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
&&
1215 !(dst
->flags
& DST_NOCACHE
)) {
1216 write_lock_bh(&neigh
->lock
);
1218 neigh_hh_init(neigh
, dst
, dst
->ops
->protocol
);
1219 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1220 neigh
->ha
, NULL
, skb
->len
);
1221 write_unlock_bh(&neigh
->lock
);
1223 read_lock_bh(&neigh
->lock
);
1224 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1225 neigh
->ha
, NULL
, skb
->len
);
1226 read_unlock_bh(&neigh
->lock
);
1229 rc
= neigh
->ops
->queue_xmit(skb
);
1236 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1237 dst
, dst
? dst
->neighbour
: NULL
);
1243 EXPORT_SYMBOL(neigh_resolve_output
);
1245 /* As fast as possible without hh cache */
1247 int neigh_connected_output(struct sk_buff
*skb
)
1250 struct dst_entry
*dst
= skb_dst(skb
);
1251 struct neighbour
*neigh
= dst
->neighbour
;
1252 struct net_device
*dev
= neigh
->dev
;
1254 __skb_pull(skb
, skb_network_offset(skb
));
1256 read_lock_bh(&neigh
->lock
);
1257 err
= dev_hard_header(skb
, dev
, ntohs(skb
->protocol
),
1258 neigh
->ha
, NULL
, skb
->len
);
1259 read_unlock_bh(&neigh
->lock
);
1261 err
= neigh
->ops
->queue_xmit(skb
);
1268 EXPORT_SYMBOL(neigh_connected_output
);
1270 static void neigh_proxy_process(unsigned long arg
)
1272 struct neigh_table
*tbl
= (struct neigh_table
*)arg
;
1273 long sched_next
= 0;
1274 unsigned long now
= jiffies
;
1275 struct sk_buff
*skb
, *n
;
1277 spin_lock(&tbl
->proxy_queue
.lock
);
1279 skb_queue_walk_safe(&tbl
->proxy_queue
, skb
, n
) {
1280 long tdif
= NEIGH_CB(skb
)->sched_next
- now
;
1283 struct net_device
*dev
= skb
->dev
;
1284 __skb_unlink(skb
, &tbl
->proxy_queue
);
1285 if (tbl
->proxy_redo
&& netif_running(dev
))
1286 tbl
->proxy_redo(skb
);
1291 } else if (!sched_next
|| tdif
< sched_next
)
1294 del_timer(&tbl
->proxy_timer
);
1296 mod_timer(&tbl
->proxy_timer
, jiffies
+ sched_next
);
1297 spin_unlock(&tbl
->proxy_queue
.lock
);
1300 void pneigh_enqueue(struct neigh_table
*tbl
, struct neigh_parms
*p
,
1301 struct sk_buff
*skb
)
1303 unsigned long now
= jiffies
;
1304 unsigned long sched_next
= now
+ (net_random() % p
->proxy_delay
);
1306 if (tbl
->proxy_queue
.qlen
> p
->proxy_qlen
) {
1311 NEIGH_CB(skb
)->sched_next
= sched_next
;
1312 NEIGH_CB(skb
)->flags
|= LOCALLY_ENQUEUED
;
1314 spin_lock(&tbl
->proxy_queue
.lock
);
1315 if (del_timer(&tbl
->proxy_timer
)) {
1316 if (time_before(tbl
->proxy_timer
.expires
, sched_next
))
1317 sched_next
= tbl
->proxy_timer
.expires
;
1321 __skb_queue_tail(&tbl
->proxy_queue
, skb
);
1322 mod_timer(&tbl
->proxy_timer
, sched_next
);
1323 spin_unlock(&tbl
->proxy_queue
.lock
);
1325 EXPORT_SYMBOL(pneigh_enqueue
);
1327 static inline struct neigh_parms
*lookup_neigh_parms(struct neigh_table
*tbl
,
1328 struct net
*net
, int ifindex
)
1330 struct neigh_parms
*p
;
1332 for (p
= &tbl
->parms
; p
; p
= p
->next
) {
1333 if ((p
->dev
&& p
->dev
->ifindex
== ifindex
&& net_eq(neigh_parms_net(p
), net
)) ||
1334 (!p
->dev
&& !ifindex
))
1341 struct neigh_parms
*neigh_parms_alloc(struct net_device
*dev
,
1342 struct neigh_table
*tbl
)
1344 struct neigh_parms
*p
, *ref
;
1345 struct net
*net
= dev_net(dev
);
1346 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1348 ref
= lookup_neigh_parms(tbl
, net
, 0);
1352 p
= kmemdup(ref
, sizeof(*p
), GFP_KERNEL
);
1355 atomic_set(&p
->refcnt
, 1);
1357 neigh_rand_reach_time(p
->base_reachable_time
);
1359 if (ops
->ndo_neigh_setup
&& ops
->ndo_neigh_setup(dev
, p
)) {
1366 write_pnet(&p
->net
, hold_net(net
));
1367 p
->sysctl_table
= NULL
;
1368 write_lock_bh(&tbl
->lock
);
1369 p
->next
= tbl
->parms
.next
;
1370 tbl
->parms
.next
= p
;
1371 write_unlock_bh(&tbl
->lock
);
1375 EXPORT_SYMBOL(neigh_parms_alloc
);
1377 static void neigh_rcu_free_parms(struct rcu_head
*head
)
1379 struct neigh_parms
*parms
=
1380 container_of(head
, struct neigh_parms
, rcu_head
);
1382 neigh_parms_put(parms
);
1385 void neigh_parms_release(struct neigh_table
*tbl
, struct neigh_parms
*parms
)
1387 struct neigh_parms
**p
;
1389 if (!parms
|| parms
== &tbl
->parms
)
1391 write_lock_bh(&tbl
