Loading...
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Forwarding Information Base.
8 *
9 * Authors: A.N.Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12#ifndef _NET_IP_FIB_H
13#define _NET_IP_FIB_H
14
15#include <net/flow.h>
16#include <linux/seq_file.h>
17#include <linux/rcupdate.h>
18#include <net/fib_notifier.h>
19#include <net/fib_rules.h>
20#include <net/inet_dscp.h>
21#include <net/inetpeer.h>
22#include <linux/percpu.h>
23#include <linux/notifier.h>
24#include <linux/refcount.h>
25#include <linux/ip.h>
26#include <linux/in_route.h>
27
28struct fib_config {
29 u8 fc_dst_len;
30 dscp_t fc_dscp;
31 u8 fc_protocol;
32 u8 fc_scope;
33 u8 fc_type;
34 u8 fc_gw_family;
35 /* 2 bytes unused */
36 u32 fc_table;
37 __be32 fc_dst;
38 union {
39 __be32 fc_gw4;
40 struct in6_addr fc_gw6;
41 };
42 int fc_oif;
43 u32 fc_flags;
44 u32 fc_priority;
45 __be32 fc_prefsrc;
46 u32 fc_nh_id;
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
58struct fib_info;
59struct rtable;
60
61struct 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
75struct 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
83struct fib_nh_common {
84 struct net_device *nhc_dev;
85 netdevice_tracker nhc_dev_tracker;
86 int nhc_oif;
87 unsigned char nhc_scope;
88 u8 nhc_family;
89 u8 nhc_gw_family;
90 unsigned char nhc_flags;
91 struct lwtunnel_state *nhc_lwtstate;
92
93 union {
94 __be32 ipv4;
95 struct in6_addr ipv6;
96 } nhc_gw;
97
98 int nhc_weight;
99 atomic_t nhc_upper_bound;
100
101 /* v4 specific, but allows fib6_nh with v4 routes */
102 struct rtable __rcu * __percpu *nhc_pcpu_rth_output;
103 struct rtable __rcu *nhc_rth_input;
104 struct fnhe_hash_bucket __rcu *nhc_exceptions;
105};
106
107struct fib_nh {
108 struct fib_nh_common nh_common;
109 struct hlist_node nh_hash;
110 struct fib_info *nh_parent;
111#ifdef CONFIG_IP_ROUTE_CLASSID
112 __u32 nh_tclassid;
113#endif
114 __be32 nh_saddr;
115 int nh_saddr_genid;
116#define fib_nh_family nh_common.nhc_family
117#define fib_nh_dev nh_common.nhc_dev
118#define fib_nh_dev_tracker nh_common.nhc_dev_tracker
119#define fib_nh_oif nh_common.nhc_oif
120#define fib_nh_flags nh_common.nhc_flags
121#define fib_nh_lws nh_common.nhc_lwtstate
122#define fib_nh_scope nh_common.nhc_scope
123#define fib_nh_gw_family nh_common.nhc_gw_family
124#define fib_nh_gw4 nh_common.nhc_gw.ipv4
125#define fib_nh_gw6 nh_common.nhc_gw.ipv6
126#define fib_nh_weight nh_common.nhc_weight
127#define fib_nh_upper_bound nh_common.nhc_upper_bound
128};
129
130/*
131 * This structure contains data shared by many of routes.
132 */
133
134struct nexthop;
135
136struct fib_info {
137 struct hlist_node fib_hash;
138 struct hlist_node fib_lhash;
139 struct list_head nh_list;
140 struct net *fib_net;
141 refcount_t fib_treeref;
142 refcount_t fib_clntref;
143 unsigned int fib_flags;
144 unsigned char fib_dead;
145 unsigned char fib_protocol;
146 unsigned char fib_scope;
147 unsigned char fib_type;
148 __be32 fib_prefsrc;
149 u32 fib_tb_id;
150 u32 fib_priority;
151 struct dst_metrics *fib_metrics;
152#define fib_mtu fib_metrics->metrics[RTAX_MTU-1]
153#define fib_window fib_metrics->metrics[RTAX_WINDOW-1]
154#define fib_rtt fib_metrics->metrics[RTAX_RTT-1]
155#define fib_advmss fib_metrics->metrics[RTAX_ADVMSS-1]
156 int fib_nhs;
157 bool fib_nh_is_v6;
158 bool nh_updated;
159 bool pfsrc_removed;
160 struct nexthop *nh;
161 struct rcu_head rcu;
162 struct fib_nh fib_nh[] __counted_by(fib_nhs);
163};
164
165
166#ifdef CONFIG_IP_MULTIPLE_TABLES
167struct fib_rule;
168#endif
169
170struct fib_table;
171struct fib_result {
172 __be32 prefix;
173 unsigned char prefixlen;
174 unsigned char nh_sel;
