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/slab.h>
19 #include <linux/types.h>
20 #include <linux/kernel.h>
21 #include <linux/module.h>
22 #include <linux/socket.h>
23 #include <linux/netdevice.h>
24 #include <linux/proc_fs.h>
26 #include <linux/sysctl.h>
28 #include <linux/times.h>
29 #include <net/net_namespace.h>
30 #include <net/neighbour.h>
33 #include <net/netevent.h>
34 #include <net/netlink.h>
35 #include <linux/rtnetlink.h>
36 #include <linux/random.h>
37 #include <linux/string.h>
38 #include <linux/log2.h>
42 #define NEIGH_PRINTK(x...) printk(x)
43 #define NEIGH_NOPRINTK(x...) do { ; } while(0)
44 #define NEIGH_PRINTK0 NEIGH_PRINTK
45 #define NEIGH_PRINTK1 NEIGH_NOPRINTK
46 #define NEIGH_PRINTK2 NEIGH_NOPRINTK
50 #define NEIGH_PRINTK1 NEIGH_PRINTK
54 #define NEIGH_PRINTK2 NEIGH_PRINTK
57 #define PNEIGH_HASHMASK 0xF
59 static void neigh_timer_handler(unsigned long arg
);
60 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
);
61 static void neigh_update_notify(struct neighbour
*neigh
);
62 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
);
64 static struct neigh_table
*neigh_tables
;
66 static const struct file_operations neigh_stat_seq_fops
;
70 Neighbour hash table buckets are protected with rwlock tbl->lock.
72 - All the scans/updates to hash buckets MUST be made under this lock.
73 - NOTHING clever should be made under this lock: no callbacks
74 to protocol backends, no attempts to send something to network.
75 It will result in deadlocks, if backend/driver wants to use neighbour
77 - If the entry requires some non-trivial actions, increase
78 its reference count and release table lock.
80 Neighbour entries are protected:
81 - with reference count.
82 - with rwlock neigh->lock
84 Reference count prevents destruction.
86 neigh->lock mainly serializes ll address data and its validity state.
87 However, the same lock is used to protect another entry fields:
91 Again, nothing clever shall be made under neigh->lock,
92 the most complicated procedure, which we allow is dev->hard_header.
93 It is supposed, that dev->hard_header is simplistic and does
94 not make callbacks to neighbour tables.
96 The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
97 list of neighbour tables. This list is used only in process context,
100 static DEFINE_RWLOCK(neigh_tbl_lock
);
102 static int neigh_blackhole(struct sk_buff
*skb
)
108 static void neigh_cleanup_and_release(struct neighbour
*neigh
)
110 if (neigh
->parms
->neigh_cleanup
)
111 neigh
->parms
->neigh_cleanup(neigh
);
113 __neigh_notify(neigh
, RTM_DELNEIGH
, 0);
114 neigh_release(neigh
);
118 * It is random distribution in the interval (1/2)*base...(3/2)*base.
119 * It corresponds to default IPv6 settings and is not overridable,
120 * because it is really reasonable choice.
123 unsigned long neigh_rand_reach_time(unsigned long base
)
125 return (base
? (net_random() % base
) + (base
>> 1) : 0);
127 EXPORT_SYMBOL(neigh_rand_reach_time
);
130 static int neigh_forced_gc(struct neigh_table
*tbl
)
135 NEIGH_CACHE_STAT_INC(tbl
, forced_gc_runs
);
137 write_lock_bh(&tbl
->lock
);
138 for (i
= 0; i
<= tbl
->hash_mask
; i
++) {
139 struct neighbour
*n
, **np
;
141 np
= &tbl
->hash_buckets
[i
];
142 while ((n
= *np
) != NULL
) {
143 /* Neighbour record may be discarded if:
144 * - nobody refers to it.
145 * - it is not permanent
147 write_lock(&n
->lock
);
148 if (atomic_read(&n
->refcnt
) == 1 &&
149 !(n
->nud_state
& NUD_PERMANENT
)) {
153 write_unlock(&n
->lock
);
154 neigh_cleanup_and_release(n
);
157 write_unlock(&n
->lock
);
162 tbl
->last_flush
= jiffies
;
164 write_unlock_bh(&tbl
->lock
);
169 static void neigh_add_timer(struct neighbour
*n
, unsigned long when
)
172 if (unlikely(mod_timer(&n
->timer
, when
))) {
173 printk("NEIGH: BUG, double timer add, state is %x\n",
179 static int neigh_del_timer(struct neighbour
*n
)
181 if ((n
->nud_state
& NUD_IN_TIMER
) &&
182 del_timer(&n
->timer
)) {
189 static void pneigh_queue_purge(struct sk_buff_head
*list
)
193 while ((skb
= skb_dequeue(list
)) != NULL
) {
199 static void neigh_flush_dev(struct neigh_table
*tbl
, struct net_device
*dev
)
203 for (i
= 0; i
<= tbl
->hash_mask
; i
++) {
204 struct neighbour
*n
, **np
= &tbl
->hash_buckets
[i
];
206 while ((n
= *np
) != NULL
) {
207 if (dev
&& n
->dev
!= dev
) {
212 write_lock(&n
->lock
);
216 if (atomic_read(&n
->refcnt
) != 1) {
217 /* The most unpleasant situation.
218 We must destroy neighbour entry,
219 but someone still uses it.
221 The destroy will be delayed until
222 the last user releases us, but
223 we must kill timers etc. and move
226 skb_queue_purge(&n
->arp_queue
);
227 n
->output
= neigh_blackhole
;
228 if (n
->nud_state
& NUD_VALID
)
229 n
->nud_state
= NUD_NOARP
;
231 n
->nud_state
= NUD_NONE
;
232 NEIGH_PRINTK2("neigh %p is stray.\n", n
);
234 write_unlock(&n
->lock
);
235 neigh_cleanup_and_release(n
);
240 void neigh_changeaddr(struct neigh_table
*tbl
, struct net_device
*dev
)
242 write_lock_bh(&tbl
->lock
);
243 neigh_flush_dev(tbl
, dev
);
244 write_unlock_bh(&tbl
->lock
);
246 EXPORT_SYMBOL(neigh_changeaddr
);
248 int neigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
250 write_lock_bh(&tbl
->lock
);
251 neigh_flush_dev(tbl
, dev
);
252 pneigh_ifdown(tbl
, dev
);
253 write_unlock_bh(&tbl
->lock
);
255 del_timer_sync(&tbl
->proxy_timer
);
256 pneigh_queue_purge(&tbl
->proxy_queue
);
259 EXPORT_SYMBOL(neigh_ifdown
);
261 static struct neighbour
*neigh_alloc(struct neigh_table
*tbl
)
263 struct neighbour
*n
= NULL
;
264 unsigned long now
= jiffies
;
267 entries
= atomic_inc_return(&tbl
->entries
) - 1;
268 if (entries
>= tbl
->gc_thresh3
||
269 (entries
>= tbl
->gc_thresh2
&&
270 time_after(now
, tbl
->last_flush
+ 5 * HZ
))) {
271 if (!neigh_forced_gc(tbl
) &&
272 entries
>= tbl
->gc_thresh3
)
276 n
= kmem_cache_zalloc(tbl
->kmem_cachep
, GFP_ATOMIC
);
280 skb_queue_head_init(&n
->arp_queue
);
281 rwlock_init(&n
->lock
);
282 n
->updated
= n
->used
= now
;
283 n
->nud_state
= NUD_NONE
;
284 n
->output
= neigh_blackhole
;
285 n
->parms
= neigh_parms_clone(&tbl
->parms
);
286 setup_timer(&n
->timer
, neigh_timer_handler
, (unsigned long)n
);
288 NEIGH_CACHE_STAT_INC(tbl
, allocs
);
290 atomic_set(&n
->refcnt
, 1);
296 atomic_dec(&tbl
->entries
);
300 static struct neighbour
**neigh_hash_alloc(unsigned int entries
)
302 unsigned long size
= entries
* sizeof(struct neighbour
*);
303 struct neighbour
**ret
;
305 if (size
<= PAGE_SIZE
) {
306 ret
= kzalloc(size
, GFP_ATOMIC
);
308 ret
= (struct neighbour
**)
309 __get_free_pages(GFP_ATOMIC
|__GFP_ZERO
, get_order(size
));
314 static void neigh_hash_free(struct neighbour
**hash
, unsigned int entries
)
316 unsigned long size
= entries
* sizeof(struct neighbour
*);
318 if (size
<= PAGE_SIZE
)
321 free_pages((unsigned long)hash
, get_order(size
));
324 static void neigh_hash_grow(struct neigh_table
*tbl
, unsigned long new_entries
)
326 struct neighbour
**new_hash
, **old_hash
;
327 unsigned int i
, new_hash_mask
, old_entries
;
329 NEIGH_CACHE_STAT_INC(tbl
, hash_grows
);
331 BUG_ON(!is_power_of_2(new_entries
));
332 new_hash
= neigh_hash_alloc(new_entries
);
336 old_entries
= tbl
->hash_mask
+ 1;
337 new_hash_mask
= new_entries
- 1;
338 old_hash
= tbl
->hash_buckets
;
340 get_random_bytes(&tbl
->hash_rnd
, sizeof(tbl
->hash_rnd
));
341 for (i
= 0; i
< old_entries
; i
++) {
342 struct neighbour
*n
, *next
;
344 for (n
= old_hash
[i
]; n
; n
= next
) {
345 unsigned int hash_val
= tbl
->hash(n
->primary_key
, n
->dev
);
347 hash_val
&= new_hash_mask
;
350 n
->next
= new_hash
[hash_val
];
351 new_hash
[hash_val
] = n
;
354 tbl
->hash_buckets
= new_hash
;
355 tbl
->hash_mask
= new_hash_mask
;
357 neigh_hash_free(old_hash
, old_entries
);
360 struct neighbour
*neigh_lookup(struct neigh_table
*tbl
, const void *pkey
,
361 struct net_device
*dev
)
364 int key_len
= tbl
->key_len
;
367 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
369 read_lock_bh(&tbl
->lock
);
370 hash_val
= tbl
->hash(pkey
, dev
);
371 for (n
= tbl
->hash_buckets
[hash_val
& tbl
->hash_mask
]; n
; n
= n
->next
) {
372 if (dev
== n
->dev
&& !memcmp(n
