Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Generic address resolution entity
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
9 * Fixes:
10 * Vitaly E. Lavrov releasing NULL neighbor in neigh_add.
11 * Harald Welte Add neighbour cache statistics like rtstat
12 */
13
14#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15
16#include <linux/slab.h>
17#include <linux/kmemleak.h>
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/arp.h>
31#include <net/dst.h>
32#include <net/sock.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>
39#include <linux/inetdevice.h>
40#include <net/addrconf.h>
41
42#include <trace/events/neigh.h>
43
44#define NEIGH_DEBUG 1
45#define neigh_dbg(level, fmt, ...) \
46do { \
47 if (level <= NEIGH_DEBUG) \
48 pr_debug(fmt, ##__VA_ARGS__); \
49} while (0)
50
51#define PNEIGH_HASHMASK 0xF
52
53static void neigh_timer_handler(struct timer_list *t);
54static void __neigh_notify(struct neighbour *n, int type, int flags,
55 u32 pid);
56static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid);
57static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
58 struct net_device *dev);
59
60#ifdef CONFIG_PROC_FS
61static const struct seq_operations neigh_stat_seq_ops;
62#endif
63
64/*
65 Neighbour hash table buckets are protected with rwlock tbl->lock.
66
67 - All the scans/updates to hash buckets MUST be made under this lock.
68 - NOTHING clever should be made under this lock: no callbacks
69 to protocol backends, no attempts to send something to network.
70 It will result in deadlocks, if backend/driver wants to use neighbour
71 cache.
72 - If the entry requires some non-trivial actions, increase
73 its reference count and release table lock.
74
75 Neighbour entries are protected:
76 - with reference count.
77 - with rwlock neigh->lock
78
79 Reference count prevents destruction.
80
81 neigh->lock mainly serializes ll address data and its validity state.
82 However, the same lock is used to protect another entry fields:
83 - timer
84 - resolution queue
85
86 Again, nothing clever shall be made under neigh->lock,
87 the most complicated procedure, which we allow is dev->hard_header.
88 It is supposed, that dev->hard_header is simplistic and does
89 not make callbacks to neighbour tables.
90 */
91
92static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
93{
94 kfree_skb(skb);
95 return -ENETDOWN;
96}
97
98static void neigh_cleanup_and_release(struct neighbour *neigh)
99{
100 trace_neigh_cleanup_and_release(neigh, 0);
101 __neigh_notify(neigh, RTM_DELNEIGH, 0, 0);
102 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
103 neigh_release(neigh);
104}
105
106/*
107 * It is random distribution in the interval (1/2)*base...(3/2)*base.
108 * It corresponds to default IPv6 settings and is not overridable,
109 * because it is really reasonable choice.
110 */
111
112unsigned long neigh_rand_reach_time(unsigned long base)
113{
114 return base ? (prandom_u32() % base) + (base >> 1) : 0;
115}
116EXPORT_SYMBOL(neigh_rand_reach_time);
117
118static void neigh_mark_dead(struct neighbour *n)
119{
120 n->dead = 1;
121 if (!list_empty(&n->gc_list)) {
122 list_del_init(&n->gc_list);
123 atomic_dec(&n->tbl->gc_entries);
124 }
125}
126
127static void neigh_update_gc_list(struct neighbour *n)
128{
129 bool on_gc_list, exempt_from_gc;
130
131 write_lock_bh(&n->tbl->lock);
132 write_lock(&n->lock);
133
134 if (n->dead)
135 goto out;
136
137 /* remove from the gc list if new state is permanent or if neighbor
138 * is externally learned; otherwise entry should be on the gc list
139 */
140 exempt_from_gc = n->nud_state & NUD_PERMANENT ||
141 n->flags & NTF_EXT_LEARNED;
142 on_gc_list = !list_empty(&n->gc_list);
143
144 if (exempt_from_gc && on_gc_list) {
145 list_del_init(&n->gc_list);
146 atomic_dec(&n->tbl->gc_entries);
147 } else if (!exempt_from_gc && !on_gc_list) {
148 /* add entries to the tail; cleaning removes from the front */
149 list_add_tail(&n->gc_list, &n->tbl->gc_list);
150 atomic_inc(&n->tbl->gc_entries);
151 }
152
153out:
154 write_unlock(&n->lock);
155 write_unlock_bh(&n->tbl->lock);
156}
157
158static bool neigh_update_ext_learned(struct neighbour *neigh, u32 flags,
159 int *notify)
160{
161 bool rc = false;
162 u8 ndm_flags;
163
164 if (!(flags & NEIGH_UPDATE_F_ADMIN))
165 return rc;
166
167 ndm_flags = (flags & NEIGH_UPDATE_F_EXT_LEARNED) ? NTF_EXT_LEARNED : 0;
168 if ((neigh->flags ^ ndm_flags) & NTF_EXT_LEARNED) {
169 if (ndm_flags & NTF_EXT_LEARNED)
170 neigh->flags |= NTF_EXT_LEARNED;
171 else
172 neigh->flags &= ~NTF_EXT_LEARNED;
173 rc = true;
174 *notify = 1;
175 }
176
177 return rc;
178}
179
180static bool neigh_del(struct neighbour *n, struct neighbour __rcu **np,
181 struct neigh_table *tbl)
182{
183 bool retval = false;
184
185 write_lock(&n->lock);
186 if (refcount_read(&n->refcnt) == 1) {
187 struct neighbour *neigh;
188
189 neigh = rcu_dereference_protected(n->next,
190 lockdep_is_held(&tbl->lock));
191 rcu_assign_pointer(*np, neigh);
192 neigh_mark_dead(n);
193 retval = true;
194 }
195 write_unlock(&n->lock);
196 if (retval)
197 neigh_cleanup_and_release(n);
198 return retval;
199}
200
201bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl)
202{
203 struct neigh_hash_table *nht;
204 void *pkey = ndel->primary_key;
205 u32 hash_val;
206 struct neighbour *n;
207 struct neighbour __rcu **np;
208
209 nht = rcu_dereference_protected(tbl->nht,
210 lockdep_is_held(&tbl->lock));
211 hash_val = tbl->hash(pkey, ndel->dev, nht->hash_rnd);
212 hash_val = hash_val >> (32 - nht->hash_shift);
213
214 np = &nht->hash_buckets[hash_val];
215 while ((n = rcu_dereference_protected(*np,
216 lockdep_is_held(&tbl->lock)))) {
217 if (n == ndel)
218 return neigh_del(n, np, tbl);
219 np = &n->next;
220 }
221 return false;
222}
223
224static int neigh_forced_gc(struct neigh_table *tbl)
225{
226 int max_clean = atomic_read(&tbl->gc_entries) - tbl->gc_thresh2;
227 unsigned long tref = jiffies - 5 * HZ;
228 struct neighbour *n, *tmp;
229 int shrunk = 0;
230
231 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
232
233 write_lock_bh(&tbl->lock);
234
235 list_for_each_entry_safe(n, tmp, &tbl->gc_list, gc_list) {
236 if (refcount_read(&n->refcnt) == 1) {
237 bool remove = false;
238
239 write_lock(&n->lock);
240 if ((n->nud_state == NUD_FAILED) ||
241 (n->nud_state == NUD_NOARP) ||
242 (tbl->is_multicast &&
243 tbl->is_multicast(n->primary_key)) ||
244 time_after(tref, n->updated))
245 remove = true;
246 write_unlock(&n->lock);
247
248 if (remove && neigh_remove_one(n, tbl))
249 shrunk++;
250 if (shrunk >= max_clean)
251 break;
252 }
253 }
254
255 tbl->last_flush = jiffies;
256
257 write_unlock_bh(&tbl->lock);
258
259 return shrunk;
260}
261
262static void neigh_add_timer(struct neighbour *n, unsigned long when)
263{
264 neigh_hold(n);
265 if (unlikely(mod_timer(&n->timer, when))) {
266 printk("NEIGH: BUG, double timer add, state is %x\n",
267 n->nud_state);
268 dump_stack();
269 }
270}
271
272static int neigh_del_timer(struct neighbour *n)
273{
274 if ((n->nud_state & NUD_IN_TIMER) &&
275 del_timer(&n->timer)) {
276 neigh_release(n);
277 return 1;
278 }
279 return 0;
280}
281
282static void pneigh_queue_purge(struct sk_buff_head *list)
283{
284 struct sk_buff *skb;
285
286 while ((skb = skb_dequeue(list)) != NULL) {
287 dev_put(skb->dev);
288 kfree_skb(skb);
289 }
290}
291
292static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev,
293 bool skip_perm)
294{
295 int i;
296 struct neigh_hash_table *nht;
297
298 nht = rcu_dereference_protected(tbl->nht,
299 lockdep_is_held(&tbl->lock));
300
301 for (i = 0; i < (1 << nht->hash_shift); i++) {
302 struct neighbour *n;
303 struct neighbour __rcu **np = &nht->hash_buckets[i];
304
305 while ((n = rcu_dereference_protected(*np,
306 lockdep_is_held(&tbl->lock))) != NULL) {
307 if (dev && n->dev != dev) {
308 np = &n->next;
309 continue;
310 }
311 if (skip_perm && n->nud_state & NUD_PERMANENT) {
312 np = &n->next;
313 continue;
314 }
315 rcu_assign_pointer(*np,
316 rcu_dereference_protected(n->next,
317 lockdep_is_held(&tbl->lock)));
318 write_lock(&n->lock);
319 neigh_del_timer(n);
320 neigh_mark_dead(n);
321 if (refcount_read(&n->refcnt) != 1) {
322 /* The most unpleasant situation.
323 We must destroy neighbour entry,
324 but someone still uses it.
325
326 The destroy will be delayed until
327 the last user releases us, but
328 we must kill timers etc. and move
329 it to safe state.
330 */
331 __skb_queue_purge(&n->arp_queue);
332 n->arp_queue_len_bytes = 0;
333 n->output = neigh_blackhole;
334 if (n->nud_state & NUD_VALID)
335 n->nud_state = NUD_NOARP;
336 else
337 n->nud_state = NUD_NONE;
338 neigh_dbg(2, "neigh %p is stray\n", n);
339 }
340 write_unlock(&n->lock);
341 neigh_cleanup_and_release(n);
342 }
343 }
344}
345
346void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
347{
348 write_lock_bh(&tbl->lock);
349 neigh_flush_dev(tbl, dev, false);
350 write_unlock_bh(&tbl->lock);
351}
352EXPORT_SYMBOL(neigh_changeaddr);
353
354static int __neigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
355 bool skip_perm)
356{
357 write_lock_bh(&tbl->lock);
358 neigh_flush_dev(tbl, dev, skip_perm);
359 pneigh_ifdown_and_unlock(tbl, dev);
360
361 del_timer_sync(&tbl->proxy_timer);
362 pneigh_queue_purge(&tbl->proxy_queue);
363 return 0;
364}
365
366int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev)
367{
368 __neigh_ifdown(tbl, dev, true);
369 return 0;
370}
371EXPORT_SYMBOL(neigh_carrier_down);
372
373int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
374{
375 __neigh_ifdown(tbl, dev, false);
376 return 0;
377}
378EXPORT_SYMBOL(neigh_ifdown);
379
380static struct neighbour *neigh_alloc(struct neigh_table *tbl,
381 struct net_device *dev,
382 bool exempt_from_gc)
383{
384 struct neighbour *n = NULL;
385 unsigned long now = jiffies;
386 int entries;
387
388 if (exempt_from_gc)
389 goto do_alloc;
390
391 entries = atomic_inc_return(&tbl->gc_entries) - 1;
392 if (entries >= tbl->gc_thresh3 ||
393 (entries >= tbl->gc_thresh2 &&
394 time_after(now, tbl->last_flush + 5 * HZ))) {
395 if (!neigh_forced_gc(tbl) &&
396 entries >= tbl->gc_thresh3) {
397 net_info_ratelimited("%s: neighbor table overflow!\n",
398 tbl->id);
399 NEIGH_CACHE_STAT_INC(tbl, table_fulls);
400 goto out_entries;
401 }
402 }
403
404do_alloc:
405 n = kzalloc(tbl->entry_size + dev->neigh_priv_len, GFP_ATOMIC);
406 if (!n)
407 goto out_entries;
408
409 __skb_queue_head_init(&n->arp_queue);
410 rwlock_init(&n->lock);
411 seqlock_init(&n->ha_lock);
412 n->updated = n->used = now;
413 n->nud_state = NUD_NONE;
414 n->output = neigh_blackhole;
415 seqlock_init(&n->hh.hh_lock);
416 n->parms = neigh_parms_clone(&tbl->parms);
417 timer_setup(&n->timer, neigh_timer_handler, 0);
418
419 NEIGH_CACHE_STAT_INC(tbl, allocs);
420 n->tbl = tbl;
421 refcount_set(&n->refcnt, 1);
422 n->dead = 1;
423 INIT_LIST_HEAD(&n->gc_list);
424
425 atomic_inc(&tbl->entries);
426out:
427 return n;
428
429out_entries:
430 if (!exempt_from_gc)
431 atomic_dec(&tbl->gc_entries);
432 goto out;
433}
434
435static void neigh_get_hash_rnd(u32 *x)
436{
437 *x = get_random_u32() | 1;
438}
439
440static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
441{
442 size_t size = (1 << shift) * sizeof(struct neighbour *);
443 struct neigh_hash_table *ret;
444 struct neighbour __rcu **buckets;
445 int i;
446
447 ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
448 if (!ret)
449 return NULL;
450 if (size <= PAGE_SIZE) {
451 buckets = kzalloc(size, GFP_ATOMIC);
452 } else {
453 buckets = (struct neighbour __rcu **)
454 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
455 get_order(size));
456 kmemleak_alloc(buckets, size, 1, GFP_ATOMIC);
457 }
458 if (!buckets) {
459 kfree(ret);
460 return NULL;
461 }
462 ret->hash_buckets = buckets;
463 ret->hash_shift = shift;
464 for (i = 0; i < NEIGH_NUM_HASH_RND; i++)
465 neigh_get_hash_rnd(&ret->hash_rnd[i]);
466 return ret;
467}
468
469static void neigh_hash_free_rcu(struct rcu_head *head)
470{
471 struct neigh_hash_table *nht = container_of(head,
472 struct neigh_hash_table,
473 rcu);
474 size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
475 struct neighbour __rcu **buckets = nht->hash_buckets;
476
477 if (size <= PAGE_SIZE) {
478 kfree(buckets);
479 } else {
480 kmemleak_free(buckets);
481 free_pages((unsigned long)buckets, get_order(size));
482 }
483 kfree(nht);
484}
485
486static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
487 unsigned long new_shift)
488{
489 unsigned int i, hash;
490 struct neigh_hash_table *new_nht, *old_nht;
491
492 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
493
494 old_nht = rcu_dereference_protected(tbl->nht,
495 lockdep_is_held(&tbl->lock));
496 new_nht = neigh_hash_alloc(new_shift);
497 if (!new_nht)
498 return old_nht;
499
500 for (i = 0; i < (1 << old_nht->hash_shift); i++) {
501 struct neighbour *n, *next;
502
503 for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
504 lockdep_is_held(&tbl->lock));
505 n != NULL;
506 n = next) {
507 hash = tbl->hash(n->primary_key, n->dev,
508 new_nht->hash_rnd);
509
510 hash >>= (32 - new_nht->hash_shift);
511 next = rcu_dereference_protected(n->next,
512 lockdep_is_held(&tbl->lock));
513
514 rcu_assign_pointer(n->next,
515 rcu_dereference_protected(
516 new_nht->hash_buckets[hash],
517 lockdep_is_held(&tbl->lock)));
518 rcu_assign_pointer(new_nht->hash_buckets[hash], n);
519 }
520 }
521
522 rcu_assign_pointer(tbl->nht, new_nht);
523 call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
524 return new_nht;
525}
526
527struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
528 struct net_device *dev)
529{
530 struct neighbour *n;
531
532 NEIGH_CACHE_STAT_INC(tbl, lookups);
533
534 rcu_read_lock_bh();
535 n = __neigh_lookup_noref(tbl, pkey, dev);
536 if (n) {
537 if (!refcount_inc_not_zero(&n->refcnt))
538 n = NULL;
539 NEIGH_CACHE_STAT_INC(tbl, hits);
540 }
541
542 rcu_read_unlock_bh();
543 return n;
544}
545EXPORT_SYMBOL(neigh_lookup);
546
547struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
548 const void *pkey)
549{
550 struct neighbour *n;
551 unsigned int key_len = tbl->key_len;
552 u32 hash_val;
553 struct neigh_hash_table *nht;
554
555 NEIGH_CACHE_STAT_INC(tbl, lookups);
556
557 rcu_read_lock_bh();
558 nht = rcu_dereference_bh(tbl->nht);
559 hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
560
561 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
562 n != NULL;
563 n = rcu_dereference_bh(n->next)) {
564 if (!memcmp(n->primary_key, pkey, key_len) &&
565 net_eq(dev_net(n->dev), net)) {
566 if (!refcount_inc_not_zero(&n->refcnt))
567 n = NULL;
568 NEIGH_CACHE_STAT_INC(tbl, hits);
569 break;
570 }
571 }
572
573 rcu_read_unlock_bh();
574 return n;
575}
576EXPORT_SYMBOL(neigh_lookup_nodev);
577
578static struct neighbour *___neigh_create(struct neigh_table *tbl,
579 const void *pkey,
580 struct net_device *dev,
581 bool exempt_from_gc, bool want_ref)
582{
583 struct neighbour *n1, *rc, *n = neigh_alloc(tbl, dev, exempt_from_gc);
584 u32 hash_val;
585 unsigned int key_len = tbl->key_len;
586 int error;
587 struct neigh_hash_table *nht;
588
589 trace_neigh_create(tbl, dev, pkey, n, exempt_from_gc);
590
591 if (!n) {
592 rc = ERR_PTR(-ENOBUFS);
593 goto out;
594 }
595
596 memcpy(n->primary_key, pkey, key_len);
597 n->dev = dev;
598 dev_hold(dev);
599
600 /* Protocol specific setup. */
601 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
602 rc = ERR_PTR(error);
603 goto out_neigh_release;
604 }
605
606 if (dev->netdev_ops->ndo_neigh_construct) {
607 error = dev->netdev_ops->ndo_neigh_construct(dev, n);
608 if (error < 0) {
609 rc = ERR_PTR(error);
610 goto out_neigh_release;
611 }
612 }
613
614 /* Device specific setup. */
615 if (n->parms->neigh_setup &&
616 (error = n->parms->neigh_setup(n)) < 0) {
617 rc = ERR_PTR(error);
618 goto out_neigh_release;
619 }
620
621 n->confirmed = jiffies - (NEIGH_VAR(n->parms, BASE_REACHABLE_TIME) << 1);
622
623 write_lock_bh(&tbl->lock);
624 nht = rcu_dereference_protected(tbl->nht,
625 lockdep_is_held(&tbl->lock));
626
627 if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
628 nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
629
630 hash_val = tbl->hash(n->primary_key, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
631
632 if (n->parms->dead) {
633 rc = ERR_PTR(-EINVAL);
634 goto out_tbl_unlock;
635 }
636
637 for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
638 lockdep_is_held(&tbl->lock));
639 n1 != NULL;
640 n1 = rcu_dereference_protected(n1->next,
641 lockdep_is_held(&tbl->lock))) {
642 if (dev == n1->dev && !memcmp(n1->primary_key, n->primary_key, key_len)) {
643 if (want_ref)
644 neigh_hold(n1);
645 rc = n1;
646 goto out_tbl_unlock;
647 }
648 }
649
650 n->dead = 0;
651 if (!exempt_from_gc)
652 list_add_tail(&n->gc_list, &n->tbl->gc_list);
653
654 if (want_ref)
655 neigh_hold(n);
656 rcu_assign_pointer(n->next,
657 rcu_dereference_protected(nht->hash_buckets[hash_val],
658 lockdep_is_held(&tbl->lock)));
659 rcu_assign_pointer(nht->hash_buckets[hash_val], n);
660 write_unlock_bh(&tbl->lock);
661 neigh_dbg(2, "neigh %p is created\n", n);
662 rc = n;
663out:
664 return rc;
665out_tbl_unlock:
666 write_unlock_bh(&tbl->lock);
667out_neigh_release:
668 if (!exempt_from_gc)
669 atomic_dec(&tbl->gc_entries);
670 neigh_release(n);
671 goto out;
672}
673
674struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
675 struct net_device *dev, bool want_ref)
676{
677 return ___neigh_create(tbl, pkey, dev, false, want_ref);
678}
679EXPORT_SYMBOL(__neigh_create);
680
681static u32 pneigh_hash(const void *pkey, unsigned int key_len)
682{
683 u32 hash_val = *(u32 *)(pkey + key_len - 4);
684 hash_val ^= (hash_val >> 16);
685 hash_val ^= hash_val >> 8;
686 hash_val ^= hash_val >> 4;
687 hash_val &= PNEIGH_HASHMASK;
688 return hash_val;
689}
690
691static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
692 struct net *net,
693 const void *pkey,
694 unsigned int key_len,
695 struct net_device *dev)
696{
697 while (n) {
698 if (!memcmp(n->key, pkey, key_len) &&
699 net_eq(pneigh_net(n), net) &&
700 (n->dev == dev || !n->dev))
701 return n;
702 n = n->next;
703 }
704 return NULL;
705}
706
707struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
708 struct net *net, const void *pkey, struct net_device *dev)
709{
710 unsigned int key_len = tbl->key_len;
711 u32 hash_val = pneigh_hash(pkey, key_len);
712
713 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
714 net, pkey, key_len, dev);
715}
716EXPORT_SYMBOL_GPL(__pneigh_lookup);
717
718struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
719 struct net *net, const void *pkey,
720 struct net_device *dev, int creat)
721{
722 struct pneigh_entry *n;
723 unsigned int key_len = tbl->key_len;
724 u32 hash_val = pneigh_hash(pkey, key_len);
725
726 read_lock_bh(&tbl->lock);
727 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
728 net, pkey, key_len, dev);
729 read_unlock_bh(&tbl->lock);
730
731 if (n || !creat)
732 goto out;
733
734 ASSERT_RTNL();
735
736 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
737 if (!n)
738 goto out;
739
740 n->protocol = 0;
741 write_pnet(&n->net, net);
742 memcpy(n->key, pkey, key_len);
743 n->dev = dev;
744 if (dev)
745 dev_hold(dev);
746
747 if (tbl->pconstructor && tbl->pconstructor(n)) {
748 if (dev)
749 dev_put(dev);
750 kfree(n);
751 n = NULL;
752 goto out;
753 }
754
755 write_lock_bh(&tbl->lock);
756 n->next = tbl->phash_buckets[hash_val];
757 tbl->phash_buckets[hash_val] = n;
758 write_unlock_bh(&tbl->lock);
759out:
760 return n;
761}
762EXPORT_SYMBOL(pneigh_lookup);
763
764
765int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
766 struct net_device *dev)
767{
768 struct pneigh_entry *n, **np;
769 unsigned int key_len = tbl->key_len;
770 u32 hash_val = pneigh_hash(pkey, key_len);
771
772 write_lock_bh(&tbl->lock);
773 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
774 np = &n->next) {
775 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
776 net_eq(pneigh_net(n), net)) {
777 *np = n->next;
778 write_unlock_bh(&tbl->lock);
779 if (tbl->pdestructor)
780 tbl->pdestructor(n);
781 if (n->dev)
782 dev_put(n->dev);
783 kfree(n);
784 return 0;
785 }
786 }
787 write_unlock_bh(&tbl->lock);
788 return -ENOENT;
789}
790
791static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
792 struct net_device *dev)
793{
794 struct pneigh_entry *n, **np, *freelist = NULL;
795 u32 h;
796
797 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
798 np = &tbl->phash_buckets[h];
799 while ((n = *np) != NULL) {
800 if (!dev || n->dev == dev) {
801 *np = n->next;
802 n->next = freelist;
803 freelist = n;
804 continue;
805 }
806 np = &n->next;
807 }
808 }
809 write_unlock_bh(&tbl->lock);
810 while ((n = freelist)) {
811 freelist = n->next;
812 n->next = NULL;
813 if (tbl->pdestructor)
814 tbl->pdestructor(n);
815 if (n->dev)
816 dev_put(n->dev);
817 kfree(n);
818 }
819 return -ENOENT;
820}
821
822static void neigh_parms_destroy(struct neigh_parms *parms);
823
824static inline void neigh_parms_put(struct neigh_parms *parms)
825{
826 if (refcount_dec_and_test(&parms->refcnt))
827 neigh_parms_destroy(parms);
828}
829
830/*
831 * neighbour must already be out of the table;
832 *
833 */
834void neigh_destroy(struct neighbour *neigh)
835{
836 struct net_device *dev = neigh->dev;
837
838 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
839
840 if (!neigh->dead) {
841 pr_warn("Destroying alive neighbour %p\n", neigh);
842 dump_stack();
843 return;
844 }
845
846 if (neigh_del_timer(neigh))
847 pr_warn("Impossible event\n");
848
849 write_lock_bh(&neigh->lock);
850 __skb_queue_purge(&neigh->arp_queue);
851 write_unlock_bh(&neigh->lock);
852 neigh->arp_queue_len_bytes = 0;
853
854 if (dev->netdev_ops->ndo_neigh_destroy)
855 dev->netdev_ops->ndo_neigh_destroy(dev, neigh);
856
857 dev_put(dev);
858 neigh_parms_put(neigh->parms);
859
860 neigh_dbg(2, "neigh %p is destroyed\n", neigh);
861
862 atomic_dec(&neigh->tbl->entries);
863 kfree_rcu(neigh, rcu);
864}
865EXPORT_SYMBOL(neigh_destroy);
866
867/* Neighbour state is suspicious;
868 disable fast path.
