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,
55 struct in6_addr fc_dst;
56 struct in6_addr fc_src;
57 struct in6_addr fc_prefsrc;
58 struct in6_addr fc_gateway;
60 unsigned long fc_expires;
66 struct nl_info fc_nlinfo;
67 struct nlattr *fc_encap;
72 struct fib6_node __rcu *parent;
73 struct fib6_node __rcu *left;
74 struct fib6_node __rcu *right;
75 #ifdef CONFIG_IPV6_SUBTREES
76 struct fib6_node __rcu *subtree;
78 struct rt6_info __rcu *leaf;
80 __u16 fn_bit; /* bit key */
83 struct rt6_info __rcu *rr_ptr;
92 #ifndef CONFIG_IPV6_SUBTREES
93 #define FIB6_SUBTREE(fn) NULL
95 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1))
104 struct in6_addr addr;
110 struct rt6_exception_bucket {
111 struct hlist_head chain;
115 struct rt6_exception {
116 struct hlist_node hlist;
117 struct rt6_info *rt6i;
122 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10
123 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT)
124 #define FIB6_MAX_DEPTH 5
127 struct in6_addr nh_gw;
128 struct net_device *nh_dev;
129 struct lwtunnel_state *nh_lwtstate;
131 unsigned int nh_flags;
132 atomic_t nh_upper_bound;
137 struct fib6_table *rt6i_table;
138 struct fib6_info __rcu *rt6_next;
139 struct fib6_node __rcu *rt6i_node;
142 * siblings is a list of fib6_info that have the the same metric/weight,
143 * destination, but not the same gateway. nsiblings is just a cache
144 * to speed up lookup.
146 struct list_head rt6i_siblings;
147 unsigned int rt6i_nsiblings;
150 struct inet6_dev *rt6i_idev;
151 unsigned long expires;
152 struct dst_metrics *fib6_metrics;
153 #define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1]
155 struct rt6key rt6i_dst;
157 struct rt6key rt6i_src;
158 struct rt6key rt6i_prefsrc;
160 struct rt6_info * __percpu *rt6i_pcpu;
161 struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
166 u8 exception_bucket_flushed:1,
173 struct fib6_nh fib6_nh;
177 struct dst_entry dst;
178 struct rt6_info __rcu *rt6_next;
179 struct rt6_info *from;
182 * Tail elements of dst_entry (__refcnt etc.)
183 * and these elements (rarely used in hot path) are in
184 * the same cache line.
186 struct fib6_table *rt6i_table;
187 struct fib6_node __rcu *rt6i_node;
189 struct in6_addr rt6i_gateway;
192 * siblings is a list of rt6_info that have the the same metric/weight,
193 * destination, but not the same gateway. nsiblings is just a cache
194 * to speed up lookup.
196 struct list_head rt6i_siblings;
197 unsigned int rt6i_nsiblings;
201 /* These are in a separate cache line. */
202 struct rt6key rt6i_dst ____cacheline_aligned_in_smp;
204 struct rt6key rt6i_src;
205 struct rt6key rt6i_prefsrc;
207 struct list_head rt6i_uncached;
208 struct uncached_list *rt6i_uncached_list;
210 struct inet6_dev *rt6i_idev;
211 struct rt6_info * __percpu *rt6i_pcpu;
212 struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
215 /* more non-fragment space at head required */
216 unsigned short rt6i_nfheader_len;
219 u8 exception_bucket_flushed:1,
226 unsigned long expires;
227 struct dst_metrics *fib6_metrics;
228 #define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1]
229 struct fib6_nh fib6_nh;
232 #define for_each_fib6_node_rt_rcu(fn) \
233 for (rt = rcu_dereference((fn)->leaf); rt; \
234 rt = rcu_dereference(rt->rt6_next))
236 #define for_each_fib6_walker_rt(w) \
237 for (rt = (w)->leaf; rt; \
238 rt = rcu_dereference_protected(rt->rt6_next, 1))
240 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
242 return ((struct rt6_info *)dst)->rt6i_idev;
245 static inline void fib6_clean_expires(struct rt6_info *f6i)
247 f6i->rt6i_flags &= ~RTF_EXPIRES;
251 static inline void fib6_set_expires(struct rt6_info *f6i,
252 unsigned long expires)
254 f6i->expires = expires;
255 f6i->rt6i_flags |= RTF_EXPIRES;
258 static inline bool fib6_check_expired(const struct rt6_info *f6i)
260 if (f6i->rt6i_flags & RTF_EXPIRES)
261 return time_after(jiffies, f6i->expires);
265 static inline void rt6_clean_expires(struct rt6_info *rt)
267 rt->rt6i_flags &= ~RTF_EXPIRES;
271 static inline void rt6_set_expires(struct rt6_info *rt, unsigned long expires)
273 rt->dst.expires = expires;
274 rt->rt6i_flags |= RTF_EXPIRES;
277 static inline void rt6_update_expires(struct rt6_info *rt0, int timeout)
279 if (!(rt0->rt6i_flags & RTF_EXPIRES) && rt0->from)
280 rt0->dst.expires = rt0->from->expires;
282 dst_set_expires(&rt0->dst, timeout);
283 rt0->rt6i_flags |= RTF_EXPIRES;
286 /* Function to safely get fn->sernum for passed in rt
287 * and store result in passed in cookie.
