2 * Generic address resolution entity
5 * Pedro Roque <roque@di.fc.ul.pt>
6 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 * Vitaly E. Lavrov releasing NULL neighbor in neigh_add.
15 * Harald Welte Add neighbour cache statistics like rtstat
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 #include <linux/slab.h>
21 #include <linux/types.h>
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/socket.h>
25 #include <linux/netdevice.h>
26 #include <linux/proc_fs.h>
28 #include <linux/sysctl.h>
30 #include <linux/times.h>
31 #include <net/net_namespace.h>
32 #include <net/neighbour.h>
35 #include <net/netevent.h>
36 #include <net/netlink.h>
37 #include <linux/rtnetlink.h>
38 #include <linux/random.h>
39 #include <linux/string.h>
40 #include <linux/log2.h>
41 #include <linux/inetdevice.h>
42 #include <net/addrconf.h>
46 #define neigh_dbg(level, fmt, ...) \
48 if (level <= NEIGH_DEBUG) \
49 pr_debug(fmt, ##__VA_ARGS__); \
52 #define PNEIGH_HASHMASK 0xF
54 static void neigh_timer_handler(unsigned long arg
);
55 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
);
56 static void neigh_update_notify(struct neighbour
*neigh
);
57 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
);
60 static const struct file_operations neigh_stat_seq_fops
;
64 Neighbour hash table buckets are protected with rwlock tbl->lock.
66 - All the scans/updates to hash buckets MUST be made under this lock.
67 - NOTHING clever should be made under this lock: no callbacks
68 to protocol backends, no attempts to send something to network.
69 It will result in deadlocks, if backend/driver wants to use neighbour
71 - If the entry requires some non-trivial actions, increase
72 its reference count and release table lock.
74 Neighbour entries are protected:
75 - with reference count.
76 - with rwlock neigh->lock
78 Reference count prevents destruction.
80 neigh->lock mainly serializes ll address data and its validity state.
81 However, the same lock is used to protect another entry fields:
85 Again, nothing clever shall be made under neigh->lock,
86 the most complicated procedure, which we allow is dev->hard_header.
87 It is supposed, that dev->hard_header is simplistic and does
88 not make callbacks to neighbour tables.
91 static int neigh_blackhole(struct neighbour
*neigh
, struct sk_buff
*skb
)
97 static void neigh_cleanup_and_release(struct neighbour
*neigh
)
99 if (neigh
->parms
->neigh_cleanup
)
100 neigh
->parms
->neigh_cleanup(neigh
);
102 __neigh_notify(neigh
, RTM_DELNEIGH
, 0);
103 neigh_release(neigh
);
107 * It is random distribution in the interval (1/2)*base...(3/2)*base.
108 * It corresponds to default IPv6 settings and is not overridable,
109 * because it is really reasonable choice.
112 unsigned long neigh_rand_reach_time(unsigned long base
)
114 return base
? (prandom_u32() % base
) + (base
>> 1) : 0;
116 EXPORT_SYMBOL(neigh_rand_reach_time
);
119 static int neigh_forced_gc(struct neigh_table
*tbl
)
123 struct neigh_hash_table
*nht
;
125 NEIGH_CACHE_STAT_INC(tbl
, forced_gc_runs
);
127 write_lock_bh(&tbl
->lock
);
128 nht
= rcu_dereference_protected(tbl
->nht
,
129 lockdep_is_held(&tbl
->lock
));
130 for (i
= 0; i
< (1 << nht
->hash_shift
); i
++) {
132 struct neighbour __rcu
**np
;
134 np
= &nht
->hash_buckets
[i
];
135 while ((n
= rcu_dereference_protected(*np
,
136 lockdep_is_held(&tbl
->lock
))) != NULL
) {
137 /* Neighbour record may be discarded if:
138 * - nobody refers to it.
139 * - it is not permanent
141 write_lock(&n
->lock
);
142 if (atomic_read(&n
->refcnt
) == 1 &&
143 !(n
->nud_state
& NUD_PERMANENT
)) {
144 rcu_assign_pointer(*np
,
145 rcu_dereference_protected(n
->next
,
146 lockdep_is_held(&tbl
->lock
)));
149 write_unlock(&n
->lock
);
150 neigh_cleanup_and_release(n
);
153 write_unlock(&n
->lock
);
158 tbl
->last_flush
= jiffies
;
160 write_unlock_bh(&tbl
->lock
);
165 static void neigh_add_timer(struct neighbour
*n
, unsigned long when
)
168 if (unlikely(mod_timer(&n
->timer
, when
))) {
169 printk("NEIGH: BUG, double timer add, state is %x\n",
175 static int neigh_del_timer(struct neighbour
*n
)
177 if ((n
->nud_state
& NUD_IN_TIMER
) &&
178 del_timer(&n
->timer
)) {
185 static void pneigh_queue_purge(struct sk_buff_head
*list
)
189 while ((skb
= skb_dequeue(list
)) != NULL
) {
195 static void neigh_flush_dev(struct neigh_table
*tbl
, struct net_device
*dev
)
198 struct neigh_hash_table
*nht
;
200 nht
= rcu_dereference_protected(tbl
->nht
,
201 lockdep_is_held(&tbl
->lock
));
203 for (i
= 0; i
< (1 << nht
->hash_shift
); i
++) {
205 struct neighbour __rcu
**np
= &nht
->hash_buckets
[i
];
207 while ((n
= rcu_dereference_protected(*np
,
208 lockdep_is_held(&tbl
->lock
))) != NULL
) {
209 if (dev
&& n
->dev
!= dev
) {
213 rcu_assign_pointer(*np
,
214 rcu_dereference_protected(n
->next
,
215 lockdep_is_held(&tbl
->lock
)));
216 write_lock(&n
->lock
);
220 if (atomic_read(&n
->refcnt
) != 1) {
221 /* The most unpleasant situation.
222 We must destroy neighbour entry,
223 but someone still uses it.
225 The destroy will be delayed until
226 the last user releases us, but
227 we must kill timers etc. and move
230 __skb_queue_purge(&n
->arp_queue
);
231 n
->arp_queue_len_bytes
= 0;
232 n
->output
= neigh_blackhole
;
233 if (n
->nud_state
& NUD_VALID
)
234 n
->nud_state
= NUD_NOARP
;
236 n
->nud_state
= NUD_NONE
;
237 neigh_dbg(2, "neigh %p is stray\n", n
);
239 write_unlock(&n
->lock
);
240 neigh_cleanup_and_release(n
);
245 void neigh_changeaddr(struct neigh_table
*tbl
, struct net_device
*dev
)
247 write_lock_bh(&tbl
->lock
);
248 neigh_flush_dev(tbl
, dev
);
249 write_unlock_bh(&tbl
->lock
);
251 EXPORT_SYMBOL(neigh_changeaddr
);
253 int neigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
255 write_lock_bh(&tbl
->lock
);
256 neigh_flush_dev(tbl
, dev
);
257 pneigh_ifdown(tbl
, dev
);
258 write_unlock_bh(&tbl
->lock
);
260 del_timer_sync(&tbl
->proxy_timer
);
261 pneigh_queue_purge(&tbl
->proxy_queue
);
264 EXPORT_SYMBOL(neigh_ifdown
);
266 static struct neighbour
*neigh_alloc(struct neigh_table
*tbl
, struct net_device
*dev
)
268 struct neighbour
*n
= NULL
;
269 unsigned long now
= jiffies
;
272 entries
= atomic_inc_return(&tbl
->entries
) - 1;
273 if (entries
>= tbl
->gc_thresh3
||
274 (entries
>= tbl
->gc_thresh2
&&
275 time_after(now
, tbl
->last_flush
+ 5 * HZ
))) {
276 if (!neigh_forced_gc(tbl
) &&
277 entries
>= tbl
->gc_thresh3
) {
278 net_info_ratelimited("%s: neighbor table overflow!\n",
280 NEIGH_CACHE_STAT_INC(tbl
, table_fulls
);
285 n
= kzalloc(tbl
->entry_size
+ dev
->neigh_priv_len
, GFP_ATOMIC
);
289 __skb_queue_head_init(&n
->arp_queue
);
290 rwlock_init(&n
->lock
);
291 seqlock_init(&n
->ha_lock
);
292 n
->updated
= n
->used
= now
;
293 n
->nud_state
= NUD_NONE
;
294 n
->output
= neigh_blackhole
;
295 seqlock_init(&n
->hh
.hh_lock
);
296 n
->parms
= neigh_parms_clone(&tbl
->parms
);
297 setup_timer(&n
->timer
, neigh_timer_handler
, (unsigned long)n
);
299 NEIGH_CACHE_STAT_INC(tbl
, allocs
);
301 atomic_set(&n
->refcnt
, 1);
307 atomic_dec(&tbl
->entries
);
311 static void neigh_get_hash_rnd(u32
*x
)
313 get_random_bytes(x
, sizeof(*x
));
317 static struct neigh_hash_table
*neigh_hash_alloc(unsigned int shift
)
319 size_t size
= (1 << shift
) * sizeof(struct neighbour
*);
320 struct neigh_hash_table
*ret
;
321 struct neighbour __rcu
**buckets
;
324 ret
= kmalloc(sizeof(*ret
), GFP_ATOMIC
);
327 if (size
<= PAGE_SIZE
)
328 buckets
= kzalloc(size
, GFP_ATOMIC
);
330 buckets
= (struct neighbour __rcu
**)
331 __get_free_pages(GFP_ATOMIC
| __GFP_ZERO
,
337 ret
->hash_buckets
= buckets
;
338 ret
->hash_shift
= shift
;
339 for (i
= 0; i
< NEIGH_NUM_HASH_RND
; i
++)
340 neigh_get_hash_rnd(&ret
->hash_rnd
[i
]);
344 static void neigh_hash_free_rcu(struct rcu_head
*head
)
346 struct neigh_hash_table
*nht
= container_of(head
,
347 struct neigh_hash_table
,
349 size_t size
= (1 << nht
->hash_shift
) * sizeof(struct neighbour
*);
350 struct neighbour __rcu
**buckets
= nht
->hash_buckets
;
352 if (size
<= PAGE_SIZE
)
355 free_pages((unsigned long)buckets
, get_order(size
));
359 static struct neigh_hash_table
*neigh_hash_grow(struct neigh_table
*tbl
,
360 unsigned long new_shift
)
362 unsigned int i
, hash
;
363 struct neigh_hash_table
*new_nht
, *old_nht
;
365 NEIGH_CACHE_STAT_INC(tbl
, hash_grows
);
367 old_nht
= rcu_dereference_protected(tbl
->nht
,
368 lockdep_is_held(&tbl
->lock
));
369 new_nht
= neigh_hash_alloc(new_shift
);
373 for (i
= 0; i
< (1 << old_nht
->hash_shift
); i
++) {
374 struct neighbour
*n
, *next
;
376 for (n
= rcu_dereference_protected(old_nht
->hash_buckets
[i
],
377 lockdep_is_held(&tbl
->lock
));
380 hash
= tbl
->hash(n
->primary_key
, n
->dev
,
383 hash
>>= (32 - new_nht
->hash_shift
);
384 next
= rcu_dereference_protected(n
->next
,
385 lockdep_is_held(&tbl
->lock
));
387 rcu_assign_pointer(n
->next
,
388 rcu_dereference_protected(
389 new_nht
->hash_buckets
[hash
],
390 lockdep_is_held(&tbl
->lock
)));
391 rcu_assign_pointer(new_nht
->hash_buckets
[hash
], n
);
395 rcu_assign_pointer(tbl
->nht
, new_nht
);
396 call_rcu(&old_nht
->rcu
, neigh_hash_free_rcu
);
400 struct neighbour
*neigh_lookup(struct neigh_table
*tbl
, const void *pkey
,
401 struct net_device
*dev
)
