[CIFS] Fix SMB uid in NTLMSSP authenticate request
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / core / neighbour.c
bloba1cbce7fdae5851f6534f433eb0aa3be3a06e1e6
1 /*
2 * Generic address resolution entity
4 * Authors:
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.
13 * Fixes:
14 * Vitaly E. Lavrov releasing NULL neighbor in neigh_add.
15 * Harald Welte Add neighbour cache statistics like rtstat
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/module.h>
21 #include <linux/socket.h>
22 #include <linux/netdevice.h>
23 #include <linux/proc_fs.h>
24 #ifdef CONFIG_SYSCTL
25 #include <linux/sysctl.h>
26 #endif
27 #include <linux/times.h>
28 #include <net/net_namespace.h>
29 #include <net/neighbour.h>
30 #include <net/dst.h>
31 #include <net/sock.h>
32 #include <net/netevent.h>
33 #include <net/netlink.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/random.h>
36 #include <linux/string.h>
37 #include <linux/log2.h>
39 #define NEIGH_DEBUG 1
41 #define NEIGH_PRINTK(x...) printk(x)
42 #define NEIGH_NOPRINTK(x...) do { ; } while(0)
43 #define NEIGH_PRINTK0 NEIGH_PRINTK
44 #define NEIGH_PRINTK1 NEIGH_NOPRINTK
45 #define NEIGH_PRINTK2 NEIGH_NOPRINTK
47 #if NEIGH_DEBUG >= 1
48 #undef NEIGH_PRINTK1
49 #define NEIGH_PRINTK1 NEIGH_PRINTK
50 #endif
51 #if NEIGH_DEBUG >= 2
52 #undef NEIGH_PRINTK2
53 #define NEIGH_PRINTK2 NEIGH_PRINTK
54 #endif
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;
64 #ifdef CONFIG_PROC_FS
65 static const struct file_operations neigh_stat_seq_fops;
66 #endif
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
75 cache.
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:
87 - timer
88 - resolution queue
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 sk_buff *skb)
103 kfree_skb(skb);
104 return -ENETDOWN;
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)
131 int shrunk = 0;
132 int i;
134 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
136 write_lock_bh(&tbl->lock);
137 for (i = 0; i <= tbl->hash_mask; i++) {
138 struct neighbour *n, **np;
140 np = &tbl->hash_buckets[i];
141 while ((n = *np) != NULL) {
142 /* Neighbour record may be discarded if:
143 * - nobody refers to it.
144 * - it is not permanent
146 write_lock(&n->lock);
147 if (atomic_read(&n->refcnt) == 1 &&
148 !(n->nud_state & NUD_PERMANENT)) {
149 *np = n->next;
150 n->dead = 1;
151 shrunk = 1;
152 write_unlock(&n->lock);
153 neigh_cleanup_and_release(n);
154 continue;
156 write_unlock(&n->lock);
157 np = &n->next;
161 tbl->last_flush = jiffies;
163 write_unlock_bh(&tbl->lock);
165 return shrunk;
168 static void neigh_add_timer(struct neighbour *n, unsigned long when)
170 neigh_hold(n);
171 if (unlikely(mod_timer(&n->timer, when))) {
172 printk("NEIGH: BUG, double timer add, state is %x\n",
173 n->nud_state);
174 dump_stack();
178 static int neigh_del_timer(struct neighbour *n)
180 if ((n->nud_state & NUD_IN_TIMER) &&
181 del_timer(&n->timer)) {
182 neigh_release(n);
183 return 1;
185 return 0;
188 static void pneigh_queue_purge(struct sk_buff_head *list)
190 struct sk_buff *skb;
192 while ((skb = skb_dequeue(list)) != NULL) {
193 dev_put(skb->dev);
194 kfree_skb(skb);
198 static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
200 int i;
202 for (i = 0; i <= tbl->hash_mask; i++) {
203 struct neighbour *n, **np = &tbl->hash_buckets[i];
205 while ((n = *np) != NULL) {
206 if (dev && n->dev != dev) {
207 np = &n->next;
208 continue;
210 *np = n->next;
211 write_lock(&n->lock);
212 neigh_del_timer(n);
213 n->dead = 1;
215 if (atomic_read(&n->refcnt) != 1) {
216 /* The most unpleasant situation.
217 We must destroy neighbour entry,
218 but someone still uses it.
220 The destroy will be delayed until
221 the last user releases us, but
222 we must kill timers etc. and move
223 it to safe state.
225 skb_queue_purge(&n->arp_queue);
226 n->output = neigh_blackhole;
227 if (n->nud_state & NUD_VALID)
228 n->nud_state = NUD_NOARP;
229 else
230 n->nud_state = NUD_NONE;
231 NEIGH_PRINTK2("neigh %p is stray.\n", n);
233 write_unlock(&n->lock);
234 neigh_cleanup_and_release(n);
239 void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
241 write_lock_bh(&tbl->lock);
242 neigh_flush_dev(tbl, dev);
243 write_unlock_bh(&tbl->lock);
245 EXPORT_SYMBOL(neigh_changeaddr);
247 int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
249 write_lock_bh(&tbl->lock);
250 neigh_flush_dev(tbl, dev);
251 pneigh_ifdown(tbl, dev);
252 write_unlock_bh(&tbl->lock);
254 del_timer_sync(&tbl->proxy_timer);
255 pneigh_queue_purge(&tbl->proxy_queue);
256 return 0;
258 EXPORT_SYMBOL(neigh_ifdown);
260 static struct neighbour *neigh_alloc(struct neigh_table *tbl)
262 struct neighbour *n = NULL;
263 unsigned long now = jiffies;
264 int entries;
266 entries = atomic_inc_return(&tbl->entries) - 1;
267 if (entries >= tbl->gc_thresh3 ||
268 (entries >= tbl->gc_thresh2 &&
269 time_after(now, tbl->last_flush + 5 * HZ))) {
270 if (!neigh_forced_gc(tbl) &&
271 entries >= tbl->gc_thresh3)
272 goto out_entries;
275 n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
276 if (!n)
277 goto out_entries;
279 skb_queue_head_init(&n->arp_queue);
280 rwlock_init(&n->lock);
281 n->updated = n->used = now;
282 n->nud_state = NUD_NONE;
283 n->output = neigh_blackhole;
284 n->parms = neigh_parms_clone(&tbl->parms);
285 setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
287 NEIGH_CACHE_STAT_INC(tbl, allocs);
288 n->tbl = tbl;
289 atomic_set(&n->refcnt, 1);
290 n->dead = 1;
291 out:
292 return n;
294 out_entries:
295 atomic_dec(&tbl->entries);
296 goto out;
299 static struct neighbour **neigh_hash_alloc(unsigned int entries)
301 unsigned long size = entries * sizeof(struct neighbour *);
302 struct neighbour **ret;
304 if (size <= PAGE_SIZE) {
305 ret = kzalloc(size, GFP_ATOMIC);
306 } else {
307 ret = (struct neighbour **)
308 __get_free_pages(GFP_ATOMIC|__GFP_ZERO, get_order(size));
310 return ret;
313 static void neigh_hash_free(struct neighbour **hash, unsigned int entries)
315 unsigned long size = entries * sizeof(struct neighbour *);
317 if (size <= PAGE_SIZE)
318 kfree(hash);
319 else
320 free_pages((unsigned long)hash, get_order(size));
323 static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries)
325 struct neighbour **new_hash, **old_hash;
326 unsigned int i, new_hash_mask, old_entries;
328 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
330 BUG_ON(!is_power_of_2(new_entries));
331 new_hash = neigh_hash_alloc(new_entries);
332 if (!new_hash)
333 return;
335 old_entries = tbl->hash_mask + 1;
336 new_hash_mask = new_entries - 1;
337 old_hash = tbl->hash_buckets;
339 get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
340 for (i = 0; i < old_entries; i++) {
341 struct neighbour *n, *next;
343 for (n = old_hash[i]; n; n = next) {
344 unsigned int hash_val = tbl->hash(n->primary_key, n->dev);
346 hash_val &= new_hash_mask;
347 next = n->next;
349 n->next = new_hash[hash_val];
350 new_hash[hash_val] = n;
353 tbl->hash_buckets = new_hash;
354 tbl->hash_mask = new_hash_mask;
356 neigh_hash_free(old_hash, old_entries);
359 struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
360 struct net_device *dev)
362 struct neighbour *n;
363 int key_len = tbl->key_len;
364 u32 hash_val;
366 NEIGH_CACHE_STAT_INC(tbl, lookups);
368 read_lock_bh(&tbl->lock);
369 hash_val = tbl->hash(pkey, dev);
370 for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
371 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
372 neigh_hold(n);
373 NEIGH_CACHE_STAT_INC(tbl, hits);
374 break;
377 read_unlock_bh(&tbl->lock);
378 return n;
380 EXPORT_SYMBOL(neigh_lookup);
382 struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
383 const void *pkey)
385 struct neighbour *n;
386 int key_len = tbl->key_len;
387 u32 hash_val;
389 NEIGH_CACHE_STAT_INC(tbl, lookups);
391 read_lock_bh(&tbl->lock);
392 hash_val = tbl->hash(pkey, NULL);
393 for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
394 if (!memcmp(n->primary_key, pkey, key_len) &&
395 net_eq(dev_net(n->dev), net)) {
396 neigh_hold(n);
397 NEIGH_CACHE_STAT_INC(tbl, hits);
398 break;
401 read_unlock_bh(&tbl->lock);
402 return n;
404 EXPORT_SYMBOL(neigh_lookup_nodev);
406 struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
407 struct net_device *dev)
