1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 */ 8 9 #ifndef _IP6_FIB_H 10 #define _IP6_FIB_H 11 12 #include <linux/ipv6_route.h> 13 #include <linux/rtnetlink.h> 14 #include <linux/spinlock.h> 15 #include <linux/notifier.h> 16 #include <net/dst.h> 17 #include <net/flow.h> 18 #include <net/ip_fib.h> 19 #include <net/netlink.h> 20 #include <net/inetpeer.h> 21 #include <net/fib_notifier.h> 22 #include <linux/indirect_call_wrapper.h> 23 #include <uapi/linux/bpf.h> 24 25 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 26 #define FIB6_TABLE_HASHSZ 256 27 #else 28 #define FIB6_TABLE_HASHSZ 1 29 #endif 30 31 #define RT6_DEBUG 2 32 33 #if RT6_DEBUG >= 3 34 #define RT6_TRACE(x...) pr_debug(x) 35 #else 36 #define RT6_TRACE(x...) do { ; } while (0) 37 #endif 38 39 struct rt6_info; 40 struct fib6_info; 41 42 struct fib6_config { 43 u32 fc_table; 44 u32 fc_metric; 45 int fc_dst_len; 46 int fc_src_len; 47 int fc_ifindex; 48 u32 fc_flags; 49 u32 fc_protocol; 50 u16 fc_type; /* only 8 bits are used */ 51 u16 fc_delete_all_nh : 1, 52 fc_ignore_dev_down:1, 53 __unused : 14; 54 u32 fc_nh_id; 55 56 struct in6_addr fc_dst; 57 struct in6_addr fc_src; 58 struct in6_addr fc_prefsrc; 59 struct in6_addr fc_gateway; 60 61 unsigned long fc_expires; 62 struct nlattr *fc_mx; 63 int fc_mx_len; 64 int fc_mp_len; 65 struct nlattr *fc_mp; 66 67 struct nl_info fc_nlinfo; 68 struct nlattr *fc_encap; 69 u16 fc_encap_type; 70 bool fc_is_fdb; 71 }; 72 73 struct fib6_node { 74 struct fib6_node __rcu *parent; 75 struct fib6_node __rcu *left; 76 struct fib6_node __rcu *right; 77 #ifdef CONFIG_IPV6_SUBTREES 78 struct fib6_node __rcu *subtree; 79 #endif 80 struct fib6_info __rcu *leaf; 81 82 __u16 fn_bit; /* bit key */ 83 __u16 fn_flags; 84 int fn_sernum; 85 struct fib6_info __rcu *rr_ptr; 86 struct rcu_head rcu; 87 }; 88 89 struct fib6_gc_args { 90 int timeout; 91 int more; 92 }; 93 94 #ifndef CONFIG_IPV6_SUBTREES 95 #define FIB6_SUBTREE(fn) NULL 96 97 static inline bool fib6_routes_require_src(const struct net *net) 98 { 99 return false; 100 } 101 102 static inline void fib6_routes_require_src_inc(struct net *net) {} 103 static inline void fib6_routes_require_src_dec(struct net *net) {} 104 105 #else 106 107 static inline bool fib6_routes_require_src(const struct net *net) 108 { 109 return net->ipv6.fib6_routes_require_src > 0; 110 } 111 112 static inline void fib6_routes_require_src_inc(struct net *net) 113 { 114 net->ipv6.fib6_routes_require_src++; 115 } 116 117 static inline void fib6_routes_require_src_dec(struct net *net) 118 { 119 net->ipv6.fib6_routes_require_src--; 120 } 121 122 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1)) 123 #endif 124 125 /* 126 * routing information 127 * 128 */ 129 130 struct rt6key { 131 struct in6_addr addr; 132 int plen; 133 }; 134 135 struct fib6_table; 136 137 struct rt6_exception_bucket { 138 struct hlist_head chain; 139 int depth; 140 }; 141 142 struct rt6_exception { 143 struct hlist_node hlist; 144 struct rt6_info *rt6i; 145 unsigned long stamp; 146 struct rcu_head rcu; 147 }; 148 149 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10 150 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT) 151 #define FIB6_MAX_DEPTH 5 152 153 struct fib6_nh { 154 struct fib_nh_common nh_common; 155 156 #ifdef CONFIG_IPV6_ROUTER_PREF 157 unsigned long last_probe; 158 #endif 159 160 struct rt6_info * __percpu *rt6i_pcpu; 161 struct rt6_exception_bucket __rcu *rt6i_exception_bucket; 162 }; 163 164 struct fib6_info { 165 struct fib6_table *fib6_table; 166 struct fib6_info __rcu *fib6_next; 167 struct fib6_node __rcu *fib6_node; 168 169 /* Multipath routes: 170 * siblings is a list of fib6_info that have the same metric/weight, 171 * destination, but not the same gateway. nsiblings is just a cache 172 * to speed up lookup. 