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_PRINTK1 NEIGH_NOPRINTK
45 #define NEIGH_PRINTK2 NEIGH_NOPRINTK
49 #define NEIGH_PRINTK1 NEIGH_PRINTK
53 #define NEIGH_PRINTK2 NEIGH_PRINTK
56 #define PNEIGH_HASHMASK 0xF
58 static void neigh_timer_handler(unsigned long arg);
59 static void __neigh_notify(struct neighbour *n, int type, int flags);
60 static void neigh_update_notify(struct neighbour *neigh);
61 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
63 static struct neigh_table *neigh_tables;
65 static const struct file_operations neigh_stat_seq_fops;
69 Neighbour hash table buckets are protected with rwlock tbl->lock.
71 - All the scans/updates to hash buckets MUST be made under this lock.
72 - NOTHING clever should be made under this lock: no callbacks
73 to protocol backends, no attempts to send something to network.
74 It will result in deadlocks, if backend/driver wants to use neighbour
76 - If the entry requires some non-trivial actions, increase
77 its reference count and release table lock.
79 Neighbour entries are protected:
80 - with reference count.
81 - with rwlock neigh->lock
83 Reference count prevents destruction.
85 neigh->lock mainly serializes ll address data and its validity state.
86 However, the same lock is used to protect another entry fields:
90 Again, nothing clever shall be made under neigh->lock,
91 the most complicated procedure, which we allow is dev->hard_header.
92 It is supposed, that dev->hard_header is simplistic and does
93 not make callbacks to neighbour tables.
95 The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
96 list of neighbour tables. This list is used only in process context,
99 static DEFINE_RWLOCK(neigh_tbl_lock);
101 static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
107 static void neigh_cleanup_and_release(struct neighbour *neigh)
109 if (neigh->parms->neigh_cleanup)
110 neigh->parms->neigh_cleanup(neigh);
112 __neigh_notify(neigh, RTM_DELNEIGH, 0);
113 neigh_release(neigh);
117 * It is random distribution in the interval (1/2)*base...(3/2)*base.
118 * It corresponds to default IPv6 settings and is not overridable,
119 * because it is really reasonable choice.
122 unsigned long neigh_rand_reach_time(unsigned long base)
124 return base ? (net_random() % base) + (base >> 1) : 0;
126 EXPORT_SYMBOL(neigh_rand_reach_time);
129 static int neigh_forced_gc(struct neigh_table *tbl)
133 struct neigh_hash_table *nht;
135 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
137 write_lock_bh(&tbl->lock);
138 nht = rcu_dereference_protected(tbl->nht,
139 lockdep_is_held(&tbl->lock));
140 for (i = 0; i < (1 << nht->hash_shift); i++) {
142 struct neighbour __rcu **np;
144 np = &nht->hash_buckets[i];
145 while ((n = rcu_dereference_protected(*np,
146 lockdep_is_held(&tbl->lock))) != NULL) {
147 /* Neighbour record may be discarded if:
148 * - nobody refers to it.
149 * - it is not permanent
151 write_lock(&n->lock);
152 if (atomic_read(&n->refcnt) == 1 &&
153 !(n->nud_state & NUD_PERMANENT)) {
154 rcu_assign_pointer(*np,
155 rcu_dereference_protected(n->next,
156 lockdep_is_held(&tbl->lock)));
159 write_unlock(&n->lock);
160 neigh_cleanup_and_release(n);
163 write_unlock(&n->lock);
168 tbl->last_flush = jiffies;
170 write_unlock_bh(&tbl->lock);
175 static void neigh_add_timer(struct neighbour *n, unsigned long when)
178 if (unlikely(mod_timer(&n->timer, when))) {
179 printk("NEIGH: BUG, double timer add, state is %x\n",
185 static int neigh_del_timer(struct neighbour *n)
187 if ((n->nud_state & NUD_IN_TIMER) &&
188 del_timer(&n->timer)) {
195 static void pneigh_queue_purge(struct sk_buff_head *list)
199 while ((skb = skb_dequeue(list)) != NULL) {
205 static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
208 struct neigh_hash_table *nht;
210 nht = rcu_dereference_protected(tbl->nht,
211 lockdep_is_held(&tbl->lock));
213 for (i = 0; i < (1 << nht->hash_shift); i++) {
215 struct neighbour __rcu **np = &nht->hash_buckets[i];
217 while ((n = rcu_dereference_protected(*np,
218 lockdep_is_held(&tbl->lock))) != NULL) {
219 if (dev && n->dev != dev) {
223 rcu_assign_pointer(*np,
224 rcu_dereference_protected(n->next,
225 lockdep_is_held(&tbl->lock)));
226 write_lock(&n->lock);
230 if (atomic_read(&n->refcnt) != 1) {
231 /* The most unpleasant situation.
232 We must destroy neighbour entry,
233 but someone still uses it.
235 The destroy will be delayed until
236 the last user releases us, but
237 we must kill timers etc. and move
240 skb_queue_purge(&n->arp_queue);
241 n->output = neigh_blackhole;
242 if (n->nud_state & NUD_VALID)
243 n->nud_state = NUD_NOARP;
245 n->nud_state = NUD_NONE;
246 NEIGH_PRINTK2("neigh %p is stray.\n", n);
248 write_unlock(&n->lock);
249 neigh_cleanup_and_release(n);
254 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
256 write_lock_bh(&tbl->lock);
257 neigh_flush_dev(tbl, dev);
258 write_unlock_bh(&tbl->lock);
260 EXPORT_SYMBOL(neigh_changeaddr);
262 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
264 write_lock_bh(&tbl->lock);
265 neigh_flush_dev(tbl, dev);
266 pneigh_ifdown(tbl, dev);
267 write_unlock_bh(&tbl->lock);
269 del_timer_sync(&tbl->proxy_timer);
270 pneigh_queue_purge(&tbl->proxy_queue);
273 EXPORT_SYMBOL(neigh_ifdown);
275 static struct neighbour *neigh_alloc(struct neigh_table *tbl)
277 struct neighbour *n = NULL;
278 unsigned long now = jiffies;
281 entries = atomic_inc_return(&tbl->entries) - 1;
282 if (entries >= tbl->gc_thresh3 ||
283 (entries >= tbl->gc_thresh2 &&
284 time_after(now, tbl->last_flush + 5 * HZ))) {
285 if (!neigh_forced_gc(tbl) &&
286 entries >= tbl->gc_thresh3)
290 n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
294 skb_queue_head_init(&n->arp_queue);
295 rwlock_init(&n->lock);
296 seqlock_init(&n->ha_lock);
297 n->updated = n->used = now;
298 n->nud_state = NUD_NONE;
299 n->output = neigh_blackhole;
300 seqlock_init(&n->hh.hh_lock);
301 n->parms = neigh_parms_clone(&tbl->parms);
302 setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
304 NEIGH_CACHE_STAT_INC(tbl, allocs);
306 atomic_set(&n->refcnt, 1);
312 atomic_dec(&tbl->entries);
316 static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
318 size_t size = (1 << shift) * sizeof(struct neighbour *);
319 struct neigh_hash_table *ret;
320 struct neighbour __rcu **buckets;
322 ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
325 if (size <= PAGE_SIZE)
326 buckets = kzalloc(size, GFP_ATOMIC);
328 buckets = (struct neighbour __rcu **)
329 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
335 ret->hash_buckets = buckets;
336 ret->hash_shift = shift;
337 get_random_bytes(&ret->hash_rnd, sizeof(ret->hash_rnd));
342 static void neigh_hash_free_rcu(struct rcu_head *head)
344 struct neigh_hash_table *nht = container_of(head,
345 struct neigh_hash_table,
347 size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
348 struct neighbour __rcu **buckets = nht->hash_buckets;
350 if (size <= PAGE_SIZE)
353 free_pages((unsigned long)buckets, get_order(size));