405 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
408 n
= __neigh_lookup_noref(tbl
, pkey
, dev
);
410 if (!atomic_inc_not_zero(&n
->refcnt
))
412 NEIGH_CACHE_STAT_INC(tbl
, hits
);
415 rcu_read_unlock_bh();
418 EXPORT_SYMBOL(neigh_lookup
);
420 struct neighbour
*neigh_lookup_nodev(struct neigh_table
*tbl
, struct net
*net
,
424 int key_len
= tbl
->key_len
;
426 struct neigh_hash_table
*nht
;
428 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
431 nht
= rcu_dereference_bh(tbl
->nht
);
432 hash_val
= tbl
->hash(pkey
, NULL
, nht
->hash_rnd
) >> (32 - nht
->hash_shift
);
434 for (n
= rcu_dereference_bh(nht
->hash_buckets
[hash_val
]);
436 n
= rcu_dereference_bh(n
->next
)) {
437 if (!memcmp(n
->primary_key
, pkey
, key_len
) &&
438 net_eq(dev_net(n
->dev
), net
)) {
439 if (!atomic_inc_not_zero(&n
->refcnt
))
441 NEIGH_CACHE_STAT_INC(tbl
, hits
);
446 rcu_read_unlock_bh();
449 EXPORT_SYMBOL(neigh_lookup_nodev
);
451 struct neighbour
*__neigh_create(struct neigh_table
*tbl
, const void *pkey
,
452 struct net_device
*dev
, bool want_ref
)
455 int key_len
= tbl
->key_len
;
457 struct neighbour
*n1
, *rc
, *n
= neigh_alloc(tbl
, dev
);
458 struct neigh_hash_table
*nht
;
461 rc
= ERR_PTR(-ENOBUFS
);
465 memcpy(n
->primary_key
, pkey
, key_len
);
469 /* Protocol specific setup. */
470 if (tbl
->constructor
&& (error
= tbl
->constructor(n
)) < 0) {
472 goto out_neigh_release
;
475 if (dev
->netdev_ops
->ndo_neigh_construct
) {
476 error
= dev
->netdev_ops
->ndo_neigh_construct(n
);
479 goto out_neigh_release
;
483 /* Device specific setup. */
484 if (n
->parms
->neigh_setup
&&
485 (error
= n
->parms
->neigh_setup(n
)) < 0) {
487 goto out_neigh_release
;
490 n
->confirmed
= jiffies
- (NEIGH_VAR(n
->parms
, BASE_REACHABLE_TIME
) << 1);
492 write_lock_bh(&tbl
->lock
);
493 nht
= rcu_dereference_protected(tbl
->nht
,
494 lockdep_is_held(&tbl
->lock
));
496 if (atomic_read(&tbl
->entries
) > (1 << nht
->hash_shift
))
497 nht
= neigh_hash_grow(tbl
, nht
->hash_shift
+ 1);
499 hash_val
= tbl
->hash(pkey
, dev
, nht
->hash_rnd
) >> (32 - nht
->hash_shift
);
501 if (n
->parms
->dead
) {
502 rc
= ERR_PTR(-EINVAL
);
506 for (n1
= rcu_dereference_protected(nht
->hash_buckets
[hash_val
],
507 lockdep_is_held(&tbl
->lock
));
509 n1
= rcu_dereference_protected(n1
->next
,
510 lockdep_is_held(&tbl
->lock
))) {
511 if (dev
== n1
->dev
&& !memcmp(n1
->primary_key
, pkey
, key_len
)) {
522 rcu_assign_pointer(n
->next
,
523 rcu_dereference_protected(nht
->hash_buckets
[hash_val
],
524 lockdep_is_held(&tbl
->lock
)));
525 rcu_assign_pointer(nht
->hash_buckets
[hash_val
], n
);
526 write_unlock_bh(&tbl
->lock
);
527 neigh_dbg(2, "neigh %p is created\n", n
);
532 write_unlock_bh(&tbl
->lock
);
537 EXPORT_SYMBOL(__neigh_create
);
539 static u32
pneigh_hash(const void *pkey
, int key_len
)
541 u32 hash_val
= *(u32
*)(pkey
+ key_len
- 4);
542 hash_val
^= (hash_val
>> 16);
543 hash_val
^= hash_val
>> 8;
544 hash_val
^= hash_val
>> 4;
545 hash_val
&= PNEIGH_HASHMASK
;
549 static struct pneigh_entry
*__pneigh_lookup_1(struct pneigh_entry
*n
,
553 struct net_device
*dev
)
556 if (!memcmp(n
->key
, pkey
, key_len
) &&
557 net_eq(pneigh_net(n
), net
) &&
558 (n
->dev
== dev
|| !n
->dev
))
565 struct pneigh_entry
*__pneigh_lookup(struct neigh_table
*tbl
,
566 struct net
*net
, const void *pkey
, struct net_device
*dev
)
568 int key_len
= tbl
->key_len
;
569 u32 hash_val
= pneigh_hash(pkey
, key_len
);
571 return __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
572 net
, pkey
, key_len
, dev
);
574 EXPORT_SYMBOL_GPL(__pneigh_lookup
);
576 struct pneigh_entry
* pneigh_lookup(struct neigh_table
*tbl
,
577 struct net
*net
, const void *pkey
,
578 struct net_device
*dev
, int creat
)
580 struct pneigh_entry
*n
;
581 int key_len
= tbl
->key_len
;
582 u32 hash_val
= pneigh_hash(pkey
, key_len
);
584 read_lock_bh(&tbl
->lock
);
585 n
= __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
586 net
, pkey
, key_len
, dev
);
587 read_unlock_bh(&tbl
->lock
);
594 n
= kmalloc(sizeof(*n
) + key_len
, GFP_KERNEL
);
598 write_pnet(&n
->net
, net
);
599 memcpy(n
->key
, pkey
, key_len
);
604 if (tbl
->pconstructor
&& tbl
->pconstructor(n
)) {
612 write_lock_bh(&tbl
->lock
);
613 n
->next
= tbl
->phash_buckets
[hash_val
];
614 tbl
->phash_buckets
[hash_val
] = n
;
615 write_unlock_bh(&tbl
->lock
);
619 EXPORT_SYMBOL(pneigh_lookup
);
622 int pneigh_delete(struct neigh_table
*tbl
, struct net
*net
, const void *pkey
,
623 struct net_device
*dev
)
625 struct pneigh_entry
*n
, **np
;
626 int key_len
= tbl
->key_len
;
627 u32 hash_val
= pneigh_hash(pkey
, key_len
);
629 write_lock_bh(&tbl
->lock
);
630 for (np
= &tbl
->phash_buckets
[hash_val
]; (n
= *np
) != NULL
;
632 if (!memcmp(n
->key
, pkey
, key_len
) && n
->dev
== dev
&&
633 net_eq(pneigh_net(n
), net
)) {
635 write_unlock_bh(&tbl
->lock
);
636 if (tbl
->pdestructor
)
644 write_unlock_bh(&tbl
->lock
);
648 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
650 struct pneigh_entry
*n
, **np
;
653 for (h
= 0; h
<= PNEIGH_HASHMASK
; h
++) {
654 np
= &tbl
->phash_buckets
[h
];
655 while ((n
= *np
) != NULL
) {
656 if (!dev
|| n
->dev
== dev
) {
658 if (tbl
->pdestructor
)
671 static void neigh_parms_destroy(struct neigh_parms
*parms
);
673 static inline void neigh_parms_put(struct neigh_parms
*parms
)
675 if (atomic_dec_and_test(&parms
->refcnt
))
676 neigh_parms_destroy(parms
);
680 * neighbour must already be out of the table;
683 void neigh_destroy(struct neighbour
*neigh
)
685 struct net_device
*dev
= neigh
->dev
;
687 NEIGH_CACHE_STAT_INC(neigh
->tbl
, destroys
);
690 pr_warn("Destroying alive neighbour %p\n", neigh
);
695 if (neigh_del_timer(neigh
))
696 pr_warn("Impossible event\n");
698 write_lock_bh(&neigh
->lock
);
699 __skb_queue_purge(&neigh
->arp_queue
);
700 write_unlock_bh(&neigh
->lock
);
701 neigh
->arp_queue_len_bytes
= 0;
703 if (dev
->netdev_ops
->ndo_neigh_destroy
)
704 dev
->netdev_ops
->ndo_neigh_destroy(neigh
);
707 neigh_parms_put(neigh
->parms
);
709 neigh_dbg(2, "neigh %p is destroyed\n", neigh
);
711 atomic_dec(&neigh
->tbl
->entries
);
712 kfree_rcu(neigh
, rcu
);
714 EXPORT_SYMBOL(neigh_destroy
);
716 /* Neighbour state is suspicious;
719 Called with write_locked neigh.
721 static void neigh_suspect(struct neighbour
*neigh
)
723 neigh_dbg(2, "neigh %p is suspected\n", neigh
);
725 neigh
->output
= neigh
->ops
->output
;
728 /* Neighbour state is OK;
731 Called with write_locked neigh.
733 static void neigh_connect(struct neighbour
*neigh
)
735 neigh_dbg(2, "neigh %p is connected\n", neigh
);
737 neigh
->output
= neigh
->ops
->connected_output
;
740 static void neigh_periodic_work(struct work_struct
*work
)
742 struct neigh_table
*tbl
= container_of(work
, struct neigh_table
, gc_work
.work
);
744 struct neighbour __rcu
**np
;
746 struct neigh_hash_table
*nht
;
748 NEIGH_CACHE_STAT_INC(tbl
, periodic_gc_runs
);
750 write_lock_bh(&tbl
->lock
);
751 nht
= rcu_dereference_protected(tbl
->nht
,
752 lockdep_is_held(&tbl
->lock
));
755 * periodically recompute ReachableTime from random function
758 if (time_after(jiffies
, tbl
->last_rand
+ 300 * HZ
)) {
759 struct neigh_parms
*p
;
760 tbl
->last_rand
= jiffies
;
761 list_for_each_entry(p
, &tbl
->parms_list
, list
)
763 neigh_rand_reach_time(NEIGH_VAR(p
, BASE_REACHABLE_TIME
));
766 if (atomic_read(&tbl
->entries
) < tbl
->gc_thresh1
)
769 for (i
= 0 ; i
< (1 << nht
->hash_shift
); i
++) {
770 np
= &nht
->hash_buckets
[i
];
772 while ((n
= rcu_dereference_protected(*np
,
773 lockdep_is_held(&tbl
->lock
))) != NULL
) {
776 write_lock(&n
->lock
);
778 state
= n
->nud_state
;
779 if (state
& (NUD_PERMANENT
| NUD_IN_TIMER
)) {
780 write_unlock(&n
->lock
);
784 if (time_before(n
->used
, n
->confirmed
))
785 n
->used
= n
->confirmed
;
787 if (atomic_read(&n
->refcnt
) == 1 &&
788 (state
== NUD_FAILED
||
789 time_after(jiffies
, n
->used
+ NEIGH_VAR(n
->parms
, GC_STALETIME
)))) {
792 write_unlock(&n
->lock
);
793 neigh_cleanup_and_release(n
);
796 write_unlock(&n
->lock
);
802 * It's fine to release lock here, even if hash table
803 * grows while we are preempted.
805 write_unlock_bh(&tbl
->lock
);
807 write_lock_bh(&tbl
->lock
);
808 nht
= rcu_dereference_protected(tbl
->nht
,
809 lockdep_is_held(&tbl
->lock
));
812 /* Cycle through all hash buckets every BASE_REACHABLE_TIME/2 ticks.
813 * ARP entry timeouts range from 1/2 BASE_REACHABLE_TIME to 3/2
814 * BASE_REACHABLE_TIME.
816 queue_delayed_work(system_power_efficient_wq
, &tbl
->gc_work
,
817 NEIGH_VAR(&tbl
->parms
, BASE_REACHABLE_TIME
) >> 1);
818 write_unlock_bh(&tbl
->lock
);
821 static __inline__
int neigh_max_probes(struct neighbour
*n
)
823 struct neigh_parms
*p
= n
->parms
;
824 return NEIGH_VAR(p
, UCAST_PROBES
) + NEIGH_VAR(p
, APP_PROBES
) +
825 (n
->nud_state
& NUD_PROBE
? NEIGH_VAR(p
, MCAST_REPROBES
) :
826 NEIGH_VAR(p
, MCAST_PROBES
));
829 static void neigh_invalidate(struct neighbour
*neigh
)
830 __releases(neigh
->lock
)
831 __acquires(neigh
->lock
)
835 NEIGH_CACHE_STAT_INC(neigh
->tbl
, res_failed
);
836 neigh_dbg(2, "neigh %p is failed\n", neigh
);
837 neigh
->updated
= jiffies
;
839 /* It is very thin place. report_unreachable is very complicated
840 routine. Particularly, it can hit the same neighbour entry!
