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);
805 nht = rcu_dereference_protected(tbl->nht,
806 lockdep_is_held(&tbl->lock));
808 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
809 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
810 * base_reachable_time.
812 schedule_delayed_work(&tbl->gc_work,
813 tbl->parms.base_reachable_time >> 1);
814 write_unlock_bh(&tbl->lock);
817 static __inline__ int neigh_max_probes(struct neighbour *n)
819 struct neigh_parms *p = n->parms;
820 return (n->nud_state & NUD_PROBE) ?
822 p->ucast_probes + p->app_probes + p->mcast_probes;
825 static void neigh_invalidate(struct neighbour *neigh)
826 __releases(neigh->lock)
827 __acquires(neigh->lock)
831 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
832 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
833 neigh->updated = jiffies;
835 /* It is very thin place. report_unreachable is very complicated
836 routine. Particularly, it can hit the same neighbour entry!
838 So that, we try to be accurate and avoid dead loop. --ANK
840 while (neigh->nud_state == NUD_FAILED &&
841 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
842 write_unlock(&neigh->lock);
843 neigh->ops->error_report(neigh, skb);
844 write_lock(&neigh->lock);
846 skb_queue_purge(&neigh->arp_queue);
849 static void neigh_probe(struct neighbour *neigh)
850 __releases(neigh->lock)
852 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
853 /* keep skb alive even if arp_queue overflows */
855 skb = skb_copy(skb, GFP_ATOMIC);
856 write_unlock(&neigh->lock);
857 neigh->ops->solicit(neigh, skb);
858 atomic_inc(&neigh->probes);
862 /* Called when a timer expires for a neighbour entry. */
864 static void neigh_timer_handler(unsigned long arg)
866 unsigned long now, next;
867 struct neighbour *neigh = (struct neighbour *)arg;
871 write_lock(&neigh->lock);
873 state = neigh->nud_state;
877 if (!(state & NUD_IN_TIMER))
880 if (state & NUD_REACHABLE) {
881 if (time_before_eq(now,
882 neigh->confirmed + neigh->parms->reachable_time)) {
883 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
884 next = neigh->confirmed + neigh->parms->reachable_time;
885 } else if (time_before_eq(now,
886 neigh->used + neigh->parms->delay_probe_time)) {
887 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
888 neigh->nud_state = NUD_DELAY;
889 neigh->updated = jiffies;
890 neigh_suspect(neigh);
891 next = now + neigh->parms->delay_probe_time;
893 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
894 neigh->nud_state = NUD_STALE;
895 neigh->updated = jiffies;
896 neigh_suspect(neigh);
899 } else if (state & NUD_DELAY) {
900 if (time_before_eq(now,
901 neigh->confirmed + neigh->parms->delay_probe_time)) {
902 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
903 neigh->nud_state = NUD_REACHABLE;
904 neigh->updated = jiffies;
905 neigh_connect(neigh);
907 next = neigh->confirmed + neigh->parms->reachable_time;
909 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
910 neigh->nud_state = NUD_PROBE;
911 neigh->updated = jiffies;
912 atomic_set(&neigh->probes, 0);
913 next = now + neigh->parms->retrans_time;
916 /* NUD_PROBE|NUD_INCOMPLETE */
917 next = now + neigh->parms->retrans_time;
920 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
921 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
922 neigh->nud_state = NUD_FAILED;
924 neigh_invalidate(neigh);
927 if (neigh->nud_state & NUD_IN_TIMER) {
928 if (time_before(next, jiffies + HZ/2))
929 next = jiffies + HZ/2;
930 if (!mod_timer(&neigh->timer, next))
933 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
937 write_unlock(&neigh->lock);
941 neigh_update_notify(neigh);
943 neigh_release(neigh);
946 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
949 bool immediate_probe = false;
951 write_lock_bh(&neigh->lock);
954 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
957 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
958 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
959 unsigned long next, now = jiffies;
961 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
962 neigh->nud_state = NUD_INCOMPLETE;
963 neigh->updated = now;
964 next = now + max(neigh->parms->retrans_time, HZ/2);
965 neigh_add_timer(neigh, next);
966 immediate_probe = true;
968 neigh->nud_state = NUD_FAILED;
969 neigh->updated = jiffies;
970 write_unlock_bh(&neigh->lock);
975 } else if (neigh->nud_state & NUD_STALE) {
976 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
977 neigh->nud_state = NUD_DELAY;
978 neigh->updated = jiffies;
979 neigh_add_timer(neigh,
980 jiffies + neigh->parms->delay_probe_time);
983 if (neigh->nud_state == NUD_INCOMPLETE) {
985 if (skb_queue_len(&neigh->arp_queue) >=
986 neigh->parms->queue_len) {
987 struct sk_buff *buff;
988 buff = __skb_dequeue(&neigh->arp_queue);
990 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
993 __skb_queue_tail(&neigh->arp_queue, skb);
1001 write_unlock(&neigh->lock);
1005 EXPORT_SYMBOL(__neigh_event_send);
1007 static void neigh_update_hhs(struct neighbour *neigh)
1009 struct hh_cache *hh;
1010 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1013 if (neigh->dev->header_ops)
1014 update = neigh->dev->header_ops->cache_update;
1019 write_seqlock_bh(&hh->hh_lock);
1020 update(hh, neigh->dev, neigh->ha);
1021 write_sequnlock_bh(&hh->hh_lock);
1028 /* Generic update routine.
1029 -- lladdr is new lladdr or NULL, if it is not supplied.
