xref: /linux/include/net/ip6_fib.h (revision f6d08d9d8543c8ee494b307804b28e2750ffedb9)
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 
23 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
24 #define FIB6_TABLE_HASHSZ 256
25 #else
26 #define FIB6_TABLE_HASHSZ 1
27 #endif
28 
29 #define RT6_DEBUG 2
30 
31 #if RT6_DEBUG >= 3
32 #define RT6_TRACE(x...) pr_debug(x)
33 #else
34 #define RT6_TRACE(x...) do { ; } while (0)
35 #endif
36 
37 struct rt6_info;
38 struct fib6_info;
39 
40 struct fib6_config {
41 	u32		fc_table;
42 	u32		fc_metric;
43 	int		fc_dst_len;
44 	int		fc_src_len;
45 	int		fc_ifindex;
46 	u32		fc_flags;
47 	u32		fc_protocol;
48 	u16		fc_type;        /* only 8 bits are used */
49 	u16		fc_delete_all_nh : 1,
50 			fc_ignore_dev_down:1,
51 			__unused : 14;
52 	u32		fc_nh_id;
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 fib6_info __rcu	*leaf;
78 
79 	__u16			fn_bit;		/* bit key */
80 	__u16			fn_flags;
81 	int			fn_sernum;
82 	struct fib6_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 /*
98  *	routing information
99  *
100  */
101 
102 struct rt6key {
103 	struct in6_addr	addr;
104 	int		plen;
105 };
106 
107 struct fib6_table;
108 
109 struct rt6_exception_bucket {
110 	struct hlist_head	chain;
111 	int			depth;
112 };
113 
114 struct rt6_exception {
115 	struct hlist_node	hlist;
116 	struct rt6_info		*rt6i;
117 	unsigned long		stamp;
118 	struct rcu_head		rcu;
119 };
120 
121 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10
122 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT)
123 #define FIB6_MAX_DEPTH 5
124 
125 struct fib6_nh {
126 	struct fib_nh_common	nh_common;
127 
128 #ifdef CONFIG_IPV6_ROUTER_PREF
129 	unsigned long		last_probe;
130 #endif
131 
132 	struct rt6_info * __percpu *rt6i_pcpu;
133 	struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
134 };
135 
136 struct fib6_info {
137 	struct fib6_table		*fib6_table;
138 	struct fib6_info __rcu		*fib6_next;
139 	struct fib6_node __rcu		*fib6_node;
140 
141 	/* Multipath routes:
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.
145 	 */
146 	union {
147 		struct list_head	fib6_siblings;
148 		struct list_head	nh_list;
149 	};
150 	unsigned int			fib6_nsiblings;
151 
152 	refcount_t			fib6_ref;
153 	unsigned long			expires;
154 	struct dst_metrics		*fib6_metrics;
155 #define fib6_pmtu		fib6_metrics->metrics[RTAX_MTU-1]
156 
157 	struct rt6key			fib6_dst;
158 	u32				fib6_flags;
159 	struct rt6key			fib6_src;
160 	struct rt6key			fib6_prefsrc;
161 
162 	u32				fib6_metric;
163 	u8				fib6_protocol;
164 	u8				fib6_type;
165 	u8				should_flush:1,
166 					dst_nocount:1,
167 					dst_nopolicy:1,
168 					dst_host:1,
169 					fib6_destroying:1,
170 					unused:3;
171 
172 	struct rcu_head			rcu;
173 	struct nexthop			*nh;
174 	struct fib6_nh			fib6_nh[0];
175 };
176 
177 struct rt6_info {
178 	struct dst_entry		dst;
179 	struct fib6_info __rcu		*from;
180 
181 	struct rt6key			rt6i_dst;
182 	struct rt6key			rt6i_src;
183 	struct in6_addr			rt6i_gateway;
184 	struct inet6_dev		*rt6i_idev;
185 	u32				rt6i_flags;
186 
187 	struct list_head		rt6i_uncached;
188 	struct uncached_list		*rt6i_uncached_list;
189 
190 	/* more non-fragment space at head required */
191 	unsigned short			rt6i_nfheader_len;
192 };
193 
194 struct fib6_result {
195 	struct fib6_nh		*nh;
196 	struct fib6_info	*f6i;
197 	u32			fib6_flags;
198 	u8			fib6_type;
199 	struct rt6_info		*rt6;
200 };
201 
202 #define for_each_fib6_node_rt_rcu(fn)					\
203 	for (rt = rcu_dereference((fn)->leaf); rt;			\