357 static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
358 unsigned long new_shift)
360 unsigned int i, hash;
361 struct neigh_hash_table *new_nht, *old_nht;
363 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
365 old_nht = rcu_dereference_protected(tbl->nht,
366 lockdep_is_held(&tbl->lock));
367 new_nht = neigh_hash_alloc(new_shift);
371 for (i = 0; i < (1 << old_nht->hash_shift); i++) {
372 struct neighbour *n, *next;
374 for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
375 lockdep_is_held(&tbl->lock));
378 hash = tbl->hash(n->primary_key, n->dev,
381 hash >>= (32 - new_nht->hash_shift);
382 next = rcu_dereference_protected(n->next,
383 lockdep_is_held(&tbl->lock));
385 rcu_assign_pointer(n->next,
386 rcu_dereference_protected(
387 new_nht->hash_buckets[hash],
388 lockdep_is_held(&tbl->lock)));
389 rcu_assign_pointer(new_nht->hash_buckets[hash], n);
393 rcu_assign_pointer(tbl->nht, new_nht);
394 call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
398 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
399 struct net_device *dev)
402 int key_len = tbl->key_len;
404 struct neigh_hash_table *nht;
406 NEIGH_CACHE_STAT_INC(tbl, lookups);
409 nht = rcu_dereference_bh(tbl->nht);
410 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
412 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
414 n = rcu_dereference_bh(n->next)) {
415 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
416 if (!atomic_inc_not_zero(&n->refcnt))
418 NEIGH_CACHE_STAT_INC(tbl, hits);
423 rcu_read_unlock_bh();
426 EXPORT_SYMBOL(neigh_lookup);
428 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
432 int key_len = tbl->key_len;
434 struct neigh_hash_table *nht;
436 NEIGH_CACHE_STAT_INC(tbl, lookups);
439 nht = rcu_dereference_bh(tbl->nht);
440 hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
442 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
444 n = rcu_dereference_bh(n->next)) {
445 if (!memcmp(n->primary_key, pkey, key_len) &&
446 net_eq(dev_net(n->dev), net)) {
447 if (!atomic_inc_not_zero(&n->refcnt))
449 NEIGH_CACHE_STAT_INC(tbl, hits);
454 rcu_read_unlock_bh();
457 EXPORT_SYMBOL(neigh_lookup_nodev);
459 struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
460 struct net_device *dev)
463 int key_len = tbl->key_len;
465 struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
466 struct neigh_hash_table *nht;
469 rc = ERR_PTR(-ENOBUFS);
473 memcpy(n->primary_key, pkey, key_len);
477 /* Protocol specific setup. */
478 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
480 goto out_neigh_release;
483 /* Device specific setup. */
484 if (n->parms->neigh_setup &&
485 (error = n->parms->neigh_setup(n)) < 0) {
487 goto out_neigh_release;
490 n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
492 write_lock_bh(&tbl->lock);
493 nht = rcu_dereference_protected(tbl->nht,
494 lockdep_is_held(&tbl->lock));
496 if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
497 nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
499 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
501 if (n->parms->dead) {
502 rc = ERR_PTR(-EINVAL);
506 for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
507 lockdep_is_held(&tbl->lock));
509 n1 = rcu_dereference_protected(n1->next,
510 lockdep_is_held(&tbl->lock))) {
511 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
520 rcu_assign_pointer(n->next,
521 rcu_dereference_protected(nht->hash_buckets[hash_val],
522 lockdep_is_held(&tbl->lock)));
523 rcu_assign_pointer(nht->hash_buckets[hash_val], n);
524 write_unlock_bh(&tbl->lock);
525 NEIGH_PRINTK2("neigh %p is created.\n", n);
530 write_unlock_bh(&tbl->lock);
535 EXPORT_SYMBOL(neigh_create);
537 static u32 pneigh_hash(const void *pkey, int key_len)
539 u32 hash_val = *(u32 *)(pkey + key_len - 4);
540 hash_val ^= (hash_val >> 16);
541 hash_val ^= hash_val >> 8;
542 hash_val ^= hash_val >> 4;
543 hash_val &= PNEIGH_HASHMASK;
547 static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
551 struct net_device *dev)
554 if (!memcmp(n->key, pkey, key_len) &&
555 net_eq(pneigh_net(n), net) &&
556 (n->dev == dev || !n->dev))
563 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
564 struct net *net, const void *pkey, struct net_device *dev)
566 int key_len = tbl->key_len;
567 u32 hash_val = pneigh_hash(pkey, key_len);
569 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
570 net, pkey, key_len, dev);
572 EXPORT_SYMBOL_GPL(__pneigh_lookup);
574 struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
575 struct net *net, const void *pkey,
576 struct net_device *dev, int creat)
578 struct pneigh_entry *n;
579 int key_len = tbl->key_len;
580 u32 hash_val = pneigh_hash(pkey, key_len);
582 read_lock_bh(&tbl->lock);
583 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
584 net, pkey, key_len, dev);
585 read_unlock_bh(&tbl->lock);
592 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
596 write_pnet(&n->net, hold_net(net));
597 memcpy(n->key, pkey, key_len);
602 if (tbl->pconstructor && tbl->pconstructor(n)) {
611 write_lock_bh(&tbl->lock);
612 n->next = tbl->phash_buckets[hash_val];
613 tbl->phash_buckets[hash_val] = n;
614 write_unlock_bh(&tbl->lock);
618 EXPORT_SYMBOL(pneigh_lookup);
621 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
622 struct net_device *dev)
624 struct pneigh_entry *n, **np;
625 int key_len = tbl->key_len;
626 u32 hash_val = pneigh_hash(pkey, key_len);
628 write_lock_bh(&tbl->lock);
629 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
631 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
632 net_eq(pneigh_net(n), net)) {
634 write_unlock_bh(&tbl->lock);
635 if (tbl->pdestructor)
639 release_net(pneigh_net(n));
644 write_unlock_bh(&tbl->lock);
648 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
650 struct pneigh_entry *n, **np;
653 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
654 np = &tbl->phash_buckets[h];
655 while ((n = *np) != NULL) {
656 if (!dev || n->dev == dev) {
658 if (tbl->pdestructor)
662 release_net(pneigh_net(n));
672 static void neigh_parms_destroy(struct neigh_parms *parms);
674 static inline void neigh_parms_put(struct neigh_parms *parms)
676 if (atomic_dec_and_test(&parms->refcnt))
677 neigh_parms_destroy(parms);
680 static void neigh_destroy_rcu(struct rcu_head *head)
682 struct neighbour *neigh = container_of(head, struct neighbour, rcu);
684 kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
687 * neighbour must already be out of the table;
690 void neigh_destroy(struct neighbour *neigh)
692 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
696 "Destroying alive neighbour %p\n", neigh);
701 if (neigh_del_timer(neigh))
702 printk(KERN_WARNING "Impossible event.\n");
704 skb_queue_purge(&neigh->arp_queue);
707 neigh_parms_put(neigh->parms);
709 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
711 atomic_dec(&neigh->tbl->entries);
712 call_rcu(&neigh->rcu, neigh_destroy_rcu);
714 EXPORT_SYMBOL(neigh_destroy);
716 /* Neighbour state is suspicious;
