1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
12 #include <linux/ipv6_route.h>
13 #include <linux/rtnetlink.h>
14 #include <linux/spinlock.h>
15 #include <linux/notifier.h>
18 #include <net/ip_fib.h>
19 #include <net/netlink.h>
20 #include <net/inetpeer.h>
21 #include <net/fib_notifier.h>
22 #include <linux/indirect_call_wrapper.h>
23 #include <uapi/linux/bpf.h>
25 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
26 #define FIB6_TABLE_HASHSZ 256
28 #define FIB6_TABLE_HASHSZ 1
44 u16 fc_type
; /* only 8 bits are used */
45 u16 fc_delete_all_nh
: 1,
50 struct in6_addr fc_dst
;
51 struct in6_addr fc_src
;
52 struct in6_addr fc_prefsrc
;
53 struct in6_addr fc_gateway
;
55 unsigned long fc_expires
;
61 struct nl_info fc_nlinfo
;
62 struct nlattr
*fc_encap
;
68 struct fib6_node __rcu
*parent
;
69 struct fib6_node __rcu
*left
;
70 struct fib6_node __rcu
*right
;
71 #ifdef CONFIG_IPV6_SUBTREES
72 struct fib6_node __rcu
*subtree
;
74 struct fib6_info __rcu
*leaf
;
76 __u16 fn_bit
; /* bit key */
79 struct fib6_info __rcu
*rr_ptr
;
88 #ifndef CONFIG_IPV6_SUBTREES
89 #define FIB6_SUBTREE(fn) NULL
91 static inline bool fib6_routes_require_src(const struct net
*net
)
96 static inline void fib6_routes_require_src_inc(struct net
*net
) {}
97 static inline void fib6_routes_require_src_dec(struct net
*net
) {}
101 static inline bool fib6_routes_require_src(const struct net
*net
)
103 return net
->ipv6
.fib6_routes_require_src
> 0;
106 static inline void fib6_routes_require_src_inc(struct net
*net
)
108 net
->ipv6
.fib6_routes_require_src
++;
111 static inline void fib6_routes_require_src_dec(struct net
*net
)
113 net
->ipv6
.fib6_routes_require_src
--;
116 #define FIB6_SUBTREE(fn) (rcu_dereference_protected((fn)->subtree, 1))
120 * routing information
125 struct in6_addr addr
;
131 struct rt6_exception_bucket
{
132 struct hlist_head chain
;
136 struct rt6_exception
{
137 struct hlist_node hlist
;
138 struct rt6_info
*rt6i
;
143 #define FIB6_EXCEPTION_BUCKET_SIZE_SHIFT 10
144 #define FIB6_EXCEPTION_BUCKET_SIZE (1 << FIB6_EXCEPTION_BUCKET_SIZE_SHIFT)
145 #define FIB6_MAX_DEPTH 5
148 struct fib_nh_common nh_common
;
150 #ifdef CONFIG_IPV6_ROUTER_PREF
151 unsigned long last_probe
;
154 struct rt6_info
* __percpu
*rt6i_pcpu
;
155 struct rt6_exception_bucket __rcu
*rt6i_exception_bucket
;
159 struct fib6_table
*fib6_table
;
160 struct fib6_info __rcu
*fib6_next
;
161 struct fib6_node __rcu
*fib6_node
;
164 * siblings is a list of fib6_info that have the same metric/weight,
165 * destination, but not the same gateway. nsiblings is just a cache
166 * to speed up lookup.