409 u32 hash_val;
410 int key_len = tbl->key_len;
411 int error;
412 struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
414 if (!n) {
415 rc = ERR_PTR(-ENOBUFS);
416 goto out;
419 memcpy(n->primary_key, pkey, key_len);
420 n->dev = dev;
421 dev_hold(dev);
423 /* Protocol specific setup. */
424 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
425 rc = ERR_PTR(error);
426 goto out_neigh_release;
429 /* Device specific setup. */
430 if (n->parms->neigh_setup &&
431 (error = n->parms->neigh_setup(n)) < 0) {
432 rc = ERR_PTR(error);
433 goto out_neigh_release;
436 n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
438 write_lock_bh(&tbl->lock);
440 if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1))
441 neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1);
443 hash_val = tbl->hash(pkey, dev) & tbl->hash_mask;
445 if (n->parms->dead) {
446 rc = ERR_PTR(-EINVAL);
447 goto out_tbl_unlock;
450 for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) {
451 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
452 neigh_hold(n1);
453 rc = n1;
454 goto out_tbl_unlock;
458 n->next = tbl->hash_buckets[hash_val];
459 tbl->hash_buckets[hash_val] = n;
460 n->dead = 0;
461 neigh_hold(n);
462 write_unlock_bh(&tbl->lock);
463 NEIGH_PRINTK2("neigh %p is created.\n", n);
464 rc = n;
465 out:
466 return rc;
467 out_tbl_unlock:
468 write_unlock_bh(&tbl->lock);
469 out_neigh_release:
470 neigh_release(n);
471 goto out;
473 EXPORT_SYMBOL(neigh_create);
475 static u32 pneigh_hash(const void *pkey, int key_len)
477 u32 hash_val = *(u32 *)(pkey + key_len - 4);
478 hash_val ^= (hash_val >> 16);
479 hash_val ^= hash_val >> 8;
480 hash_val ^= hash_val >> 4;
481 hash_val &= PNEIGH_HASHMASK;
482 return hash_val;
485 static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
486 struct net *net,
487 const void *pkey,
488 int key_len,
489 struct net_device *dev)
491 while (n) {
492 if (!memcmp(n->key, pkey, key_len) &&
493 net_eq(pneigh_net(n), net) &&
494 (n->dev == dev || !n->dev))
495 return n;
496 n = n->next;
498 return NULL;
501 struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
502 struct net *net, const void *pkey, struct net_device *dev)
504 int key_len = tbl->key_len;
505 u32 hash_val = pneigh_hash(pkey, key_len);
507 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
508 net, pkey, key_len, dev);
510 EXPORT_SYMBOL_GPL(__pneigh_lookup);
512 struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
513 struct net *net, const void *pkey,
514 struct net_device *dev, int creat)
516 struct pneigh_entry *n;
517 int key_len = tbl->key_len;
518 u32 hash_val = pneigh_hash(pkey, key_len);
520 read_lock_bh(&tbl->lock);
521 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
522 net, pkey, key_len, dev);
523 read_unlock_bh(&tbl->lock);
525 if (n || !creat)
526 goto out;
528 ASSERT_RTNL();
530 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
531 if (!n)
532 goto out;
534 write_pnet(&n->net, hold_net(net));
535 memcpy(n->key, pkey, key_len);
536 n->dev = dev;
537 if (dev)
538 dev_hold(dev);
540 if (tbl->pconstructor && tbl->pconstructor(n)) {
541 if (dev)
542 dev_put(dev);
543 release_net(net);
544 kfree(n);
545 n = NULL;
546 goto out;
549 write_lock_bh(&tbl->lock);
550 n->next = tbl->phash_buckets[hash_val];
551 tbl->phash_buckets[hash_val] = n;
552 write_unlock_bh(&tbl->lock);
553 out:
554 return n;
556 EXPORT_SYMBOL(pneigh_lookup);
559 int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
560 struct net_device *dev)
562 struct pneigh_entry *n, **np;
563 int key_len = tbl->key_len;
564 u32 hash_val = pneigh_hash(pkey, key_len);
566 write_lock_bh(&tbl->lock);
567 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
568 np = &n->next) {
569 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
570 net_eq(pneigh_net(n), net)) {
571 *np = n->next;
572 write_unlock_bh(&tbl->lock);
573 if (tbl->pdestructor)
574 tbl->pdestructor(n);
575 if (n->dev)
576 dev_put(n->dev);
577 release_net(pneigh_net(n));
578 kfree(n);
579 return 0;
582 write_unlock_bh(&tbl->lock);
583 return -ENOENT;
586 static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
588 struct pneigh_entry *n, **np;
589 u32 h;
591 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
592 np = &tbl->phash_buckets[h];
593 while ((n = *np) != NULL) {
594 if (!dev || n->dev == dev) {
595 *np = n->next;
596 if (tbl->pdestructor)
597 tbl->pdestructor(n);
598 if (n->dev)
599 dev_put(n->dev);
600 release_net(pneigh_net(n));
601 kfree(n);
602 continue;
604 np = &n->next;
607 return -ENOENT;
610 static void neigh_parms_destroy(struct neigh_parms *parms);
612 static inline void neigh_parms_put(struct neigh_parms *parms)
614 if (atomic_dec_and_test(&parms->refcnt))
615 neigh_parms_destroy(parms);
619 * neighbour must already be out of the table;
622 void neigh_destroy(struct neighbour *neigh)
624 struct hh_cache *hh;
626 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
628 if (!neigh->dead) {
629 printk(KERN_WARNING
630 "Destroying alive neighbour %p\n", neigh);
631 dump_stack();
632 return;
635 if (neigh_del_timer(neigh))
636 printk(KERN_WARNING "Impossible event.\n");
638 while ((hh = neigh->hh) != NULL) {
639 neigh->hh = hh->hh_next;
640 hh->hh_next = NULL;
642 write_seqlock_bh(&hh->hh_lock);
643 hh->hh_output = neigh_blackhole;
644 write_sequnlock_bh(&hh->hh_lock);
645 if (atomic_dec_and_test(&hh->hh_refcnt))
646 kfree(hh);
649 skb_queue_purge(&neigh->arp_queue);
651 dev_put(neigh->dev);
652 neigh_parms_put(neigh->parms);
654 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
656 atomic_dec(&neigh->tbl->entries);
657 kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
659 EXPORT_SYMBOL(neigh_destroy);
661 /* Neighbour state is suspicious;
662 disable fast path.
664 Called with write_locked neigh.
666 static void neigh_suspect(struct neighbour *neigh)
668 struct hh_cache *hh;
670 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
672 neigh->output = neigh->ops->output;
674 for (hh = neigh->hh; hh; hh = hh->hh_next)
675 hh->hh_output = neigh->ops->output;
678 /* Neighbour state is OK;
679 enable fast path.
681 Called with write_locked neigh.
683 static void neigh_connect(struct neighbour *neigh)
685 struct hh_cache *hh;
687 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
689 neigh->output = neigh->ops->connected_output;
691 for (hh = neigh->hh; hh; hh = hh->hh_next)
692 hh->hh_output = neigh->ops->hh_output;
695 static void neigh_periodic_timer(unsigned long arg)
697 struct neigh_table *tbl = (struct neigh_table *)arg;
698 struct neighbour *n, **np;
699 unsigned long expire, now = jiffies;
701 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
703 write_lock(&tbl->lock);
706 * periodically recompute ReachableTime from random function
709 if (time_after(now, tbl->last_rand + 300 * HZ)) {
710 struct neigh_parms *p;
711 tbl->last_rand = now;
712 for (p = &tbl->parms; p; p = p->next)
713 p->reachable_time =
714 neigh_rand_reach_time(p->base_reachable_time);
717 np = &tbl->hash_buckets[tbl->hash_chain_gc];
718 tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask);
720 while ((n = *np) != NULL) {
721 unsigned int state;
723 write_lock(&n->lock);
725 state = n->nud_state;
726 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
727 write_unlock(&n->lock);
728 goto next_elt;
731 if (time_before(n->used, n->confirmed))
732 n->used = n->confirmed;
734 if (atomic_read(&n->refcnt) == 1 &&
735 (state == NUD_FAILED ||
736 time_after(now, n->used + n->parms->gc_staletime))) {
737 *np = n->next;
738 n->dead = 1;
739 write_unlock(&n->lock);
740 neigh_cleanup_and_release(n);
741 continue;
743 write_unlock(&n->lock);
745 next_elt:
746 np = &n->next;
749 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
750 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
751 * base_reachable_time.
753 expire = tbl->parms.base_reachable_time >> 1;
754 expire /= (tbl->hash_mask + 1);
755 if (!expire)
756 expire = 1;
758 if (expire>HZ)
759 mod_timer(&tbl->gc_timer, round_jiffies(now + expire));
760 else
761 mod_timer(&tbl->gc_timer, now + expire);
763 write_unlock(&tbl->lock);
766 static __inline__ int neigh_max_probes(struct neighbour *n)
768 struct neigh_parms *p = n->parms;
769 return (n->nud_state & NUD_PROBE ?
