1/*
2 * Linux INET6 implementation
3 *
4 * Authors:
5 * Pedro Roque <roque@di.fc.ul.pt>
6 *
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.
11 */
12
13#ifndef _IP6_FIB_H
14#define _IP6_FIB_H
15
16#include <linux/ipv6_route.h>
17#include <linux/rtnetlink.h>
18#include <linux/spinlock.h>
19#include <linux/notifier.h>
20#include <net/dst.h>
21#include <net/flow.h>
22#include <net/netlink.h>
23#include <net/inetpeer.h>
24#include <net/fib_notifier.h>
25
26#ifdef CONFIG_IPV6_MULTIPLE_TABLES
27#define FIB6_TABLE_HASHSZ 256
28#else
29#define FIB6_TABLE_HASHSZ 1
30#endif
31
32#define RT6_DEBUG 2
33
34#if RT6_DEBUG >= 3
35#define RT6_TRACE(x...) pr_debug(x)
36#else
37#define RT6_TRACE(x...) do { ; } while (0)
38#endif
39
40struct rt6_info;
41struct fib6_info;
42
43struct fib6_config {
44 u32 fc_table;
45 u32 fc_metric;
46 int fc_dst_len;
47 int fc_src_len;
48 int fc_ifindex;
49 u32 fc_flags;
50 u32 fc_protocol;
51 u16 fc_type; /* only 8 bits are used */
52 u16 fc_delete_all_nh : 1,
53 __unused : 15;
54
55 struct in6_addr fc_dst;
56 struct in6_addr fc_src;
57 struct in6_addr fc_prefsrc;
58 struct in6_addr fc_gateway;
59
60 unsigned long fc_expires;
61 struct nlattr *fc_mx;
62 int fc_mx_len;
63 int fc_mp_len;
64 struct nlattr *fc_mp;
65
66 struct nl_info fc_nlinfo;
67 struct nlattr *fc_encap;
68 u16 fc_encap_type;
69};
70
71struct fib6_node {
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;
77#endif
78 struct fib6_info __rcu *leaf;
79
80 __u16 fn_bit; /* bit key */
81 __u16 fn_flags;
82 int fn_sernum;
83 struct fib6_info __rcu *rr_ptr;
84 struct rcu_head rcu;
85};
86
87struct fib6_gc_args {
88 int timeout;
89 int more;
90};
91
92#ifndef CONFIG_IPV6_SUBTREES
93#define FIB6_SUBTREE(fn) NULL
94#else
95#define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1))
96#endif
97
98/*
99 * routing information
100 *
101 */
102
103struct rt6key {
104 struct in6_addr addr;
105 int plen;
106};
107
108struct fib6_table;
109
110struct rt6_exception_bucket {
111 struct hlist_head chain;
112 int depth;
113};
114
115struct rt6_exception {
116 struct hlist_node hlist;
117 struct rt6_info *rt6i;
118 unsigned long stamp;
119 struct rcu_head rcu;
120};
121
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
125
126struct fib6_nh {
127 struct in6_addr nh_gw;
128 struct net_device *nh_dev;
129 struct lwtunnel_state *nh_lwtstate;
130
131 unsigned int nh_flags;
132 atomic_t nh_upper_bound;
133 int nh_weight;
134};
135
136struct 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 struct list_head fib6_siblings;
147 unsigned int fib6_nsiblings;
148
149 atomic_t fib6_ref;
150 unsigned long expires;
151 struct dst_metrics *fib6_metrics;
152#define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1]
153
154 struct rt6key fib6_dst;
155 u32 fib6_flags;
156 struct rt6key fib6_src;