719 Called with write_locked neigh.
721 static void neigh_suspect(struct neighbour *neigh)
723 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
725 neigh->output = neigh->ops->output;
728 /* Neighbour state is OK;
731 Called with write_locked neigh.
733 static void neigh_connect(struct neighbour *neigh)
735 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
737 neigh->output = neigh->ops->connected_output;
740 static void neigh_periodic_work(struct work_struct *work)
742 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
744 struct neighbour __rcu **np;
746 struct neigh_hash_table *nht;
748 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
750 write_lock_bh(&tbl->lock);
751 nht = rcu_dereference_protected(tbl->nht,
752 lockdep_is_held(&tbl->lock));
755 * periodically recompute ReachableTime from random function
758 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
759 struct neigh_parms *p;
760 tbl->last_rand = jiffies;
761 for (p = &tbl->parms; p; p = p->next)
763 neigh_rand_reach_time(p->base_reachable_time);
766 for (i = 0 ; i < (1 << nht->hash_shift); i++) {
767 np = &nht->hash_buckets[i];
769 while ((n = rcu_dereference_protected(*np,
770 lockdep_is_held(&tbl->lock))) != NULL) {
773 write_lock(&n->lock);
775 state = n->nud_state;
776 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
777 write_unlock(&n->lock);
781 if (time_before(n->used, n->confirmed))
782 n->used = n->confirmed;
784 if (atomic_read(&n->refcnt) == 1 &&
785 (state == NUD_FAILED ||
786 time_after(jiffies, n->used + n->parms->gc_staletime))) {
789 write_unlock(&n->lock);
790 neigh_cleanup_and_release(n);
793 write_unlock(&n->lock);
799 * It's fine to release lock here, even if hash table
800 * grows while we are preempted.
802 write_unlock_bh(&tbl->lock);
804 write_lock_bh(&tbl->lock);
806 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
807 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
808 * base_reachable_time.
810 schedule_delayed_work(&tbl->gc_work,
811 tbl->parms.base_reachable_time >> 1);
812 write_unlock_bh(&tbl->lock);
815 static __inline__ int neigh_max_probes(struct neighbour *n)
817 struct neigh_parms *p = n->parms;
818 return (n->nud_state & NUD_PROBE) ?
820 p->ucast_probes + p->app_probes + p->mcast_probes;
823 static void neigh_invalidate(struct neighbour *neigh)
824 __releases(neigh->lock)
825 __acquires(neigh->lock)
829 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
830 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
831 neigh->updated = jiffies;
833 /* It is very thin place. report_unreachable is very complicated
834 routine. Particularly, it can hit the same neighbour entry!
836 So that, we try to be accurate and avoid dead loop. --ANK
838 while (neigh->nud_state == NUD_FAILED &&
839 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
840 write_unlock(&neigh->lock);
841 neigh->ops->error_report(neigh, skb);
842 write_lock(&neigh->lock);
844 skb_queue_purge(&neigh->arp_queue);
847 static void neigh_probe(struct neighbour *neigh)
848 __releases(neigh->lock)
850 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
851 /* keep skb alive even if arp_queue overflows */
853 skb = skb_copy(skb, GFP_ATOMIC);
854 write_unlock(&neigh->lock);
855 neigh->ops->solicit(neigh, skb);
856 atomic_inc(&neigh->probes);
860 /* Called when a timer expires for a neighbour entry. */
862 static void neigh_timer_handler(unsigned long arg)
864 unsigned long now, next;
865 struct neighbour *neigh = (struct neighbour *)arg;
869 write_lock(&neigh->lock);
871 state = neigh->nud_state;
875 if (!(state & NUD_IN_TIMER)) {
877 printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
882 if (state & NUD_REACHABLE) {
883 if (time_before_eq(now,
884 neigh->confirmed + neigh->parms->reachable_time)) {
885 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
886 next = neigh->confirmed + neigh->parms->reachable_time;
887 } else if (time_before_eq(now,
888 neigh->used + neigh->parms->delay_probe_time)) {
889 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
890 neigh->nud_state = NUD_DELAY;
891 neigh->updated = jiffies;
892 neigh_suspect(neigh);
893 next = now + neigh->parms->delay_probe_time;
895 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
896 neigh->nud_state = NUD_STALE;
897 neigh->updated = jiffies;
898 neigh_suspect(neigh);
901 } else if (state & NUD_DELAY) {
902 if (time_before_eq(now,
903 neigh->confirmed + neigh->parms->delay_probe_time)) {
904 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
905 neigh->nud_state = NUD_REACHABLE;
906 neigh->updated = jiffies;
907 neigh_connect(neigh);
909 next = neigh->confirmed + neigh->parms->reachable_time;
911 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
912 neigh->nud_state = NUD_PROBE;
913 neigh->updated = jiffies;
914 atomic_set(&neigh->probes, 0);
915 next = now + neigh->parms->retrans_time;
918 /* NUD_PROBE|NUD_INCOMPLETE */
919 next = now + neigh->parms->retrans_time;
922 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
923 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
924 neigh->nud_state = NUD_FAILED;
926 neigh_invalidate(neigh);
929 if (neigh->nud_state & NUD_IN_TIMER) {
930 if (time_before(next, jiffies + HZ/2))
931 next = jiffies + HZ/2;
932 if (!mod_timer(&neigh->timer, next))
935 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
939 write_unlock(&neigh->lock);
943 neigh_update_notify(neigh);
945 neigh_release(neigh);
948 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
951 bool immediate_probe = false;
953 write_lock_bh(&neigh->lock);
956 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
959 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
960 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
961 unsigned long next, now = jiffies;
963 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
964 neigh->nud_state = NUD_INCOMPLETE;
965 neigh->updated = now;
966 next = now + max(neigh->parms->retrans_time, HZ/2);
967 neigh_add_timer(neigh, next);
968 immediate_probe = true;
970 neigh->nud_state = NUD_FAILED;
971 neigh->updated = jiffies;
972 write_unlock_bh(&neigh->lock);
977 } else if (neigh->nud_state & NUD_STALE) {
978 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
979 neigh->nud_state = NUD_DELAY;
980 neigh->updated = jiffies;
981 neigh_add_timer(neigh,
982 jiffies + neigh->parms->delay_probe_time);
985 if (neigh->nud_state == NUD_INCOMPLETE) {
987 if (skb_queue_len(&neigh->arp_queue) >=
988 neigh->parms->queue_len) {
989 struct sk_buff *buff;
990 buff = __skb_dequeue(&neigh->arp_queue);
992 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
995 __skb_queue_tail(&neigh->arp_queue, skb);
1000 if (immediate_probe)
1003 write_unlock(&neigh->lock);
1007 EXPORT_SYMBOL(__neigh_event_send);
1009 static void neigh_update_hhs(struct neighbour *neigh)
1011 struct hh_cache *hh;
1012 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1015 if (neigh->dev->header_ops)
1016 update = neigh->dev->header_ops->cache_update;
1021 write_seqlock_bh(&hh->hh_lock);
1022 update(hh, neigh->dev, neigh->ha);
1023 write_sequnlock_bh(&hh->hh_lock);
1030 /* Generic update routine.
1031 -- lladdr is new lladdr or NULL, if it is not supplied.
1032 -- new is new state.
1034 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1036 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1037 lladdr instead of overriding it
1039 It also allows to retain current state
1040 if lladdr is unchanged.
1041 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1043 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1045 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1048 Caller MUST hold reference count on the entry.