->lock
);
1392 for (p
= &tbl
->parms
.next
; *p
; p
= &(*p
)->next
) {
1396 write_unlock_bh(&tbl
->lock
);
1398 dev_put(parms
->dev
);
1399 call_rcu(&parms
->rcu_head
, neigh_rcu_free_parms
);
1403 write_unlock_bh(&tbl
->lock
);
1404 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1406 EXPORT_SYMBOL(neigh_parms_release
);
1408 static void neigh_parms_destroy(struct neigh_parms
*parms
)
1410 release_net(neigh_parms_net(parms
));
1414 static struct lock_class_key neigh_table_proxy_queue_class
;
1416 void neigh_table_init_no_netlink(struct neigh_table
*tbl
)
1418 unsigned long now
= jiffies
;
1419 unsigned long phsize
;
1421 write_pnet(&tbl
->parms
.net
, &init_net
);
1422 atomic_set(&tbl
->parms
.refcnt
, 1);
1423 tbl
->parms
.reachable_time
=
1424 neigh_rand_reach_time(tbl
->parms
.base_reachable_time
);
1426 if (!tbl
->kmem_cachep
)
1428 kmem_cache_create(tbl
->id
, tbl
->entry_size
, 0,
1429 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
1431 tbl
->stats
= alloc_percpu(struct neigh_statistics
);
1433 panic("cannot create neighbour cache statistics");
1435 #ifdef CONFIG_PROC_FS
1436 if (!proc_create_data(tbl
->id
, 0, init_net
.proc_net_stat
,
1437 &neigh_stat_seq_fops
, tbl
))
1438 panic("cannot create neighbour proc dir entry");
1442 tbl
->hash_buckets
= neigh_hash_alloc(tbl
->hash_mask
+ 1);
1444 phsize
= (PNEIGH_HASHMASK
+ 1) * sizeof(struct pneigh_entry
*);
1445 tbl
->phash_buckets
= kzalloc(phsize
, GFP_KERNEL
);
1447 if (!tbl
->hash_buckets
|| !tbl
->phash_buckets
)
1448 panic("cannot allocate neighbour cache hashes");
1450 get_random_bytes(&tbl
->hash_rnd
, sizeof(tbl
->hash_rnd
));
1452 rwlock_init(&tbl
->lock
);
1453 INIT_DELAYED_WORK_DEFERRABLE(&tbl
->gc_work
, neigh_periodic_work
);
1454 schedule_delayed_work(&tbl
->gc_work
, tbl
->parms
.reachable_time
);
1455 setup_timer(&tbl
->proxy_timer
, neigh_proxy_process
, (unsigned long)tbl
);
1456 skb_queue_head_init_class(&tbl
->proxy_queue
,
1457 &neigh_table_proxy_queue_class
);
1459 tbl
->last_flush
= now
;
1460 tbl
->last_rand
= now
+ tbl
->parms
.reachable_time
* 20;
1462 EXPORT_SYMBOL(neigh_table_init_no_netlink
);
1464 void neigh_table_init(struct neigh_table
*tbl
)
1466 struct neigh_table
*tmp
;
1468 neigh_table_init_no_netlink(tbl
);
1469 write_lock(&neigh_tbl_lock
);
1470 for (tmp
= neigh_tables
; tmp
; tmp
= tmp
->next
) {
1471 if (tmp
->family
== tbl
->family
)
1474 tbl
->next
= neigh_tables
;
1476 write_unlock(&neigh_tbl_lock
);
1478 if (unlikely(tmp
)) {
1479 printk(KERN_ERR
"NEIGH: Registering multiple tables for "
1480 "family %d\n", tbl
->family
);
1484 EXPORT_SYMBOL(neigh_table_init
);
1486 int neigh_table_clear(struct neigh_table
*tbl
)
1488 struct neigh_table
**tp
;
1490 /* It is not clean... Fix it to unload IPv6 module safely */
1491 cancel_delayed_work(&tbl
->gc_work
);
1492 flush_scheduled_work();
1493 del_timer_sync(&tbl
->proxy_timer
);
1494 pneigh_queue_purge(&tbl
->proxy_queue
);
1495 neigh_ifdown(tbl
, NULL
);
1496 if (atomic_read(&tbl
->entries
))
1497 printk(KERN_CRIT
"neighbour leakage\n");
1498 write_lock(&neigh_tbl_lock
);
1499 for (tp
= &neigh_tables
; *tp
; tp
= &(*tp
)->next
) {
1505 write_unlock(&neigh_tbl_lock
);
1507 neigh_hash_free(tbl
->hash_buckets
, tbl
->hash_mask
+ 1);
1508 tbl
->hash_buckets
= NULL
;
1510 kfree(tbl
->phash_buckets
);
1511 tbl
->phash_buckets
= NULL
;
1513 remove_proc_entry(tbl
->id
, init_net
.proc_net_stat
);
1515 free_percpu(tbl
->stats
);
1518 kmem_cache_destroy(tbl
->kmem_cachep
);
1519 tbl
->kmem_cachep
= NULL
;
1523 EXPORT_SYMBOL(neigh_table_clear
);
1525 static int neigh_delete(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1527 struct net
*net
= sock_net(skb
->sk
);
1529 struct nlattr
*dst_attr
;
1530 struct neigh_table
*tbl
;
1531 struct net_device
*dev
= NULL
;
1535 if (nlmsg_len(nlh
) < sizeof(*ndm
))
1538 dst_attr
= nlmsg_find_attr(nlh
, sizeof(*ndm
), NDA_DST
);
1539 if (dst_attr
== NULL
)
1542 ndm
= nlmsg_data(nlh
);
1543 if (ndm
->ndm_ifindex
) {
1544 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
1551 read_lock(&neigh_tbl_lock
);
