2 * Generic address resolution entity
5 * Pedro Roque <roque@di.fc.ul.pt>
6 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 * Vitaly E. Lavrov releasing NULL neighbor in neigh_add.
15 * Harald Welte Add neighbour cache statistics like rtstat
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 #include <linux/slab.h>
21 #include <linux/types.h>
22 #include <linux/kernel.h>
23 #include <linux/module.h>
24 #include <linux/socket.h>
25 #include <linux/netdevice.h>
26 #include <linux/proc_fs.h>
28 #include <linux/sysctl.h>
30 #include <linux/times.h>
31 #include <net/net_namespace.h>
32 #include <net/neighbour.h>
35 #include <net/netevent.h>
36 #include <net/netlink.h>
37 #include <linux/rtnetlink.h>
38 #include <linux/random.h>
39 #include <linux/string.h>
40 #include <linux/log2.h>
41 #include <linux/inetdevice.h>
42 #include <net/addrconf.h>
46 #define neigh_dbg(level, fmt, ...) \
48 if (level <= NEIGH_DEBUG) \
49 pr_debug(fmt, ##__VA_ARGS__); \
52 #define PNEIGH_HASHMASK 0xF
54 static void neigh_timer_handler(struct timer_list *t);
55 static void __neigh_notify(struct neighbour *n, int type, int flags,
57 static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid);
58 static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
59 struct net_device *dev);
62 static const struct seq_operations neigh_stat_seq_ops;
66 Neighbour hash table buckets are protected with rwlock tbl->lock.
68 - All the scans/updates to hash buckets MUST be made under this lock.
69 - NOTHING clever should be made under this lock: no callbacks
70 to protocol backends, no attempts to send something to network.
71 It will result in deadlocks, if backend/driver wants to use neighbour
73 - If the entry requires some non-trivial actions, increase
74 its reference count and release table lock.
76 Neighbour entries are protected:
77 - with reference count.
78 - with rwlock neigh->lock
80 Reference count prevents destruction.
82 neigh->lock mainly serializes ll address data and its validity state.
83 However, the same lock is used to protect another entry fields:
87 Again, nothing clever shall be made under neigh->lock,
88 the most complicated procedure, which we allow is dev->hard_header.
89 It is supposed, that dev->hard_header is simplistic and does
90 not make callbacks to neighbour tables.
93 static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
99 static void neigh_cleanup_and_release(struct neighbour *neigh)
101 if (neigh->parms->neigh_cleanup)
102 neigh->parms->neigh_cleanup(neigh);
104 __neigh_notify(neigh, RTM_DELNEIGH, 0, 0);
105 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
106 neigh_release(neigh);
110 * It is random distribution in the interval (1/2)*base...(3/2)*base.
111 * It corresponds to default IPv6 settings and is not overridable,
112 * because it is really reasonable choice.
115 unsigned long neigh_rand_reach_time(unsigned long base)
117 return base ? (prandom_u32() % base) + (base >> 1) : 0;
119 EXPORT_SYMBOL(neigh_rand_reach_time);
122 static bool neigh_del(struct neighbour *n, __u8 state, __u8 flags,
123 struct neighbour __rcu **np, struct neigh_table *tbl)
127 write_lock(&n->lock);
128 if (refcount_read(&n->refcnt) == 1 && !(n->nud_state & state) &&
129 !(n->flags & flags)) {
130 struct neighbour *neigh;
132 neigh = rcu_dereference_protected(n->next,
133 lockdep_is_held(&tbl->lock));
134 rcu_assign_pointer(*np, neigh);
138 write_unlock(&n->lock);
140 neigh_cleanup_and_release(n);
144 bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl)
146 struct neigh_hash_table *nht;
147 void *pkey = ndel->primary_key;
150 struct neighbour __rcu **np;
152 nht = rcu_dereference_protected(tbl->nht,
153 lockdep_is_held(&tbl->lock));
154 hash_val = tbl->hash(pkey, ndel->dev, nht->hash_rnd);
155 hash_val = hash_val >> (32 - nht->hash_shift);
157 np = &nht->hash_buckets[hash_val];
158 while ((n = rcu_dereference_protected(*np,
159 lockdep_is_held(&tbl->lock)))) {
161 return neigh_del(n, 0, 0, np, tbl);
167 static int neigh_forced_gc(struct neigh_table *tbl)
171 struct neigh_hash_table *nht;
173 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
175 write_lock_bh(&tbl->lock);
176 nht = rcu_dereference_protected(tbl->nht,
177 lockdep_is_held(&tbl->lock));
178 for (i = 0; i < (1 << nht->hash_shift); i++) {
180 struct neighbour __rcu **np;
182 np = &nht->hash_buckets[i];
183 while ((n = rcu_dereference_protected(*np,
184 lockdep_is_held(&tbl->lock))) != NULL) {
185 /* Neighbour record may be discarded if:
186 * - nobody refers to it.
187 * - it is not permanent
189 if (neigh_del(n, NUD_PERMANENT, NTF_EXT_LEARNED, np,
198 tbl->last_flush = jiffies;
200 write_unlock_bh(&tbl->lock);
205 static void neigh_add_timer(struct neighbour *n, unsigned long when)
208 if (unlikely(mod_timer(&n->timer, when))) {
209 printk("NEIGH: BUG, double timer add, state is %x\n",
215 static int neigh_del_timer(struct neighbour *n)
217 if ((n->nud_state & NUD_IN_TIMER) &&
218 del_timer(&n->timer)) {
225 static void pneigh_queue_purge(struct sk_buff_head *list)
229 while ((skb = skb_dequeue(list)) != NULL) {
235 static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
238 struct neigh_hash_table *nht;
240 nht = rcu_dereference_protected(tbl->nht,
241 lockdep_is_held(&tbl->lock));
243 for (i = 0; i < (1 << nht->hash_shift); i++) {
245 struct neighbour __rcu **np = &nht->hash_buckets[i];
247 while ((n = rcu_dereference_protected(*np,
248 lockdep_is_held(&tbl->lock))) != NULL) {
249 if (dev && n->dev != dev) {
253 rcu_assign_pointer(*np,
254 rcu_dereference_protected(n->next,
255 lockdep_is_held(&tbl->lock)));
256 write_lock(&n->lock);
260 if (refcount_read(&n->refcnt) != 1) {
261 /* The most unpleasant situation.
262 We must destroy neighbour entry,
263 but someone still uses it.
265 The destroy will be delayed until
266 the last user releases us, but
267 we must kill timers etc. and move
270 __skb_queue_purge(&n->arp_queue);
271 n->arp_queue_len_bytes = 0;
272 n->output = neigh_blackhole;
273 if (n->nud_state & NUD_VALID)
274 n->nud_state = NUD_NOARP;
276 n->nud_state = NUD_NONE;
277 neigh_dbg(2, "neigh %p is stray\n", n);
279 write_unlock(&n->lock);
280 neigh_cleanup_and_release(n);
285 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
287 write_lock_bh(&tbl->lock);
288 neigh_flush_dev(tbl, dev);
289 write_unlock_bh(&tbl->lock);
291 EXPORT_SYMBOL(neigh_changeaddr);
293 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
295 write_lock_bh(&tbl->lock);
296 neigh_flush_dev(tbl, dev);
297 pneigh_ifdown_and_unlock(tbl, dev);
299 del_timer_sync(&tbl->proxy_timer);
300 pneigh_queue_purge(&tbl->proxy_queue);
303 EXPORT_SYMBOL(neigh_ifdown);
305 static struct neighbour *neigh_alloc(struct neigh_table *tbl, struct net_device *dev)
307 struct neighbour *n = NULL;
308 unsigned long now = jiffies;
311 entries = atomic_inc_return(&tbl->entries) - 1;
312 if (entries >= tbl->gc_thresh3 ||
313 (entries >= tbl->gc_thresh2 &&
314 time_after(now, tbl->last_flush + 5 * HZ))) {
315 if (!neigh_forced_gc(tbl) &&
316 entries >= tbl->gc_thresh3) {
317 net_info_ratelimited("%s: neighbor table overflow!\n",
319 NEIGH_CACHE_STAT_INC(tbl, table_fulls);
324 n = kzalloc(tbl->entry_size + dev->neigh_priv_len, GFP_ATOMIC);
328 __skb_queue_head_init(&n->arp_queue);
329 rwlock_init(&n->lock);
330 seqlock_init(&n->ha_lock);
331 n->updated = n->used = now;
332 n->nud_state = NUD_NONE;
333 n->output = neigh_blackhole;
334 seqlock_init(&n->hh.hh_lock);
335 n->parms = neigh_parms_clone(&tbl->parms);
336 timer_setup(&n->timer, neigh_timer_handler, 0);
338 NEIGH_CACHE_STAT_INC(tbl, allocs);
340 refcount_set(&n->refcnt, 1);
346 atomic_dec(&tbl->entries);
350 static void neigh_get_hash_rnd(u32 *x)
352 *x = get_random_u32() | 1;
355 static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
357 size_t size = (1 << shift) * sizeof(struct neighbour *);
358 struct neigh_hash_table *ret;
359 struct neighbour __rcu **buckets;
362 ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
365 if (size <= PAGE_SIZE)
366 buckets = kzalloc(size, GFP_ATOMIC);
368 buckets = (struct neighbour __rcu **)
369 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
375 ret->hash_buckets = buckets;
376 ret->hash_shift = shift;
377 for (i = 0; i < NEIGH_NUM_HASH_RND; i++)
378 neigh_get_hash_rnd(&ret->hash_rnd[i]);
382 static void neigh_hash_free_rcu(struct rcu_head *head)
384 struct neigh_hash_table *nht = container_of(head,
385 struct neigh_hash_table,
387 size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
388 struct neighbour __rcu **buckets = nht->hash_buckets;
390 if (size <= PAGE_SIZE)
393 free_pages((unsigned long)buckets, get_order(size));
397 static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
398 unsigned long new_shift)
400 unsigned int i, hash;
401 struct neigh_hash_table *new_nht, *old_nht;
403 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
405 old_nht = rcu_dereference_protected(tbl->nht,
406 lockdep_is_held(&tbl->lock));