->primary_key
, pkey
, key_len
)) {
374 NEIGH_CACHE_STAT_INC(tbl
, hits
);
378 read_unlock_bh(&tbl
->lock
);
381 EXPORT_SYMBOL(neigh_lookup
);
383 struct neighbour
*neigh_lookup_nodev(struct neigh_table
*tbl
, struct net
*net
,
387 int key_len
= tbl
->key_len
;
390 NEIGH_CACHE_STAT_INC(tbl
, lookups
);
392 read_lock_bh(&tbl
->lock
);
393 hash_val
= tbl
->hash(pkey
, NULL
);
394 for (n
= tbl
->hash_buckets
[hash_val
& tbl
->hash_mask
]; n
; n
= n
->next
) {
395 if (!memcmp(n
->primary_key
, pkey
, key_len
) &&
396 net_eq(dev_net(n
->dev
), net
)) {
398 NEIGH_CACHE_STAT_INC(tbl
, hits
);
402 read_unlock_bh(&tbl
->lock
);
405 EXPORT_SYMBOL(neigh_lookup_nodev
);
407 struct neighbour
*neigh_create(struct neigh_table
*tbl
, const void *pkey
,
408 struct net_device
*dev
)
411 int key_len
= tbl
->key_len
;
413 struct neighbour
*n1
, *rc
, *n
= neigh_alloc(tbl
);
416 rc
= ERR_PTR(-ENOBUFS
);
420 memcpy(n
->primary_key
, pkey
, key_len
);
424 /* Protocol specific setup. */
425 if (tbl
->constructor
&& (error
= tbl
->constructor(n
)) < 0) {
427 goto out_neigh_release
;
430 /* Device specific setup. */
431 if (n
->parms
->neigh_setup
&&
432 (error
= n
->parms
->neigh_setup(n
)) < 0) {
434 goto out_neigh_release
;
437 n
->confirmed
= jiffies
- (n
->parms
->base_reachable_time
<< 1);
439 write_lock_bh(&tbl
->lock
);
441 if (atomic_read(&tbl
->entries
) > (tbl
->hash_mask
+ 1))
442 neigh_hash_grow(tbl
, (tbl
->hash_mask
+ 1) << 1);
444 hash_val
= tbl
->hash(pkey
, dev
) & tbl
->hash_mask
;
446 if (n
->parms
->dead
) {
447 rc
= ERR_PTR(-EINVAL
);
451 for (n1
= tbl
->hash_buckets
[hash_val
]; n1
; n1
= n1
->next
) {
452 if (dev
== n1
->dev
&& !memcmp(n1
->primary_key
, pkey
, key_len
)) {
459 n
->next
= tbl
->hash_buckets
[hash_val
];
460 tbl
->hash_buckets
[hash_val
] = n
;
463 write_unlock_bh(&tbl
->lock
);
464 NEIGH_PRINTK2("neigh %p is created.\n", n
);
469 write_unlock_bh(&tbl
->lock
);
474 EXPORT_SYMBOL(neigh_create
);
476 static u32
pneigh_hash(const void *pkey
, int key_len
)
478 u32 hash_val
= *(u32
*)(pkey
+ key_len
- 4);
479 hash_val
^= (hash_val
>> 16);
480 hash_val
^= hash_val
>> 8;
481 hash_val
^= hash_val
>> 4;
482 hash_val
&= PNEIGH_HASHMASK
;
486 static struct pneigh_entry
*__pneigh_lookup_1(struct pneigh_entry
*n
,
490 struct net_device
*dev
)
493 if (!memcmp(n
->key
, pkey
, key_len
) &&
494 net_eq(pneigh_net(n
), net
) &&
495 (n
->dev
== dev
|| !n
->dev
))
502 struct pneigh_entry
*__pneigh_lookup(struct neigh_table
*tbl
,
503 struct net
*net
, const void *pkey
, struct net_device
*dev
)
505 int key_len
= tbl
->key_len
;
506 u32 hash_val
= pneigh_hash(pkey
, key_len
);
508 return __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
509 net
, pkey
, key_len
, dev
);
511 EXPORT_SYMBOL_GPL(__pneigh_lookup
);
513 struct pneigh_entry
* pneigh_lookup(struct neigh_table
*tbl
,
514 struct net
*net
, const void *pkey
,
515 struct net_device
*dev
, int creat
)
517 struct pneigh_entry
*n
;
518 int key_len
= tbl
->key_len
;
519 u32 hash_val
= pneigh_hash(pkey
, key_len
);
521 read_lock_bh(&tbl
->lock
);
522 n
= __pneigh_lookup_1(tbl
->phash_buckets
[hash_val
],
523 net
, pkey
, key_len
, dev
);
524 read_unlock_bh(&tbl
->lock
);
531 n
= kmalloc(sizeof(*n
) + key_len
, GFP_KERNEL
);
535 write_pnet(&n
->net
, hold_net(net
));
536 memcpy(n
->key
, pkey
, key_len
);
541 if (tbl
->pconstructor
&& tbl
->pconstructor(n
)) {
550 write_lock_bh(&tbl
->lock
);
551 n
->next
= tbl
->phash_buckets
[hash_val
];
552 tbl
->phash_buckets
[hash_val
] = n
;
553 write_unlock_bh(&tbl
->lock
);
557 EXPORT_SYMBOL(pneigh_lookup
);
560 int pneigh_delete(struct neigh_table
*tbl
, struct net
*net
, const void *pkey
,
561 struct net_device
*dev
)
563 struct pneigh_entry
*n
, **np
;
564 int key_len
= tbl
->key_len
;
565 u32 hash_val
= pneigh_hash(pkey
, key_len
);
567 write_lock_bh(&tbl
->lock
);
568 for (np
= &tbl
->phash_buckets
[hash_val
]; (n
= *np
) != NULL
;
570 if (!memcmp(n
->key
, pkey
, key_len
) && n
->dev
== dev
&&
571 net_eq(pneigh_net(n
), net
)) {
573 write_unlock_bh(&tbl
->lock
);
574 if (tbl
->pdestructor
)
578 release_net(pneigh_net(n
));
583 write_unlock_bh(&tbl
->lock
);
587 static int pneigh_ifdown(struct neigh_table
*tbl
, struct net_device
*dev
)
589 struct pneigh_entry
*n
, **np
;
592 for (h
= 0; h
<= PNEIGH_HASHMASK
; h
++) {
593 np
= &tbl
->phash_buckets
[h
];
594 while ((n
= *np
) != NULL
) {
595 if (!dev
|| n
->dev
== dev
) {
597 if (tbl
->pdestructor
)
601 release_net(pneigh_net(n
));
611 static void neigh_parms_destroy(struct neigh_parms
*parms
);
613 static inline void neigh_parms_put(struct neigh_parms
*parms
)
615 if (atomic_dec_and_test(&parms
->refcnt
))
616 neigh_parms_destroy(parms
);
620 * neighbour must already be out of the table;
623 void neigh_destroy(struct neighbour
*neigh
)
627 NEIGH_CACHE_STAT_INC(neigh
->tbl
, destroys
);
631 "Destroying alive neighbour %p\n", neigh
);
636 if (neigh_del_timer(neigh
))
637 printk(KERN_WARNING
"Impossible event.\n");
639 while ((hh
= neigh
->hh
) != NULL
) {
640 neigh
->hh
= hh
->hh_next
;
643 write_seqlock_bh(&hh
->hh_lock
);
644 hh
->hh_output
= neigh_blackhole
;
645 write_sequnlock_bh(&hh
->hh_lock
);
646 if (atomic_dec_and_test(&hh
->hh_refcnt
))
650 skb_queue_purge(&neigh
->arp_queue
);
653 neigh_parms_put(neigh
->parms
);
655 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh
);
657 atomic_dec(&neigh
->tbl
->entries
);
658 kmem_cache_free(neigh
->tbl
->kmem_cachep
, neigh
);
660 EXPORT_SYMBOL(neigh_destroy
);
662 /* Neighbour state is suspicious;
665 Called with write_locked neigh.
667 static void neigh_suspect(struct neighbour
*neigh
)
671 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh
);
673 neigh
->output
= neigh
->ops
->output
;
675 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
)
676 hh
->hh_output
= neigh
->ops
->output
;
679 /* Neighbour state is OK;
682 Called with write_locked neigh.
684 static void neigh_connect(struct neighbour
*neigh
)
688 NEIGH_PRINTK2("neigh %p is connected.\n", neigh
);
690 neigh
->output
= neigh
->ops
->connected_output
;
692 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
)
693 hh
->hh_output
= neigh
->ops
->hh_output
;
696 static void neigh_periodic_work(struct work_struct
*work
)
698 struct neigh_table
*tbl
= container_of(work
, struct neigh_table
, gc_work
.work
);
699 struct neighbour
*n
, **np
;
702 NEIGH_CACHE_STAT_INC(tbl
, periodic_gc_runs
);
704 write_lock_bh(&tbl
->lock
);
707 * periodically recompute ReachableTime from random function
710 if (time_after(jiffies
, tbl
->last_rand
+ 300 * HZ
)) {
711 struct neigh_parms
*p
;
712 tbl
->last_rand
= jiffies
;
713 for (p
= &tbl
->parms
; p
; p
= p
->next
)
715 neigh_rand_reach_time(p
->base_reachable_time
);
718 for (i
= 0 ; i
<= tbl
->hash_mask
; i
++) {
719 np
= &tbl
->hash_buckets
[i
];
721 while ((n
= *np
) != NULL
) {
724 write_lock(&n
->lock
);
726 state
= n
->nud_state
;
727 if (state
& (NUD_PERMANENT
| NUD_IN_TIMER
)) {
728 write_unlock(&n
->lock
);
732 if (time_before(n
->used
, n
->confirmed
))
733 n
->used
= n
->confirmed
;
735 if (atomic_read(&n
->refcnt
) == 1 &&
736 (state
== NUD_FAILED
||
737 time_after(jiffies
, n
->used
+ n
->parms
->gc_staletime
))) {
740 write_unlock(&n
->lock
);
741 neigh_cleanup_and_release(n
);
744 write_unlock(&n
->lock
);
750 * It's fine to release lock here, even if hash table
751 * grows while we are preempted.
753 write_unlock_bh(&tbl
->lock
);
755 write_lock_bh(&tbl
->lock
);
757 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
758 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
759 * base_reachable_time.
761 schedule_delayed_work(&tbl
->gc_work
,
762 tbl
->parms
.base_reachable_time
>> 1);
763 write_unlock_bh(&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 static void neigh_invalidate(struct neighbour
*neigh
)
775 __releases(neigh
->lock
)
776 __acquires(neigh
->lock
)
780 NEIGH_CACHE_STAT_INC(neigh
->tbl
, res_failed
);
781 NEIGH_PRINTK2("neigh %p is failed.\n", neigh
);
782 neigh
->updated
= jiffies
;
784 /* It is very thin place. report_unreachable is very complicated
785 routine. Particularly, it can hit the same neighbour entry!
