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 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/module.h>
21 #include <linux/socket.h>
22 #include <linux/netdevice.h>
23 #include <linux/proc_fs.h>
25 #include <linux/sysctl.h>
27 #include <linux/times.h>
28 #include <net/net_namespace.h>
29 #include <net/neighbour.h>
32 #include <net/netevent.h>
33 #include <net/netlink.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/random.h>
36 #include <linux/string.h>
37 #include <linux/log2.h>
41 #define NEIGH_PRINTK(x...) printk(x)
42 #define NEIGH_NOPRINTK(x...) do { ; } while(0)
43 #define NEIGH_PRINTK0 NEIGH_PRINTK
44 #define NEIGH_PRINTK1 NEIGH_NOPRINTK
45 #define NEIGH_PRINTK2 NEIGH_NOPRINTK
49 #define NEIGH_PRINTK1 NEIGH_PRINTK
53 #define NEIGH_PRINTK2 NEIGH_PRINTK
56 #define PNEIGH_HASHMASK 0xF
58 static void neigh_timer_handler(unsigned long arg
);
59 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
);
60 static void neigh_update_notify(struct neighbour
*neigh
);
61 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
);
63 static struct neigh_table
*neigh_tables
;
65 static const struct file_operations neigh_stat_seq_fops
;
69 Neighbour hash table buckets are protected with rwlock tbl->lock.
71 - All the scans/updates to hash buckets MUST be made under this lock.
72 - NOTHING clever should be made under this lock: no callbacks
73 to protocol backends, no attempts to send something to network.
74 It will result in deadlocks, if backend/driver wants to use neighbour
76 - If the entry requires some non-trivial actions, increase
77 its reference count and release table lock.
79 Neighbour entries are protected:
80 - with reference count.
81 - with rwlock neigh->lock
83 Reference count prevents destruction.
85 neigh->lock mainly serializes ll address data and its validity state.
86 However, the same lock is used to protect another entry fields:
90 Again, nothing clever shall be made under neigh->lock,
91 the most complicated procedure, which we allow is dev->hard_header.
92 It is supposed, that dev->hard_header is simplistic and does
93 not make callbacks to neighbour tables.
95 The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
96 list of neighbour tables. This list is used only in process context,
99 static DEFINE_RWLOCK(neigh_tbl_lock
);
101 static int neigh_blackhole(struct sk_buff
*skb
)
107 static void neigh_cleanup_and_release(struct neighbour
*neigh
)
109 if (neigh
->parms
->neigh_cleanup
)
110 neigh
->parms
->neigh_cleanup(neigh
);
112 __neigh_notify(neigh
, RTM_DELNEIGH
, 0);
113 neigh_release(neigh
);
117 * It is random distribution in the interval (1/2)*base...(3/2)*base.
118 * It corresponds to default IPv6 settings and is not overridable,
119 * because it is really reasonable choice.
122 unsigned long neigh_rand_reach_time(unsigned long base
)
124 return (base
? (net_random() % base
) + (base
>> 1) : 0);
126 EXPORT_SYMBOL(neigh_rand_reach_time
);
129 static int neigh_forced_gc(struct neigh_table
*tbl
)
134 NEIGH_CACHE_STAT_INC(tbl
, forced_gc_runs
);
136 write_lock_bh(&tbl
->lock
);
137 for (i
= 0; i
<= tbl
->hash_mask
; i
++) {
138 struct neighbour
*n
, **np
;
140 np
= &tbl
->hash_buckets
[i
];
141 while ((n
= *np
) != NULL
) {
142 /* Neighbour record may be discarded if:
143 * - nobody refers to it.
144 * - it is not permanent
146 write_lock(&n
->lock
);
147 if (atomic_read(&n
->refcnt
) == 1 &&
148 !(n
->nud_state
& NUD_PERMANENT
)) {
152 write_unlock(&n
->lock
);
153 neigh_cleanup_and_release(n
);
156 write_unlock(&n
->lock
);
161 tbl
->last_flush
= jiffies
;
163 write_unlock_bh(&tbl
->lock
);
168 static void neigh_add_timer(struct neighbour
*n
, unsigned long when
)
171 if (unlikely(mod_timer(&n
->timer
, when
))) {
172 printk("NEIGH: BUG, double timer add, state is %x\n",
178 static int neigh_del_timer(struct neighbour
*n
)
180 if ((n
->nud_state
& NUD_IN_TIMER
) &&
181 del_timer(&n
->timer
)) {
188 static void pneigh_queue_purge(struct sk_buff_head
*list
)
192 while ((skb
= skb_dequeue(list
)) != NULL
) {
198 static void neigh_flush_dev(struct neigh_table
*tbl
, struct net_device
*dev
)
202 for (i
= 0; i
<= tbl
->hash_mask
; i
++) {
203 struct neighbour
*n
, **np
= &tbl
->hash_buckets
[i
];
205 while ((n
= *np
) != NULL
) {
206 if (dev
&& n
->dev
!= dev
) {
211 write_lock(&n
->lock
);
215 if (atomic_read(&n
->refcnt
) != 1) {
216 /* The most unpleasant situation.
217 We must destroy neighbour entry,
218 but someone still uses it.
220 The destroy will be delayed until
221 the last user releases us, but
222 we must kill timers etc. and move
225 skb_queue_purge(&n
->arp_queue
);
226 n
->output
= neigh_blackhole
;
227 if (n
->nud_state
& NUD_VALID
)
228 n
->nud_state
= NUD_NOARP
;
230 n
->nud_state
= NUD_NONE
;
231 NEIGH_PRINTK2("neigh %p is stray.\n", n
);
233 write_unlock(&n
->lock
);
234 neigh_cleanup_and_release(n
);
239 void neigh_changeaddr(struct neigh_table
*tbl
, struct net_device
*dev
)
241 write_lock_bh(&tbl
->lock
);
242 neigh_flush_dev(tbl
, dev
);
243 write_unlock_bh(&tbl
->lock
);
245 EXPORT_SYMBOL(neigh_changeaddr
);
247 int neigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
249 write_lock_bh(&tbl
->lock
);
250 neigh_flush_dev(tbl
, dev
);
251 pneigh_ifdown(tbl
, dev
);
252 write_unlock_bh(&tbl
->lock
);
254 del_timer_sync(&tbl
->proxy_timer
);
255 pneigh_queue_purge(&tbl
->proxy_queue
);
258 EXPORT_SYMBOL(neigh_ifdown
);
260 static struct neighbour
*neigh_alloc(struct neigh_table
*tbl
)
262 struct neighbour
*n
= NULL
;
263 unsigned long now
= jiffies
;
266 entries
= atomic_inc_return(&tbl
->entries
) - 1;
267 if (entries
>= tbl
->gc_thresh3
||
268 (entries
>= tbl
->gc_thresh2
&&
269 time_after(now
, tbl
->last_flush
+ 5 * HZ
))) {
270 if (!neigh_forced_gc(tbl
) &&
271 entries
>= tbl
->gc_thresh3
)
275 n
= kmem_cache_zalloc(tbl
->kmem_cachep
, GFP_ATOMIC
);
279 skb_queue_head_init(&n
->arp_queue
);
280 rwlock_init(&n
->lock
);
281 n
->updated
= n
->used
= now
;
282 n
->nud_state
= NUD_NONE
;
283 n
->output
= neigh_blackhole
;
284 n
->parms
= neigh_parms_clone(&tbl
->parms
);
285 setup_timer(&n
->timer
, neigh_timer_handler
, (unsigned long)n
);
287 NEIGH_CACHE_STAT_INC(tbl
, allocs
);
289 atomic_set(&n
->refcnt
, 1);
295 atomic_dec(&tbl
->entries
);
299 static struct neighbour
**neigh_hash_alloc(unsigned int entries
)
301 unsigned long size
= entries
* sizeof(struct neighbour
*);
302 struct neighbour
**ret
;
304 if (size
<= PAGE_SIZE
) {
305 ret
= kzalloc(size
, GFP_ATOMIC
);
307 ret
= (struct neighbour
**)
308 __get_free_pages(GFP_ATOMIC
|__GFP_ZERO
, get_order(size
));
313 static void neigh_hash_free(struct neighbour
**hash
, unsigned int entries
)
315 unsigned long size
= entries
* sizeof(struct neighbour
*);
317 if (size
<= PAGE_SIZE
)
320 free_pages((unsigned long)hash
, get_order(size
));
323 static void neigh_hash_grow(struct neigh_table
*tbl
, unsigned long new_entries
)
325 struct neighbour
**new_hash
, **old_hash
;
326 unsigned int i
, new_hash_mask
, old_entries
;
328 NEIGH_CACHE_STAT_INC(tbl
, hash_grows
);
330 BUG_ON(!is_power_of_2(new_entries
));
331 new_hash
= neigh_hash_alloc(new_entries
);
335 old_entries
= tbl
->hash_mask
+ 1;
336 new_hash_mask
= new_entries
- 1;
337 old_hash
= tbl
->hash_buckets
;
339 get_random_bytes(&tbl
->hash_rnd
, sizeof(tbl
->hash_rnd
));
340 for (i
= 0; i
< old_entries
; i
++) {
341 struct neighbour
*n
, *next
;
343 for (n
= old_hash
[i
]; n
; n
= next
) {
344 unsigned int hash_val
= tbl
->hash(n
->primary_key
, n
->dev
);
346 hash_val
&= new_hash_mask
;
349 n
->next
= new_hash
[hash_val
];
350 new_hash
[hash_val
] = n
;
353 tbl
->hash_buckets
= new_hash
;
354 tbl
->hash_mask
= new_hash_mask
;
356 neigh_hash_free(old_hash
, old_entries
);
359 struct neighbour
*neigh_lookup(struct neigh_table
*tbl
, const void *pkey
,
360 struct net_device
*dev
)
363 int key_len
= tbl
->key_len
;
366 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
368 read_lock_bh(&tbl
->lock
);
369 hash_val
= tbl
->hash(pkey