1030 -- new is new state.
1032 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1034 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1035 lladdr instead of overriding it
1037 It also allows to retain current state
1038 if lladdr is unchanged.
1039 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1041 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1043 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1046 Caller MUST hold reference count on the entry.
1049 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1055 struct net_device *dev;
1056 int update_isrouter = 0;
1058 write_lock_bh(&neigh->lock);
1061 old = neigh->nud_state;
1064 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1065 (old & (NUD_NOARP | NUD_PERMANENT)))
1068 if (!(new & NUD_VALID)) {
1069 neigh_del_timer(neigh);
1070 if (old & NUD_CONNECTED)
1071 neigh_suspect(neigh);
1072 neigh->nud_state = new;
1074 notify = old & NUD_VALID;
1075 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1076 (new & NUD_FAILED)) {
1077 neigh_invalidate(neigh);
1083 /* Compare new lladdr with cached one */
1084 if (!dev->addr_len) {
1085 /* First case: device needs no address. */
1087 } else if (lladdr) {
1088 /* The second case: if something is already cached
1089 and a new address is proposed:
1091 - if they are different, check override flag
1093 if ((old & NUD_VALID) &&
1094 !memcmp(lladdr, neigh->ha, dev->addr_len))
1097 /* No address is supplied; if we know something,
1098 use it, otherwise discard the request.
1101 if (!(old & NUD_VALID))
1106 if (new & NUD_CONNECTED)
1107 neigh->confirmed = jiffies;
1108 neigh->updated = jiffies;
1110 /* If entry was valid and address is not changed,
1111 do not change entry state, if new one is STALE.
1114 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1115 if (old & NUD_VALID) {
1116 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1117 update_isrouter = 0;
1118 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1119 (old & NUD_CONNECTED)) {
1125 if (lladdr == neigh->ha && new == NUD_STALE &&
1126 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1127 (old & NUD_CONNECTED))
1134 neigh_del_timer(neigh);
1135 if (new & NUD_IN_TIMER)
1136 neigh_add_timer(neigh, (jiffies +
1137 ((new & NUD_REACHABLE) ?
1138 neigh->parms->reachable_time :
1140 neigh->nud_state = new;
1143 if (lladdr != neigh->ha) {
1144 write_seqlock(&neigh->ha_lock);
1145 memcpy(&neigh->ha, lladdr, dev->addr_len);
1146 write_sequnlock(&neigh->ha_lock);
1147 neigh_update_hhs(neigh);
1148 if (!(new & NUD_CONNECTED))
1149 neigh->confirmed = jiffies -
1150 (neigh->parms->base_reachable_time << 1);
1155 if (new & NUD_CONNECTED)
1156 neigh_connect(neigh);
1158 neigh_suspect(neigh);
1159 if (!(old & NUD_VALID)) {
1160 struct sk_buff *skb;
1162 /* Again: avoid dead loop if something went wrong */
1164 while (neigh->nud_state & NUD_VALID &&
1165 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1166 struct dst_entry *dst = skb_dst(skb);
1167 struct neighbour *n2, *n1 = neigh;
1168 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);
1177 write_lock_bh(&neigh->lock);
1179 skb_queue_purge(&neigh->arp_queue);
1182 if (update_isrouter) {
1183 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1184 (neigh->flags | NTF_ROUTER) :
1185 (neigh->flags & ~NTF_ROUTER);
1187 write_unlock_bh(&neigh->lock);
1190 neigh_update_notify(neigh);
1194 EXPORT_SYMBOL(neigh_update);
1196 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1197 u8 *lladdr, void *saddr,
1198 struct net_device *dev)
1200 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1201 lladdr || !dev->addr_len);
1203 neigh_update(neigh, lladdr, NUD_STALE,
1204 NEIGH_UPDATE_F_OVERRIDE);
1207 EXPORT_SYMBOL(neigh_event_ns);
1209 /* called with read_lock_bh(&n->lock); */
1210 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst)
1212 struct net_device *dev = dst->dev;
1213 __be16 prot = dst->ops->protocol;
1214 struct hh_cache *hh = &n->hh;
1216 write_lock_bh(&n->lock);
1218 /* Only one thread can come in here and initialize the
1222 dev->header_ops->cache(n, hh, prot);
1224 write_unlock_bh(&n->lock);
1227 /* This function can be used in contexts, where only old dev_queue_xmit
1228 * worked, f.e. if you want to override normal output path (eql, shaper),
1229 * but resolution is not made yet.