1552 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1553 struct neighbour
*neigh
;
1555 if (tbl
->family
!= ndm
->ndm_family
)
1557 read_unlock(&neigh_tbl_lock
);
1559 if (nla_len(dst_attr
) < tbl
->key_len
)
1562 if (ndm
->ndm_flags
& NTF_PROXY
) {
1563 err
= pneigh_delete(tbl
, net
, nla_data(dst_attr
), dev
);
1570 neigh
= neigh_lookup(tbl
, nla_data(dst_attr
), dev
);
1571 if (neigh
== NULL
) {
1576 err
= neigh_update(neigh
, NULL
, NUD_FAILED
,
1577 NEIGH_UPDATE_F_OVERRIDE
|
1578 NEIGH_UPDATE_F_ADMIN
);
1579 neigh_release(neigh
);
1582 read_unlock(&neigh_tbl_lock
);
1583 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
;
1599 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
1604 if (tb
[NDA_DST
] == NULL
)
1607 ndm
= nlmsg_data(nlh
);
1608 if (ndm
->ndm_ifindex
) {
1609 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
1615 if (tb
[NDA_LLADDR
] && nla_len(tb
[NDA_LLADDR
]) < dev
->addr_len
)
1619 read_lock(&neigh_tbl_lock
);
1620 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1621 int flags
= NEIGH_UPDATE_F_ADMIN
| NEIGH_UPDATE_F_OVERRIDE
;
1622 struct neighbour
*neigh
;
1625 if (tbl
->family
!= ndm
->ndm_family
)
1627 read_unlock(&neigh_tbl_lock
);
1629 if (nla_len(tb
[NDA_DST
]) < tbl
->key_len
)
1631 dst
= nla_data(tb
[NDA_DST
]);
1632 lladdr
= tb
[NDA_LLADDR
] ? nla_data(tb
[NDA_LLADDR
]) : NULL
;
1634 if (ndm
->ndm_flags
& NTF_PROXY
) {
1635 struct pneigh_entry
*pn
;
1638 pn
= pneigh_lookup(tbl
, net
, dst
, dev
, 1);
1640 pn
->flags
= ndm
->ndm_flags
;
1649 neigh
= neigh_lookup(tbl
, dst
, dev
);
1650 if (neigh
== NULL
) {
1651 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1656 neigh
= __neigh_lookup_errno(tbl
, dst
, dev
);
1657 if (IS_ERR(neigh
)) {
1658 err
= PTR_ERR(neigh
);
1662 if (nlh
->nlmsg_flags
& NLM_F_EXCL
) {
1664 neigh_release(neigh
);
1668 if (!(nlh
->nlmsg_flags
& NLM_F_REPLACE
))
1669 flags
&= ~NEIGH_UPDATE_F_OVERRIDE
;
1672 if (ndm
->ndm_flags
& NTF_USE
) {
1673 neigh_event_send(neigh
, NULL
);
1676 err
= neigh_update(neigh
, lladdr
, ndm
->ndm_state
, flags
);
1677 neigh_release(neigh
);
1681 read_unlock(&neigh_tbl_lock
);
1682 err
= -EAFNOSUPPORT
;
1687 static int neightbl_fill_parms(struct sk_buff
*skb
, struct neigh_parms
*parms
)
1689 struct nlattr
*nest
;
1691 nest
= nla_nest_start(skb
, NDTA_PARMS
);
1696 NLA_PUT_U32(skb
, NDTPA_IFINDEX
, parms
->dev
->ifindex
);
1698 NLA_PUT_U32(skb
, NDTPA_REFCNT
, atomic_read(&parms
->refcnt
));
1699 NLA_PUT_U32(skb
, NDTPA_QUEUE_LEN
, parms
->queue_len
);
1700 NLA_PUT_U32(skb
, NDTPA_PROXY_QLEN
, parms
->proxy_qlen
);
1701 NLA_PUT_U32(skb
, NDTPA_APP_PROBES
, parms
->app_probes
);
1702 NLA_PUT_U32(skb
, NDTPA_UCAST_PROBES
, parms
->ucast_probes
);
1703 NLA_PUT_U32(skb
, NDTPA_MCAST_PROBES
, parms
->mcast_probes
);
1704 NLA_PUT_MSECS(skb
, NDTPA_REACHABLE_TIME
, parms
->reachable_time
);
1705 NLA_PUT_MSECS(skb
, NDTPA_BASE_REACHABLE_TIME
,
1706 parms
->base_reachable_time
);
1707 NLA_PUT_MSECS(skb
, NDTPA_GC_STALETIME
, parms
->gc_staletime
);
1708 NLA_PUT_MSECS(skb
, NDTPA_DELAY_PROBE_TIME
, parms
->delay_probe_time
);
1709 NLA_PUT_MSECS(skb
, NDTPA_RETRANS_TIME
, parms
->retrans_time
);
1710 NLA_PUT_MSECS(skb
, NDTPA_ANYCAST_DELAY
, parms
->anycast_delay
);
1711 NLA_PUT_MSECS(skb
, NDTPA_PROXY_DELAY
, parms
->proxy_delay
);
1712 NLA_PUT_MSECS(skb
, NDTPA_LOCKTIME
, parms
->locktime
);
1714 return nla_nest_end(skb
, nest
);
1717 nla_nest_cancel(skb
, nest
);
1721 static int neightbl_fill_info(struct sk_buff
*skb
, struct neigh_table
*tbl
,
1722 u32 pid
, u32 seq
, int type
, int flags
)
1724 struct nlmsghdr
*nlh
;
1725 struct ndtmsg
*ndtmsg
;
1727 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1731 ndtmsg
= nlmsg_data(nlh
);
1733 read_lock_bh(&tbl
->lock
);
1734 ndtmsg
->ndtm_family
= tbl
->family
;
1735 ndtmsg
->ndtm_pad1
= 0;
1736 ndtmsg
->ndtm_pad2
= 0;
1738 NLA_PUT_STRING(skb
, NDTA_NAME
, tbl
->id
);
1739 NLA_PUT_MSECS(skb
, NDTA_GC_INTERVAL
, tbl
->gc_interval
);
1740 NLA_PUT_U32(skb
, NDTA_THRESH1
, tbl
->gc_thresh1
);
1741 NLA_PUT_U32(skb
, NDTA_THRESH2
, tbl
->gc_thresh2
);
1742 NLA_PUT_U32(skb
, NDTA_THRESH3
, tbl