770 p->ucast_probes :
771 p->ucast_probes + p->app_probes + p->mcast_probes);
774 /* Called when a timer expires for a neighbour entry. */
776 static void neigh_timer_handler(unsigned long arg)
778 unsigned long now, next;
779 struct neighbour *neigh = (struct neighbour *)arg;
780 unsigned state;
781 int notify = 0;
783 write_lock(&neigh->lock);
785 state = neigh->nud_state;
786 now = jiffies;
787 next = now + HZ;
789 if (!(state & NUD_IN_TIMER)) {
790 #ifndef CONFIG_SMP
791 printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
792 #endif
793 goto out;
796 if (state & NUD_REACHABLE) {
797 if (time_before_eq(now,
798 neigh->confirmed + neigh->parms->reachable_time)) {
799 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
800 next = neigh->confirmed + neigh->parms->reachable_time;
801 } else if (time_before_eq(now,
802 neigh->used + neigh->parms->delay_probe_time)) {
803 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
804 neigh->nud_state = NUD_DELAY;
805 neigh->updated = jiffies;
806 neigh_suspect(neigh);
807 next = now + neigh->parms->delay_probe_time;
808 } else {
809 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
810 neigh->nud_state = NUD_STALE;
811 neigh->updated = jiffies;
812 neigh_suspect(neigh);
813 notify = 1;
815 } else if (state & NUD_DELAY) {
816 if (time_before_eq(now,
817 neigh->confirmed + neigh->parms->delay_probe_time)) {
818 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
819 neigh->nud_state = NUD_REACHABLE;
820 neigh->updated = jiffies;
821 neigh_connect(neigh);
822 notify = 1;
823 next = neigh->confirmed + neigh->parms->reachable_time;
824 } else {
825 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
826 neigh->nud_state = NUD_PROBE;
827 neigh->updated = jiffies;
828 atomic_set(&neigh->probes, 0);
829 next = now + neigh->parms->retrans_time;
831 } else {
832 /* NUD_PROBE|NUD_INCOMPLETE */
833 next = now + neigh->parms->retrans_time;
836 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
837 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
838 struct sk_buff *skb;
840 neigh->nud_state = NUD_FAILED;
841 neigh->updated = jiffies;
842 notify = 1;
843 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
844 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
846 /* It is very thin place. report_unreachable is very complicated
847 routine. Particularly, it can hit the same neighbour entry!
849 So that, we try to be accurate and avoid dead loop. --ANK
851 while (neigh->nud_state == NUD_FAILED &&
852 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
853 write_unlock(&neigh->lock);
854 neigh->ops->error_report(neigh, skb);
855 write_lock(&neigh->lock);
857 skb_queue_purge(&neigh->arp_queue);
860 if (neigh->nud_state & NUD_IN_TIMER) {
861 if (time_before(next, jiffies + HZ/2))
862 next = jiffies + HZ/2;
863 if (!mod_timer(&neigh->timer, next))
864 neigh_hold(neigh);
866 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
867 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
868 /* keep skb alive even if arp_queue overflows */
869 if (skb)
870 skb = skb_copy(skb, GFP_ATOMIC);
871 write_unlock(&neigh->lock);
872 neigh->ops->solicit(neigh, skb);
873 atomic_inc(&neigh->probes);
874 kfree_skb(skb);
875 } else {
876 out:
877 write_unlock(&neigh->lock);
880 if (notify)
881 neigh_update_notify(neigh);
883 neigh_release(neigh);
886 int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
888 int rc;
889 unsigned long now;
891 write_lock_bh(&neigh->lock);
893 rc = 0;
894 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
895 goto out_unlock_bh;
897 now = jiffies;
899 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
900 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
901 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
902 neigh->nud_state = NUD_INCOMPLETE;
903 neigh->updated = jiffies;
904 neigh_add_timer(neigh, now + 1);
905 } else {
906 neigh->nud_state = NUD_FAILED;
907 neigh->updated = jiffies;
908 write_unlock_bh(&neigh->lock);
910 kfree_skb(skb);
911 return 1;
913 } else if (neigh->nud_state & NUD_STALE) {
914 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
915 neigh->nud_state = NUD_DELAY;
916 neigh->updated = jiffies;
917 neigh_add_timer(neigh,
918 jiffies + neigh->parms->delay_probe_time);
921 if (neigh->nud_state == NUD_INCOMPLETE) {
922 if (skb) {
923 if (skb_queue_len(&neigh->arp_queue) >=
924 neigh->parms->queue_len) {
925 struct sk_buff *buff;
926 buff = __skb_dequeue(&neigh->arp_queue);
927 kfree_skb(buff);
928 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
930 __skb_queue_tail(&neigh->arp_queue, skb);
932 rc = 1;
934 out_unlock_bh:
935 write_unlock_bh(&neigh->lock);
936 return rc;
938 EXPORT_SYMBOL(__neigh_event_send);
940 static void neigh_update_hhs(struct neighbour *neigh)
942 struct hh_cache *hh;
943 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
944 = neigh->dev->header_ops->cache_update;
946 if (update) {
947 for (hh = neigh->hh; hh; hh = hh->hh_next) {
948 write_seqlock_bh(&hh->hh_lock);
949 update(hh, neigh->dev, neigh->ha);
950 write_sequnlock_bh(&hh->hh_lock);
957 /* Generic update routine.
958 -- lladdr is new lladdr or NULL, if it is not supplied.
959 -- new is new state.
960 -- flags
961 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
962 if it is different.
963 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
964 lladdr instead of overriding it
965 if it is different.
966 It also allows to retain current state
967 if lladdr is unchanged.
968 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
970 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
971 NTF_ROUTER flag.
972 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
973 a router.
975 Caller MUST hold reference count on the entry.
978 int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
979 u32 flags)
981 u8 old;
982 int err;
983 int notify = 0;
984 struct net_device *dev;
985 int update_isrouter = 0;
987 write_lock_bh(&neigh->lock);
989 dev = neigh->dev;
990 old = neigh->nud_state;
991 err = -EPERM;
993 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
994 (old & (NUD_NOARP | NUD_PERMANENT)))
995 goto out;
997 if (!(new & NUD_VALID)) {
998 neigh_del_timer(neigh);
999 if (old & NUD_CONNECTED)
1000 neigh_suspect(neigh);
1001 neigh->nud_state = new;
1002 err = 0;
1003 notify = old & NUD_VALID;
1004 goto out;
1007 /* Compare new lladdr with cached one */
1008 if (!dev->addr_len) {
1009 /* First case: device needs no address. */
1010 lladdr = neigh->ha;
1011 } else if (lladdr) {
1012 /* The second case: if something is already cached
1013 and a new address is proposed:
1014 - compare new & old
1015 - if they are different, check override flag
1017 if ((old & NUD_VALID) &&
1018 !memcmp(lladdr, neigh->ha, dev->addr_len))
1019 lladdr = neigh->ha;
1020 } else {
1021 /* No address is supplied; if we know something,
1022 use it, otherwise discard the request.
1024 err = -EINVAL;
1025 if (!(old & NUD_VALID))
1026 goto out;
1027 lladdr = neigh->ha;
1030 if (new & NUD_CONNECTED)
1031 neigh->confirmed = jiffies;
1032 neigh->updated = jiffies;
1034 /* If entry was valid and address is not changed,
1035 do not change entry state, if new one is STALE.
1037 err = 0;
1038 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1039 if (old & NUD_VALID) {
1040 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1041 update_isrouter = 0;
1042 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1043 (old & NUD_CONNECTED)) {
1044 lladdr = neigh->ha;
1045 new = NUD_STALE;
1046 } else
1047 goto out;
1048 } else {
1049 if (lladdr == neigh->ha && new == NUD_STALE &&
1050 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1051 (old & NUD_CONNECTED))
1053 new = old;
1057 if (new != old) {
1058 neigh_del_timer(neigh);
1059 if (new & NUD_IN_TIMER)
1060 neigh_add_timer(neigh, (jiffies +
1061 ((new & NUD_REACHABLE) ?
1062 neigh->parms->reachable_time :
1063 0)));
1064 neigh->nud_state = new;
1067 if (lladdr != neigh->ha) {
1068 memcpy(&neigh->ha, lladdr, dev->addr_len);
1069 neigh_update_hhs(neigh);
1070 if (!(new & NUD_CONNECTED))
1071 neigh->confirmed = jiffies -
1072 (neigh->parms->base_reachable_time << 1);
1073 notify = 1;
1075 if (new == old)
1076 goto out;
1077 if (new & NUD_CONNECTED)
1078 neigh_connect(neigh);
1079 else
1080 neigh_suspect(neigh);
1081 if (!(old & NUD_VALID)) {
1082 struct sk_buff *skb;
1084 /* Again: avoid dead loop if something went wrong */
1086 while (neigh->nud_state & NUD_VALID &&
1087 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1088 struct neighbour *n1 = neigh;
1089 write_unlock_bh(&neigh->lock);
1090 /* On shaper/eql skb->dst->neighbour != neigh :( */
1091 if (skb->dst && skb->dst->neighbour)
1092 n1 = skb->dst->neighbour;
1093 n1->output(skb);
1094 write_lock_bh(&neigh->lock);
1096 skb_queue_purge(&neigh->arp_queue);
1098 out:
1099 if (update_isrouter) {
1100 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1101 (neigh->flags | NTF_ROUTER) :
1102 (neigh->flags & ~NTF_ROUTER);
1104 write_unlock_bh(&neigh->lock);
1106 if (notify)
1107 neigh_update_notify(neigh);
1109 return err;
1111 EXPORT_SYMBOL(neigh_update);
1113 struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1114 u8 *lladdr, void *saddr,
1115 struct net_device *dev)
1117 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1118 lladdr || !dev->addr_len);
1119 if (neigh)
1120 neigh_update(neigh, lladdr, NUD_STALE,
1121 NEIGH_UPDATE_F_OVERRIDE);
1122 return neigh;
1124 EXPORT_SYMBOL(neigh_event_ns);
1126 static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst,
1127 __be16 protocol)
1129 struct hh_cache *hh;
1130 struct net_device *dev = dst->dev;
1132 for (hh = n->hh; hh; hh = hh->hh_next)
1133 if (hh->hh_type == protocol)
1134 break;
1136 if (!hh && (hh = kzalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) {
1137 seqlock_init(&hh->hh_lock);
1138 hh->hh_type = protocol;
1139 atomic_set(&hh->hh_refcnt, 0);
1140 hh->hh_next = NULL;
1142 if (dev->header_ops->cache(n, hh)) {
1143 kfree(hh);
1144 hh = NULL;
1145 } else {
1146 atomic_inc(&hh->hh_refcnt);
1147 hh->hh_next = n->hh;
1148 n->hh = hh;
1149 if (n->nud_state & NUD_CONNECTED)
1150 hh->hh_output = n->ops->hh_output;
1151 else
1152 hh->hh_output = n->ops->output;
1155 if (hh) {
1156 atomic_inc(&hh->hh_refcnt);
1157 dst->hh = hh;
1161 /* This function can be used in contexts, where only old dev_queue_xmit
1162 worked, f.e. if you want to override normal output path (eql, shaper),
1163 but resolution is not made yet.
