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)
134 struct neigh_hash_table *nht;
136 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
138 write_lock_bh(&tbl->lock);
139 nht = rcu_dereference_protected(tbl->nht,
140 lockdep_is_held(&tbl->lock));
141 for (i = 0; i <= nht->hash_mask; i++) {
142 struct neighbour *n, **np;
144 np = &nht->hash_buckets[i];
145 while ((n = *np) != NULL) {
146 /* Neighbour record may be discarded if:
147 * - nobody refers to it.
148 * - it is not permanent
150 write_lock(&n->lock);
151 if (atomic_read(&n->refcnt) == 1 &&
152 !(n->nud_state & NUD_PERMANENT)) {
156 write_unlock(&n->lock);
157 neigh_cleanup_and_release(n);
160 write_unlock(&n->lock);
165 tbl->last_flush = jiffies;
167 write_unlock_bh(&tbl->lock);
172 static void neigh_add_timer(struct neighbour *n, unsigned long when)
175 if (unlikely(mod_timer(&n->timer, when))) {
176 printk("NEIGH: BUG, double timer add, state is %x\n",
182 static int neigh_del_timer(struct neighbour *n)
184 if ((n->nud_state & NUD_IN_TIMER) &&
185 del_timer(&n->timer)) {
192 static void pneigh_queue_purge(struct sk_buff_head *list)
196 while ((skb = skb_dequeue(list)) != NULL) {
202 static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
205 struct neigh_hash_table *nht;
207 nht = rcu_dereference_protected(tbl->nht,
208 lockdep_is_held(&tbl->lock));
210 for (i = 0; i <= nht->hash_mask; i++) {
211 struct neighbour *n, **np = &nht->hash_buckets[i];
213 while ((n = *np) != NULL) {
214 if (dev && n->dev != dev) {
219 write_lock(&n->lock);
223 if (atomic_read(&n->refcnt) != 1) {
224 /* The most unpleasant situation.
225 We must destroy neighbour entry,
226 but someone still uses it.
228 The destroy will be delayed until
229 the last user releases us, but
230 we must kill timers etc. and move
233 skb_queue_purge(&n->arp_queue);
234 n->output = neigh_blackhole;
235 if (n->nud_state & NUD_VALID)
236 n->nud_state = NUD_NOARP;
238 n->nud_state = NUD_NONE;
239 NEIGH_PRINTK2("neigh %p is stray.\n", n);
241 write_unlock(&n->lock);
242 neigh_cleanup_and_release(n);
247 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
249 write_lock_bh(&tbl->lock);
250 neigh_flush_dev(tbl, dev);
251 write_unlock_bh(&tbl->lock);
253 EXPORT_SYMBOL(neigh_changeaddr);
255 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
257 write_lock_bh(&tbl->lock);
258 neigh_flush_dev(tbl, dev);
259 pneigh_ifdown(tbl, dev);
260 write_unlock_bh(&tbl->lock);
262 del_timer_sync(&tbl->proxy_timer);
263 pneigh_queue_purge(&tbl->proxy_queue);
266 EXPORT_SYMBOL(neigh_ifdown);
268 static struct neighbour *neigh_alloc(struct neigh_table *tbl)
270 struct neighbour *n = NULL;
271 unsigned long now = jiffies;
274 entries = atomic_inc_return(&tbl->entries) - 1;
275 if (entries >= tbl->gc_thresh3 ||
276 (entries >= tbl->gc_thresh2 &&
277 time_after(now, tbl->last_flush + 5 * HZ))) {
278 if (!neigh_forced_gc(tbl) &&
279 entries >= tbl->gc_thresh3)
283 n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
287 skb_queue_head_init(&n->arp_queue);
288 rwlock_init(&n->lock);
289 n->updated = n->used = now;
290 n->nud_state = NUD_NONE;
291 n->output = neigh_blackhole;
292 n->parms = neigh_parms_clone(&tbl->parms);
293 setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
295 NEIGH_CACHE_STAT_INC(tbl, allocs);
297 atomic_set(&n->refcnt, 1);
303 atomic_dec(&tbl->entries);
307 static struct neigh_hash_table *neigh_hash_alloc(unsigned int entries)
309 size_t size = entries * sizeof(struct neighbour *);
310 struct neigh_hash_table *ret;
311 struct neighbour **buckets;
313 ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
316 if (size <= PAGE_SIZE)
317 buckets = kzalloc(size, GFP_ATOMIC);
319 buckets = (struct neighbour **)
320 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
326 ret->hash_buckets = buckets;
327 ret->hash_mask = entries - 1;
328 get_random_bytes(&ret->hash_rnd, sizeof(ret->hash_rnd));
332 static void neigh_hash_free_rcu(struct rcu_head *head)
334 struct neigh_hash_table *nht = container_of(head,
335 struct neigh_hash_table,
337 size_t size = (nht->hash_mask + 1) * sizeof(struct neighbour *);
338 struct neighbour **buckets = nht->hash_buckets;
340 if (size <= PAGE_SIZE)
343 free_pages((unsigned long)buckets, get_order(size));
347 static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
348 unsigned long new_entries)
350 unsigned int i, hash;
351 struct neigh_hash_table *new_nht, *old_nht;
353 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
355 BUG_ON(!is_power_of_2(new_entries));
356 old_nht = rcu_dereference_protected(tbl->nht,
357 lockdep_is_held(&tbl->lock));
358 new_nht = neigh_hash_alloc(new_entries);
362 for (i = 0; i <= old_nht->hash_mask; i++) {
363 struct neighbour *n, *next;
365 for (n = old_nht->hash_buckets[i];
368 hash = tbl->hash(n->primary_key, n->dev,
371 hash &= new_nht->hash_mask;
374 n->next = new_nht->hash_buckets[hash];
375 new_nht->hash_buckets[hash] = n;
379 rcu_assign_pointer(tbl->nht, new_nht);
380 call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
384 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
385 struct net_device *dev)
388 int key_len = tbl->key_len;
390 struct neigh_hash_table *nht;
392 NEIGH_CACHE_STAT_INC(tbl, lookups);
395 nht = rcu_dereference_bh(tbl->nht);
396 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) & nht->hash_mask;
397 read_lock(&tbl->lock);
398 for (n = nht->hash_buckets[hash_val]; n; n = n->next) {
399 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
401 NEIGH_CACHE_STAT_INC(tbl, hits);
405 read_unlock(&tbl->lock);
406 rcu_read_unlock_bh();
409 EXPORT_SYMBOL(neigh_lookup);
411 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
415 int key_len = tbl->key_len;
417 struct neigh_hash_table *nht;
419 NEIGH_CACHE_STAT_INC(tbl, lookups);
422 nht = rcu_dereference_bh(tbl->nht);
423 hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) & nht->hash_mask;
424 read_lock(&tbl->lock);
425 for (n = nht->hash_buckets[hash_val]; n; n = n->next) {
426 if (!memcmp(n->primary_key, pkey, key_len) &&
427 net_eq(dev_net(n->dev), net)) {
429 NEIGH_CACHE_STAT_INC(tbl, hits);
433 read_unlock(&tbl->lock);
434 rcu_read_unlock_bh();
437 EXPORT_SYMBOL(neigh_lookup_nodev);
439 struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
440 struct net_device *dev)
443 int key_len = tbl->key_len;
445 struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
446 struct neigh_hash_table *nht;
449 rc = ERR_PTR(-ENOBUFS);
453 memcpy(n->primary_key, pkey, key_len);
457 /* Protocol specific setup. */
458 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
460 goto out_neigh_release;
