 802 dir = addr_bit_set(addr, fn->fn_bit); 803 pn = fn; 804 fn = dir ? 805 rcu_dereference_protected(fn->right, 806 lockdep_is_held(&table->tb6_lock)) : 807 rcu_dereference_protected(fn->left, 808 lockdep_is_held(&table->tb6_lock)); 809 } while (fn); 810 811 if (!allow_create) { 812 /* We should not create new node because 813 * NLM_F_REPLACE was specified without NLM_F_CREATE 814 * I assume it is safe to require NLM_F_CREATE when 815 * REPLACE flag is used! Later we may want to remove the 816 * check for replace_required, because according 817 * to netlink specification, NLM_F_CREATE 818 * MUST be specified if new route is created. 819 * That would keep IPv6 consistent with IPv4 820 */ 821 if (replace_required) { 822 NL_SET_ERR_MSG(extack, 823 "Can not replace route - no match found"); 824 pr_warn("Can't replace route, no match found\n"); 825 return ERR_PTR(-ENOENT); 826 } 827 pr_warn("NLM_F_CREATE should be set when creating new route\n"); 828 } 829 /* 830 * We walked to the bottom of tree. 831 * Create new leaf node without children. 832 */ 833 834 ln = node_alloc(net); 835 836 if (!ln) 837 return ERR_PTR(-ENOMEM); 838 ln->fn_bit = plen; 839 RCU_INIT_POINTER(ln->parent, pn); 840 841 if (dir) 842 rcu_assign_pointer(pn->right, ln); 843 else 844 rcu_assign_pointer(pn->left, ln); 845 846 return ln; 847 848 849 insert_above: 850 /* 851 * split since we don't have a common prefix anymore or 852 * we have a less significant route. 853 * we've to insert an intermediate node on the list 854 * this new node will point to the one we need to create 855 * and the current 856 */ 857 858 pn = rcu_dereference_protected(fn->parent, 859 lockdep_is_held(&table->tb6_lock)); 860 861 /* find 1st bit in difference between the 2 addrs. 862 863 See comment in __ipv6_addr_diff: bit may be an invalid value, 864 but if it is >= plen, the value is ignored in any case. 865 */ 866 867 bit = __ipv6_addr_diff(addr, &key->addr, sizeof(*addr)); 868 869 /* 870 * (intermediate)[in] 871 * / \ 872 * (new leaf node)[ln] (old node)[fn] 873 */ 874 if (plen > bit) { 875 in = node_alloc(net); 876 ln = node_alloc(net); 877 878 if (!in || !ln) { 879 if (in) 880 node_free_immediate(net, in); 881 if (ln) 882 node_free_immediate(net, ln); 883 return ERR_PTR(-ENOMEM); 884 } 885 886 /* 887 * new intermediate node. 888 * RTN_RTINFO will 889 * be off since that an address that chooses one of 890 * the branches would not match less specific routes 891 * in the other branch 892 */ 893 894 in->fn_bit = bit; 895 896 RCU_INIT_POINTER(in->parent, pn); 897 in->leaf = fn->leaf; 898 fib6_info_hold(rcu_dereference_protected(in->leaf, 899 lockdep_is_held(&table->tb6_lock))); 900 901 /* update parent pointer */ 902 if (dir) 903 rcu_assign_pointer(pn->right, in); 904 else 905 rcu_assign_pointer(pn->left, in); 906 907 ln->fn_bit = plen; 908 909 RCU_INIT_POINTER(ln->parent, in); 910 rcu_assign_pointer(fn->parent, in); 911 912 if (addr_bit_set(addr, bit)) { 913 rcu_assign_pointer(in->right, ln); 914 rcu_assign_pointer(in->left, fn); 915 } else { 916 rcu_assign_pointer(in->left, ln); 917 rcu_assign_pointer(in->right, fn); 918 } 919 } else { /* plen <= bit */ 920 921 /* 922 * (new leaf node)[ln] 923 * / \ 924 * (old node)[fn] NULL 925 */ 926 927 ln = node_alloc(net); 928 929 if (!ln) 930 return ERR_PTR(-ENOMEM); 931 932 ln->fn_bit = plen; 933 934 RCU_INIT_POINTER(ln->parent, pn); 935 936 if (addr_bit_set(&key->addr, plen)) 937 RCU_INIT_POINTER(ln->right, fn); 938 else 939 RCU_INIT_POINTER(ln->left, fn); 940 941 rcu_assign_pointer(fn->parent, ln); 942 943 if (dir) 944 rcu_assign_pointer(pn->right, ln); 945 else 946 rcu_assign_pointer(pn->left, ln); 947 } 948 return ln; 949 } 950 951 static void __fib6_drop_pcpu_from(struct fib6_nh *fib6_nh, 952 const struct fib6_info *match, 953 const struct fib6_table *table) 954 { 955 int cpu; 956 957 if (!fib6_nh->rt6i_pcpu) 958 return; 959 960 /* release the reference to this fib entry from 961 * all of its cached pcpu routes 962 */ 963 for_each_possible_cpu(cpu) { 964 struct rt6_info **ppcpu_rt; 965 struct rt6_info *pcpu_rt; 966 967 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu); 968 pcpu_rt = *ppcpu_rt; 969 970 /* only dropping the 'from' reference if the cached route 971 * is using 'match'. The cached pcpu_rt->from only changes 972 * from a fib6_info to NULL (ip6_dst_destroy); it can never 973 * change from one fib6_info reference to another 974 */ 975 if (pcpu_rt && rcu_access_pointer(pcpu_rt->from) == match) { 976 struct fib6_info *from; 977 978 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL); 979 fib6_info_release(from); 980 } 981 } 982 } 983 984 struct fib6_nh_pcpu_arg { 985 struct fib6_info *from; 986 const struct fib6_table *table; 987 }; 988 989 static int fib6_nh_drop_pcpu_from(struct fib6_nh *nh, void *_arg) 990 { 991 struct fib6_nh_pcpu_arg *arg = _arg; 992 993 __fib6_drop_pcpu_from(nh, arg->from, arg->table); 994 return 0; 995 } 996 997 static void fib6_drop_pcpu_from(struct fib6_info *f6i, 998 const struct fib6_table *table) 999 { 1000 /* Make sure rt6_make_pcpu_route() wont add other percpu routes 1001 * while we are cleaning them here. 1002 */ 1003 f6i->fib6_destroying = 1; 1004 mb(); /* paired with the cmpxchg() in rt6_make_pcpu_route() */ 1005 1006 if (f6i->nh) { 1007 struct fib6_nh_pcpu_arg arg = { 1008 .from = f6i, 1009 .table = table 1010 }; 1011 1012 nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_drop_pcpu_from, 1013 &arg); 1014 } else { 1015 struct fib6_nh *fib6_nh; 1016 1017 fib6_nh = f6i->fib6_nh; 1018 __fib6_drop_pcpu_from(fib6_nh, f6i, table); 1019 } 1020 } 1021 1022 static void fib6_purge_rt(struct fib6_info *rt, struct fib6_node *fn, 1023 struct net *net) 1024 { 1025 struct fib6_table *table = rt->fib6_table; 1026 1027 fib6_drop_pcpu_from(rt, table); 1028 1029 if (rt->nh && !list_empty(&rt->nh_list)) 1030 list_del_init(&rt->nh_list); 1031 1032 if (refcount_read(&rt->fib6_ref) != 1) { 1033 /* This route is used as dummy address holder in some split 1034 * nodes. It is not leaked, but it still holds other resources, 1035 * which must be released in time. So, scan ascendant nodes 1036 * and replace dummy references to this route with references 1037 * to still alive ones. 1038 */ 1039 while (fn) { 1040 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, 1041 lockdep_is_held(&table->tb6_lock)); 1042 struct fib6_info *new_leaf; 1043 if (!