1166 int neigh_compat_output(struct sk_buff *skb)
1168 struct net_device *dev = skb->dev;
1170 __skb_pull(skb, skb_network_offset(skb));
1172 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1173 skb->len) < 0 &&
1174 dev->header_ops->rebuild(skb))
1175 return 0;
1177 return dev_queue_xmit(skb);
1179 EXPORT_SYMBOL(neigh_compat_output);
1181 /* Slow and careful. */
1183 int neigh_resolve_output(struct sk_buff *skb)
1185 struct dst_entry *dst = skb->dst;
1186 struct neighbour *neigh;
1187 int rc = 0;
1189 if (!dst || !(neigh = dst->neighbour))
1190 goto discard;
1192 __skb_pull(skb, skb_network_offset(skb));
1194 if (!neigh_event_send(neigh, skb)) {
1195 int err;
1196 struct net_device *dev = neigh->dev;
1197 if (dev->header_ops->cache && !dst->hh) {
1198 write_lock_bh(&neigh->lock);
1199 if (!dst->hh)
1200 neigh_hh_init(neigh, dst, dst->ops->protocol);
1201 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1202 neigh->ha, NULL, skb->len);
1203 write_unlock_bh(&neigh->lock);
1204 } else {
1205 read_lock_bh(&neigh->lock);
1206 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1207 neigh->ha, NULL, skb->len);
1208 read_unlock_bh(&neigh->lock);
1210 if (err >= 0)
1211 rc = neigh->ops->queue_xmit(skb);
1212 else
1213 goto out_kfree_skb;
1215 out:
1216 return rc;
1217 discard:
1218 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1219 dst, dst ? dst->neighbour : NULL);
1220 out_kfree_skb:
1221 rc = -EINVAL;
1222 kfree_skb(skb);
1223 goto out;
1225 EXPORT_SYMBOL(neigh_resolve_output);
1227 /* As fast as possible without hh cache */
1229 int neigh_connected_output(struct sk_buff *skb)
1231 int err;
1232 struct dst_entry *dst = skb->dst;
1233 struct neighbour *neigh = dst->neighbour;
1234 struct net_device *dev = neigh->dev;
1236 __skb_pull(skb, skb_network_offset(skb));
1238 read_lock_bh(&neigh->lock);
1239 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1240 neigh->ha, NULL, skb->len);
1241 read_unlock_bh(&neigh->lock);
1242 if (err >= 0)
1243 err = neigh->ops->queue_xmit(skb);
1244 else {
1245 err = -EINVAL;
1246 kfree_skb(skb);
1248 return err;
1250 EXPORT_SYMBOL(neigh_connected_output);
1252 static void neigh_proxy_process(unsigned long arg)
1254 struct neigh_table *tbl = (struct neigh_table *)arg;
1255 long sched_next = 0;
1256 unsigned long now = jiffies;
1257 struct sk_buff *skb, *n;
1259 spin_lock(&tbl->proxy_queue.lock);
1261 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1262 long tdif = NEIGH_CB(skb)->sched_next - now;
1264 if (tdif <= 0) {
1265 struct net_device *dev = skb->dev;
1266 __skb_unlink(skb, &tbl->proxy_queue);
1267 if (tbl->proxy_redo && netif_running(dev))
1268 tbl->proxy_redo(skb);
1269 else
1270 kfree_skb(skb);
1272 dev_put(dev);
1273 } else if (!sched_next || tdif < sched_next)
1274 sched_next = tdif;
1276 del_timer(&tbl->proxy_timer);
1277 if (sched_next)
1278 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1279 spin_unlock(&tbl->proxy_queue.lock);
1282 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1283 struct sk_buff *skb)
1285 unsigned long now = jiffies;
1286 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1288 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1289 kfree_skb(skb);
1290 return;
1293 NEIGH_CB(skb)->sched_next = sched_next;
1294 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1296 spin_lock(&tbl->proxy_queue.lock);
1297 if (del_timer(&tbl->proxy_timer)) {
1298 if (time_before(tbl->proxy_timer.expires, sched_next))
1299 sched_next = tbl->proxy_timer.expires;
1301 dst_release(skb->dst);
1302 skb->dst = NULL;
1303 dev_hold(skb->dev);
1304 __skb_queue_tail(&tbl->proxy_queue, skb);
1305 mod_timer(&tbl->proxy_timer, sched_next);
1306 spin_unlock(&tbl->proxy_queue.lock);
1308 EXPORT_SYMBOL(pneigh_enqueue);
1310 static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl,
1311 struct net *net, int ifindex)
1313 struct neigh_parms *p;
1315 for (p = &tbl->parms; p; p = p->next) {
1316 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1317 (!p->dev && !ifindex))
1318 return p;
1321 return NULL;
1324 struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1325 struct neigh_table *tbl)
1327 struct neigh_parms *p, *ref;
1328 struct net *net = dev_net(dev);
1329 const struct net_device_ops *ops = dev->netdev_ops;
1331 ref = lookup_neigh_params(tbl, net, 0);
1332 if (!ref)
1333 return NULL;
1335 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1336 if (p) {
1337 p->tbl = tbl;
1338 atomic_set(&p->refcnt, 1);
1339 p->reachable_time =
1340 neigh_rand_reach_time(p->base_reachable_time);
1342 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1343 kfree(p);
1344 return NULL;
1347 dev_hold(dev);
1348 p->dev = dev;
1349 write_pnet(&p->net, hold_net(net));
1350 p->sysctl_table = NULL;
1351 write_lock_bh(&tbl->lock);
1352 p->next = tbl->parms.next;
1353 tbl->parms.next = p;
1354 write_unlock_bh(&tbl->lock);
1356 return p;
1358 EXPORT_SYMBOL(neigh_parms_alloc);
1360 static void neigh_rcu_free_parms(struct rcu_head *head)
1362 struct neigh_parms *parms =
1363 container_of(head, struct neigh_parms, rcu_head);
1365 neigh_parms_put(parms);
1368 void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1370 struct neigh_parms **p;
1372 if (!parms || parms == &tbl->parms)
1373 return;
1374 write_lock_bh(&tbl->lock);
1375 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1376 if (*p == parms) {
1377 *p = parms->next;
1378 parms->dead = 1;
1379 write_unlock_bh(&tbl->lock);
1380 if (parms->dev)
1381 dev_put(parms->dev);
1382 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1383 return;
1386 write_unlock_bh(&tbl->lock);
1387 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1389 EXPORT_SYMBOL(neigh_parms_release);
1391 static void neigh_parms_destroy(struct neigh_parms *parms)
1393 release_net(neigh_parms_net(parms));
1394 kfree(parms);
1397 static struct lock_class_key neigh_table_proxy_queue_class;
1399 void neigh_table_init_no_netlink(struct neigh_table *tbl)
1401 unsigned long now = jiffies;
1402 unsigned long phsize;
1404 write_pnet(&tbl->parms.net, &init_net);
1405 atomic_set(&tbl->parms.refcnt, 1);
1406 tbl->parms.reachable_time =
1407 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1409 if (!tbl->kmem_cachep)
1410 tbl->kmem_cachep =
1411 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1412 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1413 NULL);
1414 tbl->stats = alloc_percpu(struct neigh_statistics);
1415 if (!tbl->stats)
1416 panic("cannot create neighbour cache statistics");
1418 #ifdef CONFIG_PROC_FS
1419 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1420 &neigh_stat_seq_fops, tbl))
1421 panic("cannot create neighbour proc dir entry");
1422 #endif
1424 tbl->hash_mask = 1;
1425 tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1);
1427 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1428 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1430 if (!tbl->hash_buckets || !tbl->phash_buckets)
1431 panic("cannot allocate neighbour cache hashes");
1433 get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
1435 rwlock_init(&tbl->lock);
1436 setup_timer(&tbl->gc_timer, neigh_periodic_timer, (unsigned long)tbl);
1437 tbl->gc_timer.expires = now + 1;
1438 add_timer(&tbl->gc_timer);
1440 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1441 skb_queue_head_init_class(&tbl->proxy_queue,
1442 &neigh_table_proxy_queue_class);
1444 tbl->last_flush = now;
1445 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1447 EXPORT_SYMBOL(neigh_table_init_no_netlink);
1449 void neigh_table_init(struct neigh_table *tbl)
1451 struct neigh_table *tmp;
1453 neigh_table_init_no_netlink(tbl);
1454 write_lock(&neigh_tbl_lock);
1455 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1456 if (tmp->family == tbl->family)
1457 break;
1459 tbl->next = neigh_tables;
1460 neigh_tables = tbl;
1461 write_unlock(&neigh_tbl_lock);
1463 if (unlikely(tmp)) {
1464 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1465 "family %d\n", tbl->family);
1466 dump_stack();
1469 EXPORT_SYMBOL(neigh_table_init);
1471 int neigh_table_clear(struct neigh_table *tbl)
1473 struct neigh_table **tp;
1475 /* It is not clean... Fix it to unload IPv6 module safely */
1476 del_timer_sync(&tbl->gc_timer);
1477 del_timer_sync(&tbl->proxy_timer);
1478 pneigh_queue_purge(&tbl->proxy_queue);
1479 neigh_ifdown(tbl, NULL);
1480 if (atomic_read(&tbl->entries))
1481 printk(KERN_CRIT "neighbour leakage\n");
1482 write_lock(&neigh_tbl_lock);
1483 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1484 if (*tp == tbl) {
1485 *tp = tbl->next;
1486 break;
1489 write_unlock(&neigh_tbl_lock);
1491 neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1);
1492 tbl->hash_buckets = NULL;
1494 kfree(tbl->phash_buckets);
1495 tbl->phash_buckets = NULL;
1497 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1499 free_percpu(tbl->stats);
1500 tbl->stats = NULL;
1502 kmem_cache_destroy(tbl->kmem_cachep);
1503 tbl->kmem_cachep = NULL;
1505 return 0;
1507 EXPORT_SYMBOL(neigh_table_clear);
1509 static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1511 struct net *net = sock_net(skb->sk);
1512 struct ndmsg *ndm;
1513 struct nlattr *dst_attr;
1514 struct neigh_table *tbl;
1515 struct net_device *dev = NULL;
1516 int err = -EINVAL;
1518 if (nlmsg_len(nlh) < sizeof(*ndm))
1519 goto out;
1521 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1522 if (dst_attr == NULL)
1523 goto out;
1525 ndm = nlmsg_data(nlh);
1526 if (ndm->ndm_ifindex) {
1527 dev = dev_get_by_index(net, ndm->ndm_ifindex);
1528 if (dev == NULL) {
1529 err = -ENODEV;