842 So that, we try to be accurate and avoid dead loop. --ANK
844 while (neigh
->nud_state
== NUD_FAILED
&&
845 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
846 write_unlock(&neigh
->lock
);
847 neigh
->ops
->error_report(neigh
, skb
);
848 write_lock(&neigh
->lock
);
850 __skb_queue_purge(&neigh
->arp_queue
);
851 neigh
->arp_queue_len_bytes
= 0;
854 static void neigh_probe(struct neighbour
*neigh
)
855 __releases(neigh
->lock
)
857 struct sk_buff
*skb
= skb_peek_tail(&neigh
->arp_queue
);
858 /* keep skb alive even if arp_queue overflows */
860 skb
= skb_clone(skb
, GFP_ATOMIC
);
861 write_unlock(&neigh
->lock
);
862 neigh
->ops
->solicit(neigh
, skb
);
863 atomic_inc(&neigh
->probes
);
867 /* Called when a timer expires for a neighbour entry. */
869 static void neigh_timer_handler(unsigned long arg
)
871 unsigned long now
, next
;
872 struct neighbour
*neigh
= (struct neighbour
*)arg
;
876 write_lock(&neigh
->lock
);
878 state
= neigh
->nud_state
;
882 if (!(state
& NUD_IN_TIMER
))
885 if (state
& NUD_REACHABLE
) {
886 if (time_before_eq(now
,
887 neigh
->confirmed
+ neigh
->parms
->reachable_time
)) {
888 neigh_dbg(2, "neigh %p is still alive\n", neigh
);
889 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
890 } else if (time_before_eq(now
,
892 NEIGH_VAR(neigh
->parms
, DELAY_PROBE_TIME
))) {
893 neigh_dbg(2, "neigh %p is delayed\n", neigh
);
894 neigh
->nud_state
= NUD_DELAY
;
895 neigh
->updated
= jiffies
;
896 neigh_suspect(neigh
);
897 next
= now
+ NEIGH_VAR(neigh
->parms
, DELAY_PROBE_TIME
);
899 neigh_dbg(2, "neigh %p is suspected\n", neigh
);
900 neigh
->nud_state
= NUD_STALE
;
901 neigh
->updated
= jiffies
;
902 neigh_suspect(neigh
);
905 } else if (state
& NUD_DELAY
) {
906 if (time_before_eq(now
,
908 NEIGH_VAR(neigh
->parms
, DELAY_PROBE_TIME
))) {
909 neigh_dbg(2, "neigh %p is now reachable\n", neigh
);
910 neigh
->nud_state
= NUD_REACHABLE
;
911 neigh
->updated
= jiffies
;
912 neigh_connect(neigh
);
914 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
916 neigh_dbg(2, "neigh %p is probed\n", neigh
);
917 neigh
->nud_state
= NUD_PROBE
;
918 neigh
->updated
= jiffies
;
919 atomic_set(&neigh
->probes
, 0);
921 next
= now
+ NEIGH_VAR(neigh
->parms
, RETRANS_TIME
);
924 /* NUD_PROBE|NUD_INCOMPLETE */
925 next
= now
+ NEIGH_VAR(neigh
->parms
, RETRANS_TIME
);
928 if ((neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) &&
929 atomic_read(&neigh
->probes
) >= neigh_max_probes(neigh
)) {
930 neigh
->nud_state
= NUD_FAILED
;
932 neigh_invalidate(neigh
);
936 if (neigh
->nud_state
& NUD_IN_TIMER
) {
937 if (time_before(next
, jiffies
+ HZ
/2))
938 next
= jiffies
+ HZ
/2;
939 if (!mod_timer(&neigh
->timer
, next
))
942 if (neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) {
946 write_unlock(&neigh
->lock
);
950 neigh_update_notify(neigh
);
952 neigh_release(neigh
);
955 int __neigh_event_send(struct neighbour
*neigh
, struct sk_buff
*skb
)
958 bool immediate_probe
= false;
960 write_lock_bh(&neigh
->lock
);
963 if (neigh
->nud_state
& (NUD_CONNECTED
| NUD_DELAY
| NUD_PROBE
))
968 if (!(neigh
->nud_state
& (NUD_STALE
| NUD_INCOMPLETE
))) {
969 if (NEIGH_VAR(neigh
->parms
, MCAST_PROBES
) +
970 NEIGH_VAR(neigh
->parms
, APP_PROBES
)) {
971 unsigned long next
, now
= jiffies
;
973 atomic_set(&neigh
->probes
,
974 NEIGH_VAR(neigh
->parms
, UCAST_PROBES
));
975 neigh
->nud_state
= NUD_INCOMPLETE
;
976 neigh
->updated
= now
;
977 next
= now
+ max(NEIGH_VAR(neigh
->parms
, RETRANS_TIME
),
979 neigh_add_timer(neigh
, next
);
980 immediate_probe
= true;
982 neigh
->nud_state
= NUD_FAILED
;
983 neigh
->updated
= jiffies
;
984 write_unlock_bh(&neigh
->lock
);
989 } else if (neigh
->nud_state
& NUD_STALE
) {
990 neigh_dbg(2, "neigh %p is delayed\n", neigh
);
991 neigh
->nud_state
= NUD_DELAY
;
992 neigh
->updated
= jiffies
;
993 neigh_add_timer(neigh
, jiffies
+
994 NEIGH_VAR(neigh
->parms
, DELAY_PROBE_TIME
));
997 if (neigh
->nud_state
== NUD_INCOMPLETE
) {
999 while (neigh
->arp_queue_len_bytes
+ skb
->truesize
>
1000 NEIGH_VAR(neigh
->parms
, QUEUE_LEN_BYTES
)) {
1001 struct sk_buff
*buff
;
1003 buff
= __skb_dequeue(&neigh
->arp_queue
);
1006 neigh
->arp_queue_len_bytes
-= buff
->truesize
;
1008 NEIGH_CACHE_STAT_INC(neigh
->tbl
, unres_discards
);
1011 __skb_queue_tail(&neigh
->arp_queue
, skb
);
1012 neigh
->arp_queue_len_bytes
+= skb
->truesize
;
1017 if (immediate_probe
)
1020 write_unlock(&neigh
->lock
);
1025 if (neigh
->nud_state
& NUD_STALE
)
1027 write_unlock_bh(&neigh
->lock
);
1031 EXPORT_SYMBOL(__neigh_event_send
);
1033 static void neigh_update_hhs(struct neighbour
*neigh
)
1035 struct hh_cache
*hh
;
1036 void (*update
)(struct hh_cache
*, const struct net_device
*, const unsigned char *)
1039 if (neigh
->dev
->header_ops
)
1040 update
= neigh
->dev
->header_ops
->cache_update
;
1045 write_seqlock_bh(&hh
->hh_lock
);
1046 update(hh
, neigh
->dev
, neigh
->ha
);
1047 write_sequnlock_bh(&hh
->hh_lock
);
1054 /* Generic update routine.
1055 -- lladdr is new lladdr or NULL, if it is not supplied.
1056 -- new is new state.
1058 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1060 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1061 lladdr instead of overriding it
1063 It also allows to retain current state
1064 if lladdr is unchanged.
1065 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1067 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1069 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1072 Caller MUST hold reference count on the entry.
1075 int neigh_update(struct neighbour
*neigh
, const u8
*lladdr
, u8
new,
1081 struct net_device
*dev
;
1082 int update_isrouter
= 0;
1084 write_lock_bh(&neigh
->lock
);
1087 old
= neigh
->nud_state
;
1090 if (!(flags
& NEIGH_UPDATE_F_ADMIN
) &&
1091 (old
& (NUD_NOARP
| NUD_PERMANENT
)))
1096 if (!(new & NUD_VALID
)) {
1097 neigh_del_timer(neigh
);
1098 if (old
& NUD_CONNECTED
)
1099 neigh_suspect(neigh
);
1100 neigh
->nud_state
= new;
1102 notify
= old
& NUD_VALID
;
1103 if ((old
& (NUD_INCOMPLETE
| NUD_PROBE
)) &&
1104 (new & NUD_FAILED
)) {
1105 neigh_invalidate(neigh
);
1111 /* Compare new lladdr with cached one */
1112 if (!dev
->addr_len
) {
1113 /* First case: device needs no address. */
1115 } else if (lladdr
) {
1116 /* The second case: if something is already cached
1117 and a new address is proposed:
1119 - if they are different, check override flag
1121 if ((old
& NUD_VALID
) &&
1122 !memcmp(lladdr
, neigh
->ha
, dev
->addr_len
))
1125 /* No address is supplied; if we know something,
1126 use it, otherwise discard the request.
1129 if (!(old
& NUD_VALID
))
1134 if (new & NUD_CONNECTED
)
1135 neigh
->confirmed
= jiffies
;
1136 neigh
->updated
= jiffies
;
1138 /* If entry was valid and address is not changed,
1139 do not change entry state, if new one is STALE.
1142 update_isrouter
= flags
& NEIGH_UPDATE_F_OVERRIDE_ISROUTER
;
1143 if (old
& NUD_VALID
) {
1144 if (lladdr
!= neigh
->ha
&& !(flags
& NEIGH_UPDATE_F_OVERRIDE
)) {
1145 update_isrouter
= 0;
1146 if ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) &&
1147 (old
& NUD_CONNECTED
)) {
1153 if (lladdr
== neigh
->ha
&& new == NUD_STALE
&&
1154 ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) ||
1155 (old
& NUD_CONNECTED
))
1162 neigh_del_timer(neigh
);
1163 if (new & NUD_PROBE
)
1164 atomic_set(&neigh
->probes
, 0);
1165 if (new & NUD_IN_TIMER
)
1166 neigh_add_timer(neigh
, (jiffies
+
1167 ((new & NUD_REACHABLE
) ?
1168 neigh
->parms
->reachable_time
:
1170 neigh
->nud_state
= new;
1174 if (lladdr
!= neigh
->ha
) {
1175 write_seqlock(&neigh
->ha_lock
);
1176 memcpy(&neigh
->ha
, lladdr
, dev
->addr_len
);
1177 write_sequnlock(&neigh
->ha_lock
);
1178 neigh_update_hhs(neigh
);
1179 if (!(new & NUD_CONNECTED
))
1180 neigh
->confirmed
= jiffies
-
1181 (NEIGH_VAR(neigh
->parms
, BASE_REACHABLE_TIME
) << 1);
1186 if (new & NUD_CONNECTED
)
1187 neigh_connect(neigh
);
1189 neigh_suspect(neigh
);
1190 if (!(old
& NUD_VALID
)) {
1191 struct sk_buff
*skb
;
1193 /* Again: avoid dead loop if something went wrong */
1195 while (neigh
->nud_state
& NUD_VALID
&&
1196 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
1197 struct dst_entry
*dst
= skb_dst(skb
);
1198 struct neighbour
*n2
, *n1
= neigh
;
1199 write_unlock_bh(&neigh
->lock
);
1203 /* Why not just use 'neigh' as-is? The problem is that
1204 * things such as shaper, eql, and sch_teql can end up
1205 * using alternative, different, neigh objects to output
1206 * the packet in the output path. So what we need to do
1207 * here is re-lookup the top-level neigh in the path so
1208 * we can reinject the packet there.
1212 n2
= dst_neigh_lookup_skb(dst
, skb
);
1216 n1
->output(n1
, skb
);
1221 write_lock_bh(&neigh
->lock
);
1223 __skb_queue_purge(&neigh
->arp_queue
);
1224 neigh
->arp_queue_len_bytes
= 0;
1227 if (update_isrouter
) {
1228 neigh
->flags
= (flags
& NEIGH_UPDATE_F_ISROUTER
) ?
1229 (neigh
->flags
| NTF_ROUTER
) :
1230 (neigh
->flags
& ~NTF_ROUTER
);
1232 write_unlock_bh(&neigh
->lock
);
1235 neigh_update_notify(neigh
);
1239 EXPORT_SYMBOL(neigh_update
);
1241 /* Update the neigh to listen temporarily for probe responses, even if it is
1242 * in a NUD_FAILED state. The caller has to hold neigh->lock for writing.