288 * Return true if we can get cookie safely
289 * Return false if not
291 static inline bool rt6_get_cookie_safe(const struct rt6_info *rt,
294 struct fib6_node *fn;
298 fn = rcu_dereference(rt->rt6i_node);
301 *cookie = fn->fn_sernum;
302 /* pairs with smp_wmb() in fib6_update_sernum_upto_root() */
311 static inline u32 rt6_get_cookie(const struct rt6_info *rt)
315 if (rt->rt6i_flags & RTF_PCPU ||
316 (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->from))
317 rt6_get_cookie_safe(rt->from, &cookie);
322 static inline void ip6_rt_put(struct rt6_info *rt)
324 /* dst_release() accepts a NULL parameter.
325 * We rely on dst being first structure in struct rt6_info
327 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0);
328 dst_release(&rt->dst);
331 void rt6_free_pcpu(struct rt6_info *non_pcpu_rt);
333 struct rt6_info *fib6_info_alloc(gfp_t gfp_flags);
334 void fib6_info_destroy(struct rt6_info *f6i);
336 static inline void fib6_info_hold(struct rt6_info *f6i)
338 atomic_inc(&f6i->rt6i_ref);
341 static inline void fib6_info_release(struct rt6_info *f6i)
343 if (f6i && atomic_dec_and_test(&f6i->rt6i_ref))
344 fib6_info_destroy(f6i);
347 static inline void rt6_hold(struct rt6_info *rt)
349 atomic_inc(&rt->rt6i_ref);
352 static inline void rt6_release(struct rt6_info *rt)
354 if (atomic_dec_and_test(&rt->rt6i_ref)) {
356 dst_dev_put(&rt->dst);
357 dst_release(&rt->dst);
361 enum fib6_walk_state {
362 #ifdef CONFIG_IPV6_SUBTREES
373 struct fib6_node *root, *node;
374 struct rt6_info *leaf;
375 enum fib6_walk_state state;
378 int (*func)(struct fib6_walker *);
382 struct rt6_statistics {
383 __u32 fib_nodes; /* all fib6 nodes */
384 __u32 fib_route_nodes; /* intermediate nodes */
385 __u32 fib_rt_entries; /* rt entries in fib table */
386 __u32 fib_rt_cache; /* cached rt entries in exception table */
387 __u32 fib_discarded_routes; /* total number of routes delete */
389 /* The following stats are not protected by any lock */
390 atomic_t fib_rt_alloc; /* total number of routes alloced */
391 atomic_t fib_rt_uncache; /* rt entries in uncached list */
394 #define RTN_TL_ROOT 0x0001
395 #define RTN_ROOT 0x0002 /* tree root node */
396 #define RTN_RTINFO 0x0004 /* node with valid routing info */
399 * priority levels (or metrics)
405 struct hlist_node tb6_hlist;
408 struct fib6_node tb6_root;
409 struct inet_peer_base tb6_peers;
411 unsigned int fib_seq;
412 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0)
415 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC
416 #define RT6_TABLE_MAIN RT_TABLE_MAIN
417 #define RT6_TABLE_DFLT RT6_TABLE_MAIN
418 #define RT6_TABLE_INFO RT6_TABLE_MAIN
419 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN
421 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
422 #define FIB6_TABLE_MIN 1
423 #define FIB6_TABLE_MAX RT_TABLE_MAX
424 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL
426 #define FIB6_TABLE_MIN RT_TABLE_MAIN
427 #define FIB6_TABLE_MAX FIB6_TABLE_MIN
428 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN
431 typedef struct rt6_info *(*pol_lookup_t)(struct net *,
434 const struct sk_buff *, int);
436 struct fib6_entry_notifier_info {
437 struct fib_notifier_info info; /* must be first */
445 struct fib6_table *fib6_get_table(struct net *net, u32 id);
446 struct fib6_table *fib6_new_table(struct net *net, u32 id);
447 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
448 const struct sk_buff *skb,
449 int flags, pol_lookup_t lookup);
451 struct fib6_node *fib6_lookup(struct fib6_node *root,
452 const struct in6_addr *daddr,
453 const struct in6_addr *saddr);
455 struct fib6_node *fib6_locate(struct fib6_node *root,
456 const struct in6_addr *daddr, int dst_len,
457 const struct in6_addr *saddr, int src_len,
460 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
463 int fib6_add(struct fib6_node *root, struct rt6_info *rt,
464 struct nl_info *info, struct netlink_ext_ack *extack);
465 int fib6_del(struct rt6_info *rt, struct nl_info *info);
467 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
470 void fib6_run_gc(unsigned long expires, struct net *net, bool force);
472 void fib6_gc_cleanup(void);
476 int ipv6_route_open(struct inode *inode, struct file *file);
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 rt6_info *rt);
491 void fib6_update_sernum_upto_root(struct net *net, struct rt6_info *rt);
493 void fib6_metric_set(struct rt6_info *f6i, int metric, u32 val);
494 static inline bool fib6_metric_locked(struct rt6_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)