869
870 Called with write_locked neigh.
871 */
872static void neigh_suspect(struct neighbour *neigh)
873{
874 neigh_dbg(2, "neigh %p is suspected\n", neigh);
875
876 neigh->output = neigh->ops->output;
877}
878
879/* Neighbour state is OK;
880 enable fast path.
881
882 Called with write_locked neigh.
883 */
884static void neigh_connect(struct neighbour *neigh)
885{
886 neigh_dbg(2, "neigh %p is connected\n", neigh);
887
888 neigh->output = neigh->ops->connected_output;
889}
890
891static void neigh_periodic_work(struct work_struct *work)
892{
893 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
894 struct neighbour *n;
895 struct neighbour __rcu **np;
896 unsigned int i;
897 struct neigh_hash_table *nht;
898
899 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
900
901 write_lock_bh(&tbl->lock);
902 nht = rcu_dereference_protected(tbl->nht,
903 lockdep_is_held(&tbl->lock));
904
905 /*
906 * periodically recompute ReachableTime from random function
907 */
908
909 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
910 struct neigh_parms *p;
911 tbl->last_rand = jiffies;
912 list_for_each_entry(p, &tbl->parms_list, list)
913 p->reachable_time =
914 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
915 }
916
917 if (atomic_read(&tbl->entries) < tbl->gc_thresh1)
918 goto out;
919
920 for (i = 0 ; i < (1 << nht->hash_shift); i++) {
921 np = &nht->hash_buckets[i];
922
923 while ((n = rcu_dereference_protected(*np,
924 lockdep_is_held(&tbl->lock))) != NULL) {
925 unsigned int state;
926
927 write_lock(&n->lock);
928
929 state = n->nud_state;
930 if ((state & (NUD_PERMANENT | NUD_IN_TIMER)) ||
931 (n->flags & NTF_EXT_LEARNED)) {
932 write_unlock(&n->lock);
933 goto next_elt;
934 }
935
936 if (time_before(n->used, n->confirmed))
937 n->used = n->confirmed;
938
939 if (refcount_read(&n->refcnt) == 1 &&
940 (state == NUD_FAILED ||
941 time_after(jiffies, n->used + NEIGH_VAR(n->parms, GC_STALETIME)))) {
942 *np = n->next;
943 neigh_mark_dead(n);
944 write_unlock(&n->lock);
945 neigh_cleanup_and_release(n);
946 continue;
947 }
948 write_unlock(&n->lock);
949
950next_elt:
951 np = &n->next;
952 }
953 /*
954 * It's fine to release lock here, even if hash table
955 * grows while we are preempted.
956 */
957 write_unlock_bh(&tbl->lock);
958 cond_resched();
959 write_lock_bh(&tbl->lock);
960 nht = rcu_dereference_protected(tbl->nht,
961 lockdep_is_held(&tbl->lock));
962 }
963out:
964 /* Cycle through all hash buckets every BASE_REACHABLE_TIME/2 ticks.
965 * ARP entry timeouts range from 1/2 BASE_REACHABLE_TIME to 3/2
966 * BASE_REACHABLE_TIME.
967 */
968 queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
969 NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME) >> 1);
970 write_unlock_bh(&tbl->lock);
971}
972
973static __inline__ int neigh_max_probes(struct neighbour *n)
974{
975 struct neigh_parms *p = n->parms;
976 return NEIGH_VAR(p, UCAST_PROBES) + NEIGH_VAR(p, APP_PROBES) +
977 (n->nud_state & NUD_PROBE ? NEIGH_VAR(p, MCAST_REPROBES) :
978 NEIGH_VAR(p, MCAST_PROBES));
979}
980
981static void neigh_invalidate(struct neighbour *neigh)
982 __releases(neigh->lock)
983 __acquires(neigh->lock)
984{
985 struct sk_buff *skb;
986
987 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
988 neigh_dbg(2, "neigh %p is failed\n", neigh);
989 neigh->updated = jiffies;
990
991 /* It is very thin place. report_unreachable is very complicated
992 routine. Particularly, it can hit the same neighbour entry!
993
994 So that, we try to be accurate and avoid dead loop. --ANK
995 */
996 while (neigh->nud_state == NUD_FAILED &&
997 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
998 write_unlock(&neigh->lock);
999 neigh->ops->error_report(neigh, skb);
1000 write_lock(&neigh->lock);
1001 }
1002 __skb_queue_purge(&neigh->arp_queue);
1003 neigh->arp_queue_len_bytes = 0;
1004}
1005
1006static void neigh_probe(struct neighbour *neigh)
1007 __releases(neigh->lock)
1008{
1009 struct sk_buff *skb = skb_peek_tail(&neigh->arp_queue);
1010 /* keep skb alive even if arp_queue overflows */
1011 if (skb)
1012 skb = skb_clone(skb, GFP_ATOMIC);
1013 write_unlock(&neigh->lock);
1014 if (neigh->ops->solicit)
1015 neigh->ops->solicit(neigh, skb);
1016 atomic_inc(&neigh->probes);
1017 consume_skb(skb);
1018}
1019
1020/* Called when a timer expires for a neighbour entry. */
1021
1022static void neigh_timer_handler(struct timer_list *t)
1023{
1024 unsigned long now, next;
1025 struct neighbour *neigh = from_timer(neigh, t, timer);
1026 unsigned int state;
1027 int notify = 0;
1028
1029 write_lock(&neigh->lock);
1030
1031 state = neigh->nud_state;
1032 now = jiffies;
1033 next = now + HZ;
1034
1035 if (!(state & NUD_IN_TIMER))
1036 goto out;
1037
1038 if (state & NUD_REACHABLE) {
1039 if (time_before_eq(now,
1040 neigh->confirmed + neigh->parms->reachable_time)) {
1041 neigh_dbg(2, "neigh %p is still alive\n", neigh);
1042 next = neigh->confirmed + neigh->parms->reachable_time;
1043 } else if (time_before_eq(now,
1044 neigh->used +
1045 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
1046 neigh_dbg(2, "neigh %p is delayed\n", neigh);
1047 neigh->nud_state = NUD_DELAY;
1048 neigh->updated = jiffies;
1049 neigh_suspect(neigh);
1050 next = now + NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME);
1051 } else {
1052 neigh_dbg(2, "neigh %p is suspected\n", neigh);
1053 neigh->nud_state = NUD_STALE;
1054 neigh->updated = jiffies;
1055 neigh_suspect(neigh);
1056 notify = 1;
1057 }
1058 } else if (state & NUD_DELAY) {
1059 if (time_before_eq(now,
1060 neigh->confirmed +
1061 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
1062 neigh_dbg(2, "neigh %p is now reachable\n", neigh);
1063 neigh->nud_state = NUD_REACHABLE;
1064 neigh->updated = jiffies;
1065 neigh_connect(neigh);
1066 notify = 1;
1067 next = neigh->confirmed + neigh->parms->reachable_time;
1068 } else {
1069 neigh_dbg(2, "neigh %p is probed\n", neigh);
1070 neigh->nud_state = NUD_PROBE;
1071 neigh->updated = jiffies;
1072 atomic_set(&neigh->probes, 0);
1073 notify = 1;
1074 next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1075 HZ/100);
1076 }
1077 } else {
1078 /* NUD_PROBE|NUD_INCOMPLETE */
1079 next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME), HZ/100);
1080 }
1081
1082 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
1083 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
1084 neigh->nud_state = NUD_FAILED;
1085 notify = 1;
1086 neigh_invalidate(neigh);
1087 goto out;
1088 }
1089
1090 if (neigh->nud_state & NUD_IN_TIMER) {
1091 if (time_before(next, jiffies + HZ/100))
1092 next = jiffies + HZ/100;
1093 if (!mod_timer(&neigh->timer, next))
1094 neigh_hold(neigh);
1095 }
1096 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
1097 neigh_probe(neigh);
1098 } else {
1099out:
1100 write_unlock(&neigh->lock);
1101 }
1102
1103 if (notify)
1104 neigh_update_notify(neigh, 0);
1105
1106 trace_neigh_timer_handler(neigh, 0);
1107
1108 neigh_release(neigh);
1109}
1110
1111int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
1112{
1113 int rc;
1114 bool immediate_probe = false;
1115
1116 write_lock_bh(&neigh->lock);
1117
1118 rc = 0;
1119 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
1120 goto out_unlock_bh;
1121 if (neigh->dead)
1122 goto out_dead;
1123
1124 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
1125 if (NEIGH_VAR(neigh->parms, MCAST_PROBES) +
1126 NEIGH_VAR(neigh->parms, APP_PROBES)) {
1127 unsigned long next, now = jiffies;
1128
1129 atomic_set(&neigh->probes,
1130 NEIGH_VAR(neigh->parms, UCAST_PROBES));
1131 neigh_del_timer(neigh);
1132 neigh->nud_state = NUD_INCOMPLETE;
1133 neigh->updated = now;
1134 next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1135 HZ/100);
1136 neigh_add_timer(neigh, next);
1137 immediate_probe = true;
1138 } else {
1139 neigh->nud_state = NUD_FAILED;
1140 neigh->updated = jiffies;
1141 write_unlock_bh(&neigh->lock);
1142
1143 kfree_skb(skb);
1144 return 1;
1145 }
1146 } else if (neigh->nud_state & NUD_STALE) {
1147 neigh_dbg(2, "neigh %p is delayed\n", neigh);
1148 neigh_del_timer(neigh);
1149 neigh->nud_state = NUD_DELAY;
1150 neigh->updated = jiffies;
1151 neigh_add_timer(neigh, jiffies +
1152 NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME));
1153 }
1154
1155 if (neigh->nud_state == NUD_INCOMPLETE) {
1156 if (skb) {
1157 while (neigh->arp_queue_len_bytes + skb->truesize >
1158 NEIGH_VAR(neigh->parms, QUEUE_LEN_BYTES)) {
1159 struct sk_buff *buff;
1160
1161 buff = __skb_dequeue(&neigh->arp_queue);
1162 if (!buff)
1163 break;
1164 neigh->arp_queue_len_bytes -= buff->truesize;
1165 kfree_skb(buff);
1166 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
1167 }
1168 skb_dst_force(skb);
1169 __skb_queue_tail(&neigh->arp_queue, skb);
1170 neigh->arp_queue_len_bytes += skb->truesize;
1171 }
1172 rc = 1;
1173 }
1174out_unlock_bh:
1175 if (immediate_probe)
1176 neigh_probe(neigh);
1177 else
1178 write_unlock(&neigh->lock);
1179 local_bh_enable();
1180 trace_neigh_event_send_done(neigh, rc);
1181 return rc;
1182
1183out_dead:
1184 if (neigh->nud_state & NUD_STALE)
1185 goto out_unlock_bh;
1186 write_unlock_bh(&neigh->lock);
1187 kfree_skb(skb);
1188 trace_neigh_event_send_dead(neigh, 1);
1189 return 1;
1190}
1191EXPORT_SYMBOL(__neigh_event_send);
1192
1193static void neigh_update_hhs(struct neighbour *neigh)
1194{
1195 struct hh_cache *hh;
1196 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1197 = NULL;
1198
1199 if (neigh->dev->header_ops)
1200 update = neigh->dev->header_ops->cache_update;
1201
1202 if (update) {
1203 hh = &neigh->hh;
1204 if (READ_ONCE(hh->hh_len)) {
1205 write_seqlock_bh(&hh->hh_lock);
1206 update(hh, neigh->dev, neigh->ha);
1207 write_sequnlock_bh(&hh->hh_lock);
1208 }
1209 }
1210}
1211
1212
1213
1214/* Generic update routine.
1215 -- lladdr is new lladdr or NULL, if it is not supplied.
1216 -- new is new state.
1217 -- flags
1218 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1219 if it is different.
1220 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1221 lladdr instead of overriding it
1222 if it is different.
1223 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1224
1225 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1226 NTF_ROUTER flag.
1227 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1228 a router.
1229
1230 Caller MUST hold reference count on the entry.
1231 */
1232
1233static int __neigh_update(struct neighbour *neigh, const u8 *lladdr,
1234 u8 new, u32 flags, u32 nlmsg_pid,
1235 struct netlink_ext_ack *extack)
1236{
1237 bool ext_learn_change = false;
1238 u8 old;
1239 int err;
1240 int notify = 0;
1241 struct net_device *dev;
1242 int update_isrouter = 0;
1243
1244 trace_neigh_update(neigh, lladdr, new, flags, nlmsg_pid);
1245
1246 write_lock_bh(&neigh->lock);
1247
1248 dev = neigh->dev;
1249 old = neigh->nud_state;
1250 err = -EPERM;
1251
1252 if (neigh->dead) {
1253 NL_SET_ERR_MSG(extack, "Neighbor entry is now dead");
1254 new = old;
1255 goto out;
1256 }
1257 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1258 (old & (NUD_NOARP | NUD_PERMANENT)))
1259 goto out;
1260
1261 ext_learn_change = neigh_update_ext_learned(neigh, flags, ¬ify);
1262
1263 if (!(new & NUD_VALID)) {
1264 neigh_del_timer(neigh);
1265 if (old & NUD_CONNECTED)
1266 neigh_suspect(neigh);
1267 neigh->nud_state = new;
1268 err = 0;
1269 notify = old & NUD_VALID;
1270 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1271 (new & NUD_FAILED)) {
1272 neigh_invalidate(neigh);
1273 notify = 1;
1274 }
1275 goto out;
1276 }
1277
1278 /* Compare new lladdr with cached one */
1279 if (!dev->addr_len) {
1280 /* First case: device needs no address. */
1281 lladdr = neigh->ha;
1282 } else if (lladdr) {
1283 /* The second case: if something is already cached
1284 and a new address is proposed:
1285 - compare new & old
1286 - if they are different, check override flag
1287 */
1288 if ((old & NUD_VALID) &&
1289 !memcmp(lladdr, neigh->ha, dev->addr_len))
1290 lladdr = neigh->ha;
1291 } else {
1292 /* No address is supplied; if we know something,
1293 use it, otherwise discard the request.
1294 */
1295 err = -EINVAL;
1296 if (!(old & NUD_VALID)) {
1297 NL_SET_ERR_MSG(extack, "No link layer address given");
1298 goto out;
1299 }
1300 lladdr = neigh->ha;
1301 }
1302
1303 /* Update confirmed timestamp for neighbour entry after we
1304 * received ARP packet even if it doesn't change IP to MAC binding.
1305 */
1306 if (new & NUD_CONNECTED)
1307 neigh->confirmed = jiffies;
1308
1309 /* If entry was valid and address is not changed,
1310 do not change entry state, if new one is STALE.
1311 */
1312 err = 0;
1313 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1314 if (old & NUD_VALID) {
1315 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1316 update_isrouter = 0;
1317 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1318 (old & NUD_CONNECTED)) {
1319 lladdr = neigh->ha;
1320 new = NUD_STALE;
1321 } else
1322 goto out;
1323 } else {
1324 if (lladdr == neigh->ha && new == NUD_STALE &&
1325 !(flags & NEIGH_UPDATE_F_ADMIN))
1326 new = old;
1327 }
1328 }
1329
1330 /* Update timestamp only once we know we will make a change to the
1331 * neighbour entry. Otherwise we risk to move the locktime window with
1332 * noop updates and ignore relevant ARP updates.
1333 */
1334 if (new != old || lladdr != neigh->ha)
1335 neigh->updated = jiffies;
1336
1337 if (new != old) {
1338 neigh_del_timer(neigh);
1339 if (new & NUD_PROBE)
1340 atomic_set(&neigh->probes, 0);
1341 if (new & NUD_IN_TIMER)
1342 neigh_add_timer(neigh, (jiffies +
1343 ((new & NUD_REACHABLE) ?
1344 neigh->parms->reachable_time :
1345 0)));
1346 neigh->nud_state = new;
1347 notify = 1;
1348 }
1349
1350 if (lladdr != neigh->ha) {
1351 write_seqlock(&neigh->ha_lock);
1352 memcpy(&neigh->ha, lladdr, dev->addr_len);
1353 write_sequnlock(&neigh->ha_lock);
1354 neigh_update_hhs(neigh);
1355 if (!(new & NUD_CONNECTED))
1356 neigh->confirmed = jiffies -
1357 (NEIGH_VAR(neigh->parms, BASE_REACHABLE_TIME) << 1);
1358 notify = 1;
1359 }
1360 if (new == old)
1361 goto out;
1362 if (new & NUD_CONNECTED)
1363 neigh_connect(neigh);
1364 else
1365 neigh_suspect(neigh);
1366 if (!(old & NUD_VALID)) {
1367 struct sk_buff *skb;
1368
1369 /* Again: avoid dead loop if something went wrong */
1370
1371 while (neigh->nud_state & NUD_VALID &&
1372 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1373 struct dst_entry *dst = skb_dst(skb);
1374 struct neighbour *n2, *n1 = neigh;
1375 write_unlock_bh(&neigh->lock);
1376
1377 rcu_read_lock();
1378
1379 /* Why not just use 'neigh' as-is? The problem is that
1380 * things such as shaper, eql, and sch_teql can end up
1381 * using alternative, different, neigh objects to output
1382 * the packet in the output path. So what we need to do
1383 * here is re-lookup the top-level neigh in the path so
1384 * we can reinject the packet there.
1385 */
1386 n2 = NULL;
1387 if (dst && dst->obsolete != DST_OBSOLETE_DEAD) {
1388 n2 = dst_neigh_lookup_skb(dst, skb);
1389 if (n2)
1390 n1 = n2;
1391 }
1392 n1->output(n1, skb);
1393 if (n2)
1394 neigh_release(n2);
1395 rcu_read_unlock();
1396
1397 write_lock_bh(&neigh->lock);
1398 }
1399 __skb_queue_purge(&neigh->arp_queue);
1400 neigh->arp_queue_len_bytes = 0;
1401 }
1402out:
1403 if (update_isrouter)
1404 neigh_update_is_router(neigh, flags, ¬ify);
1405 write_unlock_bh(&neigh->lock);
1406
1407 if (((new ^ old) & NUD_PERMANENT) || ext_learn_change)
1408 neigh_update_gc_list(neigh);
1409
1410 if (notify)
1411 neigh_update_notify(neigh, nlmsg_pid);
1412
1413 trace_neigh_update_done(neigh, err);
1414
1415 return err;
1416}
1417
1418int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1419 u32 flags, u32 nlmsg_pid)
1420{
1421 return __neigh_update(neigh, lladdr, new, flags, nlmsg_pid, NULL);
1422}
1423EXPORT_SYMBOL(neigh_update);
1424
1425/* Update the neigh to listen temporarily for probe responses, even if it is
1426 * in a NUD_FAILED state. The caller has to hold neigh->lock for writing.
1427 */
1428void __neigh_set_probe_once(struct neighbour *neigh)
1429{
1430 if (neigh->dead)
1431 return;
1432 neigh->updated = jiffies;
1433 if (!(neigh->nud_state & NUD_FAILED))
1434 return;
1435 neigh->nud_state = NUD_INCOMPLETE;
1436 atomic_set(&neigh->probes, neigh_max_probes(neigh));
1437 neigh_add_timer(neigh,
1438 jiffies + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1439 HZ/100));
1440}
1441EXPORT_SYMBOL(__neigh_set_probe_once);
1442
1443struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1444 u8 *lladdr, void *saddr,
1445 struct net_device *dev)
1446{
1447 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1448 lladdr || !dev->addr_len);
1449 if (neigh)
1450 neigh_update(neigh, lladdr, NUD_STALE,
1451 NEIGH_UPDATE_F_OVERRIDE, 0);
1452 return neigh;
1453}
1454EXPORT_SYMBOL(neigh_event_ns);
1455
1456/* called with read_lock_bh(&n->lock); */
1457static void neigh_hh_init(struct neighbour *n)
1458{
1459 struct net_device *dev = n->dev;
1460 __be16 prot = n->tbl->protocol;
1461 struct hh_cache *hh = &n->hh;
1462
1463 write_lock_bh(&n->lock);
1464
1465 /* Only one thread can come in here and initialize the
1466 * hh_cache entry.
