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 write_lock_bh(&neigh->lock);
705 __skb_queue_purge(&neigh->arp_queue);
706 write_unlock_bh(&neigh->lock);
709 neigh_parms_put(neigh->parms);
711 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
713 atomic_dec(&neigh->tbl->entries);
714 call_rcu(&neigh->rcu, neigh_destroy_rcu);
716 EXPORT_SYMBOL(neigh_destroy);
718 /* Neighbour state is suspicious;
721 Called with write_locked neigh.
723 static void neigh_suspect(struct neighbour *neigh)
725 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
727 neigh->output = neigh->ops->output;
730 /* Neighbour state is OK;
733 Called with write_locked neigh.
735 static void neigh_connect(struct neighbour *neigh)
737 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
739 neigh->output = neigh->ops->connected_output;
742 static void neigh_periodic_work(struct work_struct *work)
744 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
746 struct neighbour __rcu **np;
748 struct neigh_hash_table *nht;
750 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
752 write_lock_bh(&tbl->lock);
753 nht = rcu_dereference_protected(tbl->nht,
754 lockdep_is_held(&tbl->lock));
757 * periodically recompute ReachableTime from random function
760 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
761 struct neigh_parms *p;
762 tbl->last_rand = jiffies;
763 for (p = &tbl->parms; p; p = p->next)
765 neigh_rand_reach_time(p->base_reachable_time);
768 for (i = 0 ; i < (1 << nht->hash_shift); i++) {
769 np = &nht->hash_buckets[i];
771 while ((n = rcu_dereference_protected(*np,
772 lockdep_is_held(&tbl->lock))) != NULL) {
775 write_lock(&n->lock);
777 state = n->nud_state;
778 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
779 write_unlock(&n->lock);
783 if (time_before(n->used, n->confirmed))
784 n->used = n->confirmed;
786 if (atomic_read(&n->refcnt) == 1 &&
787 (state == NUD_FAILED ||
788 time_after(jiffies, n->used + n->parms->gc_staletime))) {
791 write_unlock(&n->lock);
792 neigh_cleanup_and_release(n);
795 write_unlock(&n->lock);
801 * It's fine to release lock here, even if hash table
802 * grows while we are preempted.
804 write_unlock_bh(&tbl->lock);
806 write_lock_bh(&tbl->lock);
807 nht = rcu_dereference_protected(tbl->nht,
808 lockdep_is_held(&tbl->lock));
810 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
811 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
812 * base_reachable_time.
814 schedule_delayed_work(&tbl->gc_work,
815 tbl->parms.base_reachable_time >> 1);
816 write_unlock_bh(&tbl->lock);
819 static __inline__ int neigh_max_probes(struct neighbour *n)
821 struct neigh_parms *p = n->parms;
822 return (n->nud_state & NUD_PROBE) ?
824 p->ucast_probes + p->app_probes + p->mcast_probes;
827 static void neigh_invalidate(struct neighbour *neigh)
828 __releases(neigh->lock)
829 __acquires(neigh->lock)
833 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
834 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
835 neigh->updated = jiffies;
837 /* It is very thin place. report_unreachable is very complicated
838 routine. Particularly, it can hit the same neighbour entry!
840 So that, we try to be accurate and avoid dead loop. --ANK
842 while (neigh->nud_state == NUD_FAILED &&
843 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
844 write_unlock(&neigh->lock);
845 neigh->ops->error_report(neigh, skb);
846 write_lock(&neigh->lock);
848 __skb_queue_purge(&neigh->arp_queue);
851 static void neigh_probe(struct neighbour *neigh)
852 __releases(neigh->lock)
854 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
855 /* keep skb alive even if arp_queue overflows */
857 skb = skb_copy(skb, GFP_ATOMIC);
858 write_unlock(&neigh->lock);
859 neigh->ops->solicit(neigh, skb);
860 atomic_inc(&neigh->probes);
864 /* Called when a timer expires for a neighbour entry. */
866 static void neigh_timer_handler(unsigned long arg)
868 unsigned long now, next;
869 struct neighbour *neigh = (struct neighbour *)arg;
873 write_lock(&neigh->lock);
875 state = neigh->nud_state;
879 if (!(state & NUD_IN_TIMER))
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))
961 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
962 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
963 unsigned long next, now = jiffies;
965 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
966 neigh->nud_state = NUD_INCOMPLETE;
967 neigh->updated = now;
968 next = now + max(neigh->parms->retrans_time, HZ/2);
969 neigh_add_timer(neigh, next);
970 immediate_probe = true;
972 neigh->nud_state = NUD_FAILED;
973 neigh->updated = jiffies;
974 write_unlock_bh(&neigh->lock);
979 } else if (neigh->nud_state & NUD_STALE) {
980 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
981 neigh->nud_state = NUD_DELAY;
982 neigh->updated = jiffies;
983 neigh_add_timer(neigh,
984 jiffies + neigh->parms->delay_probe_time);
987 if (neigh->nud_state == NUD_INCOMPLETE) {
989 if (skb_queue_len(&neigh->arp_queue) >=
990 neigh->parms->queue_len) {
991 struct sk_buff *buff;
992 buff = __skb_dequeue(&neigh->arp_queue);
994 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
997 __skb_queue_tail(&neigh->arp_queue, skb);
1002 if (immediate_probe)
1005 write_unlock(&neigh->lock);
1010 if (neigh->nud_state & NUD_STALE)
1012 write_unlock_bh(&neigh->lock);
1016 EXPORT_SYMBOL(__neigh_event_send);
1018 static void neigh_update_hhs(struct neighbour *neigh)
1020 struct hh_cache *hh;
1021 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1024 if (neigh->dev->header_ops)
1025 update = neigh->dev->header_ops->cache_update;
1030 write_seqlock_bh(&hh->hh_lock);
1031 update(hh, neigh->dev, neigh->ha);
1032 write_sequnlock_bh(&hh->hh_lock);
1039 /* Generic update routine.