1051 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1057 struct net_device *dev;
1058 int update_isrouter = 0;
1060 write_lock_bh(&neigh->lock);
1063 old = neigh->nud_state;
1066 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1067 (old & (NUD_NOARP | NUD_PERMANENT)))
1070 if (!(new & NUD_VALID)) {
1071 neigh_del_timer(neigh);
1072 if (old & NUD_CONNECTED)
1073 neigh_suspect(neigh);
1074 neigh->nud_state = new;
1076 notify = old & NUD_VALID;
1077 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1078 (new & NUD_FAILED)) {
1079 neigh_invalidate(neigh);
1085 /* Compare new lladdr with cached one */
1086 if (!dev->addr_len) {
1087 /* First case: device needs no address. */
1089 } else if (lladdr) {
1090 /* The second case: if something is already cached
1091 and a new address is proposed:
1093 - if they are different, check override flag
1095 if ((old & NUD_VALID) &&
1096 !memcmp(lladdr, neigh->ha, dev->addr_len))
1099 /* No address is supplied; if we know something,
1100 use it, otherwise discard the request.
1103 if (!(old & NUD_VALID))
1108 if (new & NUD_CONNECTED)
1109 neigh->confirmed = jiffies;
1110 neigh->updated = jiffies;
1112 /* If entry was valid and address is not changed,
1113 do not change entry state, if new one is STALE.
1116 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1117 if (old & NUD_VALID) {
1118 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1119 update_isrouter = 0;
1120 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1121 (old & NUD_CONNECTED)) {
1127 if (lladdr == neigh->ha && new == NUD_STALE &&
1128 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1129 (old & NUD_CONNECTED))
1136 neigh_del_timer(neigh);
1137 if (new & NUD_IN_TIMER)
1138 neigh_add_timer(neigh, (jiffies +
1139 ((new & NUD_REACHABLE) ?
1140 neigh->parms->reachable_time :
1142 neigh->nud_state = new;
1145 if (lladdr != neigh->ha) {
1146 write_seqlock(&neigh->ha_lock);
1147 memcpy(&neigh->ha, lladdr, dev->addr_len);
1148 write_sequnlock(&neigh->ha_lock);
1149 neigh_update_hhs(neigh);
1150 if (!(new & NUD_CONNECTED))
1151 neigh->confirmed = jiffies -
1152 (neigh->parms->base_reachable_time << 1);
1157 if (new & NUD_CONNECTED)
1158 neigh_connect(neigh);
1160 neigh_suspect(neigh);
1161 if (!(old & NUD_VALID)) {
1162 struct sk_buff *skb;
1164 /* Again: avoid dead loop if something went wrong */
1166 while (neigh->nud_state & NUD_VALID &&
1167 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1168 struct dst_entry *dst = skb_dst(skb);
1169 struct neighbour *n2, *n1 = neigh;
1170 write_unlock_bh(&neigh->lock);
1171 /* On shaper/eql skb->dst->neighbour != neigh :( */
1172 if (dst && (n2 = dst_get_neighbour(dst)) != NULL)
1174 n1->output(n1, skb);
1175 write_lock_bh(&neigh->lock);
1177 skb_queue_purge(&neigh->arp_queue);
1180 if (update_isrouter) {
1181 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1182 (neigh->flags | NTF_ROUTER) :
1183 (neigh->flags & ~NTF_ROUTER);
1185 write_unlock_bh(&neigh->lock);
1188 neigh_update_notify(neigh);
1192 EXPORT_SYMBOL(neigh_update);
1194 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1195 u8 *lladdr, void *saddr,
1196 struct net_device *dev)
1198 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1199 lladdr || !dev->addr_len);
1201 neigh_update(neigh, lladdr, NUD_STALE,
1202 NEIGH_UPDATE_F_OVERRIDE);
1205 EXPORT_SYMBOL(neigh_event_ns);
1207 /* called with read_lock_bh(&n->lock); */
1208 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst)
1210 struct net_device *dev = dst->dev;
1211 __be16 prot = dst->ops->protocol;
1212 struct hh_cache *hh = &n->hh;
1214 write_lock_bh(&n->lock);
1216 /* Only one thread can come in here and initialize the
1220 dev->header_ops->cache(n, hh, prot);
1222 write_unlock_bh(&n->lock);
1225 /* This function can be used in contexts, where only old dev_queue_xmit
1226 * worked, f.e. if you want to override normal output path (eql, shaper),
1227 * but resolution is not made yet.
1230 int neigh_compat_output(struct neighbour *neigh, struct sk_buff *skb)
1232 struct net_device *dev = skb->dev;
1234 __skb_pull(skb, skb_network_offset(skb));
1236 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1238 dev->header_ops->rebuild(skb))
1241 return dev_queue_xmit(skb);
1243 EXPORT_SYMBOL(neigh_compat_output);
1245 /* Slow and careful. */
1247 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1249 struct dst_entry *dst = skb_dst(skb);
1255 __skb_pull(skb, skb_network_offset(skb));
1257 if (!neigh_event_send(neigh, skb)) {
1259 struct net_device *dev = neigh->dev;
1262 if (dev->header_ops->cache && !neigh->hh.hh_len)
1263 neigh_hh_init(neigh, dst);
1266 seq = read_seqbegin(&neigh->ha_lock);
1267 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1268 neigh->ha, NULL, skb->len);
1269 } while (read_seqretry(&neigh->ha_lock, seq));
1272 rc = dev_queue_xmit(skb);
1279 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1286 EXPORT_SYMBOL(neigh_resolve_output);
1288 /* As fast as possible without hh cache */
1290 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1292 struct net_device *dev = neigh->dev;
1296 __skb_pull(skb, skb_network_offset(skb));
1299 seq = read_seqbegin(&neigh->ha_lock);
1300 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1301 neigh->ha, NULL, skb->len);
1302 } while (read_seqretry(&neigh->ha_lock, seq));
1305 err = dev_queue_xmit(skb);
1312 EXPORT_SYMBOL(neigh_connected_output);
1314 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1316 return dev_queue_xmit(skb);
1318 EXPORT_SYMBOL(neigh_direct_output);
1320 static void neigh_proxy_process(unsigned long arg)
1322 struct neigh_table *tbl = (struct neigh_table *)arg;
1323 long sched_next = 0;
1324 unsigned long now = jiffies;
1325 struct sk_buff *skb, *n;
1327 spin_lock(&tbl->proxy_queue.lock);
1329 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1330 long tdif = NEIGH_CB(skb)->sched_next - now;
1333 struct net_device *dev = skb->dev;
1335 __skb_unlink(skb, &tbl->proxy_queue);
1336 if (tbl->proxy_redo && netif_running(dev)) {
1338 tbl->proxy_redo(skb);
1345 } else if (!sched_next || tdif < sched_next)
1348 del_timer(&tbl->proxy_timer);
1350 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1351 spin_unlock(&tbl->proxy_queue.lock);
1354 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1355 struct sk_buff *skb)
1357 unsigned long now = jiffies;
1358 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1360 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1365 NEIGH_CB(skb)->sched_next = sched_next;
1366 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1368 spin_lock(&tbl->proxy_queue.lock);
1369 if (del_timer(&tbl->proxy_timer)) {
1370 if (time_before(tbl->proxy_timer.expires, sched_next))
1371 sched_next = tbl->proxy_timer.expires;
1375 __skb_queue_tail(&tbl->proxy_queue, skb);
1376 mod_timer(&tbl->proxy_timer, sched_next);
1377 spin_unlock(&tbl->proxy_queue.lock);
1379 EXPORT_SYMBOL(pneigh_enqueue);
1381 static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1382 struct net *net, int ifindex)
1384 struct neigh_parms *p;
1386 for (p = &tbl->parms; p; p = p->next) {
1387 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1388 (!p->dev && !ifindex))
1395 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1396 struct neigh_table *tbl)
1398 struct neigh_parms *p, *ref;
1399 struct net *net = dev_net(dev);
1400 const struct net_device_ops *ops = dev->netdev_ops;
1402 ref = lookup_neigh_parms(tbl, net, 0);
1406 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1409 atomic_set(&p->refcnt, 1);
1411 neigh_rand_reach_time(p->base_reachable_time);