1467 */
1468 if (!hh->hh_len)
1469 dev->header_ops->cache(n, hh, prot);
1470
1471 write_unlock_bh(&n->lock);
1472}
1473
1474/* Slow and careful. */
1475
1476int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1477{
1478 int rc = 0;
1479
1480 if (!neigh_event_send(neigh, skb)) {
1481 int err;
1482 struct net_device *dev = neigh->dev;
1483 unsigned int seq;
1484
1485 if (dev->header_ops->cache && !READ_ONCE(neigh->hh.hh_len))
1486 neigh_hh_init(neigh);
1487
1488 do {
1489 __skb_pull(skb, skb_network_offset(skb));
1490 seq = read_seqbegin(&neigh->ha_lock);
1491 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1492 neigh->ha, NULL, skb->len);
1493 } while (read_seqretry(&neigh->ha_lock, seq));
1494
1495 if (err >= 0)
1496 rc = dev_queue_xmit(skb);
1497 else
1498 goto out_kfree_skb;
1499 }
1500out:
1501 return rc;
1502out_kfree_skb:
1503 rc = -EINVAL;
1504 kfree_skb(skb);
1505 goto out;
1506}
1507EXPORT_SYMBOL(neigh_resolve_output);
1508
1509/* As fast as possible without hh cache */
1510
1511int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1512{
1513 struct net_device *dev = neigh->dev;
1514 unsigned int seq;
1515 int err;
1516
1517 do {
1518 __skb_pull(skb, skb_network_offset(skb));
1519 seq = read_seqbegin(&neigh->ha_lock);
1520 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1521 neigh->ha, NULL, skb->len);
1522 } while (read_seqretry(&neigh->ha_lock, seq));
1523
1524 if (err >= 0)
1525 err = dev_queue_xmit(skb);
1526 else {
1527 err = -EINVAL;
1528 kfree_skb(skb);
1529 }
1530 return err;
1531}
1532EXPORT_SYMBOL(neigh_connected_output);
1533
1534int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1535{
1536 return dev_queue_xmit(skb);
1537}
1538EXPORT_SYMBOL(neigh_direct_output);
1539
1540static void neigh_proxy_process(struct timer_list *t)
1541{
1542 struct neigh_table *tbl = from_timer(tbl, t, proxy_timer);
1543 long sched_next = 0;
1544 unsigned long now = jiffies;
1545 struct sk_buff *skb, *n;
1546
1547 spin_lock(&tbl->proxy_queue.lock);
1548
1549 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1550 long tdif = NEIGH_CB(skb)->sched_next - now;
1551
1552 if (tdif <= 0) {
1553 struct net_device *dev = skb->dev;
1554
1555 __skb_unlink(skb, &tbl->proxy_queue);
1556 if (tbl->proxy_redo && netif_running(dev)) {
1557 rcu_read_lock();
1558 tbl->proxy_redo(skb);
1559 rcu_read_unlock();
1560 } else {
1561 kfree_skb(skb);
1562 }
1563
1564 dev_put(dev);
1565 } else if (!sched_next || tdif < sched_next)
1566 sched_next = tdif;
1567 }
1568 del_timer(&tbl->proxy_timer);
1569 if (sched_next)
1570 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1571 spin_unlock(&tbl->proxy_queue.lock);
1572}
1573
1574void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1575 struct sk_buff *skb)
1576{
1577 unsigned long sched_next = jiffies +
1578 prandom_u32_max(NEIGH_VAR(p, PROXY_DELAY));
1579
1580 if (tbl->proxy_queue.qlen > NEIGH_VAR(p, PROXY_QLEN)) {
1581 kfree_skb(skb);
1582 return;
1583 }
1584
1585 NEIGH_CB(skb)->sched_next = sched_next;
1586 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1587
1588 spin_lock(&tbl->proxy_queue.lock);
1589 if (del_timer(&tbl->proxy_timer)) {
1590 if (time_before(tbl->proxy_timer.expires, sched_next))
1591 sched_next = tbl->proxy_timer.expires;
1592 }
1593 skb_dst_drop(skb);
1594 dev_hold(skb->dev);
1595 __skb_queue_tail(&tbl->proxy_queue, skb);
1596 mod_timer(&tbl->proxy_timer, sched_next);
1597 spin_unlock(&tbl->proxy_queue.lock);
1598}
1599EXPORT_SYMBOL(pneigh_enqueue);
1600
1601static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1602 struct net *net, int ifindex)
1603{
1604 struct neigh_parms *p;
1605
1606 list_for_each_entry(p, &tbl->parms_list, list) {
1607 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1608 (!p->dev && !ifindex && net_eq(net, &init_net)))
1609 return p;
1610 }
1611
1612 return NULL;
1613}
1614
1615struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1616 struct neigh_table *tbl)
1617{
1618 struct neigh_parms *p;
1619 struct net *net = dev_net(dev);
1620 const struct net_device_ops *ops = dev->netdev_ops;
1621
1622 p = kmemdup(&tbl->parms, sizeof(*p), GFP_KERNEL);
1623 if (p) {
1624 p->tbl = tbl;
1625 refcount_set(&p->refcnt, 1);
1626 p->reachable_time =
1627 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
1628 dev_hold(dev);
1629 p->dev = dev;
1630 write_pnet(&p->net, net);
1631 p->sysctl_table = NULL;
1632
1633 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1634 dev_put(dev);
1635 kfree(p);
1636 return NULL;
1637 }
1638
1639 write_lock_bh(&tbl->lock);
1640 list_add(&p->list, &tbl->parms.list);
1641 write_unlock_bh(&tbl->lock);
1642
1643 neigh_parms_data_state_cleanall(p);
1644 }
1645 return p;
1646}
1647EXPORT_SYMBOL(neigh_parms_alloc);
1648
1649static void neigh_rcu_free_parms(struct rcu_head *head)
1650{
1651 struct neigh_parms *parms =
1652 container_of(head, struct neigh_parms, rcu_head);
1653
1654 neigh_parms_put(parms);
1655}
1656
1657void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1658{
1659 if (!parms || parms == &tbl->parms)
1660 return;
1661 write_lock_bh(&tbl->lock);
1662 list_del(&parms->list);
1663 parms->dead = 1;
1664 write_unlock_bh(&tbl->lock);
1665 if (parms->dev)
1666 dev_put(parms->dev);
1667 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1668}
1669EXPORT_SYMBOL(neigh_parms_release);
1670
1671static void neigh_parms_destroy(struct neigh_parms *parms)
1672{
1673 kfree(parms);
1674}
1675
1676static struct lock_class_key neigh_table_proxy_queue_class;
1677
1678static struct neigh_table *neigh_tables[NEIGH_NR_TABLES] __read_mostly;
1679
1680void neigh_table_init(int index, struct neigh_table *tbl)
1681{
1682 unsigned long now = jiffies;
1683 unsigned long phsize;
1684
1685 INIT_LIST_HEAD(&tbl->parms_list);
1686 INIT_LIST_HEAD(&tbl->gc_list);
1687 list_add(&tbl->parms.list, &tbl->parms_list);
1688 write_pnet(&tbl->parms.net, &init_net);
1689 refcount_set(&tbl->parms.refcnt, 1);
1690 tbl->parms.reachable_time =
1691 neigh_rand_reach_time(NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME));
1692
1693 tbl->stats = alloc_percpu(struct neigh_statistics);
1694 if (!tbl->stats)
1695 panic("cannot create neighbour cache statistics");
1696
1697#ifdef CONFIG_PROC_FS
1698 if (!proc_create_seq_data(tbl->id, 0, init_net.proc_net_stat,
1699 &neigh_stat_seq_ops, tbl))
1700 panic("cannot create neighbour proc dir entry");
1701#endif
1702
1703 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1704
1705 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1706 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1707
1708 if (!tbl->nht || !tbl->phash_buckets)
1709 panic("cannot allocate neighbour cache hashes");
1710
1711 if (!tbl->entry_size)
1712 tbl->entry_size = ALIGN(offsetof(struct neighbour, primary_key) +
1713 tbl->key_len, NEIGH_PRIV_ALIGN);
1714 else
1715 WARN_ON(tbl->entry_size % NEIGH_PRIV_ALIGN);
1716
1717 rwlock_init(&tbl->lock);
1718 INIT_DEFERRABLE_WORK(&tbl->gc_work, neigh_periodic_work);
1719 queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
1720 tbl->parms.reachable_time);
1721 timer_setup(&tbl->proxy_timer, neigh_proxy_process, 0);
1722 skb_queue_head_init_class(&tbl->proxy_queue,
1723 &neigh_table_proxy_queue_class);
1724
1725 tbl->last_flush = now;
1726 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1727
1728 neigh_tables[index] = tbl;
1729}
1730EXPORT_SYMBOL(neigh_table_init);
1731
1732int neigh_table_clear(int index, struct neigh_table *tbl)
1733{
1734 neigh_tables[index] = NULL;
1735 /* It is not clean... Fix it to unload IPv6 module safely */
1736 cancel_delayed_work_sync(&tbl->gc_work);
1737 del_timer_sync(&tbl->proxy_timer);
1738 pneigh_queue_purge(&tbl->proxy_queue);
1739 neigh_ifdown(tbl, NULL);
1740 if (atomic_read(&tbl->entries))
1741 pr_crit("neighbour leakage\n");
1742
1743 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1744 neigh_hash_free_rcu);
1745 tbl->nht = NULL;
1746
1747 kfree(tbl->phash_buckets);
1748 tbl->phash_buckets = NULL;
1749
1750 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1751
1752 free_percpu(tbl->stats);
1753 tbl->stats = NULL;
1754
1755 return 0;
1756}
1757EXPORT_SYMBOL(neigh_table_clear);
1758
1759static struct neigh_table *neigh_find_table(int family)
1760{
1761 struct neigh_table *tbl = NULL;
1762
1763 switch (family) {
1764 case AF_INET:
1765 tbl = neigh_tables[NEIGH_ARP_TABLE];
1766 break;
1767 case AF_INET6:
1768 tbl = neigh_tables[NEIGH_ND_TABLE];
1769 break;
1770 case AF_DECnet:
1771 tbl = neigh_tables[NEIGH_DN_TABLE];
1772 break;
1773 }
1774
1775 return tbl;
1776}
1777
1778const struct nla_policy nda_policy[NDA_MAX+1] = {
1779 [NDA_UNSPEC] = { .strict_start_type = NDA_NH_ID },
1780 [NDA_DST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1781 [NDA_LLADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1782 [NDA_CACHEINFO] = { .len = sizeof(struct nda_cacheinfo) },
1783 [NDA_PROBES] = { .type = NLA_U32 },
1784 [NDA_VLAN] = { .type = NLA_U16 },
1785 [NDA_PORT] = { .type = NLA_U16 },
1786 [NDA_VNI] = { .type = NLA_U32 },
1787 [NDA_IFINDEX] = { .type = NLA_U32 },
1788 [NDA_MASTER] = { .type = NLA_U32 },
1789 [NDA_PROTOCOL] = { .type = NLA_U8 },
1790 [NDA_NH_ID] = { .type = NLA_U32 },
1791 [NDA_FDB_EXT_ATTRS] = { .type = NLA_NESTED },
1792};
1793
1794static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh,
1795 struct netlink_ext_ack *extack)
1796{
1797 struct net *net = sock_net(skb->sk);
1798 struct ndmsg *ndm;
1799 struct nlattr *dst_attr;
1800 struct neigh_table *tbl;
1801 struct neighbour *neigh;
1802 struct net_device *dev = NULL;
1803 int err = -EINVAL;
1804
1805 ASSERT_RTNL();
1806 if (nlmsg_len(nlh) < sizeof(*ndm))
1807 goto out;
1808
1809 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1810 if (!dst_attr) {
1811 NL_SET_ERR_MSG(extack, "Network address not specified");
1812 goto out;
1813 }
1814
1815 ndm = nlmsg_data(nlh);
1816 if (ndm->ndm_ifindex) {
1817 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1818 if (dev == NULL) {
1819 err = -ENODEV;
1820 goto out;
1821 }
1822 }
1823
1824 tbl = neigh_find_table(ndm->ndm_family);
1825 if (tbl == NULL)
1826 return -EAFNOSUPPORT;
1827
1828 if (nla_len(dst_attr) < (int)tbl->key_len) {
1829 NL_SET_ERR_MSG(extack, "Invalid network address");
1830 goto out;
1831 }
1832
1833 if (ndm->ndm_flags & NTF_PROXY) {
1834 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1835 goto out;
1836 }
1837
1838 if (dev == NULL)
1839 goto out;
1840
1841 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1842 if (neigh == NULL) {
1843 err = -ENOENT;
1844 goto out;
1845 }
1846
1847 err = __neigh_update(neigh, NULL, NUD_FAILED,
1848 NEIGH_UPDATE_F_OVERRIDE | NEIGH_UPDATE_F_ADMIN,
1849 NETLINK_CB(skb).portid, extack);
1850 write_lock_bh(&tbl->lock);
1851 neigh_release(neigh);
1852 neigh_remove_one(neigh, tbl);
1853 write_unlock_bh(&tbl->lock);
1854
1855out:
1856 return err;
1857}
1858
1859static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh,
1860 struct netlink_ext_ack *extack)
1861{
1862 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE |
1863 NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1864 struct net *net = sock_net(skb->sk);
1865 struct ndmsg *ndm;
1866 struct nlattr *tb[NDA_MAX+1];
1867 struct neigh_table *tbl;
1868 struct net_device *dev = NULL;
1869 struct neighbour *neigh;
1870 void *dst, *lladdr;
1871 u8 protocol = 0;
1872 int err;
1873
1874 ASSERT_RTNL();
1875 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
1876 nda_policy, extack);
1877 if (err < 0)
1878 goto out;
1879
1880 err = -EINVAL;
1881 if (!tb[NDA_DST]) {
1882 NL_SET_ERR_MSG(extack, "Network address not specified");
1883 goto out;
1884 }
1885
1886 ndm = nlmsg_data(nlh);
1887 if (ndm->ndm_ifindex) {
1888 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1889 if (dev == NULL) {
1890 err = -ENODEV;
1891 goto out;
1892 }
1893
1894 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len) {
1895 NL_SET_ERR_MSG(extack, "Invalid link address");
1896 goto out;
1897 }
1898 }
1899
1900 tbl = neigh_find_table(ndm->ndm_family);
1901 if (tbl == NULL)
1902 return -EAFNOSUPPORT;
1903
1904 if (nla_len(tb[NDA_DST]) < (int)tbl->key_len) {
1905 NL_SET_ERR_MSG(extack, "Invalid network address");
1906 goto out;
1907 }
1908
1909 dst = nla_data(tb[NDA_DST]);
1910 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1911
1912 if (tb[NDA_PROTOCOL])
1913 protocol = nla_get_u8(tb[NDA_PROTOCOL]);
1914
1915 if (ndm->ndm_flags & NTF_PROXY) {
1916 struct pneigh_entry *pn;
1917
1918 err = -ENOBUFS;
1919 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1920 if (pn) {
1921 pn->flags = ndm->ndm_flags;
1922 if (protocol)
1923 pn->protocol = protocol;
1924 err = 0;
1925 }
1926 goto out;
1927 }
1928
1929 if (!dev) {
1930 NL_SET_ERR_MSG(extack, "Device not specified");
1931 goto out;
1932 }
1933
1934 if (tbl->allow_add && !tbl->allow_add(dev, extack)) {
1935 err = -EINVAL;
1936 goto out;
1937 }
1938
1939 neigh = neigh_lookup(tbl, dst, dev);
1940 if (neigh == NULL) {
1941 bool exempt_from_gc;
1942
1943 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1944 err = -ENOENT;
1945 goto out;
1946 }
1947
1948 exempt_from_gc = ndm->ndm_state & NUD_PERMANENT ||
1949 ndm->ndm_flags & NTF_EXT_LEARNED;
1950 neigh = ___neigh_create(tbl, dst, dev, exempt_from_gc, true);
1951 if (IS_ERR(neigh)) {
1952 err = PTR_ERR(neigh);
1953 goto out;
1954 }
1955 } else {
1956 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1957 err = -EEXIST;
1958 neigh_release(neigh);
1959 goto out;
1960 }
1961
1962 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1963 flags &= ~(NEIGH_UPDATE_F_OVERRIDE |
1964 NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1965 }
1966
1967 if (protocol)
1968 neigh->protocol = protocol;
1969
1970 if (ndm->ndm_flags & NTF_EXT_LEARNED)
1971 flags |= NEIGH_UPDATE_F_EXT_LEARNED;
1972
1973 if (ndm->ndm_flags & NTF_ROUTER)
1974 flags |= NEIGH_UPDATE_F_ISROUTER;
1975
1976 if (ndm->ndm_flags & NTF_USE) {
1977 neigh_event_send(neigh, NULL);
1978 err = 0;
1979 } else
1980 err = __neigh_update(neigh, lladdr, ndm->ndm_state, flags,
1981 NETLINK_CB(skb).portid, extack);
1982
1983 neigh_release(neigh);
1984
1985out:
1986 return err;
1987}
1988
1989static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1990{
1991 struct nlattr *nest;
1992
1993 nest = nla_nest_start_noflag(skb, NDTA_PARMS);
1994 if (nest == NULL)
1995 return -ENOBUFS;
1996
1997 if ((parms->dev &&
1998 nla_put_u32(skb, NDTPA_IFINDEX, parms->dev->ifindex)) ||
1999 nla_put_u32(skb, NDTPA_REFCNT, refcount_read(&parms->refcnt)) ||
2000 nla_put_u32(skb, NDTPA_QUEUE_LENBYTES,
2001 NEIGH_VAR(parms, QUEUE_LEN_BYTES)) ||
2002 /* approximative value for deprecated QUEUE_LEN (in packets) */
2003 nla_put_u32(skb, NDTPA_QUEUE_LEN,
2004 NEIGH_VAR(parms, QUEUE_LEN_BYTES) / SKB_TRUESIZE(ETH_FRAME_LEN)) ||
2005 nla_put_u32(skb, NDTPA_PROXY_QLEN, NEIGH_VAR(parms, PROXY_QLEN)) ||
2006 nla_put_u32(skb, NDTPA_APP_PROBES, NEIGH_VAR(parms, APP_PROBES)) ||
2007 nla_put_u32(skb, NDTPA_UCAST_PROBES,
2008 NEIGH_VAR(parms, UCAST_PROBES)) ||
2009 nla_put_u32(skb, NDTPA_MCAST_PROBES,
2010 NEIGH_VAR(parms, MCAST_PROBES)) ||
2011 nla_put_u32(skb, NDTPA_MCAST_REPROBES,
2012 NEIGH_VAR(parms, MCAST_REPROBES)) ||
2013 nla_put_msecs(skb, NDTPA_REACHABLE_TIME, parms->reachable_time,
2014 NDTPA_PAD) ||
2015 nla_put_msecs(skb, NDTPA_BASE_REACHABLE_TIME,
2016 NEIGH_VAR(parms, BASE_REACHABLE_TIME), NDTPA_PAD) ||
2017 nla_put_msecs(skb, NDTPA_GC_STALETIME,
2018 NEIGH_VAR(parms, GC_STALETIME), NDTPA_PAD) ||
2019 nla_put_msecs(skb, NDTPA_DELAY_PROBE_TIME,
2020 NEIGH_VAR(parms, DELAY_PROBE_TIME), NDTPA_PAD) ||
2021 nla_put_msecs(skb, NDTPA_RETRANS_TIME,
2022 NEIGH_VAR(parms, RETRANS_TIME), NDTPA_PAD) ||
2023 nla_put_msecs(skb, NDTPA_ANYCAST_DELAY,
2024 NEIGH_VAR(parms, ANYCAST_DELAY), NDTPA_PAD) ||
2025 nla_put_msecs(skb, NDTPA_PROXY_DELAY,
2026 NEIGH_VAR(parms, PROXY_DELAY), NDTPA_PAD) ||
2027 nla_put_msecs(skb, NDTPA_LOCKTIME,
2028 NEIGH_VAR(parms, LOCKTIME), NDTPA_PAD))
2029 goto nla_put_failure;
2030 return nla_nest_end(skb, nest);
2031
2032nla_put_failure:
2033 nla_nest_cancel(skb, nest);
2034 return -EMSGSIZE;
2035}
2036
2037static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
2038 u32 pid, u32 seq, int type, int flags)
2039{
2040 struct nlmsghdr *nlh;
2041 struct ndtmsg *ndtmsg;
2042
2043 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
2044 if (nlh == NULL)
2045 return -EMSGSIZE;
2046
2047 ndtmsg = nlmsg_data(nlh);
2048
2049 read_lock_bh(&tbl->lock);
2050 ndtmsg->ndtm_family = tbl->family;
2051 ndtmsg->ndtm_pad1 = 0;
2052 ndtmsg->ndtm_pad2 = 0;
2053
2054 if (nla_put_string(skb, NDTA_NAME, tbl->id) ||
2055 nla_put_msecs(skb, NDTA_GC_INTERVAL, tbl->gc_interval, NDTA_PAD) ||
2056 nla_put_u32(skb, NDTA_THRESH1, tbl->gc_thresh1) ||
2057 nla_put_u32(skb, NDTA_THRESH2, tbl->gc_thresh2) ||
2058 nla_put_u32(skb, NDTA_THRESH3, tbl->gc_thresh3))
2059 goto nla_put_failure;
2060 {
2061 unsigned long now = jiffies;
2062 long flush_delta = now - tbl->last_flush;
2063 long rand_delta = now - tbl->last_rand;
2064 struct neigh_hash_table *nht;
2065 struct ndt_config ndc = {
2066 .ndtc_key_len = tbl->key_len,
2067 .ndtc_entry_size = tbl->entry_size,
2068 .ndtc_entries = atomic_read(&tbl->entries),
2069 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
2070 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
2071 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
2072 };
2073
2074 rcu_read_lock_bh();
2075 nht = rcu_dereference_bh(tbl->nht);
2076 ndc.ndtc_hash_rnd = nht->hash_rnd[0];
2077 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
2078 rcu_read_unlock_bh();
2079
2080 if (nla_put(skb, NDTA_CONFIG, sizeof(ndc), &ndc))
2081 goto nla_put_failure;
2082 }
2083
2084 {
2085 int cpu;
2086 struct ndt_stats ndst;
2087
2088 memset(&ndst, 0, sizeof(ndst));
2089
2090 for_each_possible_cpu(cpu) {
2091 struct neigh_statistics *st;
2092
2093 st = per_cpu_ptr(tbl->stats, cpu);
2094 ndst.ndts_allocs += st->allocs;
2095 ndst.ndts_destroys += st->destroys;
2096 ndst.ndts_hash_grows += st->hash_grows;
2097 ndst.ndts_res_failed += st->res_failed;
2098 ndst.ndts_lookups += st->lookups;
2099 ndst.ndts_hits += st->hits;
2100 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
2101 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
2102 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
2103 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
2104 ndst.ndts_table_fulls += st->table_fulls;
2105 }
2106
2107 if (nla_put_64bit(skb, NDTA_STATS, sizeof(ndst), &ndst,
2108 NDTA_PAD))
2109 goto nla_put_failure;
2110 }
2111
2112 BUG_ON(tbl->parms.dev);
2113 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
2114 goto nla_put_failure;
2115
2116 read_unlock_bh(&tbl->lock);
2117 nlmsg_end(skb, nlh);
2118 return 0;
2119
2120nla_put_failure:
2121 read_unlock_bh(&tbl->lock);
2122 nlmsg_cancel(skb, nlh);
2123 return -EMSGSIZE;
2124}
2125
2126static int neightbl_fill_param_info(struct sk_buff *skb,
2127 struct neigh_table *tbl,
2128 struct neigh_parms *parms,
2129 u32 pid, u32 seq, int type,
2130 unsigned int flags)
2131{
2132 struct ndtmsg *ndtmsg;
2133 struct nlmsghdr *nlh;
2134
2135 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
2136 if (nlh == NULL)
2137 return -EMSGSIZE;
2138
2139 ndtmsg = nlmsg_data(nlh);
2140
2141 read_lock_bh(&tbl->lock);
2142 ndtmsg->ndtm_family = tbl->family;
2143 ndtmsg->ndtm_pad1 = 0;
2144 ndtmsg->ndtm_pad2 = 0;
2145
2146 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
2147 neightbl_fill_parms(skb, parms) < 0)
2148 goto errout;
2149
2150 read_unlock_bh(&tbl->lock);
2151 nlmsg_end(skb, nlh);
2152 return 0;
2153errout:
2154 read_unlock_bh(&tbl->lock);
2155 nlmsg_cancel(skb, nlh);
2156 return -EMSGSIZE;
2157}
2158
2159static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
2160 [NDTA_NAME] = { .type = NLA_STRING },
2161 [NDTA_THRESH1] = { .type = NLA_U32 },
2162 [NDTA_THRESH2] = { .type = NLA_U32 },
2163 [NDTA_THRESH3] = { .type = NLA_U32 },
2164 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
2165 [NDTA_PARMS] = { .type = NLA_NESTED },
2166};
2167
2168static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
2169 [NDTPA_IFINDEX] = { .type = NLA_U32 },
2170 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
2171 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
2172 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
2173 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
2174 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
2175 [NDTPA_MCAST_REPROBES] = { .type = NLA_U32 },
2176 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
2177 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
2178 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
2179 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
2180 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
2181 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
2182 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
2183};
2184
2185static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
2186 struct netlink_ext_ack *extack)
2187{
2188 struct net *net = sock_net(skb->sk);
2189 struct neigh_table *tbl;
2190 struct ndtmsg *ndtmsg;
2191 struct nlattr *tb[NDTA_MAX+1];
2192 bool found = false;
2193 int err, tidx;
2194
2195 err = nlmsg_parse_deprecated(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
2196 nl_neightbl_policy, extack);
2197 if (err < 0)
2198 goto errout;
2199
2200 if (tb[NDTA_NAME] == NULL) {
2201 err = -EINVAL;
2202 goto errout;
2203 }
2204
2205 ndtmsg = nlmsg_data(nlh);
2206
2207 for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2208 tbl = neigh_tables[tidx];
2209 if (!tbl)
2210 continue;
2211 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
2212 continue;
2213 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0) {
2214 found = true;
2215 break;
2216 }
2217 }
2218
2219 if (!found)
2220 return -ENOENT;
2221
2222 /*
2223 * We acquire tbl->lock to be nice to the periodic timers and
2224 * make sure they always see a consistent set of values.