175 unsigned char type;
176 unsigned char scope;
177 u32 tclassid;
178 dscp_t dscp;
179 struct fib_nh_common *nhc;
180 struct fib_info *fi;
181 struct fib_table *table;
182 struct hlist_head *fa_head;
183};
184
185struct fib_result_nl {
186 __be32 fl_addr; /* To be looked up*/
187 u32 fl_mark;
188 unsigned char fl_tos;
189 unsigned char fl_scope;
190 unsigned char tb_id_in;
191
192 unsigned char tb_id; /* Results */
193 unsigned char prefixlen;
194 unsigned char nh_sel;
195 unsigned char type;
196 unsigned char scope;
197 int err;
198};
199
200#ifdef CONFIG_IP_MULTIPLE_TABLES
201#define FIB_TABLE_HASHSZ 256
202#else
203#define FIB_TABLE_HASHSZ 2
204#endif
205
206__be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
207 unsigned char scope);
208__be32 fib_result_prefsrc(struct net *net, struct fib_result *res);
209
210#define FIB_RES_NHC(res) ((res).nhc)
211#define FIB_RES_DEV(res) (FIB_RES_NHC(res)->nhc_dev)
212#define FIB_RES_OIF(res) (FIB_RES_NHC(res)->nhc_oif)
213
214struct fib_rt_info {
215 struct fib_info *fi;
216 u32 tb_id;
217 __be32 dst;
218 int dst_len;
219 dscp_t dscp;
220 u8 type;
221 u8 offload:1,
222 trap:1,
223 offload_failed:1,
224 unused:5;
225};
226
227struct fib_entry_notifier_info {
228 struct fib_notifier_info info; /* must be first */
229 u32 dst;
230 int dst_len;
231 struct fib_info *fi;
232 dscp_t dscp;
233 u8 type;
234 u32 tb_id;
235};
236
237struct fib_nh_notifier_info {
238 struct fib_notifier_info info; /* must be first */
239 struct fib_nh *fib_nh;
240};
241
242int call_fib4_notifier(struct notifier_block *nb,
243 enum fib_event_type event_type,
244 struct fib_notifier_info *info);
245int call_fib4_notifiers(struct net *net, enum fib_event_type event_type,
246 struct fib_notifier_info *info);
247
248int __net_init fib4_notifier_init(struct net *net);
249void __net_exit fib4_notifier_exit(struct net *net);
250
251void fib_info_notify_update(struct net *net, struct nl_info *info);
252int fib_notify(struct net *net, struct notifier_block *nb,
253 struct netlink_ext_ack *extack);
254
255struct fib_table {
256 struct hlist_node tb_hlist;
257 u32 tb_id;
258 int tb_num_default;
259 struct rcu_head rcu;
260 unsigned long *tb_data;
261 unsigned long __data[];
262};
263
264struct fib_dump_filter {
265 u32 table_id;
266 /* filter_set is an optimization that an entry is set */
267 bool filter_set;
268 bool dump_routes;
269 bool dump_exceptions;
270 bool rtnl_held;
271 unsigned char protocol;
272 unsigned char rt_type;
273 unsigned int flags;
274 struct net_device *dev;
275};
276
277int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
278 struct fib_result *res, int fib_flags);
279int fib_table_insert(struct net *, struct fib_table *, struct fib_config *,
280 struct netlink_ext_ack *extack);
281int fib_table_delete(struct net *, struct fib_table *, struct fib_config *,
282 struct netlink_ext_ack *extack);
283int fib_table_dump(struct fib_table *table, struct sk_buff *skb,
284 struct netlink_callback *cb, struct fib_dump_filter *filter);
285int fib_table_flush(struct net *net, struct fib_table *table, bool flush_all);
286struct fib_table *fib_trie_unmerge(struct fib_table *main_tb);
287void fib_table_flush_external(struct fib_table *table);
288void fib_free_table(struct fib_table *tb);
289
290#ifndef CONFIG_IP_MULTIPLE_TABLES
291
292#define TABLE_LOCAL_INDEX (RT_TABLE_LOCAL & (FIB_TABLE_HASHSZ - 1))
293#define TABLE_MAIN_INDEX (RT_TABLE_MAIN & (FIB_TABLE_HASHSZ - 1))
294
295static inline struct fib_table *fib_get_table(struct net *net, u32 id)
296{
297 struct hlist_node *tb_hlist;
298 struct hlist_head *ptr;
299
300 ptr = id == RT_TABLE_LOCAL ?