->gc_thresh3
);
1745 unsigned long now
= jiffies
;
1746 unsigned int flush_delta
= now
- tbl
->last_flush
;
1747 unsigned int rand_delta
= now
- tbl
->last_rand
;
1749 struct ndt_config ndc
= {
1750 .ndtc_key_len
= tbl
->key_len
,
1751 .ndtc_entry_size
= tbl
->entry_size
,
1752 .ndtc_entries
= atomic_read(&tbl
->entries
),
1753 .ndtc_last_flush
= jiffies_to_msecs(flush_delta
),
1754 .ndtc_last_rand
= jiffies_to_msecs(rand_delta
),
1755 .ndtc_hash_rnd
= tbl
->hash_rnd
,
1756 .ndtc_hash_mask
= tbl
->hash_mask
,
1757 .ndtc_proxy_qlen
= tbl
->proxy_queue
.qlen
,
1760 NLA_PUT(skb
, NDTA_CONFIG
, sizeof(ndc
), &ndc
);
1765 struct ndt_stats ndst
;
1767 memset(&ndst
, 0, sizeof(ndst
));
1769 for_each_possible_cpu(cpu
) {
1770 struct neigh_statistics
*st
;
1772 st
= per_cpu_ptr(tbl
->stats
, cpu
);
1773 ndst
.ndts_allocs
+= st
->allocs
;
1774 ndst
.ndts_destroys
+= st
->destroys
;
1775 ndst
.ndts_hash_grows
+= st
->hash_grows
;
1776 ndst
.ndts_res_failed
+= st
->res_failed
;
1777 ndst
.ndts_lookups
+= st
->lookups
;
1778 ndst
.ndts_hits
+= st
->hits
;
1779 ndst
.ndts_rcv_probes_mcast
+= st
->rcv_probes_mcast
;
1780 ndst
.ndts_rcv_probes_ucast
+= st
->rcv_probes_ucast
;
1781 ndst
.ndts_periodic_gc_runs
+= st
->periodic_gc_runs
;
1782 ndst
.ndts_forced_gc_runs
+= st
->forced_gc_runs
;
1785 NLA_PUT(skb
, NDTA_STATS
, sizeof(ndst
), &ndst
);
1788 BUG_ON(tbl
->parms
.dev
);
1789 if (neightbl_fill_parms(skb
, &tbl
->parms
) < 0)
1790 goto nla_put_failure
;
1792 read_unlock_bh(&tbl
->lock
);
1793 return nlmsg_end(skb
, nlh
);
1796 read_unlock_bh(&tbl
->lock
);
1797 nlmsg_cancel(skb
, nlh
);
1801 static int neightbl_fill_param_info(struct sk_buff
*skb
,
1802 struct neigh_table
*tbl
,
1803 struct neigh_parms
*parms
,
1804 u32 pid
, u32 seq
, int type
,
1807 struct ndtmsg
*ndtmsg
;
1808 struct nlmsghdr
*nlh
;
1810 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndtmsg
), flags
);
1814 ndtmsg
= nlmsg_data(nlh
);
1816 read_lock_bh(&tbl
->lock
);
1817 ndtmsg
->ndtm_family
= tbl
->family
;
1818 ndtmsg
->ndtm_pad1
= 0;
1819 ndtmsg
->ndtm_pad2
= 0;
1821 if (nla_put_string(skb
, NDTA_NAME
, tbl
->id
) < 0 ||
1822 neightbl_fill_parms(skb
, parms
) < 0)
1825 read_unlock_bh(&tbl
->lock
);
1826 return nlmsg_end(skb
, nlh
);
1828 read_unlock_bh(&tbl
->lock
);
1829 nlmsg_cancel(skb
, nlh
);
1833 static const struct nla_policy nl_neightbl_policy
[NDTA_MAX
+1] = {
1834 [NDTA_NAME
] = { .type
= NLA_STRING
},
1835 [NDTA_THRESH1
] = { .type
= NLA_U32
},
1836 [NDTA_THRESH2
] = { .type
= NLA_U32
},
1837 [NDTA_THRESH3
] = { .type
= NLA_U32
},
1838 [NDTA_GC_INTERVAL
] = { .type
= NLA_U64
},
1839 [NDTA_PARMS
] = { .type
= NLA_NESTED
},
1842 static const struct nla_policy nl_ntbl_parm_policy
[NDTPA_MAX
+1] = {
1843 [NDTPA_IFINDEX
] = { .type
= NLA_U32
},
1844 [NDTPA_QUEUE_LEN
] = { .type
= NLA_U32
},
1845 [NDTPA_PROXY_QLEN
] = { .type
= NLA_U32
},
1846 [NDTPA_APP_PROBES
] = { .type
= NLA_U32
},
1847 [NDTPA_UCAST_PROBES
] = { .type
= NLA_U32
},
1848 [NDTPA_MCAST_PROBES
] = { .type
= NLA_U32
},
1849 [NDTPA_BASE_REACHABLE_TIME
] = { .type
= NLA_U64
},
1850 [NDTPA_GC_STALETIME
] = { .type
= NLA_U64
},
1851 [NDTPA_DELAY_PROBE_TIME
] = { .type
= NLA_U64
},
1852 [NDTPA_RETRANS_TIME
] = { .type
= NLA_U64
},
1853 [NDTPA_ANYCAST_DELAY
] = { .type
= NLA_U64
},
1854 [NDTPA_PROXY_DELAY
] = { .type
= NLA_U64
},
1855 [NDTPA_LOCKTIME
] = { .type
= NLA_U64
},
1858 static int neightbl_set(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1860 struct net
*net
= sock_net(skb
->sk
);
1861 struct neigh_table
*tbl
;
1862 struct ndtmsg
*ndtmsg
;
1863 struct nlattr
*tb
[NDTA_MAX
+1];
1866 err
= nlmsg_parse(nlh
, sizeof(*ndtmsg
), tb
, NDTA_MAX
,
1867 nl_neightbl_policy
);
1871 if (tb
[NDTA_NAME
] == NULL
) {
1876 ndtmsg
= nlmsg_data(nlh
);
1877 read_lock(&neigh_tbl_lock
);
1878 for (tbl
= neigh_tables
; tbl
; tbl
= tbl
->next
) {
1879 if (ndtmsg
->ndtm_family
&& tbl
->family
!= ndtmsg
->ndtm_family
)
1882 if (nla_strcmp(tb
[NDTA_NAME
], tbl
->id
) == 0)
1892 * We acquire tbl->lock to be nice to the periodic timers and
1893 * make sure they always see a consistent set of values.