1040 -- lladdr is new lladdr or NULL, if it is not supplied.
1041 -- new is new state.
1043 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1045 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1046 lladdr instead of overriding it
1048 It also allows to retain current state
1049 if lladdr is unchanged.
1050 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1052 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1054 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1057 Caller MUST hold reference count on the entry.
1060 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1066 struct net_device *dev;
1067 int update_isrouter = 0;
1069 write_lock_bh(&neigh->lock);
1072 old = neigh->nud_state;
1075 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1076 (old & (NUD_NOARP | NUD_PERMANENT)))
1081 if (!(new & NUD_VALID)) {
1082 neigh_del_timer(neigh);
1083 if (old & NUD_CONNECTED)
1084 neigh_suspect(neigh);
1085 neigh->nud_state = new;
1087 notify = old & NUD_VALID;
1088 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1089 (new & NUD_FAILED)) {
1090 neigh_invalidate(neigh);
1096 /* Compare new lladdr with cached one */
1097 if (!dev->addr_len) {
1098 /* First case: device needs no address. */
1100 } else if (lladdr) {
1101 /* The second case: if something is already cached
1102 and a new address is proposed:
1104 - if they are different, check override flag
1106 if ((old & NUD_VALID) &&
1107 !memcmp(lladdr, neigh->ha, dev->addr_len))
1110 /* No address is supplied; if we know something,
1111 use it, otherwise discard the request.
1114 if (!(old & NUD_VALID))
1119 if (new & NUD_CONNECTED)
1120 neigh->confirmed = jiffies;
1121 neigh->updated = jiffies;
1123 /* If entry was valid and address is not changed,
1124 do not change entry state, if new one is STALE.
1127 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1128 if (old & NUD_VALID) {
1129 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1130 update_isrouter = 0;
1131 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1132 (old & NUD_CONNECTED)) {
1138 if (lladdr == neigh->ha && new == NUD_STALE &&
1139 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1140 (old & NUD_CONNECTED))
1147 neigh_del_timer(neigh);
1148 if (new & NUD_IN_TIMER)
1149 neigh_add_timer(neigh, (jiffies +
1150 ((new & NUD_REACHABLE) ?
1151 neigh->parms->reachable_time :
1153 neigh->nud_state = new;
1156 if (lladdr != neigh->ha) {
1157 write_seqlock(&neigh->ha_lock);
1158 memcpy(&neigh->ha, lladdr, dev->addr_len);
1159 write_sequnlock(&neigh->ha_lock);
1160 neigh_update_hhs(neigh);
1161 if (!(new & NUD_CONNECTED))
1162 neigh->confirmed = jiffies -
1163 (neigh->parms->base_reachable_time << 1);
1168 if (new & NUD_CONNECTED)
1169 neigh_connect(neigh);
1171 neigh_suspect(neigh);
1172 if (!(old & NUD_VALID)) {
1173 struct sk_buff *skb;
1175 /* Again: avoid dead loop if something went wrong */
1177 while (neigh->nud_state & NUD_VALID &&
1178 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1179 struct dst_entry *dst = skb_dst(skb);
1180 struct neighbour *n2, *n1 = neigh;
1181 write_unlock_bh(&neigh->lock);
1184 /* On shaper/eql skb->dst->neighbour != neigh :( */
1185 if (dst && (n2 = dst_get_neighbour(dst)) != NULL)
1187 n1->output(n1, skb);
1190 write_lock_bh(&neigh->lock);
1192 __skb_queue_purge(&neigh->arp_queue);
1195 if (update_isrouter) {
1196 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1197 (neigh->flags | NTF_ROUTER) :
1198 (neigh->flags & ~NTF_ROUTER);
1200 write_unlock_bh(&neigh->lock);
1203 neigh_update_notify(neigh);
1207 EXPORT_SYMBOL(neigh_update);
1209 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1210 u8 *lladdr, void *saddr,
1211 struct net_device *dev)
1213 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1214 lladdr || !dev->addr_len);
1216 neigh_update(neigh, lladdr, NUD_STALE,
1217 NEIGH_UPDATE_F_OVERRIDE);
1220 EXPORT_SYMBOL(neigh_event_ns);
1222 /* called with read_lock_bh(&n->lock); */
1223 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst)
1225 struct net_device *dev = dst->dev;
1226 __be16 prot = dst->ops->protocol;
1227 struct hh_cache *hh = &n->hh;
1229 write_lock_bh(&n->lock);
1231 /* Only one thread can come in here and initialize the
1235 dev->header_ops->cache(n, hh, prot);
1237 write_unlock_bh(&n->lock);
1240 /* This function can be used in contexts, where only old dev_queue_xmit
1241 * worked, f.e. if you want to override normal output path (eql, shaper),
1242 * but resolution is not made yet.