, dev
);
370 for (n
= tbl
->hash_buckets
[hash_val
& tbl
->hash_mask
]; n
; n
= n
->next
) {
371 if (dev
== n
->dev
&& !memcmp(n
->primary_key
, pkey
, key_len
)) {
373 NEIGH_CACHE_STAT_INC(tbl
, hits
);
377 read_unlock_bh(&tbl
->lock
);
380 EXPORT_SYMBOL(neigh_lookup
);
382 struct neighbour
*neigh_lookup_nodev(struct neigh_table
*tbl
, struct net
*net
,
386 int key_len
= tbl
->key_len
;
389 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
391 read_lock_bh(&tbl
->lock
);
392 hash_val
= tbl
->hash(pkey
, NULL
);
393 for (n
= tbl
->hash_buckets
[hash_val
& tbl
->hash_mask
]; n
; n
= n
->next
) {
394 if (!memcmp(n
->primary_key
, pkey
, key_len
) &&
395 net_eq(dev_net(n
->dev
), net
)) {
397 NEIGH_CACHE_STAT_INC(tbl
, hits
);
401 read_unlock_bh(&tbl
->lock
);
404 EXPORT_SYMBOL(neigh_lookup_nodev
);
406 struct neighbour
*neigh_create(struct neigh_table
*tbl
, const void *pkey
,
407 struct net_device
*dev
)
410 int key_len
= tbl
->key_len
;
412 struct neighbour
*n1
, *rc
, *n
= neigh_alloc(tbl
);
415 rc
= ERR_PTR(-ENOBUFS
);
419 memcpy(n
->primary_key
, pkey
, key_len
);
423 /* Protocol specific setup. */
424 if (tbl
->constructor
&& (error
= tbl
->constructor(n
)) < 0) {
426 goto out_neigh_release
;
429 /* Device specific setup. */
430 if (n
->parms
->neigh_setup
&&
431 (error
= n
->parms
->neigh_setup(n
)) < 0) {
433 goto out_neigh_release
;
436 n
->confirmed
= jiffies
- (n
->parms
->base_reachable_time
<< 1);
438 write_lock_bh(&tbl
->lock
);
440 if (atomic_read(&tbl
->entries
) > (tbl
->hash_mask
+ 1))
441 neigh_hash_grow(tbl
, (tbl
->hash_mask
+ 1) << 1);
443 hash_val
= tbl
->hash(pkey
, dev
) & tbl
->hash_mask
;
445 if (n
->parms
->dead
) {
446 rc
= ERR_PTR(-EINVAL
);
450 for (n1
= tbl
->hash_buckets
[hash_val
]; n1
; n1
= n1
->next
) {
451 if (dev
== n1
->dev
&& !memcmp(n1
->primary_key
, pkey
, key_len
)) {
458 n
->next
= tbl
->hash_buckets
[hash_val
];
459 tbl
->hash_buckets
[hash_val
] = n
;
462 write_unlock_bh(&tbl
->lock
);
463 NEIGH_PRINTK2("neigh %p is created.\n", n
);
468 write_unlock_bh(&tbl
->lock
);
473 EXPORT_SYMBOL(neigh_create
);
475 static u32
pneigh_hash(const void *pkey
, int key_len
)
477 u32 hash_val
= *(u32
*)(pkey
+ key_len
- 4);
478 hash_val
^= (hash_val
>> 16);
479 hash_val
^= hash_val
>> 8;
480 hash_val
^= hash_val
>> 4;
481 hash_val
&= PNEIGH_HASHMASK
;
485 static struct pneigh_entry
*__pneigh_lookup_1(struct pneigh_entry
*n
,
489 struct net_device
*dev
)
492 if (!memcmp(n
->key
, pkey
, key_len
) &&
493 net_eq(pneigh_net(n
), net
) &&
494 (n
->dev
== dev
|| !n
->dev
))
501 struct pneigh_entry
*__pneigh_lookup(struct neigh_table
*tbl
,
502 struct net
*net
, const void *pkey
, struct net_device
*dev
)
504 int key_len
= tbl
->key_len
;
505 u32 hash_val
= pneigh_hash(pkey
, key_len
);
507 return __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
508 net
, pkey
, key_len
, dev
);
510 EXPORT_SYMBOL_GPL(__pneigh_lookup
);
512 struct pneigh_entry
* pneigh_lookup(struct neigh_table
*tbl
,
513 struct net
*net
, const void *pkey
,
514 struct net_device
*dev
, int creat
)
516 struct pneigh_entry
*n
;
517 int key_len
= tbl
->key_len
;
518 u32 hash_val
= pneigh_hash(pkey
, key_len
);
520 read_lock_bh(&tbl
->lock
);
521 n
= __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
522 net
, pkey
, key_len
, dev
);
523 read_unlock_bh(&tbl
->lock
);
530 n
= kmalloc(sizeof(*n
) + key_len
, GFP_KERNEL
);
534 write_pnet(&n
->net
, hold_net(net
));
535 memcpy(n
->key
, pkey
, key_len
);
540 if (tbl
->pconstructor
&& tbl
->pconstructor(n
)) {
549 write_lock_bh(&tbl
->lock
);
550 n
->next
= tbl
->phash_buckets
[hash_val
];
551 tbl
->phash_buckets
[hash_val
] = n
;
552 write_unlock_bh(&tbl
->lock
);
556 EXPORT_SYMBOL(pneigh_lookup
);
559 int pneigh_delete(struct neigh_table
*tbl
, struct net
*net
, const void *pkey
,
560 struct net_device
*dev
)
562 struct pneigh_entry
*n
, **np
;
563 int key_len
= tbl
->key_len
;
564 u32 hash_val
= pneigh_hash(pkey
, key_len
);
566 write_lock_bh(&tbl
->lock
);
567 for (np
= &tbl
->phash_buckets
[hash_val
]; (n
= *np
) != NULL
;
569 if (!memcmp(n
->key
, pkey
, key_len
) && n
->dev
== dev
&&
570 net_eq(pneigh_net(n
), net
)) {
572 write_unlock_bh(&tbl
->lock
);
573 if (tbl
->pdestructor
)
577 release_net(pneigh_net(n
));
582 write_unlock_bh(&tbl
->lock
);
586 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
588 struct pneigh_entry
*n
, **np
;
591 for (h
= 0; h
<= PNEIGH_HASHMASK
; h
++) {
592 np
= &tbl
->phash_buckets
[h
];
593 while ((n
= *np
) != NULL
) {
594 if (!dev
|| n
->dev
== dev
) {
596 if (tbl
->pdestructor
)
600 release_net(pneigh_net(n
));
610 static void neigh_parms_destroy(struct neigh_parms
*parms
);
612 static inline void neigh_parms_put(struct neigh_parms
*parms
)
614 if (atomic_dec_and_test(&parms
->refcnt
))
615 neigh_parms_destroy(parms
);
619 * neighbour must already be out of the table;
622 void neigh_destroy(struct neighbour
*neigh
)
626 NEIGH_CACHE_STAT_INC(neigh
->tbl
, destroys
);
630 "Destroying alive neighbour %p\n", neigh
);
635 if (neigh_del_timer(neigh
))
636 printk(KERN_WARNING
"Impossible event.\n");
638 while ((hh
= neigh
->hh
) != NULL
) {
639 neigh
->hh
= hh
->hh_next
;
642 write_seqlock_bh(&hh
->hh_lock
);
643 hh
->hh_output
= neigh_blackhole
;
644 write_sequnlock_bh(&hh
->hh_lock
);
645 if (atomic_dec_and_test(&hh
->hh_refcnt
))
649 skb_queue_purge(&neigh
->arp_queue
);
652 neigh_parms_put(neigh
->parms
);
654 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh
);
656 atomic_dec(&neigh
->tbl
->entries
);
657 kmem_cache_free(neigh
->tbl
->kmem_cachep
, neigh
);
659 EXPORT_SYMBOL(neigh_destroy
);
661 /* Neighbour state is suspicious;
664 Called with write_locked neigh.
666 static void neigh_suspect(struct neighbour
*neigh
)
670 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh
);
672 neigh
->output
= neigh
->ops
->output
;
674 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
)
675 hh
->hh_output
= neigh
->ops
->output
;
678 /* Neighbour state is OK;
681 Called with write_locked neigh.
683 static void neigh_connect(struct neighbour
*neigh
)
687 NEIGH_PRINTK2("neigh %p is connected.\n", neigh
);
689 neigh
->output
= neigh
->ops
->connected_output
;
691 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
)
692 hh
->hh_output
= neigh
->ops
->hh_output
;
695 static void neigh_periodic_timer(unsigned long arg
)
697 struct neigh_table
*tbl
= (struct neigh_table
*)arg
;
698 struct neighbour
*n
, **np
;
699 unsigned long expire
, now
= jiffies
;
701 NEIGH_CACHE_STAT_INC(tbl
, periodic_gc_runs
);
703 write_lock(&tbl
->lock
);
706 * periodically recompute ReachableTime from random function
709 if (time_after(now
, tbl
->last_rand
+ 300 * HZ
)) {
710 struct neigh_parms
*p
;
711 tbl
->last_rand
= now
;
712 for (p
= &tbl
->parms
; p
; p
= p
->next
)
714 neigh_rand_reach_time(p
->base_reachable_time
);
717 np
= &tbl
->hash_buckets
[tbl
->hash_chain_gc
];
718 tbl
->hash_chain_gc
= ((tbl
->hash_chain_gc
+ 1) & tbl
->hash_mask
);
720 while ((n
= *np
) != NULL
) {
723 write_lock(&n
->lock
);
725 state
= n
->nud_state
;
726 if (state
& (NUD_PERMANENT
| NUD_IN_TIMER
)) {
727 write_unlock(&n
->lock
);
731 if (time_before(n
->used
, n
->confirmed
))
732 n
->used
= n
->confirmed
;
734 if (atomic_read(&n
->refcnt
) == 1 &&
735 (state
== NUD_FAILED
||
736 time_after(now
, n
->used
+ n
->parms
->gc_staletime
))) {
739 write_unlock(&n
->lock
);
740 neigh_cleanup_and_release(n
);
743 write_unlock(&n
->lock
);
749 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
750 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
751 * base_reachable_time.
753 expire
= tbl
->parms
.base_reachable_time
>> 1;
754 expire
/= (tbl
->hash_mask
+ 1);
759 mod_timer(&tbl
->gc_timer
, round_jiffies(now
+ expire
));
761 mod_timer(&tbl
->gc_timer
, now
+ expire
);
763 write_unlock(&tbl
->lock
);
766 static __inline__
int neigh_max_probes(struct neighbour
*n
)
768 struct neigh_parms
*p
= n
->parms
;
769 return (n
->nud_state
& NUD_PROBE
?