1232 int neigh_compat_output(struct neighbour *neigh, struct sk_buff *skb)
1234 struct net_device *dev = skb->dev;
1236 __skb_pull(skb, skb_network_offset(skb));
1238 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1240 dev->header_ops->rebuild(skb))
1243 return dev_queue_xmit(skb);
1245 EXPORT_SYMBOL(neigh_compat_output);
1247 /* Slow and careful. */
1249 int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1251 struct dst_entry *dst = skb_dst(skb);
1257 __skb_pull(skb, skb_network_offset(skb));
1259 if (!neigh_event_send(neigh, skb)) {
1261 struct net_device *dev = neigh->dev;
1264 if (dev->header_ops->cache && !neigh->hh.hh_len)
1265 neigh_hh_init(neigh, dst);
1268 seq = read_seqbegin(&neigh->ha_lock);
1269 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1270 neigh->ha, NULL, skb->len);
1271 } while (read_seqretry(&neigh->ha_lock, seq));
1274 rc = dev_queue_xmit(skb);
1281 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1288 EXPORT_SYMBOL(neigh_resolve_output);
1290 /* As fast as possible without hh cache */
1292 int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1294 struct net_device *dev = neigh->dev;
1298 __skb_pull(skb, skb_network_offset(skb));
1301 seq = read_seqbegin(&neigh->ha_lock);
1302 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1303 neigh->ha, NULL, skb->len);
1304 } while (read_seqretry(&neigh->ha_lock, seq));
1307 err = dev_queue_xmit(skb);
1314 EXPORT_SYMBOL(neigh_connected_output);
1316 int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1318 return dev_queue_xmit(skb);
1320 EXPORT_SYMBOL(neigh_direct_output);
1322 static void neigh_proxy_process(unsigned long arg)
1324 struct neigh_table *tbl = (struct neigh_table *)arg;
1325 long sched_next = 0;
1326 unsigned long now = jiffies;
1327 struct sk_buff *skb, *n;
1329 spin_lock(&tbl->proxy_queue.lock);
1331 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1332 long tdif = NEIGH_CB(skb)->sched_next - now;
1335 struct net_device *dev = skb->dev;
1337 __skb_unlink(skb, &tbl->proxy_queue);
1338 if (tbl->proxy_redo && netif_running(dev)) {
1340 tbl->proxy_redo(skb);
1347 } else if (!sched_next || tdif < sched_next)
1350 del_timer(&tbl->proxy_timer);
1352 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1353 spin_unlock(&tbl->proxy_queue.lock);
1356 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1357 struct sk_buff *skb)
1359 unsigned long now = jiffies;
1360 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1362 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1367 NEIGH_CB(skb)->sched_next = sched_next;
1368 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1370 spin_lock(&tbl->proxy_queue.lock);
1371 if (del_timer(&tbl->proxy_timer)) {
1372 if (time_before(tbl->proxy_timer.expires, sched_next))
1373 sched_next = tbl->proxy_timer.expires;
1377 __skb_queue_tail(&tbl->proxy_queue, skb);
1378 mod_timer(&tbl->proxy_timer, sched_next);
1379 spin_unlock(&tbl->proxy_queue.lock);
1381 EXPORT_SYMBOL(pneigh_enqueue);
1383 static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1384 struct net *net, int ifindex)
1386 struct neigh_parms *p;
1388 for (p = &tbl->parms; p; p = p->next) {
1389 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1390 (!p->dev && !ifindex))
1397 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1398 struct neigh_table *tbl)
1400 struct neigh_parms *p, *ref;
1401 struct net *net = dev_net(dev);
1402 const struct net_device_ops *ops = dev->netdev_ops;
1404 ref = lookup_neigh_parms(tbl, net, 0);
1408 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1411 atomic_set(&p->refcnt, 1);
1413 neigh_rand_reach_time(p->base_reachable_time);
1415 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1422 write_pnet(&p->net, hold_net(net));
1423 p->sysctl_table = NULL;
1424 write_lock_bh(&tbl->lock);
1425 p->next = tbl->parms.next;
1426 tbl->parms.next = p;
1427 write_unlock_bh(&tbl->lock);
1431 EXPORT_SYMBOL(neigh_parms_alloc);
1433 static void neigh_rcu_free_parms(struct rcu_head *head)
1435 struct neigh_parms *parms =
1436 container_of(head, struct neigh_parms, rcu_head);
1438 neigh_parms_put(parms);
1441 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1443 struct neigh_parms **p;
1445 if (!parms || parms == &tbl->parms)
1447 write_lock_bh(&tbl->lock);
1448 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1452 write_unlock_bh(&tbl->lock);
1454 dev_put(parms->dev);
1455 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1459 write_unlock_bh(&tbl->lock);
1460 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1462 EXPORT_SYMBOL(neigh_parms_release);
1464 static void neigh_parms_destroy(struct neigh_parms *parms)
1466 release_net(neigh_parms_net(parms));
1470 static struct lock_class_key neigh_table_proxy_queue_class;
1472 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1474 unsigned long now = jiffies;
1475 unsigned long phsize;
1477 write_pnet(&tbl->parms.net, &init_net);
1478 atomic_set(&tbl->parms.refcnt, 1);
1479 tbl->parms.reachable_time =
1480 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1482 if (!tbl->kmem_cachep)
1484 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1485 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1487 tbl->stats = alloc_percpu(struct neigh_statistics);
1489 panic("cannot create neighbour cache statistics");
1491 #ifdef CONFIG_PROC_FS
1492 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1493 &neigh_stat_seq_fops, tbl))
1494 panic("cannot create neighbour proc dir entry");
1497 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1499 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1500 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1502 if (!tbl->nht || !tbl->phash_buckets)