301 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX] :
302 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX];
303
304 tb_hlist = rcu_dereference_rtnl(hlist_first_rcu(ptr));
305
306 return hlist_entry(tb_hlist, struct fib_table, tb_hlist);
307}
308
309static inline struct fib_table *fib_new_table(struct net *net, u32 id)
310{
311 return fib_get_table(net, id);
312}
313
314static inline int fib_lookup(struct net *net, const struct flowi4 *flp,
315 struct fib_result *res, unsigned int flags)
316{
317 struct fib_table *tb;
318 int err = -ENETUNREACH;
319
320 rcu_read_lock();
321
322 tb = fib_get_table(net, RT_TABLE_MAIN);
323 if (tb)
324 err = fib_table_lookup(tb, flp, res, flags | FIB_LOOKUP_NOREF);
325
326 if (err == -EAGAIN)
327 err = -ENETUNREACH;
328
329 rcu_read_unlock();
330
331 return err;
332}
333
334static inline bool fib4_has_custom_rules(const struct net *net)
335{
336 return false;
337}
338
339static inline bool fib4_rule_default(const struct fib_rule *rule)
340{
341 return true;
342}
343
344static inline int fib4_rules_dump(struct net *net, struct notifier_block *nb,
345 struct netlink_ext_ack *extack)
346{
347 return 0;
348}
349
350static inline unsigned int fib4_rules_seq_read(const struct net *net)
351{
352 return 0;
353}
354
355static inline bool fib4_rules_early_flow_dissect(struct net *net,
356 struct sk_buff *skb,
357 struct flowi4 *fl4,
358 struct flow_keys *flkeys)
359{
360 return false;
361}
362#else /* CONFIG_IP_MULTIPLE_TABLES */
363int __net_init fib4_rules_init(struct net *net);
364void __net_exit fib4_rules_exit(struct net *net);
365
366struct fib_table *fib_new_table(struct net *net, u32 id);
367struct fib_table *fib_get_table(struct net *net, u32 id);
368
369int __fib_lookup(struct net *net, struct flowi4 *flp,
370 struct fib_result *res, unsigned int flags);
371
372static inline int fib_lookup(struct net *net, struct flowi4 *flp,
373 struct fib_result *res, unsigned int flags)
374{
375 struct fib_table *tb;
376 int err = -ENETUNREACH;
377
378 flags |= FIB_LOOKUP_NOREF;
379 if (net->ipv4.fib_has_custom_rules)
380 return __fib_lookup(net, flp, res, flags);
381
382 rcu_read_lock();
383
384 res->tclassid = 0;
385
386 tb = rcu_dereference_rtnl(net->ipv4.fib_main);
387 if (tb)
388 err = fib_table_lookup(tb, flp, res, flags);
389
390 if (!err)
391 goto out;
392
393 tb = rcu_dereference_rtnl(net->ipv4.fib_default);
394 if (tb)
395 err = fib_table_lookup(tb, flp, res, flags);
396
397out:
398 if (err == -EAGAIN)
399 err = -ENETUNREACH;
400
401 rcu_read_unlock();
402
403 return err;
404}
405
406static inline bool fib4_has_custom_rules(const struct net *net)
407{
408 return net->ipv4.fib_has_custom_rules;
409}
410
411bool fib4_rule_default(const struct fib_rule *rule);
412int fib4_rules_dump(struct net *net, struct notifier_block *nb,
413 struct netlink_ext_ack *extack);
414unsigned int fib4_rules_seq_read(const struct net *net);
415
416static inline bool fib4_rules_early_flow_dissect(struct net *net,
417 struct sk_buff *skb,
418 struct flowi4 *fl4,
419 struct flow_keys *flkeys)
420{
421 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
422
423 if (!net->ipv4.fib_rules_require_fldissect)
424 return false;
425
426 memset(flkeys, 0, sizeof(*flkeys));
427 __skb_flow_dissect(net, skb, &flow_keys_dissector,
428 flkeys, NULL, 0, 0, 0, flag);
429
430 fl4->fl4_sport = flkeys->ports.src;
431 fl4->fl4_dport = flkeys->ports.dst;
432 fl4->flowi4_proto = flkeys->basic.ip_proto;
433
434 return true;
435}
436
437#endif /* CONFIG_IP_MULTIPLE_TABLES */
438
439static inline bool fib_dscp_masked_match(dscp_t dscp, const struct flowi4 *fl4)
440{
441 return dscp == inet_dsfield_to_dscp(RT_TOS(fl4->flowi4_tos));
442}
443
444/* Exported by fib_frontend.c */
445extern const struct nla_policy rtm_ipv4_policy[];
446void ip_fib_init(void);
447int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
448 struct netlink_ext_ack *extack);
449__be32 fib_compute_spec_dst(struct sk_buff *skb);
450bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev);
451int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
452 dscp_t dscp, int oif, struct net_device *dev,
453 struct in_device *idev, u32 *itag);
454
455static inline enum skb_drop_reason
456fib_validate_source_reason(struct sk_buff *skb, __be32 src, __be32 dst,
457 dscp_t dscp, int oif, struct net_device *dev,
458 struct in_device *idev, u32 *itag)
459{
460 int err = fib_validate_source(skb, src, dst, dscp, oif, dev, idev,
461 itag);
462 if (err < 0)
463 return -err;
464 return SKB_NOT_DROPPED_YET;
465}
466
467#ifdef CONFIG_IP_ROUTE_CLASSID
468static inline int fib_num_tclassid_users(struct net *net)
469{
470 return atomic_read(&net->ipv4.fib_num_tclassid_users);
471}
472#else
473static inline int fib_num_tclassid_users(struct net *net)
474{
475 return 0;
476}
477#endif
478int fib_unmerge(struct net *net);
479
480static inline bool nhc_l3mdev_matches_dev(const struct fib_nh_common *nhc,
481const struct net_device *dev)
482{
483 if (nhc->nhc_dev == dev ||
484 l3mdev_master_ifindex_rcu(nhc->nhc_dev) == dev->ifindex)
485 return true;
486
487 return false;
488}
489
490/* Exported by fib_semantics.c */
491int ip_fib_check_default(__be32 gw, struct net_device *dev);
492int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force);
493int fib_sync_down_addr(struct net_device *dev, __be32 local);
494int fib_sync_up(struct net_device *dev, unsigned char nh_flags);
495void fib_sync_mtu(struct net_device *dev, u32 orig_mtu);
496void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig);
497
498/* Fields used for sysctl_fib_multipath_hash_fields.
499 * Common to IPv4 and IPv6.
500 *
501 * Add new fields at the end. This is user API.
