2 * Linux INET6 implementation
5 * Pedro Roque <roque@di.fc.ul.pt>
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.
16 #include <linux/ipv6_route.h>
17 #include <linux/rtnetlink.h>
18 #include <linux/spinlock.h>
19 #include <linux/notifier.h>
22 #include <net/netlink.h>
23 #include <net/inetpeer.h>
24 #include <net/fib_notifier.h>
26 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
27 #define FIB6_TABLE_HASHSZ 256
29 #define FIB6_TABLE_HASHSZ 1
35 #define RT6_TRACE(x...) pr_debug(x)
37 #define RT6_TRACE(x...) do { ; } while (0)
51 u16 fc_type; /* only 8 bits are used */
52 u16 fc_delete_all_nh : 1,
56 struct in6_addr fc_dst;
57 struct in6_addr fc_src;
58 struct in6_addr fc_prefsrc;
59 struct in6_addr fc_gateway;
61 unsigned long fc_expires;
67 struct nl_info fc_nlinfo;
68 struct nlattr *fc_encap;
73 struct fib6_node __rcu *parent;
74 struct fib6_node __rcu *left;
75 struct fib6_node __rcu *right;
76 #ifdef CONFIG_IPV6_SUBTREES
77 struct fib6_node __rcu *subtree;
79 struct fib6_info __rcu *leaf;
81 __u16 fn_bit; /* bit key */
84 struct fib6_info __rcu *rr_ptr;
93 #ifndef CONFIG_IPV6_SUBTREES
94 #define FIB6_SUBTREE(fn) NULL
96 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1))
100 * routing information
105 struct in6_addr addr;
111 struct rt6_exception_bucket {
112 struct hlist_head chain;
116 struct rt6_exception {
117 struct hlist_node hlist;
118 struct rt6_info *rt6i;
123 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10
124 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT)
125 #define FIB6_MAX_DEPTH 5
128 struct in6_addr nh_gw;
130 struct net_device *nh_dev;
131 struct lwtunnel_state *nh_lwtstate;
133 unsigned int nh_flags;
134 atomic_t nh_upper_bound;
139 struct fib6_table *fib6_table;
140 struct fib6_info __rcu *fib6_next;
141 struct fib6_node __rcu *fib6_node;
144 * siblings is a list of fib6_info that have the the same metric/weight,
145 * destination, but not the same gateway. nsiblings is just a cache
146 * to speed up lookup.
148 struct list_head fib6_siblings;
149 unsigned int fib6_nsiblings;
152 unsigned long expires;
153 struct dst_metrics *fib6_metrics;
154 #define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1]
156 struct rt6key fib6_dst;
158 struct rt6key fib6_src;
159 struct rt6key fib6_prefsrc;
161 struct rt6_info * __percpu *rt6i_pcpu;
162 struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
164 #ifdef CONFIG_IPV6_ROUTER_PREF
165 unsigned long last_probe;
171 u8 exception_bucket_flushed:1,
178 struct fib6_nh fib6_nh;
183 struct dst_entry dst;
184 struct fib6_info __rcu *from;
186 struct rt6key rt6i_dst;
187 struct rt6key rt6i_src;
188 struct in6_addr rt6i_gateway;
189 struct inet6_dev *rt6i_idev;
192 struct list_head rt6i_uncached;
193 struct uncached_list *rt6i_uncached_list;
195 /* more non-fragment space at head required */
196 unsigned short rt6i_nfheader_len;
199 #define for_each_fib6_node_rt_rcu(fn) \
200 for (rt = rcu_dereference((fn)->leaf); rt; \
201 rt = rcu_dereference(rt->fib6_next))
203 #define for_each_fib6_walker_rt(w) \
204 for (rt = (w)->leaf; rt; \
205 rt = rcu_dereference_protected(rt->fib6_next, 1))
207 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
209 return ((struct rt6_info *)dst)->rt6i_idev;
212 static inline void fib6_clean_expires(struct fib6_info *f6i)
214 f6i->fib6_flags &= ~RTF_EXPIRES;
218 static inline void fib6_set_expires(struct fib6_info *f6i,
219 unsigned long expires)
221 f6i->expires = expires;
222 f6i->fib6_flags |= RTF_EXPIRES;
225 static inline bool fib6_check_expired(const struct fib6_info *f6i)
227 if (f6i->fib6_flags & RTF_EXPIRES)
228 return time_after(jiffies, f6i->expires);
232 /* Function to safely get fn->sernum for passed in rt
233 * and store result in passed in cookie.