169 struct list_head fib6_siblings
;
170 struct list_head nh_list
;
172 unsigned int fib6_nsiblings
;
175 unsigned long expires
;
177 struct hlist_node gc_link
;
179 struct dst_metrics
*fib6_metrics
;
180 #define fib6_pmtu fib6_metrics->metrics[RTAX_MTU-1]
182 struct rt6key fib6_dst
;
184 struct rt6key fib6_src
;
185 struct rt6key fib6_prefsrc
;
203 struct fib6_nh fib6_nh
[];
207 struct dst_entry dst
;
208 struct fib6_info __rcu
*from
;
211 struct rt6key rt6i_dst
;
212 struct rt6key rt6i_src
;
213 struct in6_addr rt6i_gateway
;
214 struct inet6_dev
*rt6i_idev
;
217 /* more non-fragment space at head required */
218 unsigned short rt6i_nfheader_len
;
223 struct fib6_info
*f6i
;
226 struct rt6_info
*rt6
;
229 #define for_each_fib6_node_rt_rcu(fn) \
230 for (rt = rcu_dereference((fn)->leaf); rt; \
231 rt = rcu_dereference(rt->fib6_next))
233 #define for_each_fib6_walker_rt(w) \
234 for (rt = (w)->leaf; rt; \
235 rt = rcu_dereference_protected(rt->fib6_next, 1))
237 #define dst_rt6_info(_ptr) container_of_const(_ptr, struct rt6_info, dst)
239 static inline struct inet6_dev
*ip6_dst_idev(const struct dst_entry
*dst
)
241 return dst_rt6_info(dst
)->rt6i_idev
;
244 static inline bool fib6_requires_src(const struct fib6_info
*rt
)
246 return rt
->fib6_src
.plen
> 0;
249 /* The callers should hold f6i->fib6_table->tb6_lock if a route has ever
250 * been added to a table before.
252 static inline void fib6_clean_expires(struct fib6_info
*f6i
)
254 f6i
->fib6_flags
&= ~RTF_EXPIRES
;
258 /* The callers should hold f6i->fib6_table->tb6_lock if a route has ever
259 * been added to a table before.
261 static inline void fib6_set_expires(struct fib6_info
*f6i
,
262 unsigned long expires
)
264 f6i
->expires
= expires
;
265 f6i
->fib6_flags
|= RTF_EXPIRES
;
268 static inline bool fib6_check_expired(const struct fib6_info
*f6i
)
270 if (f6i
->fib6_flags
& RTF_EXPIRES
)
271 return time_after(jiffies
, f6i
->expires
);
275 /* Function to safely get fn->fn_sernum for passed in rt
276 * and store result in passed in cookie.
277 * Return true if we can get cookie safely
278 * Return false if not
280 static inline bool fib6_get_cookie_safe(const struct fib6_info
*f6i
,
283 struct fib6_node
*fn
;
286 fn
= rcu_dereference(f6i
->fib6_node
);
289 *cookie
= READ_ONCE(fn
->fn_sernum
);
290 /* pairs with smp_wmb() in __fib6_update_sernum_upto_root() */
298 static inline u32
rt6_get_cookie(const struct rt6_info
*rt
)
300 struct fib6_info
*from
;
308 from
= rcu_dereference(rt
->from
);
310 fib6_get_cookie_safe(from
, &cookie
);
317 static inline void ip6_rt_put(struct rt6_info
*rt
)
319 /* dst_release() accepts a NULL parameter.