1245 int neigh_compat_output(struct neighbour *neigh, struct sk_buff *skb)
1247 struct net_device *dev = skb->dev;
1249 __skb_pull(skb, skb_network_offset(skb));
1251 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1253 dev->header_ops->rebuild(skb))
1256 return dev_queue_xmit(skb);
1258 EXPORT_SYMBOL(neigh_compat_output);
1260 /* Slow and careful. */
1262 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1264 struct dst_entry *dst = skb_dst(skb);
1270 if (!neigh_event_send(neigh, skb)) {
1272 struct net_device *dev = neigh->dev;
1275 if (dev->header_ops->cache && !neigh->hh.hh_len)
1276 neigh_hh_init(neigh, dst);
1279 __skb_pull(skb, skb_network_offset(skb));
1280 seq = read_seqbegin(&neigh->ha_lock);
1281 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1282 neigh->ha, NULL, skb->len);
1283 } while (read_seqretry(&neigh->ha_lock, seq));
1286 rc = dev_queue_xmit(skb);
1293 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1300 EXPORT_SYMBOL(neigh_resolve_output);
1302 /* As fast as possible without hh cache */
1304 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1306 struct net_device *dev = neigh->dev;
1311 __skb_pull(skb, skb_network_offset(skb));
1312 seq = read_seqbegin(&neigh->ha_lock);
1313 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1314 neigh->ha, NULL, skb->len);
1315 } while (read_seqretry(&neigh->ha_lock, seq));
1318 err = dev_queue_xmit(skb);
1325 EXPORT_SYMBOL(neigh_connected_output);
1327 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1329 return dev_queue_xmit(skb);
1331 EXPORT_SYMBOL(neigh_direct_output);
1333 static void neigh_proxy_process(unsigned long arg)
1335 struct neigh_table *tbl = (struct neigh_table *)arg;
1336 long sched_next = 0;
1337 unsigned long now = jiffies;
1338 struct sk_buff *skb, *n;
1340 spin_lock(&tbl->proxy_queue.lock);
1342 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1343 long tdif = NEIGH_CB(skb)->sched_next - now;
1346 struct net_device *dev = skb->dev;
1348 __skb_unlink(skb, &tbl->proxy_queue);
1349 if (tbl->proxy_redo && netif_running(dev)) {
1351 tbl->proxy_redo(skb);
1358 } else if (!sched_next || tdif < sched_next)
1361 del_timer(&tbl->proxy_timer);
1363 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1364 spin_unlock(&tbl->proxy_queue.lock);
1367 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1368 struct sk_buff *skb)
1370 unsigned long now = jiffies;
1371 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1373 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1378 NEIGH_CB(skb)->sched_next = sched_next;
1379 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1381 spin_lock(&tbl->proxy_queue.lock);
1382 if (del_timer(&tbl->proxy_timer)) {
1383 if (time_before(tbl->proxy_timer.expires, sched_next))
1384 sched_next = tbl->proxy_timer.expires;
1388 __skb_queue_tail(&tbl->proxy_queue, skb);
1389 mod_timer(&tbl->proxy_timer, sched_next);
1390 spin_unlock(&tbl->proxy_queue.lock);
1392 EXPORT_SYMBOL(pneigh_enqueue);
1394 static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1395 struct net *net, int ifindex)
1397 struct neigh_parms *p;
1399 for (p = &tbl->parms; p; p = p->next) {
1400 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1401 (!p->dev && !ifindex))
1408 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1409 struct neigh_table *tbl)
1411 struct neigh_parms *p, *ref;
1412 struct net *net = dev_net(dev);
1413 const struct net_device_ops *ops = dev->netdev_ops;
1415 ref = lookup_neigh_parms(tbl, net, 0);
1419 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1422 atomic_set(&p->refcnt, 1);
1424 neigh_rand_reach_time(p->base_reachable_time);
1426 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1433 write_pnet(&p->net, hold_net(net));
1434 p->sysctl_table = NULL;
1435 write_lock_bh(&tbl->lock);
1436 p->next = tbl->parms.next;
1437 tbl->parms.next = p;
1438 write_unlock_bh(&tbl->lock);
1442 EXPORT_SYMBOL(neigh_parms_alloc);
1444 static void neigh_rcu_free_parms(struct rcu_head *head)
1446 struct neigh_parms *parms =
1447 container_of(head, struct neigh_parms, rcu_head);
1449 neigh_parms_put(parms);
1452 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1454 struct neigh_parms **p;
1456 if (!parms || parms == &tbl->parms)
1458 write_lock_bh(&tbl->lock);
1459 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1463 write_unlock_bh(&tbl->lock);
1465 dev_put(parms->dev);
1466 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1470 write_unlock_bh(&tbl->lock);
1471 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1473 EXPORT_SYMBOL(neigh_parms_release);
1475 static void neigh_parms_destroy(struct neigh_parms *parms)
1477 release_net(neigh_parms_net(parms));
1481 static struct lock_class_key neigh_table_proxy_queue_class;
1483 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1485 unsigned long now = jiffies;
1486 unsigned long phsize;
1488 write_pnet(&tbl->parms.net, &init_net);
1489 atomic_set(&tbl->parms.refcnt, 1);
1490 tbl->parms.reachable_time =
1491 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1493 if (!tbl->kmem_cachep)
1495 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1496 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1498 tbl->stats = alloc_percpu(struct neigh_statistics);
1500 panic("cannot create neighbour cache statistics");
1502 #ifdef CONFIG_PROC_FS
1503 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1504 &neigh_stat_seq_fops, tbl))
1505 panic("cannot create neighbour proc dir entry");
1508 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1510 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1511 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1513 if (!tbl->nht || !tbl->phash_buckets)