204 	     rt = rcu_dereference(rt->fib6_next))
205 
206 #define for_each_fib6_walker_rt(w)					\
207 	for (rt = (w)->leaf; rt;					\
208 	     rt = rcu_dereference_protected(rt->fib6_next, 1))
209 
210 static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
211 {
212 	return ((struct rt6_info *)dst)->rt6i_idev;
213 }
214 
215 static inline void fib6_clean_expires(struct fib6_info *f6i)
216 {
217 	f6i->fib6_flags &= ~RTF_EXPIRES;
218 	f6i->expires = 0;
219 }
220 
221 static inline void fib6_set_expires(struct fib6_info *f6i,
222 				    unsigned long expires)
223 {
224 	f6i->expires = expires;
225 	f6i->fib6_flags |= RTF_EXPIRES;
226 }
227 
228 static inline bool fib6_check_expired(const struct fib6_info *f6i)
229 {
230 	if (f6i->fib6_flags & RTF_EXPIRES)
231 		return time_after(jiffies, f6i->expires);
232 	return false;
233 }
234 
235 /* Function to safely get fn->sernum for passed in rt
236  * and store result in passed in cookie.
237  * Return true if we can get cookie safely
238  * Return false if not
239  */
240 static inline bool fib6_get_cookie_safe(const struct fib6_info *f6i,
241 					u32 *cookie)
242 {
243 	struct fib6_node *fn;
244 	bool status = false;
245 
246 	fn = rcu_dereference(f6i->fib6_node);
247 
248 	if (fn) {
249 		*cookie = fn->fn_sernum;
250 		/* pairs with smp_wmb() in fib6_update_sernum_upto_root() */
251 		smp_rmb();
252 		status = true;
253 	}
254 
255 	return status;
256 }
257 
258 static inline u32 rt6_get_cookie(const struct rt6_info *rt)
259 {
260 	struct fib6_info *from;
261 	u32 cookie = 0;
262 
263 	rcu_read_lock();
264 
265 	from = rcu_dereference(rt->from);
266 	if (from)
267 		fib6_get_cookie_safe(from, &cookie);
268 
269 	rcu_read_unlock();
270 
271 	return cookie;
272 }
273 
274 static inline void ip6_rt_put(struct rt6_info *rt)
275 {
276 	/* dst_release() accepts a NULL parameter.
277 	 * We rely on dst being first structure in struct rt6_info
278 	 */
279 	BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0);
280 	dst_release(&rt->dst);
281 }
282 
283 struct fib6_info *fib6_info_alloc(gfp_t gfp_flags, bool with_fib6_nh);
284 void fib6_info_destroy_rcu(struct rcu_head *head);
285 
286 static inline void fib6_info_hold(struct fib6_info *f6i)
287 {
288 	refcount_inc(&f6i->fib6_ref);
289 }
290 
291 static inline bool fib6_info_hold_safe(struct fib6_info *f6i)
292 {
293 	return refcount_inc_not_zero(&f6i->fib6_ref);
294 }
295 
296 static inline void fib6_info_release(struct fib6_info *f6i)
297 {
298 	if (f6i && refcount_dec_and_test(&f6i->fib6_ref))
299 		call_rcu(&f6i->rcu, fib6_info_destroy_rcu);
300 }
301 
302 enum fib6_walk_state {
303 #ifdef CONFIG_IPV6_SUBTREES
304 	FWS_S,
305 #endif
306 	FWS_L,
307 	FWS_R,
308 	FWS_C,
309 	FWS_U
310 };
311 
312 struct fib6_walker {
313 	struct list_head lh;
314 	struct fib6_node *root, *node;
315 	struct fib6_info *leaf;
316 	enum fib6_walk_state state;
317 	unsigned int skip;
318 	unsigned int count;
319 	int (*func)(struct fib6_walker *);
320 	void *args;
321 };
322 
323 struct rt6_statistics {
324 	__u32		fib_nodes;		/* all fib6 nodes */
325 	__u32		fib_route_nodes;	/* intermediate nodes */
326 	__u32		fib_rt_entries;		/* rt entries in fib table */
327 	__u32		fib_rt_cache;		/* cached rt entries in exception table */
328 	__u32		fib_discarded_routes;	/* total number of routes delete */
329 
330 	/* The following stats are not protected by any lock */
331 	atomic_t	fib_rt_alloc;		/* total number of routes alloced */
332 	atomic_t	fib_rt_uncache;		/* rt entries in uncached list */
333 };
334 
335 #define RTN_TL_ROOT	0x0001