2225 */
2226 write_lock_bh(&tbl->lock);
2227
2228 if (tb[NDTA_PARMS]) {
2229 struct nlattr *tbp[NDTPA_MAX+1];
2230 struct neigh_parms *p;
2231 int i, ifindex = 0;
2232
2233 err = nla_parse_nested_deprecated(tbp, NDTPA_MAX,
2234 tb[NDTA_PARMS],
2235 nl_ntbl_parm_policy, extack);
2236 if (err < 0)
2237 goto errout_tbl_lock;
2238
2239 if (tbp[NDTPA_IFINDEX])
2240 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
2241
2242 p = lookup_neigh_parms(tbl, net, ifindex);
2243 if (p == NULL) {
2244 err = -ENOENT;
2245 goto errout_tbl_lock;
2246 }
2247
2248 for (i = 1; i <= NDTPA_MAX; i++) {
2249 if (tbp[i] == NULL)
2250 continue;
2251
2252 switch (i) {
2253 case NDTPA_QUEUE_LEN:
2254 NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2255 nla_get_u32(tbp[i]) *
2256 SKB_TRUESIZE(ETH_FRAME_LEN));
2257 break;
2258 case NDTPA_QUEUE_LENBYTES:
2259 NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2260 nla_get_u32(tbp[i]));
2261 break;
2262 case NDTPA_PROXY_QLEN:
2263 NEIGH_VAR_SET(p, PROXY_QLEN,
2264 nla_get_u32(tbp[i]));
2265 break;
2266 case NDTPA_APP_PROBES:
2267 NEIGH_VAR_SET(p, APP_PROBES,
2268 nla_get_u32(tbp[i]));
2269 break;
2270 case NDTPA_UCAST_PROBES:
2271 NEIGH_VAR_SET(p, UCAST_PROBES,
2272 nla_get_u32(tbp[i]));
2273 break;
2274 case NDTPA_MCAST_PROBES:
2275 NEIGH_VAR_SET(p, MCAST_PROBES,
2276 nla_get_u32(tbp[i]));
2277 break;
2278 case NDTPA_MCAST_REPROBES:
2279 NEIGH_VAR_SET(p, MCAST_REPROBES,
2280 nla_get_u32(tbp[i]));
2281 break;
2282 case NDTPA_BASE_REACHABLE_TIME:
2283 NEIGH_VAR_SET(p, BASE_REACHABLE_TIME,
2284 nla_get_msecs(tbp[i]));
2285 /* update reachable_time as well, otherwise, the change will
2286 * only be effective after the next time neigh_periodic_work
2287 * decides to recompute it (can be multiple minutes)
2288 */
2289 p->reachable_time =
2290 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
2291 break;
2292 case NDTPA_GC_STALETIME:
2293 NEIGH_VAR_SET(p, GC_STALETIME,
2294 nla_get_msecs(tbp[i]));
2295 break;
2296 case NDTPA_DELAY_PROBE_TIME:
2297 NEIGH_VAR_SET(p, DELAY_PROBE_TIME,
2298 nla_get_msecs(tbp[i]));
2299 call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
2300 break;
2301 case NDTPA_RETRANS_TIME:
2302 NEIGH_VAR_SET(p, RETRANS_TIME,
2303 nla_get_msecs(tbp[i]));
2304 break;
2305 case NDTPA_ANYCAST_DELAY:
2306 NEIGH_VAR_SET(p, ANYCAST_DELAY,
2307 nla_get_msecs(tbp[i]));
2308 break;
2309 case NDTPA_PROXY_DELAY:
2310 NEIGH_VAR_SET(p, PROXY_DELAY,
2311 nla_get_msecs(tbp[i]));
2312 break;
2313 case NDTPA_LOCKTIME:
2314 NEIGH_VAR_SET(p, LOCKTIME,
2315 nla_get_msecs(tbp[i]));
2316 break;
2317 }
2318 }
2319 }
2320
2321 err = -ENOENT;
2322 if ((tb[NDTA_THRESH1] || tb[NDTA_THRESH2] ||
2323 tb[NDTA_THRESH3] || tb[NDTA_GC_INTERVAL]) &&
2324 !net_eq(net, &init_net))
2325 goto errout_tbl_lock;
2326
2327 if (tb[NDTA_THRESH1])
2328 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2329
2330 if (tb[NDTA_THRESH2])
2331 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2332
2333 if (tb[NDTA_THRESH3])
2334 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2335
2336 if (tb[NDTA_GC_INTERVAL])
2337 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2338
2339 err = 0;
2340
2341errout_tbl_lock:
2342 write_unlock_bh(&tbl->lock);
2343errout:
2344 return err;
2345}
2346
2347static int neightbl_valid_dump_info(const struct nlmsghdr *nlh,
2348 struct netlink_ext_ack *extack)
2349{
2350 struct ndtmsg *ndtm;
2351
2352 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndtm))) {
2353 NL_SET_ERR_MSG(extack, "Invalid header for neighbor table dump request");
2354 return -EINVAL;
2355 }
2356
2357 ndtm = nlmsg_data(nlh);
2358 if (ndtm->ndtm_pad1 || ndtm->ndtm_pad2) {
2359 NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor table dump request");
2360 return -EINVAL;
2361 }
2362
2363 if (nlmsg_attrlen(nlh, sizeof(*ndtm))) {
2364 NL_SET_ERR_MSG(extack, "Invalid data after header in neighbor table dump request");
2365 return -EINVAL;
2366 }
2367
2368 return 0;
2369}
2370
2371static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2372{
2373 const struct nlmsghdr *nlh = cb->nlh;
2374 struct net *net = sock_net(skb->sk);
2375 int family, tidx, nidx = 0;
2376 int tbl_skip = cb->args[0];
2377 int neigh_skip = cb->args[1];
2378 struct neigh_table *tbl;
2379
2380 if (cb->strict_check) {
2381 int err = neightbl_valid_dump_info(nlh, cb->extack);
2382
2383 if (err < 0)
2384 return err;
2385 }
2386
2387 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2388
2389 for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2390 struct neigh_parms *p;
2391
2392 tbl = neigh_tables[tidx];
2393 if (!tbl)
2394 continue;
2395
2396 if (tidx < tbl_skip || (family && tbl->family != family))
2397 continue;
2398
2399 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).portid,
2400 nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2401 NLM_F_MULTI) < 0)
2402 break;
2403
2404 nidx = 0;
2405 p = list_next_entry(&tbl->parms, list);
2406 list_for_each_entry_from(p, &tbl->parms_list, list) {
2407 if (!net_eq(neigh_parms_net(p), net))
2408 continue;
2409
2410 if (nidx < neigh_skip)
2411 goto next;
2412
2413 if (neightbl_fill_param_info(skb, tbl, p,
2414 NETLINK_CB(cb->skb).portid,
2415 nlh->nlmsg_seq,
2416 RTM_NEWNEIGHTBL,
2417 NLM_F_MULTI) < 0)
2418 goto out;
2419 next:
2420 nidx++;
2421 }
2422
2423 neigh_skip = 0;
2424 }
2425out:
2426 cb->args[0] = tidx;
2427 cb->args[1] = nidx;
2428
2429 return skb->len;
2430}
2431
2432static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2433 u32 pid, u32 seq, int type, unsigned int flags)
2434{
2435 unsigned long now = jiffies;
2436 struct nda_cacheinfo ci;
2437 struct nlmsghdr *nlh;
2438 struct ndmsg *ndm;
2439
2440 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2441 if (nlh == NULL)
2442 return -EMSGSIZE;
2443
2444 ndm = nlmsg_data(nlh);
2445 ndm->ndm_family = neigh->ops->family;
2446 ndm->ndm_pad1 = 0;
2447 ndm->ndm_pad2 = 0;
2448 ndm->ndm_flags = neigh->flags;
2449 ndm->ndm_type = neigh->type;
2450 ndm->ndm_ifindex = neigh->dev->ifindex;
2451
2452 if (nla_put(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key))
2453 goto nla_put_failure;
2454
2455 read_lock_bh(&neigh->lock);
2456 ndm->ndm_state = neigh->nud_state;
2457 if (neigh->nud_state & NUD_VALID) {
2458 char haddr[MAX_ADDR_LEN];
2459
2460 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2461 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2462 read_unlock_bh(&neigh->lock);
2463 goto nla_put_failure;
2464 }
2465 }
2466
2467 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2468 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2469 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2470 ci.ndm_refcnt = refcount_read(&neigh->refcnt) - 1;
2471 read_unlock_bh(&neigh->lock);
2472
2473 if (nla_put_u32(skb, NDA_PROBES, atomic_read(&neigh->probes)) ||
2474 nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
2475 goto nla_put_failure;
2476
2477 if (neigh->protocol && nla_put_u8(skb, NDA_PROTOCOL, neigh->protocol))
2478 goto nla_put_failure;
2479
2480 nlmsg_end(skb, nlh);
2481 return 0;
2482
2483nla_put_failure:
2484 nlmsg_cancel(skb, nlh);
2485 return -EMSGSIZE;
2486}
2487
2488static int pneigh_fill_info(struct sk_buff *skb, struct pneigh_entry *pn,
2489 u32 pid, u32 seq, int type, unsigned int flags,
2490 struct neigh_table *tbl)
2491{
2492 struct nlmsghdr *nlh;
2493 struct ndmsg *ndm;
2494
2495 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2496 if (nlh == NULL)
2497 return -EMSGSIZE;
2498
2499 ndm = nlmsg_data(nlh);
2500 ndm->ndm_family = tbl->family;
2501 ndm->ndm_pad1 = 0;
2502 ndm->ndm_pad2 = 0;
2503 ndm->ndm_flags = pn->flags | NTF_PROXY;
2504 ndm->ndm_type = RTN_UNICAST;
2505 ndm->ndm_ifindex = pn->dev ? pn->dev->ifindex : 0;
2506 ndm->ndm_state = NUD_NONE;
2507
2508 if (nla_put(skb, NDA_DST, tbl->key_len, pn->key))
2509 goto nla_put_failure;
2510
2511 if (pn->protocol && nla_put_u8(skb, NDA_PROTOCOL, pn->protocol))
2512 goto nla_put_failure;
2513
2514 nlmsg_end(skb, nlh);
2515 return 0;
2516
2517nla_put_failure:
2518 nlmsg_cancel(skb, nlh);
2519 return -EMSGSIZE;
2520}
2521
2522static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid)
2523{
2524 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2525 __neigh_notify(neigh, RTM_NEWNEIGH, 0, nlmsg_pid);
2526}
2527
2528static bool neigh_master_filtered(struct net_device *dev, int master_idx)
2529{
2530 struct net_device *master;
2531
2532 if (!master_idx)
2533 return false;
2534
2535 master = dev ? netdev_master_upper_dev_get(dev) : NULL;
2536 if (!master || master->ifindex != master_idx)
2537 return true;
2538
2539 return false;
2540}
2541
2542static bool neigh_ifindex_filtered(struct net_device *dev, int filter_idx)
2543{
2544 if (filter_idx && (!dev || dev->ifindex != filter_idx))
2545 return true;
2546
2547 return false;
2548}
2549
2550struct neigh_dump_filter {
2551 int master_idx;
2552 int dev_idx;
2553};
2554
2555static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2556 struct netlink_callback *cb,
2557 struct neigh_dump_filter *filter)
2558{
2559 struct net *net = sock_net(skb->sk);
2560 struct neighbour *n;
2561 int rc, h, s_h = cb->args[1];
2562 int idx, s_idx = idx = cb->args[2];
2563 struct neigh_hash_table *nht;
2564 unsigned int flags = NLM_F_MULTI;
2565
2566 if (filter->dev_idx || filter->master_idx)
2567 flags |= NLM_F_DUMP_FILTERED;
2568
2569 rcu_read_lock_bh();
2570 nht = rcu_dereference_bh(tbl->nht);
2571
2572 for (h = s_h; h < (1 << nht->hash_shift); h++) {
2573 if (h > s_h)
2574 s_idx = 0;
2575 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2576 n != NULL;
2577 n = rcu_dereference_bh(n->next)) {
2578 if (idx < s_idx || !net_eq(dev_net(n->dev), net))
2579 goto next;
2580 if (neigh_ifindex_filtered(n->dev, filter->dev_idx) ||
2581 neigh_master_filtered(n->dev, filter->master_idx))
2582 goto next;
2583 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2584 cb->nlh->nlmsg_seq,
2585 RTM_NEWNEIGH,
2586 flags) < 0) {
2587 rc = -1;
2588 goto out;
2589 }
2590next:
2591 idx++;
2592 }
2593 }
2594 rc = skb->len;
2595out:
2596 rcu_read_unlock_bh();
2597 cb->args[1] = h;
2598 cb->args[2] = idx;
2599 return rc;
2600}
2601
2602static int pneigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2603 struct netlink_callback *cb,
2604 struct neigh_dump_filter *filter)
2605{
2606 struct pneigh_entry *n;
2607 struct net *net = sock_net(skb->sk);
2608 int rc, h, s_h = cb->args[3];
2609 int idx, s_idx = idx = cb->args[4];
2610 unsigned int flags = NLM_F_MULTI;
2611
2612 if (filter->dev_idx || filter->master_idx)
2613 flags |= NLM_F_DUMP_FILTERED;
2614
2615 read_lock_bh(&tbl->lock);
2616
2617 for (h = s_h; h <= PNEIGH_HASHMASK; h++) {
2618 if (h > s_h)
2619 s_idx = 0;
2620 for (n = tbl->phash_buckets[h], idx = 0; n; n = n->next) {
2621 if (idx < s_idx || pneigh_net(n) != net)
2622 goto next;
2623 if (neigh_ifindex_filtered(n->dev, filter->dev_idx) ||
2624 neigh_master_filtered(n->dev, filter->master_idx))
2625 goto next;
2626 if (pneigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2627 cb->nlh->nlmsg_seq,
2628 RTM_NEWNEIGH, flags, tbl) < 0) {
2629 read_unlock_bh(&tbl->lock);
2630 rc = -1;
2631 goto out;
2632 }
2633 next:
2634 idx++;
2635 }
2636 }
2637
2638 read_unlock_bh(&tbl->lock);
2639 rc = skb->len;
2640out:
2641 cb->args[3] = h;
2642 cb->args[4] = idx;
2643 return rc;
2644
2645}
2646
2647static int neigh_valid_dump_req(const struct nlmsghdr *nlh,
2648 bool strict_check,
2649 struct neigh_dump_filter *filter,
2650 struct netlink_ext_ack *extack)
2651{
2652 struct nlattr *tb[NDA_MAX + 1];
2653 int err, i;
2654
2655 if (strict_check) {
2656 struct ndmsg *ndm;
2657
2658 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
2659 NL_SET_ERR_MSG(extack, "Invalid header for neighbor dump request");
2660 return -EINVAL;
2661 }
2662
2663 ndm = nlmsg_data(nlh);
2664 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_ifindex ||
2665 ndm->ndm_state || ndm->ndm_type) {
2666 NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor dump request");
2667 return -EINVAL;
2668 }
2669
2670 if (ndm->ndm_flags & ~NTF_PROXY) {
2671 NL_SET_ERR_MSG(extack, "Invalid flags in header for neighbor dump request");
2672 return -EINVAL;
2673 }
2674
2675 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg),
2676 tb, NDA_MAX, nda_policy,
2677 extack);
2678 } else {
2679 err = nlmsg_parse_deprecated(nlh, sizeof(struct ndmsg), tb,
2680 NDA_MAX, nda_policy, extack);
2681 }
2682 if (err < 0)
2683 return err;
2684
2685 for (i = 0; i <= NDA_MAX; ++i) {
2686 if (!tb[i])
2687 continue;
2688
2689 /* all new attributes should require strict_check */
2690 switch (i) {
2691 case NDA_IFINDEX:
2692 filter->dev_idx = nla_get_u32(tb[i]);
2693 break;
2694 case NDA_MASTER:
2695 filter->master_idx = nla_get_u32(tb[i]);
2696 break;
2697 default:
2698 if (strict_check) {
2699 NL_SET_ERR_MSG(extack, "Unsupported attribute in neighbor dump request");
2700 return -EINVAL;
2701 }
2702 }
2703 }
2704
2705 return 0;
2706}
2707
2708static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2709{
2710 const struct nlmsghdr *nlh = cb->nlh;
2711 struct neigh_dump_filter filter = {};
2712 struct neigh_table *tbl;
2713 int t, family, s_t;
2714 int proxy = 0;
2715 int err;
2716
2717 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2718
2719 /* check for full ndmsg structure presence, family member is
2720 * the same for both structures
2721 */
2722 if (nlmsg_len(nlh) >= sizeof(struct ndmsg) &&
2723 ((struct ndmsg *)nlmsg_data(nlh))->ndm_flags == NTF_PROXY)
2724 proxy = 1;
2725
2726 err = neigh_valid_dump_req(nlh, cb->strict_check, &filter, cb->extack);
2727 if (err < 0 && cb->strict_check)
2728 return err;
2729
2730 s_t = cb->args[0];
2731
2732 for (t = 0; t < NEIGH_NR_TABLES; t++) {
2733 tbl = neigh_tables[t];
2734
2735 if (!tbl)
2736 continue;
2737 if (t < s_t || (family && tbl->family != family))
2738 continue;
2739 if (t > s_t)
2740 memset(&cb->args[1], 0, sizeof(cb->args) -
2741 sizeof(cb->args[0]));
2742 if (proxy)
2743 err = pneigh_dump_table(tbl, skb, cb, &filter);
2744 else
2745 err = neigh_dump_table(tbl, skb, cb, &filter);
2746 if (err < 0)
2747 break;
2748 }
2749
2750 cb->args[0] = t;
2751 return skb->len;
2752}
2753
2754static int neigh_valid_get_req(const struct nlmsghdr *nlh,
2755 struct neigh_table **tbl,
2756 void **dst, int *dev_idx, u8 *ndm_flags,
2757 struct netlink_ext_ack *extack)
2758{
2759 struct nlattr *tb[NDA_MAX + 1];
2760 struct ndmsg *ndm;
2761 int err, i;
2762
2763 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
2764 NL_SET_ERR_MSG(extack, "Invalid header for neighbor get request");
2765 return -EINVAL;
2766 }
2767
2768 ndm = nlmsg_data(nlh);
2769 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
2770 ndm->ndm_type) {
2771 NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor get request");
2772 return -EINVAL;
2773 }
2774
2775 if (ndm->ndm_flags & ~NTF_PROXY) {
2776 NL_SET_ERR_MSG(extack, "Invalid flags in header for neighbor get request");
2777 return -EINVAL;
2778 }
2779
2780 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
2781 NDA_MAX, nda_policy, extack);
2782 if (err < 0)
2783 return err;
2784
2785 *ndm_flags = ndm->ndm_flags;
2786 *dev_idx = ndm->ndm_ifindex;
2787 *tbl = neigh_find_table(ndm->ndm_family);
2788 if (*tbl == NULL) {
2789 NL_SET_ERR_MSG(extack, "Unsupported family in header for neighbor get request");
2790 return -EAFNOSUPPORT;
2791 }
2792
2793 for (i = 0; i <= NDA_MAX; ++i) {
2794 if (!tb[i])
2795 continue;
2796
2797 switch (i) {
2798 case NDA_DST:
2799 if (nla_len(tb[i]) != (int)(*tbl)->key_len) {
2800 NL_SET_ERR_MSG(extack, "Invalid network address in neighbor get request");
2801 return -EINVAL;
2802 }
2803 *dst = nla_data(tb[i]);
2804 break;
2805 default:
2806 NL_SET_ERR_MSG(extack, "Unsupported attribute in neighbor get request");
2807 return -EINVAL;
2808 }
2809 }
2810
2811 return 0;
2812}
2813
2814static inline size_t neigh_nlmsg_size(void)
2815{
2816 return NLMSG_ALIGN(sizeof(struct ndmsg))
2817 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2818 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2819 + nla_total_size(sizeof(struct nda_cacheinfo))
2820 + nla_total_size(4) /* NDA_PROBES */
2821 + nla_total_size(1); /* NDA_PROTOCOL */
2822}
2823
2824static int neigh_get_reply(struct net *net, struct neighbour *neigh,
2825 u32 pid, u32 seq)
2826{
2827 struct sk_buff *skb;
2828 int err = 0;
2829
2830 skb = nlmsg_new(neigh_nlmsg_size(), GFP_KERNEL);
2831 if (!skb)
2832 return -ENOBUFS;
2833
2834 err = neigh_fill_info(skb, neigh, pid, seq, RTM_NEWNEIGH, 0);
2835 if (err) {
2836 kfree_skb(skb);
2837 goto errout;
2838 }
2839
2840 err = rtnl_unicast(skb, net, pid);
2841errout:
2842 return err;
2843}
2844
2845static inline size_t pneigh_nlmsg_size(void)
2846{
2847 return NLMSG_ALIGN(sizeof(struct ndmsg))
2848 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2849 + nla_total_size(1); /* NDA_PROTOCOL */
2850}
2851
2852static int pneigh_get_reply(struct net *net, struct pneigh_entry *neigh,
2853 u32 pid, u32 seq, struct neigh_table *tbl)
2854{
2855 struct sk_buff *skb;
2856 int err = 0;
2857
2858 skb = nlmsg_new(pneigh_nlmsg_size(), GFP_KERNEL);
2859 if (!skb)
2860 return -ENOBUFS;
2861
2862 err = pneigh_fill_info(skb, neigh, pid, seq, RTM_NEWNEIGH, 0, tbl);
2863 if (err) {
2864 kfree_skb(skb);
2865 goto errout;
2866 }
2867
2868 err = rtnl_unicast(skb, net, pid);
2869errout:
2870 return err;
2871}
2872
2873static int neigh_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2874 struct netlink_ext_ack *extack)
2875{
2876 struct net *net = sock_net(in_skb->sk);
2877 struct net_device *dev = NULL;
2878 struct neigh_table *tbl = NULL;
2879 struct neighbour *neigh;
2880 void *dst = NULL;
2881 u8 ndm_flags = 0;
2882 int dev_idx = 0;
2883 int err;
2884
2885 err = neigh_valid_get_req(nlh, &tbl, &dst, &dev_idx, &ndm_flags,
2886 extack);
2887 if (err < 0)
2888 return err;
2889
2890 if (dev_idx) {
2891 dev = __dev_get_by_index(net, dev_idx);
2892 if (!dev) {
2893 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
2894 return -ENODEV;
2895 }
2896 }
2897
2898 if (!dst) {
2899 NL_SET_ERR_MSG(extack, "Network address not specified");
2900 return -EINVAL;
2901 }
2902
2903 if (ndm_flags & NTF_PROXY) {
2904 struct pneigh_entry *pn;
2905
2906 pn = pneigh_lookup(tbl, net, dst, dev, 0);
2907 if (!pn) {
2908 NL_SET_ERR_MSG(extack, "Proxy neighbour entry not found");
2909 return -ENOENT;
2910 }
2911 return pneigh_get_reply(net, pn, NETLINK_CB(in_skb).portid,
2912 nlh->nlmsg_seq, tbl);
2913 }
2914
2915 if (!dev) {
2916 NL_SET_ERR_MSG(extack, "No device specified");
2917 return -EINVAL;
2918 }
2919
2920 neigh = neigh_lookup(tbl, dst, dev);
2921 if (!neigh) {
2922 NL_SET_ERR_MSG(extack, "Neighbour entry not found");
2923 return -ENOENT;
2924 }
2925
2926 err = neigh_get_reply(net, neigh, NETLINK_CB(in_skb).portid,
2927 nlh->nlmsg_seq);
2928
2929 neigh_release(neigh);
2930
2931 return err;
2932}
2933
2934void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2935{
2936 int chain;
2937 struct neigh_hash_table *nht;
2938
2939 rcu_read_lock_bh();
2940 nht = rcu_dereference_bh(tbl->nht);
2941
2942 read_lock(&tbl->lock); /* avoid resizes */
2943 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2944 struct neighbour *n;
2945
2946 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2947 n != NULL;
2948 n = rcu_dereference_bh(n->next))
2949 cb(n, cookie);
2950 }
2951 read_unlock(&tbl->lock);
2952 rcu_read_unlock_bh();
2953}
2954EXPORT_SYMBOL(neigh_for_each);
2955
2956/* The tbl->lock must be held as a writer and BH disabled. */
2957void __neigh_for_each_release(struct neigh_table *tbl,
2958 int (*cb)(struct neighbour *))
2959{
2960 int chain;
2961 struct neigh_hash_table *nht;
2962
2963 nht = rcu_dereference_protected(tbl->nht,
2964 lockdep_is_held(&tbl->lock));
2965 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2966 struct neighbour *n;
2967 struct neighbour __rcu **np;
2968
2969 np = &nht->hash_buckets[chain];
2970 while ((n = rcu_dereference_protected(*np,
2971 lockdep_is_held(&tbl->lock))) != NULL) {
2972 int release;
2973
2974 write_lock(&n->lock);
2975 release = cb(n);
2976 if (release) {
2977 rcu_assign_pointer(*np,
2978 rcu_dereference_protected(n->next,
2979 lockdep_is_held(&tbl->lock)));
2980 neigh_mark_dead(n);
2981 } else
2982 np = &n->next;
2983 write_unlock(&n->lock);
2984 if (release)
2985 neigh_cleanup_and_release(n);
2986 }
2987 }
2988}
2989EXPORT_SYMBOL(__neigh_for_each_release);
2990
2991int neigh_xmit(int index, struct net_device *dev,
2992 const void *addr, struct sk_buff *skb)
2993{
2994 int err = -EAFNOSUPPORT;
2995 if (likely(index < NEIGH_NR_TABLES)) {
2996 struct neigh_table *tbl;
2997 struct neighbour *neigh;
2998
2999 tbl = neigh_tables[index];
3000 if (!tbl)
3001 goto out;
3002 rcu_read_lock_bh();
3003 if (index == NEIGH_ARP_TABLE) {
3004 u32 key = *((u32 *)addr);
3005
3006 neigh = __ipv4_neigh_lookup_noref(dev, key);
3007 } else {
3008 neigh = __neigh_lookup_noref(tbl, addr, dev);
3009 }
3010 if (!neigh)
3011 neigh = __neigh_create(tbl, addr, dev, false);
3012 err = PTR_ERR(neigh);
3013 if (IS_ERR(neigh)) {
3014 rcu_read_unlock_bh();
