1 /* 2 * Linux INET6 implementation 3 * 4 * Authors: 5 * Pedro Roque <roque@di.fc.ul.pt> 6 * 7 * This program is free software; you can redistribute it and/or 8 * modify it under the terms of the GNU General Public License 9 * as published by the Free Software Foundation; either version 10 * 2 of the License, or (at your option) any later version. 11 */ 12 13 #ifndef _IP6_FIB_H 14 #define _IP6_FIB_H 15 16 #include <linux/ipv6_route.h> 17 #include <linux/rtnetlink.h> 18 #include <linux/spinlock.h> 19 #include <linux/notifier.h> 20 #include <net/dst.h> 21 #include <net/flow.h> 22 #include <net/netlink.h> 23 #include <net/inetpeer.h> 24 #include <net/fib_notifier.h> 25 26 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 27 #define FIB6_TABLE_HASHSZ 256 28 #else 29 #define FIB6_TABLE_HASHSZ 1 30 #endif 31 32 #define RT6_DEBUG 2 33 34 #if RT6_DEBUG >= 3 35 #define RT6_TRACE(x...) pr_debug(x) 36 #else 37 #define RT6_TRACE(x...) do { ; } while (0) 38 #endif 39 40 struct rt6_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 __unused : 15; 53 54 struct in6_addr fc_dst; 55 struct in6_addr fc_src; 56 struct in6_addr fc_prefsrc; 57 struct in6_addr fc_gateway; 58 59 unsigned long fc_expires; 60 struct nlattr *fc_mx; 61 int fc_mx_len; 62 int fc_mp_len; 63 struct nlattr *fc_mp; 64 65 struct nl_info fc_nlinfo; 66 struct nlattr *fc_encap; 67 u16 fc_encap_type; 68 }; 69 70 struct fib6_node { 71 struct fib6_node __rcu *parent; 72 struct fib6_node __rcu *left; 73 struct fib6_node __rcu *right; 74 #ifdef CONFIG_IPV6_SUBTREES 75 struct fib6_node __rcu *subtree; 76 #endif 77 struct rt6_info __rcu *leaf; 78 79 __u16 fn_bit; /* bit key */ 80 __u16 fn_flags; 81 int fn_sernum; 82 struct rt6_info __rcu *rr_ptr; 83 struct rcu_head rcu; 84 }; 85 86 struct fib6_gc_args { 87 int timeout; 88 int more; 89 }; 90 91 #ifndef CONFIG_IPV6_SUBTREES 92 #define FIB6_SUBTREE(fn) NULL 93 #else 94 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1)) 95 #endif 96 97 struct mx6_config { 98 const u32 *mx; 99 DECLARE_BITMAP(mx_valid, RTAX_MAX); 100 }; 101 102 /* 103 * routing information 104 * 105 */ 106 107 struct rt6key { 108 struct in6_addr addr; 109 int plen; 110 }; 111 112 struct fib6_table; 113 114 struct rt6_exception_bucket { 115 struct hlist_head chain; 116 int depth; 117 }; 118 119 struct rt6_exception { 120 struct hlist_node hlist; 121 struct rt6_info *rt6i; 122 unsigned long stamp; 123 struct rcu_head rcu; 124 }; 125 126 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10 127 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT) 128 #define FIB6_MAX_DEPTH 5 129 130 struct rt6_info { 131 struct dst_entry dst; 132 struct rt6_info __rcu *rt6_next; 133 struct rt6_info *from; 134 135 /* 136 * Tail elements of dst_entry (__refcnt etc.) 137 * and these elements (rarely used in hot path) are in 138 * the same cache line. 139 */ 140 struct fib6_table *rt6i_table; 141 struct fib6_node __rcu *rt6i_node; 142 143 struct in6_addr rt6i_gateway; 144 145 /* Multipath routes: 146 * siblings is a list of rt6_info that have the the same metric/weight, 147 * destination, but not the same gateway. nsiblings is just a cache 148 * to speed up lookup. 149 */ 150 struct list_head rt6i_siblings; 151 unsigned int rt6i_nsiblings; 152 atomic_t rt6i_nh_upper_bound; 153 154 atomic_t rt6i_ref; 155 156 unsigned int rt6i_nh_flags; 157 158 /* These are in a separate cache line. */ 159 struct rt6key rt6i_dst ____cacheline_aligned_in_smp; 160 u32 rt6i_flags; 161 struct rt6key rt6i_src; 162 struct rt6key rt6i_prefsrc; 163 164 struct list_head rt6i_uncached; 165 struct uncached_list *rt6i_uncached_list; 166 167 struct inet6_dev *rt6i_idev; 168 struct rt6_info * __percpu *rt6i_pcpu; 169 struct rt6_exception_bucket __rcu *rt6i_exception_bucket; 170 171 u32 rt6i_metric; 172 u32 rt6i_pmtu; 173 /* more non-fragment space at head required */ 174 int rt6i_nh_weight; 175 unsigned short rt6i_nfheader_len; 176 u8 rt6i_protocol; 177 u8 exception_bucket_flushed:1, 178 should_flush:1, 179 unused:6; 180 }; 181 182 #define for_each_fib6_node_rt_rcu(fn) \ 