1413 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1420 write_pnet(&p->net, hold_net(net));
1421 p->sysctl_table = NULL;
1422 write_lock_bh(&tbl->lock);
1423 p->next = tbl->parms.next;
1424 tbl->parms.next = p;
1425 write_unlock_bh(&tbl->lock);
1429 EXPORT_SYMBOL(neigh_parms_alloc);
1431 static void neigh_rcu_free_parms(struct rcu_head *head)
1433 struct neigh_parms *parms =
1434 container_of(head, struct neigh_parms, rcu_head);
1436 neigh_parms_put(parms);
1439 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1441 struct neigh_parms **p;
1443 if (!parms || parms == &tbl->parms)
1445 write_lock_bh(&tbl->lock);
1446 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1450 write_unlock_bh(&tbl->lock);
1452 dev_put(parms->dev);
1453 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1457 write_unlock_bh(&tbl->lock);
1458 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1460 EXPORT_SYMBOL(neigh_parms_release);
1462 static void neigh_parms_destroy(struct neigh_parms *parms)
1464 release_net(neigh_parms_net(parms));
1468 static struct lock_class_key neigh_table_proxy_queue_class;
1470 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1472 unsigned long now = jiffies;
1473 unsigned long phsize;
1475 write_pnet(&tbl->parms.net, &init_net);
1476 atomic_set(&tbl->parms.refcnt, 1);
1477 tbl->parms.reachable_time =
1478 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1480 if (!tbl->kmem_cachep)
1482 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1483 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1485 tbl->stats = alloc_percpu(struct neigh_statistics);
1487 panic("cannot create neighbour cache statistics");
1489 #ifdef CONFIG_PROC_FS
1490 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1491 &neigh_stat_seq_fops, tbl))
1492 panic("cannot create neighbour proc dir entry");
1495 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1497 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1498 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1500 if (!tbl->nht || !tbl->phash_buckets)
1501 panic("cannot allocate neighbour cache hashes");
1503 rwlock_init(&tbl->lock);
1504 INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1505 schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1506 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1507 skb_queue_head_init_class(&tbl->proxy_queue,
1508 &neigh_table_proxy_queue_class);
1510 tbl->last_flush = now;
1511 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1513 EXPORT_SYMBOL(neigh_table_init_no_netlink);
1515 void neigh_table_init(struct neigh_table *tbl)
1517 struct neigh_table *tmp;
1519 neigh_table_init_no_netlink(tbl);
1520 write_lock(&neigh_tbl_lock);
1521 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1522 if (tmp->family == tbl->family)
1525 tbl->next = neigh_tables;
1527 write_unlock(&neigh_tbl_lock);
1529 if (unlikely(tmp)) {
1530 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1531 "family %d\n", tbl->family);
1535 EXPORT_SYMBOL(neigh_table_init);
1537 int neigh_table_clear(struct neigh_table *tbl)
1539 struct neigh_table **tp;
1541 /* It is not clean... Fix it to unload IPv6 module safely */
1542 cancel_delayed_work_sync(&tbl->gc_work);
1543 del_timer_sync(&tbl->proxy_timer);
1544 pneigh_queue_purge(&tbl->proxy_queue);
1545 neigh_ifdown(tbl, NULL);
1546 if (atomic_read(&tbl->entries))
1547 printk(KERN_CRIT "neighbour leakage\n");
1548 write_lock(&neigh_tbl_lock);
1549 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1555 write_unlock(&neigh_tbl_lock);
1557 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1558 neigh_hash_free_rcu);
1561 kfree(tbl->phash_buckets);
1562 tbl->phash_buckets = NULL;
1564 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1566 free_percpu(tbl->stats);
1569 kmem_cache_destroy(tbl->kmem_cachep);
1570 tbl->kmem_cachep = NULL;
1574 EXPORT_SYMBOL(neigh_table_clear);
1576 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1578 struct net *net = sock_net(skb->sk);
1580 struct nlattr *dst_attr;
1581 struct neigh_table *tbl;
1582 struct net_device *dev = NULL;
1586 if (nlmsg_len(nlh) < sizeof(*ndm))
1589 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1590 if (dst_attr == NULL)
1593 ndm = nlmsg_data(nlh);
1594 if (ndm->ndm_ifindex) {
1595 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1602 read_lock(&neigh_tbl_lock);
1603 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1604 struct neighbour *neigh;
1606 if (tbl->family != ndm->ndm_family)
1608 read_unlock(&neigh_tbl_lock);
1610 if (nla_len(dst_attr) < tbl->key_len)
1613 if (ndm->ndm_flags & NTF_PROXY) {
1614 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1621 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1622 if (neigh == NULL) {
1627 err = neigh_update(neigh, NULL, NUD_FAILED,
1628 NEIGH_UPDATE_F_OVERRIDE |
1629 NEIGH_UPDATE_F_ADMIN);
1630 neigh_release(neigh);
1633 read_unlock(&neigh_tbl_lock);
1634 err = -EAFNOSUPPORT;
1640 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1642 struct net *net = sock_net(skb->sk);
1644 struct nlattr *tb[NDA_MAX+1];
1645 struct neigh_table *tbl;
1646 struct net_device *dev = NULL;
1650 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1655 if (tb[NDA_DST] == NULL)
1658 ndm = nlmsg_data(nlh);
1659 if (ndm->ndm_ifindex) {
1660 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1666 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1670 read_lock(&neigh_tbl_lock);
1671 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1672 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1673 struct neighbour *neigh;
1676 if (tbl->family != ndm->ndm_family)
1678 read_unlock(&neigh_tbl_lock);
1680 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1682 dst = nla_data(tb[NDA_DST]);
1683 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1685 if (ndm->ndm_flags & NTF_PROXY) {
1686 struct pneigh_entry *pn;
1689 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1691 pn->flags = ndm->ndm_flags;
1700 neigh = neigh_lookup(tbl, dst, dev);
1701 if (neigh == NULL) {
1702 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1707 neigh = __neigh_lookup_errno(tbl, dst, dev);
1708 if (IS_ERR(neigh)) {
1709 err = PTR_ERR(neigh);
1713 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1715 neigh_release(neigh);
1719 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1720 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1723 if (ndm->ndm_flags & NTF_USE) {
1724 neigh_event_send(neigh, NULL);
1727 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1728 neigh_release(neigh);
1732 read_unlock(&neigh_tbl_lock);
1733 err = -EAFNOSUPPORT;
1738 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1740 struct nlattr *nest;
1742 nest = nla_nest_start(skb, NDTA_PARMS);
1747 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1749 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1750 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1751 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1752 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1753 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1754 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1755 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1756 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1757 parms->base_reachable_time);
1758 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1759 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1760 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1761 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1762 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1763 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1765 return nla_nest_end(skb, nest);