463 /* Device specific setup. */
464 if (n->parms->neigh_setup &&
465 (error = n->parms->neigh_setup(n)) < 0) {
467 goto out_neigh_release;
470 n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
472 write_lock_bh(&tbl->lock);
473 nht = rcu_dereference_protected(tbl->nht,
474 lockdep_is_held(&tbl->lock));
476 if (atomic_read(&tbl->entries) > (nht->hash_mask + 1))
477 nht = neigh_hash_grow(tbl, (nht->hash_mask + 1) << 1);
479 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) & nht->hash_mask;
481 if (n->parms->dead) {
482 rc = ERR_PTR(-EINVAL);
486 for (n1 = nht->hash_buckets[hash_val]; n1; n1 = n1->next) {
487 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
494 n->next = nht->hash_buckets[hash_val];
495 nht->hash_buckets[hash_val] = n;
498 write_unlock_bh(&tbl->lock);
499 NEIGH_PRINTK2("neigh %p is created.\n", n);
504 write_unlock_bh(&tbl->lock);
509 EXPORT_SYMBOL(neigh_create);
511 static u32 pneigh_hash(const void *pkey, int key_len)
513 u32 hash_val = *(u32 *)(pkey + key_len - 4);
514 hash_val ^= (hash_val >> 16);
515 hash_val ^= hash_val >> 8;
516 hash_val ^= hash_val >> 4;
517 hash_val &= PNEIGH_HASHMASK;
521 static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
525 struct net_device *dev)
528 if (!memcmp(n->key, pkey, key_len) &&
529 net_eq(pneigh_net(n), net) &&
530 (n->dev == dev || !n->dev))
537 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
538 struct net *net, const void *pkey, struct net_device *dev)
540 int key_len = tbl->key_len;
541 u32 hash_val = pneigh_hash(pkey, key_len);
543 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
544 net, pkey, key_len, dev);
546 EXPORT_SYMBOL_GPL(__pneigh_lookup);
548 struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
549 struct net *net, const void *pkey,
550 struct net_device *dev, int creat)
552 struct pneigh_entry *n;
553 int key_len = tbl->key_len;
554 u32 hash_val = pneigh_hash(pkey, key_len);
556 read_lock_bh(&tbl->lock);
557 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
558 net, pkey, key_len, dev);
559 read_unlock_bh(&tbl->lock);
566 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
570 write_pnet(&n->net, hold_net(net));
571 memcpy(n->key, pkey, key_len);
576 if (tbl->pconstructor && tbl->pconstructor(n)) {
585 write_lock_bh(&tbl->lock);
586 n->next = tbl->phash_buckets[hash_val];
587 tbl->phash_buckets[hash_val] = n;
588 write_unlock_bh(&tbl->lock);
592 EXPORT_SYMBOL(pneigh_lookup);
595 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
596 struct net_device *dev)
598 struct pneigh_entry *n, **np;
599 int key_len = tbl->key_len;
600 u32 hash_val = pneigh_hash(pkey, key_len);
602 write_lock_bh(&tbl->lock);
603 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
605 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
606 net_eq(pneigh_net(n), net)) {
608 write_unlock_bh(&tbl->lock);
609 if (tbl->pdestructor)
613 release_net(pneigh_net(n));
618 write_unlock_bh(&tbl->lock);
622 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
624 struct pneigh_entry *n, **np;
627 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
628 np = &tbl->phash_buckets[h];
629 while ((n = *np) != NULL) {
630 if (!dev || n->dev == dev) {
632 if (tbl->pdestructor)
636 release_net(pneigh_net(n));
646 static void neigh_parms_destroy(struct neigh_parms *parms);
648 static inline void neigh_parms_put(struct neigh_parms *parms)
650 if (atomic_dec_and_test(&parms->refcnt))
651 neigh_parms_destroy(parms);
655 * neighbour must already be out of the table;
658 void neigh_destroy(struct neighbour *neigh)
662 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
666 "Destroying alive neighbour %p\n", neigh);
671 if (neigh_del_timer(neigh))
672 printk(KERN_WARNING "Impossible event.\n");
674 while ((hh = neigh->hh) != NULL) {
675 neigh->hh = hh->hh_next;
678 write_seqlock_bh(&hh->hh_lock);
679 hh->hh_output = neigh_blackhole;
680 write_sequnlock_bh(&hh->hh_lock);
681 if (atomic_dec_and_test(&hh->hh_refcnt))
685 skb_queue_purge(&neigh->arp_queue);
688 neigh_parms_put(neigh->parms);
690 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
692 atomic_dec(&neigh->tbl->entries);
693 kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
695 EXPORT_SYMBOL(neigh_destroy);
697 /* Neighbour state is suspicious;
700 Called with write_locked neigh.
702 static void neigh_suspect(struct neighbour *neigh)
706 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
708 neigh->output = neigh->ops->output;
710 for (hh = neigh->hh; hh; hh = hh->hh_next)
711 hh->hh_output = neigh->ops->output;
714 /* Neighbour state is OK;
717 Called with write_locked neigh.
719 static void neigh_connect(struct neighbour *neigh)
723 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
725 neigh->output = neigh->ops->connected_output;
727 for (hh = neigh->hh; hh; hh = hh->hh_next)
728 hh->hh_output = neigh->ops->hh_output;
731 static void neigh_periodic_work(struct work_struct *work)
733 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
734 struct neighbour *n, **np;
736 struct neigh_hash_table *nht;
738 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
740 write_lock_bh(&tbl->lock);
741 nht = rcu_dereference_protected(tbl->nht,
742 lockdep_is_held(&tbl->lock));
745 * periodically recompute ReachableTime from random function
748 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
749 struct neigh_parms *p;
750 tbl->last_rand = jiffies;
751 for (p = &tbl->parms; p; p = p->next)
753 neigh_rand_reach_time(p->base_reachable_time);
756 for (i = 0 ; i <= nht->hash_mask; i++) {
757 np = &nht->hash_buckets[i];
759 while ((n = *np) != NULL) {
762 write_lock(&n->lock);
764 state = n->nud_state;
765 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
766 write_unlock(&n->lock);
770 if (time_before(n->used, n->confirmed))
771 n->used = n->confirmed;
773 if (atomic_read(&n->refcnt) == 1 &&
774 (state == NUD_FAILED ||
775 time_after(jiffies, n->used + n->parms->gc_staletime))) {
778 write_unlock(&n->lock);
779 neigh_cleanup_and_release(n);
782 write_unlock(&n->lock);
788 * It's fine to release lock here, even if hash table
789 * grows while we are preempted.
791 write_unlock_bh(&tbl->lock);
793 write_lock_bh(&tbl->lock);
795 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
796 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
797 * base_reachable_time.
799 schedule_delayed_work(&tbl->gc_work,
800 tbl->parms.base_reachable_time >> 1);
801 write_unlock_bh(&tbl->lock);
804 static __inline__ int neigh_max_probes(struct neighbour *n)
806 struct neigh_parms *p = n->parms;
807 return (n->nud_state & NUD_PROBE) ?