1503 panic("cannot allocate neighbour cache hashes");
1505 rwlock_init(&tbl->lock);
1506 INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1507 schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1508 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1509 skb_queue_head_init_class(&tbl->proxy_queue,
1510 &neigh_table_proxy_queue_class);
1512 tbl->last_flush = now;
1513 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1515 EXPORT_SYMBOL(neigh_table_init_no_netlink);
1517 void neigh_table_init(struct neigh_table *tbl)
1519 struct neigh_table *tmp;
1521 neigh_table_init_no_netlink(tbl);
1522 write_lock(&neigh_tbl_lock);
1523 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1524 if (tmp->family == tbl->family)
1527 tbl->next = neigh_tables;
1529 write_unlock(&neigh_tbl_lock);
1531 if (unlikely(tmp)) {
1532 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1533 "family %d\n", tbl->family);
1537 EXPORT_SYMBOL(neigh_table_init);
1539 int neigh_table_clear(struct neigh_table *tbl)
1541 struct neigh_table **tp;
1543 /* It is not clean... Fix it to unload IPv6 module safely */
1544 cancel_delayed_work_sync(&tbl->gc_work);
1545 del_timer_sync(&tbl->proxy_timer);
1546 pneigh_queue_purge(&tbl->proxy_queue);
1547 neigh_ifdown(tbl, NULL);
1548 if (atomic_read(&tbl->entries))
1549 printk(KERN_CRIT "neighbour leakage\n");
1550 write_lock(&neigh_tbl_lock);
1551 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1557 write_unlock(&neigh_tbl_lock);
1559 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1560 neigh_hash_free_rcu);
1563 kfree(tbl->phash_buckets);
1564 tbl->phash_buckets = NULL;
1566 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1568 free_percpu(tbl->stats);
1571 kmem_cache_destroy(tbl->kmem_cachep);
1572 tbl->kmem_cachep = NULL;
1576 EXPORT_SYMBOL(neigh_table_clear);
1578 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1580 struct net *net = sock_net(skb->sk);
1582 struct nlattr *dst_attr;
1583 struct neigh_table *tbl;
1584 struct net_device *dev = NULL;
1588 if (nlmsg_len(nlh) < sizeof(*ndm))
1591 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1592 if (dst_attr == NULL)
1595 ndm = nlmsg_data(nlh);
1596 if (ndm->ndm_ifindex) {
1597 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1604 read_lock(&neigh_tbl_lock);
1605 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1606 struct neighbour *neigh;
1608 if (tbl->family != ndm->ndm_family)
1610 read_unlock(&neigh_tbl_lock);
1612 if (nla_len(dst_attr) < tbl->key_len)
1615 if (ndm->ndm_flags & NTF_PROXY) {
1616 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1623 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1624 if (neigh == NULL) {
1629 err = neigh_update(neigh, NULL, NUD_FAILED,
1630 NEIGH_UPDATE_F_OVERRIDE |
1631 NEIGH_UPDATE_F_ADMIN);
1632 neigh_release(neigh);
1635 read_unlock(&neigh_tbl_lock);
1636 err = -EAFNOSUPPORT;
1642 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1644 struct net *net = sock_net(skb->sk);
1646 struct nlattr *tb[NDA_MAX+1];
1647 struct neigh_table *tbl;
1648 struct net_device *dev = NULL;
1652 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1657 if (tb[NDA_DST] == NULL)
1660 ndm = nlmsg_data(nlh);
1661 if (ndm->ndm_ifindex) {
1662 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1668 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1672 read_lock(&neigh_tbl_lock);
1673 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1674 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1675 struct neighbour *neigh;
1678 if (tbl->family != ndm->ndm_family)
1680 read_unlock(&neigh_tbl_lock);
1682 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1684 dst = nla_data(tb[NDA_DST]);
1685 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1687 if (ndm->ndm_flags & NTF_PROXY) {
1688 struct pneigh_entry *pn;
1691 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1693 pn->flags = ndm->ndm_flags;
1702 neigh = neigh_lookup(tbl, dst, dev);
1703 if (neigh == NULL) {
1704 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1709 neigh = __neigh_lookup_errno(tbl, dst, dev);
1710 if (IS_ERR(neigh)) {
1711 err = PTR_ERR(neigh);
1715 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1717 neigh_release(neigh);
1721 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1722 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1725 if (ndm->ndm_flags & NTF_USE) {
1726 neigh_event_send(neigh, NULL);
1729 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1730 neigh_release(neigh);
1734 read_unlock(&neigh_tbl_lock);
1735 err = -EAFNOSUPPORT;
1740 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1742 struct nlattr *nest;
1744 nest = nla_nest_start(skb, NDTA_PARMS);
1749 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1751 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1752 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1753 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1754 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1755 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1756 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1757 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1758 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1759 parms->base_reachable_time);
1760 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1761 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1762 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1763 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1764 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1765 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1767 return nla_nest_end(skb, nest);
1770 nla_nest_cancel(skb, nest);