1768 nla_nest_cancel(skb, nest);
1772 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1773 u32 pid, u32 seq, int type, int flags)
1775 struct nlmsghdr *nlh;
1776 struct ndtmsg *ndtmsg;
1778 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1782 ndtmsg = nlmsg_data(nlh);
1784 read_lock_bh(&tbl->lock);
1785 ndtmsg->ndtm_family = tbl->family;
1786 ndtmsg->ndtm_pad1 = 0;
1787 ndtmsg->ndtm_pad2 = 0;
1789 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1790 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1791 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1792 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1793 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1796 unsigned long now = jiffies;
1797 unsigned int flush_delta = now - tbl->last_flush;
1798 unsigned int rand_delta = now - tbl->last_rand;
1799 struct neigh_hash_table *nht;
1800 struct ndt_config ndc = {
1801 .ndtc_key_len = tbl->key_len,
1802 .ndtc_entry_size = tbl->entry_size,
1803 .ndtc_entries = atomic_read(&tbl->entries),
1804 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1805 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1806 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1810 nht = rcu_dereference_bh(tbl->nht);
1811 ndc.ndtc_hash_rnd = nht->hash_rnd;
1812 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1813 rcu_read_unlock_bh();
1815 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1820 struct ndt_stats ndst;
1822 memset(&ndst, 0, sizeof(ndst));
1824 for_each_possible_cpu(cpu) {
1825 struct neigh_statistics *st;
1827 st = per_cpu_ptr(tbl->stats, cpu);
1828 ndst.ndts_allocs += st->allocs;
1829 ndst.ndts_destroys += st->destroys;
1830 ndst.ndts_hash_grows += st->hash_grows;
1831 ndst.ndts_res_failed += st->res_failed;
1832 ndst.ndts_lookups += st->lookups;
1833 ndst.ndts_hits += st->hits;
1834 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1835 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1836 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1837 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1840 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1843 BUG_ON(tbl->parms.dev);
1844 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1845 goto nla_put_failure;
1847 read_unlock_bh(&tbl->lock);
1848 return nlmsg_end(skb, nlh);
1851 read_unlock_bh(&tbl->lock);
1852 nlmsg_cancel(skb, nlh);
1856 static int neightbl_fill_param_info(struct sk_buff *skb,
1857 struct neigh_table *tbl,
1858 struct neigh_parms *parms,
1859 u32 pid, u32 seq, int type,
1862 struct ndtmsg *ndtmsg;
1863 struct nlmsghdr *nlh;
1865 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1869 ndtmsg = nlmsg_data(nlh);
1871 read_lock_bh(&tbl->lock);
1872 ndtmsg->ndtm_family = tbl->family;
1873 ndtmsg->ndtm_pad1 = 0;
1874 ndtmsg->ndtm_pad2 = 0;
1876 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1877 neightbl_fill_parms(skb, parms) < 0)
1880 read_unlock_bh(&tbl->lock);
1881 return nlmsg_end(skb, nlh);
1883 read_unlock_bh(&tbl->lock);
1884 nlmsg_cancel(skb, nlh);
1888 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1889 [NDTA_NAME] = { .type = NLA_STRING },
1890 [NDTA_THRESH1] = { .type = NLA_U32 },
1891 [NDTA_THRESH2] = { .type = NLA_U32 },
1892 [NDTA_THRESH3] = { .type = NLA_U32 },
1893 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1894 [NDTA_PARMS] = { .type = NLA_NESTED },
1897 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1898 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1899 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1900 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1901 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1902 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1903 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1904 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1905 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1906 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1907 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1908 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1909 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1910 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1913 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1915 struct net *net = sock_net(skb->sk);
1916 struct neigh_table *tbl;
1917 struct ndtmsg *ndtmsg;
1918 struct nlattr *tb[NDTA_MAX+1];
1921 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1922 nl_neightbl_policy);
1926 if (tb[NDTA_NAME] == NULL) {
1931 ndtmsg = nlmsg_data(nlh);
1932 read_lock(&neigh_tbl_lock);
1933 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1934 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1937 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1947 * We acquire tbl->lock to be nice to the periodic timers and
1948 * make sure they always see a consistent set of values.
1950 write_lock_bh(&tbl->lock);
1952 if (tb[NDTA_PARMS]) {
1953 struct nlattr *tbp[NDTPA_MAX+1];
1954 struct neigh_parms *p;
1957 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1958 nl_ntbl_parm_policy);
1960 goto errout_tbl_lock;
1962 if (tbp[NDTPA_IFINDEX])
1963 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1965 p = lookup_neigh_parms(tbl, net, ifindex);
1968 goto errout_tbl_lock;
1971 for (i = 1; i <= NDTPA_MAX; i++) {
1976 case NDTPA_QUEUE_LEN:
1977 p->queue_len = nla_get_u32(tbp[i]);
1979 case NDTPA_PROXY_QLEN:
1980 p->proxy_qlen = nla_get_u32(tbp[i]);
1982 case NDTPA_APP_PROBES:
1983 p->app_probes = nla_get_u32(tbp[i]);
1985 case NDTPA_UCAST_PROBES:
1986 p->ucast_probes = nla_get_u32(tbp[i]);
1988 case NDTPA_MCAST_PROBES:
1989 p->mcast_probes = nla_get_u32(tbp[i]);
1991 case NDTPA_BASE_REACHABLE_TIME:
1992 p->base_reachable_time = nla_get_msecs(tbp[i]);
1994 case NDTPA_GC_STALETIME:
1995 p->gc_staletime = nla_get_msecs(tbp[i]);
1997 case NDTPA_DELAY_PROBE_TIME:
1998 p->delay_probe_time = nla_get_msecs(tbp[i]);
2000 case NDTPA_RETRANS_TIME:
2001 p->retrans_time = nla_get_msecs(tbp[i]);
2003 case NDTPA_ANYCAST_DELAY:
2004 p->anycast_delay = nla_get_msecs(tbp[i]);
2006 case NDTPA_PROXY_DELAY:
2007 p->proxy_delay = nla_get_msecs(tbp[i]);
2009 case NDTPA_LOCKTIME:
2010 p->locktime = nla_get_msecs(tbp[i]);
2016 if (tb[NDTA_THRESH1])
2017 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2019 if (tb[NDTA_THRESH2])
2020 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2022 if (tb[NDTA_THRESH3])
2023 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2025 if (tb[NDTA_GC_INTERVAL])
2026 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2031 write_unlock_bh(&tbl->lock);
2033 read_unlock(&neigh_tbl_lock);
2038 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2040 struct net *net = sock_net(skb->sk);
2041 int family, tidx, nidx = 0;
2042 int tbl_skip = cb->args[0];
2043 int neigh_skip = cb->args[1];
2044 struct neigh_table *tbl;
2046 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2048 read_lock(&neigh_tbl_lock);
2049 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
2050 struct neigh_parms *p;
2052 if (tidx < tbl_skip || (family && tbl->family != family))
2055 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2056 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2060 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
2061 if (!net_eq(neigh_parms_net(p), net))
2064 if (nidx < neigh_skip)
2067 if (neightbl_fill_param_info(skb, tbl, p,
2068 NETLINK_CB(cb->skb).pid,
2080 read_unlock(&neigh_tbl_lock);
2087 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2088 u32 pid, u32 seq, int type, unsigned int flags)
2090 unsigned long now = jiffies;