407 new_nht = neigh_hash_alloc(new_shift);
411 for (i = 0; i < (1 << old_nht->hash_shift); i++) {
412 struct neighbour *n, *next;
414 for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
415 lockdep_is_held(&tbl->lock));
418 hash = tbl->hash(n->primary_key, n->dev,
421 hash >>= (32 - new_nht->hash_shift);
422 next = rcu_dereference_protected(n->next,
423 lockdep_is_held(&tbl->lock));
425 rcu_assign_pointer(n->next,
426 rcu_dereference_protected(
427 new_nht->hash_buckets[hash],
428 lockdep_is_held(&tbl->lock)));
429 rcu_assign_pointer(new_nht->hash_buckets[hash], n);
433 rcu_assign_pointer(tbl->nht, new_nht);
434 call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
438 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
439 struct net_device *dev)
443 NEIGH_CACHE_STAT_INC(tbl, lookups);
446 n = __neigh_lookup_noref(tbl, pkey, dev);
448 if (!refcount_inc_not_zero(&n->refcnt))
450 NEIGH_CACHE_STAT_INC(tbl, hits);
453 rcu_read_unlock_bh();
456 EXPORT_SYMBOL(neigh_lookup);
458 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
462 unsigned int key_len = tbl->key_len;
464 struct neigh_hash_table *nht;
466 NEIGH_CACHE_STAT_INC(tbl, lookups);
469 nht = rcu_dereference_bh(tbl->nht);
470 hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
472 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
474 n = rcu_dereference_bh(n->next)) {
475 if (!memcmp(n->primary_key, pkey, key_len) &&
476 net_eq(dev_net(n->dev), net)) {
477 if (!refcount_inc_not_zero(&n->refcnt))
479 NEIGH_CACHE_STAT_INC(tbl, hits);
484 rcu_read_unlock_bh();
487 EXPORT_SYMBOL(neigh_lookup_nodev);
489 struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
490 struct net_device *dev, bool want_ref)
493 unsigned int key_len = tbl->key_len;
495 struct neighbour *n1, *rc, *n = neigh_alloc(tbl, dev);
496 struct neigh_hash_table *nht;
499 rc = ERR_PTR(-ENOBUFS);
503 memcpy(n->primary_key, pkey, key_len);
507 /* Protocol specific setup. */
508 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
510 goto out_neigh_release;
513 if (dev->netdev_ops->ndo_neigh_construct) {
514 error = dev->netdev_ops->ndo_neigh_construct(dev, n);
517 goto out_neigh_release;
521 /* Device specific setup. */
522 if (n->parms->neigh_setup &&
523 (error = n->parms->neigh_setup(n)) < 0) {
525 goto out_neigh_release;
528 n->confirmed = jiffies - (NEIGH_VAR(n->parms, BASE_REACHABLE_TIME) << 1);
530 write_lock_bh(&tbl->lock);
531 nht = rcu_dereference_protected(tbl->nht,
532 lockdep_is_held(&tbl->lock));
534 if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
535 nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
537 hash_val = tbl->hash(n->primary_key, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
539 if (n->parms->dead) {
540 rc = ERR_PTR(-EINVAL);
544 for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
545 lockdep_is_held(&tbl->lock));
547 n1 = rcu_dereference_protected(n1->next,
548 lockdep_is_held(&tbl->lock))) {
549 if (dev == n1->dev && !memcmp(n1->primary_key, n->primary_key, key_len)) {
560 rcu_assign_pointer(n->next,
561 rcu_dereference_protected(nht->hash_buckets[hash_val],
562 lockdep_is_held(&tbl->lock)));
563 rcu_assign_pointer(nht->hash_buckets[hash_val], n);
564 write_unlock_bh(&tbl->lock);
565 neigh_dbg(2, "neigh %p is created\n", n);
570 write_unlock_bh(&tbl->lock);
575 EXPORT_SYMBOL(__neigh_create);
577 static u32 pneigh_hash(const void *pkey, unsigned int key_len)
579 u32 hash_val = *(u32 *)(pkey + key_len - 4);
580 hash_val ^= (hash_val >> 16);
581 hash_val ^= hash_val >> 8;
582 hash_val ^= hash_val >> 4;
583 hash_val &= PNEIGH_HASHMASK;
587 static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
590 unsigned int key_len,
591 struct net_device *dev)
594 if (!memcmp(n->key, pkey, key_len) &&
595 net_eq(pneigh_net(n), net) &&
596 (n->dev == dev || !n->dev))
603 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
604 struct net *net, const void *pkey, struct net_device *dev)
606 unsigned int key_len = tbl->key_len;
607 u32 hash_val = pneigh_hash(pkey, key_len);
609 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
610 net, pkey, key_len, dev);
612 EXPORT_SYMBOL_GPL(__pneigh_lookup);
614 struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
615 struct net *net, const void *pkey,
616 struct net_device *dev, int creat)
618 struct pneigh_entry *n;
619 unsigned int key_len = tbl->key_len;
620 u32 hash_val = pneigh_hash(pkey, key_len);
622 read_lock_bh(&tbl->lock);
623 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
624 net, pkey, key_len, dev);
625 read_unlock_bh(&tbl->lock);
632 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
636 write_pnet(&n->net, net);
637 memcpy(n->key, pkey, key_len);
642 if (tbl->pconstructor && tbl->pconstructor(n)) {
650 write_lock_bh(&tbl->lock);
651 n->next = tbl->phash_buckets[hash_val];
652 tbl->phash_buckets[hash_val] = n;
653 write_unlock_bh(&tbl->lock);
657 EXPORT_SYMBOL(pneigh_lookup);
660 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
661 struct net_device *dev)
663 struct pneigh_entry *n, **np;
664 unsigned int key_len = tbl->key_len;
665 u32 hash_val = pneigh_hash(pkey, key_len);
667 write_lock_bh(&tbl->lock);
668 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
670 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
671 net_eq(pneigh_net(n), net)) {
673 write_unlock_bh(&tbl->lock);
674 if (tbl->pdestructor)
682 write_unlock_bh(&tbl->lock);
686 static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
687 struct net_device *dev)
689 struct pneigh_entry *n, **np, *freelist = NULL;
692 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
693 np = &tbl->phash_buckets[h];
694 while ((n = *np) != NULL) {
695 if (!dev || n->dev == dev) {
704 write_unlock_bh(&tbl->lock);
705 while ((n = freelist)) {
708 if (tbl->pdestructor)
717 static void neigh_parms_destroy(struct neigh_parms *parms);
719 static inline void neigh_parms_put(struct neigh_parms *parms)
721 if (refcount_dec_and_test(&parms->refcnt))
722 neigh_parms_destroy(parms);
726 * neighbour must already be out of the table;
729 void neigh_destroy(struct neighbour *neigh)
731 struct net_device *dev = neigh->dev;
733 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
736 pr_warn("Destroying alive neighbour %p\n", neigh);
741 if (neigh_del_timer(neigh))
742 pr_warn("Impossible event\n");
744 write_lock_bh(&neigh->lock);
745 __skb_queue_purge(&neigh->arp_queue);
746 write_unlock_bh(&neigh->lock);
747 neigh->arp_queue_len_bytes = 0;
749 if (dev->netdev_ops->ndo_neigh_destroy)
750 dev->netdev_ops->ndo_neigh_destroy(dev, neigh);
753 neigh_parms_put(neigh->parms);
755 neigh_dbg(2, "neigh %p is destroyed\n", neigh);
757 atomic_dec(&neigh->tbl->entries);
758 kfree_rcu(neigh, rcu);
760 EXPORT_SYMBOL(neigh_destroy);
762 /* Neighbour state is suspicious;
765 Called with write_locked neigh.
767 static void neigh_suspect(struct neighbour *neigh)
769 neigh_dbg(2, "neigh %p is suspected\n", neigh);
771 neigh->output = neigh->ops->output;
774 /* Neighbour state is OK;
777 Called with write_locked neigh.
779 static void neigh_connect(struct neighbour *neigh)
781 neigh_dbg(2, "neigh %p is connected\n", neigh);
783 neigh->output = neigh->ops->connected_output;
786 static void neigh_periodic_work(struct work_struct *work)
788 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
790 struct neighbour __rcu **np;
792 struct neigh_hash_table *nht;
794 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
796 write_lock_bh(&tbl->lock);
797 nht = rcu_dereference_protected(tbl->nht,
798 lockdep_is_held(&tbl->lock));
801 * periodically recompute ReachableTime from random function
804 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
805 struct neigh_parms *p;
806 tbl->last_rand = jiffies;
807 list_for_each_entry(p, &tbl->parms_list, list)
809 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
812 if (atomic_read(&tbl->entries) < tbl->gc_thresh1)
815 for (i = 0 ; i < (1 << nht->hash_shift); i++) {
816 np = &nht->hash_buckets[i];
818 while ((n = rcu_dereference_protected(*np,
819 lockdep_is_held(&tbl->lock))) != NULL) {
822 write_lock(&n->lock);
824 state = n->nud_state;
825 if ((state & (NUD_PERMANENT | NUD_IN_TIMER)) ||
826 (n->flags & NTF_EXT_LEARNED)) {
827 write_unlock(&n->lock);
831 if (time_before(n->used, n->confirmed))
832 n->used = n->confirmed;
834 if (refcount_read(&n->refcnt) == 1 &&
835 (state == NUD_FAILED ||
836 time_after(jiffies, n->used + NEIGH_VAR(n->parms, GC_STALETIME)))) {
839 write_unlock(&n->lock);
840 neigh_cleanup_and_release(n);
843 write_unlock(&n->lock);
849 * It's fine to release lock here, even if hash table
850 * grows while we are preempted.
852 write_unlock_bh(&tbl->lock);
854 write_lock_bh(&tbl->lock);
855 nht = rcu_dereference_protected(tbl->nht,
856 lockdep_is_held(&tbl->lock));
859 /* Cycle through all hash buckets every BASE_REACHABLE_TIME/2 ticks.
860 * ARP entry timeouts range from 1/2 BASE_REACHABLE_TIME to 3/2
861 * BASE_REACHABLE_TIME.