234 * Return true if we can get cookie safely
235 * Return false if not
237 static inline bool fib6_get_cookie_safe(const struct fib6_info *f6i,
240 struct fib6_node *fn;
243 fn = rcu_dereference(f6i->fib6_node);
246 *cookie = fn->fn_sernum;
247 /* pairs with smp_wmb() in fib6_update_sernum_upto_root() */
255 static inline u32 rt6_get_cookie(const struct rt6_info *rt)
257 struct fib6_info *from;
262 from = rcu_dereference(rt->from);
263 if (from && (rt->rt6i_flags & RTF_PCPU ||
264 unlikely(!list_empty(&rt->rt6i_uncached))))
265 fib6_get_cookie_safe(from, &cookie);
272 static inline void ip6_rt_put(struct rt6_info *rt)
274 /* dst_release() accepts a NULL parameter.
275 * We rely on dst being first structure in struct rt6_info
277 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0);
278 dst_release(&rt->dst);
281 struct fib6_info *fib6_info_alloc(gfp_t gfp_flags);
282 void fib6_info_destroy_rcu(struct rcu_head *head);
284 static inline void fib6_info_hold(struct fib6_info *f6i)
286 atomic_inc(&f6i->fib6_ref);
289 static inline bool fib6_info_hold_safe(struct fib6_info *f6i)
291 return atomic_inc_not_zero(&f6i->fib6_ref);
294 static inline void fib6_info_release(struct fib6_info *f6i)
296 if (f6i && atomic_dec_and_test(&f6i->fib6_ref))
297 call_rcu(&f6i->rcu, fib6_info_destroy_rcu);
300 enum fib6_walk_state {
301 #ifdef CONFIG_IPV6_SUBTREES
312 struct fib6_node *root, *node;
313 struct fib6_info *leaf;
314 enum fib6_walk_state state;
317 int (*func)(struct fib6_walker *);
321 struct rt6_statistics {
322 __u32 fib_nodes; /* all fib6 nodes */
323 __u32 fib_route_nodes; /* intermediate nodes */
324 __u32 fib_rt_entries; /* rt entries in fib table */
325 __u32 fib_rt_cache; /* cached rt entries in exception table */
326 __u32 fib_discarded_routes; /* total number of routes delete */
328 /* The following stats are not protected by any lock */
329 atomic_t fib_rt_alloc; /* total number of routes alloced */
330 atomic_t fib_rt_uncache; /* rt entries in uncached list */
333 #define RTN_TL_ROOT 0x0001
334 #define RTN_ROOT 0x0002 /* tree root node */
335 #define RTN_RTINFO 0x0004 /* node with valid routing info */
338 * priority levels (or metrics)
344 struct hlist_node tb6_hlist;
347 struct fib6_node tb6_root;
348 struct inet_peer_base tb6_peers;
350 unsigned int fib_seq;
351 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0)
354 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC
355 #define RT6_TABLE_MAIN RT_TABLE_MAIN
356 #define RT6_TABLE_DFLT RT6_TABLE_MAIN
357 #define RT6_TABLE_INFO RT6_TABLE_MAIN
358 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN
360 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
361 #define FIB6_TABLE_MIN 1
362 #define FIB6_TABLE_MAX RT_TABLE_MAX
363 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL
365 #define FIB6_TABLE_MIN RT_TABLE_MAIN
366 #define FIB6_TABLE_MAX FIB6_TABLE_MIN
367 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN
370 typedef struct rt6_info *(*pol_lookup_t)(struct net *,
373 const struct sk_buff *, int);
375 struct fib6_entry_notifier_info {
376 struct fib_notifier_info info; /* must be first */
377 struct fib6_info *rt;
384 struct fib6_table *fib6_get_table(struct net *net, u32 id);
385 struct fib6_table *fib6_new_table(struct net *net, u32 id);
386 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
387 const struct sk_buff *skb,
388 int flags, pol_lookup_t lookup);
390 /* called with rcu lock held; can return error pointer
391 * caller needs to select path
393 struct fib6_info *fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
396 /* called with rcu lock held; caller needs to select path */
397 struct fib6_info *fib6_table_lookup(struct net *net, struct fib6_table *table,