787 So that, we try to be accurate and avoid dead loop. --ANK
789 while (neigh
->nud_state
== NUD_FAILED
&&
790 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
791 write_unlock(&neigh
->lock
);
792 neigh
->ops
->error_report(neigh
, skb
);
793 write_lock(&neigh
->lock
);
795 skb_queue_purge(&neigh
->arp_queue
);
798 /* Called when a timer expires for a neighbour entry. */
800 static void neigh_timer_handler(unsigned long arg
)
802 unsigned long now
, next
;
803 struct neighbour
*neigh
= (struct neighbour
*)arg
;
807 write_lock(&neigh
->lock
);
809 state
= neigh
->nud_state
;
813 if (!(state
& NUD_IN_TIMER
)) {
815 printk(KERN_WARNING
"neigh: timer & !nud_in_timer\n");
820 if (state
& NUD_REACHABLE
) {
821 if (time_before_eq(now
,
822 neigh
->confirmed
+ neigh
->parms
->reachable_time
)) {
823 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh
);
824 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
825 } else if (time_before_eq(now
,
826 neigh
->used
+ neigh
->parms
->delay_probe_time
)) {
827 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh
);
828 neigh
->nud_state
= NUD_DELAY
;
829 neigh
->updated
= jiffies
;
830 neigh_suspect(neigh
);
831 next
= now
+ neigh
->parms
->delay_probe_time
;
833 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh
);
834 neigh
->nud_state
= NUD_STALE
;
835 neigh
->updated
= jiffies
;
836 neigh_suspect(neigh
);
839 } else if (state
& NUD_DELAY
) {
840 if (time_before_eq(now
,
841 neigh
->confirmed
+ neigh
->parms
->delay_probe_time
)) {
842 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh
);
843 neigh
->nud_state
= NUD_REACHABLE
;
844 neigh
->updated
= jiffies
;
845 neigh_connect(neigh
);
847 next
= neigh
->confirmed
+ neigh
->parms
->reachable_time
;
849 NEIGH_PRINTK2("neigh %p is probed.\n", neigh
);
850 neigh
->nud_state
= NUD_PROBE
;
851 neigh
->updated
= jiffies
;
852 atomic_set(&neigh
->probes
, 0);
853 next
= now
+ neigh
->parms
->retrans_time
;
856 /* NUD_PROBE|NUD_INCOMPLETE */
857 next
= now
+ neigh
->parms
->retrans_time
;
860 if ((neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) &&
861 atomic_read(&neigh
->probes
) >= neigh_max_probes(neigh
)) {
862 neigh
->nud_state
= NUD_FAILED
;
864 neigh_invalidate(neigh
);
867 if (neigh
->nud_state
& NUD_IN_TIMER
) {
868 if (time_before(next
, jiffies
+ HZ
/2))
869 next
= jiffies
+ HZ
/2;
870 if (!mod_timer(&neigh
->timer
, next
))
873 if (neigh
->nud_state
& (NUD_INCOMPLETE
| NUD_PROBE
)) {
874 struct sk_buff
*skb
= skb_peek(&neigh
->arp_queue
);
875 /* keep skb alive even if arp_queue overflows */
877 skb
= skb_copy(skb
, GFP_ATOMIC
);
878 write_unlock(&neigh
->lock
);
879 neigh
->ops
->solicit(neigh
, skb
);
880 atomic_inc(&neigh
->probes
);
884 write_unlock(&neigh
->lock
);
888 neigh_update_notify(neigh
);
890 neigh_release(neigh
);
893 int __neigh_event_send(struct neighbour
*neigh
, struct sk_buff
*skb
)
898 write_lock_bh(&neigh
->lock
);
901 if (neigh
->nud_state
& (NUD_CONNECTED
| NUD_DELAY
| NUD_PROBE
))
906 if (!(neigh
->nud_state
& (NUD_STALE
| NUD_INCOMPLETE
))) {
907 if (neigh
->parms
->mcast_probes
+ neigh
->parms
->app_probes
) {
908 atomic_set(&neigh
->probes
, neigh
->parms
->ucast_probes
);
909 neigh
->nud_state
= NUD_INCOMPLETE
;
910 neigh
->updated
= jiffies
;
911 neigh_add_timer(neigh
, now
+ 1);
913 neigh
->nud_state
= NUD_FAILED
;
914 neigh
->updated
= jiffies
;
915 write_unlock_bh(&neigh
->lock
);
920 } else if (neigh
->nud_state
& NUD_STALE
) {
921 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh
);
922 neigh
->nud_state
= NUD_DELAY
;
923 neigh
->updated
= jiffies
;
924 neigh_add_timer(neigh
,
925 jiffies
+ neigh
->parms
->delay_probe_time
);
928 if (neigh
->nud_state
== NUD_INCOMPLETE
) {
930 if (skb_queue_len(&neigh
->arp_queue
) >=
931 neigh
->parms
->queue_len
) {
932 struct sk_buff
*buff
;
933 buff
= __skb_dequeue(&neigh
->arp_queue
);
935 NEIGH_CACHE_STAT_INC(neigh
->tbl
, unres_discards
);
938 __skb_queue_tail(&neigh
->arp_queue
, skb
);
943 write_unlock_bh(&neigh
->lock
);
946 EXPORT_SYMBOL(__neigh_event_send
);
948 static void neigh_update_hhs(struct neighbour
*neigh
)
951 void (*update
)(struct hh_cache
*, const struct net_device
*, const unsigned char *)
954 if (neigh
->dev
->header_ops
)
955 update
= neigh
->dev
->header_ops
->cache_update
;
958 for (hh
= neigh
->hh
; hh
; hh
= hh
->hh_next
) {
959 write_seqlock_bh(&hh
->hh_lock
);
960 update(hh
, neigh
->dev
, neigh
->ha
);
961 write_sequnlock_bh(&hh
->hh_lock
);
968 /* Generic update routine.
969 -- lladdr is new lladdr or NULL, if it is not supplied.
972 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
974 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
975 lladdr instead of overriding it
977 It also allows to retain current state
978 if lladdr is unchanged.
979 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
981 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
983 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
986 Caller MUST hold reference count on the entry.
989 int neigh_update(struct neighbour
*neigh
, const u8
*lladdr
, u8
new,
995 struct net_device
*dev
;
996 int update_isrouter
= 0;
998 write_lock_bh(&neigh
->lock
);
1001 old
= neigh
->nud_state
;
1004 if (!(flags
& NEIGH_UPDATE_F_ADMIN
) &&
1005 (old
& (NUD_NOARP
| NUD_PERMANENT
)))
1008 if (!(new & NUD_VALID
)) {
1009 neigh_del_timer(neigh
);
1010 if (old
& NUD_CONNECTED
)
1011 neigh_suspect(neigh
);
1012 neigh
->nud_state
= new;
1014 notify
= old
& NUD_VALID
;
1015 if ((old
& (NUD_INCOMPLETE
| NUD_PROBE
)) &&
1016 (new & NUD_FAILED
)) {
1017 neigh_invalidate(neigh
);
1023 /* Compare new lladdr with cached one */
1024 if (!dev
->addr_len
) {
1025 /* First case: device needs no address. */
1027 } else if (lladdr
) {
1028 /* The second case: if something is already cached
1029 and a new address is proposed:
1031 - if they are different, check override flag
1033 if ((old
& NUD_VALID
) &&
1034 !memcmp(lladdr
, neigh
->ha
, dev
->addr_len
))
1037 /* No address is supplied; if we know something,
1038 use it, otherwise discard the request.
1041 if (!(old
& NUD_VALID
))
1046 if (new & NUD_CONNECTED
)
1047 neigh
->confirmed
= jiffies
;
1048 neigh
->updated
= jiffies
;
1050 /* If entry was valid and address is not changed,
1051 do not change entry state, if new one is STALE.
1054 update_isrouter
= flags
& NEIGH_UPDATE_F_OVERRIDE_ISROUTER
;
1055 if (old
& NUD_VALID
) {
1056 if (lladdr
!= neigh
->ha
&& !(flags
& NEIGH_UPDATE_F_OVERRIDE
)) {
1057 update_isrouter
= 0;
1058 if ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) &&
1059 (old
& NUD_CONNECTED
)) {
1065 if (lladdr
== neigh
->ha
&& new == NUD_STALE
&&
1066 ((flags
& NEIGH_UPDATE_F_WEAK_OVERRIDE
) ||
1067 (old
& NUD_CONNECTED
))
1074 neigh_del_timer(neigh
);
1075 if (new & NUD_IN_TIMER
)
1076 neigh_add_timer(neigh
, (jiffies
+
1077 ((new & NUD_REACHABLE
) ?
1078 neigh
->parms
->reachable_time
:
1080 neigh
->nud_state
= new;
1083 if (lladdr
!= neigh
->ha
) {
1084 memcpy(&neigh
->ha
, lladdr
, dev
->addr_len
);
1085 neigh_update_hhs(neigh
);
1086 if (!(new & NUD_CONNECTED
))
1087 neigh
->confirmed
= jiffies
-
1088 (neigh
->parms
->base_reachable_time
<< 1);
1093 if (new & NUD_CONNECTED
)
1094 neigh_connect(neigh
);
1096 neigh_suspect(neigh
);
1097 if (!(old
& NUD_VALID
)) {
1098 struct sk_buff
*skb
;
1100 /* Again: avoid dead loop if something went wrong */
1102 while (neigh
->nud_state
& NUD_VALID
&&
1103 (skb
= __skb_dequeue(&neigh
->arp_queue
)) != NULL
) {
1104 struct neighbour
*n1
= neigh
;
1105 write_unlock_bh(&neigh
->lock
);
1106 /* On shaper/eql skb->dst->neighbour != neigh :( */
1107 if (skb_dst(skb
) && skb_dst(skb
)->neighbour
)
1108 n1
= skb_dst(skb
)->neighbour
;
1110 write_lock_bh(&neigh
->lock
);
1112 skb_queue_purge(&neigh
->arp_queue
);
1115 if (update_isrouter
) {
1116 neigh
->flags
= (flags
& NEIGH_UPDATE_F_ISROUTER
) ?