502 */
503#define FIB_MULTIPATH_HASH_FIELD_SRC_IP BIT(0)
504#define FIB_MULTIPATH_HASH_FIELD_DST_IP BIT(1)
505#define FIB_MULTIPATH_HASH_FIELD_IP_PROTO BIT(2)
506#define FIB_MULTIPATH_HASH_FIELD_FLOWLABEL BIT(3)
507#define FIB_MULTIPATH_HASH_FIELD_SRC_PORT BIT(4)
508#define FIB_MULTIPATH_HASH_FIELD_DST_PORT BIT(5)
509#define FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP BIT(6)
510#define FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP BIT(7)
511#define FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO BIT(8)
512#define FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL BIT(9)
513#define FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT BIT(10)
514#define FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT BIT(11)
515
516#define FIB_MULTIPATH_HASH_FIELD_OUTER_MASK \
517 (FIB_MULTIPATH_HASH_FIELD_SRC_IP | \
518 FIB_MULTIPATH_HASH_FIELD_DST_IP | \
519 FIB_MULTIPATH_HASH_FIELD_IP_PROTO | \
520 FIB_MULTIPATH_HASH_FIELD_FLOWLABEL | \
521 FIB_MULTIPATH_HASH_FIELD_SRC_PORT | \
522 FIB_MULTIPATH_HASH_FIELD_DST_PORT)
523
524#define FIB_MULTIPATH_HASH_FIELD_INNER_MASK \
525 (FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP | \
526 FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP | \
527 FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO | \
528 FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL | \
529 FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT | \
530 FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
531
532#define FIB_MULTIPATH_HASH_FIELD_ALL_MASK \
533 (FIB_MULTIPATH_HASH_FIELD_OUTER_MASK | \
534 FIB_MULTIPATH_HASH_FIELD_INNER_MASK)
535
536#define FIB_MULTIPATH_HASH_FIELD_DEFAULT_MASK \
537 (FIB_MULTIPATH_HASH_FIELD_SRC_IP | \
538 FIB_MULTIPATH_HASH_FIELD_DST_IP | \
539 FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
540
541#ifdef CONFIG_IP_ROUTE_MULTIPATH
542int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,
543 const struct sk_buff *skb, struct flow_keys *flkeys);
544
545static void
546fib_multipath_hash_construct_key(siphash_key_t *key, u32 mp_seed)
547{
548 u64 mp_seed_64 = mp_seed;
549
550 key->key[0] = (mp_seed_64 << 32) | mp_seed_64;
551 key->key[1] = key->key[0];
552}
553
554static inline u32 fib_multipath_hash_from_keys(const struct net *net,
555 struct flow_keys *keys)
556{
557 siphash_aligned_key_t hash_key;
558 u32 mp_seed;
559
560 mp_seed = READ_ONCE(net->ipv4.sysctl_fib_multipath_hash_seed).mp_seed;
561 fib_multipath_hash_construct_key(&hash_key, mp_seed);
562
563 return flow_hash_from_keys_seed(keys, &hash_key);
564}
565#else
566static inline u32 fib_multipath_hash_from_keys(const struct net *net,
567 struct flow_keys *keys)
568{
569 return flow_hash_from_keys(keys);
570}
571#endif
572
573int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
574 struct netlink_ext_ack *extack);
575void fib_select_multipath(struct fib_result *res, int hash);
576void fib_select_path(struct net *net, struct fib_result *res,
577 struct flowi4 *fl4, const struct sk_buff *skb);
578
579int fib_nh_init(struct net *net, struct fib_nh *fib_nh,
580 struct fib_config *cfg, int nh_weight,
581 struct netlink_ext_ack *extack);
582void fib_nh_release(struct net *net, struct fib_nh *fib_nh);
583int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
584 struct nlattr *fc_encap, u16 fc_encap_type,
585 void *cfg, gfp_t gfp_flags,
586 struct netlink_ext_ack *extack);
587void fib_nh_common_release(struct fib_nh_common *nhc);
588
589/* Exported by fib_trie.c */
590void fib_alias_hw_flags_set(struct net *net, const struct fib_rt_info *fri);
591void fib_trie_init(void);
592struct fib_table *fib_trie_table(u32 id, struct fib_table *alias);
593bool fib_lookup_good_nhc(const struct fib_nh_common *nhc, int fib_flags,
594 const struct flowi4 *flp);
595
596static inline void fib_combine_itag(u32 *itag, const struct fib_result *res)
597{
598#ifdef CONFIG_IP_ROUTE_CLASSID
599 struct fib_nh_common *nhc = res->nhc;
600#ifdef CONFIG_IP_MULTIPLE_TABLES
601 u32 rtag;
602#endif
603 if (nhc->nhc_family == AF_INET) {
604 struct fib_nh *nh;
605
606 nh = container_of(nhc, struct fib_nh, nh_common);
607 *itag = nh->nh_tclassid << 16;
608 } else {
609 *itag = 0;
610 }
611
612#ifdef CONFIG_IP_MULTIPLE_TABLES
613 rtag = res->tclassid;
614 if (*itag == 0)
615 *itag = (rtag<<16);
616 *itag |= (rtag>>16);
617#endif
618#endif
619}
620
621void fib_flush(struct net *net);
622void free_fib_info(struct fib_info *fi);
623
624static inline void fib_info_hold(struct fib_info *fi)
625{
626 refcount_inc(&fi->fib_clntref);
627}
628
629static inline void fib_info_put(struct fib_info *fi)
630{
631 if (refcount_dec_and_test(&fi->fib_clntref))
632 free_fib_info(fi);
633}
634
635#ifdef CONFIG_PROC_FS
636int __net_init fib_proc_init(struct net *net);
637void __net_exit fib_proc_exit(struct net *net);
638#else
639static inline int fib_proc_init(struct net *net)
640{
641 return 0;
642}
643static inline void fib_proc_exit(struct net *net)
644{
645}
646#endif
647
648u32 ip_mtu_from_fib_result(struct fib_result *res, __be32 daddr);
649
650int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
651 struct fib_dump_filter *filter,
652 struct netlink_callback *cb);
653
654int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nh,
655 u8 rt_family, unsigned char *flags, bool skip_oif);
656int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nh,
657 int nh_weight, u8 rt_family, u32 nh_tclassid);
658#endif /* _NET_FIB_H */
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Forwarding Information Base.