1895 write_lock_bh(&tbl
->lock
);
1897 if (tb
[NDTA_PARMS
]) {
1898 struct nlattr
*tbp
[NDTPA_MAX
+1];
1899 struct neigh_parms
*p
;
1902 err
= nla_parse_nested(tbp
, NDTPA_MAX
, tb
[NDTA_PARMS
],
1903 nl_ntbl_parm_policy
);
1905 goto errout_tbl_lock
;
1907 if (tbp
[NDTPA_IFINDEX
])
1908 ifindex
= nla_get_u32(tbp
[NDTPA_IFINDEX
]);
1910 p
= lookup_neigh_parms(tbl
, net
, ifindex
);
1913 goto errout_tbl_lock
;
1916 for (i
= 1; i
<= NDTPA_MAX
; i
++) {
1921 case NDTPA_QUEUE_LEN
:
1922 p
->queue_len
= nla_get_u32(tbp
[i
]);
1924 case NDTPA_PROXY_QLEN
:
1925 p
->proxy_qlen
= nla_get_u32(tbp
[i
]);
1927 case NDTPA_APP_PROBES
:
1928 p
->app_probes
= nla_get_u32(tbp
[i
]);
1930 case NDTPA_UCAST_PROBES
:
1931 p
->ucast_probes
= nla_get_u32(tbp
[i
]);
1933 case NDTPA_MCAST_PROBES
:
1934 p
->mcast_probes
= nla_get_u32(tbp
[i
]);
1936 case NDTPA_BASE_REACHABLE_TIME
:
1937 p
->base_reachable_time
= nla_get_msecs(tbp
[i
]);
1939 case NDTPA_GC_STALETIME
:
1940 p
->gc_staletime
= nla_get_msecs(tbp
[i
]);
1942 case NDTPA_DELAY_PROBE_TIME
:
1943 p
->delay_probe_time
= nla_get_msecs(tbp
[i
]);
1945 case NDTPA_RETRANS_TIME
:
1946 p
->retrans_time
= nla_get_msecs(tbp
[i
]);
1948 case NDTPA_ANYCAST_DELAY
:
1949 p
->anycast_delay
= nla_get_msecs(tbp
[i
]);
1951 case NDTPA_PROXY_DELAY
:
1952 p
->proxy_delay
= nla_get_msecs(tbp
[i
]);
1954 case NDTPA_LOCKTIME
:
1955 p
->locktime
= nla_get_msecs(tbp
[i
]);
1961 if (tb
[NDTA_THRESH1
])
1962 tbl
->gc_thresh1
= nla_get_u32(tb
[NDTA_THRESH1
]);
1964 if (tb
[NDTA_THRESH2
])
1965 tbl
->gc_thresh2
= nla_get_u32(tb
[NDTA_THRESH2
]);
1967 if (tb
[NDTA_THRESH3
])
1968 tbl
->gc_thresh3
= nla_get_u32(tb
[NDTA_THRESH3
]);
1970 if (tb
[NDTA_GC_INTERVAL
])
1971 tbl
->gc_interval
= nla_get_msecs(tb
[NDTA_GC_INTERVAL
]);
1976 write_unlock_bh(&tbl
->lock
);
1978 read_unlock(&neigh_tbl_lock
);
1983 static int neightbl_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1985 struct net
*net
= sock_net(skb
->sk
);
1986 int family
, tidx
, nidx
= 0;
1987 int tbl_skip
= cb
->args
[0];
1988 int neigh_skip
= cb
->args
[1];
1989 struct neigh_table
*tbl
;
1991 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
1993 read_lock(&neigh_tbl_lock
);
1994 for (tbl
= neigh_tables
, tidx
= 0; tbl
; tbl
= tbl
->next
, tidx
++) {
1995 struct neigh_parms
*p
;
1997 if (tidx
< tbl_skip
|| (family
&& tbl
->family
!= family
))
2000 if (neightbl_fill_info(skb
, tbl
, NETLINK_CB(cb
->skb
).pid
,
2001 cb
->nlh
->nlmsg_seq
, RTM_NEWNEIGHTBL
,
2005 for (nidx
= 0, p
= tbl
->parms
.next
; p
; p
= p
->next
) {
2006 if (!net_eq(neigh_parms_net(p
), net
))
2009 if (nidx
< neigh_skip
)
2012 if (neightbl_fill_param_info(skb
, tbl
, p
,
2013 NETLINK_CB(cb
->skb
).pid
,
2025 read_unlock(&neigh_tbl_lock
);
2032 static int neigh_fill_info(struct sk_buff
*skb
, struct neighbour
*neigh
,
2033 u32 pid
, u32 seq
, int type
, unsigned int flags
)
2035 unsigned long now
= jiffies
;
2036 struct nda_cacheinfo ci
;
2037 struct nlmsghdr
*nlh
;
2040 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), flags
);
2044 ndm
= nlmsg_data(nlh
);
2045 ndm
->ndm_family
= neigh
->ops
->family
;
2048 ndm
->ndm_flags
= neigh
->flags
;
2049 ndm
->ndm_type
= neigh
->type
;
2050 ndm
->ndm_ifindex
= neigh
->dev
->ifindex
;
2052 NLA_PUT(skb
, NDA_DST
, neigh
->tbl
->key_len
, neigh
->primary_key
);
2054 read_lock_bh(&neigh
->lock
);
2055 ndm
->ndm_state
= neigh
->nud_state
;
2056 if ((neigh
->nud_state
& NUD_VALID
) &&
2057 nla_put(skb
, NDA_LLADDR
, neigh
->dev
->addr_len
, neigh
->ha
) < 0) {
2058 read_unlock_bh(&neigh
->lock
);
2059 goto nla_put_failure
;
2062 ci
.ndm_used
= jiffies_to_clock_t(now
- neigh
->used
);
2063 ci
.ndm_confirmed
= jiffies_to_clock_t(now
- neigh
->confirmed
);
2064 ci
.ndm_updated
= jiffies_to_clock_t(now
- neigh
->updated
);
2065 ci
.ndm_refcnt
= atomic_read(&neigh
->refcnt
) - 1;
2066 read_unlock_bh(&neigh
->lock
);
2068 NLA_PUT_U32(skb
, NDA_PROBES
, atomic_read(&neigh
->probes
));
2069 NLA_PUT(skb
, NDA_CACHEINFO
, sizeof(ci
), &ci
);
2071 return nlmsg_end(skb
, nlh
);
2074 nlmsg_cancel(skb
, nlh
);
2078 static void neigh_update_notify(struct neighbour
*neigh
)
2080 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE
, neigh
);
2081 __neigh_notify(neigh
, RTM_NEWNEIGH
, 0);
2084 static int neigh_dump_table(struct neigh_table
*tbl
, struct sk_buff
*skb
,
2085 struct netlink_callback
*cb
)
2087 struct net
* net
= sock_net(skb
->sk
);
2088 struct neighbour
*n
;
2089 int rc
, h
, s_h
= cb
->args
[1];
2090 int idx
, s_idx
= idx
= cb
->args
[2];
2092 read_lock_bh(&tbl
->lock
);
2093 for (h
= 0; h
<= tbl
->hash_mask
; h
++) {
2098 for (n
= tbl
->hash_buckets
[h
], idx
= 0; n