320 * We rely on dst being first structure in struct rt6_info
322 BUILD_BUG_ON(offsetof(struct rt6_info
, dst
) != 0);
323 dst_release(&rt
->dst
);
326 struct fib6_info
*fib6_info_alloc(gfp_t gfp_flags
, bool with_fib6_nh
);
327 void fib6_info_destroy_rcu(struct rcu_head
*head
);
329 static inline void fib6_info_hold(struct fib6_info
*f6i
)
331 refcount_inc(&f6i
->fib6_ref
);
334 static inline bool fib6_info_hold_safe(struct fib6_info
*f6i
)
336 return refcount_inc_not_zero(&f6i
->fib6_ref
);
339 static inline void fib6_info_release(struct fib6_info
*f6i
)
341 if (f6i
&& refcount_dec_and_test(&f6i
->fib6_ref
)) {
342 DEBUG_NET_WARN_ON_ONCE(!hlist_unhashed(&f6i
->gc_link
));
343 call_rcu_hurry(&f6i
->rcu
, fib6_info_destroy_rcu
);
347 enum fib6_walk_state
{
348 #ifdef CONFIG_IPV6_SUBTREES
359 struct fib6_node
*root
, *node
;
360 struct fib6_info
*leaf
;
361 enum fib6_walk_state state
;
364 unsigned int skip_in_node
;
365 int (*func
)(struct fib6_walker
*);
369 struct rt6_statistics
{
370 __u32 fib_nodes
; /* all fib6 nodes */
371 __u32 fib_route_nodes
; /* intermediate nodes */
372 __u32 fib_rt_entries
; /* rt entries in fib table */
373 __u32 fib_rt_cache
; /* cached rt entries in exception table */
374 __u32 fib_discarded_routes
; /* total number of routes delete */
376 /* The following stat is not protected by any lock */
377 atomic_t fib_rt_alloc
; /* total number of routes alloced */
380 #define RTN_TL_ROOT 0x0001
381 #define RTN_ROOT 0x0002 /* tree root node */
382 #define RTN_RTINFO 0x0004 /* node with valid routing info */
385 * priority levels (or metrics)
391 struct hlist_node tb6_hlist
;
394 struct fib6_node tb6_root
;
395 struct inet_peer_base tb6_peers
;
397 unsigned int fib_seq
; /* writes protected by rtnl_mutex */
398 struct hlist_head tb6_gc_hlist
; /* GC candidates */
399 #define RT6_TABLE_HAS_DFLT_ROUTER BIT(0)
402 #define RT6_TABLE_UNSPEC RT_TABLE_UNSPEC
403 #define RT6_TABLE_MAIN RT_TABLE_MAIN
404 #define RT6_TABLE_DFLT RT6_TABLE_MAIN
405 #define RT6_TABLE_INFO RT6_TABLE_MAIN
406 #define RT6_TABLE_PREFIX RT6_TABLE_MAIN
408 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
409 #define FIB6_TABLE_MIN 1
410 #define FIB6_TABLE_MAX RT_TABLE_MAX
411 #define RT6_TABLE_LOCAL RT_TABLE_LOCAL
413 #define FIB6_TABLE_MIN RT_TABLE_MAIN
414 #define FIB6_TABLE_MAX FIB6_TABLE_MIN
415 #define RT6_TABLE_LOCAL RT6_TABLE_MAIN
418 typedef struct rt6_info
*(*pol_lookup_t
)(struct net
*,
421 const struct sk_buff
*, int);
423 struct fib6_entry_notifier_info
{
424 struct fib_notifier_info info
; /* must be first */
425 struct fib6_info
*rt
;
426 unsigned int nsiblings
;
433 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
);
434 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
);
435 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi6
*fl6
,
436 const struct sk_buff
*skb
,
437 int flags
, pol_lookup_t lookup
);
439 /* called with rcu lock held; can return error pointer
440 * caller needs to select path
442 int fib6_lookup(struct net
*net
, int oif
, struct flowi6
*fl6
,
443 struct fib6_result
*res
, int flags
);
445 /* called with rcu lock held; caller needs to select path */
446 int fib6_table_lookup(struct net
*net
, struct fib6_table
*table
,
447 int oif
, struct flowi6
*fl6
, struct fib6_result
*res
,
450 void fib6_select_path(const struct net
*net
, struct fib6_result
*res
,
451 struct flowi6
*fl6