1117 (neigh
->flags
| NTF_ROUTER
) :
1118 (neigh
->flags
& ~NTF_ROUTER
);
1120 write_unlock_bh(&neigh
->lock
);
1123 neigh_update_notify(neigh
);
1127 EXPORT_SYMBOL(neigh_update
);
1129 struct neighbour
*neigh_event_ns(struct neigh_table
*tbl
,
1130 u8
*lladdr
, void *saddr
,
1131 struct net_device
*dev
)
1133 struct neighbour
*neigh
= __neigh_lookup(tbl
, saddr
, dev
,
1134 lladdr
|| !dev
->addr_len
);
1136 neigh_update(neigh
, lladdr
, NUD_STALE
,
1137 NEIGH_UPDATE_F_OVERRIDE
);
1140 EXPORT_SYMBOL(neigh_event_ns
);
1142 static void neigh_hh_init(struct neighbour
*n
, struct dst_entry
*dst
,
1145 struct hh_cache
*hh
;
1146 struct net_device
*dev
= dst
->dev
;
1148 for (hh
= n
->hh
; hh
; hh
= hh
->hh_next
)
1149 if (hh
->hh_type
== protocol
)
1152 if (!hh
&& (hh
= kzalloc(sizeof(*hh
), GFP_ATOMIC
)) != NULL
) {
1153 seqlock_init(&hh
->hh_lock
);
1154 hh
->hh_type
= protocol
;
1155 atomic_set(&hh
->hh_refcnt
, 0);
1158 if (dev
->header_ops
->cache(n
, hh
)) {
1162 atomic_inc(&hh
->hh_refcnt
);
1163 hh
->hh_next
= n
->hh
;
1165 if (n
->nud_state
& NUD_CONNECTED
)
1166 hh
->hh_output
= n
->ops
->hh_output
;
1168 hh
->hh_output
= n
->ops
->output
;
1172 atomic_inc(&hh
->hh_refcnt
);
1177 /* This function can be used in contexts, where only old dev_queue_xmit
1178 worked, f.e. if you want to override normal output path (eql, shaper),
1179 but resolution is not made yet.
1182 int neigh_compat_output(struct sk_buff
*skb
)
1184 struct net_device
*dev
= skb
->dev
;
1186 __skb_pull(skb
, skb_network_offset(skb
));
1188 if (dev_hard_header(skb
, dev
, ntohs(skb
->protocol
), NULL
, NULL
,
1190 dev
->header_ops
->rebuild(skb
))
1193 return dev_queue_xmit(skb
);
1195 EXPORT_SYMBOL(neigh_compat_output
);
1197 /* Slow and careful. */
1199 int neigh_resolve_output(struct sk_buff
*skb
)
1201 struct dst_entry
*dst
= skb_dst(skb
);
1202 struct neighbour
*neigh
;
1205 if (!dst
|| !(neigh
= dst
->neighbour
))
1208 __skb_pull(skb
, skb_network_offset(skb
));
1210 if (!neigh_event_send(neigh
, skb
)) {
1212 struct net_device
*dev
= neigh
->dev
;
1213 if (dev
->header_ops
->cache
&& !dst
->hh
) {
1214 write_lock_bh(&neigh
->lock
);
1216 neigh_hh_init(neigh
, dst
, dst
->ops
->protocol
);
1217 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1218 neigh
->ha
, NULL
, skb
->len
);
1219 write_unlock_bh(&neigh
->lock
);
1221 read_lock_bh(&neigh
->lock
);
1222 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1223 neigh
->ha
, NULL
, skb
->len
);
1224 read_unlock_bh(&neigh
->lock
);
1227 rc
= neigh
->ops
->queue_xmit(skb
);
1234 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1235 dst
, dst
? dst
->neighbour
: NULL
);
1241 EXPORT_SYMBOL(neigh_resolve_output
);
1243 /* As fast as possible without hh cache */
1245 int neigh_connected_output(struct sk_buff
*skb
)
1248 struct dst_entry
*dst
= skb_dst(skb
);
1249 struct neighbour
*neigh
= dst
->neighbour
;
1250 struct net_device
*dev
= neigh
->dev
;
1252 __skb_pull(skb
, skb_network_offset(skb
));
1254 read_lock_bh(&neigh
->lock
);
1255 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1256 neigh
->ha
, NULL
, skb
->len
);
1257 read_unlock_bh(&neigh
->lock
);
1259 err
= neigh
->ops
->queue_xmit(skb
);
1266 EXPORT_SYMBOL(neigh_connected_output
);
1268 static void neigh_proxy_process(unsigned long arg
)
1270 struct neigh_table
*tbl
= (struct neigh_table
*)arg
;
1271 long sched_next
= 0;
1272 unsigned long now
= jiffies
;
1273 struct sk_buff
*skb
, *n
;
1275 spin_lock(&tbl
->proxy_queue
.lock
);
1277 skb_queue_walk_safe(&tbl
->proxy_queue
, skb
, n
) {
1278 long tdif
= NEIGH_CB(skb
)->sched_next
- now
;
1281 struct net_device
*dev
= skb
->dev
;
1282 __skb_unlink(skb
, &tbl
->proxy_queue
);
1283 if (tbl
->proxy_redo
&& netif_running(dev
))
1284 tbl
->proxy_redo(skb
);
1289 } else if (!sched_next
|| tdif
< sched_next
)
1292 del_timer(&tbl
->proxy_timer
);
1294 mod_timer(&tbl
->proxy_timer
, jiffies
+ sched_next
);
1295 spin_unlock(&tbl
->proxy_queue
.lock
);
1298 void pneigh_enqueue(struct neigh_table
*tbl
, struct neigh_parms
*p
,
1299 struct sk_buff
*skb
)
1301 unsigned long now
= jiffies
;
1302 unsigned long sched_next
= now
+ (net_random() % p
->proxy_delay
);
1304 if (tbl
->proxy_queue
.qlen
> p
->proxy_qlen
) {
1309 NEIGH_CB(skb
)->sched_next
= sched_next
;
1310 NEIGH_CB(skb
)->flags
|= LOCALLY_ENQUEUED
;
1312 spin_lock(&tbl
->proxy_queue
.lock
);
1313 if (del_timer(&tbl
->proxy_timer
)) {
1314 if (time_before(tbl
->proxy_timer
.expires
, sched_next
))
1315 sched_next
= tbl
->proxy_timer
.expires
;
1319 __skb_queue_tail(&tbl
->proxy_queue
, skb
);
1320 mod_timer(&tbl
->proxy_timer
, sched_next
);
1321 spin_unlock(&tbl
->proxy_queue
.lock
);
1323 EXPORT_SYMBOL(pneigh_enqueue
);
1325 static inline struct neigh_parms
*lookup_neigh_parms(struct neigh_table
*tbl
,
1326 struct net
*net
, int ifindex
)
1328 struct neigh_parms
*p
;
1330 for (p
= &tbl
->parms
; p
; p
= p
->next
) {
1331 if ((p
->dev
&& p
->dev
->ifindex
== ifindex
&& net_eq(neigh_parms_net(p
), net
)) ||
1332 (!p
->dev
&& !ifindex
))
1339 struct neigh_parms
*neigh_parms_alloc(struct net_device
*dev
,
1340 struct neigh_table
*tbl
)
1342 struct neigh_parms
*p
, *ref
;
1343 struct net
*net
= dev_net(dev
);
1344 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1346 ref
= lookup_neigh_parms(tbl
, net
, 0);
1350 p
= kmemdup(ref
, sizeof(*p
), GFP_KERNEL
);
1353 atomic_set(&p
->refcnt
, 1);
1355 neigh_rand_reach_time(p
->base_reachable_time
);
1357 if (ops
->ndo_neigh_setup
&& ops
->ndo_neigh_setup(dev
, p
)) {
1364 write_pnet(&p
->net
, hold_net(net
));
1365 p
->sysctl_table
= NULL
;
1366 write_lock_bh(&tbl
->lock
);
1367 p
->next
= tbl
->parms
.next
;
1368 tbl
->parms
.next
= p
;
1369 write_unlock_bh(&tbl
->lock
);
1373 EXPORT_SYMBOL(neigh_parms_alloc
);
1375 static void neigh_rcu_free_parms(struct rcu_head
*head
)
1377 struct neigh_parms
*parms
=
1378 container_of(head
, struct neigh_parms
, rcu_head
);
1380 neigh_parms_put(parms
);
1383 void neigh_parms_release(struct neigh_table
*tbl
, struct neigh_parms
*parms
)
1385 struct neigh_parms
**p
;
1387 if (!parms
|| parms
== &tbl
->parms
)
1389 write_lock_bh(&tbl
->lock
);
1390 for (p
= &tbl
->parms
.next
; *p
; p
= &(*p
)->next
) {
1394 write_unlock_bh(&tbl
->lock
);
1396 dev_put(parms
->dev
);
1397 call_rcu(&parms
->rcu_head
, neigh_rcu_free_parms
);
1401 write_unlock_bh(&tbl
->lock
);
1402 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1404 EXPORT_SYMBOL(neigh_parms_release
);
1406 static void neigh_parms_destroy(struct neigh_parms
*parms
)
1408 release_net(neigh_parms_net(parms
));
1412 static struct lock_class_key neigh_table_proxy_queue_class
;
1414 void neigh_table_init_no_netlink(struct neigh_table
*tbl
)
1416 unsigned long now
= jiffies
;
1417 unsigned long phsize
;
1419 write_pnet(&tbl
->parms
.net
, &init_net
);
1420 atomic_set(&tbl
->parms
.refcnt
, 1);
1421 tbl
->parms
.reachable_time
=
1422 neigh_rand_reach_time(tbl
->parms
.base_reachable_time
);
1424 if (!tbl
->kmem_cachep
)
1426 kmem_cache_create(tbl
->id
, tbl
->entry_size
, 0,
1427 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
1429 tbl
->stats
= alloc_percpu(struct neigh_statistics
);
1431 panic("cannot create neighbour cache statistics");
1433 #ifdef CONFIG_PROC_FS
1434 if (!proc_create_data(tbl
->id
, 0, init_net
.proc_net_stat
,
1435 &neigh_stat_seq_fops
, tbl
))
1436 panic("cannot create neighbour proc dir entry");
1440 tbl
->hash_buckets
= neigh_hash_alloc(tbl