183 for (rt = rcu_dereference((fn)->leaf); rt; \ 184 rt = rcu_dereference(rt->rt6_next)) 185 186 #define for_each_fib6_walker_rt(w) \ 187 for (rt = (w)->leaf; rt; \ 188 rt = rcu_dereference_protected(rt->rt6_next, 1)) 189 190 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst) 191 { 192 return ((struct rt6_info *)dst)->rt6i_idev; 193 } 194 195 static inline void rt6_clean_expires(struct rt6_info *rt) 196 { 197 rt->rt6i_flags &= ~RTF_EXPIRES; 198 rt->dst.expires = 0; 199 } 200 201 static inline void rt6_set_expires(struct rt6_info *rt, unsigned long expires) 202 { 203 rt->dst.expires = expires; 204 rt->rt6i_flags |= RTF_EXPIRES; 205 } 206 207 static inline void rt6_update_expires(struct rt6_info *rt0, int timeout) 208 { 209 struct rt6_info *rt; 210 211 for (rt = rt0; rt && !(rt->rt6i_flags & RTF_EXPIRES); rt = rt->from); 212 if (rt && rt != rt0) 213 rt0->dst.expires = rt->dst.expires; 214 dst_set_expires(&rt0->dst, timeout); 215 rt0->rt6i_flags |= RTF_EXPIRES; 216 } 217 218 /* Function to safely get fn->sernum for passed in rt 219 * and store result in passed in cookie. 220 * Return true if we can get cookie safely 221 * Return false if not 222 */ 223 static inline bool rt6_get_cookie_safe(const struct rt6_info *rt, 224 u32 *cookie) 225 { 226 struct fib6_node *fn; 227 bool status = false; 228 229 rcu_read_lock(); 230 fn = rcu_dereference(rt->rt6i_node); 231 232 if (fn) { 233 *cookie = fn->fn_sernum; 234 /* pairs with smp_wmb() in fib6_update_sernum_upto_root() */ 235 smp_rmb(); 236 status = true; 237 } 238 239 rcu_read_unlock(); 240 return status; 241 } 242 243 static inline u32 rt6_get_cookie(const struct rt6_info *rt) 244 { 245 u32 cookie = 0; 246 247 if (rt->rt6i_flags & RTF_PCPU || 248 (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->from)) 249 rt = rt->from; 250 251 rt6_get_cookie_safe(rt, &cookie); 252 253 return cookie; 254 } 255 256 static inline void ip6_rt_put(struct rt6_info *rt) 257 { 258 /* dst_release() accepts a NULL parameter. 259 * We rely on dst being first structure in struct rt6_info 260 */ 261 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0); 262 dst_release(&rt->dst); 263 } 264 265 void rt6_free_pcpu(struct rt6_info *non_pcpu_rt); 266 267 static inline void rt6_hold(struct rt6_info *rt) 268 { 269 atomic_inc(&rt->rt6i_ref); 270 } 271 272 static inline void rt6_release(struct rt6_info *rt) 273 { 274 if (atomic_dec_and_test(&rt->rt6i_ref)) { 275 rt6_free_pcpu(rt); 276 dst_dev_put(&rt->dst); 277 dst_release(&rt->dst); 278 } 279 } 280 281 enum fib6_walk_state { 282 #ifdef CONFIG_IPV6_SUBTREES 283 FWS_S, 284 #endif 285 FWS_L, 286 FWS_R, 287 FWS_C, 288 FWS_U 289 }; 290 291 struct fib6_walker { 292 struct list_head lh; 293 struct fib6_node *root, *node; 294 struct rt6_info *leaf; 295 enum fib6_walk_state state; 296 unsigned int skip; 297 unsigned int count; 298 int (*func)(struct fib6_walker *); 299 void *args; 300 }; 301 302 struct rt6_statistics { 303 __u32 fib_nodes; /* all fib6 nodes */ 304 __u32 fib_route_nodes; /* intermediate nodes */ 305 __u32 fib_rt_entries; /* rt entries in fib table */ 306 __u32 fib_rt_cache; /* cached rt entries in exception table */ 307 __u32 fib_discarded_routes; /* total number of routes delete */ 308 309 /* The following stats are not protected by any lock */ 310 atomic_t fib_rt_alloc; /* total number of routes alloced */ 311 atomic_t fib_rt_uncache; /* rt entries in uncached list */ 312 }; 313 314 #define RTN_TL_ROOT 0x0001 315 #define RTN_ROOT 0x0002 /* tree root node */ 316 #define RTN_RTINFO 0x0004 /* node with valid routing info */ 317 318 /* 319 * priority levels (or metrics) 320 * 321 */ 322 323 324 struct fib6_table { 325 struct hlist_node tb6_hlist; 326 u32 tb6_id; 327 spinlock_t tb6_lock; 328 struct fib6_node tb6_root; 329 struct inet_peer_base tb6_peers; 330 unsigned int flags; 331 unsigned int fib_seq; 332 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0) 333 }; 334 335 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC 336 #define RT6_TABLE_MAIN RT_TABLE_MAIN 337 #define