1774 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1775 u32 pid, u32 seq, int type, int flags)
1777 struct nlmsghdr *nlh;
1778 struct ndtmsg *ndtmsg;
1780 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1784 ndtmsg = nlmsg_data(nlh);
1786 read_lock_bh(&tbl->lock);
1787 ndtmsg->ndtm_family = tbl->family;
1788 ndtmsg->ndtm_pad1 = 0;
1789 ndtmsg->ndtm_pad2 = 0;
1791 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1792 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1793 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1794 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1795 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1798 unsigned long now = jiffies;
1799 unsigned int flush_delta = now - tbl->last_flush;
1800 unsigned int rand_delta = now - tbl->last_rand;
1801 struct neigh_hash_table *nht;
1802 struct ndt_config ndc = {
1803 .ndtc_key_len = tbl->key_len,
1804 .ndtc_entry_size = tbl->entry_size,
1805 .ndtc_entries = atomic_read(&tbl->entries),
1806 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1807 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1808 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1812 nht = rcu_dereference_bh(tbl->nht);
1813 ndc.ndtc_hash_rnd = nht->hash_rnd;
1814 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1815 rcu_read_unlock_bh();
1817 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1822 struct ndt_stats ndst;
1824 memset(&ndst, 0, sizeof(ndst));
1826 for_each_possible_cpu(cpu) {
1827 struct neigh_statistics *st;
1829 st = per_cpu_ptr(tbl->stats, cpu);
1830 ndst.ndts_allocs += st->allocs;
1831 ndst.ndts_destroys += st->destroys;
1832 ndst.ndts_hash_grows += st->hash_grows;
1833 ndst.ndts_res_failed += st->res_failed;
1834 ndst.ndts_lookups += st->lookups;
1835 ndst.ndts_hits += st->hits;
1836 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1837 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1838 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1839 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1842 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1845 BUG_ON(tbl->parms.dev);
1846 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1847 goto nla_put_failure;
1849 read_unlock_bh(&tbl->lock);
1850 return nlmsg_end(skb, nlh);
1853 read_unlock_bh(&tbl->lock);
1854 nlmsg_cancel(skb, nlh);
1858 static int neightbl_fill_param_info(struct sk_buff *skb,
1859 struct neigh_table *tbl,
1860 struct neigh_parms *parms,
1861 u32 pid, u32 seq, int type,
1864 struct ndtmsg *ndtmsg;
1865 struct nlmsghdr *nlh;
1867 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1871 ndtmsg = nlmsg_data(nlh);
1873 read_lock_bh(&tbl->lock);
1874 ndtmsg->ndtm_family = tbl->family;
1875 ndtmsg->ndtm_pad1 = 0;
1876 ndtmsg->ndtm_pad2 = 0;
1878 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1879 neightbl_fill_parms(skb, parms) < 0)
1882 read_unlock_bh(&tbl->lock);
1883 return nlmsg_end(skb, nlh);
1885 read_unlock_bh(&tbl->lock);
1886 nlmsg_cancel(skb, nlh);
1890 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1891 [NDTA_NAME] = { .type = NLA_STRING },
1892 [NDTA_THRESH1] = { .type = NLA_U32 },
1893 [NDTA_THRESH2] = { .type = NLA_U32 },
1894 [NDTA_THRESH3] = { .type = NLA_U32 },
1895 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1896 [NDTA_PARMS] = { .type = NLA_NESTED },
1899 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1900 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1901 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1902 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1903 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1904 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1905 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1906 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1907 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1908 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1909 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1910 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1911 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1912 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1915 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1917 struct net *net = sock_net(skb->sk);
1918 struct neigh_table *tbl;
1919 struct ndtmsg *ndtmsg;
1920 struct nlattr *tb[NDTA_MAX+1];
1923 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1924 nl_neightbl_policy);
1928 if (tb[NDTA_NAME] == NULL) {
1933 ndtmsg = nlmsg_data(nlh);
1934 read_lock(&neigh_tbl_lock);
1935 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1936 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1939 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1949 * We acquire tbl->lock to be nice to the periodic timers and
1950 * make sure they always see a consistent set of values.
1952 write_lock_bh(&tbl->lock);
1954 if (tb[NDTA_PARMS]) {
1955 struct nlattr *tbp[NDTPA_MAX+1];
1956 struct neigh_parms *p;
1959 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1960 nl_ntbl_parm_policy);
1962 goto errout_tbl_lock;
1964 if (tbp[NDTPA_IFINDEX])
1965 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1967 p = lookup_neigh_parms(tbl, net, ifindex);
1970 goto errout_tbl_lock;
1973 for (i = 1; i <= NDTPA_MAX; i++) {
1978 case NDTPA_QUEUE_LEN:
1979 p->queue_len = nla_get_u32(tbp[i]);
1981 case NDTPA_PROXY_QLEN:
1982 p->proxy_qlen = nla_get_u32(tbp[i]);
1984 case NDTPA_APP_PROBES:
1985 p->app_probes = nla_get_u32(tbp[i]);
1987 case NDTPA_UCAST_PROBES:
1988 p->ucast_probes = nla_get_u32(tbp[i]);
1990 case NDTPA_MCAST_PROBES:
1991 p->mcast_probes = nla_get_u32(tbp[i]);
1993 case NDTPA_BASE_REACHABLE_TIME:
1994 p->base_reachable_time = nla_get_msecs(tbp[i]);
1996 case NDTPA_GC_STALETIME:
1997 p->gc_staletime = nla_get_msecs(tbp[i]);
1999 case NDTPA_DELAY_PROBE_TIME:
2000 p->delay_probe_time = nla_get_msecs(tbp[i]);
2002 case NDTPA_RETRANS_TIME:
2003 p->retrans_time = nla_get_msecs(tbp[i]);
2005 case NDTPA_ANYCAST_DELAY:
2006 p->anycast_delay = nla_get_msecs(tbp[i]);
2008 case NDTPA_PROXY_DELAY:
2009 p->proxy_delay = nla_get_msecs(tbp[i]);
2011 case NDTPA_LOCKTIME:
2012 p->locktime = nla_get_msecs(tbp[i]);
2018 if (tb[NDTA_THRESH1])
2019 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2021 if (tb[NDTA_THRESH2])
2022 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2024 if (tb[NDTA_THRESH3])
2025 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2027 if (tb[NDTA_GC_INTERVAL])
2028 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2033 write_unlock_bh(&tbl->lock);
2035 read_unlock(&neigh_tbl_lock);
2040 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2042 struct net *net = sock_net(skb->sk);
2043 int family, tidx, nidx = 0;
2044 int tbl_skip = cb->args[0];
2045 int neigh_skip = cb->args[1];
2046 struct neigh_table *tbl;
2048 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2050 read_lock(&neigh_tbl_lock);
2051 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
2052 struct neigh_parms *p;
2054 if (tidx < tbl_skip || (family && tbl->family != family))
2057 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2058 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2062 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
2063 if (!net_eq(neigh_parms_net(p), net))
2066 if (nidx < neigh_skip)
2069 if (neightbl_fill_param_info(skb, tbl, p,
2070 NETLINK_CB(cb->skb).pid,
2082 read_unlock(&neigh_tbl_lock);
2089 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2090 u32 pid, u32 seq, int type, unsigned int flags)
2092 unsigned long now = jiffies;
2093 struct nda_cacheinfo ci;
2094 struct nlmsghdr *nlh;
2097 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2101 ndm = nlmsg_data(nlh);
2102 ndm->ndm_family = neigh->ops->family;
2105 ndm->ndm_flags = neigh->flags;
2106 ndm->ndm_type = neigh->type;
2107 ndm->ndm_ifindex = neigh->dev->ifindex;
2109 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2111 read_lock_bh(&neigh->lock);
2112 ndm->ndm_state = neigh->nud_state;
2113 if (neigh->nud_state & NUD_VALID) {
2114 char haddr[MAX_ADDR_LEN];
2116 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2117 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2118 read_unlock_bh(&neigh->lock);
2119 goto nla_put_failure;
2123 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2124 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2125 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2126 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2127 read_unlock_bh(&neigh->lock);
2129 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2130 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2132 return nlmsg_end(skb, nlh);
2135 nlmsg_cancel(skb, nlh);
2139 static void neigh_update_notify(struct neighbour *neigh)
2141 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2142 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2145 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2146 struct netlink_callback *cb)
2148 struct net *net = sock_net(skb->sk);
2149 struct neighbour *n;
2150 int rc, h, s_h = cb->args[1];
2151 int idx, s_idx = idx = cb->args[2];
2152 struct neigh_hash_table *nht;
2155 nht = rcu_dereference_bh(tbl->nht);
2157 for (h = 0; h < (1 << nht->hash_shift); h++) {
2162 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2164 n = rcu_dereference_bh(n->next)) {
2165 if (!net_eq(dev_net(n->dev), net))
2169 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2172 NLM_F_MULTI) <= 0) {
2182 rcu_read_unlock_bh();
2188 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2190 struct neigh_table *tbl;
2193 read_lock(&neigh_tbl_lock);
2194 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2197 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2198 if (t < s_t || (family && tbl->family != family))
2201 memset(&cb->args[1], 0, sizeof(cb->args) -
2202 sizeof(cb->args[0]));
2203 if (neigh_dump_table(tbl, skb, cb) < 0)
2206 read_unlock(&neigh_tbl_lock);
2212 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2215 struct neigh_hash_table *nht;
2218 nht = rcu_dereference_bh(tbl->nht);
2220 read_lock(&tbl->lock); /* avoid resizes */
2221 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2222 struct neighbour *n;
2224 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2226 n = rcu_dereference_bh(n->next))
2229 read_unlock(&tbl->lock);
2230 rcu_read_unlock_bh();
2232 EXPORT_SYMBOL(neigh_for_each);
2234 /* The tbl->lock must be held as a writer and BH disabled. */
2235 void __neigh_for_each_release(struct neigh_table *tbl,
2236 int (*cb)(struct neighbour *))
2239 struct neigh_hash_table *nht;
2241 nht = rcu_dereference_protected(tbl->nht,
2242 lockdep_is_held(&tbl->lock));
2243 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2244 struct neighbour *n;
2245 struct neighbour __rcu **np;
2247 np = &nht->hash_buckets[chain];
2248 while ((n = rcu_dereference_protected(*np,
2249 lockdep_is_held(&tbl->lock))) != NULL) {
2252 write_lock(&n->lock);
2255 rcu_assign_pointer(*np,
2256 rcu_dereference_protected(n->next,
2257 lockdep_is_held(&tbl->lock)));
2261 write_unlock(&n->lock);
2263 neigh_cleanup_and_release(n);
2267 EXPORT_SYMBOL(__neigh_for_each_release);
2269 #ifdef CONFIG_PROC_FS
2271 static struct neighbour *neigh_get_first(struct seq_file *seq)
2273 struct neigh_seq_state *state = seq->private;
2274 struct net *net = seq_file_net(seq);
2275 struct neigh_hash_table *nht = state->nht;
2276 struct neighbour *n = NULL;
2277 int bucket = state->bucket;