771 p
->ucast_probes
+ p
->app_probes
+ p
->mcast_probes
);
774 /* Called when a timer expires for a neighbour entry. */
776 static void neigh_timer_handler(unsigned long arg
)
778 unsigned long now
, next
;
779 struct neighbour
*neigh
= (struct neighbour
*)arg
;
783 write_lock(&neigh
->lock
);
785 state
= neigh
->nud_state
;
789 if (!(state
& NUD_IN_TIMER
)) {
791 printk(KERN_WARNING
"neigh: timer & !nud_in_timer\n");
796 if (state
& NUD_REACHABLE
) {
797 if (time_before_eq(now
,
798 neigh
->confirmed
+ neigh
->parms
->reachable_time
)) {
799 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh
);
800 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
801 } else if (time_before_eq(now
,
802 neigh
->used
+ neigh
->parms
->delay_probe_time
)) {
803 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh
);
804 neigh
->nud_state
= NUD_DELAY
;
805 neigh
->updated
= jiffies
;
806 neigh_suspect(neigh
);
807 next
= now
+ neigh
->parms
->delay_probe_time
;
809 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh
);
810 neigh
->nud_state
= NUD_STALE
;
811 neigh
->updated
= jiffies
;
812 neigh_suspect(neigh
);
815 } else if (state
& NUD_DELAY
) {
816 if (time_before_eq(now
,
817 neigh
->confirmed
+ neigh
->parms
->delay_probe_time
)) {
818 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh
);
819 neigh
->nud_state
= NUD_REACHABLE
;
820 neigh
->updated
= jiffies
;
821 neigh_connect(neigh
);
823 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
825 NEIGH_PRINTK2("neigh %p is probed.\n", neigh
);
826 neigh
->nud_state
= NUD_PROBE
;
827 neigh
->updated
= jiffies
;
828 atomic_set(&neigh
->probes
, 0);
829 next
= now
+ neigh
->parms
->retrans_time
;
832 /* NUD_PROBE|NUD_INCOMPLETE */
833 next
= now
+ neigh
->parms
->retrans_time
;
836 if ((neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) &&
837 atomic_read(&neigh
->probes
) >= neigh_max_probes(neigh
)) {
840 neigh
->nud_state
= NUD_FAILED
;
841 neigh
->updated
= jiffies
;
843 NEIGH_CACHE_STAT_INC(neigh
->tbl
, res_failed
);
844 NEIGH_PRINTK2("neigh %p is failed.\n", neigh
);
846 /* It is very thin place. report_unreachable is very complicated
847 routine. Particularly, it can hit the same neighbour entry!
849 So that, we try to be accurate and avoid dead loop. --ANK
851 while (neigh
->nud_state
== NUD_FAILED
&&
852 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
853 write_unlock(&neigh
->lock
);
854 neigh
->ops
->error_report(neigh
, skb
);
855 write_lock(&neigh
->lock
);
857 skb_queue_purge(&neigh
->arp_queue
);
860 if (neigh
->nud_state
& NUD_IN_TIMER
) {
861 if (time_before(next
, jiffies
+ HZ
/2))
862 next
= jiffies
+ HZ
/2;
863 if (!mod_timer(&neigh
->timer
, next
))
866 if (neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) {
867 struct sk_buff
*skb
= skb_peek(&neigh
->arp_queue
);
868 /* keep skb alive even if arp_queue overflows */
870 skb
= skb_copy(skb
, GFP_ATOMIC
);
871 write_unlock(&neigh
->lock
);
872 neigh
->ops
->solicit(neigh
, skb
);
873 atomic_inc(&neigh
->probes
);
877 write_unlock(&neigh
->lock
);
881 neigh_update_notify(neigh
);
883 neigh_release(neigh
);
886 int __neigh_event_send(struct neighbour
*neigh
, struct sk_buff
*skb
)
891 write_lock_bh(&neigh
->lock
);
894 if (neigh
->nud_state
& (NUD_CONNECTED
| NUD_DELAY
| NUD_PROBE
))
899 if (!(neigh
->nud_state
& (NUD_STALE
| NUD_INCOMPLETE
))) {
900 if (neigh
->parms
->mcast_probes
+ neigh
->parms
->app_probes
) {
901 atomic_set(&neigh
->probes
, neigh
->parms
->ucast_probes
);
902 neigh
->nud_state
= NUD_INCOMPLETE
;
903 neigh
->updated
= jiffies
;
904 neigh_add_timer(neigh
, now
+ 1);
906 neigh
->nud_state
= NUD_FAILED
;
907 neigh
->updated
= jiffies
;
908 write_unlock_bh(&neigh
->lock
);
913 } else if (neigh
->nud_state
& NUD_STALE
) {
914 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh
);
915 neigh
->nud_state
= NUD_DELAY
;
916 neigh
->updated
= jiffies
;
917 neigh_add_timer(neigh
,
918 jiffies
+ neigh
->parms
->delay_probe_time
);
921 if (neigh
->nud_state
== NUD_INCOMPLETE
) {
923 if (skb_queue_len(&neigh
->arp_queue
) >=
924 neigh
->parms
->queue_len
) {
925 struct sk_buff
*buff
;
926 buff
= __skb_dequeue(&neigh
->arp_queue
);
928 NEIGH_CACHE_STAT_INC(neigh
->tbl
, unres_discards
);
930 __skb_queue_tail(&neigh
->arp_queue
, skb
);
935 write_unlock_bh(&neigh
->lock
);
938 EXPORT_SYMBOL(__neigh_event_send
);
940 static void neigh_update_hhs(struct neighbour
*neigh
)
943 void (*update
)(struct hh_cache
*, const struct net_device
*, const unsigned char *)
944 = neigh
->dev
->header_ops
->cache_update
;
947 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
) {
948 write_seqlock_bh(&hh
->hh_lock
);
949 update(hh
, neigh
->dev
, neigh
->ha
);
950 write_sequnlock_bh(&hh
->hh_lock
);
957 /* Generic update routine.
958 -- lladdr is new lladdr or NULL, if it is not supplied.
961 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
963 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
964 lladdr instead of overriding it
966 It also allows to retain current state
967 if lladdr is unchanged.
968 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
970 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
972 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
975 Caller MUST hold reference count on the entry.
978 int neigh_update(struct neighbour
*neigh
, const u8
*lladdr
, u8
new,
984 struct net_device
*dev
;
985 int update_isrouter
= 0;
987 write_lock_bh(&neigh
->lock
);
990 old
= neigh
->nud_state
;
993 if (!(flags
& NEIGH_UPDATE_F_ADMIN
) &&
994 (old
& (NUD_NOARP
| NUD_PERMANENT
)))
997 if (!(new & NUD_VALID
)) {
998 neigh_del_timer(neigh
);
999 if (old
& NUD_CONNECTED
)
1000 neigh_suspect(neigh
);
1001 neigh
->nud_state
= new;
1003 notify
= old
& NUD_VALID
;
1007 /* Compare new lladdr with cached one */
1008 if (!dev
->addr_len
) {
1009 /* First case: device needs no address. */
1011 } else if (lladdr
) {
1012 /* The second case: if something is already cached
1013 and a new address is proposed:
1015 - if they are different, check override flag
1017 if ((old
& NUD_VALID
) &&
1018 !memcmp(lladdr
, neigh
->ha
, dev
->addr_len
))
1021 /* No address is supplied; if we know something,
1022 use it, otherwise discard the request.
1025 if (!(old
& NUD_VALID
))
1030 if (new & NUD_CONNECTED
)
1031 neigh
->confirmed
= jiffies
;
1032 neigh
->updated
= jiffies
;
1034 /* If entry was valid and address is not changed,
1035 do not change entry state, if new one is STALE.
1038 update_isrouter
= flags
& NEIGH_UPDATE_F_OVERRIDE_ISROUTER
;
1039 if (old
& NUD_VALID
) {
1040 if (lladdr
!= neigh
->ha
&& !(flags
& NEIGH_UPDATE_F_OVERRIDE
)) {
1041 update_isrouter
= 0;
1042 if ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) &&
1043 (old
& NUD_CONNECTED
)) {
1049 if (lladdr
== neigh
->ha
&& new == NUD_STALE
&&
1050 ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) ||
1051 (old
& NUD_CONNECTED
))
1058 neigh_del_timer(neigh
);
1059 if (new & NUD_IN_TIMER
)
1060 neigh_add_timer(neigh
, (jiffies
+
1061 ((new & NUD_REACHABLE
) ?
1062 neigh
->parms
->reachable_time
:
1064 neigh
->nud_state
= new;
1067 if (lladdr
!= neigh
->ha
) {
1068 memcpy(&neigh
->ha
, lladdr
, dev
->addr_len
);
1069 neigh_update_hhs(neigh
);
1070 if (!(new & NUD_CONNECTED
))
1071 neigh
->confirmed
= jiffies
-
1072 (neigh
->parms
->base_reachable_time
<< 1);
1077 if (new & NUD_CONNECTED
)
1078 neigh_connect(neigh
);
1080 neigh_suspect(neigh
);
1081 if (!(old
& NUD_VALID
)) {
1082 struct sk_buff
*skb
;
1084 /* Again: avoid dead loop if something went wrong */
1086 while (neigh
->nud_state
& NUD_VALID
&&
1087 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
1088 struct neighbour
*n1
= neigh
;
1089 write_unlock_bh(&neigh
->lock
);
1090 /* On shaper/eql skb->dst->neighbour != neigh :( */
1091 if (skb
->dst
&& skb
->dst
->neighbour
)
1092 n1
= skb
->dst
->neighbour
;
1094 write_lock_bh(&neigh
->lock
);
1096 skb_queue_purge(&neigh
->arp_queue
);
1099 if (update_isrouter
) {
1100 neigh
->flags
= (flags
& NEIGH_UPDATE_F_ISROUTER
) ?
1101 (neigh
->flags
| NTF_ROUTER
) :
1102 (neigh
->flags
& ~NTF_ROUTER
);
1104 write_unlock_bh(&neigh
->lock
);
1107 neigh_update_notify(neigh
);
1111 EXPORT_SYMBOL(neigh_update
);
1113 struct neighbour
*neigh_event_ns(struct neigh_table
*tbl
,
1114 u8
*lladdr
, void *saddr
,
1115 struct net_device
*dev
)
1117 struct neighbour
*neigh
= __neigh_lookup(tbl
, saddr
, dev
,
1118 lladdr
|| !dev
->addr_len
);
1120 neigh_update(neigh
, lladdr
, NUD_STALE
,
1121 NEIGH_UPDATE_F_OVERRIDE
);
1124 EXPORT_SYMBOL(neigh_event_ns
);
1126 static void neigh_hh_init(struct neighbour
*n
, struct dst_entry
*dst
,
1129 struct hh_cache
*hh
;
1130 struct net_device
*dev
= dst
->dev
;
1132 for (hh
= n
->hh
; hh
; hh
= hh
->hh_next
)
1133 if (hh
->hh_type
== protocol
)
1136 if (!hh
&& (hh
= kzalloc(sizeof(*hh
), GFP_ATOMIC
)) != NULL
) {
1137 seqlock_init(&hh
->hh_lock
);
1138 hh
->hh_type
= protocol
;
1139 atomic_set(&hh
->hh_refcnt
, 0);
1142 if (dev
->header_ops
->cache(n
, hh
)) {
1146 atomic_inc(&hh
->hh_refcnt
);
1147 hh
->hh_next
= n
->hh
;
1149 if (n
->nud_state
& NUD_CONNECTED
)
1150 hh
->hh_output
= n
->ops
->hh_output
;
1152 hh
->hh_output
= n
->ops
->output
;
1156 atomic_inc(&hh
->hh_refcnt
);
1161 /* This function can be used in contexts, where only old dev_queue_xmit
1162 worked, f.e. if you want to override normal output path (eql, shaper),
1163 but resolution is not made yet.