1530 goto out;
1534 read_lock(&neigh_tbl_lock);
1535 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1536 struct neighbour *neigh;
1538 if (tbl->family != ndm->ndm_family)
1539 continue;
1540 read_unlock(&neigh_tbl_lock);
1542 if (nla_len(dst_attr) < tbl->key_len)
1543 goto out_dev_put;
1545 if (ndm->ndm_flags & NTF_PROXY) {
1546 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1547 goto out_dev_put;
1550 if (dev == NULL)
1551 goto out_dev_put;
1553 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1554 if (neigh == NULL) {
1555 err = -ENOENT;
1556 goto out_dev_put;
1559 err = neigh_update(neigh, NULL, NUD_FAILED,
1560 NEIGH_UPDATE_F_OVERRIDE |
1561 NEIGH_UPDATE_F_ADMIN);
1562 neigh_release(neigh);
1563 goto out_dev_put;
1565 read_unlock(&neigh_tbl_lock);
1566 err = -EAFNOSUPPORT;
1568 out_dev_put:
1569 if (dev)
1570 dev_put(dev);
1571 out:
1572 return err;
1575 static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1577 struct net *net = sock_net(skb->sk);
1578 struct ndmsg *ndm;
1579 struct nlattr *tb[NDA_MAX+1];
1580 struct neigh_table *tbl;
1581 struct net_device *dev = NULL;
1582 int err;
1584 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1585 if (err < 0)
1586 goto out;
1588 err = -EINVAL;
1589 if (tb[NDA_DST] == NULL)
1590 goto out;
1592 ndm = nlmsg_data(nlh);
1593 if (ndm->ndm_ifindex) {
1594 dev = dev_get_by_index(net, ndm->ndm_ifindex);
1595 if (dev == NULL) {
1596 err = -ENODEV;
1597 goto out;
1600 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1601 goto out_dev_put;
1604 read_lock(&neigh_tbl_lock);
1605 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1606 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1607 struct neighbour *neigh;
1608 void *dst, *lladdr;
1610 if (tbl->family != ndm->ndm_family)
1611 continue;
1612 read_unlock(&neigh_tbl_lock);
1614 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1615 goto out_dev_put;
1616 dst = nla_data(tb[NDA_DST]);
1617 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1619 if (ndm->ndm_flags & NTF_PROXY) {
1620 struct pneigh_entry *pn;
1622 err = -ENOBUFS;
1623 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1624 if (pn) {
1625 pn->flags = ndm->ndm_flags;
1626 err = 0;
1628 goto out_dev_put;
1631 if (dev == NULL)
1632 goto out_dev_put;
1634 neigh = neigh_lookup(tbl, dst, dev);
1635 if (neigh == NULL) {
1636 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1637 err = -ENOENT;
1638 goto out_dev_put;
1641 neigh = __neigh_lookup_errno(tbl, dst, dev);
1642 if (IS_ERR(neigh)) {
1643 err = PTR_ERR(neigh);
1644 goto out_dev_put;
1646 } else {
1647 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1648 err = -EEXIST;
1649 neigh_release(neigh);
1650 goto out_dev_put;
1653 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1654 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1657 if (ndm->ndm_flags & NTF_USE) {
1658 neigh_event_send(neigh, NULL);
1659 err = 0;
1660 } else
1661 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1662 neigh_release(neigh);
1663 goto out_dev_put;
1666 read_unlock(&neigh_tbl_lock);
1667 err = -EAFNOSUPPORT;
1669 out_dev_put:
1670 if (dev)
1671 dev_put(dev);
1672 out:
1673 return err;
1676 static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1678 struct nlattr *nest;
1680 nest = nla_nest_start(skb, NDTA_PARMS);
1681 if (nest == NULL)
1682 return -ENOBUFS;
1684 if (parms->dev)
1685 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1687 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1688 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1689 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1690 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1691 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1692 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1693 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1694 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1695 parms->base_reachable_time);
1696 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1697 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1698 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1699 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1700 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1701 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1703 return nla_nest_end(skb, nest);
1705 nla_put_failure:
1706 nla_nest_cancel(skb, nest);
1707 return -EMSGSIZE;
1710 static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1711 u32 pid, u32 seq, int type, int flags)
1713 struct nlmsghdr *nlh;
1714 struct ndtmsg *ndtmsg;
1716 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1717 if (nlh == NULL)
1718 return -EMSGSIZE;
1720 ndtmsg = nlmsg_data(nlh);
1722 read_lock_bh(&tbl->lock);
1723 ndtmsg->ndtm_family = tbl->family;
1724 ndtmsg->ndtm_pad1 = 0;
1725 ndtmsg->ndtm_pad2 = 0;
1727 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1728 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1729 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1730 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1731 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1734 unsigned long now = jiffies;
1735 unsigned int flush_delta = now - tbl->last_flush;
1736 unsigned int rand_delta = now - tbl->last_rand;
1738 struct ndt_config ndc = {
1739 .ndtc_key_len = tbl->key_len,
1740 .ndtc_entry_size = tbl->entry_size,
1741 .ndtc_entries = atomic_read(&tbl->entries),
1742 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1743 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1744 .ndtc_hash_rnd = tbl->hash_rnd,
1745 .ndtc_hash_mask = tbl->hash_mask,
1746 .ndtc_hash_chain_gc = tbl->hash_chain_gc,
1747 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1750 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
1754 int cpu;
1755 struct ndt_stats ndst;
1757 memset(&ndst, 0, sizeof(ndst));
1759 for_each_possible_cpu(cpu) {
1760 struct neigh_statistics *st;
1762 st = per_cpu_ptr(tbl->stats, cpu);
1763 ndst.ndts_allocs += st->allocs;
1764 ndst.ndts_destroys += st->destroys;
1765 ndst.ndts_hash_grows += st->hash_grows;
1766 ndst.ndts_res_failed += st->res_failed;
1767 ndst.ndts_lookups += st->lookups;
1768 ndst.ndts_hits += st->hits;
1769 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1770 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1771 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1772 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1775 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
1778 BUG_ON(tbl->parms.dev);
1779 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1780 goto nla_put_failure;
1782 read_unlock_bh(&tbl->lock);
1783 return nlmsg_end(skb, nlh);
1785 nla_put_failure:
1786 read_unlock_bh(&tbl->lock);
1787 nlmsg_cancel(skb, nlh);
1788 return -EMSGSIZE;
1791 static int neightbl_fill_param_info(struct sk_buff *skb,
1792 struct neigh_table *tbl,
1793 struct neigh_parms *parms,
1794 u32 pid, u32 seq, int type,
1795 unsigned int flags)
1797 struct ndtmsg *ndtmsg;
1798 struct nlmsghdr *nlh;
1800 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1801 if (nlh == NULL)
1802 return -EMSGSIZE;
1804 ndtmsg = nlmsg_data(nlh);
1806 read_lock_bh(&tbl->lock);
1807 ndtmsg->ndtm_family = tbl->family;
1808 ndtmsg->ndtm_pad1 = 0;
1809 ndtmsg->ndtm_pad2 = 0;
1811 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1812 neightbl_fill_parms(skb, parms) < 0)
1813 goto errout;
1815 read_unlock_bh(&tbl->lock);
1816 return nlmsg_end(skb, nlh);
1817 errout:
1818 read_unlock_bh(&tbl->lock);
1819 nlmsg_cancel(skb, nlh);
1820 return -EMSGSIZE;
1823 static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1824 [NDTA_NAME] = { .type = NLA_STRING },
1825 [NDTA_THRESH1] = { .type = NLA_U32 },
1826 [NDTA_THRESH2] = { .type = NLA_U32 },
1827 [NDTA_THRESH3] = { .type = NLA_U32 },
1828 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1829 [NDTA_PARMS] = { .type = NLA_NESTED },
1832 static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1833 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1834 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1835 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1836 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1837 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1838 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1839 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1840 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1841 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1842 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1843 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1844 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1845 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1848 static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1850 struct net *net = sock_net(skb->sk);
1851 struct neigh_table *tbl;
1852 struct ndtmsg *ndtmsg;
1853 struct nlattr *tb[NDTA_MAX+1];
1854 int err;
1856 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1857 nl_neightbl_policy);
1858 if (err < 0)
1859 goto errout;
1861 if (tb[NDTA_NAME] == NULL) {
1862 err = -EINVAL;
1863 goto errout;
1866 ndtmsg = nlmsg_data(nlh);
1867 read_lock(&neigh_tbl_lock);
1868 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1869 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1870 continue;
1872 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1873 break;
1876 if (tbl == NULL) {
1877 err = -ENOENT;
1878 goto errout_locked;
1882 * We acquire tbl->lock to be nice to the periodic timers and
1883 * make sure they always see a consistent set of values.