173 */ 174 union { 175 struct list_head fib6_siblings; 176 struct list_head nh_list; 177 }; 178 unsigned int fib6_nsiblings; 179 180 refcount_t fib6_ref; 181 unsigned long expires; 182 183 struct hlist_node gc_link; 184 185 struct dst_metrics *fib6_metrics; 186 #define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1] 187 188 struct rt6key fib6_dst; 189 u32 fib6_flags; 190 struct rt6key fib6_src; 191 struct rt6key fib6_prefsrc; 192 193 u32 fib6_metric; 194 u8 fib6_protocol; 195 u8 fib6_type; 196 197 u8 offload; 198 u8 trap; 199 u8 offload_failed; 200 201 u8 should_flush:1, 202 dst_nocount:1, 203 dst_nopolicy:1, 204 fib6_destroying:1, 205 unused:4; 206 207 struct rcu_head rcu; 208 struct nexthop *nh; 209 struct fib6_nh fib6_nh[]; 210 }; 211 212 struct rt6_info { 213 struct dst_entry dst; 214 struct fib6_info __rcu *from; 215 int sernum; 216 217 struct rt6key rt6i_dst; 218 struct rt6key rt6i_src; 219 struct in6_addr rt6i_gateway; 220 struct inet6_dev *rt6i_idev; 221 u32 rt6i_flags; 222 223 /* more non-fragment space at head required */ 224 unsigned short rt6i_nfheader_len; 225 }; 226 227 struct fib6_result { 228 struct fib6_nh *nh; 229 struct fib6_info *f6i; 230 u32 fib6_flags; 231 u8 fib6_type; 232 struct rt6_info *rt6; 233 }; 234 235 #define for_each_fib6_node_rt_rcu(fn) \ 236 for (rt = rcu_dereference((fn)->leaf); rt; \ 237 rt = rcu_dereference(rt->fib6_next)) 238 239 #define for_each_fib6_walker_rt(w) \ 240 for (rt = (w)->leaf; rt; \ 241 rt = rcu_dereference_protected(rt->fib6_next, 1)) 242 243 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst) 244 { 245 return ((struct rt6_info *)dst)->rt6i_idev; 246 } 247 248 static inline bool fib6_requires_src(const struct fib6_info *rt) 249 { 250 return rt->fib6_src.plen > 0; 251 } 252 253 static inline bool fib6_check_expired(const struct fib6_info *f6i) 254 { 255 if (f6i->fib6_flags & RTF_EXPIRES) 256 return time_after(jiffies, f6i->expires); 257 return false; 258 } 259 260 static inline bool fib6_has_expires(const struct fib6_info *f6i) 261 { 262 return f6i->fib6_flags & RTF_EXPIRES; 263 } 264 265 /* Function to safely get fn->fn_sernum for passed in rt 266 * and store result in passed in cookie. 267 * Return true if we can get cookie safely 268 * Return false if not 269 */ 270 static inline bool fib6_get_cookie_safe(const struct fib6_info *f6i, 271 u32 *cookie) 272 { 273 struct fib6_node *fn; 274 bool status = false; 275 276 fn = rcu_dereference(f6i->fib6_node); 277 278 if (fn) { 279 *cookie = READ_ONCE(fn->fn_sernum); 280 /* pairs with smp_wmb() in __fib6_update_sernum_upto_root() */ 281 smp_rmb(); 282 status = true; 283 } 284 285 return status; 286 } 287 288 static inline u32 rt6_get_cookie(const struct rt6_info *rt) 289 { 290 struct fib6_info *from; 291 u32 cookie = 0; 292 293 if (rt->sernum) 294 return rt->sernum; 295 296 rcu_read_lock(); 297 298 from = rcu_dereference(rt->from); 299 if (from) 300 fib6_get_cookie_safe(from, &cookie); 301 302 rcu_read_unlock(); 303 304 return cookie; 305 } 306 307 static inline void ip6_rt_put(struct rt6_info *rt) 308 { 309 /* dst_release() accepts a NULL parameter. 