863 queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
864 NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME) >> 1);
865 write_unlock_bh(&tbl->lock);
868 static __inline__ int neigh_max_probes(struct neighbour *n)
870 struct neigh_parms *p = n->parms;
871 return NEIGH_VAR(p, UCAST_PROBES) + NEIGH_VAR(p, APP_PROBES) +
872 (n->nud_state & NUD_PROBE ? NEIGH_VAR(p, MCAST_REPROBES) :
873 NEIGH_VAR(p, MCAST_PROBES));
876 static void neigh_invalidate(struct neighbour *neigh)
877 __releases(neigh->lock)
878 __acquires(neigh->lock)
882 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
883 neigh_dbg(2, "neigh %p is failed\n", neigh);
884 neigh->updated = jiffies;
886 /* It is very thin place. report_unreachable is very complicated
887 routine. Particularly, it can hit the same neighbour entry!
889 So that, we try to be accurate and avoid dead loop. --ANK
891 while (neigh->nud_state == NUD_FAILED &&
892 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
893 write_unlock(&neigh->lock);
894 neigh->ops->error_report(neigh, skb);
895 write_lock(&neigh->lock);
897 __skb_queue_purge(&neigh->arp_queue);
898 neigh->arp_queue_len_bytes = 0;
901 static void neigh_probe(struct neighbour *neigh)
902 __releases(neigh->lock)
904 struct sk_buff *skb = skb_peek_tail(&neigh->arp_queue);
905 /* keep skb alive even if arp_queue overflows */
907 skb = skb_clone(skb, GFP_ATOMIC);
908 write_unlock(&neigh->lock);
909 if (neigh->ops->solicit)
910 neigh->ops->solicit(neigh, skb);
911 atomic_inc(&neigh->probes);
915 /* Called when a timer expires for a neighbour entry. */
917 static void neigh_timer_handler(struct timer_list *t)
919 unsigned long now, next;
920 struct neighbour *neigh = from_timer(neigh, t, timer);
924 write_lock(&neigh->lock);
926 state = neigh->nud_state;
930 if (!(state & NUD_IN_TIMER))
933 if (state & NUD_REACHABLE) {
934 if (time_before_eq(now,
935 neigh->confirmed + neigh->parms->reachable_time)) {
936 neigh_dbg(2, "neigh %p is still alive\n", neigh);
937 next = neigh->confirmed + neigh->parms->reachable_time;
938 } else if (time_before_eq(now,
940 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
941 neigh_dbg(2, "neigh %p is delayed\n", neigh);
942 neigh->nud_state = NUD_DELAY;
943 neigh->updated = jiffies;
944 neigh_suspect(neigh);
945 next = now + NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME);
947 neigh_dbg(2, "neigh %p is suspected\n", neigh);
948 neigh->nud_state = NUD_STALE;
949 neigh->updated = jiffies;
950 neigh_suspect(neigh);
953 } else if (state & NUD_DELAY) {
954 if (time_before_eq(now,
956 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
957 neigh_dbg(2, "neigh %p is now reachable\n", neigh);
958 neigh->nud_state = NUD_REACHABLE;
959 neigh->updated = jiffies;
960 neigh_connect(neigh);
962 next = neigh->confirmed + neigh->parms->reachable_time;
964 neigh_dbg(2, "neigh %p is probed\n", neigh);
965 neigh->nud_state = NUD_PROBE;
966 neigh->updated = jiffies;
967 atomic_set(&neigh->probes, 0);
969 next = now + NEIGH_VAR(neigh->parms, RETRANS_TIME);
972 /* NUD_PROBE|NUD_INCOMPLETE */
973 next = now + NEIGH_VAR(neigh->parms, RETRANS_TIME);
976 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
977 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
978 neigh->nud_state = NUD_FAILED;
980 neigh_invalidate(neigh);
984 if (neigh->nud_state & NUD_IN_TIMER) {
985 if (time_before(next, jiffies + HZ/2))
986 next = jiffies + HZ/2;
987 if (!mod_timer(&neigh->timer, next))
990 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
994 write_unlock(&neigh->lock);
998 neigh_update_notify(neigh, 0);
1000 neigh_release(neigh);
1003 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
1006 bool immediate_probe = false;
1008 write_lock_bh(&neigh->lock);
1011 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
1016 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
1017 if (NEIGH_VAR(neigh->parms, MCAST_PROBES) +
1018 NEIGH_VAR(neigh->parms, APP_PROBES)) {
1019 unsigned long next, now = jiffies;
1021 atomic_set(&neigh->probes,
1022 NEIGH_VAR(neigh->parms, UCAST_PROBES));
1023 neigh->nud_state = NUD_INCOMPLETE;
1024 neigh->updated = now;
1025 next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1027 neigh_add_timer(neigh, next);
1028 immediate_probe = true;
1030 neigh->nud_state = NUD_FAILED;
1031 neigh->updated = jiffies;
1032 write_unlock_bh(&neigh->lock);
1037 } else if (neigh->nud_state & NUD_STALE) {
1038 neigh_dbg(2, "neigh %p is delayed\n", neigh);
1039 neigh->nud_state = NUD_DELAY;
1040 neigh->updated = jiffies;
1041 neigh_add_timer(neigh, jiffies +
1042 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME));
1045 if (neigh->nud_state == NUD_INCOMPLETE) {
1047 while (neigh->arp_queue_len_bytes + skb->truesize >
1048 NEIGH_VAR(neigh->parms, QUEUE_LEN_BYTES)) {
1049 struct sk_buff *buff;
1051 buff = __skb_dequeue(&neigh->arp_queue);
1054 neigh->arp_queue_len_bytes -= buff->truesize;
1056 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
1059 __skb_queue_tail(&neigh->arp_queue, skb);
1060 neigh->arp_queue_len_bytes += skb->truesize;
1065 if (immediate_probe)
1068 write_unlock(&neigh->lock);
1073 if (neigh->nud_state & NUD_STALE)
1075 write_unlock_bh(&neigh->lock);
1079 EXPORT_SYMBOL(__neigh_event_send);
1081 static void neigh_update_hhs(struct neighbour *neigh)
1083 struct hh_cache *hh;
1084 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1087 if (neigh->dev->header_ops)
1088 update = neigh->dev->header_ops->cache_update;
1093 write_seqlock_bh(&hh->hh_lock);
1094 update(hh, neigh->dev, neigh->ha);
1095 write_sequnlock_bh(&hh->hh_lock);
1102 /* Generic update routine.
1103 -- lladdr is new lladdr or NULL, if it is not supplied.
1104 -- new is new state.
1106 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1108 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1109 lladdr instead of overriding it
1111 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1113 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1115 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1118 Caller MUST hold reference count on the entry.
1121 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1122 u32 flags, u32 nlmsg_pid)
1127 struct net_device *dev;
1128 int update_isrouter = 0;
1130 write_lock_bh(&neigh->lock);
1133 old = neigh->nud_state;
1136 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1137 (old & (NUD_NOARP | NUD_PERMANENT)))
1142 neigh_update_ext_learned(neigh, flags, ¬ify);
1144 if (!(new & NUD_VALID)) {
1145 neigh_del_timer(neigh);
1146 if (old & NUD_CONNECTED)
1147 neigh_suspect(neigh);
1148 neigh->nud_state = new;
1150 notify = old & NUD_VALID;
1151 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1152 (new & NUD_FAILED)) {
1153 neigh_invalidate(neigh);
1159 /* Compare new lladdr with cached one */
1160 if (!dev->addr_len) {
1161 /* First case: device needs no address. */
1163 } else if (lladdr) {
1164 /* The second case: if something is already cached
1165 and a new address is proposed:
1167 - if they are different, check override flag
1169 if ((old & NUD_VALID) &&
1170 !memcmp(lladdr, neigh->ha, dev->addr_len))
1173 /* No address is supplied; if we know something,
1174 use it, otherwise discard the request.
1177 if (!(old & NUD_VALID))
1182 /* Update confirmed timestamp for neighbour entry after we
1183 * received ARP packet even if it doesn't change IP to MAC binding.
1185 if (new & NUD_CONNECTED)
1186 neigh->confirmed = jiffies;
1188 /* If entry was valid and address is not changed,
1189 do not change entry state, if new one is STALE.
1192 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1193 if (old & NUD_VALID) {
1194 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1195 update_isrouter = 0;
1196 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1197 (old & NUD_CONNECTED)) {
1203 if (lladdr == neigh->ha && new == NUD_STALE &&
1204 !(flags & NEIGH_UPDATE_F_ADMIN))
1209 /* Update timestamp only once we know we will make a change to the
1210 * neighbour entry. Otherwise we risk to move the locktime window with
1211 * noop updates and ignore relevant ARP updates.
1213 if (new != old || lladdr != neigh->ha)
1214 neigh->updated = jiffies;
1217 neigh_del_timer(neigh);
1218 if (new & NUD_PROBE)
1219 atomic_set(&neigh->probes, 0);
1220 if (new & NUD_IN_TIMER)
1221 neigh_add_timer(neigh, (jiffies +
1222 ((new & NUD_REACHABLE) ?
1223 neigh->parms->reachable_time :
1225 neigh->nud_state = new;
1229 if (lladdr != neigh->ha) {
1230 write_seqlock(&neigh->ha_lock);
1231 memcpy(&neigh->ha, lladdr, dev->addr_len);
1232 write_sequnlock(&neigh->ha_lock);
1233 neigh_update_hhs(neigh);
1234 if (!(new & NUD_CONNECTED))
1235 neigh->confirmed = jiffies -
1236 (NEIGH_VAR(neigh->parms, BASE_REACHABLE_TIME) << 1);
1241 if (new & NUD_CONNECTED)
1242 neigh_connect(neigh);
1244 neigh_suspect(neigh);
1245 if (!(old & NUD_VALID)) {
1246 struct sk_buff *skb;
1248 /* Again: avoid dead loop if something went wrong */
1250 while (neigh->nud_state & NUD_VALID &&
1251 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1252 struct dst_entry *dst = skb_dst(skb);
1253 struct neighbour *n2, *n1 = neigh;
1254 write_unlock_bh(&neigh->lock);
1258 /* Why not just use 'neigh' as-is? The problem is that
1259 * things such as shaper, eql, and sch_teql can end up
1260 * using alternative, different, neigh objects to output
1261 * the packet in the output path. So what we need to do
1262 * here is re-lookup the top-level neigh in the path so
1263 * we can reinject the packet there.