, int oif
, bool have_oif_match
,
452 const struct sk_buff
*skb
, int strict
);
453 struct fib6_node
*fib6_node_lookup(struct fib6_node
*root
,
454 const struct in6_addr
*daddr
,
455 const struct in6_addr
*saddr
);
457 struct fib6_node
*fib6_locate(struct fib6_node
*root
,
458 const struct in6_addr
*daddr
, int dst_len
,
459 const struct in6_addr
*saddr
, int src_len
,
462 void fib6_clean_all(struct net
*net
, int (*func
)(struct fib6_info
*, void *arg
),
464 void fib6_clean_all_skip_notify(struct net
*net
,
465 int (*func
)(struct fib6_info
*, void *arg
),
468 int fib6_add(struct fib6_node
*root
, struct fib6_info
*rt
,
469 struct nl_info
*info
, struct netlink_ext_ack
*extack
);
470 int fib6_del(struct fib6_info
*rt
, struct nl_info
*info
);
473 void rt6_get_prefsrc(const struct rt6_info
*rt
, struct in6_addr
*addr
)
475 const struct fib6_info
*from
;
479 from
= rcu_dereference(rt
->from
);
481 *addr
= from
->fib6_prefsrc
.addr
;
488 int fib6_nh_init(struct net
*net
, struct fib6_nh
*fib6_nh
,
489 struct fib6_config
*cfg
, gfp_t gfp_flags
,
490 struct netlink_ext_ack
*extack
);
491 void fib6_nh_release(struct fib6_nh
*fib6_nh
);
492 void fib6_nh_release_dsts(struct fib6_nh
*fib6_nh
);
494 int call_fib6_entry_notifiers(struct net
*net
,
495 enum fib_event_type event_type
,
496 struct fib6_info
*rt
,
497 struct netlink_ext_ack
*extack
);
498 int call_fib6_multipath_entry_notifiers(struct net
*net
,
499 enum fib_event_type event_type
,
500 struct fib6_info
*rt
,
501 unsigned int nsiblings
,
502 struct netlink_ext_ack
*extack
);
503 int call_fib6_entry_notifiers_replace(struct net
*net
, struct fib6_info
*rt
);
504 void fib6_rt_update(struct net
*net
, struct fib6_info
*rt
,
505 struct nl_info
*info
);
506 void inet6_rt_notify(int event
, struct fib6_info
*rt
, struct nl_info
*info
,
509 void fib6_run_gc(unsigned long expires
, struct net
*net
, bool force
);
511 void fib6_gc_cleanup(void);
515 /* Add the route to the gc list if it is not already there
517 * The callers should hold f6i->fib6_table->tb6_lock.
519 static inline void fib6_add_gc_list(struct fib6_info
*f6i
)
521 /* If fib6_node is null, the f6i is not in (or removed from) the
524 * There is a gap between finding the f6i from the table and
525 * calling this function without the protection of the tb6_lock.
526 * This check makes sure the f6i is not added to the gc list when
527 * it is not on the table.
529 if (!rcu_dereference_protected(f6i
->fib6_node
,
530 lockdep_is_held(&f6i
->fib6_table
->tb6_lock
)))
533 if (hlist_unhashed(&f6i
->gc_link
))
534 hlist_add_head(&f6i
->gc_link
, &f6i
->fib6_table
->tb6_gc_hlist
);
537 /* Remove the route from the gc list if it is on the list.
539 * The callers should hold f6i->fib6_table->tb6_lock.
541 static inline void fib6_remove_gc_list(struct fib6_info
*f6i
)
543 if (!hlist_unhashed(&f6i
->gc_link
))
544 hlist_del_init(&f6i
->gc_link
);
547 struct ipv6_route_iter
{
548 struct seq_net_private p
;
549 struct fib6_walker w
;
551 struct fib6_table
*tbl
;
555 extern const struct seq_operations ipv6_route_seq_ops
;
557 int call_fib6_notifier(struct notifier_block
*nb
,
558 enum fib_event_type event_type
,
559 struct fib_notifier_info
*info
);
560 int call_fib6_notifiers(struct net
*net
, enum fib_event_type event_type
,
561 struct fib_notifier_info
*info
);
563 int __net_init
fib6_notifier_init(struct net
*net
);