310 * We rely on dst being first structure in struct rt6_info 311 */ 312 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0); 313 dst_release(&rt->dst); 314 } 315 316 struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh); 317 void fib6_info_destroy_rcu(struct rcu_head *head); 318 319 static inline void fib6_info_hold(struct fib6_info *f6i) 320 { 321 refcount_inc(&f6i->fib6_ref); 322 } 323 324 static inline bool fib6_info_hold_safe(struct fib6_info *f6i) 325 { 326 return refcount_inc_not_zero(&f6i->fib6_ref); 327 } 328 329 static inline void fib6_info_release(struct fib6_info *f6i) 330 { 331 if (f6i && refcount_dec_and_test(&f6i->fib6_ref)) 332 call_rcu(&f6i->rcu, fib6_info_destroy_rcu); 333 } 334 335 enum fib6_walk_state { 336 #ifdef CONFIG_IPV6_SUBTREES 337 FWS_S, 338 #endif 339 FWS_L, 340 FWS_R, 341 FWS_C, 342 FWS_U 343 }; 344 345 struct fib6_walker { 346 struct list_head lh; 347 struct fib6_node *root, *node; 348 struct fib6_info *leaf; 349 enum fib6_walk_state state; 350 unsigned int skip; 351 unsigned int count; 352 unsigned int skip_in_node; 353 int (*func)(struct fib6_walker *); 354 void *args; 355 }; 356 357 struct rt6_statistics { 358 __u32 fib_nodes; /* all fib6 nodes */ 359 __u32 fib_route_nodes; /* intermediate nodes */ 360 __u32 fib_rt_entries; /* rt entries in fib table */ 361 __u32 fib_rt_cache; /* cached rt entries in exception table */ 362 __u32 fib_discarded_routes; /* total number of routes delete */ 363 364 /* The following stat is not protected by any lock */ 365 atomic_t fib_rt_alloc; /* total number of routes alloced */ 366 }; 367 368 #define RTN_TL_ROOT 0x0001 369 #define RTN_ROOT 0x0002 /* tree root node */ 370 #define RTN_RTINFO 0x0004 /* node with valid routing info */ 371 372 /* 373 * priority levels (or metrics) 374 * 375 */ 376 377 378 struct fib6_table { 379 struct hlist_node tb6_hlist; 380 u32 tb6_id; 381 spinlock_t tb6_lock; 382 struct fib6_node tb6_root; 383 struct inet_peer_base tb6_peers; 384 unsigned int flags; 385 unsigned int fib_seq; 386 struct hlist_head tb6_gc_hlist; /* GC candidates */ 387 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0) 388 }; 389 390 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC 391 #define RT6_TABLE_MAIN RT_TABLE_MAIN 392 #define RT6_TABLE_DFLT RT6_TABLE_MAIN 393 #define RT6_TABLE_INFO RT6_TABLE_MAIN 394 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN 395 396 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 397 #define FIB6_TABLE_MIN 1 398 #define FIB6_TABLE_MAX RT_TABLE_MAX 399 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL 400 #else 401 #define FIB6_TABLE_MIN RT_TABLE_MAIN 402 #define FIB6_TABLE_MAX FIB6_TABLE_MIN 403 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN 404 #endif 405 406 typedef struct rt6_info *(*pol_lookup_t)(struct net *, 407 struct fib6_table *, 408 struct flowi6 *, 409 const struct sk_buff *, int); 410 411 struct fib6_entry_notifier_info { 412 struct fib_notifier_info info; /* must be first */ 413 struct fib6_info *rt; 414 unsigned int nsiblings; 415 }; 416 417 /* 418 * exported functions 419 */ 420 421 struct fib6_table *fib6_get_table(struct net *net, u32 id); 422 struct fib6_table *fib6_new_table(struct net *net, u32 id); 423 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6, 424 const struct sk_buff *skb, 425 int flags, pol_lookup_t lookup); 426 427 /* called with rcu lock held; can return error pointer 428 * caller needs to select path 429 */ 430 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6, 431 struct fib6_result *res, int flags); 432 433 /* called with rcu lock held; caller needs to select path */ 434 int fib6_table_lookup(struct net *net, struct fib6_table *table, 435 int oif, struct flowi6 *fl6, struct fib6_result *res, 436 int strict); 437 438 void fib6_select_path(const struct net *net, struct fib6_result *res, 439 struct flowi6 *fl6, int oif, bool have_oif_match, 440 const struct sk_buff *skb, int strict); 441 struct fib6_node *fib6_node_lookup(struct fib6_node *root, 442 const struct in6_addr *daddr, 443 const struct in6_addr *saddr); 444 445 struct fib6_node *fib6_locate(struct fib6_node *root, 446 const struct in6_addr *daddr, int dst_len, 447 const struct in6_addr *saddr, int src_len, 448 bool exact_match); 449 450 