809 p->ucast_probes + p->app_probes + p->mcast_probes;
812 static void neigh_invalidate(struct neighbour *neigh)
813 __releases(neigh->lock)
814 __acquires(neigh->lock)
818 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
819 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
820 neigh->updated = jiffies;
822 /* It is very thin place. report_unreachable is very complicated
823 routine. Particularly, it can hit the same neighbour entry!
825 So that, we try to be accurate and avoid dead loop. --ANK
827 while (neigh->nud_state == NUD_FAILED &&
828 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
829 write_unlock(&neigh->lock);
830 neigh->ops->error_report(neigh, skb);
831 write_lock(&neigh->lock);
833 skb_queue_purge(&neigh->arp_queue);
836 /* Called when a timer expires for a neighbour entry. */
838 static void neigh_timer_handler(unsigned long arg)
840 unsigned long now, next;
841 struct neighbour *neigh = (struct neighbour *)arg;
845 write_lock(&neigh->lock);
847 state = neigh->nud_state;
851 if (!(state & NUD_IN_TIMER)) {
853 printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
858 if (state & NUD_REACHABLE) {
859 if (time_before_eq(now,
860 neigh->confirmed + neigh->parms->reachable_time)) {
861 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
862 next = neigh->confirmed + neigh->parms->reachable_time;
863 } else if (time_before_eq(now,
864 neigh->used + neigh->parms->delay_probe_time)) {
865 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
866 neigh->nud_state = NUD_DELAY;
867 neigh->updated = jiffies;
868 neigh_suspect(neigh);
869 next = now + neigh->parms->delay_probe_time;
871 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
872 neigh->nud_state = NUD_STALE;
873 neigh->updated = jiffies;
874 neigh_suspect(neigh);
877 } else if (state & NUD_DELAY) {
878 if (time_before_eq(now,
879 neigh->confirmed + neigh->parms->delay_probe_time)) {
880 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
881 neigh->nud_state = NUD_REACHABLE;
882 neigh->updated = jiffies;
883 neigh_connect(neigh);
885 next = neigh->confirmed + neigh->parms->reachable_time;
887 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
888 neigh->nud_state = NUD_PROBE;
889 neigh->updated = jiffies;
890 atomic_set(&neigh->probes, 0);
891 next = now + neigh->parms->retrans_time;
894 /* NUD_PROBE|NUD_INCOMPLETE */
895 next = now + neigh->parms->retrans_time;
898 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
899 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
900 neigh->nud_state = NUD_FAILED;
902 neigh_invalidate(neigh);
905 if (neigh->nud_state & NUD_IN_TIMER) {
906 if (time_before(next, jiffies + HZ/2))
907 next = jiffies + HZ/2;
908 if (!mod_timer(&neigh->timer, next))
911 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
912 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
913 /* keep skb alive even if arp_queue overflows */
915 skb = skb_copy(skb, GFP_ATOMIC);
916 write_unlock(&neigh->lock);
917 neigh->ops->solicit(neigh, skb);
918 atomic_inc(&neigh->probes);
922 write_unlock(&neigh->lock);
926 neigh_update_notify(neigh);
928 neigh_release(neigh);
931 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
936 write_lock_bh(&neigh->lock);
939 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
944 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
945 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
946 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
947 neigh->nud_state = NUD_INCOMPLETE;
948 neigh->updated = jiffies;
949 neigh_add_timer(neigh, now + 1);
951 neigh->nud_state = NUD_FAILED;
952 neigh->updated = jiffies;
953 write_unlock_bh(&neigh->lock);
958 } else if (neigh->nud_state & NUD_STALE) {
959 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
960 neigh->nud_state = NUD_DELAY;
961 neigh->updated = jiffies;
962 neigh_add_timer(neigh,
963 jiffies + neigh->parms->delay_probe_time);
966 if (neigh->nud_state == NUD_INCOMPLETE) {
968 if (skb_queue_len(&neigh->arp_queue) >=
969 neigh->parms->queue_len) {
970 struct sk_buff *buff;
971 buff = __skb_dequeue(&neigh->arp_queue);
973 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
976 __skb_queue_tail(&neigh->arp_queue, skb);
981 write_unlock_bh(&neigh->lock);
984 EXPORT_SYMBOL(__neigh_event_send);
986 static void neigh_update_hhs(struct neighbour *neigh)
989 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
992 if (neigh->dev->header_ops)
993 update = neigh->dev->header_ops->cache_update;
996 for (hh = neigh->hh; hh; hh = hh->hh_next) {
997 write_seqlock_bh(&hh->hh_lock);
998 update(hh, neigh->dev, neigh->ha);
999 write_sequnlock_bh(&hh->hh_lock);
1006 /* Generic update routine.
1007 -- lladdr is new lladdr or NULL, if it is not supplied.
1008 -- new is new state.
1010 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1012 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1013 lladdr instead of overriding it
1015 It also allows to retain current state
1016 if lladdr is unchanged.
1017 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1019 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1021 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1024 Caller MUST hold reference count on the entry.
1027 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1033 struct net_device *dev;
1034 int update_isrouter = 0;
1036 write_lock_bh(&neigh->lock);
1039 old = neigh->nud_state;
1042 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1043 (old & (NUD_NOARP | NUD_PERMANENT)))
1046 if (!(new & NUD_VALID)) {
1047 neigh_del_timer(neigh);
1048 if (old & NUD_CONNECTED)
1049 neigh_suspect(neigh);
1050 neigh->nud_state = new;
1052 notify = old & NUD_VALID;
1053 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1054 (new & NUD_FAILED)) {
1055 neigh_invalidate(neigh);
1061 /* Compare new lladdr with cached one */
1062 if (!dev->addr_len) {
1063 /* First case: device needs no address. */
1065 } else if (lladdr) {
1066 /* The second case: if something is already cached
1067 and a new address is proposed:
1069 - if they are different, check override flag
1071 if ((old & NUD_VALID) &&
1072 !memcmp(lladdr, neigh->ha, dev->addr_len))
1075 /* No address is supplied; if we know something,
1076 use it, otherwise discard the request.
1079 if (!(old & NUD_VALID))
1084 if (new & NUD_CONNECTED)
1085 neigh->confirmed = jiffies;
1086 neigh->updated = jiffies;
1088 /* If entry was valid and address is not changed,
1089 do not change entry state, if new one is STALE.
1092 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1093 if (old & NUD_VALID) {
1094 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1095 update_isrouter = 0;
1096 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1097 (old & NUD_CONNECTED)) {
1103 if (lladdr == neigh->ha && new == NUD_STALE &&
1104 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1105 (old & NUD_CONNECTED))
1112 neigh_del_timer(neigh);
1113 if (new & NUD_IN_TIMER)
1114 neigh_add_timer(neigh, (jiffies +
1115 ((new & NUD_REACHABLE) ?