1514 panic("cannot allocate neighbour cache hashes");
1516 rwlock_init(&tbl->lock);
1517 INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1518 schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1519 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1520 skb_queue_head_init_class(&tbl->proxy_queue,
1521 &neigh_table_proxy_queue_class);
1523 tbl->last_flush = now;
1524 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1526 EXPORT_SYMBOL(neigh_table_init_no_netlink);
1528 void neigh_table_init(struct neigh_table *tbl)
1530 struct neigh_table *tmp;
1532 neigh_table_init_no_netlink(tbl);
1533 write_lock(&neigh_tbl_lock);
1534 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1535 if (tmp->family == tbl->family)
1538 tbl->next = neigh_tables;
1540 write_unlock(&neigh_tbl_lock);
1542 if (unlikely(tmp)) {
1543 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1544 "family %d\n", tbl->family);
1548 EXPORT_SYMBOL(neigh_table_init);
1550 int neigh_table_clear(struct neigh_table *tbl)
1552 struct neigh_table **tp;
1554 /* It is not clean... Fix it to unload IPv6 module safely */
1555 cancel_delayed_work_sync(&tbl->gc_work);
1556 del_timer_sync(&tbl->proxy_timer);
1557 pneigh_queue_purge(&tbl->proxy_queue);
1558 neigh_ifdown(tbl, NULL);
1559 if (atomic_read(&tbl->entries))
1560 printk(KERN_CRIT "neighbour leakage\n");
1561 write_lock(&neigh_tbl_lock);
1562 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1568 write_unlock(&neigh_tbl_lock);
1570 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1571 neigh_hash_free_rcu);
1574 kfree(tbl->phash_buckets);
1575 tbl->phash_buckets = NULL;
1577 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1579 free_percpu(tbl->stats);
1582 kmem_cache_destroy(tbl->kmem_cachep);
1583 tbl->kmem_cachep = NULL;
1587 EXPORT_SYMBOL(neigh_table_clear);
1589 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1591 struct net *net = sock_net(skb->sk);
1593 struct nlattr *dst_attr;
1594 struct neigh_table *tbl;
1595 struct net_device *dev = NULL;
1599 if (nlmsg_len(nlh) < sizeof(*ndm))
1602 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1603 if (dst_attr == NULL)
1606 ndm = nlmsg_data(nlh);
1607 if (ndm->ndm_ifindex) {
1608 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1615 read_lock(&neigh_tbl_lock);
1616 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1617 struct neighbour *neigh;
1619 if (tbl->family != ndm->ndm_family)
1621 read_unlock(&neigh_tbl_lock);
1623 if (nla_len(dst_attr) < tbl->key_len)
1626 if (ndm->ndm_flags & NTF_PROXY) {
1627 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1634 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1635 if (neigh == NULL) {
1640 err = neigh_update(neigh, NULL, NUD_FAILED,
1641 NEIGH_UPDATE_F_OVERRIDE |
1642 NEIGH_UPDATE_F_ADMIN);
1643 neigh_release(neigh);
1646 read_unlock(&neigh_tbl_lock);
1647 err = -EAFNOSUPPORT;
1653 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1655 struct net *net = sock_net(skb->sk);
1657 struct nlattr *tb[NDA_MAX+1];
1658 struct neigh_table *tbl;
1659 struct net_device *dev = NULL;
1663 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1668 if (tb[NDA_DST] == NULL)
1671 ndm = nlmsg_data(nlh);
1672 if (ndm->ndm_ifindex) {
1673 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1679 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1683 read_lock(&neigh_tbl_lock);
1684 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1685 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1686 struct neighbour *neigh;
1689 if (tbl->family != ndm->ndm_family)
1691 read_unlock(&neigh_tbl_lock);
1693 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1695 dst = nla_data(tb[NDA_DST]);
1696 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1698 if (ndm->ndm_flags & NTF_PROXY) {
1699 struct pneigh_entry *pn;
1702 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1704 pn->flags = ndm->ndm_flags;
1713 neigh = neigh_lookup(tbl, dst, dev);
1714 if (neigh == NULL) {
1715 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1720 neigh = __neigh_lookup_errno(tbl, dst, dev);
1721 if (IS_ERR(neigh)) {
1722 err = PTR_ERR(neigh);
1726 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1728 neigh_release(neigh);
1732 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1733 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1736 if (ndm->ndm_flags & NTF_USE) {
1737 neigh_event_send(neigh, NULL);
1740 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1741 neigh_release(neigh);
1745 read_unlock(&neigh_tbl_lock);
1746 err = -EAFNOSUPPORT;
1751 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1753 struct nlattr *nest;
1755 nest = nla_nest_start(skb, NDTA_PARMS);
1760 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1762 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1763 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1764 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1765 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1766 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1767 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1768 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1769 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1770 parms->base_reachable_time);
1771 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1772 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1773 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1774 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1775 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1776 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1778 return nla_nest_end(skb, nest);
1781 nla_nest_cancel(skb, nest);