->hash_mask
+ 1);
1442 phsize
= (PNEIGH_HASHMASK
+ 1) * sizeof(struct pneigh_entry
*);
1443 tbl
->phash_buckets
= kzalloc(phsize
, GFP_KERNEL
);
1445 if (!tbl
->hash_buckets
|| !tbl
->phash_buckets
)
1446 panic("cannot allocate neighbour cache hashes");
1448 get_random_bytes(&tbl
->hash_rnd
, sizeof(tbl
->hash_rnd
));
1450 rwlock_init(&tbl
->lock
);
1451 INIT_DELAYED_WORK_DEFERRABLE(&tbl
->gc_work
, neigh_periodic_work
);
1452 schedule_delayed_work(&tbl
->gc_work
, tbl
->parms
.reachable_time
);
1453 setup_timer(&tbl
->proxy_timer
, neigh_proxy_process
, (unsigned long)tbl
);
1454 skb_queue_head_init_class(&tbl
->proxy_queue
,
1455 &neigh_table_proxy_queue_class
);
1457 tbl
->last_flush
= now
;
1458 tbl
->last_rand
= now
+ tbl
->parms
.reachable_time
* 20;
1460 EXPORT_SYMBOL(neigh_table_init_no_netlink
);
1462 void neigh_table_init(struct neigh_table
*tbl
)
1464 struct neigh_table
*tmp
;
1466 neigh_table_init_no_netlink(tbl
);
1467 write_lock(&neigh_tbl_lock
);
1468 for (tmp
= neigh_tables
; tmp
; tmp
= tmp
->next
) {
1469 if (tmp
->family
== tbl
->family
)
1472 tbl
->next
= neigh_tables
;
1474 write_unlock(&neigh_tbl_lock
);
1476 if (unlikely(tmp
)) {
1477 printk(KERN_ERR
"NEIGH: Registering multiple tables for "
1478 "family %d\n", tbl
->family
);
1482 EXPORT_SYMBOL(neigh_table_init
);
1484 int neigh_table_clear(struct neigh_table
*tbl
)
1486 struct neigh_table
**tp
;
1488 /* It is not clean... Fix it to unload IPv6 module safely */
1489 cancel_delayed_work(&tbl
->gc_work
);
1490 flush_scheduled_work();
1491 del_timer_sync(&tbl
->proxy_timer
);
1492 pneigh_queue_purge(&tbl
->proxy_queue
);
1493 neigh_ifdown(tbl
, NULL
);
1494 if (atomic_read(&tbl
->entries
))
1495 printk(KERN_CRIT
"neighbour leakage\n");
1496 write_lock(&neigh_tbl_lock
);
1497 for (tp
= &neigh_tables
; *tp
; tp
= &(*tp
)->next
) {
1503 write_unlock(&neigh_tbl_lock
);
1505 neigh_hash_free(tbl
->hash_buckets
, tbl
->hash_mask
+ 1);
1506 tbl
->hash_buckets
= NULL
;
1508 kfree(tbl
->phash_buckets
);
1509 tbl
->phash_buckets
= NULL
;
1511 remove_proc_entry(tbl
->id
, init_net
.proc_net_stat
);
1513 free_percpu(tbl
->stats
);
1516 kmem_cache_destroy(tbl
->kmem_cachep
);
1517 tbl
->kmem_cachep
= NULL
;
1521 EXPORT_SYMBOL(neigh_table_clear
);
1523 static int neigh_delete(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1525 struct net
*net
= sock_net(skb
->sk
);
1527 struct nlattr
*dst_attr
;
1528 struct neigh_table
*tbl
;
1529 struct net_device
*dev
= NULL
;
1532 if (nlmsg_len(nlh
) < sizeof(*ndm
))
1535 dst_attr
= nlmsg_find_attr(nlh
, sizeof(*ndm
), NDA_DST
);
1536 if (dst_attr
== NULL
)
1539 ndm
= nlmsg_data(nlh
);
1540 if (ndm
->ndm_ifindex
) {
1541 dev
= dev_get_by_index(net
, ndm
->ndm_ifindex
);
1548 read_lock(&neigh_tbl_lock
);
1549 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1550 struct neighbour
*neigh
;
1552 if (tbl
->family
!= ndm
->ndm_family
)
1554 read_unlock(&neigh_tbl_lock
);
1556 if (nla_len(dst_attr
) < tbl
->key_len
)
1559 if (ndm
->ndm_flags
& NTF_PROXY
) {
1560 err
= pneigh_delete(tbl
, net
, nla_data(dst_attr
), dev
);
1567 neigh
= neigh_lookup(tbl
, nla_data(dst_attr
), dev
);
1568 if (neigh
== NULL
) {
1573 err
= neigh_update(neigh
, NULL
, NUD_FAILED
,
1574 NEIGH_UPDATE_F_OVERRIDE
|
1575 NEIGH_UPDATE_F_ADMIN
);
1576 neigh_release(neigh
);
1579 read_unlock(&neigh_tbl_lock
);
1580 err
= -EAFNOSUPPORT
;
1589 static int neigh_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1591 struct net
*net
= sock_net(skb
->sk
);
1593 struct nlattr
*tb
[NDA_MAX
+1];
1594 struct neigh_table
*tbl
;
1595 struct net_device
*dev
= NULL
;
1598 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
1603 if (tb
[NDA_DST
] == NULL
)
1606 ndm
= nlmsg_data(nlh
);
1607 if (ndm
->ndm_ifindex
) {
1608 dev
= dev_get_by_index(net
, ndm
->ndm_ifindex
);
1614 if (tb
[NDA_LLADDR
] && nla_len(tb
[NDA_LLADDR
]) < dev
->addr_len
)
1618 read_lock(&neigh_tbl_lock
);
1619 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1620 int flags
= NEIGH_UPDATE_F_ADMIN
| NEIGH_UPDATE_F_OVERRIDE
;
1621 struct neighbour
*neigh
;
1624 if (tbl
->family
!= ndm
->ndm_family
)
1626 read_unlock(&neigh_tbl_lock
);
1628 if (nla_len(tb
[NDA_DST
]) < tbl
->key_len
)
1630 dst
= nla_data(tb
[NDA_DST
]);
1631 lladdr
= tb
[NDA_LLADDR
] ? nla_data(tb
[NDA_LLADDR
]) : NULL
;
1633 if (ndm
->ndm_flags
& NTF_PROXY
) {
1634 struct pneigh_entry
*pn
;
1637 pn
= pneigh_lookup(tbl
, net
, dst
, dev
, 1);
1639 pn
->flags
= ndm
->ndm_flags
;
1648 neigh
= neigh_lookup(tbl
, dst
, dev
);
1649 if (neigh
== NULL
) {
1650 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1655 neigh
= __neigh_lookup_errno(tbl
, dst
, dev
);
1656 if (IS_ERR(neigh
)) {
1657 err
= PTR_ERR(neigh
);
1661 if (nlh
->nlmsg_flags
& NLM_F_EXCL
) {
1663 neigh_release(neigh
);
1667 if (!(nlh
->nlmsg_flags
& NLM_F_REPLACE
))
1668 flags
&= ~NEIGH_UPDATE_F_OVERRIDE
;
1671 if (ndm
->ndm_flags
& NTF_USE
) {
1672 neigh_event_send(neigh
, NULL
);
1675 err
= neigh_update(neigh
, lladdr
, ndm
->ndm_state
, flags
);
1676 neigh_release(neigh
);
1680 read_unlock(&neigh_tbl_lock
);
1681 err
= -EAFNOSUPPORT
;
1690 static int neightbl_fill_parms(struct sk_buff
*skb
, struct neigh_parms
*parms
)
1692 struct nlattr
*nest
;
1694 nest
= nla_nest_start(skb
, NDTA_PARMS
);
1699 NLA_PUT_U32(skb
, NDTPA_IFINDEX
, parms
->dev
->ifindex
);
1701 NLA_PUT_U32(skb
, NDTPA_REFCNT
, atomic_read(&parms
->refcnt
));
1702 NLA_PUT_U32(skb
, NDTPA_QUEUE_LEN
, parms
->queue_len
);
1703 NLA_PUT_U32(skb
, NDTPA_PROXY_QLEN
, parms
->proxy_qlen
);
1704 NLA_PUT_U32(skb
, NDTPA_APP_PROBES
, parms
->app_probes
);
1705 NLA_PUT_U32(skb
, NDTPA_UCAST_PROBES
, parms
->ucast_probes
);
1706 NLA_PUT_U32(skb
, NDTPA_MCAST_PROBES
, parms
->mcast_probes
);
1707 NLA_PUT_MSECS(skb
, NDTPA_REACHABLE_TIME
, parms
->reachable_time
);
1708 NLA_PUT_MSECS(skb
, NDTPA_BASE_REACHABLE_TIME
,
1709 parms
->base_reachable_time
);
1710 NLA_PUT_MSECS(skb
, NDTPA_GC_STALETIME
, parms
->gc_staletime
);
1711 NLA_PUT_MSECS(skb
, NDTPA_DELAY_PROBE_TIME
, parms
->delay_probe_time
);
1712 NLA_PUT_MSECS(skb
, NDTPA_RETRANS_TIME
, parms
->retrans_time
);
1713 NLA_PUT_MSECS(skb
, NDTPA_ANYCAST_DELAY
, parms
->anycast_delay
);
1714 NLA_PUT_MSECS(skb
, NDTPA_PROXY_DELAY
, parms
->proxy_delay
);
1715 NLA_PUT_MSECS(skb
, NDTPA_LOCKTIME
, parms
->locktime
);
1717 return nla_nest_end(skb
, nest
);
1720 nla_nest_cancel(skb
, nest
);
1724 static int neightbl_fill_info(struct sk_buff
*skb
, struct neigh_table
*tbl
,
1725 u32 pid
, u32 seq
, int type
, int flags
)
1727 struct nlmsghdr
*nlh
;
1728 struct ndtmsg
*ndtmsg
;
1730 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1734 ndtmsg
= nlmsg_data(nlh
);
1736 read_lock_bh(&tbl
->lock
);
1737 ndtmsg
->ndtm_family
= tbl
->family
;
1738 ndtmsg
->ndtm_pad1
= 0;
1739 ndtmsg
->ndtm_pad2
= 0;
1741 NLA_PUT_STRING(skb
, NDTA_NAME
, tbl
->id
);
1742 NLA_PUT_MSECS(skb
, NDTA_GC_INTERVAL
, tbl
->gc_interval
);
1743 NLA_PUT_U32(skb
, NDTA_THRESH1
, tbl
->gc_thresh1
);
1744 NLA_PUT_U32(skb
, NDTA_THRESH2
, tbl
->gc_thresh2
);
1745 NLA_PUT_U32(skb
, NDTA_THRESH3
, tbl
->gc_thresh3
);
1748 unsigned long now
= jiffies
;
1749 unsigned int flush_delta
= now
- tbl
->last_flush
;
1750 unsigned int rand_delta
= now
- tbl
->last_rand
;
1752 struct ndt_config ndc
= {
1753 .ndtc_key_len
= tbl
->key_len
,
1754 .ndtc_entry_size
= tbl
->entry_size
,
1755 .ndtc_entries
= atomic_read(&tbl
->entries
),
1756 .ndtc_last_flush