1116 neigh->parms->reachable_time :
1118 neigh->nud_state = new;
1121 if (lladdr != neigh->ha) {
1122 memcpy(&neigh->ha, lladdr, dev->addr_len);
1123 neigh_update_hhs(neigh);
1124 if (!(new & NUD_CONNECTED))
1125 neigh->confirmed = jiffies -
1126 (neigh->parms->base_reachable_time << 1);
1131 if (new & NUD_CONNECTED)
1132 neigh_connect(neigh);
1134 neigh_suspect(neigh);
1135 if (!(old & NUD_VALID)) {
1136 struct sk_buff *skb;
1138 /* Again: avoid dead loop if something went wrong */
1140 while (neigh->nud_state & NUD_VALID &&
1141 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1142 struct neighbour *n1 = neigh;
1143 write_unlock_bh(&neigh->lock);
1144 /* On shaper/eql skb->dst->neighbour != neigh :( */
1145 if (skb_dst(skb) && skb_dst(skb)->neighbour)
1146 n1 = skb_dst(skb)->neighbour;
1148 write_lock_bh(&neigh->lock);
1150 skb_queue_purge(&neigh->arp_queue);
1153 if (update_isrouter) {
1154 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1155 (neigh->flags | NTF_ROUTER) :
1156 (neigh->flags & ~NTF_ROUTER);
1158 write_unlock_bh(&neigh->lock);
1161 neigh_update_notify(neigh);
1165 EXPORT_SYMBOL(neigh_update);
1167 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1168 u8 *lladdr, void *saddr,
1169 struct net_device *dev)
1171 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1172 lladdr || !dev->addr_len);
1174 neigh_update(neigh, lladdr, NUD_STALE,
1175 NEIGH_UPDATE_F_OVERRIDE);
1178 EXPORT_SYMBOL(neigh_event_ns);
1180 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst,
1183 struct hh_cache *hh;
1184 struct net_device *dev = dst->dev;
1186 for (hh = n->hh; hh; hh = hh->hh_next)
1187 if (hh->hh_type == protocol)
1190 if (!hh && (hh = kzalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) {
1191 seqlock_init(&hh->hh_lock);
1192 hh->hh_type = protocol;
1193 atomic_set(&hh->hh_refcnt, 0);
1196 if (dev->header_ops->cache(n, hh)) {
1200 atomic_inc(&hh->hh_refcnt);
1201 hh->hh_next = n->hh;
1203 if (n->nud_state & NUD_CONNECTED)
1204 hh->hh_output = n->ops->hh_output;
1206 hh->hh_output = n->ops->output;
1210 atomic_inc(&hh->hh_refcnt);
1215 /* This function can be used in contexts, where only old dev_queue_xmit
1216 worked, f.e. if you want to override normal output path (eql, shaper),
1217 but resolution is not made yet.
1220 int neigh_compat_output(struct sk_buff *skb)
1222 struct net_device *dev = skb->dev;
1224 __skb_pull(skb, skb_network_offset(skb));
1226 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1228 dev->header_ops->rebuild(skb))
1231 return dev_queue_xmit(skb);
1233 EXPORT_SYMBOL(neigh_compat_output);
1235 /* Slow and careful. */
1237 int neigh_resolve_output(struct sk_buff *skb)
1239 struct dst_entry *dst = skb_dst(skb);
1240 struct neighbour *neigh;
1243 if (!dst || !(neigh = dst->neighbour))
1246 __skb_pull(skb, skb_network_offset(skb));
1248 if (!neigh_event_send(neigh, skb)) {
1250 struct net_device *dev = neigh->dev;
1251 if (dev->header_ops->cache &&
1253 !(dst->flags & DST_NOCACHE)) {
1254 write_lock_bh(&neigh->lock);
1256 neigh_hh_init(neigh, dst, dst->ops->protocol);
1257 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1258 neigh->ha, NULL, skb->len);
1259 write_unlock_bh(&neigh->lock);
1261 read_lock_bh(&neigh->lock);
1262 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1263 neigh->ha, NULL, skb->len);
1264 read_unlock_bh(&neigh->lock);
1267 rc = neigh->ops->queue_xmit(skb);
1274 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1275 dst, dst ? dst->neighbour : NULL);
1281 EXPORT_SYMBOL(neigh_resolve_output);
1283 /* As fast as possible without hh cache */
1285 int neigh_connected_output(struct sk_buff *skb)
1288 struct dst_entry *dst = skb_dst(skb);
1289 struct neighbour *neigh = dst->neighbour;
1290 struct net_device *dev = neigh->dev;
1292 __skb_pull(skb, skb_network_offset(skb));
1294 read_lock_bh(&neigh->lock);
1295 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1296 neigh->ha, NULL, skb->len);
1297 read_unlock_bh(&neigh->lock);
1299 err = neigh->ops->queue_xmit(skb);
1306 EXPORT_SYMBOL(neigh_connected_output);
1308 static void neigh_proxy_process(unsigned long arg)
1310 struct neigh_table *tbl = (struct neigh_table *)arg;
1311 long sched_next = 0;
1312 unsigned long now = jiffies;
1313 struct sk_buff *skb, *n;
1315 spin_lock(&tbl->proxy_queue.lock);
1317 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1318 long tdif = NEIGH_CB(skb)->sched_next - now;
1321 struct net_device *dev = skb->dev;
1322 __skb_unlink(skb, &tbl->proxy_queue);
1323 if (tbl->proxy_redo && netif_running(dev))
1324 tbl->proxy_redo(skb);
1329 } else if (!sched_next || tdif < sched_next)
1332 del_timer(&tbl->proxy_timer);
1334 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1335 spin_unlock(&tbl->proxy_queue.lock);
1338 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1339 struct sk_buff *skb)
1341 unsigned long now = jiffies;
1342 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1344 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1349 NEIGH_CB(skb)->sched_next = sched_next;
1350 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1352 spin_lock(&tbl->proxy_queue.lock);
1353 if (del_timer(&tbl->proxy_timer)) {
1354 if (time_before(tbl->proxy_timer.expires, sched_next))
1355 sched_next = tbl->proxy_timer.expires;
1359 __skb_queue_tail(&tbl->proxy_queue, skb);
1360 mod_timer(&tbl->proxy_timer, sched_next);
1361 spin_unlock(&tbl->proxy_queue.lock);
1363 EXPORT_SYMBOL(pneigh_enqueue);
1365 static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1366 struct net *net, int ifindex)
1368 struct neigh_parms *p;
1370 for (p = &tbl->parms; p; p = p->next) {
1371 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1372 (!p->dev && !ifindex))
1379 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1380 struct neigh_table *tbl)
1382 struct neigh_parms *p, *ref;
1383 struct net *net = dev_net(dev);
1384 const struct net_device_ops *ops = dev->netdev_ops;
1386 ref = lookup_neigh_parms(tbl, net, 0);
1390 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1393 atomic_set(&p->refcnt, 1);
1395 neigh_rand_reach_time(p->base_reachable_time);
1397 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1404 write_pnet(&p->net, hold_net(net));
1405 p->sysctl_table = NULL;
1406 write_lock_bh(&tbl->lock);
1407 p->next = tbl->parms.next;
1408 tbl->parms.next = p;
1409 write_unlock_bh(&tbl->lock);
1413 EXPORT_SYMBOL(neigh_parms_alloc);
1415 static void neigh_rcu_free_parms(struct rcu_head *head)
1417 struct neigh_parms *parms =
1418 container_of(head, struct neigh_parms, rcu_head);
1420 neigh_parms_put(parms);
1423 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1425 struct neigh_parms **p;
1427 if (!parms || parms == &tbl->parms)
1429 write_lock_bh(&tbl->lock);
1430 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1434 write_unlock_bh(&tbl->lock);
1436 dev_put(parms->dev);
1437 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1441 write_unlock_bh(&tbl->lock);
1442 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1444 EXPORT_SYMBOL(neigh_parms_release);
1446 static void neigh_parms_destroy(struct neigh_parms *parms)
1448 release_net(neigh_parms_net(parms));
1452 static struct lock_class_key neigh_table_proxy_queue_class;
