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)
50 u16 fc_type; /* only 8 bits are used */
51 u16 fc_delete_all_nh : 1,
54 struct in6_addr fc_dst;
55 struct in6_addr fc_src;
56 struct in6_addr fc_prefsrc;
57 struct in6_addr fc_gateway;
59 unsigned long fc_expires;
65 struct nl_info fc_nlinfo;
66 struct nlattr *fc_encap;
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;
77 struct rt6_info __rcu *leaf;
79 __u16 fn_bit; /* bit key */
82 struct rt6_info __rcu *rr_ptr;
91 #ifndef CONFIG_IPV6_SUBTREES
92 #define FIB6_SUBTREE(fn) NULL
94 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1))
99 DECLARE_BITMAP(mx_valid, RTAX_MAX);
103 * routing information
108 struct in6_addr addr;
114 struct rt6_exception_bucket {
115 struct hlist_head chain;
119 struct rt6_exception {
120 struct hlist_node hlist;
121 struct rt6_info *rt6i;
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
131 struct dst_entry dst;
132 struct rt6_info __rcu *rt6_next;
133 struct rt6_info *from;
136 * Tail elements of dst_entry (__refcnt etc.)
137 * and these elements (rarely used in hot path) are in
138 * the same cache line.
140 struct fib6_table *rt6i_table;
141 struct fib6_node __rcu *rt6i_node;
143 struct in6_addr rt6i_gateway;
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.
150 struct list_head rt6i_siblings;
151 unsigned int rt6i_nsiblings;
152 atomic_t rt6i_nh_upper_bound;
156 unsigned int rt6i_nh_flags;
158 /* These are in a separate cache line. */
159 struct rt6key rt6i_dst ____cacheline_aligned_in_smp;
161 struct rt6key rt6i_src;
162 struct rt6key rt6i_prefsrc;
164 struct list_head rt6i_uncached;
165 struct uncached_list *rt6i_uncached_list;
167 struct inet6_dev *rt6i_idev;
168 struct rt6_info * __percpu *rt6i_pcpu;
169 struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
173 /* more non-fragment space at head required */
175 unsigned short rt6i_nfheader_len;
178 u8 exception_bucket_flushed:1,
183 #define for_each_fib6_node_rt_rcu(fn) \
184 for (rt = rcu_dereference((fn)->leaf); rt; \
185 rt = rcu_dereference(rt->rt6_next))
187 #define for_each_fib6_walker_rt(w) \
188 for (rt = (w)->leaf; rt; \
189 rt = rcu_dereference_protected(rt->rt6_next, 1))
191 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
193 return ((struct rt6_info *)dst)->rt6i_idev;
196 static inline void rt6_clean_expires(struct rt6_info *rt)
198 rt->rt6i_flags &= ~RTF_EXPIRES;
202 static inline void rt6_set_expires(struct rt6_info *rt, unsigned long expires)
204 rt->dst.expires = expires;
205 rt->rt6i_flags |= RTF_EXPIRES;
208 static inline void rt6_update_expires(struct rt6_info *rt0, int timeout)
212 for (rt = rt0; rt && !(rt->rt6i_flags & RTF_EXPIRES); rt = rt->from);
214 rt0->dst.expires = rt->dst.expires;
215 dst_set_expires(&rt0->dst, timeout);
216 rt0->rt6i_flags |= RTF_EXPIRES;
219 /* Function to safely get fn->sernum for passed in rt
220 * and store result in passed in cookie.
221 * Return true if we can get cookie safely
222 * Return false if not
224 static inline bool rt6_get_cookie_safe(const struct rt6_info *rt,
227 struct fib6_node *fn;
231 fn = rcu_dereference(rt->rt6i_node);
234 *cookie = fn->fn_sernum;
235 /* pairs with smp_wmb() in fib6_update_sernum_upto_root() */
244 static inline u32 rt6_get_cookie(const struct rt6_info *rt)
248 if (rt->rt6i_flags & RTF_PCPU ||
249 (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->from))
252 rt6_get_cookie_safe(rt, &cookie);
257 static inline void ip6_rt_put(struct rt6_info *rt)
259 /* dst_release() accepts a NULL parameter.
