xref: /linux/include/net/ip_fib.h (revision 7ff836f064e2c814a7504c91a4464eea45d475bd)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET  is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Forwarding Information Base.
7  *
8  * Authors:	A.N.Kuznetsov, <kuznet@ms2.inr.ac.ru>
9  *
10  *		This program is free software; you can redistribute it and/or
11  *		modify it under the terms of the GNU General Public License
12  *		as published by the Free Software Foundation; either version
13  *		2 of the License, or (at your option) any later version.
14  */
15 
16 #ifndef _NET_IP_FIB_H
17 #define _NET_IP_FIB_H
18 
19 #include <net/flow.h>
20 #include <linux/seq_file.h>
21 #include <linux/rcupdate.h>
22 #include <net/fib_notifier.h>
23 #include <net/fib_rules.h>
24 #include <net/inetpeer.h>
25 #include <linux/percpu.h>
26 #include <linux/notifier.h>
27 #include <linux/refcount.h>
28 
29 struct fib_config {
30 	u8			fc_dst_len;
31 	u8			fc_tos;
32 	u8			fc_protocol;
33 	u8			fc_scope;
34 	u8			fc_type;
35 	u8			fc_gw_family;
36 	/* 2 bytes unused */
37 	u32			fc_table;
38 	__be32			fc_dst;
39 	union {
40 		__be32		fc_gw4;
41 		struct in6_addr	fc_gw6;
42 	};
43 	int			fc_oif;
44 	u32			fc_flags;
45 	u32			fc_priority;
46 	__be32			fc_prefsrc;
47 	struct nlattr		*fc_mx;
48 	struct rtnexthop	*fc_mp;
49 	int			fc_mx_len;
50 	int			fc_mp_len;
51 	u32			fc_flow;
52 	u32			fc_nlflags;
53 	struct nl_info		fc_nlinfo;
54 	struct nlattr		*fc_encap;
55 	u16			fc_encap_type;
56 };
57 
58 struct fib_info;
59 struct rtable;
60 
61 struct fib_nh_exception {
62 	struct fib_nh_exception __rcu	*fnhe_next;
63 	int				fnhe_genid;
64 	__be32				fnhe_daddr;
65 	u32				fnhe_pmtu;
66 	bool				fnhe_mtu_locked;
67 	__be32				fnhe_gw;
68 	unsigned long			fnhe_expires;
69 	struct rtable __rcu		*fnhe_rth_input;
70 	struct rtable __rcu		*fnhe_rth_output;
71 	unsigned long			fnhe_stamp;
72 	struct rcu_head			rcu;
73 };
74 
75 struct fnhe_hash_bucket {
76 	struct fib_nh_exception __rcu	*chain;
77 };
78 
79 #define FNHE_HASH_SHIFT		11
80 #define FNHE_HASH_SIZE		(1 << FNHE_HASH_SHIFT)
81 #define FNHE_RECLAIM_DEPTH	5
82 
83 struct fib_nh_common {
84 	struct net_device	*nhc_dev;
85 	int			nhc_oif;
86 	unsigned char		nhc_scope;
87 	u8			nhc_family;
88 	u8			nhc_gw_family;
89 	unsigned char		nhc_flags;
90 	struct lwtunnel_state	*nhc_lwtstate;
91 
92 	union {
93 		__be32          ipv4;
94 		struct in6_addr ipv6;
95 	} nhc_gw;
96 
97 	int			nhc_weight;
98 	atomic_t		nhc_upper_bound;
99 
100 	/* v4 specific, but allows fib6_nh with v4 routes */
101 	struct rtable __rcu * __percpu *nhc_pcpu_rth_output;
102 	struct rtable __rcu     *nhc_rth_input;
103 	struct fnhe_hash_bucket	__rcu *nhc_exceptions;
104 };
105 
106 struct fib_nh {
107 	struct fib_nh_common	nh_common;
108 	struct hlist_node	nh_hash;
109 	struct fib_info		*nh_parent;
110 #ifdef CONFIG_IP_ROUTE_CLASSID
111 	__u32			nh_tclassid;
112 #endif
113 	__be32			nh_saddr;
114 	int			nh_saddr_genid;
115 #define fib_nh_family		nh_common.nhc_family
116 #define fib_nh_dev		nh_common.nhc_dev
117 #define fib_nh_oif		nh_common.nhc_oif
118 #define fib_nh_flags		nh_common.nhc_flags
119 #define fib_nh_lws		nh_common.nhc_lwtstate
120 #define fib_nh_scope		nh_common.nhc_scope
121 #define fib_nh_gw_family	nh_common.nhc_gw_family
122 #define fib_nh_gw4		nh_common.nhc_gw.ipv4
123 #define fib_nh_gw6		nh_common.nhc_gw.ipv6
124 #define fib_nh_weight		nh_common.nhc_weight
125 #define fib_nh_upper_bound	nh_common.nhc_upper_bound
126 };
127 
128 /*
129  * This structure contains data shared by many of routes.