1267 n2 = dst_neigh_lookup_skb(dst, skb);
1271 n1->output(n1, skb);
1276 write_lock_bh(&neigh->lock);
1278 __skb_queue_purge(&neigh->arp_queue);
1279 neigh->arp_queue_len_bytes = 0;
1282 if (update_isrouter) {
1283 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1284 (neigh->flags | NTF_ROUTER) :
1285 (neigh->flags & ~NTF_ROUTER);
1287 write_unlock_bh(&neigh->lock);
1290 neigh_update_notify(neigh, nlmsg_pid);
1294 EXPORT_SYMBOL(neigh_update);
1296 /* Update the neigh to listen temporarily for probe responses, even if it is
1297 * in a NUD_FAILED state. The caller has to hold neigh->lock for writing.
1299 void __neigh_set_probe_once(struct neighbour *neigh)
1303 neigh->updated = jiffies;
1304 if (!(neigh->nud_state & NUD_FAILED))
1306 neigh->nud_state = NUD_INCOMPLETE;
1307 atomic_set(&neigh->probes, neigh_max_probes(neigh));
1308 neigh_add_timer(neigh,
1309 jiffies + NEIGH_VAR(neigh->parms, RETRANS_TIME));
1311 EXPORT_SYMBOL(__neigh_set_probe_once);
1313 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1314 u8 *lladdr, void *saddr,
1315 struct net_device *dev)
1317 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1318 lladdr || !dev->addr_len);
1320 neigh_update(neigh, lladdr, NUD_STALE,
1321 NEIGH_UPDATE_F_OVERRIDE, 0);
1324 EXPORT_SYMBOL(neigh_event_ns);
1326 /* called with read_lock_bh(&n->lock); */
1327 static void neigh_hh_init(struct neighbour *n)
1329 struct net_device *dev = n->dev;
1330 __be16 prot = n->tbl->protocol;
1331 struct hh_cache *hh = &n->hh;
1333 write_lock_bh(&n->lock);
1335 /* Only one thread can come in here and initialize the
1339 dev->header_ops->cache(n, hh, prot);
1341 write_unlock_bh(&n->lock);
1344 /* Slow and careful. */
1346 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1350 if (!neigh_event_send(neigh, skb)) {
1352 struct net_device *dev = neigh->dev;
1355 if (dev->header_ops->cache && !neigh->hh.hh_len)
1356 neigh_hh_init(neigh);
1359 __skb_pull(skb, skb_network_offset(skb));
1360 seq = read_seqbegin(&neigh->ha_lock);
1361 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1362 neigh->ha, NULL, skb->len);
1363 } while (read_seqretry(&neigh->ha_lock, seq));
1366 rc = dev_queue_xmit(skb);
1377 EXPORT_SYMBOL(neigh_resolve_output);
1379 /* As fast as possible without hh cache */
1381 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1383 struct net_device *dev = neigh->dev;
1388 __skb_pull(skb, skb_network_offset(skb));
1389 seq = read_seqbegin(&neigh->ha_lock);
1390 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1391 neigh->ha, NULL, skb->len);
1392 } while (read_seqretry(&neigh->ha_lock, seq));
1395 err = dev_queue_xmit(skb);
1402 EXPORT_SYMBOL(neigh_connected_output);
1404 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1406 return dev_queue_xmit(skb);
1408 EXPORT_SYMBOL(neigh_direct_output);
1410 static void neigh_proxy_process(struct timer_list *t)
1412 struct neigh_table *tbl = from_timer(tbl, t, proxy_timer);
1413 long sched_next = 0;
1414 unsigned long now = jiffies;
1415 struct sk_buff *skb, *n;
1417 spin_lock(&tbl->proxy_queue.lock);
1419 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1420 long tdif = NEIGH_CB(skb)->sched_next - now;
1423 struct net_device *dev = skb->dev;
1425 __skb_unlink(skb, &tbl->proxy_queue);
1426 if (tbl->proxy_redo && netif_running(dev)) {
1428 tbl->proxy_redo(skb);
1435 } else if (!sched_next || tdif < sched_next)
1438 del_timer(&tbl->proxy_timer);
1440 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1441 spin_unlock(&tbl->proxy_queue.lock);
1444 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1445 struct sk_buff *skb)
1447 unsigned long now = jiffies;
1449 unsigned long sched_next = now + (prandom_u32() %
1450 NEIGH_VAR(p, PROXY_DELAY));
1452 if (tbl->proxy_queue.qlen > NEIGH_VAR(p, PROXY_QLEN)) {
1457 NEIGH_CB(skb)->sched_next = sched_next;
1458 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1460 spin_lock(&tbl->proxy_queue.lock);
1461 if (del_timer(&tbl->proxy_timer)) {
1462 if (time_before(tbl->proxy_timer.expires, sched_next))
1463 sched_next = tbl->proxy_timer.expires;
1467 __skb_queue_tail(&tbl->proxy_queue, skb);
1468 mod_timer(&tbl->proxy_timer, sched_next);
1469 spin_unlock(&tbl->proxy_queue.lock);
1471 EXPORT_SYMBOL(pneigh_enqueue);
1473 static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1474 struct net *net, int ifindex)
1476 struct neigh_parms *p;
1478 list_for_each_entry(p, &tbl->parms_list, list) {
1479 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1480 (!p->dev && !ifindex && net_eq(net, &init_net)))
1487 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1488 struct neigh_table *tbl)
1490 struct neigh_parms *p;
1491 struct net *net = dev_net(dev);
1492 const struct net_device_ops *ops = dev->netdev_ops;
1494 p = kmemdup(&tbl->parms, sizeof(*p), GFP_KERNEL);
1497 refcount_set(&p->refcnt, 1);
1499 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
1502 write_pnet(&p->net, net);
1503 p->sysctl_table = NULL;
1505 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1511 write_lock_bh(&tbl->lock);
1512 list_add(&p->list, &tbl->parms.list);
1513 write_unlock_bh(&tbl->lock);
1515 neigh_parms_data_state_cleanall(p);
1519 EXPORT_SYMBOL(neigh_parms_alloc);
1521 static void neigh_rcu_free_parms(struct rcu_head *head)
1523 struct neigh_parms *parms =
1524 container_of(head, struct neigh_parms, rcu_head);
1526 neigh_parms_put(parms);
1529 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1531 if (!parms || parms == &tbl->parms)
1533 write_lock_bh(&tbl->lock);
1534 list_del(&parms->list);
1536 write_unlock_bh(&tbl->lock);
1538 dev_put(parms->dev);
1539 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1541 EXPORT_SYMBOL(neigh_parms_release);
1543 static void neigh_parms_destroy(struct neigh_parms *parms)
1548 static struct lock_class_key neigh_table_proxy_queue_class;
1550 static struct neigh_table *neigh_tables[NEIGH_NR_TABLES] __read_mostly;
1552 void neigh_table_init(int index, struct neigh_table *tbl)
1554 unsigned long now = jiffies;
1555 unsigned long phsize;
1557 INIT_LIST_HEAD(&tbl->parms_list);
1558 list_add(&tbl->parms.list, &tbl->parms_list);
1559 write_pnet(&tbl->parms.net, &init_net);
1560 refcount_set(&tbl->parms.refcnt, 1);
1561 tbl->parms.reachable_time =
1562 neigh_rand_reach_time(NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME));
1564 tbl->stats = alloc_percpu(struct neigh_statistics);
1566 panic("cannot create neighbour cache statistics");
1568 #ifdef CONFIG_PROC_FS
1569 if (!proc_create_seq_data(tbl->id, 0, init_net.proc_net_stat,
1570 &neigh_stat_seq_ops, tbl))
1571 panic("cannot create neighbour proc dir entry");
1574 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1576 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1577 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1579 if (!tbl->nht || !tbl->phash_buckets)
1580 panic("cannot allocate neighbour cache hashes");
1582 if (!tbl->entry_size)
1583 tbl->entry_size = ALIGN(offsetof(struct neighbour, primary_key) +
1584 tbl->key_len, NEIGH_PRIV_ALIGN);
1586 WARN_ON(tbl->entry_size % NEIGH_PRIV_ALIGN);
1588 rwlock_init(&tbl->lock);
1589 INIT_DEFERRABLE_WORK(&tbl->gc_work, neigh_periodic_work);
1590 queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
1591 tbl->parms.reachable_time);
1592 timer_setup(&tbl->proxy_timer, neigh_proxy_process, 0);
1593 skb_queue_head_init_class(&tbl->proxy_queue,
1594 &neigh_table_proxy_queue_class);
1596 tbl->last_flush = now;
1597 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1599 neigh_tables[index] = tbl;
1601 EXPORT_SYMBOL(neigh_table_init);
1603 int neigh_table_clear(int index, struct neigh_table *tbl)
1605 neigh_tables[index] = NULL;
1606 /* It is not clean... Fix it to unload IPv6 module safely */
1607 cancel_delayed_work_sync(&tbl->gc_work);
1608 del_timer_sync(&tbl->proxy_timer);
1609 pneigh_queue_purge(&tbl->proxy_queue);
1610 neigh_ifdown(tbl, NULL);
1611 if (atomic_read(&tbl->entries))
1612 pr_crit("neighbour leakage\n");
1614 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1615 neigh_hash_free_rcu);
1618 kfree(tbl->phash_buckets);
1619 tbl->phash_buckets = NULL;
1621 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1623 free_percpu(tbl->stats);
1628 EXPORT_SYMBOL(neigh_table_clear);
1630 static struct neigh_table *neigh_find_table(int family)
1632 struct neigh_table *tbl = NULL;
1636 tbl = neigh_tables[NEIGH_ARP_TABLE];
1639 tbl = neigh_tables[NEIGH_ND_TABLE];
1642 tbl = neigh_tables[NEIGH_DN_TABLE];
1649 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh,
1650 struct netlink_ext_ack *extack)
1652 struct net *net = sock_net(skb->sk);
1654 struct nlattr *dst_attr;
1655 struct neigh_table *tbl;
1656 struct neighbour *neigh;
1657 struct net_device *dev = NULL;
1661 if (nlmsg_len(nlh) < sizeof(*ndm))
1664 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1665 if (dst_attr == NULL)
1668 ndm = nlmsg_data(nlh);
1669 if (ndm->ndm_ifindex) {
1670 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1677 tbl = neigh_find_table(ndm->ndm_family);