336 #define RTN_ROOT	0x0002		/* tree root node		*/
337 #define RTN_RTINFO	0x0004		/* node with valid routing info	*/
338 
339 /*
340  *	priority levels (or metrics)
341  *
342  */
343 
344 
345 struct fib6_table {
346 	struct hlist_node	tb6_hlist;
347 	u32			tb6_id;
348 	spinlock_t		tb6_lock;
349 	struct fib6_node	tb6_root;
350 	struct inet_peer_base	tb6_peers;
351 	unsigned int		flags;
352 	unsigned int		fib_seq;
353 #define RT6_TABLE_HAS_DFLT_ROUTER	BIT(0)
354 };
355 
356 #define RT6_TABLE_UNSPEC	RT_TABLE_UNSPEC
357 #define RT6_TABLE_MAIN		RT_TABLE_MAIN
358 #define RT6_TABLE_DFLT		RT6_TABLE_MAIN
359 #define RT6_TABLE_INFO		RT6_TABLE_MAIN
360 #define RT6_TABLE_PREFIX	RT6_TABLE_MAIN
361 
362 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
363 #define FIB6_TABLE_MIN		1
364 #define FIB6_TABLE_MAX		RT_TABLE_MAX
365 #define RT6_TABLE_LOCAL		RT_TABLE_LOCAL
366 #else
367 #define FIB6_TABLE_MIN		RT_TABLE_MAIN
368 #define FIB6_TABLE_MAX		FIB6_TABLE_MIN
369 #define RT6_TABLE_LOCAL		RT6_TABLE_MAIN
370 #endif
371 
372 typedef struct rt6_info *(*pol_lookup_t)(struct net *,
373 					 struct fib6_table *,
374 					 struct flowi6 *,
375 					 const struct sk_buff *, int);
376 
377 struct fib6_entry_notifier_info {
378 	struct fib_notifier_info info; /* must be first */
379 	struct fib6_info *rt;
380 	unsigned int nsiblings;
381 };
382 
383 /*
384  *	exported functions
385  */
386 
387 struct fib6_table *fib6_get_table(struct net *net, u32 id);
388 struct fib6_table *fib6_new_table(struct net *net, u32 id);
389 struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
390 				   const struct sk_buff *skb,
391 				   int flags, pol_lookup_t lookup);
392 
393 /* called with rcu lock held; can return error pointer
394  * caller needs to select path
395  */
396 int fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
397 		struct fib6_result *res, int flags);
398 
399 /* called with rcu lock held; caller needs to select path */
400 int fib6_table_lookup(struct net *net, struct fib6_table *table,
401 		      int oif, struct flowi6 *fl6, struct fib6_result *res,
402 		      int strict);
403 
404 void fib6_select_path(const struct net *net, struct fib6_result *res,
405 		      struct flowi6 *fl6, int oif, bool have_oif_match,
406 		      const struct sk_buff *skb, int strict);
407 struct fib6_node *fib6_node_lookup(struct fib6_node *root,
408 				   const struct in6_addr *daddr,
409 				   const struct in6_addr *saddr);
410 
411 struct fib6_node *fib6_locate(struct fib6_node *root,
412 			      const struct in6_addr *daddr, int dst_len,
413 			      const struct in6_addr *saddr, int src_len,
414 			      bool exact_match);
415 
416 void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *arg),
417 		    void *arg);
418 void fib6_clean_all_skip_notify(struct net *net,
419 				int (*func)(struct fib6_info *, void *arg),
420 				void *arg);
421 
422 int fib6_add(struct fib6_node *root, struct fib6_info *rt,
423 	     struct nl_info *info, struct netlink_ext_ack *extack);
424 int fib6_del(struct fib6_info *rt, struct nl_info *info);
425 
426 static inline
427 void rt6_get_prefsrc(const struct rt6_info *rt, struct in6_addr *addr)
428 {
429 	const struct fib6_info *from;
430 
431 	rcu_read_lock();
432 
433 	from = rcu_dereference(rt->from);
434 	if (from) {
435 		*addr = from->fib6_prefsrc.addr;
436 	} else {
437 		struct in6_addr in6_zero = {};
438 
439 		*addr = in6_zero;
440 	}
441 
442 	rcu_read_unlock();
443 }
444 
445 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
446 		 struct fib6_config *cfg, gfp_t gfp_flags,