2091 struct nda_cacheinfo ci;
2092 struct nlmsghdr *nlh;
2095 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2099 ndm = nlmsg_data(nlh);
2100 ndm->ndm_family = neigh->ops->family;
2103 ndm->ndm_flags = neigh->flags;
2104 ndm->ndm_type = neigh->type;
2105 ndm->ndm_ifindex = neigh->dev->ifindex;
2107 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2109 read_lock_bh(&neigh->lock);
2110 ndm->ndm_state = neigh->nud_state;
2111 if (neigh->nud_state & NUD_VALID) {
2112 char haddr[MAX_ADDR_LEN];
2114 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2115 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2116 read_unlock_bh(&neigh->lock);
2117 goto nla_put_failure;
2121 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2122 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2123 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2124 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2125 read_unlock_bh(&neigh->lock);
2127 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2128 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2130 return nlmsg_end(skb, nlh);
2133 nlmsg_cancel(skb, nlh);
2137 static void neigh_update_notify(struct neighbour *neigh)
2139 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2140 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2143 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2144 struct netlink_callback *cb)
2146 struct net *net = sock_net(skb->sk);
2147 struct neighbour *n;
2148 int rc, h, s_h = cb->args[1];
2149 int idx, s_idx = idx = cb->args[2];
2150 struct neigh_hash_table *nht;
2153 nht = rcu_dereference_bh(tbl->nht);
2155 for (h = 0; h < (1 << nht->hash_shift); h++) {
2160 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2162 n = rcu_dereference_bh(n->next)) {
2163 if (!net_eq(dev_net(n->dev), net))
2167 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2170 NLM_F_MULTI) <= 0) {
2180 rcu_read_unlock_bh();
2186 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2188 struct neigh_table *tbl;
2191 read_lock(&neigh_tbl_lock);
2192 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2195 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2196 if (t < s_t || (family && tbl->family != family))
2199 memset(&cb->args[1], 0, sizeof(cb->args) -
2200 sizeof(cb->args[0]));
2201 if (neigh_dump_table(tbl, skb, cb) < 0)
2204 read_unlock(&neigh_tbl_lock);
2210 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2213 struct neigh_hash_table *nht;
2216 nht = rcu_dereference_bh(tbl->nht);
2218 read_lock(&tbl->lock); /* avoid resizes */
2219 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2220 struct neighbour *n;
2222 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2224 n = rcu_dereference_bh(n->next))
2227 read_unlock(&tbl->lock);
2228 rcu_read_unlock_bh();
2230 EXPORT_SYMBOL(neigh_for_each);
2232 /* The tbl->lock must be held as a writer and BH disabled. */
2233 void __neigh_for_each_release(struct neigh_table *tbl,
2234 int (*cb)(struct neighbour *))
2237 struct neigh_hash_table *nht;
2239 nht = rcu_dereference_protected(tbl->nht,
2240 lockdep_is_held(&tbl->lock));
2241 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2242 struct neighbour *n;
2243 struct neighbour __rcu **np;
2245 np = &nht->hash_buckets[chain];
2246 while ((n = rcu_dereference_protected(*np,
2247 lockdep_is_held(&tbl->lock))) != NULL) {
2250 write_lock(&n->lock);
2253 rcu_assign_pointer(*np,
2254 rcu_dereference_protected(n->next,
2255 lockdep_is_held(&tbl->lock)));
2259 write_unlock(&n->lock);
2261 neigh_cleanup_and_release(n);
2265 EXPORT_SYMBOL(__neigh_for_each_release);
2267 #ifdef CONFIG_PROC_FS
2269 static struct neighbour *neigh_get_first(struct seq_file *seq)
2271 struct neigh_seq_state *state = seq->private;
2272 struct net *net = seq_file_net(seq);
2273 struct neigh_hash_table *nht = state->nht;
2274 struct neighbour *n = NULL;
2275 int bucket = state->bucket;
2277 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2278 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2279 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2282 if (!net_eq(dev_net(n->dev), net))
2284 if (state->neigh_sub_iter) {
2288 v = state->neigh_sub_iter(state, n, &fakep);
2292 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2294 if (n->nud_state & ~NUD_NOARP)
2297 n = rcu_dereference_bh(n->next);
2303 state->bucket = bucket;
2308 static struct neighbour *neigh_get_next(struct seq_file *seq,
2309 struct neighbour *n,
2312 struct neigh_seq_state *state = seq->private;
2313 struct net *net = seq_file_net(seq);
2314 struct neigh_hash_table *nht = state->nht;
2316 if (state->neigh_sub_iter) {
2317 void *v = state->neigh_sub_iter(state, n, pos);
2321 n = rcu_dereference_bh(n->next);
2325 if (!net_eq(dev_net(n->dev), net))
2327 if (state->neigh_sub_iter) {
2328 void *v = state->neigh_sub_iter(state, n, pos);
2333 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2336 if (n->nud_state & ~NUD_NOARP)
2339 n = rcu_dereference_bh(n->next);
2345 if (++state->bucket >= (1 << nht->hash_shift))
2348 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2356 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2358 struct neighbour *n = neigh_get_first(seq);
2363 n = neigh_get_next(seq, n, pos);
2368 return *pos ? NULL : n;
2371 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2373 struct neigh_seq_state *state = seq->private;
2374 struct net *net = seq_file_net(seq);
2375 struct neigh_table *tbl = state->tbl;
2376 struct pneigh_entry *pn = NULL;
2377 int bucket = state->bucket;
2379 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2380 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2381 pn = tbl->phash_buckets[bucket];
2382 while (pn && !net_eq(pneigh_net(pn), net))
2387 state->bucket = bucket;
2392 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2393 struct pneigh_entry *pn,
2396 struct neigh_seq_state *state = seq->private;
2397 struct net *net = seq_file_net(seq);
2398 struct neigh_table *tbl = state->tbl;
2402 if (++state->bucket > PNEIGH_HASHMASK)
2404 pn = tbl->phash_buckets[state->bucket];
2405 while (pn && !net_eq(pneigh_net(pn), net))
2417 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2419 struct pneigh_entry *pn = pneigh_get_first(seq);
2424 pn = pneigh_get_next(seq, pn, pos);
2429 return *pos ? NULL : pn;
2432 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2434 struct neigh_seq_state *state = seq->private;
2436 loff_t idxpos = *pos;
2438 rc = neigh_get_idx(seq, &idxpos);
2439 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2440 rc = pneigh_get_idx(seq, &idxpos);
2445 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2448 struct neigh_seq_state *state = seq->private;
2452 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2455 state->nht = rcu_dereference_bh(tbl->nht);
2457 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2459 EXPORT_SYMBOL(neigh_seq_start);
2461 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2463 struct neigh_seq_state *state;
2466 if (v == SEQ_START_TOKEN) {
2467 rc = neigh_get_first(seq);
2471 state = seq->private;
2472 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2473 rc = neigh_get_next(seq, v, NULL);
2476 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2477 rc = pneigh_get_first(seq);
2479 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2480 rc = pneigh_get_next(seq, v, NULL);
2486 EXPORT_SYMBOL(neigh_seq_next);
2488 void neigh_seq_stop(struct seq_file *seq, void *v)
2491 rcu_read_unlock_bh();
2493 EXPORT_SYMBOL(neigh_seq_stop);
2495 /* statistics via seq_file */
2497 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2499 struct neigh_table *tbl = seq->private;