= jiffies_to_msecs(flush_delta
),
1757 .ndtc_last_rand
= jiffies_to_msecs(rand_delta
),
1758 .ndtc_hash_rnd
= tbl
->hash_rnd
,
1759 .ndtc_hash_mask
= tbl
->hash_mask
,
1760 .ndtc_proxy_qlen
= tbl
->proxy_queue
.qlen
,
1763 NLA_PUT(skb
, NDTA_CONFIG
, sizeof(ndc
), &ndc
);
1768 struct ndt_stats ndst
;
1770 memset(&ndst
, 0, sizeof(ndst
));
1772 for_each_possible_cpu(cpu
) {
1773 struct neigh_statistics
*st
;
1775 st
= per_cpu_ptr(tbl
->stats
, cpu
);
1776 ndst
.ndts_allocs
+= st
->allocs
;
1777 ndst
.ndts_destroys
+= st
->destroys
;
1778 ndst
.ndts_hash_grows
+= st
->hash_grows
;
1779 ndst
.ndts_res_failed
+= st
->res_failed
;
1780 ndst
.ndts_lookups
+= st
->lookups
;
1781 ndst
.ndts_hits
+= st
->hits
;
1782 ndst
.ndts_rcv_probes_mcast
+= st
->rcv_probes_mcast
;
1783 ndst
.ndts_rcv_probes_ucast
+= st
->rcv_probes_ucast
;
1784 ndst
.ndts_periodic_gc_runs
+= st
->periodic_gc_runs
;
1785 ndst
.ndts_forced_gc_runs
+= st
->forced_gc_runs
;
1788 NLA_PUT(skb
, NDTA_STATS
, sizeof(ndst
), &ndst
);
1791 BUG_ON(tbl
->parms
.dev
);
1792 if (neightbl_fill_parms(skb
, &tbl
->parms
) < 0)
1793 goto nla_put_failure
;
1795 read_unlock_bh(&tbl
->lock
);
1796 return nlmsg_end(skb
, nlh
);
1799 read_unlock_bh(&tbl
->lock
);
1800 nlmsg_cancel(skb
, nlh
);
1804 static int neightbl_fill_param_info(struct sk_buff
*skb
,
1805 struct neigh_table
*tbl
,
1806 struct neigh_parms
*parms
,
1807 u32 pid
, u32 seq
, int type
,
1810 struct ndtmsg
*ndtmsg
;
1811 struct nlmsghdr
*nlh
;
1813 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1817 ndtmsg
= nlmsg_data(nlh
);
1819 read_lock_bh(&tbl
->lock
);
1820 ndtmsg
->ndtm_family
= tbl
->family
;
1821 ndtmsg
->ndtm_pad1
= 0;
1822 ndtmsg
->ndtm_pad2
= 0;
1824 if (nla_put_string(skb
, NDTA_NAME
, tbl
->id
) < 0 ||
1825 neightbl_fill_parms(skb
, parms
) < 0)
1828 read_unlock_bh(&tbl
->lock
);
1829 return nlmsg_end(skb
, nlh
);
1831 read_unlock_bh(&tbl
->lock
);
1832 nlmsg_cancel(skb
, nlh
);
1836 static const struct nla_policy nl_neightbl_policy
[NDTA_MAX
+1] = {
1837 [NDTA_NAME
] = { .type
= NLA_STRING
},
1838 [NDTA_THRESH1
] = { .type
= NLA_U32
},
1839 [NDTA_THRESH2
] = { .type
= NLA_U32
},
1840 [NDTA_THRESH3
] = { .type
= NLA_U32
},
1841 [NDTA_GC_INTERVAL
] = { .type
= NLA_U64
},
1842 [NDTA_PARMS
] = { .type
= NLA_NESTED
},
1845 static const struct nla_policy nl_ntbl_parm_policy
[NDTPA_MAX
+1] = {
1846 [NDTPA_IFINDEX
] = { .type
= NLA_U32
},
1847 [NDTPA_QUEUE_LEN
] = { .type
= NLA_U32
},
1848 [NDTPA_PROXY_QLEN
] = { .type
= NLA_U32
},
1849 [NDTPA_APP_PROBES
] = { .type
= NLA_U32
},
1850 [NDTPA_UCAST_PROBES
] = { .type
= NLA_U32
},
1851 [NDTPA_MCAST_PROBES
] = { .type
= NLA_U32
},
1852 [NDTPA_BASE_REACHABLE_TIME
] = { .type
= NLA_U64
},
1853 [NDTPA_GC_STALETIME
] = { .type
= NLA_U64
},
1854 [NDTPA_DELAY_PROBE_TIME
] = { .type
= NLA_U64
},
1855 [NDTPA_RETRANS_TIME
] = { .type
= NLA_U64
},
1856 [NDTPA_ANYCAST_DELAY
] = { .type
= NLA_U64
},
1857 [NDTPA_PROXY_DELAY
] = { .type
= NLA_U64
},
1858 [NDTPA_LOCKTIME
] = { .type
= NLA_U64
},
1861 static int neightbl_set(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1863 struct net
*net
= sock_net(skb
->sk
);
1864 struct neigh_table
*tbl
;
1865 struct ndtmsg
*ndtmsg
;
1866 struct nlattr
*tb
[NDTA_MAX
+1];
1869 err
= nlmsg_parse(nlh
, sizeof(*ndtmsg
), tb
, NDTA_MAX
,
1870 nl_neightbl_policy
);
1874 if (tb
[NDTA_NAME
] == NULL
) {
1879 ndtmsg
= nlmsg_data(nlh
);
1880 read_lock(&neigh_tbl_lock
);
1881 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1882 if (ndtmsg
->ndtm_family
&& tbl
->family
!= ndtmsg
->ndtm_family
)
1885 if (nla_strcmp(tb
[NDTA_NAME
], tbl
->id
) == 0)
1895 * We acquire tbl->lock to be nice to the periodic timers and
1896 * make sure they always see a consistent set of values.
1898 write_lock_bh(&tbl
->lock
);
1900 if (tb
[NDTA_PARMS
]) {
1901 struct nlattr
*tbp
[NDTPA_MAX
+1];
1902 struct neigh_parms
*p
;
1905 err
= nla_parse_nested(tbp
, NDTPA_MAX
, tb
[NDTA_PARMS
],
1906 nl_ntbl_parm_policy
);
1908 goto errout_tbl_lock
;
1910 if (tbp
[NDTPA_IFINDEX
])
1911 ifindex
= nla_get_u32(tbp
[NDTPA_IFINDEX
]);
1913 p
= lookup_neigh_parms(tbl
, net
, ifindex
);
1916 goto errout_tbl_lock
;
1919 for (i
= 1; i
<= NDTPA_MAX
; i
++) {
1924 case NDTPA_QUEUE_LEN
:
1925 p
->queue_len
= nla_get_u32(tbp
[i
]);
1927 case NDTPA_PROXY_QLEN
:
1928 p
->proxy_qlen
= nla_get_u32(tbp
[i
]);
1930 case NDTPA_APP_PROBES
:
1931 p
->app_probes
= nla_get_u32(tbp
[i
]);
1933 case NDTPA_UCAST_PROBES
:
1934 p
->ucast_probes
= nla_get_u32(tbp
[i
]);
1936 case NDTPA_MCAST_PROBES
:
1937 p
->mcast_probes
= nla_get_u32(tbp
[i
]);
1939 case NDTPA_BASE_REACHABLE_TIME
:
1940 p
->base_reachable_time
= nla_get_msecs(tbp
[i
]);
1942 case NDTPA_GC_STALETIME
:
1943 p
->gc_staletime
= nla_get_msecs(tbp
[i
]);
1945 case NDTPA_DELAY_PROBE_TIME
:
1946 p
->delay_probe_time
= nla_get_msecs(tbp
[i
]);
1948 case NDTPA_RETRANS_TIME
:
1949 p
->retrans_time
= nla_get_msecs(tbp
[i
]);
1951 case NDTPA_ANYCAST_DELAY
:
1952 p
->anycast_delay
= nla_get_msecs(tbp
[i
]);
1954 case NDTPA_PROXY_DELAY
:
1955 p
->proxy_delay
= nla_get_msecs(tbp
[i
]);
1957 case NDTPA_LOCKTIME
:
1958 p
->locktime
= nla_get_msecs(tbp
[i
]);
1964 if (tb
[NDTA_THRESH1
])
1965 tbl
->gc_thresh1
= nla_get_u32(tb
[NDTA_THRESH1
]);
1967 if (tb
[NDTA_THRESH2
])
1968 tbl
->gc_thresh2
= nla_get_u32(tb
[NDTA_THRESH2
]);
1970 if (tb
[NDTA_THRESH3
])
1971 tbl
->gc_thresh3
= nla_get_u32(tb
[NDTA_THRESH3
]);
1973 if (tb
[NDTA_GC_INTERVAL
])
1974 tbl
->gc_interval
= nla_get_msecs(tb
[NDTA_GC_INTERVAL
]);
1979 write_unlock_bh(&tbl
->lock
);
1981 read_unlock(&neigh_tbl_lock
);
1986 static int neightbl_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1988 struct net
*net
= sock_net(skb
->sk
);
1989 int family
, tidx
, nidx
= 0;
1990 int tbl_skip
= cb
->args
[0];
1991 int neigh_skip
= cb
->args
[1];
1992 struct neigh_table
*tbl
;
1994 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
1996 read_lock(&neigh_tbl_lock
);
1997 for (tbl
= neigh_tables
, tidx
= 0; tbl
; tbl
= tbl
->next
, tidx
++) {
1998 struct neigh_parms
*p
;
2000 if (tidx
< tbl_skip
|| (family
&& tbl
->family
!= family
))
2003 if (neightbl_fill_info(skb
, tbl
, NETLINK_CB(cb
->skb
).pid
,
2004 cb
->nlh
->nlmsg_seq
, RTM_NEWNEIGHTBL
,
2008 for (nidx
= 0, p
= tbl
->parms
.next
; p
; p
= p
->next
) {
2009 if (!net_eq(neigh_parms_net(p
), net
))
2012 if (nidx
< neigh_skip
)
2015 if (neightbl_fill_param_info(skb
, tbl
, p
,
2016 NETLINK_CB(cb
->skb
).pid
,
2028 read_unlock(&neigh_tbl_lock
);
2035 static int neigh_fill_info(struct sk_buff
*skb
, struct neighbour
*neigh
,
2036 u32 pid
, u32 seq
, int type
, unsigned int flags
)
2038 unsigned long now
= jiffies
;
2039 struct nda_cacheinfo ci
;
2040 struct nlmsghdr
*nlh
;
2043 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), flags
);
2047 ndm
= nlmsg_data(nlh
);
2048 ndm
->ndm_family
= neigh
->ops
->family
;
2051 ndm
->ndm_flags
= neigh
->flags
;
2052 ndm
->ndm_type
= neigh
->type
;
2053 ndm
->ndm_ifindex
= neigh
->dev
->ifindex
;
2055 NLA_PUT(skb
, NDA_DST
, neigh
->tbl
->key_len
, neigh
->primary_key
);
2057 read_lock_bh(&neigh
->lock
);
2058 ndm
->ndm_state
= neigh
->nud_state
;
2059 if ((neigh
->nud_state
& NUD_VALID
) &&
2060 nla_put(skb
, NDA_LLADDR
, neigh
->dev
->addr_len
, neigh
->ha
) < 0) {
2061 read_unlock_bh(&neigh
->lock
);
2062 goto nla_put_failure
;
2065 ci
.ndm_used
= jiffies_to_clock_t(now
- neigh
->used
);
2066 ci
.ndm_confirmed
= jiffies_to_clock_t(now
- neigh
->confirmed