1166 int neigh_compat_output(struct sk_buff
*skb
)
1168 struct net_device
*dev
= skb
->dev
;
1170 __skb_pull(skb
, skb_network_offset(skb
));
1172 if (dev_hard_header(skb
, dev
, ntohs(skb
->protocol
), NULL
, NULL
,
1174 dev
->header_ops
->rebuild(skb
))
1177 return dev_queue_xmit(skb
);
1179 EXPORT_SYMBOL(neigh_compat_output
);
1181 /* Slow and careful. */
1183 int neigh_resolve_output(struct sk_buff
*skb
)
1185 struct dst_entry
*dst
= skb
->dst
;
1186 struct neighbour
*neigh
;
1189 if (!dst
|| !(neigh
= dst
->neighbour
))
1192 __skb_pull(skb
, skb_network_offset(skb
));
1194 if (!neigh_event_send(neigh
, skb
)) {
1196 struct net_device
*dev
= neigh
->dev
;
1197 if (dev
->header_ops
->cache
&& !dst
->hh
) {
1198 write_lock_bh(&neigh
->lock
);
1200 neigh_hh_init(neigh
, dst
, dst
->ops
->protocol
);
1201 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1202 neigh
->ha
, NULL
, skb
->len
);
1203 write_unlock_bh(&neigh
->lock
);
1205 read_lock_bh(&neigh
->lock
);
1206 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1207 neigh
->ha
, NULL
, skb
->len
);
1208 read_unlock_bh(&neigh
->lock
);
1211 rc
= neigh
->ops
->queue_xmit(skb
);
1218 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1219 dst
, dst
? dst
->neighbour
: NULL
);
1225 EXPORT_SYMBOL(neigh_resolve_output
);
1227 /* As fast as possible without hh cache */
1229 int neigh_connected_output(struct sk_buff
*skb
)
1232 struct dst_entry
*dst
= skb
->dst
;
1233 struct neighbour
*neigh
= dst
->neighbour
;
1234 struct net_device
*dev
= neigh
->dev
;
1236 __skb_pull(skb
, skb_network_offset(skb
));
1238 read_lock_bh(&neigh
->lock
);
1239 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1240 neigh
->ha
, NULL
, skb
->len
);
1241 read_unlock_bh(&neigh
->lock
);
1243 err
= neigh
->ops
->queue_xmit(skb
);
1250 EXPORT_SYMBOL(neigh_connected_output
);
1252 static void neigh_proxy_process(unsigned long arg
)
1254 struct neigh_table
*tbl
= (struct neigh_table
*)arg
;
1255 long sched_next
= 0;
1256 unsigned long now
= jiffies
;
1257 struct sk_buff
*skb
, *n
;
1259 spin_lock(&tbl
->proxy_queue
.lock
);
1261 skb_queue_walk_safe(&tbl
->proxy_queue
, skb
, n
) {
1262 long tdif
= NEIGH_CB(skb
)->sched_next
- now
;
1265 struct net_device
*dev
= skb
->dev
;
1266 __skb_unlink(skb
, &tbl
->proxy_queue
);
1267 if (tbl
->proxy_redo
&& netif_running(dev
))
1268 tbl
->proxy_redo(skb
);
1273 } else if (!sched_next
|| tdif
< sched_next
)
1276 del_timer(&tbl
->proxy_timer
);
1278 mod_timer(&tbl
->proxy_timer
, jiffies
+ sched_next
);
1279 spin_unlock(&tbl
->proxy_queue
.lock
);
1282 void pneigh_enqueue(struct neigh_table
*tbl
, struct neigh_parms
*p
,
1283 struct sk_buff
*skb
)
1285 unsigned long now
= jiffies
;
1286 unsigned long sched_next
= now
+ (net_random() % p
->proxy_delay
);
1288 if (tbl
->proxy_queue
.qlen
> p
->proxy_qlen
) {
1293 NEIGH_CB(skb
)->sched_next
= sched_next
;
1294 NEIGH_CB(skb
)->flags
|= LOCALLY_ENQUEUED
;
1296 spin_lock(&tbl
->proxy_queue
.lock
);
1297 if (del_timer(&tbl
->proxy_timer
)) {
1298 if (time_before(tbl
->proxy_timer
.expires
, sched_next
))
1299 sched_next
= tbl
->proxy_timer
.expires
;
1301 dst_release(skb
->dst
);
1304 __skb_queue_tail(&tbl
->proxy_queue
, skb
);
1305 mod_timer(&tbl
->proxy_timer
, sched_next
);
1306 spin_unlock(&tbl
->proxy_queue
.lock
);
1308 EXPORT_SYMBOL(pneigh_enqueue
);
1310 static inline struct neigh_parms
*lookup_neigh_params(struct neigh_table
*tbl
,
1311 struct net
*net
, int ifindex
)
1313 struct neigh_parms
*p
;
1315 for (p
= &tbl
->parms
; p
; p
= p
->next
) {
1316 if ((p
->dev
&& p
->dev
->ifindex
== ifindex
&& net_eq(neigh_parms_net(p
), net
)) ||
1317 (!p
->dev
&& !ifindex
))
1324 struct neigh_parms
*neigh_parms_alloc(struct net_device
*dev
,
1325 struct neigh_table
*tbl
)
1327 struct neigh_parms
*p
, *ref
;
1328 struct net
*net
= dev_net(dev
);
1329 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1331 ref
= lookup_neigh_params(tbl
, net
, 0);
1335 p
= kmemdup(ref
, sizeof(*p
), GFP_KERNEL
);
1338 atomic_set(&p
->refcnt
, 1);
1340 neigh_rand_reach_time(p
->base_reachable_time
);
1342 if (ops
->ndo_neigh_setup
&& ops
->ndo_neigh_setup(dev
, p
)) {
1349 write_pnet(&p
->net
, hold_net(net
));
1350 p
->sysctl_table
= NULL
;
1351 write_lock_bh(&tbl
->lock
);
1352 p
->next
= tbl
->parms
.next
;
1353 tbl
->parms
.next
= p
;
1354 write_unlock_bh(&tbl
->lock
);
1358 EXPORT_SYMBOL(neigh_parms_alloc
);
1360 static void neigh_rcu_free_parms(struct rcu_head
*head
)
1362 struct neigh_parms
*parms
=
1363 container_of(head
, struct neigh_parms
, rcu_head
);
1365 neigh_parms_put(parms
);
1368 void neigh_parms_release(struct neigh_table
*tbl
, struct neigh_parms
*parms
)
1370 struct neigh_parms
**p
;
1372 if (!parms
|| parms
== &tbl
->parms
)
1374 write_lock_bh(&tbl
->lock
);
1375 for (p
= &tbl
->parms
.next
; *p
; p
= &(*p
)->next
) {
1379 write_unlock_bh(&tbl
->lock
);
1381 dev_put(parms
->dev
);
1382 call_rcu(&parms
->rcu_head
, neigh_rcu_free_parms
);
1386 write_unlock_bh(&tbl
->lock
);
1387 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1389 EXPORT_SYMBOL(neigh_parms_release
);
1391 static void neigh_parms_destroy(struct neigh_parms
*parms
)
1393 release_net(neigh_parms_net(parms
));
1397 static struct lock_class_key neigh_table_proxy_queue_class
;
1399 void neigh_table_init_no_netlink(struct neigh_table
*tbl
)
1401 unsigned long now
= jiffies
;
1402 unsigned long phsize
;
1404 write_pnet(&tbl
->parms
.net
, &init_net
);
1405 atomic_set(&tbl
->parms
.refcnt
, 1);
1406 tbl
->parms
.reachable_time
=
1407 neigh_rand_reach_time(tbl
->parms
.base_reachable_time
);
1409 if (!tbl
->kmem_cachep
)
1411 kmem_cache_create(tbl
->id
, tbl
->entry_size
, 0,
1412 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
1414 tbl
->stats
= alloc_percpu(struct neigh_statistics
);
1416 panic("cannot create neighbour cache statistics");
1418 #ifdef CONFIG_PROC_FS
1419 if (!proc_create_data(tbl
->id
, 0, init_net
.proc_net_stat
,
1420 &neigh_stat_seq_fops
, tbl
))
1421 panic("cannot create neighbour proc dir entry");
1425 tbl
->hash_buckets
= neigh_hash_alloc(tbl
->hash_mask
+ 1);
1427 phsize
= (PNEIGH_HASHMASK
+ 1) * sizeof(struct pneigh_entry
*);
1428 tbl
->phash_buckets
= kzalloc(phsize
, GFP_KERNEL
);
1430 if (!tbl
->hash_buckets
|| !tbl
->phash_buckets
)
1431 panic("cannot allocate neighbour cache hashes");
1433 get_random_bytes(&tbl
->hash_rnd
, sizeof(tbl
->hash_rnd
));
1435 rwlock_init(&tbl
->lock
);
1436 setup_timer(&tbl
->gc_timer
, neigh_periodic_timer
, (unsigned long)tbl
);
1437 tbl
->gc_timer
.expires
= now
+ 1;
1438 add_timer(&tbl
->gc_timer
);
1440 setup_timer(&tbl
->proxy_timer
, neigh_proxy_process
, (unsigned long)tbl
);
1441 skb_queue_head_init_class(&tbl
->proxy_queue
,
1442 &neigh_table_proxy_queue_class
);
1444 tbl
->last_flush
= now
;
1445 tbl
->last_rand
= now
+ tbl
->parms
.reachable_time
* 20;
1447 EXPORT_SYMBOL(neigh_table_init_no_netlink
);
1449 void neigh_table_init(struct neigh_table
*tbl
)
1451 struct neigh_table
*tmp
;
1453 neigh_table_init_no_netlink(tbl
);
1454 write_lock(&neigh_tbl_lock
);
1455 for (tmp
= neigh_tables
; tmp
; tmp
= tmp
->next
) {
1456 if (tmp
->family
== tbl
->family
)
1459 tbl
->next
= neigh_tables
;
1461 write_unlock(&neigh_tbl_lock
);
1463 if (unlikely(tmp
)) {
1464 printk(KERN_ERR
"NEIGH: Registering multiple tables for "
1465 "family %d\n", tbl
->family
);
1469 EXPORT_SYMBOL(neigh_table_init
);
1471 int neigh_table_clear(struct neigh_table