(fn->fn_flags & RTN_RTINFO) && leaf == rt) { 1044 new_leaf = fib6_find_prefix(net, table, fn); 1045 fib6_info_hold(new_leaf); 1046 1047 rcu_assign_pointer(fn->leaf, new_leaf); 1048 fib6_info_release(rt); 1049 } 1050 fn = rcu_dereference_protected(fn->parent, 1051 lockdep_is_held(&table->tb6_lock)); 1052 } 1053 } 1054 } 1055 1056 /* 1057 * Insert routing information in a node. 1058 */ 1059 1060 static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt, 1061 struct nl_info *info, 1062 struct netlink_ext_ack *extack) 1063 { 1064 struct fib6_info *leaf = rcu_dereference_protected(fn->leaf, 1065 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1066 struct fib6_info *iter = NULL; 1067 struct fib6_info __rcu **ins; 1068 struct fib6_info __rcu **fallback_ins = NULL; 1069 int replace = (info->nlh && 1070 (info->nlh->nlmsg_flags & NLM_F_REPLACE)); 1071 int add = (!info->nlh || 1072 (info->nlh->nlmsg_flags & NLM_F_CREATE)); 1073 int found = 0; 1074 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt); 1075 bool notify_sibling_rt = false; 1076 u16 nlflags = NLM_F_EXCL; 1077 int err; 1078 1079 if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND)) 1080 nlflags |= NLM_F_APPEND; 1081 1082 ins = &fn->leaf; 1083 1084 for (iter = leaf; iter; 1085 iter = rcu_dereference_protected(iter->fib6_next, 1086 lockdep_is_held(&rt->fib6_table->tb6_lock))) { 1087 /* 1088 * Search for duplicates 1089 */ 1090 1091 if (iter->fib6_metric == rt->fib6_metric) { 1092 /* 1093 * Same priority level 1094 */ 1095 if (info->nlh && 1096 (info->nlh->nlmsg_flags & NLM_F_EXCL)) 1097 return -EEXIST; 1098 1099 nlflags &= ~NLM_F_EXCL; 1100 if (replace) { 1101 if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) { 1102 found++; 1103 break; 1104 } 1105 fallback_ins = fallback_ins ?: ins; 1106 goto next_iter; 1107 } 1108 1109 if (rt6_duplicate_nexthop(iter, rt)) { 1110 if (rt->fib6_nsiblings) 1111 rt->fib6_nsiblings = 0; 1112 if (!(iter->fib6_flags & RTF_EXPIRES)) 1113 return -EEXIST; 1114 if (!(rt->fib6_flags & RTF_EXPIRES)) 1115 fib6_clean_expires(iter); 1116 else 1117 fib6_set_expires(iter, rt->expires); 1118 1119 if (rt->fib6_pmtu) 1120 fib6_metric_set(iter, RTAX_MTU, 1121 rt->fib6_pmtu); 1122 return -EEXIST; 1123 } 1124 /* If we have the same destination and the same metric, 1125 * but not the same gateway, then the route we try to 1126 * add is sibling to this route, increment our counter 1127 * of siblings, and later we will add our route to the 1128 * list. 1129 * Only static routes (which don't have flag 1130 * RTF_EXPIRES) are used for ECMPv6. 1131 * 1132 * To avoid long list, we only had siblings if the 1133 * route have a gateway. 1134 */ 1135 if (rt_can_ecmp && 1136 rt6_qualify_for_ecmp(iter)) 1137 rt->fib6_nsiblings++; 1138 } 1139 1140 if (iter->fib6_metric > rt->fib6_metric) 1141 break; 1142 1143 next_iter: 1144 ins = &iter->fib6_next; 1145 } 1146 1147 if (fallback_ins && !found) { 1148 /* No matching route with same ecmp-able-ness found, replace 1149 * first matching route 1150 */ 1151 ins = fallback_ins; 1152 iter = rcu_dereference_protected(*ins, 1153 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1154 found++; 1155 } 1156 1157 /* Reset round-robin state, if necessary */ 1158 if (ins == &fn->leaf) 1159 fn->rr_ptr = NULL; 1160 1161 /* Link this route to others same route. */ 1162 if (rt->fib6_nsiblings) { 1163 unsigned int fib6_nsiblings; 1164 struct fib6_info *sibling, *temp_sibling; 1165 1166 /* Find the first route that have the same metric */ 1167 sibling = leaf; 1168 notify_sibling_rt = true; 1169 while (sibling) { 1170 if (sibling->fib6_metric == rt->fib6_metric && 1171 rt6_qualify_for_ecmp(sibling)) { 1172 list_add_tail(&rt->fib6_siblings, 1173 &sibling->fib6_siblings); 1174 break; 1175 } 1176 sibling = rcu_dereference_protected(sibling->fib6_next, 1177 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1178 notify_sibling_rt = false; 1179 } 1180 /* For each sibling in the list, increment the counter of 1181 * siblings. BUG() if counters does not match, list of siblings 1182 * is broken! 1183 */ 1184 fib6_nsiblings = 0; 1185 list_for_each_entry_safe(sibling, temp_sibling, 1186 &rt->fib6_siblings, fib6_siblings) { 1187 sibling->fib6_nsiblings++; 1188 BUG_ON(sibling->fib6_nsiblings != rt->fib6_nsiblings); 1189 fib6_nsiblings++; 1190 } 1191 BUG_ON(fib6_nsiblings != rt->fib6_nsiblings); 1192 rt6_multipath_rebalance(temp_sibling); 1193 } 1194 1195 /* 1196 * insert node 1197 */ 1198 if (!replace) { 1199 if (!add) 1200 pr_warn("NLM_F_CREATE should be set when creating new route\n"); 1201 1202 add: 1203 nlflags |= NLM_F_CREATE; 1204 1205 /* The route should only be notified if it is the first 1206 * route in the node or if it is added as a sibling 1207 * route to the first route in the node. 1208 */ 1209 if (!info->skip_notify_kernel && 1210 (notify_sibling_rt || ins == &fn->leaf)) { 1211 enum fib_event_type fib_event; 1212 1213 if (notify_sibling_rt) 1214 fib_event = FIB_EVENT_ENTRY_APPEND; 1215 else 1216 fib_event = FIB_EVENT_ENTRY_REPLACE; 1217 err = call_fib6_entry_notifiers(info->nl_net, 1218 fib_event, rt, 1219 extack); 1220 if (err) { 1221 struct fib6_info *sibling, *next_sibling; 1222 1223 /* If the route has siblings, then it first 1224 * needs to be unlinked from them. 1225 */ 1226 if (!rt->fib6_nsiblings) 1227 return err; 1228 1229 list_for_each_entry_safe(sibling, next_sibling, 1230 &rt->fib6_siblings, 1231 fib6_siblings) 1232 sibling->fib6_nsiblings--; 1233 rt->fib6_nsiblings = 0; 1234 list_del_init(&rt->fib6_siblings); 1235 rt6_multipath_rebalance(next_sibling); 1236 return err; 1237 } 1238 } 1239 1240 rcu_assign_pointer(rt->fib6_next, iter); 1241 fib6_info_hold(rt); 1242 rcu_assign_pointer(rt->fib6_node, fn); 1243 rcu_assign_pointer(*ins, rt); 1244 if (!info->skip_notify) 1245 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags); 1246 info->nl_net->ipv6.rt6_stats->fib_rt_entries++; 1247 1248 if (!(fn->fn_flags & RTN_RTINFO)) { 1249 info->nl_net->ipv6.rt6_stats->fib_route_nodes++; 1250 fn->fn_flags |= RTN_RTINFO; 1251 } 1252 1253 } else { 1254 int nsiblings; 1255 1256 if (!found) { 1257 if (add) 1258 goto add; 1259 pr_warn("NLM_F_REPLACE set, but no existing node found!