1679 return -EAFNOSUPPORT;
1681 if (nla_len(dst_attr) < (int)tbl->key_len)
1684 if (ndm->ndm_flags & NTF_PROXY) {
1685 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1692 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1693 if (neigh == NULL) {
1698 err = neigh_update(neigh, NULL, NUD_FAILED,
1699 NEIGH_UPDATE_F_OVERRIDE |
1700 NEIGH_UPDATE_F_ADMIN,
1701 NETLINK_CB(skb).portid);
1702 write_lock_bh(&tbl->lock);
1703 neigh_release(neigh);
1704 neigh_remove_one(neigh, tbl);
1705 write_unlock_bh(&tbl->lock);
1711 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh,
1712 struct netlink_ext_ack *extack)
1714 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1715 struct net *net = sock_net(skb->sk);
1717 struct nlattr *tb[NDA_MAX+1];
1718 struct neigh_table *tbl;
1719 struct net_device *dev = NULL;
1720 struct neighbour *neigh;
1725 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
1730 if (tb[NDA_DST] == NULL)
1733 ndm = nlmsg_data(nlh);
1734 if (ndm->ndm_ifindex) {
1735 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1741 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1745 tbl = neigh_find_table(ndm->ndm_family);
1747 return -EAFNOSUPPORT;
1749 if (nla_len(tb[NDA_DST]) < (int)tbl->key_len)
1751 dst = nla_data(tb[NDA_DST]);
1752 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1754 if (ndm->ndm_flags & NTF_PROXY) {
1755 struct pneigh_entry *pn;
1758 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1760 pn->flags = ndm->ndm_flags;
1769 neigh = neigh_lookup(tbl, dst, dev);
1770 if (neigh == NULL) {
1771 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1776 neigh = __neigh_lookup_errno(tbl, dst, dev);
1777 if (IS_ERR(neigh)) {
1778 err = PTR_ERR(neigh);
1782 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1784 neigh_release(neigh);
1788 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1789 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1792 if (ndm->ndm_flags & NTF_EXT_LEARNED)
1793 flags |= NEIGH_UPDATE_F_EXT_LEARNED;
1795 if (ndm->ndm_flags & NTF_USE) {
1796 neigh_event_send(neigh, NULL);
1799 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags,
1800 NETLINK_CB(skb).portid);
1801 neigh_release(neigh);
1807 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1809 struct nlattr *nest;
1811 nest = nla_nest_start(skb, NDTA_PARMS);
1816 nla_put_u32(skb, NDTPA_IFINDEX, parms->dev->ifindex)) ||
1817 nla_put_u32(skb, NDTPA_REFCNT, refcount_read(&parms->refcnt)) ||
1818 nla_put_u32(skb, NDTPA_QUEUE_LENBYTES,
1819 NEIGH_VAR(parms, QUEUE_LEN_BYTES)) ||
1820 /* approximative value for deprecated QUEUE_LEN (in packets) */
1821 nla_put_u32(skb, NDTPA_QUEUE_LEN,
1822 NEIGH_VAR(parms, QUEUE_LEN_BYTES) / SKB_TRUESIZE(ETH_FRAME_LEN)) ||
1823 nla_put_u32(skb, NDTPA_PROXY_QLEN, NEIGH_VAR(parms, PROXY_QLEN)) ||
1824 nla_put_u32(skb, NDTPA_APP_PROBES, NEIGH_VAR(parms, APP_PROBES)) ||
1825 nla_put_u32(skb, NDTPA_UCAST_PROBES,
1826 NEIGH_VAR(parms, UCAST_PROBES)) ||
1827 nla_put_u32(skb, NDTPA_MCAST_PROBES,
1828 NEIGH_VAR(parms, MCAST_PROBES)) ||
1829 nla_put_u32(skb, NDTPA_MCAST_REPROBES,
1830 NEIGH_VAR(parms, MCAST_REPROBES)) ||
1831 nla_put_msecs(skb, NDTPA_REACHABLE_TIME, parms->reachable_time,
1833 nla_put_msecs(skb, NDTPA_BASE_REACHABLE_TIME,
1834 NEIGH_VAR(parms, BASE_REACHABLE_TIME), NDTPA_PAD) ||
1835 nla_put_msecs(skb, NDTPA_GC_STALETIME,
1836 NEIGH_VAR(parms, GC_STALETIME), NDTPA_PAD) ||
1837 nla_put_msecs(skb, NDTPA_DELAY_PROBE_TIME,
1838 NEIGH_VAR(parms, DELAY_PROBE_TIME), NDTPA_PAD) ||
1839 nla_put_msecs(skb, NDTPA_RETRANS_TIME,
1840 NEIGH_VAR(parms, RETRANS_TIME), NDTPA_PAD) ||
1841 nla_put_msecs(skb, NDTPA_ANYCAST_DELAY,
1842 NEIGH_VAR(parms, ANYCAST_DELAY), NDTPA_PAD) ||
1843 nla_put_msecs(skb, NDTPA_PROXY_DELAY,
1844 NEIGH_VAR(parms, PROXY_DELAY), NDTPA_PAD) ||
1845 nla_put_msecs(skb, NDTPA_LOCKTIME,
1846 NEIGH_VAR(parms, LOCKTIME), NDTPA_PAD))
1847 goto nla_put_failure;
1848 return nla_nest_end(skb, nest);
1851 nla_nest_cancel(skb, nest);
1855 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1856 u32 pid, u32 seq, int type, int flags)
1858 struct nlmsghdr *nlh;
1859 struct ndtmsg *ndtmsg;
1861 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1865 ndtmsg = nlmsg_data(nlh);
1867 read_lock_bh(&tbl->lock);
1868 ndtmsg->ndtm_family = tbl->family;
1869 ndtmsg->ndtm_pad1 = 0;
1870 ndtmsg->ndtm_pad2 = 0;
1872 if (nla_put_string(skb, NDTA_NAME, tbl->id) ||
1873 nla_put_msecs(skb, NDTA_GC_INTERVAL, tbl->gc_interval, NDTA_PAD) ||
1874 nla_put_u32(skb, NDTA_THRESH1, tbl->gc_thresh1) ||
1875 nla_put_u32(skb, NDTA_THRESH2, tbl->gc_thresh2) ||
1876 nla_put_u32(skb, NDTA_THRESH3, tbl->gc_thresh3))
1877 goto nla_put_failure;
1879 unsigned long now = jiffies;
1880 unsigned int flush_delta = now - tbl->last_flush;
1881 unsigned int rand_delta = now - tbl->last_rand;
1882 struct neigh_hash_table *nht;
1883 struct ndt_config ndc = {
1884 .ndtc_key_len = tbl->key_len,
1885 .ndtc_entry_size = tbl->entry_size,
1886 .ndtc_entries = atomic_read(&tbl->entries),
1887 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1888 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1889 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1893 nht = rcu_dereference_bh(tbl->nht);
1894 ndc.ndtc_hash_rnd = nht->hash_rnd[0];
1895 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1896 rcu_read_unlock_bh();
1898 if (nla_put(skb, NDTA_CONFIG, sizeof(ndc), &ndc))
1899 goto nla_put_failure;
1904 struct ndt_stats ndst;
1906 memset(&ndst, 0, sizeof(ndst));
1908 for_each_possible_cpu(cpu) {
1909 struct neigh_statistics *st;
1911 st = per_cpu_ptr(tbl->stats, cpu);
1912 ndst.ndts_allocs += st->allocs;
1913 ndst.ndts_destroys += st->destroys;
1914 ndst.ndts_hash_grows += st->hash_grows;
1915 ndst.ndts_res_failed += st->res_failed;
1916 ndst.ndts_lookups += st->lookups;
1917 ndst.ndts_hits += st->hits;
1918 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1919 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1920 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1921 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1922 ndst.ndts_table_fulls += st->table_fulls;
1925 if (nla_put_64bit(skb, NDTA_STATS, sizeof(ndst), &ndst,
1927 goto nla_put_failure;
1930 BUG_ON(tbl->parms.dev);
1931 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1932 goto nla_put_failure;
1934 read_unlock_bh(&tbl->lock);
1935 nlmsg_end(skb, nlh);
1939 read_unlock_bh(&tbl->lock);
1940 nlmsg_cancel(skb, nlh);
1944 static int neightbl_fill_param_info(struct sk_buff *skb,
1945 struct neigh_table *tbl,
1946 struct neigh_parms *parms,
1947 u32 pid, u32 seq, int type,
1950 struct ndtmsg *ndtmsg;
1951 struct nlmsghdr *nlh;
1953 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1957 ndtmsg = nlmsg_data(nlh);
1959 read_lock_bh(&tbl->lock);
1960 ndtmsg->ndtm_family = tbl->family;
1961 ndtmsg->ndtm_pad1 = 0;
1962 ndtmsg->ndtm_pad2 = 0;
1964 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1965 neightbl_fill_parms(skb, parms) < 0)
1968 read_unlock_bh(&tbl->lock);
1969 nlmsg_end(skb, nlh);
1972 read_unlock_bh(&tbl->lock);
1973 nlmsg_cancel(skb, nlh);
1977 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1978 [NDTA_NAME] = { .type = NLA_STRING },
1979 [NDTA_THRESH1] = { .type = NLA_U32 },
1980 [NDTA_THRESH2] = { .type = NLA_U32 },
1981 [NDTA_THRESH3] = { .type = NLA_U32 },
1982 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1983 [NDTA_PARMS] = { .type = NLA_NESTED },
1986 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1987 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1988 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1989 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1990 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1991 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1992 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1993 [NDTPA_MCAST_REPROBES] = { .type = NLA_U32 },
1994 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1995 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1996 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1997 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1998 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1999 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
2000 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
2003 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
2004 struct netlink_ext_ack *extack)
2006 struct net *net = sock_net(skb->sk);
2007 struct neigh_table *tbl;
2008 struct ndtmsg *ndtmsg;
2009 struct nlattr *tb[NDTA_MAX+1];
2013 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
2014 nl_neightbl_policy, extack);
2018 if (tb[NDTA_NAME] == NULL) {
2023 ndtmsg = nlmsg_data(nlh);
2025 for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2026 tbl = neigh_tables[tidx];
2029 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
2031 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0) {
2041 * We acquire tbl->lock to be nice to the periodic timers and
2042 * make sure they always see a consistent set of values.