);
2067 ci
.ndm_updated
= jiffies_to_clock_t(now
- neigh
->updated
);
2068 ci
.ndm_refcnt
= atomic_read(&neigh
->refcnt
) - 1;
2069 read_unlock_bh(&neigh
->lock
);
2071 NLA_PUT_U32(skb
, NDA_PROBES
, atomic_read(&neigh
->probes
));
2072 NLA_PUT(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
);
2074 return nlmsg_end(skb
, nlh
);
2077 nlmsg_cancel(skb
, nlh
);
2081 static void neigh_update_notify(struct neighbour
*neigh
)
2083 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, neigh
);
2084 __neigh_notify(neigh
, RTM_NEWNEIGH
, 0);
2087 static int neigh_dump_table(struct neigh_table
*tbl
, struct sk_buff
*skb
,
2088 struct netlink_callback
*cb
)
2090 struct net
* net
= sock_net(skb
->sk
);
2091 struct neighbour
*n
;
2092 int rc
, h
, s_h
= cb
->args
[1];
2093 int idx
, s_idx
= idx
= cb
->args
[2];
2095 read_lock_bh(&tbl
->lock
);
2096 for (h
= 0; h
<= tbl
->hash_mask
; h
++) {
2101 for (n
= tbl
->hash_buckets
[h
], idx
= 0; n
; n
= n
->next
) {
2102 if (!net_eq(dev_net(n
->dev
), net
))
2106 if (neigh_fill_info(skb
, n
, NETLINK_CB(cb
->skb
).pid
,
2109 NLM_F_MULTI
) <= 0) {
2110 read_unlock_bh(&tbl
->lock
);
2118 read_unlock_bh(&tbl
->lock
);
2126 static int neigh_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2128 struct neigh_table
*tbl
;
2131 read_lock(&neigh_tbl_lock
);
2132 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
2135 for (tbl
= neigh_tables
, t
= 0; tbl
; tbl
= tbl
->next
, t
++) {
2136 if (t
< s_t
|| (family
&& tbl
->family
!= family
))
2139 memset(&cb
->args
[1], 0, sizeof(cb
->args
) -
2140 sizeof(cb
->args
[0]));
2141 if (neigh_dump_table(tbl
, skb
, cb
) < 0)
2144 read_unlock(&neigh_tbl_lock
);
2150 void neigh_for_each(struct neigh_table
*tbl
, void (*cb
)(struct neighbour
*, void *), void *cookie
)
2154 read_lock_bh(&tbl
->lock
);
2155 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2156 struct neighbour
*n
;
2158 for (n
= tbl
->hash_buckets
[chain
]; n
; n
= n
->next
)
2161 read_unlock_bh(&tbl
->lock
);
2163 EXPORT_SYMBOL(neigh_for_each
);
2165 /* The tbl->lock must be held as a writer and BH disabled. */
2166 void __neigh_for_each_release(struct neigh_table
*tbl
,
2167 int (*cb
)(struct neighbour
*))
2171 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2172 struct neighbour
*n
, **np
;
2174 np
= &tbl
->hash_buckets
[chain
];
2175 while ((n
= *np
) != NULL
) {
2178 write_lock(&n
->lock
);
2185 write_unlock(&n
->lock
);
2187 neigh_cleanup_and_release(n
);
2191 EXPORT_SYMBOL(__neigh_for_each_release
);
2193 #ifdef CONFIG_PROC_FS
2195 static struct neighbour
*neigh_get_first(struct seq_file
*seq
)
2197 struct neigh_seq_state
*state
= seq
->private;
2198 struct net
*net
= seq_file_net(seq
);
2199 struct neigh_table
*tbl
= state
->tbl
;
2200 struct neighbour
*n
= NULL
;
2201 int bucket
= state
->bucket
;
2203 state
->flags
&= ~NEIGH_SEQ_IS_PNEIGH
;
2204 for (bucket
= 0; bucket
<= tbl
->hash_mask
; bucket
++) {
2205 n
= tbl
->hash_buckets
[bucket
];
2208 if (!net_eq(dev_net(n
->dev
), net
))
2210 if (state
->neigh_sub_iter
) {
2214 v
= state
->neigh_sub_iter(state
, n
, &fakep
);
2218 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2220 if (n
->nud_state
& ~NUD_NOARP
)
2229 state
->bucket
= bucket
;
2234 static struct neighbour
*neigh_get_next(struct seq_file
*seq
,
2235 struct neighbour
*n
,
2238 struct neigh_seq_state
*state
= seq
->private;
2239 struct net
*net
= seq_file_net(seq
);
2240 struct neigh_table
*tbl
= state
->tbl
;
2242 if (state
->neigh_sub_iter
) {
2243 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2251 if (!net_eq(dev_net(n
->dev
), net
))
2253 if (state
->neigh_sub_iter
) {
2254 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2259 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2262 if (n
->nud_state
& ~NUD_NOARP
)
2271 if (++state
->bucket
> tbl
->hash_mask
)
2274 n
= tbl
->hash_buckets
[state
->bucket
];
2282 static struct neighbour
*neigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2284 struct neighbour
*n
= neigh_get_first(seq
);
2289 n
= neigh_get_next(seq
, n
, pos
);
2294 return *pos
? NULL
: n
;
2297 static struct pneigh_entry
*pneigh_get_first(struct seq_file
*seq
)
2299 struct neigh_seq_state
*state
= seq
->private;
2300 struct net
*net
= seq_file_net(seq
);
2301 struct neigh_table
*tbl
= state
->tbl
;
2302 struct pneigh_entry
*pn
= NULL
;
2303 int bucket
= state
->bucket
;
2305 state
->flags
|= NEIGH_SEQ_IS_PNEIGH
;
2306 for (bucket
= 0; bucket
<= PNEIGH_HASHMASK
; bucket
++) {
2307 pn
= tbl
->phash_buckets
[bucket
];
2308 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2313 state
->bucket
= bucket
;
2318 static struct pneigh_entry
*pneigh_get_next(struct seq_file
*seq
,
2319 struct pneigh_entry
*pn
,
2322 struct neigh_seq_state
*state
= seq
->private;
2323 struct net
*net
= seq_file_net(seq
);
2324 struct neigh_table
*tbl
= state
->tbl
;
2328 if (++state
->bucket
> PNEIGH_HASHMASK
)
2330 pn
= tbl
->phash_buckets
[state
->bucket
];
2331 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2343 static struct pneigh_entry
*pneigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2345 struct pneigh_entry
*pn
= pneigh_get_first(seq
);
2350 pn
= pneigh_get_next(seq
, pn
, pos
);
2355 return *pos
? NULL
: pn
;
2358 static void *neigh_get_idx_any(struct seq_file
*seq
, loff_t
*pos
)
2360 struct neigh_seq_state
*state
= seq
->private;
2362 loff_t idxpos
= *pos
;
2364 rc
= neigh_get_idx(seq
, &idxpos
);
2365 if (!rc
&& !(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2366 rc
= pneigh_get_idx(seq
, &idxpos
);
2371 void *neigh_seq_start(struct seq_file
*seq
, loff_t
*pos
, struct neigh_table
*tbl
, unsigned int neigh_seq_flags
)
2372 __acquires(tbl
->lock
)
2374 struct neigh_seq_state
*state
= seq
->private;
2378 state
->flags
= (neigh_seq_flags
& ~NEIGH_SEQ_IS_PNEIGH
);
2380 read_lock_bh(&tbl
->lock
);
2382 return *pos
? neigh_get_idx_any(seq
, pos
) : SEQ_START_TOKEN
;
2384 EXPORT_SYMBOL(neigh_seq_start
);
2386 void *neigh_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2388 struct neigh_seq_state
*state
;
2391 if (v
== SEQ_START_TOKEN
) {
2392 rc
= neigh_get_first(seq
);
2396 state
= seq
->private;
2397 if (!(state
->flags
& NEIGH_SEQ_IS_PNEIGH
)) {
2398 rc
= neigh_get_next(seq
, v
, NULL
);
2401 if (!(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2402 rc
= pneigh_get_first(seq
);
2404 BUG_ON(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
);
2405 rc
= pneigh_get_next(seq
, v
, NULL
);
2411 EXPORT_SYMBOL(neigh_seq_next
);
2413 void neigh_seq_stop(struct seq_file
*seq
, void *v
)
2414 __releases(tbl
->lock
)
2416 struct neigh_seq_state
*state
= seq
->private;
2417 struct neigh_table
*tbl
= state
->tbl
;
2419 read_unlock_bh(&tbl
->lock
);
2421 EXPORT_SYMBOL(neigh_seq_stop
);
2423 /* statistics via seq_file */
2425 static void *neigh_stat_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2427 struct neigh_table
*tbl
= seq
->private;
2431 return SEQ_START_TOKEN
;
2433 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
2434 if (!cpu_possible(cpu
))
2437 return per_cpu_ptr(tbl
->stats
, cpu
);
2442 static void *neigh_stat_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2444 struct neigh_table
*tbl
= seq
->private;
2447 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
2448 if (!cpu_possible(cpu
))
2451 return per_cpu_ptr(tbl
->stats
, cpu
);
2456 static void neigh_stat_seq_stop(struct seq_file
*seq
, void *v
)
2461 static int neigh_stat_seq_show(struct seq_file