; n
= n
->next
) {
2099 if (!net_eq(dev_net(n
->dev
), net
))
2103 if (neigh_fill_info(skb
, n
, NETLINK_CB(cb
->skb
).pid
,
2106 NLM_F_MULTI
) <= 0) {
2107 read_unlock_bh(&tbl
->lock
);
2115 read_unlock_bh(&tbl
->lock
);
2123 static int neigh_dump_info(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2125 struct neigh_table
*tbl
;
2128 read_lock(&neigh_tbl_lock
);
2129 family
= ((struct rtgenmsg
*) nlmsg_data(cb
->nlh
))->rtgen_family
;
2132 for (tbl
= neigh_tables
, t
= 0; tbl
; tbl
= tbl
->next
, t
++) {
2133 if (t
< s_t
|| (family
&& tbl
->family
!= family
))
2136 memset(&cb
->args
[1], 0, sizeof(cb
->args
) -
2137 sizeof(cb
->args
[0]));
2138 if (neigh_dump_table(tbl
, skb
, cb
) < 0)
2141 read_unlock(&neigh_tbl_lock
);
2147 void neigh_for_each(struct neigh_table
*tbl
, void (*cb
)(struct neighbour
*, void *), void *cookie
)
2151 read_lock_bh(&tbl
->lock
);
2152 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2153 struct neighbour
*n
;
2155 for (n
= tbl
->hash_buckets
[chain
]; n
; n
= n
->next
)
2158 read_unlock_bh(&tbl
->lock
);
2160 EXPORT_SYMBOL(neigh_for_each
);
2162 /* The tbl->lock must be held as a writer and BH disabled. */
2163 void __neigh_for_each_release(struct neigh_table
*tbl
,
2164 int (*cb
)(struct neighbour
*))
2168 for (chain
= 0; chain
<= tbl
->hash_mask
; chain
++) {
2169 struct neighbour
*n
, **np
;
2171 np
= &tbl
->hash_buckets
[chain
];
2172 while ((n
= *np
) != NULL
) {
2175 write_lock(&n
->lock
);
2182 write_unlock(&n
->lock
);
2184 neigh_cleanup_and_release(n
);
2188 EXPORT_SYMBOL(__neigh_for_each_release
);
2190 #ifdef CONFIG_PROC_FS
2192 static struct neighbour
*neigh_get_first(struct seq_file
*seq
)
2194 struct neigh_seq_state
*state
= seq
->private;
2195 struct net
*net
= seq_file_net(seq
);
2196 struct neigh_table
*tbl
= state
->tbl
;
2197 struct neighbour
*n
= NULL
;
2198 int bucket
= state
->bucket
;
2200 state
->flags
&= ~NEIGH_SEQ_IS_PNEIGH
;
2201 for (bucket
= 0; bucket
<= tbl
->hash_mask
; bucket
++) {
2202 n
= tbl
->hash_buckets
[bucket
];
2205 if (!net_eq(dev_net(n
->dev
), net
))
2207 if (state
->neigh_sub_iter
) {
2211 v
= state
->neigh_sub_iter(state
, n
, &fakep
);
2215 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2217 if (n
->nud_state
& ~NUD_NOARP
)
2226 state
->bucket
= bucket
;
2231 static struct neighbour
*neigh_get_next(struct seq_file
*seq
,
2232 struct neighbour
*n
,
2235 struct neigh_seq_state
*state
= seq
->private;
2236 struct net
*net
= seq_file_net(seq
);
2237 struct neigh_table
*tbl
= state
->tbl
;
2239 if (state
->neigh_sub_iter
) {
2240 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2248 if (!net_eq(dev_net(n
->dev
), net
))
2250 if (state
->neigh_sub_iter
) {
2251 void *v
= state
->neigh_sub_iter(state
, n
, pos
);
2256 if (!(state
->flags
& NEIGH_SEQ_SKIP_NOARP
))
2259 if (n
->nud_state
& ~NUD_NOARP
)
2268 if (++state
->bucket
> tbl
->hash_mask
)
2271 n
= tbl
->hash_buckets
[state
->bucket
];
2279 static struct neighbour
*neigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2281 struct neighbour
*n
= neigh_get_first(seq
);
2286 n
= neigh_get_next(seq
, n
, pos
);
2291 return *pos
? NULL
: n
;
2294 static struct pneigh_entry
*pneigh_get_first(struct seq_file
*seq
)
2296 struct neigh_seq_state
*state
= seq
->private;
2297 struct net
*net
= seq_file_net(seq
);
2298 struct neigh_table
*tbl
= state
->tbl
;
2299 struct pneigh_entry
*pn
= NULL
;
2300 int bucket
= state
->bucket
;
2302 state
->flags
|= NEIGH_SEQ_IS_PNEIGH
;
2303 for (bucket
= 0; bucket
<= PNEIGH_HASHMASK
; bucket
++) {
2304 pn
= tbl
->phash_buckets
[bucket
];
2305 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2310 state
->bucket
= bucket
;
2315 static struct pneigh_entry
*pneigh_get_next(struct seq_file
*seq
,
2316 struct pneigh_entry
*pn
,
2319 struct neigh_seq_state
*state
= seq
->private;
2320 struct net
*net
= seq_file_net(seq
);
2321 struct neigh_table
*tbl
= state
->tbl
;
2325 if (++state
->bucket
> PNEIGH_HASHMASK
)
2327 pn
= tbl
->phash_buckets
[state
->bucket
];
2328 while (pn
&& !net_eq(pneigh_net(pn
), net
))
2340 static struct pneigh_entry
*pneigh_get_idx(struct seq_file
*seq
, loff_t
*pos
)
2342 struct pneigh_entry
*pn
= pneigh_get_first(seq
);
2347 pn
= pneigh_get_next(seq
, pn
, pos
);
2352 return *pos
? NULL
: pn
;
2355 static void *neigh_get_idx_any(struct seq_file
*seq
, loff_t
*pos
)
2357 struct neigh_seq_state
*state
= seq
->private;
2359 loff_t idxpos
= *pos
;
2361 rc
= neigh_get_idx(seq
, &idxpos
);
2362 if (!rc
&& !(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2363 rc
= pneigh_get_idx(seq
, &idxpos
);
2368 void *neigh_seq_start(struct seq_file
*seq
, loff_t