2279 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2280 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2281 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2284 if (!net_eq(dev_net(n->dev), net))
2286 if (state->neigh_sub_iter) {
2290 v = state->neigh_sub_iter(state, n, &fakep);
2294 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2296 if (n->nud_state & ~NUD_NOARP)
2299 n = rcu_dereference_bh(n->next);
2305 state->bucket = bucket;
2310 static struct neighbour *neigh_get_next(struct seq_file *seq,
2311 struct neighbour *n,
2314 struct neigh_seq_state *state = seq->private;
2315 struct net *net = seq_file_net(seq);
2316 struct neigh_hash_table *nht = state->nht;
2318 if (state->neigh_sub_iter) {
2319 void *v = state->neigh_sub_iter(state, n, pos);
2323 n = rcu_dereference_bh(n->next);
2327 if (!net_eq(dev_net(n->dev), net))
2329 if (state->neigh_sub_iter) {
2330 void *v = state->neigh_sub_iter(state, n, pos);
2335 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2338 if (n->nud_state & ~NUD_NOARP)
2341 n = rcu_dereference_bh(n->next);
2347 if (++state->bucket >= (1 << nht->hash_shift))
2350 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2358 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2360 struct neighbour *n = neigh_get_first(seq);
2365 n = neigh_get_next(seq, n, pos);
2370 return *pos ? NULL : n;
2373 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2375 struct neigh_seq_state *state = seq->private;
2376 struct net *net = seq_file_net(seq);
2377 struct neigh_table *tbl = state->tbl;
2378 struct pneigh_entry *pn = NULL;
2379 int bucket = state->bucket;
2381 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2382 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2383 pn = tbl->phash_buckets[bucket];
2384 while (pn && !net_eq(pneigh_net(pn), net))
2389 state->bucket = bucket;
2394 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2395 struct pneigh_entry *pn,
2398 struct neigh_seq_state *state = seq->private;
2399 struct net *net = seq_file_net(seq);
2400 struct neigh_table *tbl = state->tbl;
2404 } while (pn && !net_eq(pneigh_net(pn), net));
2407 if (++state->bucket > PNEIGH_HASHMASK)
2409 pn = tbl->phash_buckets[state->bucket];
2410 while (pn && !net_eq(pneigh_net(pn), net))
2422 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2424 struct pneigh_entry *pn = pneigh_get_first(seq);
2429 pn = pneigh_get_next(seq, pn, pos);
2434 return *pos ? NULL : pn;
2437 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2439 struct neigh_seq_state *state = seq->private;
2441 loff_t idxpos = *pos;
2443 rc = neigh_get_idx(seq, &idxpos);
2444 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2445 rc = pneigh_get_idx(seq, &idxpos);
2450 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2453 struct neigh_seq_state *state = seq->private;
2457 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2460 state->nht = rcu_dereference_bh(tbl->nht);
2462 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2464 EXPORT_SYMBOL(neigh_seq_start);
2466 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2468 struct neigh_seq_state *state;
2471 if (v == SEQ_START_TOKEN) {
2472 rc = neigh_get_first(seq);
2476 state = seq->private;
2477 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2478 rc = neigh_get_next(seq, v, NULL);
2481 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2482 rc = pneigh_get_first(seq);
2484 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2485 rc = pneigh_get_next(seq, v, NULL);
2491 EXPORT_SYMBOL(neigh_seq_next);
2493 void neigh_seq_stop(struct seq_file *seq, void *v)
2496 rcu_read_unlock_bh();
2498 EXPORT_SYMBOL(neigh_seq_stop);
2500 /* statistics via seq_file */
2502 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2504 struct neigh_table *tbl = seq->private;
2508 return SEQ_START_TOKEN;
2510 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2511 if (!cpu_possible(cpu))
2514 return per_cpu_ptr(tbl->stats, cpu);
2519 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2521 struct neigh_table *tbl = seq->private;
2524 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2525 if (!cpu_possible(cpu))
2528 return per_cpu_ptr(tbl->stats, cpu);
2533 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2538 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2540 struct neigh_table *tbl = seq->private;
2541 struct neigh_statistics *st = v;
2543 if (v == SEQ_START_TOKEN) {
2544 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");
2548 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2549 "%08lx %08lx %08lx %08lx %08lx\n",
2550 atomic_read(&tbl->entries),
2561 st->rcv_probes_mcast,
2562 st->rcv_probes_ucast,
2564 st->periodic_gc_runs,
2572 static const struct seq_operations neigh_stat_seq_ops = {
2573 .start = neigh_stat_seq_start,
2574 .next = neigh_stat_seq_next,
2575 .stop = neigh_stat_seq_stop,
2576 .show = neigh_stat_seq_show,
2579 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2581 int ret = seq_open(file, &neigh_stat_seq_ops);
2584 struct seq_file *sf = file->private_data;
2585 sf->private = PDE(inode)->data;
2590 static const struct file_operations neigh_stat_seq_fops = {
2591 .owner = THIS_MODULE,
2592 .open = neigh_stat_seq_open,
2594 .llseek = seq_lseek,
2595 .release = seq_release,
2598 #endif /* CONFIG_PROC_FS */
2600 static inline size_t neigh_nlmsg_size(void)
2602 return NLMSG_ALIGN(sizeof(struct ndmsg))
2603 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2604 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2605 + nla_total_size(sizeof(struct nda_cacheinfo))
2606 + nla_total_size(4); /* NDA_PROBES */
2609 static void __neigh_notify(struct neighbour *n, int type, int flags)