*seq
, void *v
)
2463 struct neigh_table
*tbl
= seq
->private;
2464 struct neigh_statistics
*st
= v
;
2466 if (v
== SEQ_START_TOKEN
) {
2467 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");
2471 seq_printf(seq
, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2472 "%08lx %08lx %08lx %08lx %08lx\n",
2473 atomic_read(&tbl
->entries
),
2484 st
->rcv_probes_mcast
,
2485 st
->rcv_probes_ucast
,
2487 st
->periodic_gc_runs
,
2495 static const struct seq_operations neigh_stat_seq_ops
= {
2496 .start
= neigh_stat_seq_start
,
2497 .next
= neigh_stat_seq_next
,
2498 .stop
= neigh_stat_seq_stop
,
2499 .show
= neigh_stat_seq_show
,
2502 static int neigh_stat_seq_open(struct inode
*inode
, struct file
*file
)
2504 int ret
= seq_open(file
, &neigh_stat_seq_ops
);
2507 struct seq_file
*sf
= file
->private_data
;
2508 sf
->private = PDE(inode
)->data
;
2513 static const struct file_operations neigh_stat_seq_fops
= {
2514 .owner
= THIS_MODULE
,
2515 .open
= neigh_stat_seq_open
,
2517 .llseek
= seq_lseek
,
2518 .release
= seq_release
,
2521 #endif /* CONFIG_PROC_FS */
2523 static inline size_t neigh_nlmsg_size(void)
2525 return NLMSG_ALIGN(sizeof(struct ndmsg
))
2526 + nla_total_size(MAX_ADDR_LEN
) /* NDA_DST */
2527 + nla_total_size(MAX_ADDR_LEN
) /* NDA_LLADDR */
2528 + nla_total_size(sizeof(struct nda_cacheinfo
))
2529 + nla_total_size(4); /* NDA_PROBES */
2532 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
)
2534 struct net
*net
= dev_net(n
->dev
);
2535 struct sk_buff
*skb
;
2538 skb
= nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC
);
2542 err
= neigh_fill_info(skb
, n
, 0, 0, type
, flags
);
2544 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2545 WARN_ON(err
== -EMSGSIZE
);
2549 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2553 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2557 void neigh_app_ns(struct neighbour
*n
)
2559 __neigh_notify(n
, RTM_GETNEIGH
, NLM_F_REQUEST
);
2561 EXPORT_SYMBOL(neigh_app_ns
);
2562 #endif /* CONFIG_ARPD */
2564 #ifdef CONFIG_SYSCTL
2566 #define NEIGH_VARS_MAX 19
2568 static struct neigh_sysctl_table
{
2569 struct ctl_table_header
*sysctl_header
;
2570 struct ctl_table neigh_vars
[NEIGH_VARS_MAX
];
2572 } neigh_sysctl_template __read_mostly
= {
2575 .procname
= "mcast_solicit",
2576 .maxlen
= sizeof(int),
2578 .proc_handler
= proc_dointvec
,
2581 .procname
= "ucast_solicit",
2582 .maxlen
= sizeof(int),
2584 .proc_handler
= proc_dointvec
,
2587 .procname
= "app_solicit",
2588 .maxlen
= sizeof(int),
2590 .proc_handler
= proc_dointvec
,
2593 .procname
= "retrans_time",
2594 .maxlen
= sizeof(int),
2596 .proc_handler
= proc_dointvec_userhz_jiffies
,
2599 .procname
= "base_reachable_time",
2600 .maxlen
= sizeof(int),
2602 .proc_handler
= proc_dointvec_jiffies
,
2605 .procname
= "delay_first_probe_time",
2606 .maxlen
= sizeof(int),
2608 .proc_handler
= proc_dointvec_jiffies
,
2611 .procname
= "gc_stale_time",
2612 .maxlen
= sizeof(int),
2614 .proc_handler
= proc_dointvec_jiffies
,
2617 .procname
= "unres_qlen",
2618 .maxlen
= sizeof(int),
2620 .proc_handler
= proc_dointvec
,
2623 .procname
= "proxy_qlen",
2624 .maxlen
= sizeof(int),
2626 .proc_handler
= proc_dointvec
,
2629 .procname
= "anycast_delay",
2630 .maxlen
= sizeof(int),
2632 .proc_handler
= proc_dointvec_userhz_jiffies
,
2635 .procname
= "proxy_delay",
2636 .maxlen
= sizeof(int),
2638 .proc_handler
= proc_dointvec_userhz_jiffies
,
2641 .procname
= "locktime",
2642 .maxlen
= sizeof(int),
2644 .proc_handler
= proc_dointvec_userhz_jiffies
,
2647 .procname
= "retrans_time_ms",
2648 .maxlen
= sizeof(int),
2650 .proc_handler
= proc_dointvec_ms_jiffies
,
2653 .procname
= "base_reachable_time_ms",
2654 .maxlen
= sizeof(int),
2656 .proc_handler
= proc_dointvec_ms_jiffies
,
2659 .procname
= "gc_interval",
2660 .maxlen
= sizeof(int),
2662 .proc_handler
= proc_dointvec_jiffies
,
2665 .procname
= "gc_thresh1",
2666 .maxlen
= sizeof(int),
2668 .proc_handler
= proc_dointvec
,
2671 .procname
= "gc_thresh2",
2672 .maxlen
= sizeof(int),
2674 .proc_handler
= proc_dointvec
,
2677 .procname
= "gc_thresh3",
2678 .maxlen
= sizeof(int),
2680 .proc_handler
= proc_dointvec
,
2686 int neigh_sysctl_register(struct net_device
*dev
, struct neigh_parms
*p
,
2687 char *p_name
, proc_handler
*handler
)
2689 struct neigh_sysctl_table
*t
;
2690 const char *dev_name_source
= NULL
;
2692 #define NEIGH_CTL_PATH_ROOT 0
2693 #define NEIGH_CTL_PATH_PROTO 1
2694 #define NEIGH_CTL_PATH_NEIGH 2
2695 #define NEIGH_CTL_PATH_DEV 3
2697 struct ctl_path neigh_path
[] = {
2698 { .procname
= "net", },
2699 { .procname
= "proto", },
2700 { .procname
= "neigh", },
2701 { .procname
= "default", },
2705 t
= kmemdup(&neigh_sysctl_template
, sizeof(*t
), GFP_KERNEL
);
2709 t
->neigh_vars
[0].data
= &p
->mcast_probes
;
2710 t
->neigh_vars
[1].data
= &p
->ucast_probes
;
2711 t
->neigh_vars
[2].data
= &p
->app_probes
;
2712 t
->neigh_vars
[3].data
= &p
->retrans_time
;
2713 t
->neigh_vars
[4].data
= &p
->base_reachable_time
;
2714 t
->neigh_vars
[5].data
= &p
->delay_probe_time
;
2715 t
->neigh_vars
[6].data
= &p
->gc_staletime
;
2716 t
->neigh_vars
[7].data
= &p
->queue_len
;
2717 t
->neigh_vars
[8].data
= &p
->proxy_qlen
;
2718 t
->neigh_vars
[9].data
= &p
->anycast_delay
;
2719 t
->neigh_vars
[10].data
= &p
->proxy_delay
;
2720 t
->neigh_vars
[11].data
= &p
->locktime
;
2721 t
->neigh_vars
[12].data
= &p
->retrans_time
;
2722 t
->neigh_vars
[13].data
= &p
->base_reachable_time
;
2725 dev_name_source
= dev
->name
;
2726 /* Terminate the table early */
2727 memset(&t
->neigh_vars
[14], 0, sizeof(t
->neigh_vars
[14]));
2729 dev_name_source
= neigh_path
[NEIGH_CTL_PATH_DEV
].procname
;
2730 t
->neigh_vars
[14].data
= (int *)(p
+ 1);
2731 t
->neigh_vars
[15].data
= (int *)(p
+ 1) + 1;
2732 t
->neigh_vars
[16].data
= (int *)(p
+ 1) + 2;
2733 t
->neigh_vars
[17].data
= (int *)(p
+ 1) + 3;
2739 t
->neigh_vars
[3].proc_handler
= handler
;
2740 t
->neigh_vars
[3].extra1
= dev
;
2742 t
->neigh_vars
[4].proc_handler
= handler
;
2743 t
->neigh_vars
[4].extra1
= dev
;
2744 /* RetransTime (in milliseconds)*/
2745 t
->neigh_vars
[12].proc_handler
= handler
;
2746 t
->neigh_vars
[12].extra1
= dev
;
2747 /* ReachableTime (in milliseconds) */
2748 t
->neigh_vars
[13].proc_handler
= handler
;
2749 t
->neigh_vars
[13].extra1
= dev
;
2752 t
->dev_name
= kstrdup(dev_name_source
, GFP_KERNEL
);
2756 neigh_path
[NEIGH_CTL_PATH_DEV
].procname
= t
->dev_name
;
2757 neigh_path
[NEIGH_CTL_PATH_PROTO
].procname
= p_name
;
2760 register_net_sysctl_table(neigh_parms_net(p
), neigh_path
, t
->neigh_vars
);
2761 if (!t
->sysctl_header
)
2764 p
->sysctl_table
= t
;
2774 EXPORT_SYMBOL(neigh_sysctl_register
);
2776 void neigh_sysctl_unregister(struct neigh_parms
*p
)
2778 if (p
->sysctl_table
) {
2779 struct neigh_sysctl_table
*t
= p
->sysctl_table
;
2780 p
->sysctl_table
= NULL
;
2781 unregister_sysctl_table(t
->sysctl_header
);
2786 EXPORT_SYMBOL(neigh_sysctl_unregister
);
2788 #endif /* CONFIG_SYSCTL */
2790 static int __init
neigh_init(void)
2792 rtnl_register(PF_UNSPEC
, RTM_NEWNEIGH
, neigh_add
, NULL
);
2793 rtnl_register(PF_UNSPEC
, RTM_DELNEIGH
, neigh_delete
, NULL
);
2794 rtnl_register(PF_UNSPEC
, RTM_GETNEIGH
, NULL
, neigh_dump_info
);
2796 rtnl_register(PF_UNSPEC
, RTM_GETNEIGHTBL
, NULL
, neightbl_dump_info
);
2797 rtnl_register(PF_UNSPEC
, RTM_SETNEIGHTBL
, neightbl_set
, NULL
);
2802 subsys_initcall(neigh_init
);