*pos
, struct neigh_table
*tbl
, unsigned int neigh_seq_flags
)
2369 __acquires(tbl
->lock
)
2371 struct neigh_seq_state
*state
= seq
->private;
2375 state
->flags
= (neigh_seq_flags
& ~NEIGH_SEQ_IS_PNEIGH
);
2377 read_lock_bh(&tbl
->lock
);
2379 return *pos
? neigh_get_idx_any(seq
, pos
) : SEQ_START_TOKEN
;
2381 EXPORT_SYMBOL(neigh_seq_start
);
2383 void *neigh_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2385 struct neigh_seq_state
*state
;
2388 if (v
== SEQ_START_TOKEN
) {
2389 rc
= neigh_get_first(seq
);
2393 state
= seq
->private;
2394 if (!(state
->flags
& NEIGH_SEQ_IS_PNEIGH
)) {
2395 rc
= neigh_get_next(seq
, v
, NULL
);
2398 if (!(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
))
2399 rc
= pneigh_get_first(seq
);
2401 BUG_ON(state
->flags
& NEIGH_SEQ_NEIGH_ONLY
);
2402 rc
= pneigh_get_next(seq
, v
, NULL
);
2408 EXPORT_SYMBOL(neigh_seq_next
);
2410 void neigh_seq_stop(struct seq_file
*seq
, void *v
)
2411 __releases(tbl
->lock
)
2413 struct neigh_seq_state
*state
= seq
->private;
2414 struct neigh_table
*tbl
= state
->tbl
;
2416 read_unlock_bh(&tbl
->lock
);
2418 EXPORT_SYMBOL(neigh_seq_stop
);
2420 /* statistics via seq_file */
2422 static void *neigh_stat_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2424 struct neigh_table
*tbl
= seq
->private;
2428 return SEQ_START_TOKEN
;
2430 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
2431 if (!cpu_possible(cpu
))
2434 return per_cpu_ptr(tbl
->stats
, cpu
);
2439 static void *neigh_stat_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2441 struct neigh_table
*tbl
= seq
->private;
2444 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
2445 if (!cpu_possible(cpu
))
2448 return per_cpu_ptr(tbl
->stats
, cpu
);
2453 static void neigh_stat_seq_stop(struct seq_file
*seq
, void *v
)
2458 static int neigh_stat_seq_show(struct seq_file
*seq
, void *v
)
2460 struct neigh_table
*tbl
= seq
->private;
2461 struct neigh_statistics
*st
= v
;
2463 if (v
== SEQ_START_TOKEN
) {
2464 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");
2468 seq_printf(seq
, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2469 "%08lx %08lx %08lx %08lx %08lx\n",
2470 atomic_read(&tbl
->entries
),
2481 st
->rcv_probes_mcast
,
2482 st
->rcv_probes_ucast
,
2484 st
->periodic_gc_runs
,
2492 static const struct seq_operations neigh_stat_seq_ops
= {
2493 .start
= neigh_stat_seq_start
,
2494 .next
= neigh_stat_seq_next
,
2495 .stop
= neigh_stat_seq_stop
,
2496 .show
= neigh_stat_seq_show
,
2499 static int neigh_stat_seq_open(struct inode
*inode
, struct file
*file
)
2501 int ret
= seq_open(file
, &neigh_stat_seq_ops
);
2504 struct seq_file
*sf
= file
->private_data
;
2505 sf
->private = PDE(inode
)->data
;
2510 static const struct file_operations neigh_stat_seq_fops
= {
2511 .owner
= THIS_MODULE
,
2512 .open
= neigh_stat_seq_open
,
2514 .llseek
= seq_lseek
,
2515 .release
= seq_release
,
2518 #endif /* CONFIG_PROC_FS */
2520 static inline size_t neigh_nlmsg_size(void)
2522 return NLMSG_ALIGN(sizeof(struct ndmsg
))
2523 + nla_total_size(MAX_ADDR_LEN
) /* NDA_DST */
2524 + nla_total_size(MAX_ADDR_LEN
) /* NDA_LLADDR */
2525 + nla_total_size(sizeof(struct nda_cacheinfo
))
2526 + nla_total_size(4); /* NDA_PROBES */
2529 static void __neigh_notify(struct neighbour
*n
, int type
, int flags
)
2531 struct net
*net
= dev_net(n
->dev
);
2532 struct sk_buff
*skb
;
2535 skb
= nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC
);
2539 err
= neigh_fill_info(skb
, n
, 0, 0, type
, flags
);
2541 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2542 WARN_ON(err
== -EMSGSIZE
);
2546 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2550 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2554 void neigh_app_ns(struct neighbour
*n
)
2556 __neigh_notify(n
, RTM_GETNEIGH
, NLM_F_REQUEST
);
2558 EXPORT_SYMBOL(neigh_app_ns
);
2559 #endif /* CONFIG_ARPD */
2561 #ifdef CONFIG_SYSCTL
2563 #define NEIGH_VARS_MAX 19
2565 static struct neigh_sysctl_table
{
2566 struct ctl_table_header
*sysctl_header
;
2567 struct ctl_table neigh_vars
[NEIGH_VARS_MAX
];
2569 } neigh_sysctl_template __read_mostly
= {
2572 .procname
= "mcast_solicit",
2573 .maxlen
= sizeof(int),
2575 .proc_handler
= proc_dointvec
,
2578 .procname
= "ucast_solicit",
2579 .maxlen
= sizeof(int),
2581 .proc_handler
= proc_dointvec
,
2584 .procname
= "app_solicit",
2585 .maxlen
= sizeof(int),
2587 .proc_handler
= proc_dointvec
,
2590 .procname
= "retrans_time",
2591 .maxlen
= sizeof(int),
2593 .proc_handler
= proc_dointvec_userhz_jiffies
,
2596 .procname
= "base_reachable_time",
2597 .maxlen
= sizeof(int),
2599 .proc_handler
= proc_dointvec_jiffies