2503 return SEQ_START_TOKEN;
2505 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2506 if (!cpu_possible(cpu))
2509 return per_cpu_ptr(tbl->stats, cpu);
2514 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2516 struct neigh_table *tbl = seq->private;
2519 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2520 if (!cpu_possible(cpu))
2523 return per_cpu_ptr(tbl->stats, cpu);
2528 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2533 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2535 struct neigh_table *tbl = seq->private;
2536 struct neigh_statistics *st = v;
2538 if (v == SEQ_START_TOKEN) {
2539 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");
2543 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2544 "%08lx %08lx %08lx %08lx %08lx\n",
2545 atomic_read(&tbl->entries),
2556 st->rcv_probes_mcast,
2557 st->rcv_probes_ucast,
2559 st->periodic_gc_runs,
2567 static const struct seq_operations neigh_stat_seq_ops = {
2568 .start = neigh_stat_seq_start,
2569 .next = neigh_stat_seq_next,
2570 .stop = neigh_stat_seq_stop,
2571 .show = neigh_stat_seq_show,
2574 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2576 int ret = seq_open(file, &neigh_stat_seq_ops);
2579 struct seq_file *sf = file->private_data;
2580 sf->private = PDE(inode)->data;
2585 static const struct file_operations neigh_stat_seq_fops = {
2586 .owner = THIS_MODULE,
2587 .open = neigh_stat_seq_open,
2589 .llseek = seq_lseek,
2590 .release = seq_release,
2593 #endif /* CONFIG_PROC_FS */
2595 static inline size_t neigh_nlmsg_size(void)
2597 return NLMSG_ALIGN(sizeof(struct ndmsg))
2598 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2599 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2600 + nla_total_size(sizeof(struct nda_cacheinfo))
2601 + nla_total_size(4); /* NDA_PROBES */
2604 static void __neigh_notify(struct neighbour *n, int type, int flags)
2606 struct net *net = dev_net(n->dev);
2607 struct sk_buff *skb;
2610 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2614 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2616 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2617 WARN_ON(err == -EMSGSIZE);
2621 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2625 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2629 void neigh_app_ns(struct neighbour *n)
2631 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2633 EXPORT_SYMBOL(neigh_app_ns);
2634 #endif /* CONFIG_ARPD */
2636 #ifdef CONFIG_SYSCTL
2638 #define NEIGH_VARS_MAX 19
2640 static struct neigh_sysctl_table {
2641 struct ctl_table_header *sysctl_header;
2642 struct ctl_table neigh_vars[NEIGH_VARS_MAX];
2644 } neigh_sysctl_template __read_mostly = {
2647 .procname = "mcast_solicit",
2648 .maxlen = sizeof(int),
2650 .proc_handler = proc_dointvec,
2653 .procname = "ucast_solicit",
2654 .maxlen = sizeof(int),
2656 .proc_handler = proc_dointvec,
2659 .procname = "app_solicit",
2660 .maxlen = sizeof(int),
2662 .proc_handler = proc_dointvec,
2665 .procname = "retrans_time",
2666 .maxlen = sizeof(int),
2668 .proc_handler = proc_dointvec_userhz_jiffies,
2671 .procname = "base_reachable_time",
2672 .maxlen = sizeof(int),
2674 .proc_handler = proc_dointvec_jiffies,
2677 .procname = "delay_first_probe_time",
2678 .maxlen = sizeof(int),
2680 .proc_handler = proc_dointvec_jiffies,
2683 .procname = "gc_stale_time",
2684 .maxlen = sizeof(int),
2686 .proc_handler = proc_dointvec_jiffies,
2689 .procname = "unres_qlen",
2690 .maxlen = sizeof(int),
2692 .proc_handler = proc_dointvec,
2695 .procname = "proxy_qlen",
2696 .maxlen = sizeof(int),
2698 .proc_handler = proc_dointvec,
2701 .procname = "anycast_delay",
2702 .maxlen = sizeof(int),
2704 .proc_handler = proc_dointvec_userhz_jiffies,
2707 .procname = "proxy_delay",
2708 .maxlen = sizeof(int),
2710 .proc_handler = proc_dointvec_userhz_jiffies,
2713 .procname = "locktime",
2714 .maxlen = sizeof(int),
2716 .proc_handler = proc_dointvec_userhz_jiffies,
2719 .procname = "retrans_time_ms",
2720 .maxlen = sizeof(int),
2722 .proc_handler = proc_dointvec_ms_jiffies,
2725 .procname = "base_reachable_time_ms",
2726 .maxlen = sizeof(int),
2728 .proc_handler = proc_dointvec_ms_jiffies,
2731 .procname = "gc_interval",
2732 .maxlen = sizeof(int),
2734 .proc_handler = proc_dointvec_jiffies,
2737 .procname = "gc_thresh1",
2738 .maxlen = sizeof(int),
2740 .proc_handler = proc_dointvec,
2743 .procname = "gc_thresh2",
2744 .maxlen = sizeof(int),
2746 .proc_handler = proc_dointvec,
2749 .procname = "gc_thresh3",
2750 .maxlen = sizeof(int),
2752 .proc_handler = proc_dointvec,
2758 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2759 char *p_name, proc_handler *handler)
2761 struct neigh_sysctl_table *t;
2762 const char *dev_name_source = NULL;
2764 #define NEIGH_CTL_PATH_ROOT 0
2765 #define NEIGH_CTL_PATH_PROTO 1
2766 #define NEIGH_CTL_PATH_NEIGH 2
2767 #define NEIGH_CTL_PATH_DEV 3
2769 struct ctl_path neigh_path[] = {
2770 { .procname = "net", },
2771 { .procname = "proto", },
2772 { .procname = "neigh", },
2773 { .procname = "default", },
2777 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2781 t->neigh_vars[0].data = &p->mcast_probes;
2782 t->neigh_vars[1].data = &p->ucast_probes;
2783 t->neigh_vars[2].data = &p->app_probes;
2784 t->neigh_vars[3].data = &p->retrans_time;
2785 t->neigh_vars[4].data = &p->base_reachable_time;
2786 t->neigh_vars[5].data = &p->delay_probe_time;
2787 t->neigh_vars[6].data = &p->gc_staletime;
2788 t->neigh_vars[7].data = &p->queue_len;
2789 t->neigh_vars[8].data = &p->proxy_qlen;
2790 t->neigh_vars[9].data = &p->anycast_delay;
2791 t->neigh_vars[10].data = &p->proxy_delay;
2792 t->neigh_vars[11].data = &p->locktime;
2793 t->neigh_vars[12].data = &p->retrans_time;
2794 t->neigh_vars[13].data = &p->base_reachable_time;
2797 dev_name_source = dev->name;
2798 /* Terminate the table early */
2799 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
2801 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2802 t->neigh_vars[14].data = (int *)(p + 1);
2803 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2804 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2805 t->neigh_vars[17].data = (int *)(p + 1) + 3;
2811 t->neigh_vars[3].proc_handler = handler;
2812 t->neigh_vars[3].extra1 = dev;
2814 t->neigh_vars[4].proc_handler = handler;
2815 t->neigh_vars[4].extra1 = dev;
2816 /* RetransTime (in milliseconds)*/
2817 t->neigh_vars[12].proc_handler = handler;
2818 t->neigh_vars[12].extra1 = dev;
2819 /* ReachableTime (in milliseconds) */
2820 t->neigh_vars[13].proc_handler = handler;
2821 t->neigh_vars[13].extra1 = dev;
2824 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2828 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2829 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2832 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2833 if (!t->sysctl_header)
2836 p->sysctl_table = t;
2846 EXPORT_SYMBOL(neigh_sysctl_register);
2848 void neigh_sysctl_unregister(struct neigh_parms *p)
2850 if (p->sysctl_table) {
2851 struct neigh_sysctl_table *t = p->sysctl_table;
2852 p->sysctl_table = NULL;
2853 unregister_sysctl_table(t->sysctl_header);
2858 EXPORT_SYMBOL(neigh_sysctl_unregister);
2860 #endif /* CONFIG_SYSCTL */
2862 static int __init neigh_init(void)
2864 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, NULL);
2865 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, NULL);
2866 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, NULL);
2868 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
2870 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, NULL);
2875 subsys_initcall(neigh_init);