void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *arg), 451 void *arg); 452 void fib6_clean_all_skip_notify(struct net *net, 453 int (*func)(struct fib6_info *, void *arg), 454 void *arg); 455 456 int fib6_add(struct fib6_node *root, struct fib6_info *rt, 457 struct nl_info *info, struct netlink_ext_ack *extack); 458 int fib6_del(struct fib6_info *rt, struct nl_info *info); 459 460 static inline 461 void rt6_get_prefsrc(const struct rt6_info *rt, struct in6_addr *addr) 462 { 463 const struct fib6_info *from; 464 465 rcu_read_lock(); 466 467 from = rcu_dereference(rt->from); 468 if (from) 469 *addr = from->fib6_prefsrc.addr; 470 else 471 *addr = in6addr_any; 472 473 rcu_read_unlock(); 474 } 475 476 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh, 477 struct fib6_config *cfg, gfp_t gfp_flags, 478 struct netlink_ext_ack *extack); 479 void fib6_nh_release(struct fib6_nh *fib6_nh); 480 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh); 481 482 int call_fib6_entry_notifiers(struct net *net, 483 enum fib_event_type event_type, 484 struct fib6_info *rt, 485 struct netlink_ext_ack *extack); 486 int call_fib6_multipath_entry_notifiers(struct net *net, 487 enum fib_event_type event_type, 488 struct fib6_info *rt, 489 unsigned int nsiblings, 490 struct netlink_ext_ack *extack); 491 int call_fib6_entry_notifiers_replace(struct net *net, struct fib6_info *rt); 492 void fib6_rt_update(struct net *net, struct fib6_info *rt, 493 struct nl_info *info); 494 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info, 495 unsigned int flags); 496 497 void fib6_run_gc(unsigned long expires, struct net *net, bool force); 498 499 void fib6_gc_cleanup(void); 500 501 int fib6_init(void); 502 503 /* fib6_info must be locked by the caller, and fib6_info->fib6_table can be 504 * NULL. 505 */ 506 static inline void fib6_set_expires_locked(struct fib6_info *f6i, 507 unsigned long expires) 508 { 509 struct fib6_table *tb6; 510 511 tb6 = f6i->fib6_table; 512 f6i->expires = expires; 513 if (tb6 && !fib6_has_expires(f6i)) 514 hlist_add_head(&f6i->gc_link, &tb6->tb6_gc_hlist); 515 f6i->fib6_flags |= RTF_EXPIRES; 516 } 517 518 /* fib6_info must be locked by the caller, and fib6_info->fib6_table can be 519 * NULL. If fib6_table is NULL, the fib6_info will no be inserted into the 520 * list of GC candidates until it is inserted into a table. 521 */ 522 static inline void fib6_set_expires(struct fib6_info *f6i, 523 unsigned long expires) 524 { 525 spin_lock_bh(&f6i->fib6_table->tb6_lock); 526 fib6_set_expires_locked(f6i, expires); 527 spin_unlock_bh(&f6i->fib6_table->tb6_lock); 528 } 529 530 static inline void fib6_clean_expires_locked(struct fib6_info *f6i) 531 { 532 if (fib6_has_expires(f6i)) 533 hlist_del_init(&f6i->gc_link); 534 f6i->fib6_flags &= ~RTF_EXPIRES; 535 f6i->expires = 0; 536 } 537 538 static inline void fib6_clean_expires(struct fib6_info *f6i) 539 { 540 spin_lock_bh(&f6i->fib6_table->tb6_lock); 541 fib6_clean_expires_locked(f6i); 542 spin_unlock_bh(&f6i->fib6_table->tb6_lock); 543 } 544 545 struct ipv6_route_iter { 546 struct seq_net_private p; 547 struct fib6_walker w; 548 loff_t skip; 549 struct fib6_table *tbl; 550 int sernum; 551 }; 552 553 extern const struct seq_operations ipv6_route_seq_ops; 554 555 int call_fib6_notifier(struct notifier_block *nb, 556 enum fib_event_type event_type, 557 struct fib_notifier_info *info); 558 int call_fib6_notifiers(struct net *net, enum fib_event_type event_type, 559 struct fib_notifier_info *info); 560 561 int __net_init fib6_notifier_init(struct net *net); 562 void __net_exit fib6_notifier_exit(struct net *net); 563 564 unsigned int fib6_tables_seq_read(struct net *net); 565 int fib6_tables_dump(struct net *net, struct notifier_block *nb, 566 struct netlink_ext_ack *extack); 567 568 void fib6_update_sernum(struct net *net, struct fib6_info *rt); 569 void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt); 570 void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i); 571 572 void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val); 573 static inline bool fib6_metric_locked(struct fib6_info *f6i, int metric) 574 { 575 return !!