2044 write_lock_bh(&tbl->lock);
2046 if (tb[NDTA_PARMS]) {
2047 struct nlattr *tbp[NDTPA_MAX+1];
2048 struct neigh_parms *p;
2051 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
2052 nl_ntbl_parm_policy, extack);
2054 goto errout_tbl_lock;
2056 if (tbp[NDTPA_IFINDEX])
2057 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
2059 p = lookup_neigh_parms(tbl, net, ifindex);
2062 goto errout_tbl_lock;
2065 for (i = 1; i <= NDTPA_MAX; i++) {
2070 case NDTPA_QUEUE_LEN:
2071 NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2072 nla_get_u32(tbp[i]) *
2073 SKB_TRUESIZE(ETH_FRAME_LEN));
2075 case NDTPA_QUEUE_LENBYTES:
2076 NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2077 nla_get_u32(tbp[i]));
2079 case NDTPA_PROXY_QLEN:
2080 NEIGH_VAR_SET(p, PROXY_QLEN,
2081 nla_get_u32(tbp[i]));
2083 case NDTPA_APP_PROBES:
2084 NEIGH_VAR_SET(p, APP_PROBES,
2085 nla_get_u32(tbp[i]));
2087 case NDTPA_UCAST_PROBES:
2088 NEIGH_VAR_SET(p, UCAST_PROBES,
2089 nla_get_u32(tbp[i]));
2091 case NDTPA_MCAST_PROBES:
2092 NEIGH_VAR_SET(p, MCAST_PROBES,
2093 nla_get_u32(tbp[i]));
2095 case NDTPA_MCAST_REPROBES:
2096 NEIGH_VAR_SET(p, MCAST_REPROBES,
2097 nla_get_u32(tbp[i]));
2099 case NDTPA_BASE_REACHABLE_TIME:
2100 NEIGH_VAR_SET(p, BASE_REACHABLE_TIME,
2101 nla_get_msecs(tbp[i]));
2102 /* update reachable_time as well, otherwise, the change will
2103 * only be effective after the next time neigh_periodic_work
2104 * decides to recompute it (can be multiple minutes)
2107 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
2109 case NDTPA_GC_STALETIME:
2110 NEIGH_VAR_SET(p, GC_STALETIME,
2111 nla_get_msecs(tbp[i]));
2113 case NDTPA_DELAY_PROBE_TIME:
2114 NEIGH_VAR_SET(p, DELAY_PROBE_TIME,
2115 nla_get_msecs(tbp[i]));
2116 call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
2118 case NDTPA_RETRANS_TIME:
2119 NEIGH_VAR_SET(p, RETRANS_TIME,
2120 nla_get_msecs(tbp[i]));
2122 case NDTPA_ANYCAST_DELAY:
2123 NEIGH_VAR_SET(p, ANYCAST_DELAY,
2124 nla_get_msecs(tbp[i]));
2126 case NDTPA_PROXY_DELAY:
2127 NEIGH_VAR_SET(p, PROXY_DELAY,
2128 nla_get_msecs(tbp[i]));
2130 case NDTPA_LOCKTIME:
2131 NEIGH_VAR_SET(p, LOCKTIME,
2132 nla_get_msecs(tbp[i]));
2139 if ((tb[NDTA_THRESH1] || tb[NDTA_THRESH2] ||
2140 tb[NDTA_THRESH3] || tb[NDTA_GC_INTERVAL]) &&
2141 !net_eq(net, &init_net))
2142 goto errout_tbl_lock;
2144 if (tb[NDTA_THRESH1])
2145 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2147 if (tb[NDTA_THRESH2])
2148 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2150 if (tb[NDTA_THRESH3])
2151 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2153 if (tb[NDTA_GC_INTERVAL])
2154 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2159 write_unlock_bh(&tbl->lock);
2164 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2166 struct net *net = sock_net(skb->sk);
2167 int family, tidx, nidx = 0;
2168 int tbl_skip = cb->args[0];
2169 int neigh_skip = cb->args[1];
2170 struct neigh_table *tbl;
2172 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2174 for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2175 struct neigh_parms *p;
2177 tbl = neigh_tables[tidx];
2181 if (tidx < tbl_skip || (family && tbl->family != family))
2184 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).portid,
2185 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2190 p = list_next_entry(&tbl->parms, list);
2191 list_for_each_entry_from(p, &tbl->parms_list, list) {
2192 if (!net_eq(neigh_parms_net(p), net))
2195 if (nidx < neigh_skip)
2198 if (neightbl_fill_param_info(skb, tbl, p,
2199 NETLINK_CB(cb->skb).portid,
2217 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2218 u32 pid, u32 seq, int type, unsigned int flags)
2220 unsigned long now = jiffies;
2221 struct nda_cacheinfo ci;
2222 struct nlmsghdr *nlh;
2225 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2229 ndm = nlmsg_data(nlh);
2230 ndm->ndm_family = neigh->ops->family;
2233 ndm->ndm_flags = neigh->flags;
2234 ndm->ndm_type = neigh->type;
2235 ndm->ndm_ifindex = neigh->dev->ifindex;
2237 if (nla_put(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key))
2238 goto nla_put_failure;
2240 read_lock_bh(&neigh->lock);
2241 ndm->ndm_state = neigh->nud_state;
2242 if (neigh->nud_state & NUD_VALID) {
2243 char haddr[MAX_ADDR_LEN];
2245 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2246 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2247 read_unlock_bh(&neigh->lock);
2248 goto nla_put_failure;
2252 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2253 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2254 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2255 ci.ndm_refcnt = refcount_read(&neigh->refcnt) - 1;
2256 read_unlock_bh(&neigh->lock);
2258 if (nla_put_u32(skb, NDA_PROBES, atomic_read(&neigh->probes)) ||
2259 nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
2260 goto nla_put_failure;
2262 nlmsg_end(skb, nlh);
2266 nlmsg_cancel(skb, nlh);
2270 static int pneigh_fill_info(struct sk_buff *skb, struct pneigh_entry *pn,
2271 u32 pid, u32 seq, int type, unsigned int flags,
2272 struct neigh_table *tbl)
2274 struct nlmsghdr *nlh;
2277 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2281 ndm = nlmsg_data(nlh);
2282 ndm->ndm_family = tbl->family;
2285 ndm->ndm_flags = pn->flags | NTF_PROXY;
2286 ndm->ndm_type = RTN_UNICAST;
2287 ndm->ndm_ifindex = pn->dev ? pn->dev->ifindex : 0;
2288 ndm->ndm_state = NUD_NONE;
2290 if (nla_put(skb, NDA_DST, tbl->key_len, pn->key))
2291 goto nla_put_failure;
2293 nlmsg_end(skb, nlh);
2297 nlmsg_cancel(skb, nlh);
2301 static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid)
2303 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2304 __neigh_notify(neigh, RTM_NEWNEIGH, 0, nlmsg_pid);
2307 static bool neigh_master_filtered(struct net_device *dev, int master_idx)
2309 struct net_device *master;
2314 master = netdev_master_upper_dev_get(dev);
2315 if (!master || master->ifindex != master_idx)
2321 static bool neigh_ifindex_filtered(struct net_device *dev, int filter_idx)
2323 if (filter_idx && dev->ifindex != filter_idx)
2329 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2330 struct netlink_callback *cb)
2332 struct net *net = sock_net(skb->sk);
2333 const struct nlmsghdr *nlh = cb->nlh;
2334 struct nlattr *tb[NDA_MAX + 1];
2335 struct neighbour *n;
2336 int rc, h, s_h = cb->args[1];
2337 int idx, s_idx = idx = cb->args[2];
2338 struct neigh_hash_table *nht;
2339 int filter_master_idx = 0, filter_idx = 0;
2340 unsigned int flags = NLM_F_MULTI;
2343 err = nlmsg_parse(nlh, sizeof(struct ndmsg), tb, NDA_MAX, NULL, NULL);
2345 if (tb[NDA_IFINDEX]) {
2346 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
2348 filter_idx = nla_get_u32(tb[NDA_IFINDEX]);
2350 if (tb[NDA_MASTER]) {
2351 if (nla_len(tb[NDA_MASTER]) != sizeof(u32))
2353 filter_master_idx = nla_get_u32(tb[NDA_MASTER]);
2355 if (filter_idx || filter_master_idx)
2356 flags |= NLM_F_DUMP_FILTERED;
2360 nht = rcu_dereference_bh(tbl->nht);
2362 for (h = s_h; h < (1 << nht->hash_shift); h++) {
2365 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2367 n = rcu_dereference_bh(n->next)) {
2368 if (idx < s_idx || !net_eq(dev_net(n->dev), net))
2370 if (neigh_ifindex_filtered(n->dev, filter_idx) ||
2371 neigh_master_filtered(n->dev, filter_master_idx))
2373 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2386 rcu_read_unlock_bh();
2392 static int pneigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2393 struct netlink_callback *cb)
2395 struct pneigh_entry *n;
2396 struct net *net = sock_net(skb->sk);
2397 int rc, h, s_h = cb->args[3];
2398 int idx, s_idx = idx = cb->args[4];
2400 read_lock_bh(&tbl->lock);
2402 for (h = s_h; h <= PNEIGH_HASHMASK; h++) {
2405 for (n = tbl->phash_buckets[h], idx = 0; n; n = n->next) {
2406 if (idx < s_idx || pneigh_net(n) != net)
2408 if (pneigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2411 NLM_F_MULTI, tbl) < 0) {
2412 read_unlock_bh(&tbl->lock);
2421 read_unlock_bh(&tbl->lock);
2430 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2432 struct neigh_table *tbl;
2437 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2439 /* check for full ndmsg structure presence, family member is
2440 * the same for both structures
2442 if (nlmsg_len(cb->nlh) >= sizeof(struct ndmsg) &&
2443 ((struct ndmsg *) nlmsg_data(cb->nlh))->ndm_flags == NTF_PROXY)
2448 for (t = 0; t < NEIGH_NR_TABLES; t++) {
2449 tbl = neigh_tables[t];
2453 if (t < s_t || (family && tbl->family != family))
2456 memset(&cb->args[1], 0, sizeof(cb->args) -
2457 sizeof(cb->args[0]));
2459 err = pneigh_dump_table(tbl, skb, cb);
2461 err = neigh_dump_table(tbl, skb, cb);
2470 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2473 struct neigh_hash_table *nht;
2476 nht = rcu_dereference_bh(tbl->nht);
2478 read_lock(&tbl->lock); /* avoid resizes */
2479 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2480 struct neighbour *n;
2482 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2484 n = rcu_dereference_bh(n->next))
2487 read_unlock(&tbl->lock);
2488 rcu_read_unlock_bh();