\n"); 1260 return -ENOENT; 1261 } 1262 1263 if (!info->skip_notify_kernel && ins == &fn->leaf) { 1264 err = call_fib6_entry_notifiers(info->nl_net, 1265 FIB_EVENT_ENTRY_REPLACE, 1266 rt, extack); 1267 if (err) 1268 return err; 1269 } 1270 1271 fib6_info_hold(rt); 1272 rcu_assign_pointer(rt->fib6_node, fn); 1273 rt->fib6_next = iter->fib6_next; 1274 rcu_assign_pointer(*ins, rt); 1275 if (!info->skip_notify) 1276 inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE); 1277 if (!(fn->fn_flags & RTN_RTINFO)) { 1278 info->nl_net->ipv6.rt6_stats->fib_route_nodes++; 1279 fn->fn_flags |= RTN_RTINFO; 1280 } 1281 nsiblings = iter->fib6_nsiblings; 1282 iter->fib6_node = NULL; 1283 fib6_purge_rt(iter, fn, info->nl_net); 1284 if (rcu_access_pointer(fn->rr_ptr) == iter) 1285 fn->rr_ptr = NULL; 1286 fib6_info_release(iter); 1287 1288 if (nsiblings) { 1289 /* Replacing an ECMP route, remove all siblings */ 1290 ins = &rt->fib6_next; 1291 iter = rcu_dereference_protected(*ins, 1292 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1293 while (iter) { 1294 if (iter->fib6_metric > rt->fib6_metric) 1295 break; 1296 if (rt6_qualify_for_ecmp(iter)) { 1297 *ins = iter->fib6_next; 1298 iter->fib6_node = NULL; 1299 fib6_purge_rt(iter, fn, info->nl_net); 1300 if (rcu_access_pointer(fn->rr_ptr) == iter) 1301 fn->rr_ptr = NULL; 1302 fib6_info_release(iter); 1303 nsiblings--; 1304 info->nl_net->ipv6.rt6_stats->fib_rt_entries--; 1305 } else { 1306 ins = &iter->fib6_next; 1307 } 1308 iter = rcu_dereference_protected(*ins, 1309 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1310 } 1311 WARN_ON(nsiblings != 0); 1312 } 1313 } 1314 1315 return 0; 1316 } 1317 1318 static void fib6_start_gc(struct net *net, struct fib6_info *rt) 1319 { 1320 if (!timer_pending(&net->ipv6.ip6_fib_timer) && 1321 (rt->fib6_flags & RTF_EXPIRES)) 1322 mod_timer(&net->ipv6.ip6_fib_timer, 1323 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); 1324 } 1325 1326 void fib6_force_start_gc(struct net *net) 1327 { 1328 if (!timer_pending(&net->ipv6.ip6_fib_timer)) 1329 mod_timer(&net->ipv6.ip6_fib_timer, 1330 jiffies + net->ipv6.sysctl.ip6_rt_gc_interval); 1331 } 1332 1333 static void __fib6_update_sernum_upto_root(struct fib6_info *rt, 1334 int sernum) 1335 { 1336 struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node, 1337 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1338 1339 /* paired with smp_rmb() in rt6_get_cookie_safe() */ 1340 smp_wmb(); 1341 while (fn) { 1342 fn->fn_sernum = sernum; 1343 fn = rcu_dereference_protected(fn->parent, 1344 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1345 } 1346 } 1347 1348 void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt) 1349 { 1350 __fib6_update_sernum_upto_root(rt, fib6_new_sernum(net)); 1351 } 1352 1353 /* allow ipv4 to update sernum via ipv6_stub */ 1354 void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i) 1355 { 1356 spin_lock_bh(&f6i->fib6_table->tb6_lock); 1357 fib6_update_sernum_upto_root(net, f6i); 1358 spin_unlock_bh(&f6i->fib6_table->tb6_lock); 1359 } 1360 1361 /* 1362 * Add routing information to the routing tree. 1363 * <destination addr>/<source addr> 1364 * with source addr info in sub-trees 1365 * Need to own table->tb6_lock 1366 */ 1367 1368 int fib6_add(struct fib6_node *root, struct fib6_info *rt, 1369 struct nl_info *info, struct netlink_ext_ack *extack) 1370 { 1371 struct fib6_table *table = rt->fib6_table; 1372 struct fib6_node *fn, *pn = NULL; 1373 int err = -ENOMEM; 1374 int allow_create = 1; 1375 int replace_required = 0; 1376 int sernum = fib6_new_sernum(info->nl_net); 1377 1378 if (info->nlh) { 1379 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE)) 1380 allow_create = 0; 1381 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 1382 replace_required = 1; 1383 } 1384 if (!allow_create && !replace_required) 1385 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n"); 1386 1387 fn = fib6_add_1(info->nl_net, table, root, 1388 &rt->fib6_dst.addr, rt->fib6_dst.plen, 1389 offsetof(struct fib6_info, fib6_dst), allow_create, 1390 replace_required, extack); 1391 if (IS_ERR(fn)) { 1392 err = PTR_ERR(fn); 1393 fn = NULL; 1394 goto out; 1395 } 1396 1397 pn = fn; 1398 1399 #ifdef CONFIG_IPV6_SUBTREES 1400 if (rt->fib6_src.plen) { 1401 struct fib6_node *sn; 1402 1403 if (!rcu_access_pointer(fn->subtree)) { 1404 struct fib6_node *sfn; 1405 1406 /* 1407 * Create subtree. 1408 * 1409 * fn[main tree] 1410 * | 1411 * sfn[subtree root] 1412 * \ 1413 * sn[new leaf node] 1414 */ 1415 1416 /* Create subtree root node */ 1417 sfn = node_alloc(info->nl_net); 1418 if (!sfn) 1419 goto failure; 1420 1421 fib6_info_hold(info->nl_net->ipv6.fib6_null_entry); 1422 rcu_assign_pointer(sfn->leaf, 1423 info->nl_net->ipv6.fib6_null_entry); 1424 sfn->fn_flags = RTN_ROOT; 1425 1426 /* Now add the first leaf node to new subtree */ 1427 1428 sn = fib6_add_1(info->nl_net, table, sfn, 1429 &rt->fib6_src.addr, rt->fib6_src.plen, 1430 offsetof(struct fib6_info, fib6_src), 1431 allow_create, replace_required, extack); 1432 1433 if (IS_ERR(sn)) { 1434 /* If it is failed, discard just allocated 1435 root, and then (in failure) stale node 1436 in main tree. 1437 */ 1438 node_free_immediate(info->nl_net, sfn); 1439 err = PTR_ERR(sn); 1440 goto failure; 1441 } 1442 1443 /* Now link new subtree to main tree */ 1444 rcu_assign_pointer(sfn->parent, fn); 1445 rcu_assign_pointer(fn->subtree, sfn); 1446 } else { 1447 sn = fib6_add_1(info->nl_net, table, FIB6_SUBTREE(fn), 1448 &rt->fib6_src.addr, rt->fib6_src.plen, 1449 offsetof(struct fib6_info, fib6_src), 1450 allow_create, replace_required, extack); 1451 1452 if (IS_ERR(sn)) { 1453 err = PTR_ERR(sn); 1454 goto failure; 1455 } 1456 } 1457 1458 if (!rcu_access_pointer(fn->leaf)) { 1459 if (fn->fn_flags & RTN_TL_ROOT) { 1460 /* put back null_entry for root node */ 1461 rcu_assign_pointer(fn->leaf, 1462 info->nl_net->ipv6.fib6_null_entry); 1463 } else { 1464 fib6_info_hold(rt); 1465 rcu_assign_pointer(fn->leaf, rt); 1466 } 1467 } 1468 fn = sn; 1469 } 1470 #endif 1471 1472 err = fib6_add_rt2node(fn, rt, info, extack); 1473 if (!err) { 1474 if (rt->nh) 1475 list_add(&rt->nh_list, &rt->nh->f6i_list); 1476 __fib6_update_sernum_upto_root(rt, sernum); 1477 fib6_start_gc(info->nl_net, rt); 1478 } 1479 1480 out: 1481 if (err) { 1482 #ifdef CONFIG_IPV6_SUBTREES 1483 /* 1484 * If fib6_add_1 has cleared the old leaf pointer in the 1485 * super-tree leaf node we have to find a new one for it. 1486 */ 1487 if (pn != fn) { 1488 struct fib6_info *pn_leaf = 1489 rcu_dereference_protected(pn->leaf, 1490 lockdep_is_held(&table->tb6_lock)); 1491 if (pn_leaf == rt) { 1492 pn_leaf = NULL; 1493 RCU_INIT_POINTER(pn->leaf, NULL); 1494 fib6_info_release(rt); 1495 } 1496 if (!pn_leaf && !