*tbl
)
1473 struct neigh_table
**tp
;
1475 /* It is not clean... Fix it to unload IPv6 module safely */
1476 del_timer_sync(&tbl
->gc_timer
);
1477 del_timer_sync(&tbl
->proxy_timer
);
1478 pneigh_queue_purge(&tbl
->proxy_queue
);
1479 neigh_ifdown(tbl
, NULL
);
1480 if (atomic_read(&tbl
->entries
))
1481 printk(KERN_CRIT
"neighbour leakage\n");
1482 write_lock(&neigh_tbl_lock
);
1483 for (tp
= &neigh_tables
; *tp
; tp
= &(*tp
)->next
) {
1489 write_unlock(&neigh_tbl_lock
);
1491 neigh_hash_free(tbl
->hash_buckets
, tbl
->hash_mask
+ 1);
1492 tbl
->hash_buckets
= NULL
;
1494 kfree(tbl
->phash_buckets
);
1495 tbl
->phash_buckets
= NULL
;
1497 remove_proc_entry(tbl
->id
, init_net
.proc_net_stat
);
1499 free_percpu(tbl
->stats
);
1502 kmem_cache_destroy(tbl
->kmem_cachep
);
1503 tbl
->kmem_cachep
= NULL
;
1507 EXPORT_SYMBOL(neigh_table_clear
);
1509 static int neigh_delete(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1511 struct net
*net
= sock_net(skb
->sk
);
1513 struct nlattr
*dst_attr
;
1514 struct neigh_table
*tbl
;
1515 struct net_device
*dev
= NULL
;
1518 if (nlmsg_len(nlh
) < sizeof(*ndm
))
1521 dst_attr
= nlmsg_find_attr(nlh
, sizeof(*ndm
), NDA_DST
);
1522 if (dst_attr
== NULL
)
1525 ndm
= nlmsg_data(nlh
);
1526 if (ndm
->ndm_ifindex
) {
1527 dev
= dev_get_by_index(net
, ndm
->ndm_ifindex
);
1534 read_lock(&neigh_tbl_lock
);
1535 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1536 struct neighbour
*neigh
;
1538 if (tbl
->family
!= ndm
->ndm_family
)
1540 read_unlock(&neigh_tbl_lock
);
1542 if (nla_len(dst_attr
) < tbl
->key_len
)
1545 if (ndm
->ndm_flags
& NTF_PROXY
) {
1546 err
= pneigh_delete(tbl
, net
, nla_data(dst_attr
), dev
);
1553 neigh
= neigh_lookup(tbl
, nla_data(dst_attr
), dev
);
1554 if (neigh
== NULL
) {
1559 err
= neigh_update(neigh
, NULL
, NUD_FAILED
,
1560 NEIGH_UPDATE_F_OVERRIDE
|
1561 NEIGH_UPDATE_F_ADMIN
);
1562 neigh_release(neigh
);
1565 read_unlock(&neigh_tbl_lock
);
1566 err
= -EAFNOSUPPORT
;
1575 static int neigh_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1577 struct net
*net
= sock_net(skb
->sk
);
1579 struct nlattr
*tb
[NDA_MAX
+1];
1580 struct neigh_table
*tbl
;
1581 struct net_device
*dev
= NULL
;
1584 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
1589 if (tb
[NDA_DST
] == NULL
)
1592 ndm
= nlmsg_data(nlh
);
1593 if (ndm
->ndm_ifindex
) {
1594 dev
= dev_get_by_index(net
, ndm
->ndm_ifindex
);
1600 if (tb
[NDA_LLADDR
] && nla_len(tb
[NDA_LLADDR
]) < dev
->addr_len
)
1604 read_lock(&neigh_tbl_lock
);
1605 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1606 int flags
= NEIGH_UPDATE_F_ADMIN
| NEIGH_UPDATE_F_OVERRIDE
;
1607 struct neighbour
*neigh
;
1610 if (tbl
->family
!= ndm
->ndm_family
)
1612 read_unlock(&neigh_tbl_lock
);
1614 if (nla_len(tb
[NDA_DST
]) < tbl
->key_len
)
1616 dst
= nla_data(tb
[NDA_DST
]);
1617 lladdr
= tb
[NDA_LLADDR
] ? nla_data(tb
[NDA_LLADDR
]) : NULL
;
1619 if (ndm
->ndm_flags
& NTF_PROXY
) {
1620 struct pneigh_entry
*pn
;
1623 pn
= pneigh_lookup(tbl
, net
, dst
, dev
, 1);
1625 pn
->flags
= ndm
->ndm_flags
;
1634 neigh
= neigh_lookup(tbl
, dst
, dev
);
1635 if (neigh
== NULL
) {
1636 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1641 neigh
= __neigh_lookup_errno(tbl
, dst
, dev
);
1642 if (IS_ERR(neigh
)) {
1643 err
= PTR_ERR(neigh
);
1647 if (nlh
->nlmsg_flags
& NLM_F_EXCL
) {
1649 neigh_release(neigh
);
1653 if (!(nlh
->nlmsg_flags
& NLM_F_REPLACE
))
1654 flags
&= ~NEIGH_UPDATE_F_OVERRIDE
;
1657 err
= neigh_update(neigh
, lladdr
, ndm
->ndm_state
, flags
);
1658 neigh_release(neigh
);
1662 read_unlock(&neigh_tbl_lock
);
1663 err
= -EAFNOSUPPORT
;
1672 static int neightbl_fill_parms(struct sk_buff
*skb
, struct neigh_parms
*parms
)
1674 struct nlattr
*nest
;
1676 nest
= nla_nest_start(skb
, NDTA_PARMS
);
1681 NLA_PUT_U32(skb
, NDTPA_IFINDEX
, parms
->dev
->ifindex
);
1683 NLA_PUT_U32(skb
, NDTPA_REFCNT
, atomic_read(&parms
->refcnt
));
1684 NLA_PUT_U32(skb
, NDTPA_QUEUE_LEN
, parms
->queue_len
);
1685 NLA_PUT_U32(skb
, NDTPA_PROXY_QLEN
, parms
->proxy_qlen
);
1686 NLA_PUT_U32(skb
, NDTPA_APP_PROBES
, parms
->app_probes
);
1687 NLA_PUT_U32(skb
, NDTPA_UCAST_PROBES
, parms
->ucast_probes
);
1688 NLA_PUT_U32(skb
, NDTPA_MCAST_PROBES
, parms
->mcast_probes
);
1689 NLA_PUT_MSECS(skb
, NDTPA_REACHABLE_TIME
, parms
->reachable_time
);
1690 NLA_PUT_MSECS(skb
, NDTPA_BASE_REACHABLE_TIME
,
1691 parms
->base_reachable_time
);
1692 NLA_PUT_MSECS(skb
, NDTPA_GC_STALETIME
, parms
->gc_staletime
);
1693 NLA_PUT_MSECS(skb
, NDTPA_DELAY_PROBE_TIME
, parms
->delay_probe_time
);
1694 NLA_PUT_MSECS(skb
, NDTPA_RETRANS_TIME
, parms
->retrans_time
);
1695 NLA_PUT_MSECS(skb
, NDTPA_ANYCAST_DELAY
, parms
->anycast_delay
);
1696 NLA_PUT_MSECS(skb
, NDTPA_PROXY_DELAY
, parms
->proxy_delay
);
1697 NLA_PUT_MSECS(skb
, NDTPA_LOCKTIME
, parms
->locktime
);
1699 return nla_nest_end(skb
, nest
);
1702 nla_nest_cancel(skb
, nest
);
1706 static int neightbl_fill_info(struct sk_buff
*skb
, struct neigh_table
*tbl
,
1707 u32 pid
, u32 seq
, int type
, int flags
)
1709 struct nlmsghdr
*nlh
;
1710 struct ndtmsg
*ndtmsg
;
1712 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1716 ndtmsg
= nlmsg_data(nlh
);
1718 read_lock_bh(&tbl
->lock
);
1719 ndtmsg
->ndtm_family
= tbl
->family
;
1720 ndtmsg
->ndtm_pad1
= 0;
1721 ndtmsg
->ndtm_pad2
= 0;
1723 NLA_PUT_STRING(skb
, NDTA_NAME
, tbl
->id
);
1724 NLA_PUT_MSECS(skb
, NDTA_GC_INTERVAL
, tbl
->gc_interval
);
1725 NLA_PUT_U32(skb
, NDTA_THRESH1
, tbl
->gc_thresh1
);
1726 NLA_PUT_U32(skb
, NDTA_THRESH2
, tbl
->gc_thresh2
);
1727 NLA_PUT_U32(skb
, NDTA_THRESH3
, tbl
->gc_thresh3
);
1730 unsigned long now
= jiffies
;
1731 unsigned int flush_delta
= now
- tbl
->last_flush
;
1732 unsigned int rand_delta
= now
- tbl
->last_rand
;
1734 struct ndt_config ndc
= {
1735 .ndtc_key_len
= tbl
->key_len
,
1736 .ndtc_entry_size
= tbl
->entry_size
,
1737 .ndtc_entries
= atomic_read(&tbl
->entries
),
1738 .ndtc_last_flush
= jiffies_to_msecs(flush_delta
),
1739 .ndtc_last_rand
= jiffies_to_msecs(rand_delta
),
1740 .ndtc_hash_rnd
= tbl
->hash_rnd
,
1741 .ndtc_hash_mask
= tbl
->hash_mask
,
1742 .ndtc_hash_chain_gc
= tbl
->hash_chain_gc
,
1743 .ndtc_proxy_qlen
= tbl
->proxy_queue
.qlen
,
1746 NLA_PUT(skb
, NDTA_CONFIG
, sizeof(ndc
), &ndc
);
1751 struct ndt_stats ndst
;
1753 memset(&ndst
, 0, sizeof(ndst
));
1755 for_each_possible_cpu(cpu
) {
1756 struct neigh_statistics
*st
;
1758 st
= per_cpu_ptr(tbl
->stats
, cpu
);
1759 ndst
.ndts_allocs
+= st
->allocs
;
1760 ndst
.ndts_destroys
+= st
->destroys
;
1761 ndst
.ndts_hash_grows
+= st
->hash_grows
;
1762 ndst
.ndts_res_failed
+= st
->res_failed
;
1763 ndst
.ndts_lookups
+= st
->lookups
;
1764 ndst
.ndts_hits
+= st
->hits
;
1765 ndst
.ndts_rcv_probes_mcast
+= st
->rcv_probes_mcast
;
1766 ndst
.ndts_rcv_probes_ucast
+= st
->rcv_probes_ucast
;
1767 ndst
.ndts_periodic_gc_runs
+= st
->periodic_gc_runs
;
1768 ndst
.ndts_forced_gc_runs
+= st
->forced_gc_runs
;
1771 NLA_PUT(skb
, NDTA_STATS
, sizeof(ndst
), &ndst
);
1774 BUG_ON(tbl
->parms
.dev
);
1775 if (neightbl_fill_parms(skb
, &tbl
->parms
) < 0)
1776 goto nla_put_failure
;
1778 read_unlock_bh(&tbl
->lock
);
1779 return nlmsg_end(skb
, nlh
);
1782 read_unlock_bh(&tbl
->lock
);
1783 nlmsg_cancel(skb
, nlh
);
1787 static int neightbl_fill_param_info(struct sk_buff
*skb
,
1788 struct neigh_table
*tbl
,
1789 struct neigh_parms
*parms
,
1790 u32 pid
, u32 seq
, int type
,
1793 struct ndtmsg