157 struct rt6key fib6_prefsrc;
158
159 struct rt6_info * __percpu *rt6i_pcpu;
160 struct rt6_exception_bucket __rcu *rt6i_exception_bucket;
161
162#ifdef CONFIG_IPV6_ROUTER_PREF
163 unsigned long last_probe;
164#endif
165
166 u32 fib6_metric;
167 u8 fib6_protocol;
168 u8 fib6_type;
169 u8 exception_bucket_flushed:1,
170 should_flush:1,
171 dst_nocount:1,
172 dst_nopolicy:1,
173 dst_host:1,
174 unused:3;
175
176 struct fib6_nh fib6_nh;
177 struct rcu_head rcu;
178};
179
180struct rt6_info {
181 struct dst_entry dst;
182 struct fib6_info __rcu *from;
183
184 struct rt6key rt6i_dst;
185 struct rt6key rt6i_src;
186 struct in6_addr rt6i_gateway;
187 struct inet6_dev *rt6i_idev;
188 u32 rt6i_flags;
189
190 struct list_head rt6i_uncached;
191 struct uncached_list *rt6i_uncached_list;
192
193 /* more non-fragment space at head required */
194 unsigned short rt6i_nfheader_len;
195};
196
197#define for_each_fib6_node_rt_rcu(fn) \
198 for (rt = rcu_dereference((fn)->leaf); rt; \
199 rt = rcu_dereference(rt->fib6_next))
200
201#define for_each_fib6_walker_rt(w) \
202 for (rt = (w)->leaf; rt; \
203 rt = rcu_dereference_protected(rt->fib6_next, 1))
204
205static inline struct inet6_dev *ip6_dst_idev(struct dst_entry *dst)
206{
207 return ((struct rt6_info *)dst)->rt6i_idev;
208}
209
210static inline void fib6_clean_expires(struct fib6_info *f6i)
211{
212 f6i->fib6_flags &= ~RTF_EXPIRES;
213 f6i->expires = 0;
214}
215
216static inline void fib6_set_expires(struct fib6_info *f6i,
217 unsigned long expires)
218{
219 f6i->expires = expires;
220 f6i->fib6_flags |= RTF_EXPIRES;
221}
222
223static inline bool fib6_check_expired(const struct fib6_info *f6i)
224{
225 if (f6i->fib6_flags & RTF_EXPIRES)
226 return time_after(jiffies, f6i->expires);
227 return false;
228}
229
230/* Function to safely get fn->sernum for passed in rt
231 * and store result in passed in cookie.
232 * Return true if we can get cookie safely
233 * Return false if not
234 */
235static inline bool fib6_get_cookie_safe(const struct fib6_info *f6i,
236 u32 *cookie)
237{
238 struct fib6_node *fn;
239 bool status = false;
240
241 fn = rcu_dereference(f6i->fib6_node);
242
243 if (fn) {
244 *cookie = fn->fn_sernum;
245 /* pairs with smp_wmb() in fib6_update_sernum_upto_root() */
246 smp_rmb();
247 status = true;
248 }
249
250 return status;
251}
252
253static inline u32 rt6_get_cookie(const struct rt6_info *rt)
254{
255 struct fib6_info *from;
256 u32 cookie = 0;
257
258 rcu_read_lock();
259
260 from = rcu_dereference(rt->from);
261 if (from && (rt->rt6i_flags & RTF_PCPU ||
262 unlikely(!list_empty(&rt->rt6i_uncached))))
263 fib6_get_cookie_safe(from, &cookie);
264
265 rcu_read_unlock();
266
267 return cookie;
268}
269
270static inline void ip6_rt_put(struct rt6_info *rt)
271{
272 /* dst_release() accepts a NULL parameter.