3015 goto out_kfree_skb;
3016 }
3017 err = neigh->output(neigh, skb);
3018 rcu_read_unlock_bh();
3019 }
3020 else if (index == NEIGH_LINK_TABLE) {
3021 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
3022 addr, NULL, skb->len);
3023 if (err < 0)
3024 goto out_kfree_skb;
3025 err = dev_queue_xmit(skb);
3026 }
3027out:
3028 return err;
3029out_kfree_skb:
3030 kfree_skb(skb);
3031 goto out;
3032}
3033EXPORT_SYMBOL(neigh_xmit);
3034
3035#ifdef CONFIG_PROC_FS
3036
3037static struct neighbour *neigh_get_first(struct seq_file *seq)
3038{
3039 struct neigh_seq_state *state = seq->private;
3040 struct net *net = seq_file_net(seq);
3041 struct neigh_hash_table *nht = state->nht;
3042 struct neighbour *n = NULL;
3043 int bucket;
3044
3045 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
3046 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
3047 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
3048
3049 while (n) {
3050 if (!net_eq(dev_net(n->dev), net))
3051 goto next;
3052 if (state->neigh_sub_iter) {
3053 loff_t fakep = 0;
3054 void *v;
3055
3056 v = state->neigh_sub_iter(state, n, &fakep);
3057 if (!v)
3058 goto next;
3059 }
3060 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
3061 break;
3062 if (n->nud_state & ~NUD_NOARP)
3063 break;
3064next:
3065 n = rcu_dereference_bh(n->next);
3066 }
3067
3068 if (n)
3069 break;
3070 }
3071 state->bucket = bucket;
3072
3073 return n;
3074}
3075
3076static struct neighbour *neigh_get_next(struct seq_file *seq,
3077 struct neighbour *n,
3078 loff_t *pos)
3079{
3080 struct neigh_seq_state *state = seq->private;
3081 struct net *net = seq_file_net(seq);
3082 struct neigh_hash_table *nht = state->nht;
3083
3084 if (state->neigh_sub_iter) {
3085 void *v = state->neigh_sub_iter(state, n, pos);
3086 if (v)
3087 return n;
3088 }
3089 n = rcu_dereference_bh(n->next);
3090
3091 while (1) {
3092 while (n) {
3093 if (!net_eq(dev_net(n->dev), net))
3094 goto next;
3095 if (state->neigh_sub_iter) {
3096 void *v = state->neigh_sub_iter(state, n, pos);
3097 if (v)
3098 return n;
3099 goto next;
3100 }
3101 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
3102 break;
3103
3104 if (n->nud_state & ~NUD_NOARP)
3105 break;
3106next:
3107 n = rcu_dereference_bh(n->next);
3108 }
3109
3110 if (n)
3111 break;
3112
3113 if (++state->bucket >= (1 << nht->hash_shift))
3114 break;
3115
3116 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
3117 }
3118
3119 if (n && pos)
3120 --(*pos);
3121 return n;
3122}
3123
3124static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
3125{
3126 struct neighbour *n = neigh_get_first(seq);
3127
3128 if (n) {
3129 --(*pos);
3130 while (*pos) {
3131 n = neigh_get_next(seq, n, pos);
3132 if (!n)
3133 break;
3134 }
3135 }
3136 return *pos ? NULL : n;
3137}
3138
3139static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
3140{
3141 struct neigh_seq_state *state = seq->private;
3142 struct net *net = seq_file_net(seq);
3143 struct neigh_table *tbl = state->tbl;
3144 struct pneigh_entry *pn = NULL;
3145 int bucket;
3146
3147 state->flags |= NEIGH_SEQ_IS_PNEIGH;
3148 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
3149 pn = tbl->phash_buckets[bucket];
3150 while (pn && !net_eq(pneigh_net(pn), net))
3151 pn = pn->next;
3152 if (pn)
3153 break;
3154 }
3155 state->bucket = bucket;
3156
3157 return pn;
3158}
3159
3160static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
3161 struct pneigh_entry *pn,
3162 loff_t *pos)
3163{
3164 struct neigh_seq_state *state = seq->private;
3165 struct net *net = seq_file_net(seq);
3166 struct neigh_table *tbl = state->tbl;
3167
3168 do {
3169 pn = pn->next;
3170 } while (pn && !net_eq(pneigh_net(pn), net));
3171
3172 while (!pn) {
3173 if (++state->bucket > PNEIGH_HASHMASK)
3174 break;
3175 pn = tbl->phash_buckets[state->bucket];
3176 while (pn && !net_eq(pneigh_net(pn), net))
3177 pn = pn->next;
3178 if (pn)
3179 break;
3180 }
3181
3182 if (pn && pos)
3183 --(*pos);
3184
3185 return pn;
3186}
3187
3188static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
3189{
3190 struct pneigh_entry *pn = pneigh_get_first(seq);
3191
3192 if (pn) {
3193 --(*pos);
3194 while (*pos) {
3195 pn = pneigh_get_next(seq, pn, pos);
3196 if (!pn)
3197 break;
3198 }
3199 }
3200 return *pos ? NULL : pn;
3201}
3202
3203static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
3204{
3205 struct neigh_seq_state *state = seq->private;
3206 void *rc;
3207 loff_t idxpos = *pos;
3208
3209 rc = neigh_get_idx(seq, &idxpos);
3210 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
3211 rc = pneigh_get_idx(seq, &idxpos);
3212
3213 return rc;
3214}
3215
3216void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
3217 __acquires(tbl->lock)
3218 __acquires(rcu_bh)
3219{
3220 struct neigh_seq_state *state = seq->private;
3221
3222 state->tbl = tbl;
3223 state->bucket = 0;
3224 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
3225
3226 rcu_read_lock_bh();
3227 state->nht = rcu_dereference_bh(tbl->nht);
3228 read_lock(&tbl->lock);
3229
3230 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
3231}
3232EXPORT_SYMBOL(neigh_seq_start);
3233
3234void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3235{
3236 struct neigh_seq_state *state;
3237 void *rc;
3238
3239 if (v == SEQ_START_TOKEN) {
3240 rc = neigh_get_first(seq);
3241 goto out;
3242 }
3243
3244 state = seq->private;
3245 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
3246 rc = neigh_get_next(seq, v, NULL);
3247 if (rc)
3248 goto out;
3249 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
3250 rc = pneigh_get_first(seq);
3251 } else {
3252 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
3253 rc = pneigh_get_next(seq, v, NULL);
3254 }
3255out:
3256 ++(*pos);
3257 return rc;
3258}
3259EXPORT_SYMBOL(neigh_seq_next);
3260
3261void neigh_seq_stop(struct seq_file *seq, void *v)
3262 __releases(tbl->lock)
3263 __releases(rcu_bh)
3264{
3265 struct neigh_seq_state *state = seq->private;
3266 struct neigh_table *tbl = state->tbl;
3267
3268 read_unlock(&tbl->lock);
3269 rcu_read_unlock_bh();
3270}
3271EXPORT_SYMBOL(neigh_seq_stop);
3272
3273/* statistics via seq_file */
3274
3275static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
3276{
3277 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3278 int cpu;
3279
3280 if (*pos == 0)
3281 return SEQ_START_TOKEN;
3282
3283 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
3284 if (!cpu_possible(cpu))
3285 continue;
3286 *pos = cpu+1;
3287 return per_cpu_ptr(tbl->stats, cpu);
3288 }
3289 return NULL;
3290}
3291
3292static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3293{
3294 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3295 int cpu;
3296
3297 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
3298 if (!cpu_possible(cpu))
3299 continue;
3300 *pos = cpu+1;
3301 return per_cpu_ptr(tbl->stats, cpu);
3302 }
3303 (*pos)++;
3304 return NULL;
3305}
3306
3307static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
3308{
3309
3310}
3311
3312static int neigh_stat_seq_show(struct seq_file *seq, void *v)
3313{
3314 struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3315 struct neigh_statistics *st = v;
3316
3317 if (v == SEQ_START_TOKEN) {
3318 seq_printf(seq, "entries allocs destroys hash_grows lookups hits res_failed rcv_probes_mcast rcv_probes_ucast periodic_gc_runs forced_gc_runs unresolved_discards table_fulls\n");
3319 return 0;
3320 }
3321
3322 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
3323 "%08lx %08lx %08lx %08lx %08lx %08lx\n",
3324 atomic_read(&tbl->entries),
3325
3326 st->allocs,
3327 st->destroys,
3328 st->hash_grows,
3329
3330 st->lookups,
3331 st->hits,
3332
3333 st->res_failed,
3334
3335 st->rcv_probes_mcast,
3336 st->rcv_probes_ucast,
3337
3338 st->periodic_gc_runs,
3339 st->forced_gc_runs,
3340 st->unres_discards,
3341 st->table_fulls
3342 );
3343
3344 return 0;
3345}
3346
3347static const struct seq_operations neigh_stat_seq_ops = {
3348 .start = neigh_stat_seq_start,
3349 .next = neigh_stat_seq_next,
3350 .stop = neigh_stat_seq_stop,
3351 .show = neigh_stat_seq_show,
3352};
3353#endif /* CONFIG_PROC_FS */
3354
3355static void __neigh_notify(struct neighbour *n, int type, int flags,
3356 u32 pid)
3357{
3358 struct net *net = dev_net(n->dev);
3359 struct sk_buff *skb;
3360 int err = -ENOBUFS;
3361
3362 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
3363 if (skb == NULL)
3364 goto errout;
3365
3366 err = neigh_fill_info(skb, n, pid, 0, type, flags);
3367 if (err < 0) {
3368 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
3369 WARN_ON(err == -EMSGSIZE);
3370 kfree_skb(skb);
3371 goto errout;
3372 }
3373 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3374 return;
3375errout:
3376 if (err < 0)
3377 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3378}
3379
3380void neigh_app_ns(struct neighbour *n)
3381{
3382 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST, 0);
3383}
3384EXPORT_SYMBOL(neigh_app_ns);
3385
3386#ifdef CONFIG_SYSCTL
3387static int unres_qlen_max = INT_MAX / SKB_TRUESIZE(ETH_FRAME_LEN);
3388
3389static int proc_unres_qlen(struct ctl_table *ctl, int write,
3390 void *buffer, size_t *lenp, loff_t *ppos)
3391{
3392 int size, ret;
3393 struct ctl_table tmp = *ctl;
3394
3395 tmp.extra1 = SYSCTL_ZERO;
3396 tmp.extra2 = &unres_qlen_max;
3397 tmp.data = &size;
3398
3399 size = *(int *)ctl->data / SKB_TRUESIZE(ETH_FRAME_LEN);
3400 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
3401
3402 if (write && !ret)
3403 *(int *)ctl->data = size * SKB_TRUESIZE(ETH_FRAME_LEN);
3404 return ret;
3405}
3406
3407static struct neigh_parms *neigh_get_dev_parms_rcu(struct net_device *dev,
3408 int family)
3409{
3410 switch (family) {
3411 case AF_INET:
3412 return __in_dev_arp_parms_get_rcu(dev);
3413 case AF_INET6:
3414 return __in6_dev_nd_parms_get_rcu(dev);
3415 }
3416 return NULL;
3417}
3418
3419static void neigh_copy_dflt_parms(struct net *net, struct neigh_parms *p,
3420 int index)
3421{
3422 struct net_device *dev;
3423 int family = neigh_parms_family(p);
3424
3425 rcu_read_lock();
3426 for_each_netdev_rcu(net, dev) {
3427 struct neigh_parms *dst_p =
3428 neigh_get_dev_parms_rcu(dev, family);
3429
3430 if (dst_p && !test_bit(index, dst_p->data_state))
3431 dst_p->data[index] = p->data[index];
3432 }
3433 rcu_read_unlock();
3434}
3435
3436static void neigh_proc_update(struct ctl_table *ctl, int write)
3437{
3438 struct net_device *dev = ctl->extra1;
3439 struct neigh_parms *p = ctl->extra2;
3440 struct net *net = neigh_parms_net(p);
3441 int index = (int *) ctl->data - p->data;
3442
3443 if (!write)
3444 return;
3445
3446 set_bit(index, p->data_state);
3447 if (index == NEIGH_VAR_DELAY_PROBE_TIME)
3448 call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
3449 if (!dev) /* NULL dev means this is default value */
3450 neigh_copy_dflt_parms(net, p, index);
3451}
3452
3453static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write,
3454 void *buffer, size_t *lenp,
3455 loff_t *ppos)
3456{
3457 struct ctl_table tmp = *ctl;
3458 int ret;
3459
3460 tmp.extra1 = SYSCTL_ZERO;
3461 tmp.extra2 = SYSCTL_INT_MAX;
3462
3463 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
3464 neigh_proc_update(ctl, write);
3465 return ret;
3466}
3467
3468int neigh_proc_dointvec(struct ctl_table *ctl, int write, void *buffer,
3469 size_t *lenp, loff_t *ppos)
3470{
3471 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
3472
3473 neigh_proc_update(ctl, write);
3474 return ret;
3475}
3476EXPORT_SYMBOL(neigh_proc_dointvec);
3477
3478int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write, void *buffer,
3479 size_t *lenp, loff_t *ppos)
3480{
3481 int ret = proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3482
3483 neigh_proc_update(ctl, write);
3484 return ret;
3485}
3486EXPORT_SYMBOL(neigh_proc_dointvec_jiffies);
3487
3488static int neigh_proc_dointvec_userhz_jiffies(struct ctl_table *ctl, int write,
3489 void *buffer, size_t *lenp,
3490 loff_t *ppos)
3491{
3492 int ret = proc_dointvec_userhz_jiffies(ctl, write, buffer, lenp, ppos);
3493
3494 neigh_proc_update(ctl, write);
3495 return ret;
3496}
3497
3498int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write,
3499 void *buffer, size_t *lenp, loff_t *ppos)
3500{
3501 int ret = proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3502
3503 neigh_proc_update(ctl, write);
3504 return ret;
3505}
3506EXPORT_SYMBOL(neigh_proc_dointvec_ms_jiffies);
3507
3508static int neigh_proc_dointvec_unres_qlen(struct ctl_table *ctl, int write,
3509 void *buffer, size_t *lenp,
3510 loff_t *ppos)
3511{
3512 int ret = proc_unres_qlen(ctl, write, buffer, lenp, ppos);
3513
3514 neigh_proc_update(ctl, write);
3515 return ret;
3516}
3517
3518static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,
3519 void *buffer, size_t *lenp,
3520 loff_t *ppos)
3521{
3522 struct neigh_parms *p = ctl->extra2;
3523 int ret;
3524
3525 if (strcmp(ctl->procname, "base_reachable_time") == 0)
3526 ret = neigh_proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3527 else if (strcmp(ctl->procname, "base_reachable_time_ms") == 0)
3528 ret = neigh_proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3529 else
3530 ret = -1;
3531
3532 if (write && ret == 0) {
3533 /* update reachable_time as well, otherwise, the change will
3534 * only be effective after the next time neigh_periodic_work
3535 * decides to recompute it
3536 */
3537 p->reachable_time =
3538 neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
3539 }
3540 return ret;
3541}
3542
3543#define NEIGH_PARMS_DATA_OFFSET(index) \
3544 (&((struct neigh_parms *) 0)->data[index])
3545
3546#define NEIGH_SYSCTL_ENTRY(attr, data_attr, name, mval, proc) \
3547 [NEIGH_VAR_ ## attr] = { \
3548 .procname = name, \
3549 .data = NEIGH_PARMS_DATA_OFFSET(NEIGH_VAR_ ## data_attr), \
3550 .maxlen = sizeof(int), \
3551 .mode = mval, \
3552 .proc_handler = proc, \
3553 }
3554
3555#define NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(attr, name) \
3556 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_zero_intmax)
3557
3558#define NEIGH_SYSCTL_JIFFIES_ENTRY(attr, name) \
3559 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_jiffies)
3560
3561#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(attr, name) \
3562 NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)
3563
3564#define NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(attr, data_attr, name) \
3565 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3566
3567#define NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(attr, data_attr, name) \
3568 NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_unres_qlen)
3569
3570static struct neigh_sysctl_table {
3571 struct ctl_table_header *sysctl_header;
3572 struct ctl_table neigh_vars[NEIGH_VAR_MAX + 1];
3573} neigh_sysctl_template __read_mostly = {
3574 .neigh_vars = {
3575 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_PROBES, "mcast_solicit"),
3576 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(UCAST_PROBES, "ucast_solicit"),
3577 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(APP_PROBES, "app_solicit"),
3578 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_REPROBES, "mcast_resolicit"),
3579 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(RETRANS_TIME, "retrans_time"),
3580 NEIGH_SYSCTL_JIFFIES_ENTRY(BASE_REACHABLE_TIME, "base_reachable_time"),
3581 NEIGH_SYSCTL_JIFFIES_ENTRY(DELAY_PROBE_TIME, "delay_first_probe_time"),
3582 NEIGH_SYSCTL_JIFFIES_ENTRY(GC_STALETIME, "gc_stale_time"),
3583 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(QUEUE_LEN_BYTES, "unres_qlen_bytes"),
3584 NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(PROXY_QLEN, "proxy_qlen"),
3585 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(ANYCAST_DELAY, "anycast_delay"),
3586 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(PROXY_DELAY, "proxy_delay"),
3587 NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(LOCKTIME, "locktime"),
3588 NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(QUEUE_LEN, QUEUE_LEN_BYTES, "unres_qlen"),
3589 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(RETRANS_TIME_MS, RETRANS_TIME, "retrans_time_ms"),
3590 NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(BASE_REACHABLE_TIME_MS, BASE_REACHABLE_TIME, "base_reachable_time_ms"),
3591 [NEIGH_VAR_GC_INTERVAL] = {
3592 .procname = "gc_interval",
3593 .maxlen = sizeof(int),
3594 .mode = 0644,
3595 .proc_handler = proc_dointvec_jiffies,
3596 },
3597 [NEIGH_VAR_GC_THRESH1] = {
3598 .procname = "gc_thresh1",
3599 .maxlen = sizeof(int),
3600 .mode = 0644,
3601 .extra1 = SYSCTL_ZERO,
3602 .extra2 = SYSCTL_INT_MAX,
3603 .proc_handler = proc_dointvec_minmax,
3604 },
3605 [NEIGH_VAR_GC_THRESH2] = {
3606 .procname = "gc_thresh2",
3607 .maxlen = sizeof(int),
3608 .mode = 0644,
3609 .extra1 = SYSCTL_ZERO,
3610 .extra2 = SYSCTL_INT_MAX,
3611 .proc_handler = proc_dointvec_minmax,
3612 },
3613 [NEIGH_VAR_GC_THRESH3] = {
3614 .procname = "gc_thresh3",
3615 .maxlen = sizeof(int),
3616 .mode = 0644,
3617 .extra1 = SYSCTL_ZERO,
3618 .extra2 = SYSCTL_INT_MAX,
3619 .proc_handler = proc_dointvec_minmax,
3620 },
3621 {},
3622 },
3623};
3624
3625int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
3626 proc_handler *handler)
3627{
3628 int i;
3629 struct neigh_sysctl_table *t;
3630 const char *dev_name_source;
3631 char neigh_path[ sizeof("net//neigh/") + IFNAMSIZ + IFNAMSIZ ];
3632 char *p_name;
3633
3634 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
3635 if (!t)
3636 goto err;
3637
3638 for (i = 0; i < NEIGH_VAR_GC_INTERVAL; i++) {
3639 t->neigh_vars[i].data += (long) p;
3640 t->neigh_vars[i].extra1 = dev;
3641 t->neigh_vars[i].extra2 = p;
3642 }
3643
3644 if (dev) {
3645 dev_name_source = dev->name;
3646 /* Terminate the table early */
3647 memset(&t->neigh_vars[NEIGH_VAR_GC_INTERVAL], 0,
3648 sizeof(t->neigh_vars[NEIGH_VAR_GC_INTERVAL]));
3649 } else {
3650 struct neigh_table *tbl = p->tbl;
3651 dev_name_source = "default";
3652 t->neigh_vars[NEIGH_VAR_GC_INTERVAL].data = &tbl->gc_interval;
3653 t->neigh_vars[NEIGH_VAR_GC_THRESH1].data = &tbl->gc_thresh1;
3654 t->neigh_vars[NEIGH_VAR_GC_THRESH2].data = &tbl->gc_thresh2;
3655 t->neigh_vars[NEIGH_VAR_GC_THRESH3].data = &tbl->gc_thresh3;
3656 }
3657
3658 if (handler) {
3659 /* RetransTime */
3660 t->neigh_vars[NEIGH_VAR_RETRANS_TIME].proc_handler = handler;
3661 /* ReachableTime */
3662 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler = handler;
3663 /* RetransTime (in milliseconds)*/
3664 t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].proc_handler = handler;
3665 /* ReachableTime (in milliseconds) */
3666 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler = handler;
3667 } else {
3668 /* Those handlers will update p->reachable_time after
3669 * base_reachable_time(_ms) is set to ensure the new timer starts being
3670 * applied after the next neighbour update instead of waiting for
3671 * neigh_periodic_work to update its value (can be multiple minutes)
3672 * So any handler that replaces them should do this as well
3673 */
3674 /* ReachableTime */
3675 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler =
3676 neigh_proc_base_reachable_time;
3677 /* ReachableTime (in milliseconds) */
3678 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler =
3679 neigh_proc_base_reachable_time;
3680 }
3681
3682 /* Don't export sysctls to unprivileged users */
3683 if (neigh_parms_net(p)->user_ns != &init_user_ns)
3684 t->neigh_vars[0].procname = NULL;
3685
3686 switch (neigh_parms_family(p)) {
3687 case AF_INET:
3688 p_name = "ipv4";
3689 break;
3690 case AF_INET6:
3691 p_name = "ipv6";
3692 break;
3693 default:
3694 BUG();
3695 }
3696
3697 snprintf(neigh_path, sizeof(neigh_path), "net/%s/neigh/%s",
3698 p_name, dev_name_source);
3699 t->sysctl_header =
3700 register_net_sysctl(neigh_parms_net(p), neigh_path, t->neigh_vars);
3701 if (!t->sysctl_header)
3702 goto free;
3703
3704 p->sysctl_table = t;
3705 return 0;
3706
3707free:
3708 kfree(t);
3709err:
3710 return -ENOBUFS;
3711}
3712EXPORT_SYMBOL(neigh_sysctl_register);
3713
3714void neigh_sysctl_unregister(struct neigh_parms *p)
3715{
3716 if (p->sysctl_table) {
3717 struct neigh_sysctl_table *t = p->sysctl_table;
3718 p->sysctl_table = NULL;
3719 unregister_net_sysctl_table(t->sysctl_header);
3720 kfree(t);
3721 }
3722}
3723EXPORT_SYMBOL(neigh_sysctl_unregister);
3724
3725#endif /* CONFIG_SYSCTL */
3726
3727static int __init neigh_init(void)
3728{
3729 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, 0);
3730 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, 0);
3731 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, neigh_get, neigh_dump_info, 0);
3732
3733 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
3734 0);
3735 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, 0);
3736
3737 return 0;
3738}
3739
3740subsys_initcall(neigh_init);
1/*
2 * Generic address resolution entity
3 *
4 * Authors:
5 * Pedro Roque <roque@di.fc.ul.pt>
6 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
7 *
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.