*ndtmsg
;
1794 struct nlmsghdr
*nlh
;
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
) < 0 ||
1808 neightbl_fill_parms(skb
, parms
) < 0)
1811 read_unlock_bh(&tbl
->lock
);
1812 return nlmsg_end(skb
, nlh
);
1814 read_unlock_bh(&tbl
->lock
);
1815 nlmsg_cancel(skb
, nlh
);
1819 static const struct nla_policy nl_neightbl_policy
[NDTA_MAX
+1] = {
1820 [NDTA_NAME
] = { .type
= NLA_STRING
},
1821 [NDTA_THRESH1
] = { .type
= NLA_U32
},
1822 [NDTA_THRESH2
] = { .type
= NLA_U32
},
1823 [NDTA_THRESH3
] = { .type
= NLA_U32
},
1824 [NDTA_GC_INTERVAL
] = { .type
= NLA_U64
},
1825 [NDTA_PARMS
] = { .type
= NLA_NESTED
},
1828 static const struct nla_policy nl_ntbl_parm_policy
[NDTPA_MAX
+1] = {
1829 [NDTPA_IFINDEX
] = { .type
= NLA_U32
},
1830 [NDTPA_QUEUE_LEN
] = { .type
= NLA_U32
},
1831 [NDTPA_PROXY_QLEN
] = { .type
= NLA_U32
},
1832 [NDTPA_APP_PROBES
] = { .type
= NLA_U32
},
1833 [NDTPA_UCAST_PROBES
] = { .type
= NLA_U32
},
1834 [NDTPA_MCAST_PROBES
] = { .type
= NLA_U32
},
1835 [NDTPA_BASE_REACHABLE_TIME
] = { .type
= NLA_U64
},
1836 [NDTPA_GC_STALETIME
] = { .type
= NLA_U64
},
1837 [NDTPA_DELAY_PROBE_TIME
] = { .type
= NLA_U64
},
1838 [NDTPA_RETRANS_TIME
] = { .type
= NLA_U64
},
1839 [NDTPA_ANYCAST_DELAY
] = { .type
= NLA_U64
},
1840 [NDTPA_PROXY_DELAY
] = { .type
= NLA_U64
},
1841 [NDTPA_LOCKTIME
] = { .type
= NLA_U64
},
1844 static int neightbl_set(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1846 struct net
*net
= sock_net(skb
->sk
);
1847 struct neigh_table
*tbl
;
1848 struct ndtmsg
*ndtmsg
;
1849 struct nlattr
*tb
[NDTA_MAX
+1];
1852 err
= nlmsg_parse(nlh
, sizeof(*ndtmsg
), tb
, NDTA_MAX
,
1853 nl_neightbl_policy
);
1857 if (tb
[NDTA_NAME
] == NULL
) {
1862 ndtmsg
= nlmsg_data(nlh
);
1863 read_lock(&neigh_tbl_lock
);
1864 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1865 if (ndtmsg
->ndtm_family
&& tbl
->family
!= ndtmsg
->ndtm_family
)
1868 if (nla_strcmp(tb
[NDTA_NAME
], tbl
->id
) == 0)
1878 * We acquire tbl->lock to be nice to the periodic timers and
1879 * make sure they always see a consistent set of values.
1881 write_lock_bh(&tbl
->lock
);
1883 if (tb
[NDTA_PARMS
]) {
1884 struct nlattr
*tbp
[NDTPA_MAX
+1];
1885 struct neigh_parms
*p
;
1888 err
= nla_parse_nested(tbp
, NDTPA_MAX
, tb
[NDTA_PARMS
],
1889 nl_ntbl_parm_policy
);
1891 goto errout_tbl_lock
;
1893 if (tbp
[NDTPA_IFINDEX
])
1894 ifindex
= nla_get_u32(tbp
[NDTPA_IFINDEX
]);
1896 p
= lookup_neigh_params(tbl
, net
, ifindex
);
1899 goto errout_tbl_lock
;
1902 for (i
= 1; i
<= NDTPA_MAX
; i
++) {
1907 case NDTPA_QUEUE_LEN
:
1908 p
->queue_len
= nla_get_u32(tbp
[i
]);
1910 case NDTPA_PROXY_QLEN
:
1911 p
->proxy_qlen
= nla_get_u32(tbp
[i
]);
1913 case NDTPA_APP_PROBES
:
1914 p
->app_probes
= nla_get_u32(tbp
[i
]);
1916 case NDTPA_UCAST_PROBES
:
1917 p
->ucast_probes
= nla_get_u32(tbp
[i
]);
1919 case NDTPA_MCAST_PROBES
:
1920 p
->mcast_probes
= nla_get_u32(tbp
[i
]);
1922 case NDTPA_BASE_REACHABLE_TIME
:
1923 p
->base_reachable_time
= nla_get_msecs(tbp
[i
]);
1925 case NDTPA_GC_STALETIME
:
1926 p
->gc_staletime
= nla_get_msecs(tbp
[i
]);
1928 case NDTPA_DELAY_PROBE_TIME
:
1929 p
->delay_probe_time
= nla_get_msecs(tbp
[i
]);
1931 case NDTPA_RETRANS_TIME
:
1932 p
->retrans_time
= nla_get_msecs(tbp
[i
]);
1934 case NDTPA_ANYCAST_DELAY
:
1935 p
->anycast_delay
= nla_get_msecs(tbp
[i
]);
1937 case NDTPA_PROXY_DELAY
:
1938 p
->proxy_delay
= nla_get_msecs(tbp
[i
]);
1940 case NDTPA_LOCKTIME
:
1941 p
->locktime
= nla_get_msecs(tbp
[i
]);
1947 if (tb
[NDTA_THRESH1
])
1948 tbl
->gc_thresh1
= nla_get_u32(tb
[NDTA_THRESH1
]);
1950 if (tb
[NDTA_THRESH2
])
1951 tbl
->gc_thresh2
= nla_get_u32(tb
[NDTA_THRESH2
]);
1953 if (tb
[NDTA_THRESH3
])
1954 tbl
->gc_thresh3
= nla_get_u32(tb
[NDTA_THRESH3
]);
1956 if (tb
[NDTA_GC_INTERVAL
])
1957 tbl
->gc_interval
= nla_get_msecs(tb
[NDTA_GC_INTERVAL
]);
1962 write_unlock_bh(&tbl
->lock
);
1964 read_unlock(&neigh_tbl_lock
);
1969 static int neightbl_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1971 struct net
*net
= sock_net(skb
->sk
);
1972 int family
, tidx
, nidx
= 0;
1973 int tbl_skip
= cb
->args
[0];
1974 int neigh_skip
= cb
->args
[1];
1975 struct neigh_table
*tbl
;
1977 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
1979 read_lock(&neigh_tbl_lock
);
1980 for (tbl
= neigh_tables
, tidx
= 0; tbl
; tbl
= tbl
->next
, tidx
++) {
1981 struct neigh_parms
*p
;
1983 if (tidx
< tbl_skip
|| (family
&& tbl
->family
!= family
))
1986 if (neightbl_fill_info(skb
, tbl
, NETLINK_CB(cb
->skb
).pid
,
1987 cb
->nlh
->nlmsg_seq
, RTM_NEWNEIGHTBL
,
1991 for (nidx
= 0, p
= tbl
->parms
.next
; p
; p
= p
->next
) {
1992 if (!net_eq(neigh_parms_net(p
), net
))
1995 if (nidx
< neigh_skip
)
1998 if (neightbl_fill_param_info(skb
, tbl
, p
,
1999 NETLINK_CB(cb
->skb
).pid
,
2011 read_unlock(&neigh_tbl_lock
);
2018 static int neigh_fill_info(struct sk_buff
*skb
, struct neighbour
*neigh
,
2019 u32 pid
, u32 seq
, int type
, unsigned int flags
)
2021 unsigned long now
= jiffies
;
2022 struct nda_cacheinfo ci
;
2023 struct nlmsghdr
*nlh
;
2026 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), flags
);
2030 ndm
= nlmsg_data(nlh
);
2031 ndm
->ndm_family
= neigh
->ops
->family
;
2034 ndm
->ndm_flags
= neigh
->flags
;
2035 ndm
->ndm_type
= neigh
->type
;
2036 ndm
->ndm_ifindex
= neigh
->dev
->ifindex
;
2038 NLA_PUT(skb
, NDA_DST
, neigh
->tbl
->key_len
, neigh
->primary_key
);
2040 read_lock_bh(&neigh
->lock
);
2041 ndm
->ndm_state
= neigh
->nud_state
;
2042 if ((neigh
->nud_state
& NUD_VALID
) &&
2043 nla_put(skb
, NDA_LLADDR
, neigh
->dev
->addr_len
, neigh
->ha
) < 0) {
2044 read_unlock_bh(&neigh
->lock
);
2045 goto nla_put_failure
;
2048 ci
.ndm_used
= jiffies_to_clock_t(now
- neigh
->used
);
2049 ci
.ndm_confirmed
= jiffies_to_clock_t(now
- neigh
->confirmed
);
2050 ci
.ndm_updated
= jiffies_to_clock_t(now
- neigh
->updated
);
2051 ci
.ndm_refcnt
= atomic_read(&neigh
->refcnt
) - 1;
2052 read_unlock_bh(&neigh
->lock
);
2054 NLA_PUT_U32(skb
, NDA_PROBES
, atomic_read(&neigh
->probes
));
2055 NLA_PUT(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
);
2057 return nlmsg_end(skb
, nlh
);
2060 nlmsg_cancel(skb
, nlh
);
2064 static void neigh_update_notify(struct neighbour
*neigh
)
2066 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, neigh
);
2067 __neigh_notify(neigh
, RTM_NEWNEIGH
, 0);
2070 static int neigh_dump_table(struct neigh_table
*tbl
, struct sk_buff
*skb
,
2071 struct netlink_callback
*cb
)
2073 struct net
* net
= sock_net(skb
->sk
);
2074 struct neighbour
*n
;
2075 int rc
, h
, s_h
= cb
->args
[1];
2076 int idx
, s_idx
= idx
= cb
->args
[2];
2078 read_lock_bh(&tbl
->lock
);
2079 for (h
= 0; h
<= tbl
->hash_mask
; h
++) {
2084 for (n
= tbl
->hash_buckets
[h
], idx
= 0; n
; n
= n
->next
) {
2085 if (dev_net(n
->dev
) != net
)
2089 if (neigh_fill_info(skb
, n
, NETLINK_CB(cb
->skb
).pid
,
2092 NLM_F_MULTI
) <= 0) {
2093 read_unlock_bh(&tbl
->lock
);
2101 read_unlock_bh(&tbl
->lock
);
2109 static int neigh_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2111 struct neigh_table
*tbl
;
2114 read_lock(&neigh_tbl_lock
);
2115 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
2118 for (tbl
= neigh_tables
, t
= 0; tbl
; tbl
= tbl
->next
, t
++) {
2119 if (t
< s_t
|| (family
&& tbl
->family
!= family
))
2122 memset(&cb
->args
[1], 0, sizeof(cb
->args
) -
2123 sizeof(cb
->args
[0]));
2124 if (neigh_dump_table(tbl
, skb
, cb
) < 0)
2127 read_unlock(&neigh_tbl_lock