1885 write_lock_bh(&tbl->lock);
1887 if (tb[NDTA_PARMS]) {
1888 struct nlattr *tbp[NDTPA_MAX+1];
1889 struct neigh_parms *p;
1890 int i, ifindex = 0;
1892 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1893 nl_ntbl_parm_policy);
1894 if (err < 0)
1895 goto errout_tbl_lock;
1897 if (tbp[NDTPA_IFINDEX])
1898 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1900 p = lookup_neigh_params(tbl, net, ifindex);
1901 if (p == NULL) {
1902 err = -ENOENT;
1903 goto errout_tbl_lock;
1906 for (i = 1; i <= NDTPA_MAX; i++) {
1907 if (tbp[i] == NULL)
1908 continue;
1910 switch (i) {
1911 case NDTPA_QUEUE_LEN:
1912 p->queue_len = nla_get_u32(tbp[i]);
1913 break;
1914 case NDTPA_PROXY_QLEN:
1915 p->proxy_qlen = nla_get_u32(tbp[i]);
1916 break;
1917 case NDTPA_APP_PROBES:
1918 p->app_probes = nla_get_u32(tbp[i]);
1919 break;
1920 case NDTPA_UCAST_PROBES:
1921 p->ucast_probes = nla_get_u32(tbp[i]);
1922 break;
1923 case NDTPA_MCAST_PROBES:
1924 p->mcast_probes = nla_get_u32(tbp[i]);
1925 break;
1926 case NDTPA_BASE_REACHABLE_TIME:
1927 p->base_reachable_time = nla_get_msecs(tbp[i]);
1928 break;
1929 case NDTPA_GC_STALETIME:
1930 p->gc_staletime = nla_get_msecs(tbp[i]);
1931 break;
1932 case NDTPA_DELAY_PROBE_TIME:
1933 p->delay_probe_time = nla_get_msecs(tbp[i]);
1934 break;
1935 case NDTPA_RETRANS_TIME:
1936 p->retrans_time = nla_get_msecs(tbp[i]);
1937 break;
1938 case NDTPA_ANYCAST_DELAY:
1939 p->anycast_delay = nla_get_msecs(tbp[i]);
1940 break;
1941 case NDTPA_PROXY_DELAY:
1942 p->proxy_delay = nla_get_msecs(tbp[i]);
1943 break;
1944 case NDTPA_LOCKTIME:
1945 p->locktime = nla_get_msecs(tbp[i]);
1946 break;
1951 if (tb[NDTA_THRESH1])
1952 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
1954 if (tb[NDTA_THRESH2])
1955 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
1957 if (tb[NDTA_THRESH3])
1958 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
1960 if (tb[NDTA_GC_INTERVAL])
1961 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
1963 err = 0;
1965 errout_tbl_lock:
1966 write_unlock_bh(&tbl->lock);
1967 errout_locked:
1968 read_unlock(&neigh_tbl_lock);
1969 errout:
1970 return err;
1973 static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
1975 struct net *net = sock_net(skb->sk);
1976 int family, tidx, nidx = 0;
1977 int tbl_skip = cb->args[0];
1978 int neigh_skip = cb->args[1];
1979 struct neigh_table *tbl;
1981 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
1983 read_lock(&neigh_tbl_lock);
1984 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
1985 struct neigh_parms *p;
1987 if (tidx < tbl_skip || (family && tbl->family != family))
1988 continue;
1990 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
1991 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
1992 NLM_F_MULTI) <= 0)
1993 break;
1995 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
1996 if (!net_eq(neigh_parms_net(p), net))
1997 continue;
1999 if (nidx < neigh_skip)
2000 goto next;
2002 if (neightbl_fill_param_info(skb, tbl, p,
2003 NETLINK_CB(cb->skb).pid,
2004 cb->nlh->nlmsg_seq,
2005 RTM_NEWNEIGHTBL,
2006 NLM_F_MULTI) <= 0)
2007 goto out;
2008 next:
2009 nidx++;
2012 neigh_skip = 0;
2014 out:
2015 read_unlock(&neigh_tbl_lock);
2016 cb->args[0] = tidx;
2017 cb->args[1] = nidx;
2019 return skb->len;
2022 static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2023 u32 pid, u32 seq, int type, unsigned int flags)
2025 unsigned long now = jiffies;
2026 struct nda_cacheinfo ci;
2027 struct nlmsghdr *nlh;
2028 struct ndmsg *ndm;
2030 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2031 if (nlh == NULL)
2032 return -EMSGSIZE;
2034 ndm = nlmsg_data(nlh);
2035 ndm->ndm_family = neigh->ops->family;
2036 ndm->ndm_pad1 = 0;
2037 ndm->ndm_pad2 = 0;
2038 ndm->ndm_flags = neigh->flags;
2039 ndm->ndm_type = neigh->type;
2040 ndm->ndm_ifindex = neigh->dev->ifindex;
2042 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2044 read_lock_bh(&neigh->lock);
2045 ndm->ndm_state = neigh->nud_state;
2046 if ((neigh->nud_state & NUD_VALID) &&
2047 nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, neigh->ha) < 0) {
2048 read_unlock_bh(&neigh->lock);
2049 goto nla_put_failure;
2052 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2053 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2054 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2055 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2056 read_unlock_bh(&neigh->lock);
2058 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2059 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2061 return nlmsg_end(skb, nlh);
2063 nla_put_failure:
2064 nlmsg_cancel(skb, nlh);
2065 return -EMSGSIZE;
2068 static void neigh_update_notify(struct neighbour *neigh)
2070 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2071 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2074 static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2075 struct netlink_callback *cb)
2077 struct net * net = sock_net(skb->sk);
2078 struct neighbour *n;
2079 int rc, h, s_h = cb->args[1];
2080 int idx, s_idx = idx = cb->args[2];
2082 read_lock_bh(&tbl->lock);
2083 for (h = 0; h <= tbl->hash_mask; h++) {
2084 if (h < s_h)
2085 continue;
2086 if (h > s_h)
2087 s_idx = 0;
2088 for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next) {
2089 if (dev_net(n->dev) != net)
2090 continue;
2091 if (idx < s_idx)
2092 goto next;
2093 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2094 cb->nlh->nlmsg_seq,
2095 RTM_NEWNEIGH,
2096 NLM_F_MULTI) <= 0) {
2097 read_unlock_bh(&tbl->lock);
2098 rc = -1;
2099 goto out;
2101 next:
2102 idx++;
2105 read_unlock_bh(&tbl->lock);
2106 rc = skb->len;
2107 out:
2108 cb->args[1] = h;
2109 cb->args[2] = idx;
2110 return rc;
2113 static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2115 struct neigh_table *tbl;
2116 int t, family, s_t;
2118 read_lock(&neigh_tbl_lock);
2119 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2120 s_t = cb->args[0];
2122 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2123 if (t < s_t || (family && tbl->family != family))
2124 continue;
2125 if (t > s_t)
2126 memset(&cb->args[1], 0, sizeof(cb->args) -
2127 sizeof(cb->args[0]));
2128 if (neigh_dump_table(tbl, skb, cb) < 0)
2129 break;
2131 read_unlock(&neigh_tbl_lock);
2133 cb->args[0] = t;
2134 return skb->len;
2137 void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2139 int chain;
2141 read_lock_bh(&tbl->lock);
2142 for (chain = 0; chain <= tbl->hash_mask; chain++) {
2143 struct neighbour *n;
2145 for (n = tbl->hash_buckets[chain]; n; n = n->next)
2146 cb(n, cookie);
2148 read_unlock_bh(&tbl->lock);
2150 EXPORT_SYMBOL(neigh_for_each);
2152 /* The tbl->lock must be held as a writer and BH disabled. */
2153 void __neigh_for_each_release(struct neigh_table *tbl,
2154 int (*cb)(struct neighbour *))
2156 int chain;
2158 for (chain = 0; chain <= tbl->hash_mask; chain++) {
2159 struct neighbour *n, **np;
2161 np = &tbl->hash_buckets[chain];
2162 while ((n = *np) != NULL) {
2163 int release;
2165 write_lock(&n->lock);
2166 release = cb(n);
2167 if (release) {
2168 *np = n->next;
2169 n->dead = 1;
2170 } else
2171 np = &n->next;
2172 write_unlock(&n->lock);
2173 if (release)
2174 neigh_cleanup_and_release(n);
2178 EXPORT_SYMBOL(__neigh_for_each_release);
2180 #ifdef CONFIG_PROC_FS
2182 static struct neighbour *neigh_get_first(struct seq_file *seq)
2184 struct neigh_seq_state *state = seq->private;
2185 struct net *net = seq_file_net(seq);
2186 struct neigh_table *tbl = state->tbl;
2187 struct neighbour *n = NULL;
2188 int bucket = state->bucket;
2190 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2191 for (bucket = 0; bucket <= tbl->hash_mask; bucket++) {