130  */
131 
132 struct fib_info {
133 	struct hlist_node	fib_hash;
134 	struct hlist_node	fib_lhash;
135 	struct net		*fib_net;
136 	int			fib_treeref;
137 	refcount_t		fib_clntref;
138 	unsigned int		fib_flags;
139 	unsigned char		fib_dead;
140 	unsigned char		fib_protocol;
141 	unsigned char		fib_scope;
142 	unsigned char		fib_type;
143 	__be32			fib_prefsrc;
144 	u32			fib_tb_id;
145 	u32			fib_priority;
146 	struct dst_metrics	*fib_metrics;
147 #define fib_mtu fib_metrics->metrics[RTAX_MTU-1]
148 #define fib_window fib_metrics->metrics[RTAX_WINDOW-1]
149 #define fib_rtt fib_metrics->metrics[RTAX_RTT-1]
150 #define fib_advmss fib_metrics->metrics[RTAX_ADVMSS-1]
151 	int			fib_nhs;
152 	bool			fib_nh_is_v6;
153 	bool			nh_updated;
154 	struct rcu_head		rcu;
155 	struct fib_nh		fib_nh[0];
156 #define fib_dev		fib_nh[0].fib_nh_dev
157 };
158 
159 
160 #ifdef CONFIG_IP_MULTIPLE_TABLES
161 struct fib_rule;
162 #endif
163 
164 struct fib_table;
165 struct fib_result {
166 	__be32			prefix;
167 	unsigned char		prefixlen;
168 	unsigned char		nh_sel;
169 	unsigned char		type;
170 	unsigned char		scope;
171 	u32			tclassid;
172 	struct fib_nh_common	*nhc;
173 	struct fib_info		*fi;
174 	struct fib_table	*table;
175 	struct hlist_head	*fa_head;
176 };
177 
178 struct fib_result_nl {
179 	__be32		fl_addr;   /* To be looked up*/
180 	u32		fl_mark;
181 	unsigned char	fl_tos;
182 	unsigned char   fl_scope;
183 	unsigned char   tb_id_in;
184 
185 	unsigned char   tb_id;      /* Results */
186 	unsigned char	prefixlen;
187 	unsigned char	nh_sel;
188 	unsigned char	type;
189 	unsigned char	scope;
190 	int             err;
191 };
192 
193 static inline struct fib_nh_common *fib_info_nhc(struct fib_info *fi, int nhsel)
194 {
195 	return &fi->fib_nh[nhsel].nh_common;
196 }
197 
198 #ifdef CONFIG_IP_MULTIPLE_TABLES
199 #define FIB_TABLE_HASHSZ 256
200 #else
201 #define FIB_TABLE_HASHSZ 2
202 #endif
203 
204 __be32 fib_info_update_nh_saddr(struct net *net, struct fib_nh *nh,
205 				unsigned char scope);
206 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res);
207 
208 #define FIB_RES_NHC(res)		((res).nhc)
209 #define FIB_RES_DEV(res)	(FIB_RES_NHC(res)->nhc_dev)
210 #define FIB_RES_OIF(res)	(FIB_RES_NHC(res)->nhc_oif)
211 
212 struct fib_entry_notifier_info {
213 	struct fib_notifier_info info; /* must be first */
214 	u32 dst;
215 	int dst_len;