);
2133 void neigh_for_each(struct neigh_table
*tbl
, void (*cb
)(struct neighbour
*, void *), void *cookie
)
2137 read_lock_bh(&tbl
->lock
);
2138 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2139 struct neighbour
*n
;
2141 for (n
= tbl
->hash_buckets
[chain
]; n
; n
= n
->next
)
2144 read_unlock_bh(&tbl
->lock
);
2146 EXPORT_SYMBOL(neigh_for_each
);
2148 /* The tbl->lock must be held as a writer and BH disabled. */
2149 void __neigh_for_each_release(struct neigh_table
*tbl
,
2150 int (*cb
)(struct neighbour
*))
2154 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2155 struct neighbour
*n
, **np
;
2157 np
= &tbl
->hash_buckets
[chain
];
2158 while ((n
= *np
) != NULL
) {
2161 write_lock(&n
->lock
);
2168 write_unlock(&n
->lock
);
2170 neigh_cleanup_and_release(n
);
2174 EXPORT_SYMBOL(__neigh_for_each_release
);
2176 #ifdef CONFIG_PROC_FS
2178 static struct neighbour
*neigh_get_first(struct seq_file
*seq
)
2180 struct neigh_seq_state
*state
= seq
->private;
2181 struct net
*net
= seq_file_net(seq
);
2182 struct neigh_table
*tbl
= state
->tbl
;
2183 struct neighbour
*n
= NULL
;
2184 int bucket
= state
->bucket
;
2186 state
->flags
&= ~NEIGH_SEQ_IS_PNEIGH
;
2187 for (bucket
= 0; bucket
<= tbl
->hash_mask
; bucket
++) {
2188 n
= tbl
->hash_buckets
[bucket
];
2191 if (!net_eq(dev_net(n
->dev
), net
))
2193 if (state
->neigh_sub_iter
) {
2197 v
= state
->neigh_sub_iter(state
, n
, &fakep
);
2201 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2203 if (n
->nud_state
& ~NUD_NOARP
)
2212 state
->bucket
= bucket
;
2217 static struct neighbour
*neigh_get_next(struct seq_file
*seq
,
2218 struct neighbour
*n
,
2221 struct neigh_seq_state
*state
= seq
->private;
2222 struct net
*net
= seq_file_net(seq
);
2223 struct neigh_table
*tbl
= state
->tbl
;
2225 if (state
->neigh_sub_iter
) {
2226 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2234 if (!net_eq(dev_net(n
->dev
), net
))
2236 if (state
->neigh_sub_iter
) {
2237 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2242 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2245 if (n
->nud_state
& ~NUD_NOARP
)
2254 if (++state
->bucket
> tbl
->hash_mask
)
2257 n
= tbl
->hash_buckets
[state
->bucket
];
2265 static struct neighbour
*neigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2267 struct neighbour
*n
= neigh_get_first(seq
);
2272 n
= neigh_get_next(seq
, n
, pos
);
2277 return *pos
? NULL
: n
;
2280 static struct pneigh_entry
*pneigh_get_first(struct seq_file
*seq
)
2282 struct neigh_seq_state
*state
= seq
->private;
2283 struct net
*net
= seq_file_net(seq
);
2284 struct neigh_table
*tbl
= state
->tbl
;
2285 struct pneigh_entry
*pn
= NULL
;
2286 int bucket
= state
->bucket
;
2288 state
->flags
|= NEIGH_SEQ_IS_PNEIGH
;
2289 for (bucket
= 0; bucket
<= PNEIGH_HASHMASK
; bucket
++) {
2290 pn
= tbl
->phash_buckets
[bucket
];
2291 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2296 state
->bucket
= bucket
;
2301 static struct pneigh_entry
*pneigh_get_next(struct seq_file
*seq
,
2302 struct pneigh_entry
*pn
,
2305 struct neigh_seq_state
*state
= seq
->private;
2306 struct net
*net
= seq_file_net(seq
);
2307 struct neigh_table
*tbl
= state
->tbl
;
2311 if (++state
->bucket
> PNEIGH_HASHMASK
)
2313 pn
= tbl
->phash_buckets
[state
->bucket
];
2314 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2326 static struct pneigh_entry
*pneigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2328 struct pneigh_entry
*pn
= pneigh_get_first(seq
);
2333 pn
= pneigh_get_next(seq
, pn
, pos
);
2338 return *pos
? NULL
: pn
;
2341 static void *neigh_get_idx_any(struct seq_file
*seq
, loff_t
*pos
)
2343 struct neigh_seq_state
*state
= seq
->private;
2345 loff_t idxpos
= *pos
;
2347 rc
= neigh_get_idx(seq
, &idxpos
);
2348 if (!rc
&& !(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2349 rc
= pneigh_get_idx(seq
, &idxpos
);
2354 void *neigh_seq_start(struct seq_file
*seq
, loff_t
*pos
, struct neigh_table
*tbl
, unsigned int neigh_seq_flags
)
2355 __acquires(tbl
->lock
)
2357 struct neigh_seq_state
*state
= seq
->private;
2361 state
->flags
= (neigh_seq_flags
& ~NEIGH_SEQ_IS_PNEIGH
);
2363 read_lock_bh(&tbl
->lock
);
2365 return *pos
? neigh_get_idx_any(seq
, pos
) : SEQ_START_TOKEN
;
2367 EXPORT_SYMBOL(neigh_seq_start
);
2369 void *neigh_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2371 struct neigh_seq_state
*state
;
2374 if (v
== SEQ_START_TOKEN
) {
2375 rc
= neigh_get_first(seq
);
2379 state
= seq
->private;
2380 if (!(state
->flags
& NEIGH_SEQ_IS_PNEIGH
)) {
2381 rc
= neigh_get_next(seq
, v
, NULL
);
2384 if (!(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2385 rc
= pneigh_get_first(seq
);
2387 BUG_ON(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
);
2388 rc
= pneigh_get_next(seq
, v
, NULL
);
2394 EXPORT_SYMBOL(neigh_seq_next
);
2396 void neigh_seq_stop(struct seq_file
*seq
, void *v
)
2397 __releases(tbl
->lock
)
2399 struct neigh_seq_state
*state
= seq
->private;
2400 struct neigh_table
*tbl
= state
->tbl
;
2402 read_unlock_bh(&tbl
->lock
);
2404 EXPORT_SYMBOL(neigh_seq_stop
);
2406 /* statistics via seq_file */
2408 static void *neigh_stat_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2410 struct proc_dir_entry
*pde
= seq
->private;
2411 struct neigh_table
*tbl
= pde
->data
;
2415 return SEQ_START_TOKEN
;
2417 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
2418 if (!cpu_possible(cpu
))
2421 return per_cpu_ptr(tbl
->stats
, cpu
);
2426 static void *neigh_stat_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2428 struct proc_dir_entry
*pde
= seq
->private;
2429 struct neigh_table
*tbl
= pde
->data
;
2432 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
2433 if (!cpu_possible(cpu
))
2436 return per_cpu_ptr(tbl
->stats
, cpu
);
2441 static void neigh_stat_seq_stop(struct seq_file
*seq
, void *v
)
2446 static int neigh_stat_seq_show(struct seq_file
*seq
, void *v
)
2448 struct proc_dir_entry
*pde
= seq
->private;
2449 struct neigh_table
*tbl
= pde
->data
;
2450 struct neigh_statistics
*st
= v
;
2452 if (v
== SEQ_START_TOKEN
) {
2453 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\n");
2457 seq_printf(seq
, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2458 "%08lx %08lx %08lx %08lx %08lx\n",
2459 atomic_read(&tbl
->entries
),
2470 st
->rcv_probes_mcast
,
2471 st
->rcv_probes_ucast
,
2473 st
->periodic_gc_runs
,
2481 static const struct seq_operations neigh_stat_seq_ops
= {
2482 .start
= neigh_stat_seq_start
,
2483 .next
= neigh_stat_seq_next
,
2484 .stop
= neigh_stat_seq_stop
,
2485 .show
= neigh_stat_seq_show
,
2488 static int neigh_stat_seq_open(struct inode
*inode
, struct file
*file
)
2490 int ret
= seq_open(file
, &neigh_stat_seq_ops
);
2493 struct seq_file
*sf
= file
->private_data
;
2494 sf
->private = PDE(inode
);
2499 static const struct file_operations neigh_stat_seq_fops
= {
2500 .owner
= THIS_MODULE
,
2501 .open
= neigh_stat_seq_open
,
2503 .llseek
= seq_lseek
,
2504 .release
= seq_release
,
2507 #endif /* CONFIG_PROC_FS */
2509 static inline size_t neigh_nlmsg_size(void)
2511 return NLMSG_ALIGN(sizeof(struct ndmsg
))
2512 + nla_total_size(MAX_ADDR_LEN
) /* NDA_DST */
2513 + nla_total_size(MAX_ADDR_LEN
) /* NDA_LLADDR */
2514 + nla_total_size(sizeof(struct nda_cacheinfo
))
2515 + nla_total_size(4); /* NDA_PROBES */
2518 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
)
2520 struct net
*net
= dev_net(n
->dev
);
2521 struct sk_buff
*skb
;