(f6i->fib6_metrics->metrics[RTAX_LOCK - 1] & (1 << metric)); 576 } 577 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i, 578 bool offload, bool trap, bool offload_failed); 579 580 #if IS_BUILTIN(CONFIG_IPV6) && defined(CONFIG_BPF_SYSCALL) 581 struct bpf_iter__ipv6_route { 582 __bpf_md_ptr(struct bpf_iter_meta *, meta); 583 __bpf_md_ptr(struct fib6_info *, rt); 584 }; 585 #endif 586 587 INDIRECT_CALLABLE_DECLARE(struct rt6_info *ip6_pol_route_output(struct net *net, 588 struct fib6_table *table, 589 struct flowi6 *fl6, 590 const struct sk_buff *skb, 591 int flags)); 592 INDIRECT_CALLABLE_DECLARE(struct rt6_info *ip6_pol_route_input(struct net *net, 593 struct fib6_table *table, 594 struct flowi6 *fl6, 595 const struct sk_buff *skb, 596 int flags)); 597 INDIRECT_CALLABLE_DECLARE(struct rt6_info *__ip6_route_redirect(struct net *net, 598 struct fib6_table *table, 599 struct flowi6 *fl6, 600 const struct sk_buff *skb, 601 int flags)); 602 INDIRECT_CALLABLE_DECLARE(struct rt6_info *ip6_pol_route_lookup(struct net *net, 603 struct fib6_table *table, 604 struct flowi6 *fl6, 605 const struct sk_buff *skb, 606 int flags)); 607 static inline struct rt6_info *pol_lookup_func(pol_lookup_t lookup, 608 struct net *net, 609 struct fib6_table *table, 610 struct flowi6 *fl6, 611 const struct sk_buff *skb, 612 int flags) 613 { 614 return INDIRECT_CALL_4(lookup, 615 ip6_pol_route_output, 616 ip6_pol_route_input, 617 ip6_pol_route_lookup, 618 __ip6_route_redirect, 619 net, table, fl6, skb, flags); 620 } 621 622 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 623 static inline bool fib6_has_custom_rules(const struct net *net) 624 { 625 return net->ipv6.fib6_has_custom_rules; 626 } 627 628 int fib6_rules_init(void); 629 void fib6_rules_cleanup(void); 630 bool fib6_rule_default(const struct fib_rule *rule); 631 int fib6_rules_dump(struct net *net, struct notifier_block *nb, 632 struct netlink_ext_ack *extack); 633 unsigned int fib6_rules_seq_read(struct net *net); 634 635 static inline bool fib6_rules_early_flow_dissect(struct net *net, 636 struct sk_buff *skb, 637 struct flowi6 *fl6, 638 struct flow_keys *flkeys) 639 { 640 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP; 641 642 if (!net->ipv6.fib6_rules_require_fldissect) 643 return false; 644 645 memset(flkeys, 0, sizeof(*flkeys)); 646 __skb_flow_dissect(net, skb, &flow_keys_dissector, 647 flkeys, NULL, 0, 0, 0, flag); 648 649 fl6->fl6_sport = flkeys->ports.src; 650 fl6->fl6_dport = flkeys->ports.dst; 651 fl6->flowi6_proto = flkeys->basic.ip_proto; 652 653 return true; 654 } 655 #else 656 static inline bool fib6_has_custom_rules(const struct net *net) 657 { 658 return false; 659 } 660 static inline int fib6_rules_init(void) 661 { 662 return 0; 663 } 664 static inline void fib6_rules_cleanup(void) 665 { 666 return ; 667 } 668 static inline bool fib6_rule_default(const struct fib_rule *rule) 669 { 670 return true; 671 } 672 static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb, 673 struct netlink_ext_ack *extack) 674 { 675 return 0; 676 } 677 static inline unsigned int fib6_rules_seq_read(struct net *net) 678 { 679 return 0; 680 } 681 static inline bool fib6_rules_early_flow_dissect(struct net *net, 682 struct sk_buff *skb, 683 struct flowi6 *fl6, 684 struct flow_keys *flkeys) 685 { 686 return false; 687 } 688 #endif 689 #endif 690