216 	struct fib_info *fi;
217 	u8 tos;
218 	u8 type;
219 	u32 tb_id;
220 };
221 
222 struct fib_nh_notifier_info {
223 	struct fib_notifier_info info; /* must be first */
224 	struct fib_nh *fib_nh;
225 };
226 
227 int call_fib4_notifier(struct notifier_block *nb, struct net *net,
228 		       enum fib_event_type event_type,
229 		       struct fib_notifier_info *info);
230 int call_fib4_notifiers(struct net *net, enum fib_event_type event_type,
231 			struct fib_notifier_info *info);
232 
233 int __net_init fib4_notifier_init(struct net *net);
234 void __net_exit fib4_notifier_exit(struct net *net);
235 
236 void fib_info_notify_update(struct net *net, struct nl_info *info);
237 void fib_notify(struct net *net, struct notifier_block *nb);
238 
239 struct fib_table {
240 	struct hlist_node	tb_hlist;
241 	u32			tb_id;
242 	int			tb_num_default;
243 	struct rcu_head		rcu;
244 	unsigned long 		*tb_data;
245 	unsigned long		__data[0];
246 };
247 
248 struct fib_dump_filter {
249 	u32			table_id;
250 	/* filter_set is an optimization that an entry is set */
251 	bool			filter_set;
252 	bool			dump_all_families;
253 	unsigned char		protocol;
254 	unsigned char		rt_type;
255 	unsigned int		flags;
256 	struct net_device	*dev;
257 };
258 
259 int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
260 		     struct fib_result *res, int fib_flags);
261 int fib_table_insert(struct net *, struct fib_table *, struct fib_config *,
262 		     struct netlink_ext_ack *extack);
263 int fib_table_delete(struct net *, struct fib_table *, struct fib_config *,
264 		     struct netlink_ext_ack *extack);
265 int fib_table_dump(struct fib_table *table, struct sk_buff *skb,
266 		   struct netlink_callback *cb, struct fib_dump_filter *filter);
267 int fib_table_flush(struct net *net, struct fib_table *table, bool flush_all);
268 struct fib_table *fib_trie_unmerge(struct fib_table *main_tb);
269 void fib_table_flush_external(struct fib_table *table);
270 void fib_free_table(struct fib_table *tb);
271 
272 #ifndef CONFIG_IP_MULTIPLE_TABLES
273 
274 #define TABLE_LOCAL_INDEX	(RT_TABLE_LOCAL & (FIB_TABLE_HASHSZ - 1))
275 #define TABLE_MAIN_INDEX	(RT_TABLE_MAIN  & (FIB_TABLE_HASHSZ - 1))
276 
277 static inline struct fib_table *fib_get_table(struct net *net, u32 id)
278 {
279 	struct hlist_node *tb_hlist;
280 	struct hlist_head *ptr;
281 
282 	ptr = id == RT_TABLE_LOCAL ?