2192 n = tbl->hash_buckets[bucket];
2194 while (n) {
2195 if (!net_eq(dev_net(n->dev), net))
2196 goto next;
2197 if (state->neigh_sub_iter) {
2198 loff_t fakep = 0;
2199 void *v;
2201 v = state->neigh_sub_iter(state, n, &fakep);
2202 if (!v)
2203 goto next;
2205 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2206 break;
2207 if (n->nud_state & ~NUD_NOARP)
2208 break;
2209 next:
2210 n = n->next;
2213 if (n)
2214 break;
2216 state->bucket = bucket;
2218 return n;
2221 static struct neighbour *neigh_get_next(struct seq_file *seq,
2222 struct neighbour *n,
2223 loff_t *pos)
2225 struct neigh_seq_state *state = seq->private;
2226 struct net *net = seq_file_net(seq);
2227 struct neigh_table *tbl = state->tbl;
2229 if (state->neigh_sub_iter) {
2230 void *v = state->neigh_sub_iter(state, n, pos);
2231 if (v)
2232 return n;
2234 n = n->next;
2236 while (1) {
2237 while (n) {
2238 if (!net_eq(dev_net(n->dev), net))
2239 goto next;
2240 if (state->neigh_sub_iter) {
2241 void *v = state->neigh_sub_iter(state, n, pos);
2242 if (v)
2243 return n;
2244 goto next;
2246 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2247 break;
2249 if (n->nud_state & ~NUD_NOARP)
2250 break;
2251 next:
2252 n = n->next;
2255 if (n)
2256 break;
2258 if (++state->bucket > tbl->hash_mask)
2259 break;
2261 n = tbl->hash_buckets[state->bucket];
2264 if (n && pos)
2265 --(*pos);
2266 return n;
2269 static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2271 struct neighbour *n = neigh_get_first(seq);
2273 if (n) {
2274 --(*pos);
2275 while (*pos) {
2276 n = neigh_get_next(seq, n, pos);
2277 if (!n)
2278 break;
2281 return *pos ? NULL : n;
2284 static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2286 struct neigh_seq_state *state = seq->private;
2287 struct net *net = seq_file_net(seq);
2288 struct neigh_table *tbl = state->tbl;
2289 struct pneigh_entry *pn = NULL;
2290 int bucket = state->bucket;
2292 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2293 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2294 pn = tbl->phash_buckets[bucket];
2295 while (pn && !net_eq(pneigh_net(pn), net))
2296 pn = pn->next;
2297 if (pn)
2298 break;
2300 state->bucket = bucket;
2302 return pn;
2305 static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2306 struct pneigh_entry *pn,
2307 loff_t *pos)
2309 struct neigh_seq_state *state = seq->private;
2310 struct net *net = seq_file_net(seq);
2311 struct neigh_table *tbl = state->tbl;
2313 pn = pn->next;
2314 while (!pn) {
2315 if (++state->bucket > PNEIGH_HASHMASK)
2316 break;
2317 pn = tbl->phash_buckets[state->bucket];
2318 while (pn && !net_eq(pneigh_net(pn), net))
2319 pn = pn->next;
2320 if (pn)
2321 break;
2324 if (pn && pos)
2325 --(*pos);
2327 return pn;
2330 static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2332 struct pneigh_entry *pn = pneigh_get_first(seq);
2334 if (pn) {
2335 --(*pos);
2336 while (*pos) {
2337 pn = pneigh_get_next(seq, pn, pos);
2338 if (!pn)
2339 break;
2342 return *pos ? NULL : pn;
2345 static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2347 struct neigh_seq_state *state = seq->private;
2348 void *rc;
2349 loff_t idxpos = *pos;
2351 rc = neigh_get_idx(seq, &idxpos);
2352 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2353 rc = pneigh_get_idx(seq, &idxpos);
2355 return rc;
2358 void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2359 __acquires(tbl->lock)
2361 struct neigh_seq_state *state = seq->private;
2363 state->tbl = tbl;
2364 state->bucket = 0;
2365 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2367 read_lock_bh(&tbl->lock);
2369 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2371 EXPORT_SYMBOL(neigh_seq_start);
2373 void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2375 struct neigh_seq_state *state;
2376 void *rc;
2378 if (v == SEQ_START_TOKEN) {
2379 rc = neigh_get_first(seq);
2380 goto out;
2383 state = seq->private;
2384 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2385 rc = neigh_get_next(seq, v, NULL);
2386 if (rc)
2387 goto out;
2388 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2389 rc = pneigh_get_first(seq);
2390 } else {
2391 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2392 rc = pneigh_get_next(seq, v, NULL);
2394 out:
2395 ++(*pos);
2396 return rc;
2398 EXPORT_SYMBOL(neigh_seq_next);
2400 void neigh_seq_stop(struct seq_file *seq, void *v)
2401 __releases(tbl->lock)
2403 struct neigh_seq_state *state = seq->private;
2404 struct neigh_table *tbl = state->tbl;
2406 read_unlock_bh(&tbl->lock);
2408 EXPORT_SYMBOL(neigh_seq_stop);
2410 /* statistics via seq_file */
2412 static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2414 struct proc_dir_entry *pde = seq->private;
2415 struct neigh_table *tbl = pde->data;
2416 int cpu;
2418 if (*pos == 0)
2419 return SEQ_START_TOKEN;
2421 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2422 if (!cpu_possible(cpu))
2423 continue;
2424 *pos = cpu+1;
2425 return per_cpu_ptr(tbl->stats, cpu);
2427 return NULL;
2430 static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2432 struct proc_dir_entry *pde = seq->private;
2433 struct neigh_table *tbl = pde->data;
2434 int cpu;
2436 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2437 if (!cpu_possible(cpu))
2438 continue;
2439 *pos = cpu+1;
2440 return per_cpu_ptr(tbl->stats, cpu);
2442 return NULL;
2445 static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2450 static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2452 struct proc_dir_entry *pde = seq->private;
2453 struct neigh_table *tbl = pde->data;
2454 struct neigh_statistics *st = v;
2456 if (v == SEQ_START_TOKEN) {
2457 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");
2458 return 0;
2461 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2462 "%08lx %08lx %08lx %08lx %08lx\n",
2463 atomic_read(&tbl->entries),
2465 st->allocs,
2466 st->destroys,
2467 st->hash_grows,
2469 st->lookups,
2470 st->hits,
2472 st->res_failed,
2474 st->rcv_probes_mcast,
2475 st->rcv_probes_ucast,
2477 st->periodic_gc_runs,
2478 st->forced_gc_runs,
2479 st->unres_discards
2482 return 0;
2485 static const struct seq_operations neigh_stat_seq_ops = {
2486 .start = neigh_stat_seq_start,
2487 .next = neigh_stat_seq_next,
2488 .stop = neigh_stat_seq_stop,
2489 .show = neigh_stat_seq_show,
2492 static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2494 int ret = seq_open(file, &neigh_stat_seq_ops);
2496 if (!ret) {
2497 struct seq_file *sf = file->private_data;
2498 sf->private = PDE(inode);
2500 return ret;
2503 static const struct file_operations neigh_stat_seq_fops = {
2504 .owner = THIS_MODULE,
2505 .open = neigh_stat_seq_open,
2506 .read = seq_read,
2507 .llseek = seq_lseek,
2508 .release = seq_release,
2511 #endif /* CONFIG_PROC_FS */
2513 static inline size_t neigh_nlmsg_size(void)
2515 return NLMSG_ALIGN(sizeof(struct ndmsg))
2516 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2517 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2518 + nla_total_size(sizeof(struct nda_cacheinfo))
2519 + nla_total_size(4); /* NDA_PROBES */
2522 static void __neigh_notify(struct neighbour *n, int type, int flags)
2524 struct net *net = dev_net(n->dev);
2525 struct sk_buff *skb;
2526 int err = -ENOBUFS;
2528 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2529 if (skb == NULL)
2530 goto errout;
2532 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2533 if (err < 0) {
2534 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2535 WARN_ON(err == -EMSGSIZE);
2536 kfree_skb(skb);
2537 goto errout;
2539 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2540 return;
2541 errout:
2542 if (err < 0)