2490 EXPORT_SYMBOL(neigh_for_each);
2492 /* The tbl->lock must be held as a writer and BH disabled. */
2493 void __neigh_for_each_release(struct neigh_table *tbl,
2494 int (*cb)(struct neighbour *))
2497 struct neigh_hash_table *nht;
2499 nht = rcu_dereference_protected(tbl->nht,
2500 lockdep_is_held(&tbl->lock));
2501 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2502 struct neighbour *n;
2503 struct neighbour __rcu **np;
2505 np = &nht->hash_buckets[chain];
2506 while ((n = rcu_dereference_protected(*np,
2507 lockdep_is_held(&tbl->lock))) != NULL) {
2510 write_lock(&n->lock);
2513 rcu_assign_pointer(*np,
2514 rcu_dereference_protected(n->next,
2515 lockdep_is_held(&tbl->lock)));
2519 write_unlock(&n->lock);
2521 neigh_cleanup_and_release(n);
2525 EXPORT_SYMBOL(__neigh_for_each_release);
2527 int neigh_xmit(int index, struct net_device *dev,
2528 const void *addr, struct sk_buff *skb)
2530 int err = -EAFNOSUPPORT;
2531 if (likely(index < NEIGH_NR_TABLES)) {
2532 struct neigh_table *tbl;
2533 struct neighbour *neigh;
2535 tbl = neigh_tables[index];
2539 neigh = __neigh_lookup_noref(tbl, addr, dev);
2541 neigh = __neigh_create(tbl, addr, dev, false);
2542 err = PTR_ERR(neigh);
2543 if (IS_ERR(neigh)) {
2544 rcu_read_unlock_bh();
2547 err = neigh->output(neigh, skb);
2548 rcu_read_unlock_bh();
2550 else if (index == NEIGH_LINK_TABLE) {
2551 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
2552 addr, NULL, skb->len);
2555 err = dev_queue_xmit(skb);
2563 EXPORT_SYMBOL(neigh_xmit);
2565 #ifdef CONFIG_PROC_FS
2567 static struct neighbour *neigh_get_first(struct seq_file *seq)
2569 struct neigh_seq_state *state = seq->private;
2570 struct net *net = seq_file_net(seq);
2571 struct neigh_hash_table *nht = state->nht;
2572 struct neighbour *n = NULL;
2573 int bucket = state->bucket;
2575 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2576 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2577 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2580 if (!net_eq(dev_net(n->dev), net))
2582 if (state->neigh_sub_iter) {
2586 v = state->neigh_sub_iter(state, n, &fakep);
2590 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2592 if (n->nud_state & ~NUD_NOARP)
2595 n = rcu_dereference_bh(n->next);
2601 state->bucket = bucket;
2606 static struct neighbour *neigh_get_next(struct seq_file *seq,
2607 struct neighbour *n,
2610 struct neigh_seq_state *state = seq->private;
2611 struct net *net = seq_file_net(seq);
2612 struct neigh_hash_table *nht = state->nht;
2614 if (state->neigh_sub_iter) {
2615 void *v = state->neigh_sub_iter(state, n, pos);
2619 n = rcu_dereference_bh(n->next);
2623 if (!net_eq(dev_net(n->dev), net))
2625 if (state->neigh_sub_iter) {
2626 void *v = state->neigh_sub_iter(state, n, pos);
2631 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2634 if (n->nud_state & ~NUD_NOARP)
2637 n = rcu_dereference_bh(n->next);
2643 if (++state->bucket >= (1 << nht->hash_shift))
2646 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2654 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2656 struct neighbour *n = neigh_get_first(seq);
2661 n = neigh_get_next(seq, n, pos);
2666 return *pos ? NULL : n;
2669 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2671 struct neigh_seq_state *state = seq->private;
2672 struct net *net = seq_file_net(seq);
2673 struct neigh_table *tbl = state->tbl;
2674 struct pneigh_entry *pn = NULL;
2675 int bucket = state->bucket;
2677 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2678 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2679 pn = tbl->phash_buckets[bucket];
2680 while (pn && !net_eq(pneigh_net(pn), net))
2685 state->bucket = bucket;
2690 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2691 struct pneigh_entry *pn,
2694 struct neigh_seq_state *state = seq->private;
2695 struct net *net = seq_file_net(seq);
2696 struct neigh_table *tbl = state->tbl;
2700 } while (pn && !net_eq(pneigh_net(pn), net));
2703 if (++state->bucket > PNEIGH_HASHMASK)
2705 pn = tbl->phash_buckets[state->bucket];
2706 while (pn && !net_eq(pneigh_net(pn), net))
2718 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2720 struct pneigh_entry *pn = pneigh_get_first(seq);
2725 pn = pneigh_get_next(seq, pn, pos);
2730 return *pos ? NULL : pn;
2733 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2735 struct neigh_seq_state *state = seq->private;
2737 loff_t idxpos = *pos;
2739 rc = neigh_get_idx(seq, &idxpos);
2740 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2741 rc = pneigh_get_idx(seq, &idxpos);
2746 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2749 struct neigh_seq_state *state = seq->private;
2753 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2756 state->nht = rcu_dereference_bh(tbl->nht);
2758 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2760 EXPORT_SYMBOL(neigh_seq_start);
2762 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2764 struct neigh_seq_state *state;
2767 if (v == SEQ_START_TOKEN) {
2768 rc = neigh_get_first(seq);
2772 state = seq->private;
2773 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2774 rc = neigh_get_next(seq, v, NULL);
2777 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2778 rc = pneigh_get_first(seq);
2780 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2781 rc = pneigh_get_next(seq, v, NULL);
2787 EXPORT_SYMBOL(neigh_seq_next);
2789 void neigh_seq_stop(struct seq_file *seq, void *v)
2792 rcu_read_unlock_bh();
2794 EXPORT_SYMBOL(neigh_seq_stop);
2796 /* statistics via seq_file */
2798 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2800 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
2804 return SEQ_START_TOKEN;
2806 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2807 if (!cpu_possible(cpu))
2810 return per_cpu_ptr(tbl->stats, cpu);
2815 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2817 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
2820 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2821 if (!cpu_possible(cpu))
2824 return per_cpu_ptr(tbl->stats, cpu);
2829 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2834 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2836 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
2837 struct neigh_statistics *st = v;
2839 if (v == SEQ_START_TOKEN) {
2840 seq_printf(seq, "entries allocs destroys hash_grows lookups hits res_failed rcv_probes_mcast rcv_probes_ucast periodic_gc_runs forced_gc_runs unresolved_discards table_fulls\n");
2844 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2845 "%08lx %08lx %08lx %08lx %08lx %08lx\n",
2846 atomic_read(&tbl->entries),
2857 st->rcv_probes_mcast,
2858 st->rcv_probes_ucast,
2860 st->periodic_gc_runs,
2869 static const struct seq_operations neigh_stat_seq_ops = {
2870 .start = neigh_stat_seq_start,
2871 .next = neigh_stat_seq_next,
2872 .stop = neigh_stat_seq_stop,
2873 .show = neigh_stat_seq_show,
2875 #endif /* CONFIG_PROC_FS */
2877 static inline size_t neigh_nlmsg_size(void)
2879 return NLMSG_ALIGN(sizeof(struct ndmsg))
2880 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2881 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2882 + nla_total_size(sizeof(struct nda_cacheinfo))
2883 + nla_total_size(4); /* NDA_PROBES */
2886 static void __neigh_notify(struct neighbour *n, int type, int flags,
2889 struct net *net = dev_net(n->dev);
2890 struct sk_buff *skb;
2893 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2897 err = neigh_fill_info(skb, n, pid, 0, type, flags);
2899 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2900 WARN_ON(err == -EMSGSIZE);
2904 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2908 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2911 void neigh_app_ns(struct neighbour *n)
2913 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST, 0);
2915 EXPORT_SYMBOL(neigh_app_ns);
2917 #ifdef CONFIG_SYSCTL
2919 static int int_max = INT_MAX;
2920 static int unres_qlen_max = INT_MAX / SKB_TRUESIZE(ETH_FRAME_LEN);
2922 static int proc_unres_qlen(struct ctl_table *ctl, int write,
2923 void __user *buffer, size_t *lenp, loff_t *ppos)
2926 struct ctl_table tmp = *ctl;
2929 tmp.extra2 = &unres_qlen_max;
2932 size = *(int *)ctl->data / SKB_TRUESIZE(ETH_FRAME_LEN);
2933 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2936 *(int *)ctl->data = size * SKB_TRUESIZE(ETH_FRAME_LEN);
2940 static struct neigh_parms *neigh_get_dev_parms_rcu(struct net_device *dev,
2945 return __in_dev_arp_parms_get_rcu(dev);
2947 return __in6_dev_nd_parms_get_rcu(dev);
2952 static void neigh_copy_dflt_parms(struct net *net, struct neigh_parms *p,
2955 struct net_device *dev;
2956 int family = neigh_parms_family(p);
2959 for_each_netdev_rcu(net, dev) {
2960 struct neigh_parms *dst_p =
2961 neigh_get_dev_parms_rcu(dev, family);
2963 if (dst_p && !test_bit(index, dst_p->data_state))
2964 dst_p->data[index] = p->data[index];
2969 static void neigh_proc_update(struct ctl_table *ctl, int write)
2971 struct net_device *dev = ctl->extra1;
2972 struct neigh_parms *p = ctl->extra2;