8 *
9 * Authors: A.N.Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 */
11
12#ifndef _NET_IP_FIB_H
13#define _NET_IP_FIB_H
14
15#include <net/flow.h>
16#include <linux/seq_file.h>
17#include <linux/rcupdate.h>
18#include <net/fib_notifier.h>
19#include <net/fib_rules.h>
20#include <net/inet_dscp.h>
21#include <net/inetpeer.h>
22#include <linux/percpu.h>
23#include <linux/notifier.h>
24#include <linux/refcount.h>
25
26struct fib_config {
27 u8 fc_dst_len;
28 dscp_t fc_dscp;
29 u8 fc_protocol;
30 u8 fc_scope;
31 u8 fc_type;
32 u8 fc_gw_family;
33 /* 2 bytes unused */
34 u32 fc_table;
35 __be32 fc_dst;
36 union {
37 __be32 fc_gw4;
38 struct in6_addr fc_gw6;
39 };
40 int fc_oif;
41 u32 fc_flags;
42 u32 fc_priority;
43 __be32 fc_prefsrc;
44 u32 fc_nh_id;
45 struct nlattr *fc_mx;
46 struct rtnexthop *fc_mp;
47 int fc_mx_len;
48 int fc_mp_len;
49 u32 fc_flow;
50 u32 fc_nlflags;
51 struct nl_info fc_nlinfo;
52 struct nlattr *fc_encap;
53 u16 fc_encap_type;
54};
55
56struct fib_info;
57struct rtable;
58
59struct fib_nh_exception {
60 struct fib_nh_exception __rcu *fnhe_next;
61 int fnhe_genid;
62 __be32 fnhe_daddr;
63 u32 fnhe_pmtu;
64 bool fnhe_mtu_locked;
65 __be32 fnhe_gw;
66 unsigned long fnhe_expires;
67 struct rtable __rcu *fnhe_rth_input;
68 struct rtable __rcu *fnhe_rth_output;
69 unsigned long fnhe_stamp;
70 struct rcu_head rcu;
71};
72
73struct fnhe_hash_bucket {
74 struct fib_nh_exception __rcu *chain;
75};
76
77#define FNHE_HASH_SHIFT 11
78#define FNHE_HASH_SIZE (1 << FNHE_HASH_SHIFT)
79#define FNHE_RECLAIM_DEPTH 5
80
81struct fib_nh_common {
82 struct net_device *nhc_dev;
83 netdevice_tracker nhc_dev_tracker;
84 int nhc_oif;
85 unsigned char nhc_scope;
86 u8 nhc_family;
87 u8 nhc_gw_family;
88 unsigned char nhc_flags;
89 struct lwtunnel_state *nhc_lwtstate;
90
91 union {
92 __be32 ipv4;
93 struct in6_addr ipv6;
94 } nhc_gw;
95
96 int nhc_weight;
97 atomic_t nhc_upper_bound;
98
99 /* v4 specific, but allows fib6_nh with v4 routes */
100 struct rtable __rcu * __percpu *nhc_pcpu_rth_output;
101 struct rtable __rcu *nhc_rth_input;
102 struct fnhe_hash_bucket __rcu *nhc_exceptions;
103};
104
105struct fib_nh {
106 struct fib_nh_common nh_common;
107 struct hlist_node nh_hash;
108 struct fib_info *nh_parent;
109#ifdef CONFIG_IP_ROUTE_CLASSID
110 __u32 nh_tclassid;
111#endif
112 __be32 nh_saddr;
113 int nh_saddr_genid;
114#define fib_nh_family nh_common.nhc_family
115#define fib_nh_dev nh_common.nhc_dev
116#define fib_nh_dev_tracker nh_common.nhc_dev_tracker
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
132struct nexthop;
133
134struct fib_info {
135 struct hlist_node fib_hash;
136 struct hlist_node fib_lhash;
137 struct list_head nh_list;
138 struct net *fib_net;
139 refcount_t fib_treeref;
140 refcount_t fib_clntref;
141 unsigned int fib_flags;
142 unsigned char fib_dead;
143 unsigned char fib_protocol;
144 unsigned char fib_scope;
145 unsigned char fib_type;
146 __be32 fib_prefsrc;
147 u32 fib_tb_id;
148 u32 fib_priority;
149 struct dst_metrics *fib_metrics;
150#define fib_mtu fib_metrics->metrics[RTAX_MTU-1]
151#define fib_window fib_metrics->metrics[RTAX_WINDOW-1]
152#define fib_rtt fib_metrics->metrics[RTAX_RTT-1]
153#define fib_advmss fib_metrics->metrics[RTAX_ADVMSS-1]
154 int fib_nhs;
155 bool fib_nh_is_v6;
156 bool nh_updated;
157 bool pfsrc_removed;
158 struct nexthop *nh;
159 struct rcu_head rcu;
160 struct fib_nh fib_nh[] __counted_by(fib_nhs);
161};
162
163
164#ifdef CONFIG_IP_MULTIPLE_TABLES
165struct fib_rule;
166#endif
167
168struct fib_table;
169struct fib_result {
170 __be32 prefix;
171 unsigned char prefixlen;
172 unsigned char nh_sel;