1785 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1786 u32 pid, u32 seq, int type, int flags)
1788 struct nlmsghdr *nlh;
1789 struct ndtmsg *ndtmsg;
1791 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1795 ndtmsg = nlmsg_data(nlh);
1797 read_lock_bh(&tbl->lock);
1798 ndtmsg->ndtm_family = tbl->family;
1799 ndtmsg->ndtm_pad1 = 0;
1800 ndtmsg->ndtm_pad2 = 0;
1802 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1803 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1804 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1805 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1806 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1809 unsigned long now = jiffies;
1810 unsigned int flush_delta = now - tbl->last_flush;
1811 unsigned int rand_delta = now - tbl->last_rand;
1812 struct neigh_hash_table *nht;
1813 struct ndt_config ndc = {
1814 .ndtc_key_len = tbl->key_len,
1815 .ndtc_entry_size = tbl->entry_size,
1816 .ndtc_entries = atomic_read(&tbl->entries),
1817 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1818 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1819 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1823 nht = rcu_dereference_bh(tbl->nht);
1824 ndc.ndtc_hash_rnd = nht->hash_rnd;
1825 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1826 rcu_read_unlock_bh();
1828 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1833 struct ndt_stats ndst;
1835 memset(&ndst, 0, sizeof(ndst));
1837 for_each_possible_cpu(cpu) {
1838 struct neigh_statistics *st;
1840 st = per_cpu_ptr(tbl->stats, cpu);
1841 ndst.ndts_allocs += st->allocs;
1842 ndst.ndts_destroys += st->destroys;
1843 ndst.ndts_hash_grows += st->hash_grows;
1844 ndst.ndts_res_failed += st->res_failed;
1845 ndst.ndts_lookups += st->lookups;
1846 ndst.ndts_hits += st->hits;
1847 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1848 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1849 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1850 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1853 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1856 BUG_ON(tbl->parms.dev);
1857 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1858 goto nla_put_failure;
1860 read_unlock_bh(&tbl->lock);
1861 return nlmsg_end(skb, nlh);
1864 read_unlock_bh(&tbl->lock);
1865 nlmsg_cancel(skb, nlh);
1869 static int neightbl_fill_param_info(struct sk_buff *skb,
1870 struct neigh_table *tbl,
1871 struct neigh_parms *parms,
1872 u32 pid, u32 seq, int type,
1875 struct ndtmsg *ndtmsg;
1876 struct nlmsghdr *nlh;
1878 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1882 ndtmsg = nlmsg_data(nlh);
1884 read_lock_bh(&tbl->lock);
1885 ndtmsg->ndtm_family = tbl->family;
1886 ndtmsg->ndtm_pad1 = 0;
1887 ndtmsg->ndtm_pad2 = 0;
1889 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1890 neightbl_fill_parms(skb, parms) < 0)
1893 read_unlock_bh(&tbl->lock);
1894 return nlmsg_end(skb, nlh);
1896 read_unlock_bh(&tbl->lock);
1897 nlmsg_cancel(skb, nlh);
1901 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1902 [NDTA_NAME] = { .type = NLA_STRING },
1903 [NDTA_THRESH1] = { .type = NLA_U32 },
1904 [NDTA_THRESH2] = { .type = NLA_U32 },
1905 [NDTA_THRESH3] = { .type = NLA_U32 },
1906 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1907 [NDTA_PARMS] = { .type = NLA_NESTED },
1910 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1911 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1912 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1913 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1914 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1915 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1916 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1917 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1918 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1919 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1920 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1921 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1922 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1923 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1926 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1928 struct net *net = sock_net(skb->sk);
1929 struct neigh_table *tbl;
1930 struct ndtmsg *ndtmsg;
1931 struct nlattr *tb[NDTA_MAX+1];
1934 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1935 nl_neightbl_policy);
1939 if (tb[NDTA_NAME] == NULL) {
1944 ndtmsg = nlmsg_data(nlh);
1945 read_lock(&neigh_tbl_lock);
1946 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1947 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1950 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1960 * We acquire tbl->lock to be nice to the periodic timers and
1961 * make sure they always see a consistent set of values.
1963 write_lock_bh(&tbl->lock);
1965 if (tb[NDTA_PARMS]) {
1966 struct nlattr *tbp[NDTPA_MAX+1];
1967 struct neigh_parms *p;
1970 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1971 nl_ntbl_parm_policy);
1973 goto errout_tbl_lock;
1975 if (tbp[NDTPA_IFINDEX])
1976 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1978 p = lookup_neigh_parms(tbl, net, ifindex);
1981 goto errout_tbl_lock;
1984 for (i = 1; i <= NDTPA_MAX; i++) {
1989 case NDTPA_QUEUE_LEN:
1990 p->queue_len = nla_get_u32(tbp[i]);
1992 case NDTPA_PROXY_QLEN:
1993 p->proxy_qlen = nla_get_u32(tbp[i]);
1995 case NDTPA_APP_PROBES:
1996 p->app_probes = nla_get_u32(tbp[i]);
1998 case NDTPA_UCAST_PROBES:
1999 p->ucast_probes = nla_get_u32(tbp[i]);
2001 case NDTPA_MCAST_PROBES:
2002 p->mcast_probes = nla_get_u32(tbp[i]);
2004 case NDTPA_BASE_REACHABLE_TIME:
2005 p->base_reachable_time = nla_get_msecs(tbp[i]);
2007 case NDTPA_GC_STALETIME:
2008 p->gc_staletime = nla_get_msecs(tbp[i]);
2010 case NDTPA_DELAY_PROBE_TIME:
2011 p->delay_probe_time = nla_get_msecs(tbp[i]);
2013 case NDTPA_RETRANS_TIME:
2014 p->retrans_time = nla_get_msecs(tbp[i]);
2016 case NDTPA_ANYCAST_DELAY:
2017 p->anycast_delay = nla_get_msecs(tbp[i]);
2019 case NDTPA_PROXY_DELAY:
2020 p->proxy_delay = nla_get_msecs(tbp[i]);
2022 case NDTPA_LOCKTIME:
2023 p->locktime = nla_get_msecs(tbp[i]);
2029 if (tb[NDTA_THRESH1])
2030 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2032 if (tb[NDTA_THRESH2])
2033 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2035 if (tb[NDTA_THRESH3])
2036 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2038 if (tb[NDTA_GC_INTERVAL])
2039 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2044 write_unlock_bh(&tbl->lock);
2046 read_unlock(&neigh_tbl_lock);
2051 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2053 struct net *net = sock_net(skb->sk);
2054 int family, tidx, nidx = 0;
2055 int tbl_skip = cb->args[0];
2056 int neigh_skip = cb->args[1];
2057 struct neigh_table *tbl;
2059 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2061 read_lock(&neigh_tbl_lock);
2062 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
2063 struct neigh_parms *p;
2065 if (tidx < tbl_skip || (family && tbl->family != family))
2068 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2069 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2073 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
2074 if (!net_eq(neigh_parms_net(p), net))
2077 if (nidx < neigh_skip)
2080 if (neightbl_fill_param_info(skb, tbl, p,
2081 NETLINK_CB(cb->skb).pid,
2093 read_unlock(&neigh_tbl_lock);
2100 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2101 u32 pid, u32 seq, int type, unsigned int flags)
2103 unsigned long now = jiffies;
2104 struct nda_cacheinfo ci;
2105 struct nlmsghdr *nlh;
2108 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2112 ndm = nlmsg_data(nlh);
2113 ndm->ndm_family = neigh->ops->family;
2116 ndm->ndm_flags = neigh->flags;
2117 ndm->ndm_type = neigh->type;
2118 ndm->ndm_ifindex = neigh->dev->ifindex;
2120 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2122 read_lock_bh(&neigh->lock);
2123 ndm->ndm_state = neigh->nud_state;
2124 if (neigh->nud_state & NUD_VALID) {
2125 char haddr[MAX_ADDR_LEN];
2127 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2128 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2129 read_unlock_bh(&neigh->lock);
2130 goto nla_put_failure;
2134 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2135 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2136 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2137 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2138 read_unlock_bh(&neigh->lock);
2140 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2141 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2143 return nlmsg_end(skb, nlh);
2146 nlmsg_cancel(skb, nlh);
2150 static void neigh_update_notify(struct neighbour *neigh)
2152 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2153 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2156 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2157 struct netlink_callback *cb)
2159 struct net *net = sock_net(skb->sk);
2160 struct neighbour *n;
2161 int rc, h, s_h = cb->args[1];
2162 int idx, s_idx = idx = cb->args[2];
2163 struct neigh_hash_table *nht;
2166 nht = rcu_dereference_bh(tbl->nht);
2168 for (h = 0; h < (1 << nht->hash_shift); h++) {
2173 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2175 n = rcu_dereference_bh(n->next)) {
2176 if (!net_eq(dev_net(n->dev), net))
2180 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2183 NLM_F_MULTI) <= 0) {
2193 rcu_read_unlock_bh();
2199 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2201 struct neigh_table *tbl;
2204 read_lock(&neigh_tbl_lock);
2205 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2208 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2209 if (t < s_t || (family && tbl->family != family))
2212 memset(&cb->args[1], 0, sizeof(cb->args) -
2213 sizeof(cb->args[0]));
2214 if (neigh_dump_table(tbl, skb, cb) < 0)
2217 read_unlock(&neigh_tbl_lock);
2223 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2226 struct neigh_hash_table *nht;
2229 nht = rcu_dereference_bh(tbl->nht);
2231 read_lock(&tbl->lock); /* avoid resizes */
2232 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2233 struct neighbour *n;
2235 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2237 n = rcu_dereference_bh(n->next))
2240 read_unlock(&tbl->lock);
2241 rcu_read_unlock_bh();
2243 EXPORT_SYMBOL(neigh_for_each);
2245 /* The tbl->lock must be held as a writer and BH disabled. */
2246 void __neigh_for_each_release(struct neigh_table *tbl,
2247 int (*cb)(struct neighbour *))
2250 struct neigh_hash_table *nht;
2252 nht = rcu_dereference_protected(tbl->nht,
2253 lockdep_is_held(&tbl->lock));
2254 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2255 struct neighbour *n;
2256 struct neighbour __rcu **np;
2258 np = &nht->hash_buckets[chain];
2259 while ((n = rcu_dereference_protected(*np,
2260 lockdep_is_held(&tbl->lock))) != NULL) {
2263 write_lock(&n->lock);
2266 rcu_assign_pointer(*np,
2267 rcu_dereference_protected(n->next,
2268 lockdep_is_held(&tbl->lock)));
2272 write_unlock(&n->lock);
2274 neigh_cleanup_and_release(n);
2278 EXPORT_SYMBOL(__neigh_for_each_release);
2280 #ifdef CONFIG_PROC_FS
2282 static struct neighbour *neigh_get_first(struct seq_file *seq)
2284 struct neigh_seq_state *state = seq->private;
2285 struct net *net = seq_file_net(seq);
2286 struct neigh_hash_table *nht = state->nht;
2287 struct neighbour *n = NULL;
2288 int bucket = state->bucket;