1454 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1456 unsigned long now = jiffies;
1457 unsigned long phsize;
1459 write_pnet(&tbl->parms.net, &init_net);
1460 atomic_set(&tbl->parms.refcnt, 1);
1461 tbl->parms.reachable_time =
1462 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1464 if (!tbl->kmem_cachep)
1466 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1467 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1469 tbl->stats = alloc_percpu(struct neigh_statistics);
1471 panic("cannot create neighbour cache statistics");
1473 #ifdef CONFIG_PROC_FS
1474 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1475 &neigh_stat_seq_fops, tbl))
1476 panic("cannot create neighbour proc dir entry");
1479 tbl->nht = neigh_hash_alloc(8);
1481 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1482 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1484 if (!tbl->nht || !tbl->phash_buckets)
1485 panic("cannot allocate neighbour cache hashes");
1487 rwlock_init(&tbl->lock);
1488 INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1489 schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1490 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1491 skb_queue_head_init_class(&tbl->proxy_queue,
1492 &neigh_table_proxy_queue_class);
1494 tbl->last_flush = now;
1495 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1497 EXPORT_SYMBOL(neigh_table_init_no_netlink);
1499 void neigh_table_init(struct neigh_table *tbl)
1501 struct neigh_table *tmp;
1503 neigh_table_init_no_netlink(tbl);
1504 write_lock(&neigh_tbl_lock);
1505 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1506 if (tmp->family == tbl->family)
1509 tbl->next = neigh_tables;
1511 write_unlock(&neigh_tbl_lock);
1513 if (unlikely(tmp)) {
1514 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1515 "family %d\n", tbl->family);
1519 EXPORT_SYMBOL(neigh_table_init);
1521 int neigh_table_clear(struct neigh_table *tbl)
1523 struct neigh_table **tp;
1525 /* It is not clean... Fix it to unload IPv6 module safely */
1526 cancel_delayed_work(&tbl->gc_work);
1527 flush_scheduled_work();
1528 del_timer_sync(&tbl->proxy_timer);
1529 pneigh_queue_purge(&tbl->proxy_queue);
1530 neigh_ifdown(tbl, NULL);
1531 if (atomic_read(&tbl->entries))
1532 printk(KERN_CRIT "neighbour leakage\n");
1533 write_lock(&neigh_tbl_lock);
1534 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1540 write_unlock(&neigh_tbl_lock);
1542 call_rcu(&tbl->nht->rcu, neigh_hash_free_rcu);
1545 kfree(tbl->phash_buckets);
1546 tbl->phash_buckets = NULL;
1548 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1550 free_percpu(tbl->stats);
1553 kmem_cache_destroy(tbl->kmem_cachep);
1554 tbl->kmem_cachep = NULL;
1558 EXPORT_SYMBOL(neigh_table_clear);
1560 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1562 struct net *net = sock_net(skb->sk);
1564 struct nlattr *dst_attr;
1565 struct neigh_table *tbl;
1566 struct net_device *dev = NULL;
1570 if (nlmsg_len(nlh) < sizeof(*ndm))
1573 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1574 if (dst_attr == NULL)
1577 ndm = nlmsg_data(nlh);
1578 if (ndm->ndm_ifindex) {
1579 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1586 read_lock(&neigh_tbl_lock);
1587 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1588 struct neighbour *neigh;
1590 if (tbl->family != ndm->ndm_family)
1592 read_unlock(&neigh_tbl_lock);
1594 if (nla_len(dst_attr) < tbl->key_len)
1597 if (ndm->ndm_flags & NTF_PROXY) {
1598 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1605 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1606 if (neigh == NULL) {
1611 err = neigh_update(neigh, NULL, NUD_FAILED,
1612 NEIGH_UPDATE_F_OVERRIDE |
1613 NEIGH_UPDATE_F_ADMIN);
1614 neigh_release(neigh);
1617 read_unlock(&neigh_tbl_lock);
1618 err = -EAFNOSUPPORT;
1624 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1626 struct net *net = sock_net(skb->sk);
1628 struct nlattr *tb[NDA_MAX+1];
1629 struct neigh_table *tbl;
1630 struct net_device *dev = NULL;
1634 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1639 if (tb[NDA_DST] == NULL)
1642 ndm = nlmsg_data(nlh);
1643 if (ndm->ndm_ifindex) {
1644 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1650 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1654 read_lock(&neigh_tbl_lock);
1655 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1656 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1657 struct neighbour *neigh;
1660 if (tbl->family != ndm->ndm_family)
1662 read_unlock(&neigh_tbl_lock);
1664 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1666 dst = nla_data(tb[NDA_DST]);
1667 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1669 if (ndm->ndm_flags & NTF_PROXY) {
1670 struct pneigh_entry *pn;
1673 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1675 pn->flags = ndm->ndm_flags;
1684 neigh = neigh_lookup(tbl, dst, dev);
1685 if (neigh == NULL) {
1686 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1691 neigh = __neigh_lookup_errno(tbl, dst, dev);
1692 if (IS_ERR(neigh)) {
1693 err = PTR_ERR(neigh);
1697 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1699 neigh_release(neigh);
1703 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1704 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1707 if (ndm->ndm_flags & NTF_USE) {
1708 neigh_event_send(neigh, NULL);
1711 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1712 neigh_release(neigh);
1716 read_unlock(&neigh_tbl_lock);
1717 err = -EAFNOSUPPORT;
1722 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1724 struct nlattr *nest;
1726 nest = nla_nest_start(skb, NDTA_PARMS);
1731 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1733 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1734 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1735 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1736 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1737 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1738 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1739 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1740 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1741 parms->base_reachable_time);
1742 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1743 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1744 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1745 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1746 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1747 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1749 return nla_nest_end(skb, nest);
1752 nla_nest_cancel(skb, nest);
1756 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1757 u32 pid, u32 seq, int type, int flags)
1759 struct nlmsghdr *nlh;
1760 struct ndtmsg *ndtmsg;
1762 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1766 ndtmsg = nlmsg_data(nlh);
1768 read_lock_bh(&tbl->lock);
1769 ndtmsg->ndtm_family = tbl->family;
1770 ndtmsg->ndtm_pad1 = 0;
1771 ndtmsg->ndtm_pad2 = 0;
1773 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1774 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1775 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1776 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1777 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1780 unsigned long now = jiffies;
1781 unsigned int flush_delta = now - tbl->last_flush;