173 unsigned char type;
174 unsigned char scope;
175 u32 tclassid;
176 struct fib_nh_common *nhc;
177 struct fib_info *fi;
178 struct fib_table *table;
179 struct hlist_head *fa_head;
180};
181
182struct fib_result_nl {
183 __be32 fl_addr; /* To be looked up*/
184 u32 fl_mark;
185 unsigned char fl_tos;
186 unsigned char fl_scope;
187 unsigned char tb_id_in;
188
189 unsigned char tb_id; /* Results */
190 unsigned char prefixlen;
191 unsigned char nh_sel;
192 unsigned char type;
193 unsigned char scope;
194 int err;
195};
196
197#ifdef CONFIG_IP_MULTIPLE_TABLES
198#define FIB_TABLE_HASHSZ 256
199#else
200#define FIB_TABLE_HASHSZ 2
201#endif
202
203__be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
204 unsigned char scope);
205__be32 fib_result_prefsrc(struct net *net, struct fib_result *res);
206
207#define FIB_RES_NHC(res) ((res).nhc)
208#define FIB_RES_DEV(res) (FIB_RES_NHC(res)->nhc_dev)
209#define FIB_RES_OIF(res) (FIB_RES_NHC(res)->nhc_oif)
210
211struct fib_rt_info {
212 struct fib_info *fi;
213 u32 tb_id;
214 __be32 dst;
215 int dst_len;
216 dscp_t dscp;
217 u8 type;
218 u8 offload:1,
219 trap:1,
220 offload_failed:1,
221 unused:5;
222};
223
224struct fib_entry_notifier_info {
225 struct fib_notifier_info info; /* must be first */
226 u32 dst;
227 int dst_len;
228 struct fib_info *fi;
229 dscp_t dscp;
230 u8 type;
231 u32 tb_id;
232};
233
234struct fib_nh_notifier_info {
235 struct fib_notifier_info info; /* must be first */
236 struct fib_nh *fib_nh;
237};
238
239int call_fib4_notifier(struct notifier_block *nb,
240 enum fib_event_type event_type,
241 struct fib_notifier_info *info);
242int call_fib4_notifiers(struct net *net, enum fib_event_type event_type,
243 struct fib_notifier_info *info);
244
245int __net_init fib4_notifier_init(struct net *net);
246void __net_exit fib4_notifier_exit(struct net *net);
247
248void fib_info_notify_update(struct net *net, struct nl_info *info);
249int fib_notify(struct net *net, struct notifier_block *nb,
250 struct netlink_ext_ack *extack);
251
252struct fib_table {
253 struct hlist_node tb_hlist;
254 u32 tb_id;
255 int tb_num_default;
256 struct rcu_head rcu;
257 unsigned long *tb_data;
258 unsigned long __data[];
259};
260
261struct fib_dump_filter {
262 u32 table_id;
263 /* filter_set is an optimization that an entry is set */
264 bool filter_set;
265 bool dump_routes;
266 bool dump_exceptions;
267 bool rtnl_held;
268 unsigned char protocol;
269 unsigned char rt_type;
270 unsigned int flags;
271 struct net_device *dev;
272};
273
274int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
275 struct fib_result *res, int fib_flags);
276int fib_table_insert(struct net *, struct fib_table *, struct fib_config *,
277 struct netlink_ext_ack *extack);
278int fib_table_delete(struct net *, struct fib_table *, struct fib_config *,
279 struct netlink_ext_ack *extack);
280int fib_table_dump(struct fib_table *table, struct sk_buff *skb,
281 struct netlink_callback *cb, struct fib_dump_filter *filter);
282int fib_table_flush(struct net *net, struct fib_table *table, bool flush_all);
283struct fib_table *fib_trie_unmerge(struct fib_table *main_tb);
284void fib_table_flush_external(struct fib_table *table);
285void fib_free_table(struct fib_table *tb);
286
287#ifndef CONFIG_IP_MULTIPLE_TABLES
288
289#define TABLE_LOCAL_INDEX (RT_TABLE_LOCAL & (FIB_TABLE_HASHSZ - 1))
290#define TABLE_MAIN_INDEX (RT_TABLE_MAIN & (FIB_TABLE_HASHSZ - 1))
291
292static inline struct fib_table *fib_get_table(struct net *net, u32 id)
293{
294 struct hlist_node *tb_hlist;
295 struct hlist_head *ptr;
296
297 ptr = id == RT_TABLE_LOCAL ?