2290 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2291 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2292 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2295 if (!net_eq(dev_net(n->dev), net))
2297 if (state->neigh_sub_iter) {
2301 v = state->neigh_sub_iter(state, n, &fakep);
2305 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2307 if (n->nud_state & ~NUD_NOARP)
2310 n = rcu_dereference_bh(n->next);
2316 state->bucket = bucket;
2321 static struct neighbour *neigh_get_next(struct seq_file *seq,
2322 struct neighbour *n,
2325 struct neigh_seq_state *state = seq->private;
2326 struct net *net = seq_file_net(seq);
2327 struct neigh_hash_table *nht = state->nht;
2329 if (state->neigh_sub_iter) {
2330 void *v = state->neigh_sub_iter(state, n, pos);
2334 n = rcu_dereference_bh(n->next);
2338 if (!net_eq(dev_net(n->dev), net))
2340 if (state->neigh_sub_iter) {
2341 void *v = state->neigh_sub_iter(state, n, pos);
2346 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2349 if (n->nud_state & ~NUD_NOARP)
2352 n = rcu_dereference_bh(n->next);
2358 if (++state->bucket >= (1 << nht->hash_shift))
2361 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2369 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2371 struct neighbour *n = neigh_get_first(seq);
2376 n = neigh_get_next(seq, n, pos);
2381 return *pos ? NULL : n;
2384 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2386 struct neigh_seq_state *state = seq->private;
2387 struct net *net = seq_file_net(seq);
2388 struct neigh_table *tbl = state->tbl;
2389 struct pneigh_entry *pn = NULL;
2390 int bucket = state->bucket;
2392 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2393 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2394 pn = tbl->phash_buckets[bucket];
2395 while (pn && !net_eq(pneigh_net(pn), net))
2400 state->bucket = bucket;
2405 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2406 struct pneigh_entry *pn,
2409 struct neigh_seq_state *state = seq->private;
2410 struct net *net = seq_file_net(seq);
2411 struct neigh_table *tbl = state->tbl;
2415 } while (pn && !net_eq(pneigh_net(pn), net));
2418 if (++state->bucket > PNEIGH_HASHMASK)
2420 pn = tbl->phash_buckets[state->bucket];
2421 while (pn && !net_eq(pneigh_net(pn), net))
2433 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2435 struct pneigh_entry *pn = pneigh_get_first(seq);
2440 pn = pneigh_get_next(seq, pn, pos);
2445 return *pos ? NULL : pn;
2448 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2450 struct neigh_seq_state *state = seq->private;
2452 loff_t idxpos = *pos;
2454 rc = neigh_get_idx(seq, &idxpos);
2455 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2456 rc = pneigh_get_idx(seq, &idxpos);
2461 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2464 struct neigh_seq_state *state = seq->private;
2468 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2471 state->nht = rcu_dereference_bh(tbl->nht);
2473 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2475 EXPORT_SYMBOL(neigh_seq_start);
2477 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2479 struct neigh_seq_state *state;
2482 if (v == SEQ_START_TOKEN) {
2483 rc = neigh_get_first(seq);
2487 state = seq->private;
2488 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2489 rc = neigh_get_next(seq, v, NULL);
2492 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2493 rc = pneigh_get_first(seq);
2495 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2496 rc = pneigh_get_next(seq, v, NULL);
2502 EXPORT_SYMBOL(neigh_seq_next);
2504 void neigh_seq_stop(struct seq_file *seq, void *v)
2507 rcu_read_unlock_bh();
2509 EXPORT_SYMBOL(neigh_seq_stop);
2511 /* statistics via seq_file */
2513 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2515 struct neigh_table *tbl = seq->private;
2519 return SEQ_START_TOKEN;
2521 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2522 if (!cpu_possible(cpu))
2525 return per_cpu_ptr(tbl->stats, cpu);
2530 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2532 struct neigh_table *tbl = seq->private;
2535 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2536 if (!cpu_possible(cpu))
2539 return per_cpu_ptr(tbl->stats, cpu);
2544 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2549 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2551 struct neigh_table *tbl = seq->private;
2552 struct neigh_statistics *st = v;
2554 if (v == SEQ_START_TOKEN) {
2555 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");
2559 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2560 "%08lx %08lx %08lx %08lx %08lx\n",
2561 atomic_read(&tbl->entries),
2572 st->rcv_probes_mcast,
2573 st->rcv_probes_ucast,
2575 st->periodic_gc_runs,
2583 static const struct seq_operations neigh_stat_seq_ops = {
2584 .start = neigh_stat_seq_start,
2585 .next = neigh_stat_seq_next,
2586 .stop = neigh_stat_seq_stop,
2587 .show = neigh_stat_seq_show,
2590 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2592 int ret = seq_open(file, &neigh_stat_seq_ops);
2595 struct seq_file *sf = file->private_data;
2596 sf->private = PDE(inode)->data;
2601 static const struct file_operations neigh_stat_seq_fops = {
2602 .owner = THIS_MODULE,
2603 .open = neigh_stat_seq_open,
2605 .llseek = seq_lseek,
2606 .release = seq_release,
2609 #endif /* CONFIG_PROC_FS */
2611 static inline size_t neigh_nlmsg_size(void)
2613 return NLMSG_ALIGN(sizeof(struct ndmsg))
2614 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2615 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2616 + nla_total_size(sizeof(struct nda_cacheinfo))
2617 + nla_total_size(4); /* NDA_PROBES */
2620 static void __neigh_notify(struct neighbour *n, int type, int flags)