RT6_TABLE_DFLT RT6_TABLE_MAIN 338 #define RT6_TABLE_INFO RT6_TABLE_MAIN 339 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN 340 341 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 342 #define FIB6_TABLE_MIN 1 343 #define FIB6_TABLE_MAX RT_TABLE_MAX 344 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL 345 #else 346 #define FIB6_TABLE_MIN RT_TABLE_MAIN 347 #define FIB6_TABLE_MAX FIB6_TABLE_MIN 348 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN 349 #endif 350 351 typedef struct rt6_info *(*pol_lookup_t)(struct net *, 352 struct fib6_table *, 353 struct flowi6 *, 354 const struct sk_buff *, int); 355 356 struct fib6_entry_notifier_info { 357 struct fib_notifier_info info; /* must be first */ 358 struct rt6_info *rt; 359 }; 360 361 /* 362 * exported functions 363 */ 364 365 struct fib6_table *fib6_get_table(struct net *net, u32 id); 366 struct fib6_table *fib6_new_table(struct net *net, u32 id); 367 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6, 368 const struct sk_buff *skb, 369 int flags, pol_lookup_t lookup); 370 371 struct fib6_node *fib6_lookup(struct fib6_node *root, 372 const struct in6_addr *daddr, 373 const struct in6_addr *saddr); 374 375 struct fib6_node *fib6_locate(struct fib6_node *root, 376 const struct in6_addr *daddr, int dst_len, 377 const struct in6_addr *saddr, int src_len, 378 bool exact_match); 379 380 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg), 381 void *arg); 382 383 int fib6_add(struct fib6_node *root, struct rt6_info *rt, 384 struct nl_info *info, struct mx6_config *mxc, 385 struct netlink_ext_ack *extack); 386 int fib6_del(struct rt6_info *rt, struct nl_info *info); 387 388 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info, 389 unsigned int flags); 390 391 void fib6_run_gc(unsigned long expires, struct net *net, bool force); 392 393 void fib6_gc_cleanup(void); 394 395 int fib6_init(void); 396 397 int ipv6_route_open(struct inode *inode, struct file *file); 398 399 int call_fib6_notifier(struct notifier_block *nb, struct net *net, 400 enum fib_event_type event_type, 401 struct fib_notifier_info *info); 402 int call_fib6_notifiers(struct net *net, enum fib_event_type event_type, 403 struct fib_notifier_info *info); 404 405 int __net_init fib6_notifier_init(struct net *net); 406 void __net_exit fib6_notifier_exit(struct net *net); 407 408 unsigned int fib6_tables_seq_read(struct net *net); 409 int fib6_tables_dump(struct net *net, struct notifier_block *nb); 410 411 void fib6_update_sernum(struct rt6_info *rt); 412 void fib6_update_sernum_upto_root(struct net *net, struct rt6_info *rt); 413 414 #ifdef CONFIG_IPV6_MULTIPLE_TABLES 415 int fib6_rules_init(void); 416 void fib6_rules_cleanup(void); 417 bool fib6_rule_default(const struct fib_rule *rule); 418 int fib6_rules_dump(struct net *net, struct notifier_block *nb); 419 unsigned int fib6_rules_seq_read(struct net *net); 420 421 static inline bool fib6_rules_early_flow_dissect(struct net *net, 422 struct sk_buff *skb, 423 struct flowi6 *fl6, 424 struct flow_keys *flkeys) 425 { 426 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP; 427 428 if (!net->ipv6.fib6_rules_require_fldissect) 429 return false; 430 431 skb_flow_dissect_flow_keys(skb, flkeys, flag); 432 fl6->fl6_sport = flkeys->ports.src; 433 fl6->fl6_dport = flkeys->ports.dst; 434 fl6->flowi6_proto = flkeys->basic.ip_proto; 435 436 return true; 437 } 438 #else 439 static inline int fib6_rules_init(void) 440 { 441 return 0; 442 } 443 static inline void fib6_rules_cleanup(void) 444 { 445 return ; 446 } 447 static inline bool fib6_rule_default(const struct fib_rule *rule) 448 { 449 return true; 450 } 451 static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb) 452 { 453 return 0; 454 } 455 static inline unsigned int fib6_rules_seq_read(struct net *net) 456 { 457 return 0; 458 } 459 static inline bool fib6_rules_early_flow_dissect(struct net *net, 460 struct sk_buff *skb, 461 struct flowi6 *fl6, 462 struct flow_keys *flkeys) 463 { 464 return false; 465 } 466 #endif 467 #endif 468