564 void __net_exit
fib6_notifier_exit(struct net
*net
);
566 unsigned int fib6_tables_seq_read(const struct net
*net
);
567 int fib6_tables_dump(struct net
*net
, struct notifier_block
*nb
,
568 struct netlink_ext_ack
*extack
);
570 void fib6_update_sernum(struct net
*net
, struct fib6_info
*rt
);
571 void fib6_update_sernum_upto_root(struct net
*net
, struct fib6_info
*rt
);
572 void fib6_update_sernum_stub(struct net
*net
, struct fib6_info
*f6i
);
574 void fib6_metric_set(struct fib6_info
*f6i
, int metric
, u32 val
);
575 static inline bool fib6_metric_locked(struct fib6_info
*f6i
, int metric
)
577 return !!(f6i
->fib6_metrics
->metrics
[RTAX_LOCK
- 1] & (1 << metric
));
579 void fib6_info_hw_flags_set(struct net
*net
, struct fib6_info
*f6i
,
580 bool offload
, bool trap
, bool offload_failed
);
582 #if IS_BUILTIN(CONFIG_IPV6) && defined(CONFIG_BPF_SYSCALL)
583 struct bpf_iter__ipv6_route
{
584 __bpf_md_ptr(struct bpf_iter_meta
*, meta
);
585 __bpf_md_ptr(struct fib6_info
*, rt
);
589 INDIRECT_CALLABLE_DECLARE(struct rt6_info
*ip6_pol_route_output(struct net
*net
,
590 struct fib6_table
*table
,
592 const struct sk_buff
*skb
,
594 INDIRECT_CALLABLE_DECLARE(struct rt6_info
*ip6_pol_route_input(struct net
*net
,
595 struct fib6_table
*table
,
597 const struct sk_buff
*skb
,
599 INDIRECT_CALLABLE_DECLARE(struct rt6_info
*__ip6_route_redirect(struct net
*net
,
600 struct fib6_table
*table
,
602 const struct sk_buff
*skb
,
604 INDIRECT_CALLABLE_DECLARE(struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
605 struct fib6_table
*table
,
607 const struct sk_buff
*skb
,
609 static inline struct rt6_info
*pol_lookup_func(pol_lookup_t lookup
,
611 struct fib6_table
*table
,
613 const struct sk_buff
*skb
,
616 return INDIRECT_CALL_4(lookup
,
617 ip6_pol_route_output
,
619 ip6_pol_route_lookup
,
620 __ip6_route_redirect
,
621 net
, table
, fl6
, skb
, flags
);
624 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
625 static inline bool fib6_has_custom_rules(const struct net
*net
)
627 return net
->ipv6
.fib6_has_custom_rules
;
630 int fib6_rules_init(void);
631 void fib6_rules_cleanup(void);
632 bool fib6_rule_default(const struct fib_rule
*rule
);
633 int fib6_rules_dump(struct net
*net
, struct notifier_block
*nb
,
634 struct netlink_ext_ack
*extack
);
635 unsigned int fib6_rules_seq_read(const struct net
*net
);
637 static inline bool fib6_rules_early_flow_dissect(struct net
*net
,
640 struct flow_keys
*flkeys
)
642 unsigned int flag
= FLOW_DISSECTOR_F_STOP_AT_ENCAP
;
644 if (!net
->ipv6
.fib6_rules_require_fldissect
)
647 memset(flkeys
, 0, sizeof(*flkeys
));
648 __skb_flow_dissect(net
, skb
, &flow_keys_dissector
,
649 flkeys
, NULL
, 0, 0, 0, flag
);
651 fl6
->fl6_sport
= flkeys
->ports
.src
;
652 fl6
->fl6_dport
= flkeys
->ports
.dst
;
653 fl6
->flowi6_proto
= flkeys
->basic
.ip_proto
;
658 static inline bool fib6_has_custom_rules(const struct net
*net
)
662 static inline int fib6_rules_init(void)
666 static inline void fib6_rules_cleanup(void)
670 static inline bool fib6_rule_default(const struct fib_rule
*rule
)
674 static inline int fib6_rules_dump(struct net
*net
, struct notifier_block
*nb
,
675 struct netlink_ext_ack
*extack
)
679 static inline unsigned int fib6_rules_seq_read(const struct net
*net
)
683 static inline bool fib6_rules_early_flow_dissect(struct net
*net
,
686 struct flow_keys
*flkeys
)