2611 struct net *net = dev_net(n->dev);
2612 struct sk_buff *skb;
2615 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2619 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2621 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2622 WARN_ON(err == -EMSGSIZE);
2626 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2630 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2634 void neigh_app_ns(struct neighbour *n)
2636 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2638 EXPORT_SYMBOL(neigh_app_ns);
2639 #endif /* CONFIG_ARPD */
2641 #ifdef CONFIG_SYSCTL
2643 #define NEIGH_VARS_MAX 19
2645 static struct neigh_sysctl_table {
2646 struct ctl_table_header *sysctl_header;
2647 struct ctl_table neigh_vars[NEIGH_VARS_MAX];
2649 } neigh_sysctl_template __read_mostly = {
2652 .procname = "mcast_solicit",
2653 .maxlen = sizeof(int),
2655 .proc_handler = proc_dointvec,
2658 .procname = "ucast_solicit",
2659 .maxlen = sizeof(int),
2661 .proc_handler = proc_dointvec,
2664 .procname = "app_solicit",
2665 .maxlen = sizeof(int),
2667 .proc_handler = proc_dointvec,
2670 .procname = "retrans_time",
2671 .maxlen = sizeof(int),
2673 .proc_handler = proc_dointvec_userhz_jiffies,
2676 .procname = "base_reachable_time",
2677 .maxlen = sizeof(int),
2679 .proc_handler = proc_dointvec_jiffies,
2682 .procname = "delay_first_probe_time",
2683 .maxlen = sizeof(int),
2685 .proc_handler = proc_dointvec_jiffies,
2688 .procname = "gc_stale_time",
2689 .maxlen = sizeof(int),
2691 .proc_handler = proc_dointvec_jiffies,
2694 .procname = "unres_qlen",
2695 .maxlen = sizeof(int),
2697 .proc_handler = proc_dointvec,
2700 .procname = "proxy_qlen",
2701 .maxlen = sizeof(int),
2703 .proc_handler = proc_dointvec,
2706 .procname = "anycast_delay",
2707 .maxlen = sizeof(int),
2709 .proc_handler = proc_dointvec_userhz_jiffies,
2712 .procname = "proxy_delay",
2713 .maxlen = sizeof(int),
2715 .proc_handler = proc_dointvec_userhz_jiffies,
2718 .procname = "locktime",
2719 .maxlen = sizeof(int),
2721 .proc_handler = proc_dointvec_userhz_jiffies,
2724 .procname = "retrans_time_ms",
2725 .maxlen = sizeof(int),
2727 .proc_handler = proc_dointvec_ms_jiffies,
2730 .procname = "base_reachable_time_ms",
2731 .maxlen = sizeof(int),
2733 .proc_handler = proc_dointvec_ms_jiffies,
2736 .procname = "gc_interval",
2737 .maxlen = sizeof(int),
2739 .proc_handler = proc_dointvec_jiffies,
2742 .procname = "gc_thresh1",
2743 .maxlen = sizeof(int),
2745 .proc_handler = proc_dointvec,
2748 .procname = "gc_thresh2",
2749 .maxlen = sizeof(int),
2751 .proc_handler = proc_dointvec,
2754 .procname = "gc_thresh3",
2755 .maxlen = sizeof(int),
2757 .proc_handler = proc_dointvec,
2763 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2764 char *p_name, proc_handler *handler)
2766 struct neigh_sysctl_table *t;
2767 const char *dev_name_source = NULL;
2769 #define NEIGH_CTL_PATH_ROOT 0
2770 #define NEIGH_CTL_PATH_PROTO 1
2771 #define NEIGH_CTL_PATH_NEIGH 2
2772 #define NEIGH_CTL_PATH_DEV 3
2774 struct ctl_path neigh_path[] = {
2775 { .procname = "net", },
2776 { .procname = "proto", },
2777 { .procname = "neigh", },
2778 { .procname = "default", },
2782 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2786 t->neigh_vars[0].data = &p->mcast_probes;
2787 t->neigh_vars[1].data = &p->ucast_probes;
2788 t->neigh_vars[2].data = &p->app_probes;
2789 t->neigh_vars[3].data = &p->retrans_time;
2790 t->neigh_vars[4].data = &p->base_reachable_time;
2791 t->neigh_vars[5].data = &p->delay_probe_time;
2792 t->neigh_vars[6].data = &p->gc_staletime;
2793 t->neigh_vars[7].data = &p->queue_len;
2794 t->neigh_vars[8].data = &p->proxy_qlen;
2795 t->neigh_vars[9].data = &p->anycast_delay;
2796 t->neigh_vars[10].data = &p->proxy_delay;
2797 t->neigh_vars[11].data = &p->locktime;
2798 t->neigh_vars[12].data = &p->retrans_time;
2799 t->neigh_vars[13].data = &p->base_reachable_time;
2802 dev_name_source = dev->name;
2803 /* Terminate the table early */
2804 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
2806 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2807 t->neigh_vars[14].data = (int *)(p + 1);
2808 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2809 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2810 t->neigh_vars[17].data = (int *)(p + 1) + 3;
2816 t->neigh_vars[3].proc_handler = handler;
2817 t->neigh_vars[3].extra1 = dev;
2819 t->neigh_vars[4].proc_handler = handler;
2820 t->neigh_vars[4].extra1 = dev;
2821 /* RetransTime (in milliseconds)*/
2822 t->neigh_vars[12].proc_handler = handler;
2823 t->neigh_vars[12].extra1 = dev;
2824 /* ReachableTime (in milliseconds) */
2825 t->neigh_vars[13].proc_handler = handler;
2826 t->neigh_vars[13].extra1 = dev;
2829 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2833 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2834 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2837 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2838 if (!t->sysctl_header)
2841 p->sysctl_table = t;
2851 EXPORT_SYMBOL(neigh_sysctl_register);
2853 void neigh_sysctl_unregister(struct neigh_parms *p)
2855 if (p->sysctl_table) {
2856 struct neigh_sysctl_table *t = p->sysctl_table;
2857 p->sysctl_table = NULL;
2858 unregister_sysctl_table(t->sysctl_header);
2863 EXPORT_SYMBOL(neigh_sysctl_unregister);
2865 #endif /* CONFIG_SYSCTL */
2867 static int __init neigh_init(void)
2869 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, NULL);
2870 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, NULL);
2871 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, NULL);
2873 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
2875 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, NULL);
2880 subsys_initcall(neigh_init);