1244 void __neigh_set_probe_once(struct neighbour
*neigh
)
1248 neigh
->updated
= jiffies
;
1249 if (!(neigh
->nud_state
& NUD_FAILED
))
1251 neigh
->nud_state
= NUD_INCOMPLETE
;
1252 atomic_set(&neigh
->probes
, neigh_max_probes(neigh
));
1253 neigh_add_timer(neigh
,
1254 jiffies
+ NEIGH_VAR(neigh
->parms
, RETRANS_TIME
));
1256 EXPORT_SYMBOL(__neigh_set_probe_once
);
1258 struct neighbour
*neigh_event_ns(struct neigh_table
*tbl
,
1259 u8
*lladdr
, void *saddr
,
1260 struct net_device
*dev
)
1262 struct neighbour
*neigh
= __neigh_lookup(tbl
, saddr
, dev
,
1263 lladdr
|| !dev
->addr_len
);
1265 neigh_update(neigh
, lladdr
, NUD_STALE
,
1266 NEIGH_UPDATE_F_OVERRIDE
);
1269 EXPORT_SYMBOL(neigh_event_ns
);
1271 /* called with read_lock_bh(&n->lock); */
1272 static void neigh_hh_init(struct neighbour
*n
)
1274 struct net_device
*dev
= n
->dev
;
1275 __be16 prot
= n
->tbl
->protocol
;
1276 struct hh_cache
*hh
= &n
->hh
;
1278 write_lock_bh(&n
->lock
);
1280 /* Only one thread can come in here and initialize the
1284 dev
->header_ops
->cache(n
, hh
, prot
);
1286 write_unlock_bh(&n
->lock
);
1289 /* Slow and careful. */
1291 int neigh_resolve_output(struct neighbour
*neigh
, struct sk_buff
*skb
)
1295 if (!neigh_event_send(neigh
, skb
)) {
1297 struct net_device
*dev
= neigh
->dev
;
1300 if (dev
->header_ops
->cache
&& !neigh
->hh
.hh_len
)
1301 neigh_hh_init(neigh
);
1304 __skb_pull(skb
, skb_network_offset(skb
));
1305 seq
= read_seqbegin(&neigh
->ha_lock
);
1306 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1307 neigh
->ha
, NULL
, skb
->len
);
1308 } while (read_seqretry(&neigh
->ha_lock
, seq
));
1311 rc
= dev_queue_xmit(skb
);
1322 EXPORT_SYMBOL(neigh_resolve_output
);
1324 /* As fast as possible without hh cache */
1326 int neigh_connected_output(struct neighbour
*neigh
, struct sk_buff
*skb
)
1328 struct net_device
*dev
= neigh
->dev
;
1333 __skb_pull(skb
, skb_network_offset(skb
));
1334 seq
= read_seqbegin(&neigh
->ha_lock
);
1335 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1336 neigh
->ha
, NULL
, skb
->len
);
1337 } while (read_seqretry(&neigh
->ha_lock
, seq
));
1340 err
= dev_queue_xmit(skb
);
1347 EXPORT_SYMBOL(neigh_connected_output
);
1349 int neigh_direct_output(struct neighbour
*neigh
, struct sk_buff
*skb
)
1351 return dev_queue_xmit(skb
);
1353 EXPORT_SYMBOL(neigh_direct_output
);
1355 static void neigh_proxy_process(unsigned long arg
)
1357 struct neigh_table
*tbl
= (struct neigh_table
*)arg
;
1358 long sched_next
= 0;
1359 unsigned long now
= jiffies
;
1360 struct sk_buff
*skb
, *n
;
1362 spin_lock(&tbl
->proxy_queue
.lock
);
1364 skb_queue_walk_safe(&tbl
->proxy_queue
, skb
, n
) {
1365 long tdif
= NEIGH_CB(skb
)->sched_next
- now
;
1368 struct net_device
*dev
= skb
->dev
;
1370 __skb_unlink(skb
, &tbl
->proxy_queue
);
1371 if (tbl
->proxy_redo
&& netif_running(dev
)) {
1373 tbl
->proxy_redo(skb
);
1380 } else if (!sched_next
|| tdif
< sched_next
)
1383 del_timer(&tbl
->proxy_timer
);
1385 mod_timer(&tbl
->proxy_timer
, jiffies
+ sched_next
);
1386 spin_unlock(&tbl
->proxy_queue
.lock
);
1389 void pneigh_enqueue(struct neigh_table
*tbl
, struct neigh_parms
*p
,
1390 struct sk_buff
*skb
)
1392 unsigned long now
= jiffies
;
1394 unsigned long sched_next
= now
+ (prandom_u32() %
1395 NEIGH_VAR(p
, PROXY_DELAY
));
1397 if (tbl
->proxy_queue
.qlen
> NEIGH_VAR(p
, PROXY_QLEN
)) {
1402 NEIGH_CB(skb
)->sched_next
= sched_next
;
1403 NEIGH_CB(skb
)->flags
|= LOCALLY_ENQUEUED
;
1405 spin_lock(&tbl
->proxy_queue
.lock
);
1406 if (del_timer(&tbl
->proxy_timer
)) {
1407 if (time_before(tbl
->proxy_timer
.expires
, sched_next
))
1408 sched_next
= tbl
->proxy_timer
.expires
;
1412 __skb_queue_tail(&tbl
->proxy_queue
, skb
);
1413 mod_timer(&tbl
->proxy_timer
, sched_next
);
1414 spin_unlock(&tbl
->proxy_queue
.lock
);
1416 EXPORT_SYMBOL(pneigh_enqueue
);
1418 static inline struct neigh_parms
*lookup_neigh_parms(struct neigh_table
*tbl
,
1419 struct net
*net
, int ifindex
)
1421 struct neigh_parms
*p
;
1423 list_for_each_entry(p
, &tbl
->parms_list
, list
) {
1424 if ((p
->dev
&& p
->dev
->ifindex
== ifindex
&& net_eq(neigh_parms_net(p
), net
)) ||
1425 (!p
->dev
&& !ifindex
&& net_eq(net
, &init_net
)))
1432 struct neigh_parms
*neigh_parms_alloc(struct net_device
*dev
,
1433 struct neigh_table
*tbl
)
1435 struct neigh_parms
*p
;
1436 struct net
*net
= dev_net(dev
);
1437 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1439 p
= kmemdup(&tbl
->parms
, sizeof(*p
), GFP_KERNEL
);
1442 atomic_set(&p
->refcnt
, 1);
1444 neigh_rand_reach_time(NEIGH_VAR(p
, BASE_REACHABLE_TIME
));
1447 write_pnet(&p
->net
, net
);
1448 p
->sysctl_table
= NULL
;
1450 if (ops
->ndo_neigh_setup
&& ops
->ndo_neigh_setup(dev
, p
)) {
1456 write_lock_bh(&tbl
->lock
);
1457 list_add(&p
->list
, &tbl
->parms
.list
);
1458 write_unlock_bh(&tbl
->lock
);
1460 neigh_parms_data_state_cleanall(p
);
1464 EXPORT_SYMBOL(neigh_parms_alloc
);
1466 static void neigh_rcu_free_parms(struct rcu_head
*head
)
1468 struct neigh_parms
*parms
=
1469 container_of(head
, struct neigh_parms
, rcu_head
);
1471 neigh_parms_put(parms
);
1474 void neigh_parms_release(struct neigh_table
*tbl
, struct neigh_parms
*parms
)
1476 if (!parms
|| parms
== &tbl
->parms
)
1478 write_lock_bh(&tbl
->lock
);
1479 list_del(&parms
->list
);
1481 write_unlock_bh(&tbl
->lock
);
1483 dev_put(parms
->dev
);
1484 call_rcu(&parms
->rcu_head
, neigh_rcu_free_parms
);
1486 EXPORT_SYMBOL(neigh_parms_release
);
1488 static void neigh_parms_destroy(struct neigh_parms
*parms
)
1493 static struct lock_class_key neigh_table_proxy_queue_class
;
1495 static struct neigh_table
*neigh_tables
[NEIGH_NR_TABLES
] __read_mostly
;
1497 void neigh_table_init(int index
, struct neigh_table
*tbl
)
1499 unsigned long now
= jiffies
;
1500 unsigned long phsize
;
1502 INIT_LIST_HEAD(&tbl
->parms_list
);
1503 list_add(&tbl
->parms
.list
, &tbl
->parms_list
);
1504 write_pnet(&tbl
->parms
.net
, &init_net
);
1505 atomic_set(&tbl
->parms
.refcnt
, 1);
1506 tbl
->parms
.reachable_time
=
1507 neigh_rand_reach_time(NEIGH_VAR(&tbl
->parms
, BASE_REACHABLE_TIME
));
1509 tbl
->stats
= alloc_percpu(struct neigh_statistics
);
1511 panic("cannot create neighbour cache statistics");
1513 #ifdef CONFIG_PROC_FS
1514 if (!proc_create_data(tbl
->id
, 0, init_net
.proc_net_stat
,
1515 &neigh_stat_seq_fops
, tbl
))
1516 panic("cannot create neighbour proc dir entry");
1519 RCU_INIT_POINTER(tbl
->nht
, neigh_hash_alloc(3));
1521 phsize
= (PNEIGH_HASHMASK
+ 1) * sizeof(struct pneigh_entry
*);
1522 tbl
->phash_buckets
= kzalloc(phsize
, GFP_KERNEL
);
1524 if (!tbl
->nht
|| !tbl
->phash_buckets
)
1525 panic("cannot allocate neighbour cache hashes");
1527 if (!tbl
->entry_size
)
1528 tbl
->entry_size
= ALIGN(offsetof(struct neighbour
, primary_key
) +
1529 tbl
->key_len
, NEIGH_PRIV_ALIGN
);
1531 WARN_ON(tbl
->entry_size
% NEIGH_PRIV_ALIGN
);
1533 rwlock_init(&tbl
->lock
);
1534 INIT_DEFERRABLE_WORK(&tbl
->gc_work
, neigh_periodic_work
);
1535 queue_delayed_work(system_power_efficient_wq
, &tbl
->gc_work
,
1536 tbl
->parms
.reachable_time
);
1537 setup_timer(&tbl
->proxy_timer
, neigh_proxy_process
, (unsigned long)tbl
);
1538 skb_queue_head_init_class(&tbl
->proxy_queue
,
1539 &neigh_table_proxy_queue_class
);
1541 tbl
->last_flush
= now
;
1542 tbl
->last_rand
= now
+ tbl
->parms
.reachable_time
* 20;
1544 neigh_tables
[index
] = tbl
;
1546 EXPORT_SYMBOL(neigh_table_init
);
1548 int neigh_table_clear(int index
, struct neigh_table
*tbl
)
1550 neigh_tables
[index
] = NULL
;
1551 /* It is not clean... Fix it to unload IPv6 module safely */
1552 cancel_delayed_work_sync(&tbl
->gc_work
);
1553 del_timer_sync(&tbl
->proxy_timer
);
1554 pneigh_queue_purge(&tbl
->proxy_queue
);
1555 neigh_ifdown(tbl
, NULL
);
1556 if (atomic_read(&tbl
->entries
))
1557 pr_crit("neighbour leakage\n");
1559 call_rcu(&rcu_dereference_protected(tbl
->nht
, 1)->rcu
,
1560 neigh_hash_free_rcu
);
1563 kfree(tbl
->phash_buckets
);
1564 tbl
->phash_buckets
= NULL
;
1566 remove_proc_entry(tbl
->id
, init_net
.proc_net_stat
);
1568 free_percpu(tbl
->stats
);
1573 EXPORT_SYMBOL(neigh_table_clear
);
1575 static struct neigh_table
*neigh_find_table(int family
)
1577 struct neigh_table
*tbl
= NULL
;
1581 tbl
= neigh_tables
[NEIGH_ARP_TABLE
];
1584 tbl
= neigh_tables
[NEIGH_ND_TABLE
];
1587 tbl
= neigh_tables
[NEIGH_DN_TABLE
];
1594 static int neigh_delete(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1596 struct net
*net
= sock_net(skb
->sk
);
1598 struct nlattr
*dst_attr
;
1599 struct neigh_table
*tbl
;
1600 struct neighbour
*neigh
;
1601 struct net_device
*dev
= NULL
;
1605 if (nlmsg_len(nlh
) < sizeof(*ndm
))
1608 dst_attr
= nlmsg_find_attr(nlh
, sizeof(*ndm
), NDA_DST
);
1609 if (dst_attr
== NULL
)
1612 ndm
= nlmsg_data(nlh
);
1613 if (ndm
->ndm_ifindex
) {
1614 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
1621 tbl
= neigh_find_table(ndm
->ndm_family
);
1623 return -EAFNOSUPPORT