(pn->fn_flags & RTN_RTINFO)) { 1497 pn_leaf = fib6_find_prefix(info->nl_net, table, 1498 pn); 1499 #if RT6_DEBUG >= 2 1500 if (!pn_leaf) { 1501 WARN_ON(!pn_leaf); 1502 pn_leaf = 1503 info->nl_net->ipv6.fib6_null_entry; 1504 } 1505 #endif 1506 fib6_info_hold(pn_leaf); 1507 rcu_assign_pointer(pn->leaf, pn_leaf); 1508 } 1509 } 1510 #endif 1511 goto failure; 1512 } else if (fib6_requires_src(rt)) { 1513 fib6_routes_require_src_inc(info->nl_net); 1514 } 1515 return err; 1516 1517 failure: 1518 /* fn->leaf could be NULL and fib6_repair_tree() needs to be called if: 1519 * 1. fn is an intermediate node and we failed to add the new 1520 * route to it in both subtree creation failure and fib6_add_rt2node() 1521 * failure case. 1522 * 2. fn is the root node in the table and we fail to add the first 1523 * default route to it. 1524 */ 1525 if (fn && 1526 (!(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)) || 1527 (fn->fn_flags & RTN_TL_ROOT && 1528 !rcu_access_pointer(fn->leaf)))) 1529 fib6_repair_tree(info->nl_net, table, fn); 1530 return err; 1531 } 1532 1533 /* 1534 * Routing tree lookup 1535 * 1536 */ 1537 1538 struct lookup_args { 1539 int offset; /* key offset on fib6_info */ 1540 const struct in6_addr *addr; /* search key */ 1541 }; 1542 1543 static struct fib6_node *fib6_node_lookup_1(struct fib6_node *root, 1544 struct lookup_args *args) 1545 { 1546 struct fib6_node *fn; 1547 __be32 dir; 1548 1549 if (unlikely(args->offset == 0)) 1550 return NULL; 1551 1552 /* 1553 * Descend on a tree 1554 */ 1555 1556 fn = root; 1557 1558 for (;;) { 1559 struct fib6_node *next; 1560 1561 dir = addr_bit_set(args->addr, fn->fn_bit); 1562 1563 next = dir ? rcu_dereference(fn->right) : 1564 rcu_dereference(fn->left); 1565 1566 if (next) { 1567 fn = next; 1568 continue; 1569 } 1570 break; 1571 } 1572 1573 while (fn) { 1574 struct fib6_node *subtree = FIB6_SUBTREE(fn); 1575 1576 if (subtree || fn->fn_flags & RTN_RTINFO) { 1577 struct fib6_info *leaf = rcu_dereference(fn->leaf); 1578 struct rt6key *key; 1579 1580 if (!leaf) 1581 goto backtrack; 1582 1583 key = (struct rt6key *) ((u8 *)leaf + args->offset); 1584 1585 if (ipv6_prefix_equal(&key->addr, args->addr, key->plen)) { 1586 #ifdef CONFIG_IPV6_SUBTREES 1587 if (subtree) { 1588 struct fib6_node *sfn; 1589 sfn = fib6_node_lookup_1(subtree, 1590 args + 1); 1591 if (!sfn) 1592 goto backtrack; 1593 fn = sfn; 1594 } 1595 #endif 1596 if (fn->fn_flags & RTN_RTINFO) 1597 return fn; 1598 } 1599 } 1600 backtrack: 1601 if (fn->fn_flags & RTN_ROOT) 1602 break; 1603 1604 fn = rcu_dereference(fn->parent); 1605 } 1606 1607 return NULL; 1608 } 1609 1610 /* called with rcu_read_lock() held 1611 */ 1612 struct fib6_node *fib6_node_lookup(struct fib6_node *root, 1613 const struct in6_addr *daddr, 1614 const struct in6_addr *saddr) 1615 { 1616 struct fib6_node *fn; 1617 struct lookup_args args[] = { 1618 { 1619 .offset = offsetof(struct fib6_info, fib6_dst), 1620 .addr = daddr, 1621 }, 1622 #ifdef CONFIG_IPV6_SUBTREES 1623 { 1624 .offset = offsetof(struct fib6_info, fib6_src), 1625 .addr = saddr, 1626 }, 1627 #endif 1628 { 1629 .offset = 0, /* sentinel */ 1630 } 1631 }; 1632 1633 fn = fib6_node_lookup_1(root, daddr ? args : args + 1); 1634 if (!fn || fn->fn_flags & RTN_TL_ROOT) 1635 fn = root; 1636 1637 return fn; 1638 } 1639 1640 /* 1641 * Get node with specified destination prefix (and source prefix, 1642 * if subtrees are used) 1643 * exact_match == true means we try to find fn with exact match of 1644 * the passed in prefix addr 1645 * exact_match == false means we try to find fn with longest prefix 1646 * match of the passed in prefix addr. This is useful for finding fn 1647 * for cached route as it will be stored in the exception table under 1648 * the node with longest prefix length. 1649 */ 1650 1651 1652 static struct fib6_node *fib6_locate_1(struct fib6_node *root, 1653 const struct in6_addr *addr, 1654 int plen, int offset, 1655 bool exact_match) 1656 { 1657 struct fib6_node *fn, *prev = NULL; 1658 1659 for (fn = root; fn ; ) { 1660 struct fib6_info *leaf = rcu_dereference(fn->leaf); 1661 struct rt6key *key; 1662 1663 /* This node is being deleted */ 1664 if (!leaf) { 1665 if (plen <= fn->fn_bit) 1666 goto out; 1667 else 1668 goto next; 1669 } 1670 1671 key = (struct rt6key *)((u8 *)leaf + offset); 1672 1673 /* 1674 * Prefix match 1675 */ 1676 if (plen < fn->fn_bit || 1677 !ipv6_prefix_equal(&key->addr, addr, fn->fn_bit)) 1678 goto out; 1679 1680 if (plen == fn->fn_bit) 1681 return fn; 1682 1683 if (fn->fn_flags & RTN_RTINFO) 1684 prev = fn; 1685 1686 next: 1687 /* 1688 * We have more bits to go 1689 */ 1690 if (addr_bit_set(addr, fn->fn_bit)) 1691 fn = rcu_dereference(fn->right); 1692 else 1693 fn = rcu_dereference(fn->left); 1694 } 1695 out: 1696 if (exact_match) 1697 return NULL; 1698 else 1699 return prev; 1700 } 1701 1702 struct fib6_node *fib6_locate(struct fib6_node *root, 1703 const struct in6_addr *daddr, int dst_len, 1704 const struct in6_addr *saddr, int src_len, 1705 bool exact_match) 1706 { 1707 struct fib6_node *fn; 1708 1709 fn = fib6_locate_1(root, daddr, dst_len, 1710 offsetof(struct fib6_info, fib6_dst), 1711 exact_match); 1712 1713 #ifdef CONFIG_IPV6_SUBTREES 1714 if (src_len) { 1715 WARN_ON(saddr == NULL); 1716 if (fn) { 1717 struct fib6_node *subtree = FIB6_SUBTREE(fn); 1718 1719 if (subtree) { 1720 fn = fib6_locate_1(subtree, saddr, src_len, 1721 offsetof(struct fib6_info, fib6_src), 1722 exact_match); 1723 } 1724 } 1725 } 1726 #endif 1727 1728 if (fn && fn->fn_flags & RTN_RTINFO) 1729 return fn; 1730 1731 return NULL; 1732 } 1733 1734 1735 /* 1736 * Deletion 1737 * 1738 */ 1739 1740 static struct fib6_info *fib6_find_prefix(struct net *net, 1741 struct fib6_table *table, 1742 struct fib6_node *fn) 1743 { 1744 struct fib6_node *child_left, *child_right; 1745 1746 if (fn->fn_flags & RTN_ROOT) 1747 return net->ipv6.fib6_null_entry; 1748 1749 while (fn) { 1750 child_left = rcu_dereference_protected(fn->left, 1751 lockdep_is_held(&table->tb6_lock)); 1752 child_right = rcu_dereference_protected(fn->right, 1753 lockdep_is_held(&table->tb6_lock)); 1754 if (child_left) 1755 return rcu_dereference_protected(child_left->leaf, 1756 lockdep_is_held(&table->tb6_lock)); 1757 if (child_right) 1758 return rcu_dereference_protected(child_right->leaf, 1759 lockdep_is_held(&table->tb6_lock)); 1760 1761 fn = FIB6_SUBTREE(fn); 1762 } 1763 return NULL; 1764 } 1765 1766 /* 1767 * Called to trim the tree of intermediate nodes when possible. "fn" 1768 * is the node we want to try and remove. 