2543 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2546 #ifdef CONFIG_ARPD
2547 void neigh_app_ns(struct neighbour *n)
2549 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2551 EXPORT_SYMBOL(neigh_app_ns);
2552 #endif /* CONFIG_ARPD */
2554 #ifdef CONFIG_SYSCTL
2556 static struct neigh_sysctl_table {
2557 struct ctl_table_header *sysctl_header;
2558 struct ctl_table neigh_vars[__NET_NEIGH_MAX];
2559 char *dev_name;
2560 } neigh_sysctl_template __read_mostly = {
2561 .neigh_vars = {
2563 .ctl_name = NET_NEIGH_MCAST_SOLICIT,
2564 .procname = "mcast_solicit",
2565 .maxlen = sizeof(int),
2566 .mode = 0644,
2567 .proc_handler = proc_dointvec,
2570 .ctl_name = NET_NEIGH_UCAST_SOLICIT,
2571 .procname = "ucast_solicit",
2572 .maxlen = sizeof(int),
2573 .mode = 0644,
2574 .proc_handler = proc_dointvec,
2577 .ctl_name = NET_NEIGH_APP_SOLICIT,
2578 .procname = "app_solicit",
2579 .maxlen = sizeof(int),
2580 .mode = 0644,
2581 .proc_handler = proc_dointvec,
2584 .procname = "retrans_time",
2585 .maxlen = sizeof(int),
2586 .mode = 0644,
2587 .proc_handler = proc_dointvec_userhz_jiffies,
2590 .ctl_name = NET_NEIGH_REACHABLE_TIME,
2591 .procname = "base_reachable_time",
2592 .maxlen = sizeof(int),
2593 .mode = 0644,
2594 .proc_handler = proc_dointvec_jiffies,
2595 .strategy = sysctl_jiffies,
2598 .ctl_name = NET_NEIGH_DELAY_PROBE_TIME,
2599 .procname = "delay_first_probe_time",
2600 .maxlen = sizeof(int),
2601 .mode = 0644,
2602 .proc_handler = proc_dointvec_jiffies,
2603 .strategy = sysctl_jiffies,
2606 .ctl_name = NET_NEIGH_GC_STALE_TIME,
2607 .procname = "gc_stale_time",
2608 .maxlen = sizeof(int),
2609 .mode = 0644,
2610 .proc_handler = proc_dointvec_jiffies,
2611 .strategy = sysctl_jiffies,
2614 .ctl_name = NET_NEIGH_UNRES_QLEN,
2615 .procname = "unres_qlen",
2616 .maxlen = sizeof(int),
2617 .mode = 0644,
2618 .proc_handler = proc_dointvec,
2621 .ctl_name = NET_NEIGH_PROXY_QLEN,
2622 .procname = "proxy_qlen",
2623 .maxlen = sizeof(int),
2624 .mode = 0644,
2625 .proc_handler = proc_dointvec,
2628 .procname = "anycast_delay",
2629 .maxlen = sizeof(int),
2630 .mode = 0644,
2631 .proc_handler = proc_dointvec_userhz_jiffies,
2634 .procname = "proxy_delay",
2635 .maxlen = sizeof(int),
2636 .mode = 0644,
2637 .proc_handler = proc_dointvec_userhz_jiffies,
2640 .procname = "locktime",
2641 .maxlen = sizeof(int),
2642 .mode = 0644,
2643 .proc_handler = proc_dointvec_userhz_jiffies,
2646 .ctl_name = NET_NEIGH_RETRANS_TIME_MS,
2647 .procname = "retrans_time_ms",
2648 .maxlen = sizeof(int),
2649 .mode = 0644,
2650 .proc_handler = proc_dointvec_ms_jiffies,
2651 .strategy = sysctl_ms_jiffies,
2654 .ctl_name = NET_NEIGH_REACHABLE_TIME_MS,
2655 .procname = "base_reachable_time_ms",
2656 .maxlen = sizeof(int),
2657 .mode = 0644,
2658 .proc_handler = proc_dointvec_ms_jiffies,
2659 .strategy = sysctl_ms_jiffies,
2662 .ctl_name = NET_NEIGH_GC_INTERVAL,
2663 .procname = "gc_interval",
2664 .maxlen = sizeof(int),
2665 .mode = 0644,
2666 .proc_handler = proc_dointvec_jiffies,
2667 .strategy = sysctl_jiffies,
2670 .ctl_name = NET_NEIGH_GC_THRESH1,
2671 .procname = "gc_thresh1",
2672 .maxlen = sizeof(int),
2673 .mode = 0644,
2674 .proc_handler = proc_dointvec,
2677 .ctl_name = NET_NEIGH_GC_THRESH2,
2678 .procname = "gc_thresh2",
2679 .maxlen = sizeof(int),
2680 .mode = 0644,
2681 .proc_handler = proc_dointvec,
2684 .ctl_name = NET_NEIGH_GC_THRESH3,
2685 .procname = "gc_thresh3",
2686 .maxlen = sizeof(int),
2687 .mode = 0644,
2688 .proc_handler = proc_dointvec,
2694 int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2695 int p_id, int pdev_id, char *p_name,
2696 proc_handler *handler, ctl_handler *strategy)
2698 struct neigh_sysctl_table *t;
2699 const char *dev_name_source = NULL;
2701 #define NEIGH_CTL_PATH_ROOT 0
2702 #define NEIGH_CTL_PATH_PROTO 1
2703 #define NEIGH_CTL_PATH_NEIGH 2
2704 #define NEIGH_CTL_PATH_DEV 3
2706 struct ctl_path neigh_path[] = {
2707 { .procname = "net", .ctl_name = CTL_NET, },
2708 { .procname = "proto", .ctl_name = 0, },
2709 { .procname = "neigh", .ctl_name = 0, },
2710 { .procname = "default", .ctl_name = NET_PROTO_CONF_DEFAULT, },
2711 { },
2714 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2715 if (!t)
2716 goto err;
2718 t->neigh_vars[0].data = &p->mcast_probes;
2719 t->neigh_vars[1].data = &p->ucast_probes;
2720 t->neigh_vars[2].data = &p->app_probes;
2721 t->neigh_vars[3].data = &p->retrans_time;
2722 t->neigh_vars[4].data = &p->base_reachable_time;
2723 t->neigh_vars[5].data = &p->delay_probe_time;
2724 t->neigh_vars[6].data = &p->gc_staletime;
2725 t->neigh_vars[7].data = &p->queue_len;
2726 t->neigh_vars[8].data = &p->proxy_qlen;
2727 t->neigh_vars[9].data = &p->anycast_delay;
2728 t->neigh_vars[10].data = &p->proxy_delay;
2729 t->neigh_vars[11].data = &p->locktime;
2730 t->neigh_vars[12].data = &p->retrans_time;
2731 t->neigh_vars[13].data = &p->base_reachable_time;
2733 if (dev) {
2734 dev_name_source = dev->name;
2735 neigh_path[NEIGH_CTL_PATH_DEV].ctl_name = dev->ifindex;
2736 /* Terminate the table early */
2737 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
2738 } else {
2739 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2740 t->neigh_vars[14].data = (int *)(p + 1);
2741 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2742 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2743 t->neigh_vars[17].data = (int *)(p + 1) + 3;
2747 if (handler || strategy) {
2748 /* RetransTime */
2749 t->neigh_vars[3].proc_handler = handler;
2750 t->neigh_vars[3].strategy = strategy;
2751 t->neigh_vars[3].extra1 = dev;
2752 if (!strategy)
2753 t->neigh_vars[3].ctl_name = CTL_UNNUMBERED;
2754 /* ReachableTime */
2755 t->neigh_vars[4].proc_handler = handler;
2756 t->neigh_vars[4].strategy = strategy;
2757 t->neigh_vars[4].extra1 = dev;
2758 if (!strategy)
2759 t->neigh_vars[4].ctl_name = CTL_UNNUMBERED;
2760 /* RetransTime (in milliseconds)*/
2761 t->neigh_vars[12].proc_handler = handler;
2762 t->neigh_vars[12].strategy = strategy;
2763 t->neigh_vars[12].extra1 = dev;
2764 if (!strategy)
2765 t->neigh_vars[12].ctl_name = CTL_UNNUMBERED;
2766 /* ReachableTime (in milliseconds) */
2767 t->neigh_vars[13].proc_handler = handler;
2768 t->neigh_vars[13].strategy = strategy;
2769 t->neigh_vars[13].extra1 = dev;
2770 if (!strategy)
2771 t->neigh_vars[13].ctl_name = CTL_UNNUMBERED;
2774 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2775 if (!t->dev_name)
2776 goto free;
2778 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2779 neigh_path[NEIGH_CTL_PATH_NEIGH].ctl_name = pdev_id;
2780 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2781 neigh_path[NEIGH_CTL_PATH_PROTO].ctl_name = p_id;
2783 t->sysctl_header =
2784 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2785 if (!t->sysctl_header)
2786 goto free_procname;
2788 p->sysctl_table = t;
2789 return 0;
2791 free_procname:
2792 kfree(t->dev_name);
2793 free:
2794 kfree(t);
2795 err:
2796 return -ENOBUFS;
2798 EXPORT_SYMBOL(neigh_sysctl_register);
2800 void neigh_sysctl_unregister(struct neigh_parms *p)
2802 if (p->sysctl_table) {
2803 struct neigh_sysctl_table *t = p->sysctl_table;
2804 p->sysctl_table = NULL;
2805 unregister_sysctl_table(t->sysctl_header);
2806 kfree(t->dev_name);
2807 kfree(t);
2810 EXPORT_SYMBOL(neigh_sysctl_unregister);
2812 #endif /* CONFIG_SYSCTL */
2814 static int __init neigh_init(void)
2816 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL);
2817 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL);
2818 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info);
2820 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info);
2821 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL);
2823 return 0;
2826 subsys_initcall(neigh_init);