12 *
13 * Fixes:
14 * Vitaly E. Lavrov releasing NULL neighbor in neigh_add.
15 * Harald Welte Add neighbour cache statistics like rtstat
16 */
17
18#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
19
20#include <linux/slab.h>
21#include <linux/types.h>
22#include <linux/kernel.h>
23#include <linux/module.h>
24#include <linux/socket.h>
25#include <linux/netdevice.h>
26#include <linux/proc_fs.h>
27#ifdef CONFIG_SYSCTL
28#include <linux/sysctl.h>
29#endif
30#include <linux/times.h>
31#include <net/net_namespace.h>
32#include <net/neighbour.h>
33#include <net/dst.h>
34#include <net/sock.h>
35#include <net/netevent.h>
36#include <net/netlink.h>
37#include <linux/rtnetlink.h>
38#include <linux/random.h>
39#include <linux/string.h>
40#include <linux/log2.h>
41
42#define NEIGH_DEBUG 1
43
44#define NEIGH_PRINTK(x...) printk(x)
45#define NEIGH_NOPRINTK(x...) do { ; } while(0)
46#define NEIGH_PRINTK1 NEIGH_NOPRINTK
47#define NEIGH_PRINTK2 NEIGH_NOPRINTK
48
49#if NEIGH_DEBUG >= 1
50#undef NEIGH_PRINTK1
51#define NEIGH_PRINTK1 NEIGH_PRINTK
52#endif
53#if NEIGH_DEBUG >= 2
54#undef NEIGH_PRINTK2
55#define NEIGH_PRINTK2 NEIGH_PRINTK
56#endif
57
58#define PNEIGH_HASHMASK 0xF
59
60static void neigh_timer_handler(unsigned long arg);
61static void __neigh_notify(struct neighbour *n, int type, int flags);
62static void neigh_update_notify(struct neighbour *neigh);
63static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
64
65static struct neigh_table *neigh_tables;
66#ifdef CONFIG_PROC_FS
67static const struct file_operations neigh_stat_seq_fops;
68#endif
69
70/*
71 Neighbour hash table buckets are protected with rwlock tbl->lock.
72
73 - All the scans/updates to hash buckets MUST be made under this lock.
74 - NOTHING clever should be made under this lock: no callbacks
75 to protocol backends, no attempts to send something to network.
76 It will result in deadlocks, if backend/driver wants to use neighbour
77 cache.
78 - If the entry requires some non-trivial actions, increase
79 its reference count and release table lock.
80
81 Neighbour entries are protected:
82 - with reference count.
83 - with rwlock neigh->lock
84
85 Reference count prevents destruction.
86
87 neigh->lock mainly serializes ll address data and its validity state.
88 However, the same lock is used to protect another entry fields:
89 - timer
90 - resolution queue
91
92 Again, nothing clever shall be made under neigh->lock,
93 the most complicated procedure, which we allow is dev->hard_header.
94 It is supposed, that dev->hard_header is simplistic and does
95 not make callbacks to neighbour tables.
96
97 The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
98 list of neighbour tables. This list is used only in process context,
99 */
100
101static DEFINE_RWLOCK(neigh_tbl_lock);
102
103static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
104{
105 kfree_skb(skb);
106 return -ENETDOWN;
107}
108
109static void neigh_cleanup_and_release(struct neighbour *neigh)
110{
111 if (neigh->parms->neigh_cleanup)
112 neigh->parms->neigh_cleanup(neigh);
113
114 __neigh_notify(neigh, RTM_DELNEIGH, 0);
115 neigh_release(neigh);
116}
117
118/*
119 * It is random distribution in the interval (1/2)*base...(3/2)*base.
120 * It corresponds to default IPv6 settings and is not overridable,
121 * because it is really reasonable choice.
122 */
123
124unsigned long neigh_rand_reach_time(unsigned long base)
125{
126 return base ? (net_random() % base) + (base >> 1) : 0;
127}
128EXPORT_SYMBOL(neigh_rand_reach_time);
129
130
131static int neigh_forced_gc(struct neigh_table *tbl)
132{
133 int shrunk = 0;
134 int i;
135 struct neigh_hash_table *nht;
136
137 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
138
139 write_lock_bh(&tbl->lock);
140 nht = rcu_dereference_protected(tbl->nht,
141 lockdep_is_held(&tbl->lock));
142 for (i = 0; i < (1 << nht->hash_shift); i++) {
143 struct neighbour *n;
144 struct neighbour __rcu **np;
145
146 np = &nht->hash_buckets[i];
147 while ((n = rcu_dereference_protected(*np,
148 lockdep_is_held(&tbl->lock))) != NULL) {
149 /* Neighbour record may be discarded if:
150 * - nobody refers to it.
151 * - it is not permanent
152 */
153 write_lock(&n->lock);
154 if (atomic_read(&n->refcnt) == 1 &&
155 !(n->nud_state & NUD_PERMANENT)) {
156 rcu_assign_pointer(*np,
157 rcu_dereference_protected(n->next,
158 lockdep_is_held(&tbl->lock)));
159 n->dead = 1;
160 shrunk = 1;
161 write_unlock(&n->lock);
162 neigh_cleanup_and_release(n);
163 continue;
164 }
165 write_unlock(&n->lock);
166 np = &n->next;
167 }
168 }
169
170 tbl->last_flush = jiffies;
171
172 write_unlock_bh(&tbl->lock);
173
174 return shrunk;
175}
176
177static void neigh_add_timer(struct neighbour *n, unsigned long when)
178{
179 neigh_hold(n);
180 if (unlikely(mod_timer(&n->timer, when))) {
181 printk("NEIGH: BUG, double timer add, state is %x\n",
182 n->nud_state);
183 dump_stack();
184 }
185}
186
187static int neigh_del_timer(struct neighbour *n)
188{
189 if ((n->nud_state & NUD_IN_TIMER) &&
190 del_timer(&n->timer)) {
191 neigh_release(n);
192 return 1;
193 }
194 return 0;
195}
196
197static void pneigh_queue_purge(struct sk_buff_head *list)
198{
199 struct sk_buff *skb;
200
201 while ((skb = skb_dequeue(list)) != NULL) {
202 dev_put(skb->dev);
203 kfree_skb(skb);
204 }
205}
206
207static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
208{
209 int i;
210 struct neigh_hash_table *nht;
211
212 nht = rcu_dereference_protected(tbl->nht,
213 lockdep_is_held(&tbl->lock));
214
215 for (i = 0; i < (1 << nht->hash_shift); i++) {
216 struct neighbour *n;
217 struct neighbour __rcu **np = &nht->hash_buckets[i];
218
219 while ((n = rcu_dereference_protected(*np,
220 lockdep_is_held(&tbl->lock))) != NULL) {
221 if (dev && n->dev != dev) {
222 np = &n->next;
223 continue;
224 }
225 rcu_assign_pointer(*np,
226 rcu_dereference_protected(n->next,
227 lockdep_is_held(&tbl->lock)));
228 write_lock(&n->lock);
229 neigh_del_timer(n);
230 n->dead = 1;
231
232 if (atomic_read(&n->refcnt) != 1) {
233 /* The most unpleasant situation.
234 We must destroy neighbour entry,
235 but someone still uses it.
236
237 The destroy will be delayed until
238 the last user releases us, but
239 we must kill timers etc. and move
240 it to safe state.
241 */
242 skb_queue_purge(&n->arp_queue);
243 n->arp_queue_len_bytes = 0;
244 n->output = neigh_blackhole;
245 if (n->nud_state & NUD_VALID)
246 n->nud_state = NUD_NOARP;
247 else
248 n->nud_state = NUD_NONE;
249 NEIGH_PRINTK2("neigh %p is stray.\n", n);
250 }
251 write_unlock(&n->lock);
252 neigh_cleanup_and_release(n);
253 }
254 }
255}
256
257void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
258{
259 write_lock_bh(&tbl->lock);
260 neigh_flush_dev(tbl, dev);
261 write_unlock_bh(&tbl->lock);
262}
263EXPORT_SYMBOL(neigh_changeaddr);
264
265int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
266{
267 write_lock_bh(&tbl->lock);
268 neigh_flush_dev(tbl, dev);
269 pneigh_ifdown(tbl, dev);
270 write_unlock_bh(&tbl->lock);
271
272 del_timer_sync(&tbl->proxy_timer);
273 pneigh_queue_purge(&tbl->proxy_queue);
274 return 0;
275}
276EXPORT_SYMBOL(neigh_ifdown);
277
278static struct neighbour *neigh_alloc(struct neigh_table *tbl, struct net_device *dev)
279{
280 struct neighbour *n = NULL;
281 unsigned long now = jiffies;
282 int entries;
283
284 entries = atomic_inc_return(&tbl->entries) - 1;
285 if (entries >= tbl->gc_thresh3 ||
286 (entries >= tbl->gc_thresh2 &&
287 time_after(now, tbl->last_flush + 5 * HZ))) {
288 if (!neigh_forced_gc(tbl) &&
289 entries >= tbl->gc_thresh3)
290 goto out_entries;
291 }
292
293 if (tbl->entry_size)
294 n = kzalloc(tbl->entry_size, GFP_ATOMIC);
295 else {
296 int sz = sizeof(*n) + tbl->key_len;
297
298 sz = ALIGN(sz, NEIGH_PRIV_ALIGN);
299 sz += dev->neigh_priv_len;
300 n = kzalloc(sz, GFP_ATOMIC);
301 }
302 if (!n)
303 goto out_entries;
304
305 skb_queue_head_init(&n->arp_queue);
306 rwlock_init(&n->lock);
307 seqlock_init(&n->ha_lock);
308 n->updated = n->used = now;
309 n->nud_state = NUD_NONE;
310 n->output = neigh_blackhole;
311 seqlock_init(&n->hh.hh_lock);
312 n->parms = neigh_parms_clone(&tbl->parms);
313 setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
314
315 NEIGH_CACHE_STAT_INC(tbl, allocs);
316 n->tbl = tbl;
317 atomic_set(&n->refcnt, 1);
318 n->dead = 1;
319out:
320 return n;
321
322out_entries:
323 atomic_dec(&tbl->entries);
324 goto out;
325}
326
327static void neigh_get_hash_rnd(u32 *x)
328{
329 get_random_bytes(x, sizeof(*x));
330 *x |= 1;
331}
332
333static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
334{
335 size_t size = (1 << shift) * sizeof(struct neighbour *);
336 struct neigh_hash_table *ret;
337 struct neighbour __rcu **buckets;
338 int i;
339
340 ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
341 if (!ret)
342 return NULL;
343 if (size <= PAGE_SIZE)
344 buckets = kzalloc(size, GFP_ATOMIC);
345 else
346 buckets = (struct neighbour __rcu **)
347 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
348 get_order(size));
349 if (!buckets) {
350 kfree(ret);
351 return NULL;
352 }
353 ret->hash_buckets = buckets;
354 ret->hash_shift = shift;
355 for (i = 0; i < NEIGH_NUM_HASH_RND; i++)
356 neigh_get_hash_rnd(&ret->hash_rnd[i]);
357 return ret;
358}
359
360static void neigh_hash_free_rcu(struct rcu_head *head)
361{
362 struct neigh_hash_table *nht = container_of(head,
363 struct neigh_hash_table,
364 rcu);
365 size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
366 struct neighbour __rcu **buckets = nht->hash_buckets;
367
368 if (size <= PAGE_SIZE)
369 kfree(buckets);
370 else
371 free_pages((unsigned long)buckets, get_order(size));
372 kfree(nht);
373}
374
375static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
376 unsigned long new_shift)
377{
378 unsigned int i, hash;
379 struct neigh_hash_table *new_nht, *old_nht;
380
381 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
382
383 old_nht = rcu_dereference_protected(tbl->nht,
384 lockdep_is_held(&tbl->lock));
385 new_nht = neigh_hash_alloc(new_shift);
386 if (!new_nht)
387 return old_nht;
388
389 for (i = 0; i < (1 << old_nht->hash_shift); i++) {
390 struct neighbour *n, *next;
391
392 for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
393 lockdep_is_held(&tbl->lock));
394 n != NULL;
395 n = next) {
396 hash = tbl->hash(n->primary_key, n->dev,
397 new_nht->hash_rnd);
398
399 hash >>= (32 - new_nht->hash_shift);
400 next = rcu_dereference_protected(n->next,
401 lockdep_is_held(&tbl->lock));
402
403 rcu_assign_pointer(n->next,
404 rcu_dereference_protected(
405 new_nht->hash_buckets[hash],
406 lockdep_is_held(&tbl->lock)));
407 rcu_assign_pointer(new_nht->hash_buckets[hash], n);
408 }
409 }
410
411 rcu_assign_pointer(tbl->nht, new_nht);
412 call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
413 return new_nht;
414}
415
416struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
417 struct net_device *dev)
418{
419 struct neighbour *n;
420 int key_len = tbl->key_len;
421 u32 hash_val;
422 struct neigh_hash_table *nht;
423
424 NEIGH_CACHE_STAT_INC(tbl, lookups);
425
426 rcu_read_lock_bh();
427 nht = rcu_dereference_bh(tbl->nht);
428 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
429
430 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
431 n != NULL;
432 n = rcu_dereference_bh(n->next)) {
433 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
434 if (!atomic_inc_not_zero(&n->refcnt))
435 n = NULL;
436 NEIGH_CACHE_STAT_INC(tbl, hits);
437 break;
438 }
439 }
440
441 rcu_read_unlock_bh();
442 return n;
443}
444EXPORT_SYMBOL(neigh_lookup);
445
446struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
447 const void *pkey)
448{
449 struct neighbour *n;
450 int key_len = tbl->key_len;
451 u32 hash_val;
452 struct neigh_hash_table *nht;
453
454 NEIGH_CACHE_STAT_INC(tbl, lookups);
455
456 rcu_read_lock_bh();
457 nht = rcu_dereference_bh(tbl->nht);
458 hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
459
460 for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
461 n != NULL;
462 n = rcu_dereference_bh(n->next)) {
463 if (!memcmp(n->primary_key, pkey, key_len) &&
464 net_eq(dev_net(n->dev), net)) {
465 if (!atomic_inc_not_zero(&n->refcnt))
466 n = NULL;
467 NEIGH_CACHE_STAT_INC(tbl, hits);
468 break;
469 }
470 }
471
472 rcu_read_unlock_bh();
473 return n;
474}
475EXPORT_SYMBOL(neigh_lookup_nodev);
476
477struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
478 struct net_device *dev)
479{
480 u32 hash_val;
481 int key_len = tbl->key_len;
482 int error;
483 struct neighbour *n1, *rc, *n = neigh_alloc(tbl, dev);
484 struct neigh_hash_table *nht;
485
486 if (!n) {
487 rc = ERR_PTR(-ENOBUFS);
488 goto out;
489 }
490
491 memcpy(n->primary_key, pkey, key_len);
492 n->dev = dev;
493 dev_hold(dev);
494
495 /* Protocol specific setup. */
496 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
497 rc = ERR_PTR(error);
498 goto out_neigh_release;
499 }
500
501 if (dev->netdev_ops->ndo_neigh_construct) {
502 error = dev->netdev_ops->ndo_neigh_construct(n);
503 if (error < 0) {
504 rc = ERR_PTR(error);
505 goto out_neigh_release;
506 }
507 }
508
509 /* Device specific setup. */
510 if (n->parms->neigh_setup &&
511 (error = n->parms->neigh_setup(n)) < 0) {
512 rc = ERR_PTR(error);
513 goto out_neigh_release;
514 }
515
516 n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
517
518 write_lock_bh(&tbl->lock);
519 nht = rcu_dereference_protected(tbl->nht,
520 lockdep_is_held(&tbl->lock));
521
522 if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
523 nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
524
525 hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
526
527 if (n->parms->dead) {
528 rc = ERR_PTR(-EINVAL);
529 goto out_tbl_unlock;
530 }
531
532 for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
533 lockdep_is_held(&tbl->lock));
534 n1 != NULL;
535 n1 = rcu_dereference_protected(n1->next,
536 lockdep_is_held(&tbl->lock))) {
537 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
538 neigh_hold(n1);
539 rc = n1;
540 goto out_tbl_unlock;
541 }
542 }
543
544 n->dead = 0;
545 neigh_hold(n);
546 rcu_assign_pointer(n->next,
547 rcu_dereference_protected(nht->hash_buckets[hash_val],
548 lockdep_is_held(&tbl->lock)));
549 rcu_assign_pointer(nht->hash_buckets[hash_val], n);
550 write_unlock_bh(&tbl->lock);
551 NEIGH_PRINTK2("neigh %p is created.\n", n);
552 rc = n;
553out:
554 return rc;
555out_tbl_unlock:
556 write_unlock_bh(&tbl->lock);
557out_neigh_release:
558 neigh_release(n);
559 goto out;
560}
561EXPORT_SYMBOL(neigh_create);
562
563static u32 pneigh_hash(const void *pkey, int key_len)
564{
565 u32 hash_val = *(u32 *)(pkey + key_len - 4);
566 hash_val ^= (hash_val >> 16);
567 hash_val ^= hash_val >> 8;
568 hash_val ^= hash_val >> 4;
569 hash_val &= PNEIGH_HASHMASK;
570 return hash_val;
571}
572
573static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
574 struct net *net,
575 const void *pkey,
576 int key_len,
577 struct net_device *dev)
578{
579 while (n) {
580 if (!memcmp(n->key, pkey, key_len) &&
581 net_eq(pneigh_net(n), net) &&
582 (n->dev == dev || !n->dev))
583 return n;
584 n = n->next;
585 }
586 return NULL;
587}
588
589struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
590 struct net *net, const void *pkey, struct net_device *dev)
591{
592 int key_len = tbl->key_len;
593 u32 hash_val = pneigh_hash(pkey, key_len);
594
595 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
596 net, pkey, key_len, dev);
597}
598EXPORT_SYMBOL_GPL(__pneigh_lookup);
599
600struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
601 struct net *net, const void *pkey,
602 struct net_device *dev, int creat)
603{
604 struct pneigh_entry *n;
605 int key_len = tbl->key_len;
606 u32 hash_val = pneigh_hash(pkey, key_len);
607
608 read_lock_bh(&tbl->lock);
609 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
610 net, pkey, key_len, dev);
611 read_unlock_bh(&tbl->lock);
612
613 if (n || !creat)
614 goto out;
615
616 ASSERT_RTNL();
617
618 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
619 if (!n)
620 goto out;
621
622 write_pnet(&n->net, hold_net(net));
623 memcpy(n->key, pkey, key_len);
624 n->dev = dev;
625 if (dev)
626 dev_hold(dev);
627
628 if (tbl->pconstructor && tbl->pconstructor(n)) {
629 if (dev)
630 dev_put(dev);
631 release_net(net);
632 kfree(n);
633 n = NULL;
634 goto out;
635 }
636
637 write_lock_bh(&tbl->lock);
638 n->next = tbl->phash_buckets[hash_val];
639 tbl->phash_buckets[hash_val] = n;
640 write_unlock_bh(&tbl->lock);
641out:
642 return n;
643}
644EXPORT_SYMBOL(pneigh_lookup);
645
646
647int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
648 struct net_device *dev)
649{
650 struct pneigh_entry *n, **np;
651 int key_len = tbl->key_len;
652 u32 hash_val = pneigh_hash(pkey, key_len);
653
654 write_lock_bh(&tbl->lock);
655 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
656 np = &n->next) {
657 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
658 net_eq(pneigh_net(n), net)) {
659 *np = n->next;
660 write_unlock_bh(&tbl->lock);
661 if (tbl->pdestructor)
662 tbl->pdestructor(n);
663 if (n->dev)
664 dev_put(n->dev);
665 release_net(pneigh_net(n));
666 kfree(n);
667 return 0;
668 }
669 }
670 write_unlock_bh(&tbl->lock);
671 return -ENOENT;
672}
673
674static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
675{
676 struct pneigh_entry *n, **np;
677 u32 h;
678
679 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
680 np = &tbl->phash_buckets[h];
681 while ((n = *np) != NULL) {
682 if (!dev || n->dev == dev) {
683 *np = n->next;
684 if (tbl->pdestructor)
685 tbl->pdestructor(n);
686 if (n->dev)
687 dev_put(n->dev);
688 release_net(pneigh_net(n));
689 kfree(n);
690 continue;
691 }
692 np = &n->next;
693 }
694 }
695 return -ENOENT;
696}
697
698static void neigh_parms_destroy(struct neigh_parms *parms);
699
700static inline void neigh_parms_put(struct neigh_parms *parms)
701{
702 if (atomic_dec_and_test(&parms->refcnt))
703 neigh_parms_destroy(parms);
704}
705
706/*
707 * neighbour must already be out of the table;
708 *
709 */
710void neigh_destroy(struct neighbour *neigh)
711{
712 struct net_device *dev = neigh->dev;
713
714 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
715
716 if (!neigh->dead) {
717 pr_warn("Destroying alive neighbour %p\n", neigh);
718 dump_stack();
719 return;
720 }
721
722 if (neigh_del_timer(neigh))
723 pr_warn("Impossible event\n");
724
725 skb_queue_purge(&neigh->arp_queue);
726 neigh->arp_queue_len_bytes = 0;
727
728 if (dev->netdev_ops->ndo_neigh_destroy)
729 dev->netdev_ops->ndo_neigh_destroy(neigh);
730
731 dev_put(dev);
732 neigh_parms_put(neigh->parms);
733
734 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
735
736 atomic_dec(&neigh->tbl->entries);
737 kfree_rcu(neigh, rcu);
738}
739EXPORT_SYMBOL(neigh_destroy);
740
741/* Neighbour state is suspicious;
742 disable fast path.
743
744 Called with write_locked neigh.
745 */
746static void neigh_suspect(struct neighbour *neigh)
747{
748 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
749
750 neigh->output = neigh->ops->output;
751}
752
753/* Neighbour state is OK;
754 enable fast path.