2524 skb
= nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC
);
2528 err
= neigh_fill_info(skb
, n
, 0, 0, type
, flags
);
2530 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2531 WARN_ON(err
== -EMSGSIZE
);
2535 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2539 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2543 void neigh_app_ns(struct neighbour
*n
)
2545 __neigh_notify(n
, RTM_GETNEIGH
, NLM_F_REQUEST
);
2547 EXPORT_SYMBOL(neigh_app_ns
);
2548 #endif /* CONFIG_ARPD */
2550 #ifdef CONFIG_SYSCTL
2552 static struct neigh_sysctl_table
{
2553 struct ctl_table_header
*sysctl_header
;
2554 struct ctl_table neigh_vars
[__NET_NEIGH_MAX
];
2556 } neigh_sysctl_template __read_mostly
= {
2559 .ctl_name
= NET_NEIGH_MCAST_SOLICIT
,
2560 .procname
= "mcast_solicit",
2561 .maxlen
= sizeof(int),
2563 .proc_handler
= proc_dointvec
,
2566 .ctl_name
= NET_NEIGH_UCAST_SOLICIT
,
2567 .procname
= "ucast_solicit",
2568 .maxlen
= sizeof(int),
2570 .proc_handler
= proc_dointvec
,
2573 .ctl_name
= NET_NEIGH_APP_SOLICIT
,
2574 .procname
= "app_solicit",
2575 .maxlen
= sizeof(int),
2577 .proc_handler
= proc_dointvec
,
2580 .procname
= "retrans_time",
2581 .maxlen
= sizeof(int),
2583 .proc_handler
= proc_dointvec_userhz_jiffies
,
2586 .ctl_name
= NET_NEIGH_REACHABLE_TIME
,
2587 .procname
= "base_reachable_time",
2588 .maxlen
= sizeof(int),
2590 .proc_handler
= proc_dointvec_jiffies
,
2591 .strategy
= sysctl_jiffies
,
2594 .ctl_name
= NET_NEIGH_DELAY_PROBE_TIME
,
2595 .procname
= "delay_first_probe_time",
2596 .maxlen
= sizeof(int),
2598 .proc_handler
= proc_dointvec_jiffies
,
2599 .strategy
= sysctl_jiffies
,
2602 .ctl_name
= NET_NEIGH_GC_STALE_TIME
,
2603 .procname
= "gc_stale_time",
2604 .maxlen
= sizeof(int),
2606 .proc_handler
= proc_dointvec_jiffies
,
2607 .strategy
= sysctl_jiffies
,
2610 .ctl_name
= NET_NEIGH_UNRES_QLEN
,
2611 .procname
= "unres_qlen",
2612 .maxlen
= sizeof(int),
2614 .proc_handler
= proc_dointvec
,
2617 .ctl_name
= NET_NEIGH_PROXY_QLEN
,
2618 .procname
= "proxy_qlen",
2619 .maxlen
= sizeof(int),
2621 .proc_handler
= proc_dointvec
,
2624 .procname
= "anycast_delay",
2625 .maxlen
= sizeof(int),
2627 .proc_handler
= proc_dointvec_userhz_jiffies
,
2630 .procname
= "proxy_delay",
2631 .maxlen
= sizeof(int),
2633 .proc_handler
= proc_dointvec_userhz_jiffies
,
2636 .procname
= "locktime",
2637 .maxlen
= sizeof(int),
2639 .proc_handler
= proc_dointvec_userhz_jiffies
,
2642 .ctl_name
= NET_NEIGH_RETRANS_TIME_MS
,
2643 .procname
= "retrans_time_ms",
2644 .maxlen
= sizeof(int),
2646 .proc_handler
= proc_dointvec_ms_jiffies
,
2647 .strategy
= sysctl_ms_jiffies
,
2650 .ctl_name
= NET_NEIGH_REACHABLE_TIME_MS
,
2651 .procname
= "base_reachable_time_ms",
2652 .maxlen
= sizeof(int),
2654 .proc_handler
= proc_dointvec_ms_jiffies
,
2655 .strategy
= sysctl_ms_jiffies
,
2658 .ctl_name
= NET_NEIGH_GC_INTERVAL
,
2659 .procname
= "gc_interval",
2660 .maxlen
= sizeof(int),
2662 .proc_handler
= proc_dointvec_jiffies
,
2663 .strategy
= sysctl_jiffies
,
2666 .ctl_name
= NET_NEIGH_GC_THRESH1
,
2667 .procname
= "gc_thresh1",
2668 .maxlen
= sizeof(int),
2670 .proc_handler
= proc_dointvec
,
2673 .ctl_name
= NET_NEIGH_GC_THRESH2
,
2674 .procname
= "gc_thresh2",
2675 .maxlen
= sizeof(int),
2677 .proc_handler
= proc_dointvec
,
2680 .ctl_name
= NET_NEIGH_GC_THRESH3
,
2681 .procname
= "gc_thresh3",
2682 .maxlen
= sizeof(int),
2684 .proc_handler
= proc_dointvec
,
2690 int neigh_sysctl_register(struct net_device
*dev
, struct neigh_parms
*p
,
2691 int p_id
, int pdev_id
, char *p_name
,
2692 proc_handler
*handler
, ctl_handler
*strategy
)
2694 struct neigh_sysctl_table
*t
;
2695 const char *dev_name_source
= NULL
;
2697 #define NEIGH_CTL_PATH_ROOT 0
2698 #define NEIGH_CTL_PATH_PROTO 1
2699 #define NEIGH_CTL_PATH_NEIGH 2
2700 #define NEIGH_CTL_PATH_DEV 3
2702 struct ctl_path neigh_path
[] = {
2703 { .procname
= "net", .ctl_name
= CTL_NET
, },
2704 { .procname
= "proto", .ctl_name
= 0, },
2705 { .procname
= "neigh", .ctl_name
= 0, },
2706 { .procname
= "default", .ctl_name
= NET_PROTO_CONF_DEFAULT
, },
2710 t
= kmemdup(&neigh_sysctl_template
, sizeof(*t
), GFP_KERNEL
);
2714 t
->neigh_vars
[0].data
= &p
->mcast_probes
;
2715 t
->neigh_vars
[1].data
= &p
->ucast_probes
;
2716 t
->neigh_vars
[2].data
= &p
->app_probes
;
2717 t
->neigh_vars
[3].data
= &p
->retrans_time
;
2718 t
->neigh_vars
[4].data
= &p
->base_reachable_time
;
2719 t
->neigh_vars
[5].data
= &p
->delay_probe_time
;
2720 t
->neigh_vars
[6].data
= &p
->gc_staletime
;
2721 t
->neigh_vars
[7].data
= &p
->queue_len
;
2722 t
->neigh_vars
[8].data
= &p
->proxy_qlen
;
2723 t
->neigh_vars
[9].data
= &p
->anycast_delay
;
2724 t
->neigh_vars
[10].data
= &p
->proxy_delay
;
2725 t
->neigh_vars
[11].data
= &p
->locktime
;
2726 t
->neigh_vars
[12].data
= &p
->retrans_time
;
2727 t
->neigh_vars
[13].data
= &p
->base_reachable_time
;
2730 dev_name_source
= dev
->name
;
2731 neigh_path
[NEIGH_CTL_PATH_DEV
].ctl_name
= dev
->ifindex
;
2732 /* Terminate the table early */
2733 memset(&t
->neigh_vars
[14], 0, sizeof(t
->neigh_vars
[14]));
2735 dev_name_source
= neigh_path
[NEIGH_CTL_PATH_DEV
].procname
;
2736 t
->neigh_vars
[14].data
= (int *)(p
+ 1);
2737 t
->neigh_vars
[15].data
= (int *)(p
+ 1) + 1;
2738 t
->neigh_vars
[16].data
= (int *)(p
+ 1) + 2;
2739 t
->neigh_vars
[17].data
= (int *)(p
+ 1) + 3;
2743 if (handler
|| strategy
) {
2745 t
->neigh_vars
[3].proc_handler
= handler
;
2746 t
->neigh_vars
[3].strategy
= strategy
;
2747 t
->neigh_vars
[3].extra1
= dev
;
2749 t
->neigh_vars
[3].ctl_name
= CTL_UNNUMBERED
;
2751 t
->neigh_vars
[4].proc_handler
= handler
;
2752 t
->neigh_vars
[4].strategy
= strategy
;
2753 t
->neigh_vars
[4].extra1
= dev
;
2755 t
->neigh_vars
[4].ctl_name
= CTL_UNNUMBERED
;
2756 /* RetransTime (in milliseconds)*/
2757 t
->neigh_vars
[12].proc_handler
= handler
;
2758 t
->neigh_vars
[12].strategy
= strategy
;
2759 t
->neigh_vars
[12].extra1
= dev
;
2761 t
->neigh_vars
[12].ctl_name
= CTL_UNNUMBERED
;
2762 /* ReachableTime (in milliseconds) */
2763 t
->neigh_vars
[13].proc_handler
= handler
;
2764 t
->neigh_vars
[13].strategy
= strategy
;
2765 t
->neigh_vars
[13].extra1
= dev
;
2767 t
->neigh_vars
[13].ctl_name
= CTL_UNNUMBERED
;
2770 t
->dev_name
= kstrdup(dev_name_source
, GFP_KERNEL
);
2774 neigh_path
[NEIGH_CTL_PATH_DEV
].procname
= t
->dev_name
;
2775 neigh_path
[NEIGH_CTL_PATH_NEIGH
].ctl_name
= pdev_id
;
2776 neigh_path
[NEIGH_CTL_PATH_PROTO
].procname
= p_name
;
2777 neigh_path
[NEIGH_CTL_PATH_PROTO
].ctl_name
= p_id
;
2780 register_net_sysctl_table(neigh_parms_net(p
), neigh_path
, t
->neigh_vars
);
2781 if (!t
->sysctl_header
)
2784 p
->sysctl_table
= t
;
2794 EXPORT_SYMBOL(neigh_sysctl_register
);
2796 void neigh_sysctl_unregister(struct neigh_parms
*p
)
2798 if (p
->sysctl_table
) {
2799 struct neigh_sysctl_table
*t
= p
->sysctl_table
;
2800 p
->sysctl_table
= NULL
;
2801 unregister_sysctl_table(t
->sysctl_header
);
2806 EXPORT_SYMBOL(neigh_sysctl_unregister
);
2808 #endif /* CONFIG_SYSCTL */
2810 static int __init
neigh_init(void)
2812 rtnl_register(PF_UNSPEC
, RTM_NEWNEIGH
, neigh_add
, NULL
);
2813 rtnl_register(PF_UNSPEC
, RTM_DELNEIGH
, neigh_delete
, NULL
);
2814 rtnl_register(PF_UNSPEC
, RTM_GETNEIGH
, NULL
, neigh_dump_info
);
2816 rtnl_register(PF_UNSPEC
, RTM_GETNEIGHTBL
, NULL
, neightbl_dump_info
);
2817 rtnl_register(PF_UNSPEC
, RTM_SETNEIGHTBL
, neightbl_set
, NULL
);
2822 subsys_initcall(neigh_init
);