;
1625 if (nla_len(dst_attr
) < tbl
->key_len
)
1628 if (ndm
->ndm_flags
& NTF_PROXY
) {
1629 err
= pneigh_delete(tbl
, net
, nla_data(dst_attr
), dev
);
1636 neigh
= neigh_lookup(tbl
, nla_data(dst_attr
), dev
);
1637 if (neigh
== NULL
) {
1642 err
= neigh_update(neigh
, NULL
, NUD_FAILED
,
1643 NEIGH_UPDATE_F_OVERRIDE
|
1644 NEIGH_UPDATE_F_ADMIN
);
1645 neigh_release(neigh
);
1651 static int neigh_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1653 int flags
= NEIGH_UPDATE_F_ADMIN
| NEIGH_UPDATE_F_OVERRIDE
;
1654 struct net
*net
= sock_net(skb
->sk
);
1656 struct nlattr
*tb
[NDA_MAX
+1];
1657 struct neigh_table
*tbl
;
1658 struct net_device
*dev
= NULL
;
1659 struct neighbour
*neigh
;
1664 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
1669 if (tb
[NDA_DST
] == NULL
)
1672 ndm
= nlmsg_data(nlh
);
1673 if (ndm
->ndm_ifindex
) {
1674 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
1680 if (tb
[NDA_LLADDR
] && nla_len(tb
[NDA_LLADDR
]) < dev
->addr_len
)
1684 tbl
= neigh_find_table(ndm
->ndm_family
);
1686 return -EAFNOSUPPORT
;
1688 if (nla_len(tb
[NDA_DST
]) < tbl
->key_len
)
1690 dst
= nla_data(tb
[NDA_DST
]);
1691 lladdr
= tb
[NDA_LLADDR
] ? nla_data(tb
[NDA_LLADDR
]) : NULL
;
1693 if (ndm
->ndm_flags
& NTF_PROXY
) {
1694 struct pneigh_entry
*pn
;
1697 pn
= pneigh_lookup(tbl
, net
, dst
, dev
, 1);
1699 pn
->flags
= ndm
->ndm_flags
;
1708 neigh
= neigh_lookup(tbl
, dst
, dev
);
1709 if (neigh
== NULL
) {
1710 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1715 neigh
= __neigh_lookup_errno(tbl
, dst
, dev
);
1716 if (IS_ERR(neigh
)) {
1717 err
= PTR_ERR(neigh
);
1721 if (nlh
->nlmsg_flags
& NLM_F_EXCL
) {
1723 neigh_release(neigh
);
1727 if (!(nlh
->nlmsg_flags
& NLM_F_REPLACE
))
1728 flags
&= ~NEIGH_UPDATE_F_OVERRIDE
;
1731 if (ndm
->ndm_flags
& NTF_USE
) {
1732 neigh_event_send(neigh
, NULL
);
1735 err
= neigh_update(neigh
, lladdr
, ndm
->ndm_state
, flags
);
1736 neigh_release(neigh
);
1742 static int neightbl_fill_parms(struct sk_buff
*skb
, struct neigh_parms
*parms
)
1744 struct nlattr
*nest
;
1746 nest
= nla_nest_start(skb
, NDTA_PARMS
);
1751 nla_put_u32(skb
, NDTPA_IFINDEX
, parms
->dev
->ifindex
)) ||
1752 nla_put_u32(skb
, NDTPA_REFCNT
, atomic_read(&parms
->refcnt
)) ||
1753 nla_put_u32(skb
, NDTPA_QUEUE_LENBYTES
,
1754 NEIGH_VAR(parms
, QUEUE_LEN_BYTES
)) ||
1755 /* approximative value for deprecated QUEUE_LEN (in packets) */
1756 nla_put_u32(skb
, NDTPA_QUEUE_LEN
,
1757 NEIGH_VAR(parms
, QUEUE_LEN_BYTES
) / SKB_TRUESIZE(ETH_FRAME_LEN
)) ||
1758 nla_put_u32(skb
, NDTPA_PROXY_QLEN
, NEIGH_VAR(parms
, PROXY_QLEN
)) ||
1759 nla_put_u32(skb
, NDTPA_APP_PROBES
, NEIGH_VAR(parms
, APP_PROBES
)) ||
1760 nla_put_u32(skb
, NDTPA_UCAST_PROBES
,
1761 NEIGH_VAR(parms
, UCAST_PROBES
)) ||
1762 nla_put_u32(skb
, NDTPA_MCAST_PROBES
,
1763 NEIGH_VAR(parms
, MCAST_PROBES
)) ||
1764 nla_put_u32(skb
, NDTPA_MCAST_REPROBES
,
1765 NEIGH_VAR(parms
, MCAST_REPROBES
)) ||
1766 nla_put_msecs(skb
, NDTPA_REACHABLE_TIME
, parms
->reachable_time
,
1768 nla_put_msecs(skb
, NDTPA_BASE_REACHABLE_TIME
,
1769 NEIGH_VAR(parms
, BASE_REACHABLE_TIME
), NDTPA_PAD
) ||
1770 nla_put_msecs(skb
, NDTPA_GC_STALETIME
,
1771 NEIGH_VAR(parms
, GC_STALETIME
), NDTPA_PAD
) ||
1772 nla_put_msecs(skb
, NDTPA_DELAY_PROBE_TIME
,
1773 NEIGH_VAR(parms
, DELAY_PROBE_TIME
), NDTPA_PAD
) ||
1774 nla_put_msecs(skb
, NDTPA_RETRANS_TIME
,
1775 NEIGH_VAR(parms
, RETRANS_TIME
), NDTPA_PAD
) ||
1776 nla_put_msecs(skb
, NDTPA_ANYCAST_DELAY
,
1777 NEIGH_VAR(parms
, ANYCAST_DELAY
), NDTPA_PAD
) ||
1778 nla_put_msecs(skb
, NDTPA_PROXY_DELAY
,
1779 NEIGH_VAR(parms
, PROXY_DELAY
), NDTPA_PAD
) ||
1780 nla_put_msecs(skb
, NDTPA_LOCKTIME
,
1781 NEIGH_VAR(parms
, LOCKTIME
), NDTPA_PAD
))
1782 goto nla_put_failure
;
1783 return nla_nest_end(skb
, nest
);
1786 nla_nest_cancel(skb
, nest
);
1790 static int neightbl_fill_info(struct sk_buff
*skb
, struct neigh_table
*tbl
,
1791 u32 pid
, u32 seq
, int type
, int flags
)
1793 struct nlmsghdr
*nlh
;
1794 struct ndtmsg
*ndtmsg
;
1796 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1800 ndtmsg
= nlmsg_data(nlh
);
1802 read_lock_bh(&tbl
->lock
);
1803 ndtmsg
->ndtm_family
= tbl
->family
;
1804 ndtmsg
->ndtm_pad1
= 0;
1805 ndtmsg
->ndtm_pad2
= 0;
1807 if (nla_put_string(skb
, NDTA_NAME
, tbl
->id
) ||
1808 nla_put_msecs(skb
, NDTA_GC_INTERVAL
, tbl
->gc_interval
, NDTA_PAD
) ||
1809 nla_put_u32(skb
, NDTA_THRESH1
, tbl
->gc_thresh1
) ||
1810 nla_put_u32(skb
, NDTA_THRESH2
, tbl
->gc_thresh2
) ||
1811 nla_put_u32(skb
, NDTA_THRESH3
, tbl
->gc_thresh3
))
1812 goto nla_put_failure
;
1814 unsigned long now
= jiffies
;
1815 unsigned int flush_delta
= now
- tbl
->last_flush
;
1816 unsigned int rand_delta
= now
- tbl
->last_rand
;
1817 struct neigh_hash_table
*nht
;
1818 struct ndt_config ndc
= {
1819 .ndtc_key_len
= tbl
->key_len
,
1820 .ndtc_entry_size
= tbl
->entry_size
,
1821 .ndtc_entries
= atomic_read(&tbl
->entries
),
1822 .ndtc_last_flush
= jiffies_to_msecs(flush_delta
),
1823 .ndtc_last_rand
= jiffies_to_msecs(rand_delta
),
1824 .ndtc_proxy_qlen
= tbl
->proxy_queue
.qlen
,
1828 nht
= rcu_dereference_bh(tbl
->nht
);
1829 ndc
.ndtc_hash_rnd
= nht
->hash_rnd
[0];
1830 ndc
.ndtc_hash_mask
= ((1 << nht
->hash_shift
) - 1);
1831 rcu_read_unlock_bh();
1833 if (nla_put(skb
, NDTA_CONFIG
, sizeof(ndc
), &ndc
))
1834 goto nla_put_failure
;
1839 struct ndt_stats ndst
;
1841 memset(&ndst
, 0, sizeof(ndst
));
1843 for_each_possible_cpu(cpu
) {
1844 struct neigh_statistics
*st
;
1846 st
= per_cpu_ptr(tbl
->stats
, cpu
);
1847 ndst
.ndts_allocs
+= st
->allocs
;
1848 ndst
.ndts_destroys
+= st
->destroys
;
1849 ndst
.ndts_hash_grows
+= st
->hash_grows
;
1850 ndst
.ndts_res_failed
+= st
->res_failed
;
1851 ndst
.ndts_lookups
+= st
->lookups
;
1852 ndst
.ndts_hits
+= st
->hits
;
1853 ndst
.ndts_rcv_probes_mcast
+= st
->rcv_probes_mcast
;
1854 ndst
.ndts_rcv_probes_ucast
+= st
->rcv_probes_ucast
;
1855 ndst
.ndts_periodic_gc_runs
+= st
->periodic_gc_runs
;
1856 ndst
.ndts_forced_gc_runs
+= st
->forced_gc_runs
;
1857 ndst
.ndts_table_fulls
+= st
->table_fulls
;
1860 if (nla_put(skb
, NDTA_STATS
, sizeof(ndst
), &ndst
))
1861 goto nla_put_failure
;
1864 BUG_ON(tbl
->parms
.dev
);
1865 if (neightbl_fill_parms(skb
, &tbl
->parms
) < 0)
1866 goto nla_put_failure
;
1868 read_unlock_bh(&tbl
->lock
);
1869 nlmsg_end(skb
, nlh
);
1873 read_unlock_bh(&tbl
->lock
);
1874 nlmsg_cancel(skb
, nlh
);
1878 static int neightbl_fill_param_info(struct sk_buff
*skb
,
1879 struct neigh_table
*tbl
,
1880 struct neigh_parms
*parms
,
1881 u32 pid
, u32 seq
, int type
,
1884 struct ndtmsg
*ndtmsg
;
1885 struct nlmsghdr
*nlh
;
1887 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1891 ndtmsg
= nlmsg_data(nlh
);
1893 read_lock_bh(&tbl
->lock
);
1894 ndtmsg
->ndtm_family
= tbl
->family
;
1895 ndtmsg
->ndtm_pad1
= 0;
1896 ndtmsg
->ndtm_pad2
= 0;
1898 if (nla_put_string(skb
, NDTA_NAME
, tbl
->id
) < 0 ||
1899 neightbl_fill_parms(skb
, parms
) < 0)
1902 read_unlock_bh(&tbl
->lock
);
1903 nlmsg_end(skb
, nlh
);
1906 read_unlock_bh(&tbl
->lock
);
1907 nlmsg_cancel(skb
, nlh
);
1911 static const struct nla_policy nl_neightbl_policy
[NDTA_MAX
+1] = {
1912 [NDTA_NAME
] = { .type
= NLA_STRING
},
1913 [NDTA_THRESH1
] = { .type
= NLA_U32
},
1914 [NDTA_THRESH2
] = { .type
= NLA_U32
},
1915 [NDTA_THRESH3
] = { .type
= NLA_U32
},
1916 [NDTA_GC_INTERVAL
] = { .type
= NLA_U64
},
1917 [NDTA_PARMS
] = { .type
= NLA_NESTED
},
1920 static const struct nla_policy nl_ntbl_parm_policy
[NDTPA_MAX
+1] = {
1921 [NDTPA_IFINDEX
] = { .type
= NLA_U32
},
1922 [NDTPA_QUEUE_LEN
] = { .type
= NLA_U32
},
1923 [NDTPA_PROXY_QLEN
] = { .type
= NLA_U32
},
1924 [NDTPA_APP_PROBES
] = { .type
= NLA_U32
},
1925 [NDTPA_UCAST_PROBES
] = { .type
= NLA_U32
},
1926 [NDTPA_MCAST_PROBES
] = { .type
= NLA_U32
},
1927 [NDTPA_MCAST_REPROBES
] = { .type
= NLA_U32
},
1928 [NDTPA_BASE_REACHABLE_TIME
] = { .type
= NLA_U64
},
1929 [NDTPA_GC_STALETIME
] = { .type
= NLA_U64
},
1930 [NDTPA_DELAY_PROBE_TIME
] = { .type
= NLA_U64
},
1931 [NDTPA_RETRANS_TIME
] = { .type
= NLA_U64
},
1932 [NDTPA_ANYCAST_DELAY
] = { .type
= NLA_U64
},
1933 [NDTPA_PROXY_DELAY
] = { .type
= NLA_U64
},
1934 [NDTPA_LOCKTIME
] = { .type
= NLA_U64
},
1937 static int neightbl_set(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1939 struct net
*net
= sock_net(skb
->sk
);
1940 struct neigh_table
*tbl
;
1941 struct ndtmsg
*ndtmsg
;
1942 struct nlattr
*tb
[NDTA_MAX
+1];
1946 err
= nlmsg_parse(nlh
, sizeof(*ndtmsg
), tb
, NDTA_MAX
,
1947 nl_neightbl_policy
);
1951 if (tb
[NDTA_NAME
] == NULL
) {
1956 ndtmsg
= nlmsg_data(nlh
);
1958 for (tidx
= 0; tidx
< NEIGH_NR_TABLES
; tidx
++) {
1959 tbl
= neigh_tables
[tidx
];
1962 if (ndtmsg
->ndtm_family
&& tbl
->family
!= ndtmsg
->ndtm_family
)
1964 if (nla_strcmp(tb
[NDTA_NAME
], tbl
->id
) == 0) {
1974 * We acquire tbl->lock to be nice to the periodic timers and
1975 * make sure they always see a consistent set of values.