755
756 Called with write_locked neigh.
757 */
758static void neigh_connect(struct neighbour *neigh)
759{
760 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
761
762 neigh->output = neigh->ops->connected_output;
763}
764
765static void neigh_periodic_work(struct work_struct *work)
766{
767 struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
768 struct neighbour *n;
769 struct neighbour __rcu **np;
770 unsigned int i;
771 struct neigh_hash_table *nht;
772
773 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
774
775 write_lock_bh(&tbl->lock);
776 nht = rcu_dereference_protected(tbl->nht,
777 lockdep_is_held(&tbl->lock));
778
779 /*
780 * periodically recompute ReachableTime from random function
781 */
782
783 if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
784 struct neigh_parms *p;
785 tbl->last_rand = jiffies;
786 for (p = &tbl->parms; p; p = p->next)
787 p->reachable_time =
788 neigh_rand_reach_time(p->base_reachable_time);
789 }
790
791 for (i = 0 ; i < (1 << nht->hash_shift); i++) {
792 np = &nht->hash_buckets[i];
793
794 while ((n = rcu_dereference_protected(*np,
795 lockdep_is_held(&tbl->lock))) != NULL) {
796 unsigned int state;
797
798 write_lock(&n->lock);
799
800 state = n->nud_state;
801 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
802 write_unlock(&n->lock);
803 goto next_elt;
804 }
805
806 if (time_before(n->used, n->confirmed))
807 n->used = n->confirmed;
808
809 if (atomic_read(&n->refcnt) == 1 &&
810 (state == NUD_FAILED ||
811 time_after(jiffies, n->used + n->parms->gc_staletime))) {
812 *np = n->next;
813 n->dead = 1;
814 write_unlock(&n->lock);
815 neigh_cleanup_and_release(n);
816 continue;
817 }
818 write_unlock(&n->lock);
819
820next_elt:
821 np = &n->next;
822 }
823 /*
824 * It's fine to release lock here, even if hash table
825 * grows while we are preempted.
826 */
827 write_unlock_bh(&tbl->lock);
828 cond_resched();
829 write_lock_bh(&tbl->lock);
830 nht = rcu_dereference_protected(tbl->nht,
831 lockdep_is_held(&tbl->lock));
832 }
833 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
834 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
835 * base_reachable_time.
836 */
837 schedule_delayed_work(&tbl->gc_work,
838 tbl->parms.base_reachable_time >> 1);
839 write_unlock_bh(&tbl->lock);
840}
841
842static __inline__ int neigh_max_probes(struct neighbour *n)
843{
844 struct neigh_parms *p = n->parms;
845 return (n->nud_state & NUD_PROBE) ?
846 p->ucast_probes :
847 p->ucast_probes + p->app_probes + p->mcast_probes;
848}
849
850static void neigh_invalidate(struct neighbour *neigh)
851 __releases(neigh->lock)
852 __acquires(neigh->lock)
853{
854 struct sk_buff *skb;
855
856 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
857 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
858 neigh->updated = jiffies;
859
860 /* It is very thin place. report_unreachable is very complicated
861 routine. Particularly, it can hit the same neighbour entry!
862
863 So that, we try to be accurate and avoid dead loop. --ANK
864 */
865 while (neigh->nud_state == NUD_FAILED &&
866 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
867 write_unlock(&neigh->lock);
868 neigh->ops->error_report(neigh, skb);
869 write_lock(&neigh->lock);
870 }
871 skb_queue_purge(&neigh->arp_queue);
872 neigh->arp_queue_len_bytes = 0;
873}
874
875static void neigh_probe(struct neighbour *neigh)
876 __releases(neigh->lock)
877{
878 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
879 /* keep skb alive even if arp_queue overflows */
880 if (skb)
881 skb = skb_copy(skb, GFP_ATOMIC);
882 write_unlock(&neigh->lock);
883 neigh->ops->solicit(neigh, skb);
884 atomic_inc(&neigh->probes);
885 kfree_skb(skb);
886}
887
888/* Called when a timer expires for a neighbour entry. */
889
890static void neigh_timer_handler(unsigned long arg)
891{
892 unsigned long now, next;
893 struct neighbour *neigh = (struct neighbour *)arg;
894 unsigned int state;
895 int notify = 0;
896
897 write_lock(&neigh->lock);
898
899 state = neigh->nud_state;
900 now = jiffies;
901 next = now + HZ;
902
903 if (!(state & NUD_IN_TIMER))
904 goto out;
905
906 if (state & NUD_REACHABLE) {
907 if (time_before_eq(now,
908 neigh->confirmed + neigh->parms->reachable_time)) {
909 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
910 next = neigh->confirmed + neigh->parms->reachable_time;
911 } else if (time_before_eq(now,
912 neigh->used + neigh->parms->delay_probe_time)) {
913 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
914 neigh->nud_state = NUD_DELAY;
915 neigh->updated = jiffies;
916 neigh_suspect(neigh);
917 next = now + neigh->parms->delay_probe_time;
918 } else {
919 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
920 neigh->nud_state = NUD_STALE;
921 neigh->updated = jiffies;
922 neigh_suspect(neigh);
923 notify = 1;
924 }
925 } else if (state & NUD_DELAY) {
926 if (time_before_eq(now,
927 neigh->confirmed + neigh->parms->delay_probe_time)) {
928 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
929 neigh->nud_state = NUD_REACHABLE;
930 neigh->updated = jiffies;
931 neigh_connect(neigh);
932 notify = 1;
933 next = neigh->confirmed + neigh->parms->reachable_time;
934 } else {
935 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
936 neigh->nud_state = NUD_PROBE;
937 neigh->updated = jiffies;
938 atomic_set(&neigh->probes, 0);
939 next = now + neigh->parms->retrans_time;
940 }
941 } else {
942 /* NUD_PROBE|NUD_INCOMPLETE */
943 next = now + neigh->parms->retrans_time;
944 }
945
946 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
947 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
948 neigh->nud_state = NUD_FAILED;
949 notify = 1;
950 neigh_invalidate(neigh);
951 }
952
953 if (neigh->nud_state & NUD_IN_TIMER) {
954 if (time_before(next, jiffies + HZ/2))
955 next = jiffies + HZ/2;
956 if (!mod_timer(&neigh->timer, next))
957 neigh_hold(neigh);
958 }
959 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
960 neigh_probe(neigh);
961 } else {
962out:
963 write_unlock(&neigh->lock);
964 }
965
966 if (notify)
967 neigh_update_notify(neigh);
968
969 neigh_release(neigh);
970}
971
972int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
973{
974 int rc;
975 bool immediate_probe = false;
976
977 write_lock_bh(&neigh->lock);
978
979 rc = 0;
980 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
981 goto out_unlock_bh;
982
983 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
984 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
985 unsigned long next, now = jiffies;
986
987 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
988 neigh->nud_state = NUD_INCOMPLETE;
989 neigh->updated = now;
990 next = now + max(neigh->parms->retrans_time, HZ/2);
991 neigh_add_timer(neigh, next);
992 immediate_probe = true;
993 } else {
994 neigh->nud_state = NUD_FAILED;
995 neigh->updated = jiffies;
996 write_unlock_bh(&neigh->lock);
997
998 kfree_skb(skb);
999 return 1;
1000 }
1001 } else if (neigh->nud_state & NUD_STALE) {
1002 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
1003 neigh->nud_state = NUD_DELAY;
1004 neigh->updated = jiffies;
1005 neigh_add_timer(neigh,
1006 jiffies + neigh->parms->delay_probe_time);
1007 }
1008
1009 if (neigh->nud_state == NUD_INCOMPLETE) {
1010 if (skb) {
1011 while (neigh->arp_queue_len_bytes + skb->truesize >
1012 neigh->parms->queue_len_bytes) {
1013 struct sk_buff *buff;
1014
1015 buff = __skb_dequeue(&neigh->arp_queue);
1016 if (!buff)
1017 break;
1018 neigh->arp_queue_len_bytes -= buff->truesize;
1019 kfree_skb(buff);
1020 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
1021 }
1022 skb_dst_force(skb);
1023 __skb_queue_tail(&neigh->arp_queue, skb);
1024 neigh->arp_queue_len_bytes += skb->truesize;
1025 }
1026 rc = 1;
1027 }
1028out_unlock_bh:
1029 if (immediate_probe)
1030 neigh_probe(neigh);
1031 else
1032 write_unlock(&neigh->lock);
1033 local_bh_enable();
1034 return rc;
1035}
1036EXPORT_SYMBOL(__neigh_event_send);
1037
1038static void neigh_update_hhs(struct neighbour *neigh)
1039{
1040 struct hh_cache *hh;
1041 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1042 = NULL;
1043
1044 if (neigh->dev->header_ops)
1045 update = neigh->dev->header_ops->cache_update;
1046
1047 if (update) {
1048 hh = &neigh->hh;
1049 if (hh->hh_len) {
1050 write_seqlock_bh(&hh->hh_lock);
1051 update(hh, neigh->dev, neigh->ha);
1052 write_sequnlock_bh(&hh->hh_lock);
1053 }
1054 }
1055}
1056
1057
1058
1059/* Generic update routine.
1060 -- lladdr is new lladdr or NULL, if it is not supplied.
1061 -- new is new state.
1062 -- flags
1063 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1064 if it is different.
1065 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1066 lladdr instead of overriding it
1067 if it is different.
1068 It also allows to retain current state
1069 if lladdr is unchanged.
1070 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
1071
1072 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1073 NTF_ROUTER flag.
1074 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
1075 a router.
1076
1077 Caller MUST hold reference count on the entry.
1078 */
1079
1080int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1081 u32 flags)
1082{
1083 u8 old;
1084 int err;
1085 int notify = 0;
1086 struct net_device *dev;
1087 int update_isrouter = 0;
1088
1089 write_lock_bh(&neigh->lock);
1090
1091 dev = neigh->dev;
1092 old = neigh->nud_state;
1093 err = -EPERM;
1094
1095 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1096 (old & (NUD_NOARP | NUD_PERMANENT)))
1097 goto out;
1098
1099 if (!(new & NUD_VALID)) {
1100 neigh_del_timer(neigh);
1101 if (old & NUD_CONNECTED)
1102 neigh_suspect(neigh);
1103 neigh->nud_state = new;
1104 err = 0;
1105 notify = old & NUD_VALID;
1106 if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1107 (new & NUD_FAILED)) {
1108 neigh_invalidate(neigh);
1109 notify = 1;
1110 }
1111 goto out;
1112 }
1113
1114 /* Compare new lladdr with cached one */
1115 if (!dev->addr_len) {
1116 /* First case: device needs no address. */
1117 lladdr = neigh->ha;
1118 } else if (lladdr) {
1119 /* The second case: if something is already cached
1120 and a new address is proposed:
1121 - compare new & old
1122 - if they are different, check override flag
1123 */
1124 if ((old & NUD_VALID) &&
1125 !memcmp(lladdr, neigh->ha, dev->addr_len))
1126 lladdr = neigh->ha;
1127 } else {
1128 /* No address is supplied; if we know something,
1129 use it, otherwise discard the request.
1130 */
1131 err = -EINVAL;
1132 if (!(old & NUD_VALID))
1133 goto out;
1134 lladdr = neigh->ha;
1135 }
1136
1137 if (new & NUD_CONNECTED)
1138 neigh->confirmed = jiffies;
1139 neigh->updated = jiffies;
1140
1141 /* If entry was valid and address is not changed,
1142 do not change entry state, if new one is STALE.
1143 */
1144 err = 0;
1145 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1146 if (old & NUD_VALID) {
1147 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1148 update_isrouter = 0;
1149 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1150 (old & NUD_CONNECTED)) {
1151 lladdr = neigh->ha;
1152 new = NUD_STALE;
1153 } else
1154 goto out;
1155 } else {
1156 if (lladdr == neigh->ha && new == NUD_STALE &&
1157 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1158 (old & NUD_CONNECTED))
1159 )
1160 new = old;
1161 }
1162 }
1163
1164 if (new != old) {
1165 neigh_del_timer(neigh);
1166 if (new & NUD_IN_TIMER)
1167 neigh_add_timer(neigh, (jiffies +
1168 ((new & NUD_REACHABLE) ?
1169 neigh->parms->reachable_time :
1170 0)));
1171 neigh->nud_state = new;
1172 }
1173
1174 if (lladdr != neigh->ha) {
1175 write_seqlock(&neigh->ha_lock);
1176 memcpy(&neigh->ha, lladdr, dev->addr_len);
1177 write_sequnlock(&neigh->ha_lock);
1178 neigh_update_hhs(neigh);
1179 if (!(new & NUD_CONNECTED))
1180 neigh->confirmed = jiffies -
1181 (neigh->parms->base_reachable_time << 1);
1182 notify = 1;
1183 }
1184 if (new == old)
1185 goto out;
1186 if (new & NUD_CONNECTED)
1187 neigh_connect(neigh);
1188 else
1189 neigh_suspect(neigh);
1190 if (!(old & NUD_VALID)) {
1191 struct sk_buff *skb;
1192
1193 /* Again: avoid dead loop if something went wrong */
1194
1195 while (neigh->nud_state & NUD_VALID &&
1196 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1197 struct dst_entry *dst = skb_dst(skb);
1198 struct neighbour *n2, *n1 = neigh;
1199 write_unlock_bh(&neigh->lock);
1200
1201 rcu_read_lock();
1202 /* On shaper/eql skb->dst->neighbour != neigh :( */
1203 if (dst && (n2 = dst_get_neighbour_noref(dst)) != NULL)
1204 n1 = n2;
1205 n1->output(n1, skb);
1206 rcu_read_unlock();
1207
1208 write_lock_bh(&neigh->lock);
1209 }
1210 skb_queue_purge(&neigh->arp_queue);
1211 neigh->arp_queue_len_bytes = 0;
1212 }
1213out:
1214 if (update_isrouter) {
1215 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1216 (neigh->flags | NTF_ROUTER) :
1217 (neigh->flags & ~NTF_ROUTER);
1218 }
1219 write_unlock_bh(&neigh->lock);
1220
1221 if (notify)
1222 neigh_update_notify(neigh);
1223
1224 return err;
1225}
1226EXPORT_SYMBOL(neigh_update);
1227
1228struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1229 u8 *lladdr, void *saddr,
1230 struct net_device *dev)
1231{
1232 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1233 lladdr || !dev->addr_len);
1234 if (neigh)
1235 neigh_update(neigh, lladdr, NUD_STALE,
1236 NEIGH_UPDATE_F_OVERRIDE);
1237 return neigh;
1238}
1239EXPORT_SYMBOL(neigh_event_ns);
1240
1241/* called with read_lock_bh(&n->lock); */
1242static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst)
1243{
1244 struct net_device *dev = dst->dev;
1245 __be16 prot = dst->ops->protocol;
1246 struct hh_cache *hh = &n->hh;
1247
1248 write_lock_bh(&n->lock);
1249
1250 /* Only one thread can come in here and initialize the
1251 * hh_cache entry.
1252 */
1253 if (!hh->hh_len)
1254 dev->header_ops->cache(n, hh, prot);
1255
1256 write_unlock_bh(&n->lock);
1257}
1258
1259/* This function can be used in contexts, where only old dev_queue_xmit
1260 * worked, f.e. if you want to override normal output path (eql, shaper),
1261 * but resolution is not made yet.
1262 */
1263
1264int neigh_compat_output(struct neighbour *neigh, struct sk_buff *skb)
1265{
1266 struct net_device *dev = skb->dev;
1267
1268 __skb_pull(skb, skb_network_offset(skb));
1269
1270 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1271 skb->len) < 0 &&
1272 dev->header_ops->rebuild(skb))
1273 return 0;
1274
1275 return dev_queue_xmit(skb);
1276}
1277EXPORT_SYMBOL(neigh_compat_output);
1278
1279/* Slow and careful. */
1280
1281int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1282{
1283 struct dst_entry *dst = skb_dst(skb);
1284 int rc = 0;
1285
1286 if (!dst)
1287 goto discard;
1288
1289 __skb_pull(skb, skb_network_offset(skb));
1290
1291 if (!neigh_event_send(neigh, skb)) {
1292 int err;
1293 struct net_device *dev = neigh->dev;
1294 unsigned int seq;
1295
1296 if (dev->header_ops->cache && !neigh->hh.hh_len)
1297 neigh_hh_init(neigh, dst);
1298
1299 do {
1300 seq = read_seqbegin(&neigh->ha_lock);
1301 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1302 neigh->ha, NULL, skb->len);
1303 } while (read_seqretry(&neigh->ha_lock, seq));
1304
1305 if (err >= 0)
1306 rc = dev_queue_xmit(skb);
1307 else
1308 goto out_kfree_skb;
1309 }
1310out:
1311 return rc;
1312discard:
1313 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1314 dst, neigh);
1315out_kfree_skb:
1316 rc = -EINVAL;
1317 kfree_skb(skb);
1318 goto out;
1319}
1320EXPORT_SYMBOL(neigh_resolve_output);
1321
1322/* As fast as possible without hh cache */
1323
1324int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1325{
1326 struct net_device *dev = neigh->dev;
1327 unsigned int seq;
1328 int err;
1329
1330 __skb_pull(skb, skb_network_offset(skb));
1331
1332 do {
1333 seq = read_seqbegin(&neigh->ha_lock);
1334 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1335 neigh->ha, NULL, skb->len);
1336 } while (read_seqretry(&neigh->ha_lock, seq));
1337
1338 if (err >= 0)
1339 err = dev_queue_xmit(skb);
1340 else {
1341 err = -EINVAL;
1342 kfree_skb(skb);
1343 }
1344 return err;
1345}
1346EXPORT_SYMBOL(neigh_connected_output);
1347
1348int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1349{
1350 return dev_queue_xmit(skb);
1351}
1352EXPORT_SYMBOL(neigh_direct_output);
1353
1354static void neigh_proxy_process(unsigned long arg)
1355{
1356 struct neigh_table *tbl = (struct neigh_table *)arg;
1357 long sched_next = 0;
1358 unsigned long now = jiffies;
1359 struct sk_buff *skb, *n;
1360
1361 spin_lock(&tbl->proxy_queue.lock);
1362
1363 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1364 long tdif = NEIGH_CB(skb)->sched_next - now;
1365
1366 if (tdif <= 0) {
1367 struct net_device *dev = skb->dev;
1368
1369 __skb_unlink(skb, &tbl->proxy_queue);
1370 if (tbl->proxy_redo && netif_running(dev)) {
1371 rcu_read_lock();
1372 tbl->proxy_redo(skb);
1373 rcu_read_unlock();
1374 } else {
1375 kfree_skb(skb);
1376 }
1377
1378 dev_put(dev);
1379 } else if (!sched_next || tdif < sched_next)
1380 sched_next = tdif;
1381 }
1382 del_timer(&tbl->proxy_timer);
1383 if (sched_next)
1384 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1385 spin_unlock(&tbl->proxy_queue.lock);
1386}
1387
1388void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1389 struct sk_buff *skb)
1390{
1391 unsigned long now = jiffies;
1392 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1393
1394 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1395 kfree_skb(skb);
1396 return;
1397 }
1398
1399 NEIGH_CB(skb)->sched_next = sched_next;
1400 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1401
1402 spin_lock(&tbl->proxy_queue.lock);
1403 if (del_timer(&tbl->proxy_timer)) {
1404 if (time_before(tbl->proxy_timer.expires, sched_next))
1405 sched_next = tbl->proxy_timer.expires;
1406 }
1407 skb_dst_drop(skb);
1408 dev_hold(skb->dev);
1409 __skb_queue_tail(&tbl->proxy_queue, skb);
1410 mod_timer(&tbl->proxy_timer, sched_next);
1411 spin_unlock(&tbl->proxy_queue.lock);
1412}
1413EXPORT_SYMBOL(pneigh_enqueue);
1414
1415static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1416 struct net *net, int ifindex)
1417{
1418 struct neigh_parms *p;
1419
1420 for (p = &tbl->parms; p; p = p->next) {
1421 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1422 (!p->dev && !ifindex))
1423 return p;
1424 }
1425
1426 return NULL;
1427}
1428
1429struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1430 struct neigh_table *tbl)
1431{
1432 struct neigh_parms *p, *ref;
1433 struct net *net = dev_net(dev);
1434 const struct net_device_ops *ops = dev->netdev_ops;
1435
1436 ref = lookup_neigh_parms(tbl, net, 0);
1437 if (!ref)
1438 return NULL;
1439
1440 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1441 if (p) {
1442 p->tbl = tbl;
1443 atomic_set(&p->refcnt, 1);
1444 p->reachable_time =
1445 neigh_rand_reach_time(p->base_reachable_time);
1446
1447 if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1448 kfree(p);
1449 return NULL;
1450 }
1451
1452 dev_hold(dev);
1453 p->dev = dev;
1454 write_pnet(&p->net, hold_net(net));
1455 p->sysctl_table = NULL;
1456 write_lock_bh(&tbl->lock);
1457 p->next = tbl->parms.next;
1458 tbl->parms.next = p;
1459 write_unlock_bh(&tbl->lock);
1460 }
1461 return p;
1462}
1463EXPORT_SYMBOL(neigh_parms_alloc);
1464
1465static void neigh_rcu_free_parms(struct rcu_head *head)
1466{
1467 struct neigh_parms *parms =
1468 container_of(head, struct neigh_parms, rcu_head);
1469
1470 neigh_parms_put(parms);
1471}
1472
1473void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1474{
1475 struct neigh_parms **p;
1476
1477 if (!parms || parms == &tbl->parms)
1478 return;
1479 write_lock_bh(&tbl->lock);
1480 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1481 if (*p == parms) {
1482 *p = parms->next;
1483 parms->dead = 1;
1484 write_unlock_bh(&tbl->lock);
1485 if (parms->dev)
1486 dev_put(parms->dev);
1487 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1488 return;
1489 }
1490 }
1491 write_unlock_bh(&tbl->lock);
1492 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1493}
1494EXPORT_SYMBOL(neigh_parms_release);
1495
1496static void neigh_parms_destroy(struct neigh_parms *parms)
1497{
1498 release_net(neigh_parms_net(parms));
1499 kfree(parms);
1500}
1501
1502static struct lock_class_key neigh_table_proxy_queue_class;
1503
1504static void neigh_table_init_no_netlink(struct neigh_table *tbl)
1505{
1506 unsigned long now = jiffies;
1507 unsigned long phsize;
1508
1509 write_pnet(&tbl->parms.net, &init_net);
1510 atomic_set(&tbl->parms.refcnt, 1);
1511 tbl->parms.reachable_time =
1512 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1513
1514 tbl->stats = alloc_percpu(struct neigh_statistics);
1515 if (!tbl->stats)
1516 panic("cannot create neighbour cache statistics");
1517
1518#ifdef CONFIG_PROC_FS
1519 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1520 &neigh_stat_seq_fops, tbl))
1521 panic("cannot create neighbour proc dir entry");
1522#endif
1523
1524 RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1525
1526 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1527 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1528
1529 if (!tbl->nht || !tbl->phash_buckets)
1530 panic("cannot allocate neighbour cache hashes");
1531
1532 rwlock_init(&tbl->lock);
1533 INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1534 schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1535 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
1536 skb_queue_head_init_class(&tbl->proxy_queue,
1537 &neigh_table_proxy_queue_class);
1538
1539 tbl->last_flush = now;
1540 tbl->last_rand = now + tbl->parms.reachable_time * 20;
1541}
1542
1543void neigh_table_init(struct neigh_table *tbl)
1544{
1545 struct neigh_table *tmp;