1769 * Need to own table->tb6_lock 1770 */ 1771 1772 static struct fib6_node *fib6_repair_tree(struct net *net, 1773 struct fib6_table *table, 1774 struct fib6_node *fn) 1775 { 1776 int children; 1777 int nstate; 1778 struct fib6_node *child; 1779 struct fib6_walker *w; 1780 int iter = 0; 1781 1782 /* Set fn->leaf to null_entry for root node. */ 1783 if (fn->fn_flags & RTN_TL_ROOT) { 1784 rcu_assign_pointer(fn->leaf, net->ipv6.fib6_null_entry); 1785 return fn; 1786 } 1787 1788 for (;;) { 1789 struct fib6_node *fn_r = rcu_dereference_protected(fn->right, 1790 lockdep_is_held(&table->tb6_lock)); 1791 struct fib6_node *fn_l = rcu_dereference_protected(fn->left, 1792 lockdep_is_held(&table->tb6_lock)); 1793 struct fib6_node *pn = rcu_dereference_protected(fn->parent, 1794 lockdep_is_held(&table->tb6_lock)); 1795 struct fib6_node *pn_r = rcu_dereference_protected(pn->right, 1796 lockdep_is_held(&table->tb6_lock)); 1797 struct fib6_node *pn_l = rcu_dereference_protected(pn->left, 1798 lockdep_is_held(&table->tb6_lock)); 1799 struct fib6_info *fn_leaf = rcu_dereference_protected(fn->leaf, 1800 lockdep_is_held(&table->tb6_lock)); 1801 struct fib6_info *pn_leaf = rcu_dereference_protected(pn->leaf, 1802 lockdep_is_held(&table->tb6_lock)); 1803 struct fib6_info *new_fn_leaf; 1804 1805 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn->fn_bit, iter); 1806 iter++; 1807 1808 WARN_ON(fn->fn_flags & RTN_RTINFO); 1809 WARN_ON(fn->fn_flags & RTN_TL_ROOT); 1810 WARN_ON(fn_leaf); 1811 1812 children = 0; 1813 child = NULL; 1814 if (fn_r) 1815 child = fn_r, children |= 1; 1816 if (fn_l) 1817 child = fn_l, children |= 2; 1818 1819 if (children == 3 || FIB6_SUBTREE(fn) 1820 #ifdef CONFIG_IPV6_SUBTREES 1821 /* Subtree root (i.e. fn) may have one child */ 1822 || (children && fn->fn_flags & RTN_ROOT) 1823 #endif 1824 ) { 1825 new_fn_leaf = fib6_find_prefix(net, table, fn); 1826 #if RT6_DEBUG >= 2 1827 if (!new_fn_leaf) { 1828 WARN_ON(!new_fn_leaf); 1829 new_fn_leaf = net->ipv6.fib6_null_entry; 1830 } 1831 #endif 1832 fib6_info_hold(new_fn_leaf); 1833 rcu_assign_pointer(fn->leaf, new_fn_leaf); 1834 return pn; 1835 } 1836 1837 #ifdef CONFIG_IPV6_SUBTREES 1838 if (FIB6_SUBTREE(pn) == fn) { 1839 WARN_ON(!(fn->fn_flags & RTN_ROOT)); 1840 RCU_INIT_POINTER(pn->subtree, NULL); 1841 nstate = FWS_L; 1842 } else { 1843 WARN_ON(fn->fn_flags & RTN_ROOT); 1844 #endif 1845 if (pn_r == fn) 1846 rcu_assign_pointer(pn->right, child); 1847 else if (pn_l == fn) 1848 rcu_assign_pointer(pn->left, child); 1849 #if RT6_DEBUG >= 2 1850 else 1851 WARN_ON(1); 1852 #endif 1853 if (child) 1854 rcu_assign_pointer(child->parent, pn); 1855 nstate = FWS_R; 1856 #ifdef CONFIG_IPV6_SUBTREES 1857 } 1858 #endif 1859 1860 read_lock(&net->ipv6.fib6_walker_lock); 1861 FOR_WALKERS(net, w) { 1862 if (!child) { 1863 if (w->node == fn) { 1864 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w, w->state, nstate); 1865 w->node = pn; 1866 w->state = nstate; 1867 } 1868 } else { 1869 if (w->node == fn) { 1870 w->node = child; 1871 if (children&2) { 1872 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); 1873 w->state = w->state >= FWS_R ? FWS_U : FWS_INIT; 1874 } else { 1875 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w, w->state); 1876 w->state = w->state >= FWS_C ? FWS_U : FWS_INIT; 1877 } 1878 } 1879 } 1880 } 1881 read_unlock(&net->ipv6.fib6_walker_lock); 1882 1883 node_free(net, fn); 1884 if (pn->fn_flags & RTN_RTINFO || FIB6_SUBTREE(pn)) 1885 return pn; 1886 1887 RCU_INIT_POINTER(pn->leaf, NULL); 1888 fib6_info_release(pn_leaf); 1889 fn = pn; 1890 } 1891 } 1892 1893 static void fib6_del_route(struct fib6_table *table, struct fib6_node *fn, 1894 struct fib6_info __rcu **rtp, struct nl_info *info) 1895 { 1896 struct fib6_info *leaf, *replace_rt = NULL; 1897 struct fib6_walker *w; 1898 struct fib6_info *rt = rcu_dereference_protected(*rtp, 1899 lockdep_is_held(&table->tb6_lock)); 1900 struct net *net = info->nl_net; 1901 bool notify_del = false; 1902 1903 RT6_TRACE("fib6_del_route\n"); 1904 1905 /* If the deleted route is the first in the node and it is not part of 1906 * a multipath route, then we need to replace it with the next route 1907 * in the node, if exists. 1908 */ 1909 leaf = rcu_dereference_protected(fn->leaf, 1910 lockdep_is_held(&table->tb6_lock)); 1911 if (leaf == rt && !rt->fib6_nsiblings) { 1912 if (rcu_access_pointer(rt->fib6_next)) 1913 replace_rt = rcu_dereference_protected(rt->fib6_next, 1914 lockdep_is_held(&table->tb6_lock)); 1915 else 1916 notify_del = true; 1917 } 1918 1919 /* Unlink it */ 1920 *rtp = rt->fib6_next; 1921 rt->fib6_node = NULL; 1922 net->ipv6.rt6_stats->fib_rt_entries--; 1923 net->ipv6.rt6_stats->fib_discarded_routes++; 1924 1925 /* Flush all cached dst in exception table */ 1926 rt6_flush_exceptions(rt); 1927 1928 /* Reset round-robin state, if necessary */ 1929 if (rcu_access_pointer(fn->rr_ptr) == rt) 1930 fn->rr_ptr = NULL; 1931 1932 /* Remove this entry from other siblings */ 1933 if (rt->fib6_nsiblings) { 1934 struct fib6_info *sibling, *next_sibling; 1935 1936 /* The route is deleted from a multipath route. If this 1937 * multipath route is the first route in the node, then we need 1938 * to emit a delete notification. Otherwise, we need to skip 1939 * the notification. 