,
2602 .procname
= "delay_first_probe_time",
2603 .maxlen
= sizeof(int),
2605 .proc_handler
= proc_dointvec_jiffies
,
2608 .procname
= "gc_stale_time",
2609 .maxlen
= sizeof(int),
2611 .proc_handler
= proc_dointvec_jiffies
,
2614 .procname
= "unres_qlen",
2615 .maxlen
= sizeof(int),
2617 .proc_handler
= proc_dointvec
,
2620 .procname
= "proxy_qlen",
2621 .maxlen
= sizeof(int),
2623 .proc_handler
= proc_dointvec
,
2626 .procname
= "anycast_delay",
2627 .maxlen
= sizeof(int),
2629 .proc_handler
= proc_dointvec_userhz_jiffies
,
2632 .procname
= "proxy_delay",
2633 .maxlen
= sizeof(int),
2635 .proc_handler
= proc_dointvec_userhz_jiffies
,
2638 .procname
= "locktime",
2639 .maxlen
= sizeof(int),
2641 .proc_handler
= proc_dointvec_userhz_jiffies
,
2644 .procname
= "retrans_time_ms",
2645 .maxlen
= sizeof(int),
2647 .proc_handler
= proc_dointvec_ms_jiffies
,
2650 .procname
= "base_reachable_time_ms",
2651 .maxlen
= sizeof(int),
2653 .proc_handler
= proc_dointvec_ms_jiffies
,
2656 .procname
= "gc_interval",
2657 .maxlen
= sizeof(int),
2659 .proc_handler
= proc_dointvec_jiffies
,
2662 .procname
= "gc_thresh1",
2663 .maxlen
= sizeof(int),
2665 .proc_handler
= proc_dointvec
,
2668 .procname
= "gc_thresh2",
2669 .maxlen
= sizeof(int),
2671 .proc_handler
= proc_dointvec
,
2674 .procname
= "gc_thresh3",
2675 .maxlen
= sizeof(int),
2677 .proc_handler
= proc_dointvec
,
2683 int neigh_sysctl_register(struct net_device
*dev
, struct neigh_parms
*p
,
2684 char *p_name
, proc_handler
*handler
)
2686 struct neigh_sysctl_table
*t
;
2687 const char *dev_name_source
= NULL
;
2689 #define NEIGH_CTL_PATH_ROOT 0
2690 #define NEIGH_CTL_PATH_PROTO 1
2691 #define NEIGH_CTL_PATH_NEIGH 2
2692 #define NEIGH_CTL_PATH_DEV 3
2694 struct ctl_path neigh_path
[] = {
2695 { .procname
= "net", },
2696 { .procname
= "proto", },
2697 { .procname
= "neigh", },
2698 { .procname
= "default", },
2702 t
= kmemdup(&neigh_sysctl_template
, sizeof(*t
), GFP_KERNEL
);
2706 t
->neigh_vars
[0].data
= &p
->mcast_probes
;
2707 t
->neigh_vars
[1].data
= &p
->ucast_probes
;
2708 t
->neigh_vars
[2].data
= &p
->app_probes
;
2709 t
->neigh_vars
[3].data
= &p
->retrans_time
;
2710 t
->neigh_vars
[4].data
= &p
->base_reachable_time
;
2711 t
->neigh_vars
[5].data
= &p
->delay_probe_time
;
2712 t
->neigh_vars
[6].data
= &p
->gc_staletime
;
2713 t
->neigh_vars
[7].data
= &p
->queue_len
;
2714 t
->neigh_vars
[8].data
= &p
->proxy_qlen
;
2715 t
->neigh_vars
[9].data
= &p
->anycast_delay
;
2716 t
->neigh_vars
[10].data
= &p
->proxy_delay
;
2717 t
->neigh_vars
[11].data
= &p
->locktime
;
2718 t
->neigh_vars
[12].data
= &p
->retrans_time
;
2719 t
->neigh_vars
[13].data
= &p
->base_reachable_time
;
2722 dev_name_source
= dev
->name
;
2723 /* Terminate the table early */
2724 memset(&t
->neigh_vars
[14], 0, sizeof(t
->neigh_vars
[14]));
2726 dev_name_source
= neigh_path
[NEIGH_CTL_PATH_DEV
].procname
;
2727 t
->neigh_vars
[14].data
= (int *)(p
+ 1);
2728 t
->neigh_vars
[15].data
= (int *)(p
+ 1) + 1;
2729 t
->neigh_vars
[16].data
= (int *)(p
+ 1) + 2;
2730 t
->neigh_vars
[17].data
= (int *)(p
+ 1) + 3;
2736 t
->neigh_vars
[3].proc_handler
= handler
;
2737 t
->neigh_vars
[3].extra1
= dev
;
2739 t
->neigh_vars
[4].proc_handler
= handler
;
2740 t
->neigh_vars
[4].extra1
= dev
;
2741 /* RetransTime (in milliseconds)*/
2742 t
->neigh_vars
[12].proc_handler
= handler
;
2743 t
->neigh_vars
[12].extra1
= dev
;
2744 /* ReachableTime (in milliseconds) */
2745 t
->neigh_vars
[13].proc_handler
= handler
;
2746 t
->neigh_vars
[13].extra1
= dev
;
2749 t
->dev_name
= kstrdup(dev_name_source
, GFP_KERNEL
);
2753 neigh_path
[NEIGH_CTL_PATH_DEV
].procname
= t
->dev_name
;
2754 neigh_path
[NEIGH_CTL_PATH_PROTO
].procname
= p_name
;
2757 register_net_sysctl_table(neigh_parms_net(p
), neigh_path
, t
->neigh_vars
);
2758 if (!t
->sysctl_header
)
2761 p
->sysctl_table
= t
;
2771 EXPORT_SYMBOL(neigh_sysctl_register
);
2773 void neigh_sysctl_unregister(struct neigh_parms
*p
)
2775 if (p
->sysctl_table
) {
2776 struct neigh_sysctl_table
*t
= p
->sysctl_table
;
2777 p
->sysctl_table
= NULL
;
2778 unregister_sysctl_table(t
->sysctl_header
);
2783 EXPORT_SYMBOL(neigh_sysctl_unregister
);
2785 #endif /* CONFIG_SYSCTL */
2787 static int __init
neigh_init(void)
2789 rtnl_register(PF_UNSPEC
, RTM_NEWNEIGH
, neigh_add
, NULL
);
2790 rtnl_register(PF_UNSPEC
, RTM_DELNEIGH
, neigh_delete
, NULL
);
2791 rtnl_register(PF_UNSPEC
, RTM_GETNEIGH
, NULL
, neigh_dump_info
);
2793 rtnl_register(PF_UNSPEC
, RTM_GETNEIGHTBL
, NULL
, neightbl_dump_info
);
2794 rtnl_register(PF_UNSPEC
, RTM_SETNEIGHTBL
, neightbl_set
, NULL
);
2799 subsys_initcall(neigh_init
);