273 * We rely on dst being first structure in struct rt6_info
274 */
275 BUILD_BUG_ON(offsetof(struct rt6_info, dst) != 0);
276 dst_release(&rt->dst);
277}
278
279struct fib6_info *fib6_info_alloc(gfp_t gfp_flags);
280void fib6_info_destroy_rcu(struct rcu_head *head);
281
282static inline void fib6_info_hold(struct fib6_info *f6i)
283{
284 atomic_inc(&f6i->fib6_ref);
285}
286
287static inline bool fib6_info_hold_safe(struct fib6_info *f6i)
288{
289 return atomic_inc_not_zero(&f6i->fib6_ref);
290}
291
292static inline void fib6_info_release(struct fib6_info *f6i)
293{
294 if (f6i && atomic_dec_and_test(&f6i->fib6_ref))
295 call_rcu(&f6i->rcu, fib6_info_destroy_rcu);
296}
297
298enum fib6_walk_state {
299#ifdef CONFIG_IPV6_SUBTREES
300 FWS_S,
301#endif
302 FWS_L,
303 FWS_R,
304 FWS_C,
305 FWS_U
306};
307
308struct fib6_walker {
309 struct list_head lh;
310 struct fib6_node *root, *node;
311 struct fib6_info *leaf;
312 enum fib6_walk_state state;
313 unsigned int skip;
314 unsigned int count;
315 int (*func)(struct fib6_walker *);
316 void *args;
317};
318
319struct rt6_statistics {
320 __u32 fib_nodes; /* all fib6 nodes */
321 __u32 fib_route_nodes; /* intermediate nodes */
322 __u32 fib_rt_entries; /* rt entries in fib table */
323 __u32 fib_rt_cache; /* cached rt entries in exception table */
324 __u32 fib_discarded_routes; /* total number of routes delete */
325
326 /* The following stats are not protected by any lock */
327 atomic_t fib_rt_alloc; /* total number of routes alloced */
328 atomic_t fib_rt_uncache; /* rt entries in uncached list */
329};
330
331#define RTN_TL_ROOT 0x0001
332#define RTN_ROOT 0x0002 /* tree root node */
333#define RTN_RTINFO 0x0004 /* node with valid routing info */
334
335/*
336 * priority levels (or metrics)
337 *
338 */
339
340
341struct fib6_table {
342 struct hlist_node tb6_hlist;
343 u32 tb6_id;
344 spinlock_t tb6_lock;
345 struct fib6_node tb6_root;
346 struct inet_peer_base tb6_peers;
347 unsigned int flags;
348 unsigned int fib_seq;
349#define RT6_TABLE_HAS_DFLT_ROUTER BIT(0)
350};
351
352#define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC
353#define RT6_TABLE_MAIN RT_TABLE_MAIN
354#define RT6_TABLE_DFLT RT6_TABLE_MAIN
355#define RT6_TABLE_INFO RT6_TABLE_MAIN
356#define RT6_TABLE_PREFIX RT6_TABLE_MAIN
357
358#ifdef CONFIG_IPV6_MULTIPLE_TABLES
359#define FIB6_TABLE_MIN 1
360#define FIB6_TABLE_MAX RT_TABLE_MAX
361#define RT6_TABLE_LOCAL RT_TABLE_LOCAL
362#else
363#define FIB6_TABLE_MIN RT_TABLE_MAIN
364#define FIB6_TABLE_MAX FIB6_TABLE_MIN
365#define RT6_TABLE_LOCAL RT6_TABLE_MAIN
366#endif
367
368typedef struct rt6_info *(*pol_lookup_t)(struct net *,
369 struct fib6_table *,
370 struct flowi6 *,
371 const struct sk_buff *, int);
372
373struct fib6_entry_notifier_info {
374 struct fib_notifier_info info; /* must be first */
375 struct fib6_info *rt;
376};
377
378/*
379 * exported functions
380 */
381
382struct fib6_table *fib6_get_table(struct net *net, u32 id);
383struct fib6_table *fib6_new_table(struct net *net, u32 id);
384struct dst_entry *fib6_rule_lookup(struct net *net, struct flowi6 *fl6,
385 const struct sk_buff *skb,
386 int flags, pol_lookup_t lookup);
387
388/* called with rcu lock held; can return error pointer
389 * caller needs to select path
390 */
391struct fib6_info *fib6_lookup(struct net *net, int oif, struct flowi6 *fl6,
392 int flags);
393
394/* called with rcu lock held; caller needs to select path */
395struct fib6_info *fib6_table_lookup(struct net *net, struct fib6_table *table,
396 int oif, struct flowi6 *fl6, int strict);
397
398struct fib6_info *fib6_multipath_select(const struct net *net,
399 struct fib6_info *match,
400 struct flowi6 *fl6, int oif,
401 const struct sk_buff *skb, int strict);
402
403struct fib6_node *fib6_node_lookup(struct fib6_node *root,
404 const struct in6_addr *daddr,
405 const struct in6_addr *saddr);
406
407struct fib6_node *fib6_locate(struct fib6_node *root,