1977 write_lock_bh(&tbl
->lock
);
1979 if (tb
[NDTA_PARMS
]) {
1980 struct nlattr
*tbp
[NDTPA_MAX
+1];
1981 struct neigh_parms
*p
;
1984 err
= nla_parse_nested(tbp
, NDTPA_MAX
, tb
[NDTA_PARMS
],
1985 nl_ntbl_parm_policy
);
1987 goto errout_tbl_lock
;
1989 if (tbp
[NDTPA_IFINDEX
])
1990 ifindex
= nla_get_u32(tbp
[NDTPA_IFINDEX
]);
1992 p
= lookup_neigh_parms(tbl
, net
, ifindex
);
1995 goto errout_tbl_lock
;
1998 for (i
= 1; i
<= NDTPA_MAX
; i
++) {
2003 case NDTPA_QUEUE_LEN
:
2004 NEIGH_VAR_SET(p
, QUEUE_LEN_BYTES
,
2005 nla_get_u32(tbp
[i
]) *
2006 SKB_TRUESIZE(ETH_FRAME_LEN
));
2008 case NDTPA_QUEUE_LENBYTES
:
2009 NEIGH_VAR_SET(p
, QUEUE_LEN_BYTES
,
2010 nla_get_u32(tbp
[i
]));
2012 case NDTPA_PROXY_QLEN
:
2013 NEIGH_VAR_SET(p
, PROXY_QLEN
,
2014 nla_get_u32(tbp
[i
]));
2016 case NDTPA_APP_PROBES
:
2017 NEIGH_VAR_SET(p
, APP_PROBES
,
2018 nla_get_u32(tbp
[i
]));
2020 case NDTPA_UCAST_PROBES
:
2021 NEIGH_VAR_SET(p
, UCAST_PROBES
,
2022 nla_get_u32(tbp
[i
]));
2024 case NDTPA_MCAST_PROBES
:
2025 NEIGH_VAR_SET(p
, MCAST_PROBES
,
2026 nla_get_u32(tbp
[i
]));
2028 case NDTPA_MCAST_REPROBES
:
2029 NEIGH_VAR_SET(p
, MCAST_REPROBES
,
2030 nla_get_u32(tbp
[i
]));
2032 case NDTPA_BASE_REACHABLE_TIME
:
2033 NEIGH_VAR_SET(p
, BASE_REACHABLE_TIME
,
2034 nla_get_msecs(tbp
[i
]));
2035 /* update reachable_time as well, otherwise, the change will
2036 * only be effective after the next time neigh_periodic_work
2037 * decides to recompute it (can be multiple minutes)
2040 neigh_rand_reach_time(NEIGH_VAR(p
, BASE_REACHABLE_TIME
));
2042 case NDTPA_GC_STALETIME
:
2043 NEIGH_VAR_SET(p
, GC_STALETIME
,
2044 nla_get_msecs(tbp
[i
]));
2046 case NDTPA_DELAY_PROBE_TIME
:
2047 NEIGH_VAR_SET(p
, DELAY_PROBE_TIME
,
2048 nla_get_msecs(tbp
[i
]));
2050 case NDTPA_RETRANS_TIME
:
2051 NEIGH_VAR_SET(p
, RETRANS_TIME
,
2052 nla_get_msecs(tbp
[i
]));
2054 case NDTPA_ANYCAST_DELAY
:
2055 NEIGH_VAR_SET(p
, ANYCAST_DELAY
,
2056 nla_get_msecs(tbp
[i
]));
2058 case NDTPA_PROXY_DELAY
:
2059 NEIGH_VAR_SET(p
, PROXY_DELAY
,
2060 nla_get_msecs(tbp
[i
]));
2062 case NDTPA_LOCKTIME
:
2063 NEIGH_VAR_SET(p
, LOCKTIME
,
2064 nla_get_msecs(tbp
[i
]));
2071 if ((tb
[NDTA_THRESH1
] || tb
[NDTA_THRESH2
] ||
2072 tb
[NDTA_THRESH3
] || tb
[NDTA_GC_INTERVAL
]) &&
2073 !net_eq(net
, &init_net
))
2074 goto errout_tbl_lock
;
2076 if (tb
[NDTA_THRESH1
])
2077 tbl
->gc_thresh1
= nla_get_u32(tb
[NDTA_THRESH1
]);
2079 if (tb
[NDTA_THRESH2
])
2080 tbl
->gc_thresh2
= nla_get_u32(tb
[NDTA_THRESH2
]);
2082 if (tb
[NDTA_THRESH3
])
2083 tbl
->gc_thresh3
= nla_get_u32(tb
[NDTA_THRESH3
]);
2085 if (tb
[NDTA_GC_INTERVAL
])
2086 tbl
->gc_interval
= nla_get_msecs(tb
[NDTA_GC_INTERVAL
]);
2091 write_unlock_bh(&tbl
->lock
);
2096 static int neightbl_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2098 struct net
*net
= sock_net(skb
->sk
);
2099 int family
, tidx
, nidx
= 0;
2100 int tbl_skip
= cb
->args
[0];
2101 int neigh_skip
= cb
->args
[1];
2102 struct neigh_table
*tbl
;
2104 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
2106 for (tidx
= 0; tidx
< NEIGH_NR_TABLES
; tidx
++) {
2107 struct neigh_parms
*p
;
2109 tbl
= neigh_tables
[tidx
];
2113 if (tidx
< tbl_skip
|| (family
&& tbl
->family
!= family
))
2116 if (neightbl_fill_info(skb
, tbl
, NETLINK_CB(cb
->skb
).portid
,
2117 cb
->nlh
->nlmsg_seq
, RTM_NEWNEIGHTBL
,
2122 p
= list_next_entry(&tbl
->parms
, list
);
2123 list_for_each_entry_from(p
, &tbl
->parms_list
, list
) {
2124 if (!net_eq(neigh_parms_net(p
), net
))
2127 if (nidx
< neigh_skip
)
2130 if (neightbl_fill_param_info(skb
, tbl
, p
,
2131 NETLINK_CB(cb
->skb
).portid
,
2149 static int neigh_fill_info(struct sk_buff
*skb
, struct neighbour
*neigh
,
2150 u32 pid
, u32 seq
, int type
, unsigned int flags
)
2152 unsigned long now
= jiffies
;
2153 struct nda_cacheinfo ci
;
2154 struct nlmsghdr
*nlh
;
2157 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), flags
);
2161 ndm
= nlmsg_data(nlh
);
2162 ndm
->ndm_family
= neigh
->ops
->family
;
2165 ndm
->ndm_flags
= neigh
->flags
;
2166 ndm
->ndm_type
= neigh
->type
;
2167 ndm
->ndm_ifindex
= neigh
->dev
->ifindex
;
2169 if (nla_put(skb
, NDA_DST
, neigh
->tbl
->key_len
, neigh
->primary_key
))
2170 goto nla_put_failure
;
2172 read_lock_bh(&neigh
->lock
);
2173 ndm
->ndm_state
= neigh
->nud_state
;
2174 if (neigh
->nud_state
& NUD_VALID
) {
2175 char haddr
[MAX_ADDR_LEN
];
2177 neigh_ha_snapshot(haddr
, neigh
, neigh
->dev
);
2178 if (nla_put(skb
, NDA_LLADDR
, neigh
->dev
->addr_len
, haddr
) < 0) {
2179 read_unlock_bh(&neigh
->lock
);
2180 goto nla_put_failure
;
2184 ci
.ndm_used
= jiffies_to_clock_t(now
- neigh
->used
);
2185 ci
.ndm_confirmed
= jiffies_to_clock_t(now
- neigh
->confirmed
);
2186 ci
.ndm_updated
= jiffies_to_clock_t(now
- neigh
->updated
);
2187 ci
.ndm_refcnt
= atomic_read(&neigh
->refcnt
) - 1;
2188 read_unlock_bh(&neigh
->lock
);
2190 if (nla_put_u32(skb
, NDA_PROBES
, atomic_read(&neigh
->probes
)) ||
2191 nla_put(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
))
2192 goto nla_put_failure
;
2194 nlmsg_end(skb
, nlh
);
2198 nlmsg_cancel(skb
, nlh
);
2202 static int pneigh_fill_info(struct sk_buff
*skb
, struct pneigh_entry
*pn
,
2203 u32 pid
, u32 seq
, int type
, unsigned int flags
,
2204 struct neigh_table
*tbl
)
2206 struct nlmsghdr
*nlh
;
2209 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), flags
);
2213 ndm
= nlmsg_data(nlh
);
2214 ndm
->ndm_family
= tbl
->family
;
2217 ndm
->ndm_flags
= pn
->flags
| NTF_PROXY
;
2218 ndm
->ndm_type
= RTN_UNICAST
;
2219 ndm
->ndm_ifindex
= pn
->dev
? pn
->dev
->ifindex
: 0;
2220 ndm
->ndm_state
= NUD_NONE
;
2222 if (nla_put(skb
, NDA_DST
, tbl
->key_len
, pn
->key
))
2223 goto nla_put_failure
;
2225 nlmsg_end(skb
, nlh
);
2229 nlmsg_cancel(skb
, nlh
);
2233 static void neigh_update_notify(struct neighbour
*neigh
)
2235 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, neigh
);
2236 __neigh_notify(neigh
, RTM_NEWNEIGH
, 0);
2239 static bool neigh_master_filtered(struct net_device
*dev
, int master_idx
)
2241 struct net_device
*master
;
2246 master
= netdev_master_upper_dev_get(dev
);
2247 if (!master
|| master
->ifindex
!= master_idx
)
2253 static bool neigh_ifindex_filtered(struct net_device
*dev
, int filter_idx
)
2255 if (filter_idx
&& dev
->ifindex
!= filter_idx
)
2261 static int neigh_dump_table(struct neigh_table
*tbl
, struct sk_buff
*skb
,
2262 struct netlink_callback
*cb
)
2264 struct net
*net
= sock_net(skb
->sk
);
2265 const struct nlmsghdr
*nlh
= cb
->nlh
;
2266 struct nlattr
*tb
[NDA_MAX
+ 1];
2267 struct neighbour
*n
;
2268 int rc
, h
, s_h
= cb
->args
[1];
2269 int idx
, s_idx
= idx
= cb
->args
[2];
2270 struct neigh_hash_table
*nht
;
2271 int filter_master_idx
= 0, filter_idx
= 0;
2272 unsigned int flags
= NLM_F_MULTI
;
2275 err
= nlmsg_parse(nlh
, sizeof(struct ndmsg
), tb
, NDA_MAX
, NULL
);
2277 if (tb
[NDA_IFINDEX
])
2278 filter_idx
= nla_get_u32(tb
[NDA_IFINDEX
]);
2281 filter_master_idx
= nla_get_u32(tb
[NDA_MASTER
]);
2283 if (filter_idx
|| filter_master_idx
)
2284 flags
|= NLM_F_DUMP_FILTERED
;
2288 nht
= rcu_dereference_bh(tbl
->nht
);
2290 for (h
= s_h
; h
< (1 << nht
->hash_shift
); h
++) {
2293 for (n
= rcu_dereference_bh(nht
->hash_buckets
[h
]), idx
= 0;
2295 n
= rcu_dereference_bh(n
->next
)) {
2296 if (!net_eq(dev_net(n
->dev
), net
))
2298 if (neigh_ifindex_filtered(n
->dev
, filter_idx
))
2300 if (neigh_master_filtered(n
->dev
, filter_master_idx
))
2304 if (neigh_fill_info(skb
, n
, NETLINK_CB(cb
->skb
).portid
,
2317 rcu_read_unlock_bh();
2323 static int pneigh_dump_table(struct neigh_table
*tbl
, struct sk_buff
*skb
,
2324 struct netlink_callback
*cb
)
2326 struct pneigh_entry
*n
;
2327 struct net
*net
= sock_net(skb
->sk
);
2328 int rc
, h
, s_h
= cb
->args
[3];
2329 int idx
, s_idx
= idx
= cb
->args
[4];
2331 read_lock_bh(&tbl
->lock
);
2333 for (h
= s_h
; h
<= PNEIGH_HASHMASK
; h
++) {
2336 for (n
= tbl
->phash_buckets
[h
], idx
= 0; n
; n
= n
->next
) {
2337 if (pneigh_net(n
) != net
)
2341 if (pneigh_fill_info(skb
, n
, NETLINK_CB(cb
->skb
).portid
,
2344 NLM_F_MULTI
, tbl
) < 0) {
2345 read_unlock_bh(&tbl
->lock
);
2354 read_unlock_bh(&tbl
->lock
);
2363 static int neigh_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2365 struct neigh_table
*tbl
;
2370 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
2372 /* check for full ndmsg structure presence, family member is
2373 * the same for both structures
2375 if (nlmsg_len(cb
->nlh
) >= sizeof(struct ndmsg
) &&
2376 ((struct ndmsg
*) nlmsg_data(cb
->nlh
))->ndm_flags
== NTF_PROXY
)
2381 for (t
= 0; t
< NEIGH_NR_TABLES
; t
++) {
2382 tbl
= neigh_tables
[t
];
2386 if (t
< s_t
|| (family
&& tbl
->family
!= family
))
2389 memset(&cb
->args
[1], 0, sizeof(cb
->args
) -
2390 sizeof(cb
->args
[0]));
2392 err
= pneigh_dump_table(tbl
, skb
, cb
);
2394 err
= neigh_dump_table(tbl
, skb
, cb
);
2403 void neigh_for_each(struct neigh_table
*tbl
, void (*cb
)(struct neighbour
*, void *), void *cookie
)
2406 struct neigh_hash_table
*nht
;
2409 nht
= rcu_dereference_bh(tbl
->nht
);
2411 read_lock(&tbl
->lock
); /* avoid resizes */
2412 for (chain
= 0; chain
< (1 << nht
->hash_shift
); chain
++) {
2413 struct neighbour
*n
;
2415 for (n
= rcu_dereference_bh(nht
->hash_buckets
[chain
]);
2417 n
= rcu_dereference_bh(n
->next
))
2420 read_unlock(&tbl
->lock
);
2421 rcu_read_unlock_bh();
2423 EXPORT_SYMBOL(neigh_for_each
);
2425 /* The tbl->lock must be held as a writer and BH disabled. */
2426 void __neigh_for_each_release(struct neigh_table
*tbl
,
2427 int (*cb