1782 unsigned int rand_delta = now - tbl->last_rand;
1783 struct neigh_hash_table *nht;
1784 struct ndt_config ndc = {
1785 .ndtc_key_len = tbl->key_len,
1786 .ndtc_entry_size = tbl->entry_size,
1787 .ndtc_entries = atomic_read(&tbl->entries),
1788 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1789 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1790 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1794 nht = rcu_dereference_bh(tbl->nht);
1795 ndc.ndtc_hash_rnd = nht->hash_rnd;
1796 ndc.ndtc_hash_mask = nht->hash_mask;
1797 rcu_read_unlock_bh();
1799 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1804 struct ndt_stats ndst;
1806 memset(&ndst, 0, sizeof(ndst));
1808 for_each_possible_cpu(cpu) {
1809 struct neigh_statistics *st;
1811 st = per_cpu_ptr(tbl->stats, cpu);
1812 ndst.ndts_allocs += st->allocs;
1813 ndst.ndts_destroys += st->destroys;
1814 ndst.ndts_hash_grows += st->hash_grows;
1815 ndst.ndts_res_failed += st->res_failed;
1816 ndst.ndts_lookups += st->lookups;
1817 ndst.ndts_hits += st->hits;
1818 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1819 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1820 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1821 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1824 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1827 BUG_ON(tbl->parms.dev);
1828 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1829 goto nla_put_failure;
1831 read_unlock_bh(&tbl->lock);
1832 return nlmsg_end(skb, nlh);
1835 read_unlock_bh(&tbl->lock);
1836 nlmsg_cancel(skb, nlh);
1840 static int neightbl_fill_param_info(struct sk_buff *skb,
1841 struct neigh_table *tbl,
1842 struct neigh_parms *parms,
1843 u32 pid, u32 seq, int type,
1846 struct ndtmsg *ndtmsg;
1847 struct nlmsghdr *nlh;
1849 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1853 ndtmsg = nlmsg_data(nlh);
1855 read_lock_bh(&tbl->lock);
1856 ndtmsg->ndtm_family = tbl->family;
1857 ndtmsg->ndtm_pad1 = 0;
1858 ndtmsg->ndtm_pad2 = 0;
1860 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1861 neightbl_fill_parms(skb, parms) < 0)
1864 read_unlock_bh(&tbl->lock);
1865 return nlmsg_end(skb, nlh);
1867 read_unlock_bh(&tbl->lock);
1868 nlmsg_cancel(skb, nlh);
1872 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1873 [NDTA_NAME] = { .type = NLA_STRING },
1874 [NDTA_THRESH1] = { .type = NLA_U32 },
1875 [NDTA_THRESH2] = { .type = NLA_U32 },
1876 [NDTA_THRESH3] = { .type = NLA_U32 },
1877 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1878 [NDTA_PARMS] = { .type = NLA_NESTED },
1881 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1882 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1883 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1884 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1885 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1886 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1887 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1888 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1889 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1890 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1891 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1892 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1893 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1894 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1897 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1899 struct net *net = sock_net(skb->sk);
1900 struct neigh_table *tbl;
1901 struct ndtmsg *ndtmsg;
1902 struct nlattr *tb[NDTA_MAX+1];
1905 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1906 nl_neightbl_policy);
1910 if (tb[NDTA_NAME] == NULL) {
1915 ndtmsg = nlmsg_data(nlh);
1916 read_lock(&neigh_tbl_lock);
1917 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1918 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1921 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1931 * We acquire tbl->lock to be nice to the periodic timers and
1932 * make sure they always see a consistent set of values.
1934 write_lock_bh(&tbl->lock);
1936 if (tb[NDTA_PARMS]) {
1937 struct nlattr *tbp[NDTPA_MAX+1];
1938 struct neigh_parms *p;
1941 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1942 nl_ntbl_parm_policy);
1944 goto errout_tbl_lock;
1946 if (tbp[NDTPA_IFINDEX])
1947 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1949 p = lookup_neigh_parms(tbl, net, ifindex);
1952 goto errout_tbl_lock;
1955 for (i = 1; i <= NDTPA_MAX; i++) {
1960 case NDTPA_QUEUE_LEN:
1961 p->queue_len = nla_get_u32(tbp[i]);
1963 case NDTPA_PROXY_QLEN:
1964 p->proxy_qlen = nla_get_u32(tbp[i]);
1966 case NDTPA_APP_PROBES:
1967 p->app_probes = nla_get_u32(tbp[i]);
1969 case NDTPA_UCAST_PROBES:
1970 p->ucast_probes = nla_get_u32(tbp[i]);
1972 case NDTPA_MCAST_PROBES:
1973 p->mcast_probes = nla_get_u32(tbp[i]);
1975 case NDTPA_BASE_REACHABLE_TIME:
1976 p->base_reachable_time = nla_get_msecs(tbp[i]);
1978 case NDTPA_GC_STALETIME:
1979 p->gc_staletime = nla_get_msecs(tbp[i]);
1981 case NDTPA_DELAY_PROBE_TIME:
1982 p->delay_probe_time = nla_get_msecs(tbp[i]);
1984 case NDTPA_RETRANS_TIME:
1985 p->retrans_time = nla_get_msecs(tbp[i]);
1987 case NDTPA_ANYCAST_DELAY:
1988 p->anycast_delay = nla_get_msecs(tbp[i]);
1990 case NDTPA_PROXY_DELAY:
1991 p->proxy_delay = nla_get_msecs(tbp[i]);
1993 case NDTPA_LOCKTIME:
1994 p->locktime = nla_get_msecs(tbp[i]);
2000 if (tb[NDTA_THRESH1])
2001 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2003 if (tb[NDTA_THRESH2])
2004 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2006 if (tb[NDTA_THRESH3])
2007 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2009 if (tb[NDTA_GC_INTERVAL])
2010 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2015 write_unlock_bh(&tbl->lock);
2017 read_unlock(&neigh_tbl_lock);
2022 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2024 struct net *net = sock_net(skb->sk);
2025 int family, tidx, nidx = 0;
2026 int tbl_skip = cb->args[0];
2027 int neigh_skip = cb->args[1];
2028 struct neigh_table *tbl;
2030 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2032 read_lock(&neigh_tbl_lock);
2033 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
2034 struct neigh_parms *p;
2036 if (tidx < tbl_skip || (family && tbl->family != family))
2039 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2040 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2044 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
2045 if (!net_eq(neigh_parms_net(p), net))
2048 if (nidx < neigh_skip)
2051 if (neightbl_fill_param_info(skb, tbl, p,
2052 NETLINK_CB(cb->skb).pid,
2064 read_unlock(&neigh_tbl_lock);
2071 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2072 u32 pid, u32 seq, int type, unsigned int flags)
2074 unsigned long now = jiffies;
2075 struct nda_cacheinfo ci;
2076 struct nlmsghdr *nlh;
2079 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2083 ndm = nlmsg_data(nlh);
2084 ndm->ndm_family = neigh->ops->family;
2087 ndm->ndm_flags = neigh->flags;
2088 ndm->ndm_type = neigh->type;
2089 ndm->ndm_ifindex = neigh->dev->ifindex;
2091 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2093 read_lock_bh(&neigh->lock);
2094 ndm->ndm_state = neigh->nud_state;
2095 if ((neigh->nud_state & NUD_VALID) &&