1940 */ 1941 if (rt->fib6_metric == leaf->fib6_metric && 1942 rt6_qualify_for_ecmp(leaf)) 1943 notify_del = true; 1944 list_for_each_entry_safe(sibling, next_sibling, 1945 &rt->fib6_siblings, fib6_siblings) 1946 sibling->fib6_nsiblings--; 1947 rt->fib6_nsiblings = 0; 1948 list_del_init(&rt->fib6_siblings); 1949 rt6_multipath_rebalance(next_sibling); 1950 } 1951 1952 /* Adjust walkers */ 1953 read_lock(&net->ipv6.fib6_walker_lock); 1954 FOR_WALKERS(net, w) { 1955 if (w->state == FWS_C && w->leaf == rt) { 1956 RT6_TRACE("walker %p adjusted by delroute\n", w); 1957 w->leaf = rcu_dereference_protected(rt->fib6_next, 1958 lockdep_is_held(&table->tb6_lock)); 1959 if (!w->leaf) 1960 w->state = FWS_U; 1961 } 1962 } 1963 read_unlock(&net->ipv6.fib6_walker_lock); 1964 1965 /* If it was last route, call fib6_repair_tree() to: 1966 * 1. For root node, put back null_entry as how the table was created. 1967 * 2. For other nodes, expunge its radix tree node. 1968 */ 1969 if (!rcu_access_pointer(fn->leaf)) { 1970 if (!(fn->fn_flags & RTN_TL_ROOT)) { 1971 fn->fn_flags &= ~RTN_RTINFO; 1972 net->ipv6.rt6_stats->fib_route_nodes--; 1973 } 1974 fn = fib6_repair_tree(net, table, fn); 1975 } 1976 1977 fib6_purge_rt(rt, fn, net); 1978 1979 if (!info->skip_notify_kernel) { 1980 if (notify_del) 1981 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, 1982 rt, NULL); 1983 else if (replace_rt) 1984 call_fib6_entry_notifiers_replace(net, replace_rt); 1985 } 1986 if (!info->skip_notify) 1987 inet6_rt_notify(RTM_DELROUTE, rt, info, 0); 1988 1989 fib6_info_release(rt); 1990 } 1991 1992 /* Need to own table->tb6_lock */ 1993 int fib6_del(struct fib6_info *rt, struct nl_info *info) 1994 { 1995 struct fib6_node *fn = rcu_dereference_protected(rt->fib6_node, 1996 lockdep_is_held(&rt->fib6_table->tb6_lock)); 1997 struct fib6_table *table = rt->fib6_table; 1998 struct net *net = info->nl_net; 1999 struct fib6_info __rcu **rtp; 2000 struct fib6_info __rcu **rtp_next; 2001 2002 if (!fn || rt == net->ipv6.fib6_null_entry) 2003 return -ENOENT; 2004 2005 WARN_ON(!(fn->fn_flags & RTN_RTINFO)); 2006 2007 /* 2008 * Walk the leaf entries looking for ourself 2009 */ 2010 2011 for (rtp = &fn->leaf; *rtp; rtp = rtp_next) { 2012 struct fib6_info *cur = rcu_dereference_protected(*rtp, 2013 lockdep_is_held(&table->tb6_lock)); 2014 if (rt == cur) { 2015 if (fib6_requires_src(cur)) 2016 fib6_routes_require_src_dec(info->nl_net); 2017 fib6_del_route(table, fn, rtp, info); 2018 return 0; 2019 } 2020 rtp_next = &cur->fib6_next; 2021 } 2022 return -ENOENT; 2023 } 2024 2025 /* 2026 * Tree traversal function. 2027 * 2028 * Certainly, it is not interrupt safe. 2029 * However, it is internally reenterable wrt itself and fib6_add/fib6_del. 2030 * It means, that we can modify tree during walking 2031 * and use this function for garbage collection, clone pruning, 2032 * cleaning tree when a device goes down etc. etc. 2033 * 2034 * It guarantees that every node will be traversed, 2035 * and that it will be traversed only once. 2036 * 2037 * Callback function w->func may return: 2038 * 0 -> continue walking. 2039 * positive value -> walking is suspended (used by tree dumps, 2040 * and probably by gc, if it will be split to several slices) 2041 * negative value -> terminate walking. 2042 * 2043 * The function itself returns: 2044 * 0 -> walk is complete. 2045 * >0 -> walk is incomplete (i.e. suspended) 2046 * <0 -> walk is terminated by an error. 2047 * 2048 * This function is called with tb6_lock held. 2049 */ 2050 2051 static int fib6_walk_continue(struct fib6_walker *w) 2052 { 2053 struct fib6_node *fn, *pn, *left, *right; 2054 2055 /* w->root should always be table->tb6_root */ 2056 WARN_ON_ONCE(!(w->root->fn_flags & RTN_TL_ROOT)); 2057 2058 for (;;) { 2059 fn = w->node; 2060 if (!fn) 2061 return 0; 2062 2063 switch (w->state) { 2064 #ifdef CONFIG_IPV6_SUBTREES 2065 case FWS_S: 2066 if (FIB6_SUBTREE(fn)) { 2067 w->node = FIB6_SUBTREE(fn); 2068 continue; 2069 } 2070 w->state = FWS_L; 2071 #endif 2072 /* fall through */ 2073 case FWS_L: 2074 left = rcu_dereference_protected(fn->left, 1); 2075 if (left) { 2076 w->node = left; 2077 w->state = FWS_INIT; 2078 continue; 2079 } 2080 w->state = FWS_R; 2081 /* fall through */ 2082 case FWS_R: 2083 right = rcu_dereference_protected(fn->right, 1); 2084 if (right) { 2085 w->node = right; 2086 w->state = FWS_INIT; 2087 continue; 2088 } 2089 w->state = FWS_C; 2090 w->leaf = rcu_dereference_protected(fn->leaf, 1); 2091 /* fall through */ 2092 case FWS_C: 2093 if (w->leaf && fn->fn_flags & RTN_RTINFO) { 2094 int err; 2095 2096 if (w->skip) { 2097 w->skip--; 2098 goto skip; 2099 } 2100 2101 err = w->func(w); 2102 if (err) 2103 return err; 2104 2105 w->count++; 2106 continue; 2107 } 2108 skip: 2109 w->state = FWS_U; 2110 /* fall through */ 2111 case FWS_U: 2112 if (fn == w->root) 2113 return 0; 2114 pn = rcu_dereference_protected(fn->parent, 1); 2115 left = rcu_dereference_protected(pn->left, 1); 2116 right = rcu_dereference_protected(pn->right, 1); 2117 w->node = pn; 2118 #ifdef CONFIG_IPV6_SUBTREES 2119 if (FIB6_SUBTREE(pn) == fn) { 2120 WARN_ON(!(fn->fn_flags & RTN_ROOT)); 2121 w->state = FWS_L; 2122 continue; 2123 } 2124 #endif 2125 if (left == fn) { 2126 w->state = FWS_R; 2127 continue; 2128 } 2129 if (right == fn) { 2130 w->state = FWS_C; 2131 w->leaf = rcu_dereference_protected(w->node->leaf, 1); 2132 continue; 2133 } 2134 #if RT6_DEBUG >= 2 2135 WARN_ON(1); 2136 #endif 2137 } 2138 } 2139 } 2140 2141 static int fib6_walk(struct net *net, struct fib6_walker *w) 2142 { 2143 int res; 2144 2145 w->state = FWS_INIT; 2146 w->node = w->root; 2147 2148 fib6_walker_link(net, w); 2149 res = fib6_walk_continue(w); 2150 if (res <= 0) 2151 fib6_walker_unlink(net, w); 2152 return res; 2153 } 2154 2155 static int fib6_clean_node(struct fib6_walker *w) 2156 { 2157 int res; 2158 struct fib6_info *rt; 2159 struct fib6_cleaner *c = container_of(w, struct fib6_cleaner, w); 2160 struct nl_info info = { 2161 .nl_net = c->net, 2162 .skip_notify = c->skip_notify, 2163 }; 2164 2165 if (c->sernum != FIB6_NO_SERNUM_CHANGE && 2166 w->node->fn_sernum != c->sernum) 2167 w->node->fn_sernum = c->sernum; 2168 2169 if (!c->func) { 2170 WARN_ON_ONCE(c->sernum == FIB6_NO_SERNUM_CHANGE); 2171 w->leaf = NULL; 2172 return 0; 2173 } 2174 2175 for_each_fib6_walker_rt(w) { 2176 res = c->func(rt, c->arg); 2177 if (res == -1) { 2178 w->leaf = rt; 2179 res = fib6_del(rt, &info); 2180 if (res) { 2181 #if RT6_DEBUG >= 2 2182 pr_debug("%s: del failed: rt=%p@%p err=%d\n", 2183 __func__, rt, 2184 rcu_access_pointer(rt->fib6_node), 2185 res); 2186 #endif 2187 continue; 2188 } 2189 return 0; 2190 } else if (res == -2) { 2191 if (WARN_ON(!rt->fib6_nsiblings)) 2192 continue; 2193 rt = list_last_entry(&rt->fib6_siblings, 2194 struct fib6_info, fib6_siblings); 2195 continue; 2196 } 2197 WARN_ON(res != 0); 2198 } 2199 w->leaf = rt; 2200 return 0; 2201 } 2202 2203 /* 2204 * Convenient frontend to tree walker. 