)(struct neighbour
*))
2430 struct neigh_hash_table
*nht
;
2432 nht
= rcu_dereference_protected(tbl
->nht
,
2433 lockdep_is_held(&tbl
->lock
));
2434 for (chain
= 0; chain
< (1 << nht
->hash_shift
); chain
++) {
2435 struct neighbour
*n
;
2436 struct neighbour __rcu
**np
;
2438 np
= &nht
->hash_buckets
[chain
];
2439 while ((n
= rcu_dereference_protected(*np
,
2440 lockdep_is_held(&tbl
->lock
))) != NULL
) {
2443 write_lock(&n
->lock
);
2446 rcu_assign_pointer(*np
,
2447 rcu_dereference_protected(n
->next
,
2448 lockdep_is_held(&tbl
->lock
)));
2452 write_unlock(&n
->lock
);
2454 neigh_cleanup_and_release(n
);
2458 EXPORT_SYMBOL(__neigh_for_each_release
);
2460 int neigh_xmit(int index
, struct net_device
*dev
,
2461 const void *addr
, struct sk_buff
*skb
)
2463 int err
= -EAFNOSUPPORT
;
2464 if (likely(index
< NEIGH_NR_TABLES
)) {
2465 struct neigh_table
*tbl
;
2466 struct neighbour
*neigh
;
2468 tbl
= neigh_tables
[index
];
2471 neigh
= __neigh_lookup_noref(tbl
, addr
, dev
);
2473 neigh
= __neigh_create(tbl
, addr
, dev
, false);
2474 err
= PTR_ERR(neigh
);
2477 err
= neigh
->output(neigh
, skb
);
2479 else if (index
== NEIGH_LINK_TABLE
) {
2480 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
2481 addr
, NULL
, skb
->len
);
2484 err
= dev_queue_xmit(skb
);
2492 EXPORT_SYMBOL(neigh_xmit
);
2494 #ifdef CONFIG_PROC_FS
2496 static struct neighbour
*neigh_get_first(struct seq_file
*seq
)
2498 struct neigh_seq_state
*state
= seq
->private;
2499 struct net
*net
= seq_file_net(seq
);
2500 struct neigh_hash_table
*nht
= state
->nht
;
2501 struct neighbour
*n
= NULL
;
2502 int bucket
= state
->bucket
;
2504 state
->flags
&= ~NEIGH_SEQ_IS_PNEIGH
;
2505 for (bucket
= 0; bucket
< (1 << nht
->hash_shift
); bucket
++) {
2506 n
= rcu_dereference_bh(nht
->hash_buckets
[bucket
]);
2509 if (!net_eq(dev_net(n
->dev
), net
))
2511 if (state
->neigh_sub_iter
) {
2515 v
= state
->neigh_sub_iter(state
, n
, &fakep
);
2519 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2521 if (n
->nud_state
& ~NUD_NOARP
)
2524 n
= rcu_dereference_bh(n
->next
);
2530 state
->bucket
= bucket
;
2535 static struct neighbour
*neigh_get_next(struct seq_file
*seq
,
2536 struct neighbour
*n
,
2539 struct neigh_seq_state
*state
= seq
->private;
2540 struct net
*net
= seq_file_net(seq
);
2541 struct neigh_hash_table
*nht
= state
->nht
;
2543 if (state
->neigh_sub_iter
) {
2544 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2548 n
= rcu_dereference_bh(n
->next
);
2552 if (!net_eq(dev_net(n
->dev
), net
))
2554 if (state
->neigh_sub_iter
) {
2555 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2560 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2563 if (n
->nud_state
& ~NUD_NOARP
)
2566 n
= rcu_dereference_bh(n
->next
);
2572 if (++state
->bucket
>= (1 << nht
->hash_shift
))
2575 n
= rcu_dereference_bh(nht
->hash_buckets
[state
->bucket
]);
2583 static struct neighbour
*neigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2585 struct neighbour
*n
= neigh_get_first(seq
);
2590 n
= neigh_get_next(seq
, n
, pos
);
2595 return *pos
? NULL
: n
;
2598 static struct pneigh_entry
*pneigh_get_first(struct seq_file
*seq
)
2600 struct neigh_seq_state
*state
= seq
->private;
2601 struct net
*net
= seq_file_net(seq
);
2602 struct neigh_table
*tbl
= state
->tbl
;
2603 struct pneigh_entry
*pn
= NULL
;
2604 int bucket
= state
->bucket
;
2606 state
->flags
|= NEIGH_SEQ_IS_PNEIGH
;
2607 for (bucket
= 0; bucket
<= PNEIGH_HASHMASK
; bucket
++) {
2608 pn
= tbl
->phash_buckets
[bucket
];
2609 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2614 state
->bucket
= bucket
;
2619 static struct pneigh_entry
*pneigh_get_next(struct seq_file
*seq
,
2620 struct pneigh_entry
*pn
,
2623 struct neigh_seq_state
*state
= seq
->private;
2624 struct net
*net
= seq_file_net(seq
);
2625 struct neigh_table
*tbl
= state
->tbl
;
2629 } while (pn
&& !net_eq(pneigh_net(pn
), net
));
2632 if (++state
->bucket
> PNEIGH_HASHMASK
)
2634 pn
= tbl
->phash_buckets
[state
->bucket
];
2635 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2647 static struct pneigh_entry
*pneigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2649 struct pneigh_entry
*pn
= pneigh_get_first(seq
);
2654 pn
= pneigh_get_next(seq
, pn
, pos
);
2659 return *pos
? NULL
: pn
;
2662 static void *neigh_get_idx_any(struct seq_file
*seq
, loff_t
*pos
)
2664 struct neigh_seq_state
*state
= seq
->private;
2666 loff_t idxpos
= *pos
;
2668 rc
= neigh_get_idx(seq
, &idxpos
);
2669 if (!rc
&& !(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2670 rc
= pneigh_get_idx(seq
, &idxpos
);
2675 void *neigh_seq_start(struct seq_file
*seq
, loff_t
*pos
, struct neigh_table
*tbl
, unsigned int neigh_seq_flags
)
2678 struct neigh_seq_state
*state
= seq
->private;
2682 state
->flags
= (neigh_seq_flags
& ~NEIGH_SEQ_IS_PNEIGH
);
2685 state
->nht
= rcu_dereference_bh(tbl
->nht
);
2687 return *pos
? neigh_get_idx_any(seq
, pos
) : SEQ_START_TOKEN
;
2689 EXPORT_SYMBOL(neigh_seq_start
);
2691 void *neigh_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2693 struct neigh_seq_state
*state
;
2696 if (v
== SEQ_START_TOKEN
) {
2697 rc
= neigh_get_first(seq
);
2701 state
= seq
->private;
2702 if (!(state
->flags
& NEIGH_SEQ_IS_PNEIGH
)) {
2703 rc
= neigh_get_next(seq
, v
, NULL
);
2706 if (!(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2707 rc
= pneigh_get_first(seq
);
2709 BUG_ON(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
);
2710 rc
= pneigh_get_next(seq
, v
, NULL
);
2716 EXPORT_SYMBOL(neigh_seq_next
);
2718 void neigh_seq_stop(struct seq_file
*seq
, void *v
)
2721 rcu_read_unlock_bh();
2723 EXPORT_SYMBOL(neigh_seq_stop
);
2725 /* statistics via seq_file */
2727 static void *neigh_stat_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2729 struct neigh_table
*tbl
= seq
->private;
2733 return SEQ_START_TOKEN
;
2735 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
2736 if (!cpu_possible(cpu
))
2739 return per_cpu_ptr(tbl
->stats
, cpu
);
2744 static void *neigh_stat_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2746 struct neigh_table
*tbl
= seq
->private;
2749 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
2750 if (!cpu_possible(cpu
))
2753 return per_cpu_ptr(tbl
->stats
, cpu
);
2758 static void neigh_stat_seq_stop(struct seq_file
*seq
, void *v
)
2763 static int neigh_stat_seq_show(struct seq_file
*seq
, void *v
)
2765 struct neigh_table
*tbl
= seq
->private;
2766 struct neigh_statistics
*st
= v
;
2768 if (v
== SEQ_START_TOKEN
) {
2769 seq_printf(seq
, "entries allocs destroys hash_grows lookups hits res_failed rcv_probes_mcast rcv_probes_ucast periodic_gc_runs forced_gc_runs unresolved_discards table_fulls\n");
2773 seq_printf(seq
, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2774 "%08lx %08lx %08lx %08lx %08lx %08lx\n",
2775 atomic_read(&tbl
->entries
),
2786 st
->rcv_probes_mcast
,
2787 st
->rcv_probes_ucast
,
2789 st
->periodic_gc_runs
,
2798 static const struct seq_operations neigh_stat_seq_ops
= {
2799 .start
= neigh_stat_seq_start
,
2800 .next
= neigh_stat_seq_next
,
2801 .stop
= neigh_stat_seq_stop
,
2802 .show
= neigh_stat_seq_show
,
2805 static int neigh_stat_seq_open(struct inode
*inode
, struct file
*file
)
2807 int ret
= seq_open(file
, &neigh_stat_seq_ops
);
2810 struct seq_file
*sf
= file
->private_data
;
2811 sf
->private = PDE_DATA(inode
);
2816 static const struct file_operations neigh_stat_seq_fops
= {
2817 .owner
= THIS_MODULE
,
2818 .open
= neigh_stat_seq_open
,
2820 .llseek
= seq_lseek
,
2821 .release
= seq_release
,
2824 #endif /* CONFIG_PROC_FS */
2826 static inline size_t neigh_nlmsg_size(void)
2828 return NLMSG_ALIGN(sizeof(struct ndmsg
))
2829 + nla_total_size(MAX_ADDR_LEN
) /* NDA_DST */
2830 + nla_total_size(MAX_ADDR_LEN
) /* NDA_LLADDR */
2831 + nla_total_size(sizeof(struct nda_cacheinfo
))
2832 + nla_total_size(4); /* NDA_PROBES */
2835 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
)
2837 struct net
*net
= dev_net(n
->dev
);
2838 struct sk_buff
*skb
;
2841 skb
= nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC
);
2845 err
= neigh_fill_info(skb
, n
, 0, 0, type
, flags
);
2847 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2848 WARN_ON(err
== -EMSGSIZE
);
2852 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2856 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2859 void neigh_app_ns(struct neighbour
*n
)
2861 __neigh_notify(n
, RTM_GETNEIGH
, NLM_F_REQUEST
);
2863 EXPORT_SYMBOL(neigh_app_ns
);
2865 #ifdef CONFIG_SYSCTL
2867 static int int_max
= INT_MAX
;
2868 static int unres_qlen_max
= INT_MAX
/ SKB_TRUESIZE(ETH_FRAME_LEN
);
2870 static int proc_unres_qlen(struct ctl_table
*ctl
, int write
,
2871 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2874 struct ctl_table tmp
= *ctl
;
2877 tmp
.extra2
= &unres_qlen_max
;
2880 size
= *(int *)ctl
->data
/ SKB_TRUESIZE(ETH_FRAME_LEN
);
2881 ret
= proc_dointvec_minmax(&tmp
, write
, buffer
, lenp
, ppos
);
2884 *(int *)ctl
->data
= size
* SKB_TRUESIZE(ETH_FRAME_LEN
);
2888 static struct neigh_parms
*neigh_get_dev_parms_rcu(struct net_device
*dev
,
2893 return __in_dev_arp_parms_get_rcu(dev
);
2895 return __in6_dev_nd_parms_get_rcu(dev
);
2900 static void neigh_copy_dflt_parms(struct net
*net
, struct neigh_parms
*p
,
2903 struct net_device
*dev
;
2904 int family
= neigh_parms_family(p
);
2907 for_each_netdev_rcu(net
, dev
) {
2908 struct neigh_parms
*dst_p
=
2909 neigh_get_dev_parms_rcu(dev
, family
);
2911 if (dst_p
&& !test_bit(index
, dst_p
->data_state
))
2912 dst_p
->data