298 &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX] :
299 &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX];
300
301 tb_hlist = rcu_dereference_rtnl(hlist_first_rcu(ptr));
302
303 return hlist_entry(tb_hlist, struct fib_table, tb_hlist);
304}
305
306static inline struct fib_table *fib_new_table(struct net *net, u32 id)
307{
308 return fib_get_table(net, id);
309}
310
311static inline int fib_lookup(struct net *net, const struct flowi4 *flp,
312 struct fib_result *res, unsigned int flags)
313{
314 struct fib_table *tb;
315 int err = -ENETUNREACH;
316
317 rcu_read_lock();
318
319 tb = fib_get_table(net, RT_TABLE_MAIN);
320 if (tb)
321 err = fib_table_lookup(tb, flp, res, flags | FIB_LOOKUP_NOREF);
322
323 if (err == -EAGAIN)
324 err = -ENETUNREACH;
325
326 rcu_read_unlock();
327
328 return err;
329}
330
331static inline bool fib4_has_custom_rules(const struct net *net)
332{
333 return false;
334}
335
336static inline bool fib4_rule_default(const struct fib_rule *rule)
337{
338 return true;
339}
340
341static inline int fib4_rules_dump(struct net *net, struct notifier_block *nb,
342 struct netlink_ext_ack *extack)
343{
344 return 0;
345}
346
347static inline unsigned int fib4_rules_seq_read(struct net *net)
348{
349 return 0;
350}
351
352static inline bool fib4_rules_early_flow_dissect(struct net *net,
353 struct sk_buff *skb,
354 struct flowi4 *fl4,
355 struct flow_keys *flkeys)
356{
357 return false;
358}
359#else /* CONFIG_IP_MULTIPLE_TABLES */
360int __net_init fib4_rules_init(struct net *net);
361void __net_exit fib4_rules_exit(struct net *net);
362
363struct fib_table *fib_new_table(struct net *net, u32 id);
364struct fib_table *fib_get_table(struct net *net, u32 id);
365
366int __fib_lookup(struct net *net, struct flowi4 *flp,
367 struct fib_result *res, unsigned int flags);
368
369static inline int fib_lookup(struct net *net, struct flowi4 *flp,
370 struct fib_result *res, unsigned int flags)
371{
372 struct fib_table *tb;
373 int err = -ENETUNREACH;
374
375 flags |= FIB_LOOKUP_NOREF;
376 if (net->ipv4.fib_has_custom_rules)
377 return __fib_lookup(net, flp, res, flags);
378
379 rcu_read_lock();
380
381 res->tclassid = 0;
382
383 tb = rcu_dereference_rtnl(net->ipv4.fib_main);
384 if (tb)
385 err = fib_table_lookup(tb, flp, res, flags);
386
387 if (!err)
388 goto out;
389
390 tb = rcu_dereference_rtnl(net->ipv4.fib_default);
391 if (tb)
392 err = fib_table_lookup(tb, flp, res, flags);
393
394out:
395 if (err == -EAGAIN)
396 err = -ENETUNREACH;
397
398 rcu_read_unlock();
399
400 return err;
401}
402
403static inline bool fib4_has_custom_rules(const struct net *net)
404{
405 return net->ipv4.fib_has_custom_rules;
406}
407
408bool fib4_rule_default(const struct fib_rule *rule);
409int fib4_rules_dump(struct net *net, struct notifier_block *nb,
410 struct netlink_ext_ack *extack);
411unsigned int fib4_rules_seq_read(struct net *net);
412
413static inline bool fib4_rules_early_flow_dissect(struct net *net,
414 struct sk_buff *skb,
415 struct flowi4 *fl4,
416 struct flow_keys *flkeys)
417{
418 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
419
420 if (!net->ipv4.fib_rules_require_fldissect)
421 return false;
422
423 memset(flkeys, 0, sizeof(*flkeys));
424 __skb_flow_dissect(net, skb, &flow_keys_dissector,
425 flkeys, NULL, 0, 0, 0, flag);
426
427 fl4->fl4_sport = flkeys->ports.src;
428 fl4->fl4_dport = flkeys->ports.dst;
429 fl4->flowi4_proto = flkeys->basic.ip_proto;
430
431 return true;
432}
433
434#endif /* CONFIG_IP_MULTIPLE_TABLES */
435
436/* Exported by fib_frontend.c */
437extern const struct nla_policy rtm_ipv4_policy[];
438void ip_fib_init(void);
439int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
440 struct netlink_ext_ack *extack);
441__be32 fib_compute_spec_dst(struct sk_buff *skb);
442bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev);
443int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
444 u8 tos, int oif, struct net_device *dev,
445 struct in_device *idev, u32 *itag);
446#ifdef CONFIG_IP_ROUTE_CLASSID
447static inline int fib_num_tclassid_users(struct net *net)
448{
449 return atomic_read(&net->ipv4.fib_num_tclassid_users);
450}
451#else
452static inline int fib_num_tclassid_users(struct net *net)
453{
454 return 0;
455}
456#endif
457int fib_unmerge(struct net *net);
458
459static inline bool nhc_l3mdev_matches_dev(const struct fib_nh_common *nhc,
460const struct net_device *dev)
461{
462 if (nhc->nhc_dev == dev ||
463 l3mdev_master_ifindex_rcu(nhc->nhc_dev) == dev->ifindex)
464 return true;
465
466 return false;
467}
468
469/* Exported by fib_semantics.c */
470int ip_fib_check_default(__be32 gw, struct net_device *dev);
471int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force);
472int fib_sync_down_addr(struct net_device *dev, __be32 local);
473int fib_sync_up(struct net_device *dev, unsigned char nh_flags);
474void fib_sync_mtu(struct net_device *dev, u32 orig_mtu);
475void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig);
476
477/* Fields used for sysctl_fib_multipath_hash_fields.
478 * Common to IPv4 and IPv6.
479 *
480 * Add new fields at the end. This is user API.