283 		&net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX] :
284 		&net->ipv4.fib_table_hash[TABLE_MAIN_INDEX];
285 
286 	tb_hlist = rcu_dereference_rtnl(hlist_first_rcu(ptr));
287 
288 	return hlist_entry(tb_hlist, struct fib_table, tb_hlist);
289 }
290 
291 static inline struct fib_table *fib_new_table(struct net *net, u32 id)
292 {
293 	return fib_get_table(net, id);
294 }
295 
296 static inline int fib_lookup(struct net *net, const struct flowi4 *flp,
297 			     struct fib_result *res, unsigned int flags)
298 {
299 	struct fib_table *tb;
300 	int err = -ENETUNREACH;
301 
302 	rcu_read_lock();
303 
304 	tb = fib_get_table(net, RT_TABLE_MAIN);
305 	if (tb)
306 		err = fib_table_lookup(tb, flp, res, flags | FIB_LOOKUP_NOREF);
307 
308 	if (err == -EAGAIN)
309 		err = -ENETUNREACH;
310 
311 	rcu_read_unlock();
312 
313 	return err;
314 }
315 
316 static inline bool fib4_rule_default(const struct fib_rule *rule)
317 {
318 	return true;
319 }
320 
321 static inline int fib4_rules_dump(struct net *net, struct notifier_block *nb)
322 {
323 	return 0;
324 }
325 
326 static inline unsigned int fib4_rules_seq_read(struct net *net)
327 {
328 	return 0;
329 }
330 
331 static inline bool fib4_rules_early_flow_dissect(struct net *net,
332 						 struct sk_buff *skb,
333 						 struct flowi4 *fl4,
334 						 struct flow_keys *flkeys)
335 {
336 	return false;
337 }
338 #else /* CONFIG_IP_MULTIPLE_TABLES */
339 int __net_init fib4_rules_init(struct net *net);
340 void __net_exit fib4_rules_exit(struct net *net);
341 
342 struct fib_table *fib_new_table(struct net *net, u32 id);
343 struct fib_table *fib_get_table(struct net *net, u32 id);
344 
345 int __fib_lookup(struct net *net, struct flowi4 *flp,
346 		 struct fib_result *res, unsigned int flags);
347 
348 static inline int fib_lookup(struct net *net, struct flowi4 *flp,
349 			     struct fib_result *res, unsigned int flags)
350 {
351 	struct fib_table *tb;
352 	int err = -ENETUNREACH;
353 
354 	flags |= FIB_LOOKUP_NOREF;
355 	if (net->ipv4.fib_has_custom_rules)
356 		return __fib_lookup(net, flp, res, flags);
357 
358 	rcu_read_lock();
359 
360 	res->tclassid = 0;
361 
362 	tb = rcu_dereference_rtnl(net->ipv4.fib_main);
363 	if (tb)
364 		err = fib_table_lookup(tb, flp, res, flags);
365 
366 	if (!err)
367 		goto out;
368 
369 	tb = rcu_dereference_rtnl(net->ipv4.fib_default);
370 	if (tb)
371 		err = fib_table_lookup(tb, flp, res, flags);
372 
373 out:
374 	if (err == -EAGAIN)
375 		err = -ENETUNREACH;
376 
377 	rcu_read_unlock();
378 
379 	return err;
380 }
381 
382 bool fib4_rule_default(const struct fib_rule *rule);
383 int fib4_rules_dump(struct net *net, struct notifier_block *nb);
384 unsigned int fib4_rules_seq_read(struct net *net);
385 
386 static inline bool fib4_rules_early_flow_dissect(struct net *net,
387 						 struct sk_buff *skb,
388 						 struct flowi4 *fl4,
389 						 struct flow_keys *flkeys)
390 {
391 	unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
392 
393 	if (!net->ipv4.fib_rules_require_fldissect)
394 		return false;
395 
396 	skb_flow_dissect_flow_keys(skb, flkeys, flag);
397 	fl4->fl4_sport = flkeys->ports.src;
398 	fl4->fl4_dport = flkeys->ports.dst;
399 	fl4->flowi4_proto = flkeys->basic.ip_proto;
400 
401 	return true;
402 }
403 
404 #endif /* CONFIG_IP_MULTIPLE_TABLES */
405 
406 /* Exported by fib_frontend.c */
407 extern const struct nla_policy rtm_ipv4_policy[];
408 void ip_fib_init(void);
409 int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
410 		    struct netlink_ext_ack *extack);
411 __be32 fib_compute_spec_dst(struct sk_buff *skb);