2973 struct net *net = neigh_parms_net(p);
2974 int index = (int *) ctl->data - p->data;
2979 set_bit(index, p->data_state);
2980 if (index == NEIGH_VAR_DELAY_PROBE_TIME)
2981 call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
2982 if (!dev) /* NULL dev means this is default value */
2983 neigh_copy_dflt_parms(net, p, index);
2986 static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write,
2987 void __user *buffer,
2988 size_t *lenp, loff_t *ppos)
2990 struct ctl_table tmp = *ctl;
2994 tmp.extra2 = &int_max;
2996 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2997 neigh_proc_update(ctl, write);
3001 int neigh_proc_dointvec(struct ctl_table *ctl, int write,
3002 void __user *buffer, size_t *lenp, loff_t *ppos)
3004 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
3006 neigh_proc_update(ctl, write);
3009 EXPORT_SYMBOL(neigh_proc_dointvec);
3011 int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write,
3012 void __user *buffer,
3013 size_t *lenp, loff_t *ppos)
3015 int ret = proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3017 neigh_proc_update(ctl, write);
3020 EXPORT_SYMBOL(neigh_proc_dointvec_jiffies);
3022 static int neigh_proc_dointvec_userhz_jiffies(struct ctl_table *ctl, int write,
3023 void __user *buffer,
3024 size_t *lenp, loff_t *ppos)
3026 int ret = proc_dointvec_userhz_jiffies(ctl, write, buffer, lenp, ppos);
3028 neigh_proc_update(ctl, write);
3032 int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write,
3033 void __user *buffer,
3034 size_t *lenp, loff_t *ppos)
3036 int ret = proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3038 neigh_proc_update(ctl, write);
3041 EXPORT_SYMBOL(neigh_proc_dointvec_ms_jiffies);
3043 static int neigh_proc_dointvec_unres_qlen(struct ctl_table *ctl, int write,
3044 void __user *buffer,
3045 size_t *lenp, loff_t *ppos)
3047 int ret = proc_unres_qlen(ctl, write, buffer, lenp, ppos);
3049 neigh_proc_update(ctl, write);
3053 static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,
3054 void __user *buffer,
3055 size_t *lenp, loff_t *ppos)
3057 struct neigh_parms *p = ctl->extra2;
3060 if (strcmp(ctl->procname, "base_reachable_time") == 0)
3061 ret = neigh_proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3062 else if (strcmp(ctl->procname, "base_reachable_time_ms") == 0)
3063 ret = neigh_proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3067 if (write && ret == 0) {
3068 /* update reachable_time as well, otherwise, the change will
3069 * only be effective after the next time neigh_periodic_work
3070 * decides to recompute it
3073 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
3078 #define NEIGH_PARMS_DATA_OFFSET(index) \
3079 (&((struct neigh_parms *) 0)->data[index])
3081 #define NEIGH_SYSCTL_ENTRY(attr, data_attr, name, mval, proc) \
3082 [NEIGH_VAR_ ## attr] = { \
3084 .data = NEIGH_PARMS_DATA_OFFSET(NEIGH_VAR_ ## data_attr), \
3085 .maxlen = sizeof(int), \
3087 .proc_handler = proc, \
3090 #define NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(attr, name) \
3091 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_zero_intmax)
3093 #define NEIGH_SYSCTL_JIFFIES_ENTRY(attr, name) \
3094 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_jiffies)
3096 #define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(attr, name) \
3097 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)
3099 #define NEIGH_SYSCTL_MS_JIFFIES_ENTRY(attr, name) \
3100 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3102 #define NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(attr, data_attr, name) \
3103 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3105 #define NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(attr, data_attr, name) \
3106 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_unres_qlen)
3108 static struct neigh_sysctl_table {
3109 struct ctl_table_header *sysctl_header;
3110 struct ctl_table neigh_vars[NEIGH_VAR_MAX + 1];
3111 } neigh_sysctl_template __read_mostly = {
3113 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_PROBES, "mcast_solicit"),
3114 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(UCAST_PROBES, "ucast_solicit"),
3115 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(APP_PROBES, "app_solicit"),
3116 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_REPROBES, "mcast_resolicit"),
3117 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(RETRANS_TIME, "retrans_time"),
3118 NEIGH_SYSCTL_JIFFIES_ENTRY(BASE_REACHABLE_TIME, "base_reachable_time"),
3119 NEIGH_SYSCTL_JIFFIES_ENTRY(DELAY_PROBE_TIME, "delay_first_probe_time"),
3120 NEIGH_SYSCTL_JIFFIES_ENTRY(GC_STALETIME, "gc_stale_time"),
3121 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(QUEUE_LEN_BYTES, "unres_qlen_bytes"),
3122 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(PROXY_QLEN, "proxy_qlen"),
3123 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(ANYCAST_DELAY, "anycast_delay"),
3124 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(PROXY_DELAY, "proxy_delay"),
3125 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(LOCKTIME, "locktime"),
3126 NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(QUEUE_LEN, QUEUE_LEN_BYTES, "unres_qlen"),
3127 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(RETRANS_TIME_MS, RETRANS_TIME, "retrans_time_ms"),
3128 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(BASE_REACHABLE_TIME_MS, BASE_REACHABLE_TIME, "base_reachable_time_ms"),
3129 [NEIGH_VAR_GC_INTERVAL] = {
3130 .procname = "gc_interval",
3131 .maxlen = sizeof(int),
3133 .proc_handler = proc_dointvec_jiffies,
3135 [NEIGH_VAR_GC_THRESH1] = {
3136 .procname = "gc_thresh1",
3137 .maxlen = sizeof(int),
3141 .proc_handler = proc_dointvec_minmax,
3143 [NEIGH_VAR_GC_THRESH2] = {
3144 .procname = "gc_thresh2",
3145 .maxlen = sizeof(int),
3149 .proc_handler = proc_dointvec_minmax,
3151 [NEIGH_VAR_GC_THRESH3] = {
3152 .procname = "gc_thresh3",
3153 .maxlen = sizeof(int),
3157 .proc_handler = proc_dointvec_minmax,
3163 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
3164 proc_handler *handler)
3167 struct neigh_sysctl_table *t;
3168 const char *dev_name_source;
3169 char neigh_path[ sizeof("net//neigh/") + IFNAMSIZ + IFNAMSIZ ];
3172 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
3176 for (i = 0; i < NEIGH_VAR_GC_INTERVAL; i++) {
3177 t->neigh_vars[i].data += (long) p;
3178 t->neigh_vars[i].extra1 = dev;
3179 t->neigh_vars[i].extra2 = p;
3183 dev_name_source = dev->name;
3184 /* Terminate the table early */
3185 memset(&t->neigh_vars[NEIGH_VAR_GC_INTERVAL], 0,
3186 sizeof(t->neigh_vars[NEIGH_VAR_GC_INTERVAL]));
3188 struct neigh_table *tbl = p->tbl;
3189 dev_name_source = "default";
3190 t->neigh_vars[NEIGH_VAR_GC_INTERVAL].data = &tbl->gc_interval;
3191 t->neigh_vars[NEIGH_VAR_GC_THRESH1].data = &tbl->gc_thresh1;
3192 t->neigh_vars[NEIGH_VAR_GC_THRESH2].data = &tbl->gc_thresh2;
3193 t->neigh_vars[NEIGH_VAR_GC_THRESH3].data = &tbl->gc_thresh3;
3198 t->neigh_vars[NEIGH_VAR_RETRANS_TIME].proc_handler = handler;
3200 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler = handler;
3201 /* RetransTime (in milliseconds)*/
3202 t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].proc_handler = handler;
3203 /* ReachableTime (in milliseconds) */
3204 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler = handler;
3206 /* Those handlers will update p->reachable_time after
3207 * base_reachable_time(_ms) is set to ensure the new timer starts being
3208 * applied after the next neighbour update instead of waiting for
3209 * neigh_periodic_work to update its value (can be multiple minutes)
3210 * So any handler that replaces them should do this as well
3213 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler =
3214 neigh_proc_base_reachable_time;
3215 /* ReachableTime (in milliseconds) */
3216 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler =
3217 neigh_proc_base_reachable_time;
3220 /* Don't export sysctls to unprivileged users */
3221 if (neigh_parms_net(p)->user_ns != &init_user_ns)
3222 t->neigh_vars[0].procname = NULL;
3224 switch (neigh_parms_family(p)) {
3235 snprintf(neigh_path, sizeof(neigh_path), "net/%s/neigh/%s",
3236 p_name, dev_name_source);
3238 register_net_sysctl(neigh_parms_net(p), neigh_path, t->neigh_vars);
3239 if (!t->sysctl_header)
3242 p->sysctl_table = t;
3250 EXPORT_SYMBOL(neigh_sysctl_register);
3252 void neigh_sysctl_unregister(struct neigh_parms *p)
3254 if (p->sysctl_table) {
3255 struct neigh_sysctl_table *t = p->sysctl_table;
3256 p->sysctl_table = NULL;
3257 unregister_net_sysctl_table(t->sysctl_header);
3261 EXPORT_SYMBOL(neigh_sysctl_unregister);
3263 #endif /* CONFIG_SYSCTL */
3265 static int __init neigh_init(void)
3267 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, 0);
3268 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, 0);
3269 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, 0);
3271 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
3273 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, 0);
3278 subsys_initcall(neigh_init);