398 int oif, struct flowi6 *fl6, int strict);
400 struct fib6_info *fib6_multipath_select(const struct net *net,
401 struct fib6_info *match,
402 struct flowi6 *fl6, int oif,
403 const struct sk_buff *skb, int strict);
405 struct fib6_node *fib6_node_lookup(struct fib6_node *root,
406 const struct in6_addr *daddr,
407 const struct in6_addr *saddr);
409 struct fib6_node *fib6_locate(struct fib6_node *root,
410 const struct in6_addr *daddr, int dst_len,
411 const struct in6_addr *saddr, int src_len,
414 void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *arg),
416 void fib6_clean_all_skip_notify(struct net *net,
417 int (*func)(struct fib6_info *, void *arg),
420 int fib6_add(struct fib6_node *root, struct fib6_info *rt,
421 struct nl_info *info, struct netlink_ext_ack *extack);
422 int fib6_del(struct fib6_info *rt, struct nl_info *info);
425 void rt6_get_prefsrc(const struct rt6_info *rt, struct in6_addr *addr)
427 const struct fib6_info *from;
431 from = rcu_dereference(rt->from);
433 *addr = from->fib6_prefsrc.addr;
435 struct in6_addr in6_zero = {};
443 static inline struct net_device *fib6_info_nh_dev(const struct fib6_info *f6i)
445 return f6i->fib6_nh.nh_dev;
448 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
449 struct fib6_config *cfg, gfp_t gfp_flags,
450 struct netlink_ext_ack *extack);
451 void fib6_nh_release(struct fib6_nh *fib6_nh);
454 struct lwtunnel_state *fib6_info_nh_lwt(const struct fib6_info *f6i)
456 return f6i->fib6_nh.nh_lwtstate;
459 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
462 void fib6_run_gc(unsigned long expires, struct net *net, bool force);
464 void fib6_gc_cleanup(void);
468 struct ipv6_route_iter {
469 struct seq_net_private p;
470 struct fib6_walker w;
472 struct fib6_table *tbl;
476 extern const struct seq_operations ipv6_route_seq_ops;
478 int call_fib6_notifier(struct notifier_block *nb, struct net *net,
479 enum fib_event_type event_type,
480 struct fib_notifier_info *info);
481 int call_fib6_notifiers(struct net *net, enum fib_event_type event_type,
482 struct fib_notifier_info *info);
484 int __net_init fib6_notifier_init(struct net *net);
485 void __net_exit fib6_notifier_exit(struct net *net);
487 unsigned int fib6_tables_seq_read(struct net *net);
488 int fib6_tables_dump(struct net *net, struct notifier_block *nb);
490 void fib6_update_sernum(struct net *net, struct fib6_info *rt);
491 void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt);
493 void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val);
494 static inline bool fib6_metric_locked(struct fib6_info *f6i, int metric)
496 return !!(f6i->fib6_metrics->metrics[RTAX_LOCK - 1] & (1 << metric));
499 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
500 int fib6_rules_init(void);
501 void fib6_rules_cleanup(void);
502 bool fib6_rule_default(const struct fib_rule *rule);
503 int fib6_rules_dump(struct net *net, struct notifier_block *nb);
504 unsigned int fib6_rules_seq_read(struct net *net);
506 static inline bool fib6_rules_early_flow_dissect(struct net *net,
509 struct flow_keys *flkeys)
511 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
513 if (!net->ipv6.fib6_rules_require_fldissect)
516 skb_flow_dissect_flow_keys(skb, flkeys, flag);
517 fl6->fl6_sport = flkeys->ports.src;
518 fl6->fl6_dport = flkeys->ports.dst;
519 fl6->flowi6_proto = flkeys->basic.ip_proto;
524 static inline int fib6_rules_init(void)
528 static inline void fib6_rules_cleanup(void)
532 static inline bool fib6_rule_default(const struct fib_rule *rule)
536 static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb)
540 static inline unsigned int fib6_rules_seq_read(struct net *net)
544 static inline bool fib6_rules_early_flow_dissect(struct net *net,
547 struct flow_keys *flkeys)