412 bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev);
413 int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
414 			u8 tos, int oif, struct net_device *dev,
415 			struct in_device *idev, u32 *itag);
416 #ifdef CONFIG_IP_ROUTE_CLASSID
417 static inline int fib_num_tclassid_users(struct net *net)
418 {
419 	return net->ipv4.fib_num_tclassid_users;
420 }
421 #else
422 static inline int fib_num_tclassid_users(struct net *net)
423 {
424 	return 0;
425 }
426 #endif
427 int fib_unmerge(struct net *net);
428 
429 /* Exported by fib_semantics.c */
430 int ip_fib_check_default(__be32 gw, struct net_device *dev);
431 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force);
432 int fib_sync_down_addr(struct net_device *dev, __be32 local);
433 int fib_sync_up(struct net_device *dev, unsigned char nh_flags);
434 void fib_sync_mtu(struct net_device *dev, u32 orig_mtu);
435 void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig);
436 
437 #ifdef CONFIG_IP_ROUTE_MULTIPATH
438 int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,
439 		       const struct sk_buff *skb, struct flow_keys *flkeys);
440 #endif
441 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
442 		 struct netlink_ext_ack *extack);
443 void fib_select_multipath(struct fib_result *res, int hash);
444 void fib_select_path(struct net *net, struct fib_result *res,
445 		     struct flowi4 *fl4, const struct sk_buff *skb);
446 
447 int fib_nh_init(struct net *net, struct fib_nh *fib_nh,
448 		struct fib_config *cfg, int nh_weight,
449 		struct netlink_ext_ack *extack);
450 void fib_nh_release(struct net *net, struct fib_nh *fib_nh);
451 int fib_nh_common_init(struct fib_nh_common *nhc, struct nlattr *fc_encap,
452 		       u16 fc_encap_type, void *cfg, gfp_t gfp_flags,
453 		       struct netlink_ext_ack *extack);
454 void fib_nh_common_release(struct fib_nh_common *nhc);
455 
456 /* Exported by fib_trie.c */
457 void fib_trie_init(void);
458 struct fib_table *fib_trie_table(u32 id, struct fib_table *alias);
459 
460 static inline void fib_combine_itag(u32 *itag, const struct fib_result *res)
461 {
462 #ifdef CONFIG_IP_ROUTE_CLASSID
463 	struct fib_nh_common *nhc = res->nhc;
464 	struct fib_nh *nh = container_of(nhc, struct fib_nh, nh_common);
465 #ifdef CONFIG_IP_MULTIPLE_TABLES
466 	u32 rtag;
467 #endif
468 	*itag = nh->nh_tclassid << 16;
469 #ifdef CONFIG_IP_MULTIPLE_TABLES
470 	rtag = res->tclassid;
471 	if (*itag == 0)
472 		*itag = (rtag<<16);
473 	*itag |= (rtag>>16);
474 #endif
475 #endif
476 }
477 
478 void fib_flush(struct net *net);
479 void free_fib_info(struct fib_info *fi);
480 
481 static inline void fib_info_hold(struct fib_info *fi)
482 {
483 	refcount_inc(&fi->fib_clntref);
484 }
485 
486 static inline void fib_info_put(struct fib_info *fi)
487 {
488 	if (refcount_dec_and_test(&fi->fib_clntref))
489 		free_fib_info(fi);
490 }
491 
492 #ifdef CONFIG_PROC_FS
493 int __net_init fib_proc_init(struct net *net);
494 void __net_exit fib_proc_exit(struct net *net);
495 #else
496 static inline int fib_proc_init(struct net *net)
497 {
498 	return 0;
499 }
500 static inline void fib_proc_exit(struct net *net)
501 {
502 }
503 #endif
504 
505 u32 ip_mtu_from_fib_result(struct fib_result *res, __be32 daddr);
506 
507 int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
508 			  struct fib_dump_filter *filter,
509 			  struct netlink_callback *cb);
510 
511 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nh,
512 		     unsigned char *flags, bool skip_oif);
513 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nh,
514 		    int nh_weight);
515 #endif  /* _NET_FIB_H */
516