447 		 struct netlink_ext_ack *extack);
448 void fib6_nh_release(struct fib6_nh *fib6_nh);
449 
450 int call_fib6_entry_notifiers(struct net *net,
451 			      enum fib_event_type event_type,
452 			      struct fib6_info *rt,
453 			      struct netlink_ext_ack *extack);
454 int call_fib6_multipath_entry_notifiers(struct net *net,
455 					enum fib_event_type event_type,
456 					struct fib6_info *rt,
457 					unsigned int nsiblings,
458 					struct netlink_ext_ack *extack);
459 void fib6_rt_update(struct net *net, struct fib6_info *rt,
460 		    struct nl_info *info);
461 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
462 		     unsigned int flags);
463 
464 void fib6_run_gc(unsigned long expires, struct net *net, bool force);
465 
466 void fib6_gc_cleanup(void);
467 
468 int fib6_init(void);
469 
470 struct ipv6_route_iter {
471 	struct seq_net_private p;
472 	struct fib6_walker w;
473 	loff_t skip;
474 	struct fib6_table *tbl;
475 	int sernum;
476 };
477 
478 extern const struct seq_operations ipv6_route_seq_ops;
479 
480 int call_fib6_notifier(struct notifier_block *nb, struct net *net,
481 		       enum fib_event_type event_type,
482 		       struct fib_notifier_info *info);
483 int call_fib6_notifiers(struct net *net, enum fib_event_type event_type,
484 			struct fib_notifier_info *info);
485 
486 int __net_init fib6_notifier_init(struct net *net);
487 void __net_exit fib6_notifier_exit(struct net *net);
488 
489 unsigned int fib6_tables_seq_read(struct net *net);
490 int fib6_tables_dump(struct net *net, struct notifier_block *nb);
491 
492 void fib6_update_sernum(struct net *net, struct fib6_info *rt);
493 void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt);
494 void fib6_update_sernum_stub(struct net *net, struct fib6_info *f6i);
495 
496 void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val);
497 static inline bool fib6_metric_locked(struct fib6_info *f6i, int metric)
498 {
499 	return !!(f6i->fib6_metrics->metrics[RTAX_LOCK - 1] & (1 << metric));
500 }
501 
502 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
503 int fib6_rules_init(void);
504 void fib6_rules_cleanup(void);
505 bool fib6_rule_default(const struct fib_rule *rule);
506 int fib6_rules_dump(struct net *net, struct notifier_block *nb);
507 unsigned int fib6_rules_seq_read(struct net *net);
508 
509 static inline bool fib6_rules_early_flow_dissect(struct net *net,
510 						 struct sk_buff *skb,
511 						 struct flowi6 *fl6,
512 						 struct flow_keys *flkeys)
513 {
514 	unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
515 
516 	if (!net->ipv6.fib6_rules_require_fldissect)
517 		return false;
518 
519 	skb_flow_dissect_flow_keys(skb, flkeys, flag);
520 	fl6->fl6_sport = flkeys->ports.src;
521 	fl6->fl6_dport = flkeys->ports.dst;
522 	fl6->flowi6_proto = flkeys->basic.ip_proto;
523 
524 	return true;
525 }
526 #else
527 static inline int               fib6_rules_init(void)
528 {
529 	return 0;
530 }
531 static inline void              fib6_rules_cleanup(void)
532 {
533 	return ;
534 }
535 static inline bool fib6_rule_default(const struct fib_rule *rule)
536 {
537 	return true;
538 }
539 static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb)
540 {
541 	return 0;
542 }
543 static inline unsigned int fib6_rules_seq_read(struct net *net)
544 {
545 	return 0;
546 }
547 static inline bool fib6_rules_early_flow_dissect(struct net *net,
548 						 struct sk_buff *skb,
549 						 struct flowi6 *fl6,
550 						 struct flow_keys *flkeys)
551 {
552 	return false;
553 }
554 #endif
555 #endif
556