2096 nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, neigh->ha) < 0) {
2097 read_unlock_bh(&neigh->lock);
2098 goto nla_put_failure;
2101 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2102 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2103 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2104 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2105 read_unlock_bh(&neigh->lock);
2107 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2108 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2110 return nlmsg_end(skb, nlh);
2113 nlmsg_cancel(skb, nlh);
2117 static void neigh_update_notify(struct neighbour *neigh)
2119 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2120 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2123 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2124 struct netlink_callback *cb)
2126 struct net * net = sock_net(skb->sk);
2127 struct neighbour *n;
2128 int rc, h, s_h = cb->args[1];
2129 int idx, s_idx = idx = cb->args[2];
2130 struct neigh_hash_table *nht;
2133 nht = rcu_dereference_bh(tbl->nht);
2135 read_lock(&tbl->lock);
2136 for (h = 0; h <= nht->hash_mask; h++) {
2141 for (n = nht->hash_buckets[h], idx = 0; n; n = n->next) {
2142 if (!net_eq(dev_net(n->dev), net))
2146 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2149 NLM_F_MULTI) <= 0) {
2159 read_unlock(&tbl->lock);
2160 rcu_read_unlock_bh();
2166 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2168 struct neigh_table *tbl;
2171 read_lock(&neigh_tbl_lock);
2172 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2175 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2176 if (t < s_t || (family && tbl->family != family))
2179 memset(&cb->args[1], 0, sizeof(cb->args) -
2180 sizeof(cb->args[0]));
2181 if (neigh_dump_table(tbl, skb, cb) < 0)
2184 read_unlock(&neigh_tbl_lock);
2190 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2193 struct neigh_hash_table *nht;
2196 nht = rcu_dereference_bh(tbl->nht);
2198 read_lock(&tbl->lock);
2199 for (chain = 0; chain <= nht->hash_mask; chain++) {
2200 struct neighbour *n;
2202 for (n = nht->hash_buckets[chain]; n; n = n->next)
2205 read_unlock(&tbl->lock);
2206 rcu_read_unlock_bh();
2208 EXPORT_SYMBOL(neigh_for_each);
2210 /* The tbl->lock must be held as a writer and BH disabled. */
2211 void __neigh_for_each_release(struct neigh_table *tbl,
2212 int (*cb)(struct neighbour *))
2215 struct neigh_hash_table *nht;
2217 nht = rcu_dereference_protected(tbl->nht,
2218 lockdep_is_held(&tbl->lock));
2219 for (chain = 0; chain <= nht->hash_mask; chain++) {
2220 struct neighbour *n, **np;
2222 np = &nht->hash_buckets[chain];
2223 while ((n = *np) != NULL) {
2226 write_lock(&n->lock);
2233 write_unlock(&n->lock);
2235 neigh_cleanup_and_release(n);
2239 EXPORT_SYMBOL(__neigh_for_each_release);
2241 #ifdef CONFIG_PROC_FS
2243 static struct neighbour *neigh_get_first(struct seq_file *seq)
2245 struct neigh_seq_state *state = seq->private;
2246 struct net *net = seq_file_net(seq);
2247 struct neigh_hash_table *nht = state->nht;
2248 struct neighbour *n = NULL;
2249 int bucket = state->bucket;
2251 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2252 for (bucket = 0; bucket <= nht->hash_mask; bucket++) {
2253 n = nht->hash_buckets[bucket];
2256 if (!net_eq(dev_net(n->dev), net))
2258 if (state->neigh_sub_iter) {
2262 v = state->neigh_sub_iter(state, n, &fakep);
2266 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2268 if (n->nud_state & ~NUD_NOARP)
2277 state->bucket = bucket;
2282 static struct neighbour *neigh_get_next(struct seq_file *seq,
2283 struct neighbour *n,
2286 struct neigh_seq_state *state = seq->private;
2287 struct net *net = seq_file_net(seq);
2288 struct neigh_hash_table *nht = state->nht;
2290 if (state->neigh_sub_iter) {
2291 void *v = state->neigh_sub_iter(state, n, pos);
2299 if (!net_eq(dev_net(n->dev), net))
2301 if (state->neigh_sub_iter) {
2302 void *v = state->neigh_sub_iter(state, n, pos);
2307 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2310 if (n->nud_state & ~NUD_NOARP)
2319 if (++state->bucket > nht->hash_mask)
2322 n = nht->hash_buckets[state->bucket];
2330 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2332 struct neighbour *n = neigh_get_first(seq);
2337 n = neigh_get_next(seq, n, pos);
2342 return *pos ? NULL : n;
2345 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2347 struct neigh_seq_state *state = seq->private;
2348 struct net *net = seq_file_net(seq);
2349 struct neigh_table *tbl = state->tbl;
2350 struct pneigh_entry *pn = NULL;
2351 int bucket = state->bucket;
2353 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2354 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2355 pn = tbl->phash_buckets[bucket];
2356 while (pn && !net_eq(pneigh_net(pn), net))
2361 state->bucket = bucket;
2366 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2367 struct pneigh_entry *pn,
2370 struct neigh_seq_state *state = seq->private;
2371 struct net *net = seq_file_net(seq);
2372 struct neigh_table *tbl = state->tbl;
2376 if (++state->bucket > PNEIGH_HASHMASK)
2378 pn = tbl->phash_buckets[state->bucket];
2379 while (pn && !net_eq(pneigh_net(pn), net))
2391 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2393 struct pneigh_entry *pn = pneigh_get_first(seq);
2398 pn = pneigh_get_next(seq, pn, pos);
2403 return *pos ? NULL : pn;
2406 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2408 struct neigh_seq_state *state = seq->private;
2410 loff_t idxpos = *pos;
2412 rc = neigh_get_idx(seq, &idxpos);
2413 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2414 rc = pneigh_get_idx(seq, &idxpos);
2419 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2420 __acquires(tbl->lock)
2423 struct neigh_seq_state *state = seq->private;
2427 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2430 state->nht = rcu_dereference_bh(tbl->nht);
2431 read_lock(&tbl->lock);
2432 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2434 EXPORT_SYMBOL(neigh_seq_start);
2436 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2438 struct neigh_seq_state *state;
2441 if (v == SEQ_START_TOKEN) {
2442 rc = neigh_get_first(seq);
2446 state = seq->private;
2447 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2448 rc = neigh_get_next(seq, v, NULL);
2451 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2452 rc = pneigh_get_first(seq);
2454 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2455 rc = pneigh_get_next(seq, v, NULL);
2461 EXPORT_SYMBOL(neigh_seq_next);
2463 void neigh_seq_stop(struct seq_file *seq, void *v)
2464 __releases(tbl->lock)
2467 struct neigh_seq_state *state = seq->private;
2468 struct neigh_table *tbl = state->tbl;
2470 read_unlock(&tbl->lock);
2471 rcu_read_unlock_bh();
2473 EXPORT_SYMBOL(neigh_seq_stop);
2475 /* statistics via seq_file */
2477 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2479 struct neigh_table *tbl = seq->private;
2483 return SEQ_START_TOKEN;
2485 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2486 if (!cpu_possible(cpu))
2489 return per_cpu_ptr(tbl->stats, cpu);
2494 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2496 struct neigh_table *tbl = seq->private;
2499 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2500 if (!cpu_possible(cpu))