408 const struct in6_addr *daddr, int dst_len,
409 const struct in6_addr *saddr, int src_len,
410 bool exact_match);
411
412void fib6_clean_all(struct net *net, int (*func)(struct fib6_info *, void *arg),
413 void *arg);
414void fib6_clean_all_skip_notify(struct net *net,
415 int (*func)(struct fib6_info *, void *arg),
416 void *arg);
417
418int fib6_add(struct fib6_node *root, struct fib6_info *rt,
419 struct nl_info *info, struct netlink_ext_ack *extack);
420int fib6_del(struct fib6_info *rt, struct nl_info *info);
421
422static inline
423void rt6_get_prefsrc(const struct rt6_info *rt, struct in6_addr *addr)
424{
425 const struct fib6_info *from;
426
427 rcu_read_lock();
428
429 from = rcu_dereference(rt->from);
430 if (from) {
431 *addr = from->fib6_prefsrc.addr;
432 } else {
433 struct in6_addr in6_zero = {};
434
435 *addr = in6_zero;
436 }
437
438 rcu_read_unlock();
439}
440
441static inline struct net_device *fib6_info_nh_dev(const struct fib6_info *f6i)
442{
443 return f6i->fib6_nh.nh_dev;
444}
445
446static inline
447struct lwtunnel_state *fib6_info_nh_lwt(const struct fib6_info *f6i)
448{
449 return f6i->fib6_nh.nh_lwtstate;
450}
451
452void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
453 unsigned int flags);
454
455void fib6_run_gc(unsigned long expires, struct net *net, bool force);
456
457void fib6_gc_cleanup(void);
458
459int fib6_init(void);
460
461struct ipv6_route_iter {
462 struct seq_net_private p;
463 struct fib6_walker w;
464 loff_t skip;
465 struct fib6_table *tbl;
466 int sernum;
467};
468
469extern const struct seq_operations ipv6_route_seq_ops;
470
471int call_fib6_notifier(struct notifier_block *nb, struct net *net,
472 enum fib_event_type event_type,
473 struct fib_notifier_info *info);
474int call_fib6_notifiers(struct net *net, enum fib_event_type event_type,
475 struct fib_notifier_info *info);
476
477int __net_init fib6_notifier_init(struct net *net);
478void __net_exit fib6_notifier_exit(struct net *net);
479
480unsigned int fib6_tables_seq_read(struct net *net);
481int fib6_tables_dump(struct net *net, struct notifier_block *nb);
482
483void fib6_update_sernum(struct net *net, struct fib6_info *rt);
484void fib6_update_sernum_upto_root(struct net *net, struct fib6_info *rt);
485
486void fib6_metric_set(struct fib6_info *f6i, int metric, u32 val);
487static inline bool fib6_metric_locked(struct fib6_info *f6i, int metric)
488{
489 return !!(f6i->fib6_metrics->metrics[RTAX_LOCK - 1] & (1 << metric));
490}
491
492#ifdef CONFIG_IPV6_MULTIPLE_TABLES
493int fib6_rules_init(void);
494void fib6_rules_cleanup(void);
495bool fib6_rule_default(const struct fib_rule *rule);
496int fib6_rules_dump(struct net *net, struct notifier_block *nb);
497unsigned int fib6_rules_seq_read(struct net *net);
498
499static inline bool fib6_rules_early_flow_dissect(struct net *net,
500 struct sk_buff *skb,
501 struct flowi6 *fl6,
502 struct flow_keys *flkeys)
503{
504 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
505
506 if (!net->ipv6.fib6_rules_require_fldissect)
507 return false;
508
509 skb_flow_dissect_flow_keys(skb, flkeys, flag);
510 fl6->fl6_sport = flkeys->ports.src;
511 fl6->fl6_dport = flkeys->ports.dst;
512 fl6->flowi6_proto = flkeys->basic.ip_proto;
513
514 return true;
515}
516#else
517static inline int fib6_rules_init(void)
518{
519 return 0;
520}
521static inline void fib6_rules_cleanup(void)
522{
523 return ;
524}
525static inline bool fib6_rule_default(const struct fib_rule *rule)
526{
527 return true;
528}
529static inline int fib6_rules_dump(struct net *net, struct notifier_block *nb)
530{
531 return 0;
532}
533static inline unsigned int fib6_rules_seq_read(struct net *net)
534{
535 return 0;
536}
537static inline bool fib6_rules_early_flow_dissect(struct net *net,
538 struct sk_buff *skb,
539 struct flowi6 *fl6,
540 struct flow_keys *flkeys)
541{
542 return false;
543}
544#endif
545#endif
546