[index
] = p
->data
[index
];
2917 static void neigh_proc_update(struct ctl_table
*ctl
, int write
)
2919 struct net_device
*dev
= ctl
->extra1
;
2920 struct neigh_parms
*p
= ctl
->extra2
;
2921 struct net
*net
= neigh_parms_net(p
);
2922 int index
= (int *) ctl
->data
- p
->data
;
2927 set_bit(index
, p
->data_state
);
2928 if (!dev
) /* NULL dev means this is default value */
2929 neigh_copy_dflt_parms(net
, p
, index
);
2932 static int neigh_proc_dointvec_zero_intmax(struct ctl_table
*ctl
, int write
,
2933 void __user
*buffer
,
2934 size_t *lenp
, loff_t
*ppos
)
2936 struct ctl_table tmp
= *ctl
;
2940 tmp
.extra2
= &int_max
;
2942 ret
= proc_dointvec_minmax(&tmp
, write
, buffer
, lenp
, ppos
);
2943 neigh_proc_update(ctl
, write
);
2947 int neigh_proc_dointvec(struct ctl_table
*ctl
, int write
,
2948 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2950 int ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2952 neigh_proc_update(ctl
, write
);
2955 EXPORT_SYMBOL(neigh_proc_dointvec
);
2957 int neigh_proc_dointvec_jiffies(struct ctl_table
*ctl
, int write
,
2958 void __user
*buffer
,
2959 size_t *lenp
, loff_t
*ppos
)
2961 int ret
= proc_dointvec_jiffies(ctl
, write
, buffer
, lenp
, ppos
);
2963 neigh_proc_update(ctl
, write
);
2966 EXPORT_SYMBOL(neigh_proc_dointvec_jiffies
);
2968 static int neigh_proc_dointvec_userhz_jiffies(struct ctl_table
*ctl
, int write
,
2969 void __user
*buffer
,
2970 size_t *lenp
, loff_t
*ppos
)
2972 int ret
= proc_dointvec_userhz_jiffies(ctl
, write
, buffer
, lenp
, ppos
);
2974 neigh_proc_update(ctl
, write
);
2978 int neigh_proc_dointvec_ms_jiffies(struct ctl_table
*ctl
, int write
,
2979 void __user
*buffer
,
2980 size_t *lenp
, loff_t
*ppos
)
2982 int ret
= proc_dointvec_ms_jiffies(ctl
, write
, buffer
, lenp
, ppos
);
2984 neigh_proc_update(ctl
, write
);
2987 EXPORT_SYMBOL(neigh_proc_dointvec_ms_jiffies
);
2989 static int neigh_proc_dointvec_unres_qlen(struct ctl_table
*ctl
, int write
,
2990 void __user
*buffer
,
2991 size_t *lenp
, loff_t
*ppos
)
2993 int ret
= proc_unres_qlen(ctl
, write
, buffer
, lenp
, ppos
);
2995 neigh_proc_update(ctl
, write
);
2999 static int neigh_proc_base_reachable_time(struct ctl_table
*ctl
, int write
,
3000 void __user
*buffer
,
3001 size_t *lenp
, loff_t
*ppos
)
3003 struct neigh_parms
*p
= ctl
->extra2
;
3006 if (strcmp(ctl
->procname
, "base_reachable_time") == 0)
3007 ret
= neigh_proc_dointvec_jiffies(ctl
, write
, buffer
, lenp
, ppos
);
3008 else if (strcmp(ctl
->procname
, "base_reachable_time_ms") == 0)
3009 ret
= neigh_proc_dointvec_ms_jiffies(ctl
, write
, buffer
, lenp
, ppos
);
3013 if (write
&& ret
== 0) {
3014 /* update reachable_time as well, otherwise, the change will
3015 * only be effective after the next time neigh_periodic_work
3016 * decides to recompute it
3019 neigh_rand_reach_time(NEIGH_VAR(p
, BASE_REACHABLE_TIME
));
3024 #define NEIGH_PARMS_DATA_OFFSET(index) \
3025 (&((struct neigh_parms *) 0)->data[index])
3027 #define NEIGH_SYSCTL_ENTRY(attr, data_attr, name, mval, proc) \
3028 [NEIGH_VAR_ ## attr] = { \
3030 .data = NEIGH_PARMS_DATA_OFFSET(NEIGH_VAR_ ## data_attr), \
3031 .maxlen = sizeof(int), \
3033 .proc_handler = proc, \
3036 #define NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(attr, name) \
3037 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_zero_intmax)
3039 #define NEIGH_SYSCTL_JIFFIES_ENTRY(attr, name) \
3040 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_jiffies)
3042 #define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(attr, name) \
3043 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)
3045 #define NEIGH_SYSCTL_MS_JIFFIES_ENTRY(attr, name) \
3046 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3048 #define NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(attr, data_attr, name) \
3049 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3051 #define NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(attr, data_attr, name) \
3052 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_unres_qlen)
3054 static struct neigh_sysctl_table
{
3055 struct ctl_table_header
*sysctl_header
;
3056 struct ctl_table neigh_vars
[NEIGH_VAR_MAX
+ 1];
3057 } neigh_sysctl_template __read_mostly
= {
3059 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_PROBES
, "mcast_solicit"),
3060 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(UCAST_PROBES
, "ucast_solicit"),
3061 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(APP_PROBES
, "app_solicit"),
3062 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_REPROBES
, "mcast_resolicit"),
3063 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(RETRANS_TIME
, "retrans_time"),
3064 NEIGH_SYSCTL_JIFFIES_ENTRY(BASE_REACHABLE_TIME
, "base_reachable_time"),
3065 NEIGH_SYSCTL_JIFFIES_ENTRY(DELAY_PROBE_TIME
, "delay_first_probe_time"),
3066 NEIGH_SYSCTL_JIFFIES_ENTRY(GC_STALETIME
, "gc_stale_time"),
3067 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(QUEUE_LEN_BYTES
, "unres_qlen_bytes"),
3068 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(PROXY_QLEN
, "proxy_qlen"),
3069 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(ANYCAST_DELAY
, "anycast_delay"),
3070 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(PROXY_DELAY
, "proxy_delay"),
3071 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(LOCKTIME
, "locktime"),
3072 NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(QUEUE_LEN
, QUEUE_LEN_BYTES
, "unres_qlen"),
3073 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(RETRANS_TIME_MS
, RETRANS_TIME
, "retrans_time_ms"),
3074 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(BASE_REACHABLE_TIME_MS
, BASE_REACHABLE_TIME
, "base_reachable_time_ms"),
3075 [NEIGH_VAR_GC_INTERVAL
] = {
3076 .procname
= "gc_interval",
3077 .maxlen
= sizeof(int),
3079 .proc_handler
= proc_dointvec_jiffies
,
3081 [NEIGH_VAR_GC_THRESH1
] = {
3082 .procname
= "gc_thresh1",
3083 .maxlen
= sizeof(int),
3087 .proc_handler
= proc_dointvec_minmax
,
3089 [NEIGH_VAR_GC_THRESH2
] = {
3090 .procname
= "gc_thresh2",
3091 .maxlen
= sizeof(int),
3095 .proc_handler
= proc_dointvec_minmax
,
3097 [NEIGH_VAR_GC_THRESH3
] = {
3098 .procname
= "gc_thresh3",
3099 .maxlen
= sizeof(int),
3103 .proc_handler
= proc_dointvec_minmax
,
3109 int neigh_sysctl_register(struct net_device
*dev
, struct neigh_parms
*p
,
3110 proc_handler
*handler
)
3113 struct neigh_sysctl_table
*t
;
3114 const char *dev_name_source
;
3115 char neigh_path
[ sizeof("net//neigh/") + IFNAMSIZ
+ IFNAMSIZ
];
3118 t
= kmemdup(&neigh_sysctl_template
, sizeof(*t
), GFP_KERNEL
);
3122 for (i
= 0; i
< NEIGH_VAR_GC_INTERVAL
; i
++) {
3123 t
->neigh_vars
[i
].data
+= (long) p
;
3124 t
->neigh_vars
[i
].extra1
= dev
;
3125 t
->neigh_vars
[i
].extra2
= p
;
3129 dev_name_source
= dev
->name
;
3130 /* Terminate the table early */
3131 memset(&t
->neigh_vars
[NEIGH_VAR_GC_INTERVAL
], 0,
3132 sizeof(t
->neigh_vars
[NEIGH_VAR_GC_INTERVAL
]));
3134 struct neigh_table
*tbl
= p
->tbl
;
3135 dev_name_source
= "default";
3136 t
->neigh_vars
[NEIGH_VAR_GC_INTERVAL
].data
= &tbl
->gc_interval
;
3137 t
->neigh_vars
[NEIGH_VAR_GC_THRESH1
].data
= &tbl
->gc_thresh1
;
3138 t
->neigh_vars
[NEIGH_VAR_GC_THRESH2
].data
= &tbl
->gc_thresh2
;
3139 t
->neigh_vars
[NEIGH_VAR_GC_THRESH3
].data
= &tbl
->gc_thresh3
;
3144 t
->neigh_vars
[NEIGH_VAR_RETRANS_TIME
].proc_handler
= handler
;
3146 t
->neigh_vars
[NEIGH_VAR_BASE_REACHABLE_TIME
].proc_handler
= handler
;
3147 /* RetransTime (in milliseconds)*/
3148 t
->neigh_vars
[NEIGH_VAR_RETRANS_TIME_MS
].proc_handler
= handler
;
3149 /* ReachableTime (in milliseconds) */
3150 t
->neigh_vars
[NEIGH_VAR_BASE_REACHABLE_TIME_MS
].proc_handler
= handler
;
3152 /* Those handlers will update p->reachable_time after
3153 * base_reachable_time(_ms) is set to ensure the new timer starts being
3154 * applied after the next neighbour update instead of waiting for
3155 * neigh_periodic_work to update its value (can be multiple minutes)
3156 * So any handler that replaces them should do this as well
3159 t
->neigh_vars
[NEIGH_VAR_BASE_REACHABLE_TIME
].proc_handler
=
3160 neigh_proc_base_reachable_time
;
3161 /* ReachableTime (in milliseconds) */
3162 t
->neigh_vars
[NEIGH_VAR_BASE_REACHABLE_TIME_MS
].proc_handler
=
3163 neigh_proc_base_reachable_time
;
3166 /* Don't export sysctls to unprivileged users */
3167 if (neigh_parms_net(p
)->user_ns
!= &init_user_ns
)
3168 t
->neigh_vars
[0].procname
= NULL
;
3170 switch (neigh_parms_family(p
)) {
3181 snprintf(neigh_path
, sizeof(neigh_path
), "net/%s/neigh/%s",
3182 p_name
, dev_name_source
);
3184 register_net_sysctl(neigh_parms_net(p
), neigh_path
, t
->neigh_vars
);
3185 if (!t
->sysctl_header
)
3188 p
->sysctl_table
= t
;
3196 EXPORT_SYMBOL(neigh_sysctl_register
);
3198 void neigh_sysctl_unregister(struct neigh_parms
*p
)
3200 if (p
->sysctl_table
) {
3201 struct neigh_sysctl_table
*t
= p
->sysctl_table
;
3202 p
->sysctl_table
= NULL
;
3203 unregister_net_sysctl_table(t
->sysctl_header
);
3207 EXPORT_SYMBOL(neigh_sysctl_unregister
);
3209 #endif /* CONFIG_SYSCTL */
3211 static int __init
neigh_init(void)
3213 rtnl_register(PF_UNSPEC
, RTM_NEWNEIGH
, neigh_add
, NULL
, NULL
);
3214 rtnl_register(PF_UNSPEC
, RTM_DELNEIGH
, neigh_delete
, NULL
, NULL
);
3215 rtnl_register(PF_UNSPEC
, RTM_GETNEIGH
, NULL
, neigh_dump_info
, NULL
);
3217 rtnl_register(PF_UNSPEC
, RTM_GETNEIGHTBL
, NULL
, neightbl_dump_info
,
3219 rtnl_register(PF_UNSPEC
, RTM_SETNEIGHTBL
, neightbl_set
, NULL
, NULL
);
3224 subsys_initcall(neigh_init
);