2503 return per_cpu_ptr(tbl->stats, cpu);
2508 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2513 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2515 struct neigh_table *tbl = seq->private;
2516 struct neigh_statistics *st = v;
2518 if (v == SEQ_START_TOKEN) {
2519 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");
2523 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2524 "%08lx %08lx %08lx %08lx %08lx\n",
2525 atomic_read(&tbl->entries),
2536 st->rcv_probes_mcast,
2537 st->rcv_probes_ucast,
2539 st->periodic_gc_runs,
2547 static const struct seq_operations neigh_stat_seq_ops = {
2548 .start = neigh_stat_seq_start,
2549 .next = neigh_stat_seq_next,
2550 .stop = neigh_stat_seq_stop,
2551 .show = neigh_stat_seq_show,
2554 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2556 int ret = seq_open(file, &neigh_stat_seq_ops);
2559 struct seq_file *sf = file->private_data;
2560 sf->private = PDE(inode)->data;
2565 static const struct file_operations neigh_stat_seq_fops = {
2566 .owner = THIS_MODULE,
2567 .open = neigh_stat_seq_open,
2569 .llseek = seq_lseek,
2570 .release = seq_release,
2573 #endif /* CONFIG_PROC_FS */
2575 static inline size_t neigh_nlmsg_size(void)
2577 return NLMSG_ALIGN(sizeof(struct ndmsg))
2578 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2579 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2580 + nla_total_size(sizeof(struct nda_cacheinfo))
2581 + nla_total_size(4); /* NDA_PROBES */
2584 static void __neigh_notify(struct neighbour *n, int type, int flags)
2586 struct net *net = dev_net(n->dev);
2587 struct sk_buff *skb;
2590 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2594 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2596 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2597 WARN_ON(err == -EMSGSIZE);
2601 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2605 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2609 void neigh_app_ns(struct neighbour *n)
2611 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2613 EXPORT_SYMBOL(neigh_app_ns);
2614 #endif /* CONFIG_ARPD */
2616 #ifdef CONFIG_SYSCTL
2618 #define NEIGH_VARS_MAX 19
2620 static struct neigh_sysctl_table {
2621 struct ctl_table_header *sysctl_header;
2622 struct ctl_table neigh_vars[NEIGH_VARS_MAX];
2624 } neigh_sysctl_template __read_mostly = {
2627 .procname = "mcast_solicit",
2628 .maxlen = sizeof(int),
2630 .proc_handler = proc_dointvec,
2633 .procname = "ucast_solicit",
2634 .maxlen = sizeof(int),
2636 .proc_handler = proc_dointvec,
2639 .procname = "app_solicit",
2640 .maxlen = sizeof(int),
2642 .proc_handler = proc_dointvec,
2645 .procname = "retrans_time",
2646 .maxlen = sizeof(int),
2648 .proc_handler = proc_dointvec_userhz_jiffies,
2651 .procname = "base_reachable_time",
2652 .maxlen = sizeof(int),
2654 .proc_handler = proc_dointvec_jiffies,
2657 .procname = "delay_first_probe_time",
2658 .maxlen = sizeof(int),
2660 .proc_handler = proc_dointvec_jiffies,
2663 .procname = "gc_stale_time",
2664 .maxlen = sizeof(int),
2666 .proc_handler = proc_dointvec_jiffies,
2669 .procname = "unres_qlen",
2670 .maxlen = sizeof(int),
2672 .proc_handler = proc_dointvec,
2675 .procname = "proxy_qlen",
2676 .maxlen = sizeof(int),
2678 .proc_handler = proc_dointvec,
2681 .procname = "anycast_delay",
2682 .maxlen = sizeof(int),
2684 .proc_handler = proc_dointvec_userhz_jiffies,
2687 .procname = "proxy_delay",
2688 .maxlen = sizeof(int),
2690 .proc_handler = proc_dointvec_userhz_jiffies,
2693 .procname = "locktime",
2694 .maxlen = sizeof(int),
2696 .proc_handler = proc_dointvec_userhz_jiffies,
2699 .procname = "retrans_time_ms",
2700 .maxlen = sizeof(int),
2702 .proc_handler = proc_dointvec_ms_jiffies,
2705 .procname = "base_reachable_time_ms",
2706 .maxlen = sizeof(int),
2708 .proc_handler = proc_dointvec_ms_jiffies,
2711 .procname = "gc_interval",
2712 .maxlen = sizeof(int),
2714 .proc_handler = proc_dointvec_jiffies,
2717 .procname = "gc_thresh1",
2718 .maxlen = sizeof(int),
2720 .proc_handler = proc_dointvec,
2723 .procname = "gc_thresh2",
2724 .maxlen = sizeof(int),
2726 .proc_handler = proc_dointvec,
2729 .procname = "gc_thresh3",
2730 .maxlen = sizeof(int),
2732 .proc_handler = proc_dointvec,
2738 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2739 char *p_name, proc_handler *handler)
2741 struct neigh_sysctl_table *t;
2742 const char *dev_name_source = NULL;
2744 #define NEIGH_CTL_PATH_ROOT 0
2745 #define NEIGH_CTL_PATH_PROTO 1
2746 #define NEIGH_CTL_PATH_NEIGH 2
2747 #define NEIGH_CTL_PATH_DEV 3
2749 struct ctl_path neigh_path[] = {
2750 { .procname = "net", },
2751 { .procname = "proto", },
2752 { .procname = "neigh", },
2753 { .procname = "default", },
2757 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2761 t->neigh_vars[0].data = &p->mcast_probes;
2762 t->neigh_vars[1].data = &p->ucast_probes;
2763 t->neigh_vars[2].data = &p->app_probes;
2764 t->neigh_vars[3].data = &p->retrans_time;
2765 t->neigh_vars[4].data = &p->base_reachable_time;
2766 t->neigh_vars[5].data = &p->delay_probe_time;
2767 t->neigh_vars[6].data = &p->gc_staletime;
2768 t->neigh_vars[7].data = &p->queue_len;
2769 t->neigh_vars[8].data = &p->proxy_qlen;
2770 t->neigh_vars[9].data = &p->anycast_delay;
2771 t->neigh_vars[10].data = &p->proxy_delay;
2772 t->neigh_vars[11].data = &p->locktime;
2773 t->neigh_vars[12].data = &p->retrans_time;
2774 t->neigh_vars[13].data = &p->base_reachable_time;
2777 dev_name_source = dev->name;
2778 /* Terminate the table early */
2779 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
2781 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2782 t->neigh_vars[14].data = (int *)(p + 1);
2783 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2784 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2785 t->neigh_vars[17].data = (int *)(p + 1) + 3;
2791 t->neigh_vars[3].proc_handler = handler;
2792 t->neigh_vars[3].extra1 = dev;
2794 t->neigh_vars[4].proc_handler = handler;
2795 t->neigh_vars[4].extra1 = dev;
2796 /* RetransTime (in milliseconds)*/
2797 t->neigh_vars[12].proc_handler = handler;
2798 t->neigh_vars[12].extra1 = dev;
2799 /* ReachableTime (in milliseconds) */
2800 t->neigh_vars[13].proc_handler = handler;
2801 t->neigh_vars[13].extra1 = dev;
2804 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2808 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2809 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2812 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2813 if (!t->sysctl_header)
2816 p->sysctl_table = t;
2826 EXPORT_SYMBOL(neigh_sysctl_register);
2828 void neigh_sysctl_unregister(struct neigh_parms *p)
2830 if (p->sysctl_table) {
2831 struct neigh_sysctl_table *t = p->sysctl_table;
2832 p->sysctl_table = NULL;
2833 unregister_sysctl_table(t->sysctl_header);
2838 EXPORT_SYMBOL(neigh_sysctl_unregister);
2840 #endif /* CONFIG_SYSCTL */
2842 static int __init neigh_init(void)
2844 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL);
2845 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL);
2846 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info);
2848 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info);
2849 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL);
2854 subsys_initcall(neigh_init);