2205 * 2206 * func is called on each route. 2207 * It may return -2 -> skip multipath route. 2208 * -1 -> delete this route. 2209 * 0 -> continue walking 2210 */ 2211 2212 static void fib6_clean_tree(struct net *net, struct fib6_node *root, 2213 int (*func)(struct fib6_info *, void *arg), 2214 int sernum, void *arg, bool skip_notify) 2215 { 2216 struct fib6_cleaner c; 2217 2218 c.w.root = root; 2219 c.w.func = fib6_clean_node; 2220 c.w.count = 0; 2221 c.w.skip = 0; 2222 c.w.skip_in_node = 0; 2223 c.func = func; 2224 c.sernum = sernum; 2225 c.arg = arg; 2226 c.net = net; 2227 c.skip_notify = skip_notify; 2228 2229 fib6_walk(net, &c.w); 2230 } 2231 2232 static void __fib6_clean_all(struct net *net, 2233 int (*func)(struct fib6_info *, void *), 2234 int sernum, void *arg, bool skip_notify) 2235 { 2236 struct fib6_table *table; 2237 struct hlist_head *head; 2238 unsigned int h; 2239 2240 rcu_read_lock(); 2241 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { 2242 head = &net->ipv6.fib_table_hash[h]; 2243 hlist_for_each_entry_rcu(table, head, tb6_hlist) { 2244 spin_lock_bh(&table->tb6_lock); 2245 fib6_clean_tree(net, &table->tb6_root, 2246 func, sernum, arg, skip_notify); 2247 spin_unlock_bh(&table->tb6_lock); 2248 } 2249 } 2250 rcu_read_unlock(); 2251 } 2252 2253 void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *), 2254 void *arg) 2255 { 2256 __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, false); 2257 } 2258 2259 void fib6_clean_all_skip_notify(struct net *net, 2260 int (*func)(struct fib6_info *, void *), 2261 void *arg) 2262 { 2263 __fib6_clean_all(net, func, FIB6_NO_SERNUM_CHANGE, arg, true); 2264 } 2265 2266 static void fib6_flush_trees(struct net *net) 2267 { 2268 int new_sernum = fib6_new_sernum(net); 2269 2270 __fib6_clean_all(net, NULL, new_sernum, NULL, false); 2271 } 2272 2273 /* 2274 * Garbage collection 2275 */ 2276 2277 static int fib6_age(struct fib6_info *rt, void *arg) 2278 { 2279 struct fib6_gc_args *gc_args = arg; 2280 unsigned long now = jiffies; 2281 2282 /* 2283 * check addrconf expiration here. 2284 * Routes are expired even if they are in use. 2285 */ 2286 2287 if (rt->fib6_flags & RTF_EXPIRES && rt->expires) { 2288 if (time_after(now, rt->expires)) { 2289 RT6_TRACE("expiring %p\n", rt); 2290 return -1; 2291 } 2292 gc_args->more++; 2293 } 2294 2295 /* Also age clones in the exception table. 2296 * Note, that clones are aged out 2297 * only if they are not in use now. 2298 */ 2299 rt6_age_exceptions(rt, gc_args, now); 2300 2301 return 0; 2302 } 2303 2304 void fib6_run_gc(unsigned long expires, struct net *net, bool force) 2305 { 2306 struct fib6_gc_args gc_args; 2307 unsigned long now; 2308 2309 if (force) { 2310 spin_lock_bh(&net->ipv6.fib6_gc_lock); 2311 } else if (!spin_trylock_bh(&net->ipv6.fib6_gc_lock)) { 2312 mod_timer(&net->ipv6.ip6_fib_timer, jiffies + HZ); 2313 return; 2314 } 2315 gc_args.timeout = expires ? (int)expires : 2316 net->ipv6.sysctl.ip6_rt_gc_interval; 2317 gc_args.more = 0; 2318 2319 fib6_clean_all(net, fib6_age, &gc_args); 2320 now = jiffies; 2321 net->ipv6.ip6_rt_last_gc = now; 2322 2323 if (gc_args.more) 2324 mod_timer(&net->ipv6.ip6_fib_timer, 2325 round_jiffies(now 2326 + net->ipv6.sysctl.ip6_rt_gc_interval)); 2327 else 2328 del_timer(&net->ipv6.ip6_fib_timer); 2329 spin_unlock_bh(&net->ipv6.fib6_gc_lock); 2330 } 2331 2332 static void fib6_gc_timer_cb(struct timer_list *t) 2333 { 2334 struct net *arg = from_timer(arg, t, ipv6.ip6_fib_timer); 2335 2336 fib6_run_gc(0, arg, true); 2337 } 2338 2339 static int __net_init fib6_net_init(struct net *net) 2340 { 2341 size_t size = sizeof(struct hlist_head) * FIB6_TABLE_HASHSZ; 2342 int err; 2343 2344 err = fib6_notifier_init(net); 2345 if (err) 2346 return err; 2347 2348 spin_lock_init(&net->ipv6.fib6_gc_lock); 2349 rwlock_init(&net->ipv6.fib6_walker_lock); 2350 INIT_LIST_HEAD(&net->ipv6.fib6_walkers); 2351 timer_setup(&net->ipv6.ip6_fib_timer, fib6_gc_timer_cb, 0); 2352 2353 net->ipv6.rt6_stats = kzalloc(sizeof(*net->ipv6.rt6_stats), GFP_KERNEL); 2354 if (!net->ipv6.rt6_stats) 2355 goto out_timer; 2356 2357 /* Avoid false sharing : Use at least a full cache line */ 2358 size = max_t(size_t, size, L1_CACHE_BYTES); 2359 2360 net->ipv6.fib_table_hash = kzalloc(size, GFP_KERNEL); 2361 if (!net->ipv6.fib_table_hash) 2362 goto out_rt6_stats; 2363 2364 net->ipv6.fib6_main_tbl = kzalloc(sizeof(*net->ipv6.fib6_main_tbl), 2365 GFP_KERNEL); 2366 if (!net->ipv6.fib6_main_tbl) 2367 goto out_fib_table_hash; 2368 2369 net->ipv6.fib6_main_tbl->tb6_id = RT6_TABLE_MAIN; 2370 rcu_assign_pointer(net->ipv6.fib6_main_tbl->tb6_root.leaf, 2371 net->ipv6.fib6_null_entry); 2372 net->ipv6.fib6_main_tbl->tb6_root.fn_flags = 2373 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; 2374 inet_peer_base_init(&net->ipv6.fib6_main_tbl->tb6_peers); 2375 2376 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 2377 net->ipv6.fib6_local_tbl = kzalloc(sizeof(*net->ipv6.fib6_local_tbl), 2378 GFP_KERNEL); 2379 if (!net->ipv6.fib6_local_tbl) 2380 goto out_fib6_main_tbl; 2381 net->ipv6.fib6_local_tbl->tb6_id = RT6_TABLE_LOCAL; 2382 rcu_assign_pointer(net->ipv6.fib6_local_tbl->tb6_root.leaf, 2383 net->ipv6.fib6_null_entry); 2384 net->ipv6.fib6_local_tbl->tb6_root.fn_flags = 2385 RTN_ROOT | RTN_TL_ROOT | RTN_RTINFO; 2386 inet_peer_base_init(&net->ipv6.fib6_local_tbl->tb6_peers); 2387 #endif 2388 fib6_tables_init(net); 2389 2390 return 0; 2391 2392 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 2393 out_fib6_main_tbl: 2394 kfree(net->ipv6.fib6_main_tbl); 2395 #endif 2396 out_fib_table_hash: 2397 kfree(net->ipv6.fib_table_hash); 2398 out_rt6_stats: 2399 kfree(net->ipv6.rt6_stats); 2400 out_timer: 2401 fib6_notifier_exit(net); 2402 return -ENOMEM; 2403 } 2404 2405 static void fib6_net_exit(struct net *net) 2406 { 2407 unsigned int i; 2408 2409 del_timer_sync(&net->ipv6.ip6_fib_timer); 2410 2411 for (i = 0; i < FIB6_TABLE_HASHSZ; i++) { 2412 struct hlist_head *head = &net->ipv6.fib_table_hash[i]; 2413 struct hlist_node *tmp; 2414 struct fib6_table *tb; 2415 2416 hlist_for_each_entry_safe(tb, tmp, head, tb6_hlist) { 2417 hlist_del(&tb->tb6_hlist); 2418 fib6_free_table(tb); 2419 } 2420 } 2421 2422 kfree(net->ipv6.fib_table_hash); 2423 kfree(net->ipv6.rt6_stats); 2424 fib6_notifier_exit(net); 2425 } 2426 2427 static struct pernet_operations fib6_net_ops = { 2428 .init = fib6_net_init, 2429 .exit = fib6_net_exit, 2430 }; 2431 2432 int __init fib6_init(void) 2433 { 2434 int ret = -ENOMEM; 2435 2436 fib6_node_kmem = kmem_cache_create("fib6_nodes", 2437 sizeof(struct fib6_node), 2438 0, SLAB_HWCACHE_ALIGN, 2439 NULL); 2440 if (!fib6_node_kmem) 2441 goto out; 2442 2443 ret = register_pernet_subsys(&fib6_net_ops); 2444 if (ret) 2445 goto out_kmem_cache_create; 2446 2447 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL, 2448 inet6_dump_fib, 0); 2449 if (ret) 2450 goto out_unregister_subsys; 2451 2452 __fib6_flush_trees = fib6_flush_trees; 2453 out: 2454 return ret; 2455 2456 out_unregister_subsys: 2457 unregister_pernet_subsys(&fib6_net_ops); 2458 out_kmem_cache_create: 2459 kmem_cache_destroy(fib6_node_kmem); 2460 goto out; 2461 } 2462 2463 void fib6_gc_cleanup(void) 2464 { 2465 unregister_pernet_subsys(&fib6_net_ops); 2466 kmem_cache_destroy(fib6_node_kmem); 2467 } 2468 2469 #ifdef CONFIG_PROC_FS 2470 static int ipv6_route_seq_show(struct seq_file *seq, void *v) 2471 { 2472 struct fib6_info *rt = v; 2473 struct ipv6_route_iter *iter = seq->private; 2474 struct fib6_nh *fib6_nh = rt->fib6_nh; 2475 unsigned int flags = rt->fib6_flags; 2476 const struct net_device *dev; 2477 2478 if (rt->nh) 2479 fib6_nh = nexthop_fib6_nh(rt->nh); 2480 2481 seq_printf(seq, "%pi6 %02x ", &rt->fib6_dst.addr, rt->fib6_dst.plen); 2482 2483 #ifdef CONFIG_IPV6_SUBTREES 2484 seq_printf(seq, "%pi6 %02x ", &rt->fib6_src.addr, rt->fib6_src.plen); 2485 #else 2486 seq_puts(seq, "00000000000000000000000000000000 00 "); 2487 #endif 2488 if (fib6_nh->fib_nh_gw_family) { 2489 flags |= RTF_GATEWAY; 2490 seq_printf(seq, "%pi6", &fib6_nh->fib_nh_gw6); 2491 } else { 2492 seq_puts(seq, "00000000000000000000000000000000"); 2493 } 2494 2495 dev = fib6_nh->fib_nh_dev; 2496 seq_printf(seq, " %08x %08x %08x %08x %8s\n", 2497 rt->fib6_metric, refcount_read(&rt->fib6_ref), 0, 2498 flags, dev ? dev->name : ""); 2499 iter->w.leaf = NULL; 2500 return 0; 2501 } 2502 2503 static int ipv6_route_yield(struct fib6_walker *w) 2504 { 2505 struct ipv6_route_iter *iter = w->args; 2506 2507 if (!iter->skip) 2508 return 1; 2509 2510 do { 2511 iter->w.leaf = rcu_dereference_protected( 2512 iter->w.leaf->fib6_next, 2513 lockdep_is_held(&iter->tbl->tb6_lock)); 2514 iter->skip--; 2515 if (!iter->skip && iter->w.leaf) 2516 return 1; 2517 } while (iter->w.leaf); 2518 2519 return 0; 2520 } 2521 2522 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter *iter, 2523 struct net *net) 2524 { 2525 memset(&iter->w, 0, sizeof(iter->w)); 2526 iter->w.func = ipv6_route_yield; 2527 iter->w.root = &iter->tbl->tb6_root; 2528 iter->w.state = FWS_INIT; 2529 iter->w.node = iter->w.root; 2530 iter->w.args = iter; 2531 iter->sernum = iter->w.root->fn_sernum; 2532 INIT_LIST_HEAD(&iter->w.lh); 2533 fib6_walker_link(net, &iter->w); 2534 } 2535 2536 static struct fib6_table *ipv6_route_seq_next_table(struct fib6_table *tbl, 2537 struct net *net) 2538 { 2539 unsigned int h; 2540 struct hlist_node *node; 2541 2542 if (tbl) { 2543 h = (tbl->tb6_id & (FIB6_TABLE_HASHSZ - 1)) + 1; 2544 node = rcu_dereference_bh(hlist_next_rcu(&tbl->tb6_hlist)); 2545 } else { 2546 h = 0; 2547 node = NULL; 2548 } 2549 2550 while (!node && h < FIB6_TABLE_HASHSZ) { 2551 node = rcu_dereference_bh( 2552 hlist_first_rcu(&net->ipv6.fib_table_hash[h++])); 2553 } 2554 return hlist_entry_safe(node, struct fib6_table, tb6_hlist); 2555 } 2556 2557 static void ipv6_route_check_sernum(struct ipv6_route_iter *iter) 2558 { 2559 if (iter->sernum != iter->w.root->fn_sernum) { 2560 iter->sernum = iter->w.root->fn_sernum; 2561 iter->w.state = FWS_INIT; 2562 iter->w.node = iter->w.root; 2563 WARN_ON(iter->w.skip); 2564 iter->w.skip = iter->w.count; 2565 } 2566 } 2567 2568 static void *ipv6_route_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2569 { 2570 int r; 2571 struct fib6_info *n; 2572 struct net *net = seq_file_net(seq); 2573 struct ipv6_route_iter *iter = seq->private; 2574 2575 ++(*pos); 2576 if (!v) 2577 goto iter_table; 2578 2579 n = rcu_dereference_bh(((struct fib6_info *)v)->fib6_next); 2580 if (n) 2581 return n; 2582 2583 iter_table: 2584 ipv6_route_check_sernum(iter); 2585 spin_lock_bh(&iter->tbl->tb6_lock); 2586 r = fib6_walk_continue(&iter->w); 2587 spin_unlock_bh(&iter->tbl->tb6_lock); 2588 if (r > 0) { 2589 return iter->w.leaf; 2590 } else if (r < 0) { 2591 fib6_walker_unlink(net, &iter->w); 2592 return NULL; 2593 } 2594 fib6_walker_unlink(net, &iter->w); 2595 2596 iter->tbl = ipv6_route_seq_next_table(iter->tbl, net); 2597 if (!iter->tbl) 2598 return NULL; 2599 2600 ipv6_route_seq_setup_walk(iter, net); 2601 goto iter_table; 2602 } 2603 2604 static void *ipv6_route_seq_start(struct seq_file *seq, loff_t *pos) 2605 __acquires(RCU_BH) 2606 { 2607 struct net *net = seq_file_net(seq); 2608 struct ipv6_route_iter *iter = seq->private; 2609 2610 rcu_read_lock_bh(); 2611 iter->tbl = ipv6_route_seq_next_table(NULL, net); 2612 iter->skip = *pos; 2613 2614 if (iter->tbl) { 2615 ipv6_route_seq_setup_walk(iter, net); 2616 return ipv6_route_seq_next(seq, NULL, pos); 2617 } else { 2618 return NULL; 2619 } 2620 } 2621 2622 static bool ipv6_route_iter_active(struct ipv6_route_iter *iter) 2623 { 2624 struct fib6_walker *w = &iter->w; 2625 return w->node && !(w->state == FWS_U && w->node == w->root); 2626 } 2627 2628 static void ipv6_route_seq_stop(struct seq_file *seq, void *v) 2629 __releases(RCU_BH) 2630 { 2631 struct net *net = seq_file_net(seq); 2632 struct ipv6_route_iter *iter = seq->private; 2633 2634 if (ipv6_route_iter_active(iter)) 2635 fib6_walker_unlink(net, &iter->w); 2636 2637 rcu_read_unlock_bh(); 2638 } 2639 2640 const struct seq_operations ipv6_route_seq_ops = { 2641 .start = ipv6_route_seq_start, 2642 .next = ipv6_route_seq_next, 2643 .stop = ipv6_route_seq_stop, 2644 .show = ipv6_route_seq_show 2645 }; 2646 #endif /* CONFIG_PROC_FS */ 2647