481 */
482#define FIB_MULTIPATH_HASH_FIELD_SRC_IP BIT(0)
483#define FIB_MULTIPATH_HASH_FIELD_DST_IP BIT(1)
484#define FIB_MULTIPATH_HASH_FIELD_IP_PROTO BIT(2)
485#define FIB_MULTIPATH_HASH_FIELD_FLOWLABEL BIT(3)
486#define FIB_MULTIPATH_HASH_FIELD_SRC_PORT BIT(4)
487#define FIB_MULTIPATH_HASH_FIELD_DST_PORT BIT(5)
488#define FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP BIT(6)
489#define FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP BIT(7)
490#define FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO BIT(8)
491#define FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL BIT(9)
492#define FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT BIT(10)
493#define FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT BIT(11)
494
495#define FIB_MULTIPATH_HASH_FIELD_OUTER_MASK \
496 (FIB_MULTIPATH_HASH_FIELD_SRC_IP | \
497 FIB_MULTIPATH_HASH_FIELD_DST_IP | \
498 FIB_MULTIPATH_HASH_FIELD_IP_PROTO | \
499 FIB_MULTIPATH_HASH_FIELD_FLOWLABEL | \
500 FIB_MULTIPATH_HASH_FIELD_SRC_PORT | \
501 FIB_MULTIPATH_HASH_FIELD_DST_PORT)
502
503#define FIB_MULTIPATH_HASH_FIELD_INNER_MASK \
504 (FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP | \
505 FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP | \
506 FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO | \
507 FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL | \
508 FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT | \
509 FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
510
511#define FIB_MULTIPATH_HASH_FIELD_ALL_MASK \
512 (FIB_MULTIPATH_HASH_FIELD_OUTER_MASK | \
513 FIB_MULTIPATH_HASH_FIELD_INNER_MASK)
514
515#define FIB_MULTIPATH_HASH_FIELD_DEFAULT_MASK \
516 (FIB_MULTIPATH_HASH_FIELD_SRC_IP | \
517 FIB_MULTIPATH_HASH_FIELD_DST_IP | \
518 FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
519
520#ifdef CONFIG_IP_ROUTE_MULTIPATH
521int fib_multipath_hash(const struct net *net, const struct flowi4 *fl4,
522 const struct sk_buff *skb, struct flow_keys *flkeys);
523#endif
524int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
525 struct netlink_ext_ack *extack);
526void fib_select_multipath(struct fib_result *res, int hash);
527void fib_select_path(struct net *net, struct fib_result *res,
528 struct flowi4 *fl4, const struct sk_buff *skb);
529
530int fib_nh_init(struct net *net, struct fib_nh *fib_nh,
531 struct fib_config *cfg, int nh_weight,
532 struct netlink_ext_ack *extack);
533void fib_nh_release(struct net *net, struct fib_nh *fib_nh);
534int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
535 struct nlattr *fc_encap, u16 fc_encap_type,
536 void *cfg, gfp_t gfp_flags,
537 struct netlink_ext_ack *extack);
538void fib_nh_common_release(struct fib_nh_common *nhc);
539
540/* Exported by fib_trie.c */
541void fib_alias_hw_flags_set(struct net *net, const struct fib_rt_info *fri);
542void fib_trie_init(void);
543struct fib_table *fib_trie_table(u32 id, struct fib_table *alias);
544bool fib_lookup_good_nhc(const struct fib_nh_common *nhc, int fib_flags,
545 const struct flowi4 *flp);
546
547static inline void fib_combine_itag(u32 *itag, const struct fib_result *res)
548{
549#ifdef CONFIG_IP_ROUTE_CLASSID
550 struct fib_nh_common *nhc = res->nhc;
551#ifdef CONFIG_IP_MULTIPLE_TABLES
552 u32 rtag;
553#endif
554 if (nhc->nhc_family == AF_INET) {
555 struct fib_nh *nh;
556
557 nh = container_of(nhc, struct fib_nh, nh_common);
558 *itag = nh->nh_tclassid << 16;
559 } else {
560 *itag = 0;
561 }
562
563#ifdef CONFIG_IP_MULTIPLE_TABLES
564 rtag = res->tclassid;
565 if (*itag == 0)
566 *itag = (rtag<<16);
567 *itag |= (rtag>>16);
568#endif
569#endif
570}
571
572void fib_flush(struct net *net);
573void free_fib_info(struct fib_info *fi);
574
575static inline void fib_info_hold(struct fib_info *fi)
576{
577 refcount_inc(&fi->fib_clntref);
578}
579
580static inline void fib_info_put(struct fib_info *fi)
581{
582 if (refcount_dec_and_test(&fi->fib_clntref))
583 free_fib_info(fi);
584}
585
586#ifdef CONFIG_PROC_FS
587int __net_init fib_proc_init(struct net *net);
588void __net_exit fib_proc_exit(struct net *net);
589#else
590static inline int fib_proc_init(struct net *net)
591{
592 return 0;
593}
594static inline void fib_proc_exit(struct net *net)
595{
596}
597#endif
598
599u32 ip_mtu_from_fib_result(struct fib_result *res, __be32 daddr);
600
601int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
602 struct fib_dump_filter *filter,
603 struct netlink_callback *cb);
604
605int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nh,
606 u8 rt_family, unsigned char *flags, bool skip_oif);
607int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nh,
608 int nh_weight, u8 rt_family, u32 nh_tclassid);
609#endif /* _NET_FIB_H */