2622 struct net *net = dev_net(n->dev);
2623 struct sk_buff *skb;
2626 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2630 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2632 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2633 WARN_ON(err == -EMSGSIZE);
2637 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2641 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2645 void neigh_app_ns(struct neighbour *n)
2647 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2649 EXPORT_SYMBOL(neigh_app_ns);
2650 #endif /* CONFIG_ARPD */
2652 #ifdef CONFIG_SYSCTL
2654 #define NEIGH_VARS_MAX 19
2656 static struct neigh_sysctl_table {
2657 struct ctl_table_header *sysctl_header;
2658 struct ctl_table neigh_vars[NEIGH_VARS_MAX];
2660 } neigh_sysctl_template __read_mostly = {
2663 .procname = "mcast_solicit",
2664 .maxlen = sizeof(int),
2666 .proc_handler = proc_dointvec,
2669 .procname = "ucast_solicit",
2670 .maxlen = sizeof(int),
2672 .proc_handler = proc_dointvec,
2675 .procname = "app_solicit",
2676 .maxlen = sizeof(int),
2678 .proc_handler = proc_dointvec,
2681 .procname = "retrans_time",
2682 .maxlen = sizeof(int),
2684 .proc_handler = proc_dointvec_userhz_jiffies,
2687 .procname = "base_reachable_time",
2688 .maxlen = sizeof(int),
2690 .proc_handler = proc_dointvec_jiffies,
2693 .procname = "delay_first_probe_time",
2694 .maxlen = sizeof(int),
2696 .proc_handler = proc_dointvec_jiffies,
2699 .procname = "gc_stale_time",
2700 .maxlen = sizeof(int),
2702 .proc_handler = proc_dointvec_jiffies,
2705 .procname = "unres_qlen",
2706 .maxlen = sizeof(int),
2708 .proc_handler = proc_dointvec,
2711 .procname = "proxy_qlen",
2712 .maxlen = sizeof(int),
2714 .proc_handler = proc_dointvec,
2717 .procname = "anycast_delay",
2718 .maxlen = sizeof(int),
2720 .proc_handler = proc_dointvec_userhz_jiffies,
2723 .procname = "proxy_delay",
2724 .maxlen = sizeof(int),
2726 .proc_handler = proc_dointvec_userhz_jiffies,
2729 .procname = "locktime",
2730 .maxlen = sizeof(int),
2732 .proc_handler = proc_dointvec_userhz_jiffies,
2735 .procname = "retrans_time_ms",
2736 .maxlen = sizeof(int),
2738 .proc_handler = proc_dointvec_ms_jiffies,
2741 .procname = "base_reachable_time_ms",
2742 .maxlen = sizeof(int),
2744 .proc_handler = proc_dointvec_ms_jiffies,
2747 .procname = "gc_interval",
2748 .maxlen = sizeof(int),
2750 .proc_handler = proc_dointvec_jiffies,
2753 .procname = "gc_thresh1",
2754 .maxlen = sizeof(int),
2756 .proc_handler = proc_dointvec,
2759 .procname = "gc_thresh2",
2760 .maxlen = sizeof(int),
2762 .proc_handler = proc_dointvec,
2765 .procname = "gc_thresh3",
2766 .maxlen = sizeof(int),
2768 .proc_handler = proc_dointvec,
2774 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2775 char *p_name, proc_handler *handler)
2777 struct neigh_sysctl_table *t;
2778 const char *dev_name_source = NULL;
2780 #define NEIGH_CTL_PATH_ROOT 0
2781 #define NEIGH_CTL_PATH_PROTO 1
2782 #define NEIGH_CTL_PATH_NEIGH 2
2783 #define NEIGH_CTL_PATH_DEV 3
2785 struct ctl_path neigh_path[] = {
2786 { .procname = "net", },
2787 { .procname = "proto", },
2788 { .procname = "neigh", },
2789 { .procname = "default", },
2793 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2797 t->neigh_vars[0].data = &p->mcast_probes;
2798 t->neigh_vars[1].data = &p->ucast_probes;
2799 t->neigh_vars[2].data = &p->app_probes;
2800 t->neigh_vars[3].data = &p->retrans_time;
2801 t->neigh_vars[4].data = &p->base_reachable_time;
2802 t->neigh_vars[5].data = &p->delay_probe_time;
2803 t->neigh_vars[6].data = &p->gc_staletime;
2804 t->neigh_vars[7].data = &p->queue_len;
2805 t->neigh_vars[8].data = &p->proxy_qlen;
2806 t->neigh_vars[9].data = &p->anycast_delay;
2807 t->neigh_vars[10].data = &p->proxy_delay;
2808 t->neigh_vars[11].data = &p->locktime;
2809 t->neigh_vars[12].data = &p->retrans_time;
2810 t->neigh_vars[13].data = &p->base_reachable_time;
2813 dev_name_source = dev->name;
2814 /* Terminate the table early */
2815 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
2817 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2818 t->neigh_vars[14].data = (int *)(p + 1);
2819 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2820 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2821 t->neigh_vars[17].data = (int *)(p + 1) + 3;
2827 t->neigh_vars[3].proc_handler = handler;
2828 t->neigh_vars[3].extra1 = dev;
2830 t->neigh_vars[4].proc_handler = handler;
2831 t->neigh_vars[4].extra1 = dev;
2832 /* RetransTime (in milliseconds)*/
2833 t->neigh_vars[12].proc_handler = handler;
2834 t->neigh_vars[12].extra1 = dev;
2835 /* ReachableTime (in milliseconds) */
2836 t->neigh_vars[13].proc_handler = handler;
2837 t->neigh_vars[13].extra1 = dev;
2840 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2844 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2845 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2848 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2849 if (!t->sysctl_header)
2852 p->sysctl_table = t;
2862 EXPORT_SYMBOL(neigh_sysctl_register);
2864 void neigh_sysctl_unregister(struct neigh_parms *p)
2866 if (p->sysctl_table) {
2867 struct neigh_sysctl_table *t = p->sysctl_table;
2868 p->sysctl_table = NULL;
2869 unregister_sysctl_table(t->sysctl_header);
2874 EXPORT_SYMBOL(neigh_sysctl_unregister);
2876 #endif /* CONFIG_SYSCTL */
2878 static int __init neigh_init(void)
2880 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, NULL);
2881 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, NULL);
2882 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, NULL);
2884 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
2886 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, NULL);
2891 subsys_initcall(neigh_init);