260 * We rely on dst being first structure in struct rt6_info
262 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0);
263 dst_release(&rt->dst);
266 void rt6_free_pcpu(struct rt6_info *non_pcpu_rt);
268 static inline void rt6_hold(struct rt6_info *rt)
270 atomic_inc(&rt->rt6i_ref);
273 static inline void rt6_release(struct rt6_info *rt)
275 if (atomic_dec_and_test(&rt->rt6i_ref)) {
277 dst_dev_put(&rt->dst);
278 dst_release(&rt->dst);
282 enum fib6_walk_state {
283 #ifdef CONFIG_IPV6_SUBTREES
294 struct fib6_node *root, *node;
295 struct rt6_info *leaf;
296 enum fib6_walk_state state;
299 int (*func)(struct fib6_walker *);
303 struct rt6_statistics {
304 __u32 fib_nodes; /* all fib6 nodes */
305 __u32 fib_route_nodes; /* intermediate nodes */
306 __u32 fib_rt_entries; /* rt entries in fib table */
307 __u32 fib_rt_cache; /* cached rt entries in exception table */
308 __u32 fib_discarded_routes; /* total number of routes delete */
310 /* The following stats are not protected by any lock */
311 atomic_t fib_rt_alloc; /* total number of routes alloced */
312 atomic_t fib_rt_uncache; /* rt entries in uncached list */
315 #define RTN_TL_ROOT 0x0001
316 #define RTN_ROOT 0x0002 /* tree root node */
317 #define RTN_RTINFO 0x0004 /* node with valid routing info */
320 * priority levels (or metrics)
326 struct hlist_node tb6_hlist;
329 struct fib6_node tb6_root;
330 struct inet_peer_base tb6_peers;
332 unsigned int fib_seq;
333 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0)
336 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC
337 #define RT6_TABLE_MAIN RT_TABLE_MAIN
338 #define RT6_TABLE_DFLT RT6_TABLE_MAIN
339 #define RT6_TABLE_INFO RT6_TABLE_MAIN
340 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN
342 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
343 #define FIB6_TABLE_MIN 1
344 #define FIB6_TABLE_MAX RT_TABLE_MAX
345 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL
347 #define FIB6_TABLE_MIN RT_TABLE_MAIN
348 #define FIB6_TABLE_MAX FIB6_TABLE_MIN
349 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN
352 typedef struct rt6_info *(*pol_lookup_t)(struct net *,
355 const struct sk_buff *, int);
357 struct fib6_entry_notifier_info {
358 struct fib_notifier_info info; /* must be first */
366 struct fib6_table *fib6_get_table(struct net *net, u32 id);
367 struct fib6_table *fib6_new_table(struct net *net, u32 id);
368 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
369 const struct sk_buff *skb,
370 int flags, pol_lookup_t lookup);
372 struct fib6_node *fib6_lookup(struct fib6_node *root,
373 const struct in6_addr *daddr,
374 const struct in6_addr *saddr);
376 struct fib6_node *fib6_locate(struct fib6_node *root,
377 const struct in6_addr *daddr, int dst_len,
378 const struct in6_addr *saddr, int src_len,
381 void fib6_clean_all(struct net *net, int (*func)(struct rt6_info *, void *arg),
384 int fib6_add(struct fib6_node *root, struct rt6_info *rt,
385 struct nl_info *info, struct mx6_config *mxc,
386 struct netlink_ext_ack *extack);
387 int fib6_del(struct rt6_info *rt, struct nl_info *info);
389 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
392 void fib6_run_gc(unsigned long expires, struct net *net, bool force);
394 void fib6_gc_cleanup(void);
398 int ipv6_route_open(struct inode *inode, struct file *file);
400 int call_fib6_notifier(struct notifier_block *nb, struct net *net,
401 enum fib_event_type event_type,
402 struct fib_notifier_info *info);
403 int call_fib6_notifiers(struct net *net, enum fib_event_type event_type,
404 struct fib_notifier_info *info);
406 int __net_init fib6_notifier_init(struct net *net);
407 void __net_exit fib6_notifier_exit(struct net *net);
409 unsigned int fib6_tables_seq_read(struct net *net);
410 int fib6_tables_dump(struct net *net, struct notifier_block *nb);
412 void fib6_update_sernum(struct net *net, struct rt6_info *rt);
413 void fib6_update_sernum_upto_root(struct net *net, struct rt6_info *rt);
415 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
416 int fib6_rules_init(void);
417 void fib6_rules_cleanup(void);
418 bool fib6_rule_default(const struct fib_rule *rule);
419 int fib6_rules_dump(struct net *net, struct notifier_block *nb);
420 unsigned int fib6_rules_seq_read(struct net *net);
422 static inline bool fib6_rules_early_flow_dissect(struct net *net,
425 struct flow_keys *flkeys)
427 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
429 if (!net->ipv6.fib6_rules_require_fldissect)
432 skb_flow_dissect_flow_keys(skb, flkeys, flag);
433 fl6->fl6_sport = flkeys->ports.src;
434 fl6->fl6_dport = flkeys->ports.dst;
435 fl6->flowi6_proto = flkeys->basic.ip_proto;
440 static inline int fib6_rules_init(void)
444 static inline void fib6_rules_cleanup(void)
448 static inline bool fib6_rule_default(const struct fib_rule *rule)
452 static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb)
456 static inline unsigned int fib6_rules_seq_read(struct net *net)
460 static inline bool fib6_rules_early_flow_dissect(struct net *net,
463 struct flow_keys *flkeys)