1546
1547 neigh_table_init_no_netlink(tbl);
1548 write_lock(&neigh_tbl_lock);
1549 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1550 if (tmp->family == tbl->family)
1551 break;
1552 }
1553 tbl->next = neigh_tables;
1554 neigh_tables = tbl;
1555 write_unlock(&neigh_tbl_lock);
1556
1557 if (unlikely(tmp)) {
1558 pr_err("Registering multiple tables for family %d\n",
1559 tbl->family);
1560 dump_stack();
1561 }
1562}
1563EXPORT_SYMBOL(neigh_table_init);
1564
1565int neigh_table_clear(struct neigh_table *tbl)
1566{
1567 struct neigh_table **tp;
1568
1569 /* It is not clean... Fix it to unload IPv6 module safely */
1570 cancel_delayed_work_sync(&tbl->gc_work);
1571 del_timer_sync(&tbl->proxy_timer);
1572 pneigh_queue_purge(&tbl->proxy_queue);
1573 neigh_ifdown(tbl, NULL);
1574 if (atomic_read(&tbl->entries))
1575 pr_crit("neighbour leakage\n");
1576 write_lock(&neigh_tbl_lock);
1577 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1578 if (*tp == tbl) {
1579 *tp = tbl->next;
1580 break;
1581 }
1582 }
1583 write_unlock(&neigh_tbl_lock);
1584
1585 call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1586 neigh_hash_free_rcu);
1587 tbl->nht = NULL;
1588
1589 kfree(tbl->phash_buckets);
1590 tbl->phash_buckets = NULL;
1591
1592 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1593
1594 free_percpu(tbl->stats);
1595 tbl->stats = NULL;
1596
1597 return 0;
1598}
1599EXPORT_SYMBOL(neigh_table_clear);
1600
1601static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1602{
1603 struct net *net = sock_net(skb->sk);
1604 struct ndmsg *ndm;
1605 struct nlattr *dst_attr;
1606 struct neigh_table *tbl;
1607 struct net_device *dev = NULL;
1608 int err = -EINVAL;
1609
1610 ASSERT_RTNL();
1611 if (nlmsg_len(nlh) < sizeof(*ndm))
1612 goto out;
1613
1614 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1615 if (dst_attr == NULL)
1616 goto out;
1617
1618 ndm = nlmsg_data(nlh);
1619 if (ndm->ndm_ifindex) {
1620 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1621 if (dev == NULL) {
1622 err = -ENODEV;
1623 goto out;
1624 }
1625 }
1626
1627 read_lock(&neigh_tbl_lock);
1628 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1629 struct neighbour *neigh;
1630
1631 if (tbl->family != ndm->ndm_family)
1632 continue;
1633 read_unlock(&neigh_tbl_lock);
1634
1635 if (nla_len(dst_attr) < tbl->key_len)
1636 goto out;
1637
1638 if (ndm->ndm_flags & NTF_PROXY) {
1639 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1640 goto out;
1641 }
1642
1643 if (dev == NULL)
1644 goto out;
1645
1646 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1647 if (neigh == NULL) {
1648 err = -ENOENT;
1649 goto out;
1650 }
1651
1652 err = neigh_update(neigh, NULL, NUD_FAILED,
1653 NEIGH_UPDATE_F_OVERRIDE |
1654 NEIGH_UPDATE_F_ADMIN);
1655 neigh_release(neigh);
1656 goto out;
1657 }
1658 read_unlock(&neigh_tbl_lock);
1659 err = -EAFNOSUPPORT;
1660
1661out:
1662 return err;
1663}
1664
1665static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1666{
1667 struct net *net = sock_net(skb->sk);
1668 struct ndmsg *ndm;
1669 struct nlattr *tb[NDA_MAX+1];
1670 struct neigh_table *tbl;
1671 struct net_device *dev = NULL;
1672 int err;
1673
1674 ASSERT_RTNL();
1675 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1676 if (err < 0)
1677 goto out;
1678
1679 err = -EINVAL;
1680 if (tb[NDA_DST] == NULL)
1681 goto out;
1682
1683 ndm = nlmsg_data(nlh);
1684 if (ndm->ndm_ifindex) {
1685 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1686 if (dev == NULL) {
1687 err = -ENODEV;
1688 goto out;
1689 }
1690
1691 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1692 goto out;
1693 }
1694
1695 read_lock(&neigh_tbl_lock);
1696 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1697 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1698 struct neighbour *neigh;
1699 void *dst, *lladdr;
1700
1701 if (tbl->family != ndm->ndm_family)
1702 continue;
1703 read_unlock(&neigh_tbl_lock);
1704
1705 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1706 goto out;
1707 dst = nla_data(tb[NDA_DST]);
1708 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1709
1710 if (ndm->ndm_flags & NTF_PROXY) {
1711 struct pneigh_entry *pn;
1712
1713 err = -ENOBUFS;
1714 pn = pneigh_lookup(tbl, net, dst, dev, 1);
1715 if (pn) {
1716 pn->flags = ndm->ndm_flags;
1717 err = 0;
1718 }
1719 goto out;
1720 }
1721
1722 if (dev == NULL)
1723 goto out;
1724
1725 neigh = neigh_lookup(tbl, dst, dev);
1726 if (neigh == NULL) {
1727 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1728 err = -ENOENT;
1729 goto out;
1730 }
1731
1732 neigh = __neigh_lookup_errno(tbl, dst, dev);
1733 if (IS_ERR(neigh)) {
1734 err = PTR_ERR(neigh);
1735 goto out;
1736 }
1737 } else {
1738 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1739 err = -EEXIST;
1740 neigh_release(neigh);
1741 goto out;
1742 }
1743
1744 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1745 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1746 }
1747
1748 if (ndm->ndm_flags & NTF_USE) {
1749 neigh_event_send(neigh, NULL);
1750 err = 0;
1751 } else
1752 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1753 neigh_release(neigh);
1754 goto out;
1755 }
1756
1757 read_unlock(&neigh_tbl_lock);
1758 err = -EAFNOSUPPORT;
1759out:
1760 return err;
1761}
1762
1763static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1764{
1765 struct nlattr *nest;
1766
1767 nest = nla_nest_start(skb, NDTA_PARMS);
1768 if (nest == NULL)
1769 return -ENOBUFS;
1770
1771 if ((parms->dev &&
1772 nla_put_u32(skb, NDTPA_IFINDEX, parms->dev->ifindex)) ||
1773 nla_put_u32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt)) ||
1774 nla_put_u32(skb, NDTPA_QUEUE_LENBYTES, parms->queue_len_bytes) ||
1775 /* approximative value for deprecated QUEUE_LEN (in packets) */
1776 nla_put_u32(skb, NDTPA_QUEUE_LEN,
1777 DIV_ROUND_UP(parms->queue_len_bytes,
1778 SKB_TRUESIZE(ETH_FRAME_LEN))) ||
1779 nla_put_u32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen) ||
1780 nla_put_u32(skb, NDTPA_APP_PROBES, parms->app_probes) ||
1781 nla_put_u32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes) ||
1782 nla_put_u32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes) ||
1783 nla_put_msecs(skb, NDTPA_REACHABLE_TIME, parms->reachable_time) ||
1784 nla_put_msecs(skb, NDTPA_BASE_REACHABLE_TIME,
1785 parms->base_reachable_time) ||
1786 nla_put_msecs(skb, NDTPA_GC_STALETIME, parms->gc_staletime) ||
1787 nla_put_msecs(skb, NDTPA_DELAY_PROBE_TIME,
1788 parms->delay_probe_time) ||
1789 nla_put_msecs(skb, NDTPA_RETRANS_TIME, parms->retrans_time) ||
1790 nla_put_msecs(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay) ||
1791 nla_put_msecs(skb, NDTPA_PROXY_DELAY, parms->proxy_delay) ||
1792 nla_put_msecs(skb, NDTPA_LOCKTIME, parms->locktime))
1793 goto nla_put_failure;
1794 return nla_nest_end(skb, nest);
1795
1796nla_put_failure:
1797 nla_nest_cancel(skb, nest);
1798 return -EMSGSIZE;
1799}
1800
1801static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1802 u32 pid, u32 seq, int type, int flags)
1803{
1804 struct nlmsghdr *nlh;
1805 struct ndtmsg *ndtmsg;
1806
1807 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1808 if (nlh == NULL)
1809 return -EMSGSIZE;
1810
1811 ndtmsg = nlmsg_data(nlh);
1812
1813 read_lock_bh(&tbl->lock);
1814 ndtmsg->ndtm_family = tbl->family;
1815 ndtmsg->ndtm_pad1 = 0;
1816 ndtmsg->ndtm_pad2 = 0;
1817
1818 if (nla_put_string(skb, NDTA_NAME, tbl->id) ||
1819 nla_put_msecs(skb, NDTA_GC_INTERVAL, tbl->gc_interval) ||
1820 nla_put_u32(skb, NDTA_THRESH1, tbl->gc_thresh1) ||
1821 nla_put_u32(skb, NDTA_THRESH2, tbl->gc_thresh2) ||
1822 nla_put_u32(skb, NDTA_THRESH3, tbl->gc_thresh3))
1823 goto nla_put_failure;
1824 {
1825 unsigned long now = jiffies;
1826 unsigned int flush_delta = now - tbl->last_flush;
1827 unsigned int rand_delta = now - tbl->last_rand;
1828 struct neigh_hash_table *nht;
1829 struct ndt_config ndc = {
1830 .ndtc_key_len = tbl->key_len,
1831 .ndtc_entry_size = tbl->entry_size,
1832 .ndtc_entries = atomic_read(&tbl->entries),
1833 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1834 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1835 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1836 };
1837
1838 rcu_read_lock_bh();
1839 nht = rcu_dereference_bh(tbl->nht);
1840 ndc.ndtc_hash_rnd = nht->hash_rnd[0];
1841 ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1842 rcu_read_unlock_bh();
1843
1844 if (nla_put(skb, NDTA_CONFIG, sizeof(ndc), &ndc))
1845 goto nla_put_failure;
1846 }
1847
1848 {
1849 int cpu;
1850 struct ndt_stats ndst;
1851
1852 memset(&ndst, 0, sizeof(ndst));
1853
1854 for_each_possible_cpu(cpu) {
1855 struct neigh_statistics *st;
1856
1857 st = per_cpu_ptr(tbl->stats, cpu);
1858 ndst.ndts_allocs += st->allocs;
1859 ndst.ndts_destroys += st->destroys;
1860 ndst.ndts_hash_grows += st->hash_grows;
1861 ndst.ndts_res_failed += st->res_failed;
1862 ndst.ndts_lookups += st->lookups;
1863 ndst.ndts_hits += st->hits;
1864 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1865 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1866 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1867 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1868 }
1869
1870 if (nla_put(skb, NDTA_STATS, sizeof(ndst), &ndst))
1871 goto nla_put_failure;
1872 }
1873
1874 BUG_ON(tbl->parms.dev);
1875 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1876 goto nla_put_failure;
1877
1878 read_unlock_bh(&tbl->lock);
1879 return nlmsg_end(skb, nlh);
1880
1881nla_put_failure:
1882 read_unlock_bh(&tbl->lock);
1883 nlmsg_cancel(skb, nlh);
1884 return -EMSGSIZE;
1885}
1886
1887static int neightbl_fill_param_info(struct sk_buff *skb,
1888 struct neigh_table *tbl,
1889 struct neigh_parms *parms,
1890 u32 pid, u32 seq, int type,
1891 unsigned int flags)
1892{
1893 struct ndtmsg *ndtmsg;
1894 struct nlmsghdr *nlh;
1895
1896 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1897 if (nlh == NULL)
1898 return -EMSGSIZE;
1899
1900 ndtmsg = nlmsg_data(nlh);
1901
1902 read_lock_bh(&tbl->lock);
1903 ndtmsg->ndtm_family = tbl->family;
1904 ndtmsg->ndtm_pad1 = 0;
1905 ndtmsg->ndtm_pad2 = 0;
1906
1907 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1908 neightbl_fill_parms(skb, parms) < 0)
1909 goto errout;
1910
1911 read_unlock_bh(&tbl->lock);
1912 return nlmsg_end(skb, nlh);
1913errout:
1914 read_unlock_bh(&tbl->lock);
1915 nlmsg_cancel(skb, nlh);
1916 return -EMSGSIZE;
1917}
1918
1919static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1920 [NDTA_NAME] = { .type = NLA_STRING },
1921 [NDTA_THRESH1] = { .type = NLA_U32 },
1922 [NDTA_THRESH2] = { .type = NLA_U32 },
1923 [NDTA_THRESH3] = { .type = NLA_U32 },
1924 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1925 [NDTA_PARMS] = { .type = NLA_NESTED },
1926};
1927
1928static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1929 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1930 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1931 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1932 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1933 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1934 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1935 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1936 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1937 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1938 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1939 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1940 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1941 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1942};
1943
1944static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1945{
1946 struct net *net = sock_net(skb->sk);
1947 struct neigh_table *tbl;
1948 struct ndtmsg *ndtmsg;
1949 struct nlattr *tb[NDTA_MAX+1];
1950 int err;
1951
1952 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1953 nl_neightbl_policy);
1954 if (err < 0)
1955 goto errout;
1956
1957 if (tb[NDTA_NAME] == NULL) {
1958 err = -EINVAL;
1959 goto errout;
1960 }
1961
1962 ndtmsg = nlmsg_data(nlh);
1963 read_lock(&neigh_tbl_lock);
1964 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1965 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1966 continue;
1967
1968 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1969 break;
1970 }
1971
1972 if (tbl == NULL) {
1973 err = -ENOENT;
1974 goto errout_locked;
1975 }
1976
1977 /*
1978 * We acquire tbl->lock to be nice to the periodic timers and
1979 * make sure they always see a consistent set of values.
1980 */
1981 write_lock_bh(&tbl->lock);
1982
1983 if (tb[NDTA_PARMS]) {
1984 struct nlattr *tbp[NDTPA_MAX+1];
1985 struct neigh_parms *p;
1986 int i, ifindex = 0;
1987
1988 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1989 nl_ntbl_parm_policy);
1990 if (err < 0)
1991 goto errout_tbl_lock;
1992
1993 if (tbp[NDTPA_IFINDEX])
1994 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1995
1996 p = lookup_neigh_parms(tbl, net, ifindex);
1997 if (p == NULL) {
1998 err = -ENOENT;
1999 goto errout_tbl_lock;
2000 }
2001
2002 for (i = 1; i <= NDTPA_MAX; i++) {
2003 if (tbp[i] == NULL)
2004 continue;
2005
2006 switch (i) {
2007 case NDTPA_QUEUE_LEN:
2008 p->queue_len_bytes = nla_get_u32(tbp[i]) *
2009 SKB_TRUESIZE(ETH_FRAME_LEN);
2010 break;
2011 case NDTPA_QUEUE_LENBYTES:
2012 p->queue_len_bytes = nla_get_u32(tbp[i]);
2013 break;
2014 case NDTPA_PROXY_QLEN:
2015 p->proxy_qlen = nla_get_u32(tbp[i]);
2016 break;
2017 case NDTPA_APP_PROBES:
2018 p->app_probes = nla_get_u32(tbp[i]);
2019 break;
2020 case NDTPA_UCAST_PROBES:
2021 p->ucast_probes = nla_get_u32(tbp[i]);
2022 break;
2023 case NDTPA_MCAST_PROBES:
2024 p->mcast_probes = nla_get_u32(tbp[i]);
2025 break;
2026 case NDTPA_BASE_REACHABLE_TIME:
2027 p->base_reachable_time = nla_get_msecs(tbp[i]);
2028 break;
2029 case NDTPA_GC_STALETIME:
2030 p->gc_staletime = nla_get_msecs(tbp[i]);
2031 break;
2032 case NDTPA_DELAY_PROBE_TIME:
2033 p->delay_probe_time = nla_get_msecs(tbp[i]);
2034 break;
2035 case NDTPA_RETRANS_TIME:
2036 p->retrans_time = nla_get_msecs(tbp[i]);
2037 break;
2038 case NDTPA_ANYCAST_DELAY:
2039 p->anycast_delay = nla_get_msecs(tbp[i]);
2040 break;
2041 case NDTPA_PROXY_DELAY:
2042 p->proxy_delay = nla_get_msecs(tbp[i]);
2043 break;
2044 case NDTPA_LOCKTIME:
2045 p->locktime = nla_get_msecs(tbp[i]);
2046 break;
2047 }
2048 }
2049 }
2050
2051 if (tb[NDTA_THRESH1])
2052 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2053
2054 if (tb[NDTA_THRESH2])
2055 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2056
2057 if (tb[NDTA_THRESH3])
2058 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2059
2060 if (tb[NDTA_GC_INTERVAL])
2061 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2062
2063 err = 0;
2064
2065errout_tbl_lock:
2066 write_unlock_bh(&tbl->lock);
2067errout_locked:
2068 read_unlock(&neigh_tbl_lock);
2069errout:
2070 return err;
2071}
2072
2073static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2074{
2075 struct net *net = sock_net(skb->sk);
2076 int family, tidx, nidx = 0;
2077 int tbl_skip = cb->args[0];
2078 int neigh_skip = cb->args[1];
2079 struct neigh_table *tbl;
2080
2081 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2082
2083 read_lock(&neigh_tbl_lock);
2084 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
2085 struct neigh_parms *p;
2086
2087 if (tidx < tbl_skip || (family && tbl->family != family))
2088 continue;
2089
2090 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2091 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2092 NLM_F_MULTI) <= 0)
2093 break;
2094
2095 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
2096 if (!net_eq(neigh_parms_net(p), net))
2097 continue;
2098
2099 if (nidx < neigh_skip)
2100 goto next;
2101
2102 if (neightbl_fill_param_info(skb, tbl, p,
2103 NETLINK_CB(cb->skb).pid,
2104 cb->nlh->nlmsg_seq,
2105 RTM_NEWNEIGHTBL,
2106 NLM_F_MULTI) <= 0)
2107 goto out;
2108 next:
2109 nidx++;
2110 }
2111
2112 neigh_skip = 0;
2113 }
2114out:
2115 read_unlock(&neigh_tbl_lock);
2116 cb->args[0] = tidx;
2117 cb->args[1] = nidx;
2118
2119 return skb->len;
2120}
2121
2122static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2123 u32 pid, u32 seq, int type, unsigned int flags)
2124{
2125 unsigned long now = jiffies;
2126 struct nda_cacheinfo ci;
2127 struct nlmsghdr *nlh;
2128 struct ndmsg *ndm;
2129
2130 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2131 if (nlh == NULL)
2132 return -EMSGSIZE;
2133
2134 ndm = nlmsg_data(nlh);
2135 ndm->ndm_family = neigh->ops->family;
2136 ndm->ndm_pad1 = 0;
2137 ndm->ndm_pad2 = 0;
2138 ndm->ndm_flags = neigh->flags;
2139 ndm->ndm_type = neigh->type;
2140 ndm->ndm_ifindex = neigh->dev->ifindex;
2141
2142 if (nla_put(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key))
2143 goto nla_put_failure;
2144
2145 read_lock_bh(&neigh->lock);
2146 ndm->ndm_state = neigh->nud_state;
2147 if (neigh->nud_state & NUD_VALID) {
2148 char haddr[MAX_ADDR_LEN];
2149
2150 neigh_ha_snapshot(haddr, neigh, neigh->dev);
2151 if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2152 read_unlock_bh(&neigh->lock);
2153 goto nla_put_failure;
2154 }
2155 }
2156
2157 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2158 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2159 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
2160 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2161 read_unlock_bh(&neigh->lock);
2162
2163 if (nla_put_u32(skb, NDA_PROBES, atomic_read(&neigh->probes)) ||
2164 nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
2165 goto nla_put_failure;
2166
2167 return nlmsg_end(skb, nlh);
2168
2169nla_put_failure:
2170 nlmsg_cancel(skb, nlh);
2171 return -EMSGSIZE;
2172}
2173
2174static int pneigh_fill_info(struct sk_buff *skb, struct pneigh_entry *pn,
2175 u32 pid, u32 seq, int type, unsigned int flags,
2176 struct neigh_table *tbl)
2177{
2178 struct nlmsghdr *nlh;
2179 struct ndmsg *ndm;
2180
2181 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2182 if (nlh == NULL)
2183 return -EMSGSIZE;
2184
2185 ndm = nlmsg_data(nlh);
2186 ndm->ndm_family = tbl->family;
2187 ndm->ndm_pad1 = 0;
2188 ndm->ndm_pad2 = 0;
2189 ndm->ndm_flags = pn->flags | NTF_PROXY;
2190 ndm->ndm_type = NDA_DST;
2191 ndm->ndm_ifindex = pn->dev->ifindex;
2192 ndm->ndm_state = NUD_NONE;
2193
2194 if (nla_put(skb, NDA_DST, tbl->key_len, pn->key))
2195 goto nla_put_failure;
2196
2197 return nlmsg_end(skb, nlh);
2198
2199nla_put_failure:
2200 nlmsg_cancel(skb, nlh);
2201 return -EMSGSIZE;
2202}
2203
2204static void neigh_update_notify(struct neighbour *neigh)
2205{
2206 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2207 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2208}
2209
2210static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2211 struct netlink_callback *cb)
2212{
2213 struct net *net = sock_net(skb->sk);
2214 struct neighbour *n;
2215 int rc, h, s_h = cb->args[1];
2216 int idx, s_idx = idx = cb->args[2];
2217 struct neigh_hash_table *nht;
2218
2219 rcu_read_lock_bh();
2220 nht = rcu_dereference_bh(tbl->nht);
2221
2222 for (h = s_h; h < (1 << nht->hash_shift); h++) {
2223 if (h > s_h)
2224 s_idx = 0;
2225 for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2226 n != NULL;
2227 n = rcu_dereference_bh(n->next)) {
2228 if (!net_eq(dev_net(n->dev), net))
2229 continue;
2230 if (idx < s_idx)
2231 goto next;
2232 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2233 cb->nlh->nlmsg_seq,
2234 RTM_NEWNEIGH,
2235 NLM_F_MULTI) <= 0) {
2236 rc = -1;
2237 goto out;
2238 }
2239next:
2240 idx++;
2241 }
2242 }
2243 rc = skb->len;
2244out:
2245 rcu_read_unlock_bh();
2246 cb->args[1] = h;
2247 cb->args[2] = idx;
2248 return rc;
2249}
2250
2251static int pneigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2252 struct netlink_callback *cb)
2253{
2254 struct pneigh_entry *n;
2255 struct net *net = sock_net(skb->sk);
2256 int rc, h, s_h = cb->args[3];
2257 int idx, s_idx = idx = cb->args[4];
2258
2259 read_lock_bh(&tbl->lock);
2260
2261 for (h = s_h; h <= PNEIGH_HASHMASK; h++) {
2262 if (h > s_h)
2263 s_idx = 0;
2264 for (n = tbl->phash_buckets[h], idx = 0; n; n = n->next) {
2265 if (dev_net(n->dev) != net)
2266 continue;
2267 if (idx < s_idx)
2268 goto next;
2269 if (pneigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2270 cb->nlh->nlmsg_seq,
2271 RTM_NEWNEIGH,
2272 NLM_F_MULTI, tbl) <= 0) {
2273 read_unlock_bh(&tbl->lock);
2274 rc = -1;
2275 goto out;
2276 }
2277 next:
2278 idx++;
2279 }
2280 }
2281
2282 read_unlock_bh(&tbl->lock);
2283 rc = skb->len;
2284out:
2285 cb->args[3] = h;
2286 cb->args[4] = idx;
2287 return rc;
2288
2289}
2290
2291static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2292{
2293 struct neigh_table *tbl;
2294 int t, family, s_t;
2295 int proxy = 0;
2296 int err;
2297
2298 read_lock(&neigh_tbl_lock);
2299 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2300
2301 /* check for full ndmsg structure presence, family member is
2302 * the same for both structures
2303 */
2304 if (nlmsg_len(cb->nlh) >= sizeof(struct ndmsg) &&
2305 ((struct ndmsg *) nlmsg_data(cb->nlh))->ndm_flags == NTF_PROXY)
2306 proxy = 1;
2307
2308 s_t = cb->args[0];
2309
2310 for (tbl = neigh_tables, t = 0; tbl;
2311 tbl = tbl->next, t++) {
2312 if (t < s_t || (family && tbl->family != family))
2313 continue;
2314 if (t > s_t)
2315 memset(&cb->args[1], 0, sizeof(cb->args) -
2316 sizeof(cb->args[0]));
2317 if (proxy)
2318 err = pneigh_dump_table(tbl, skb, cb);
2319 else
2320 err = neigh_dump_table(tbl, skb, cb);
2321 if (err < 0)
2322 break;
2323 }
2324 read_unlock(&neigh_tbl_lock);
2325
2326 cb->args[0] = t;
2327 return skb->len;
2328}
2329
2330void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2331{
2332 int chain;
2333 struct neigh_hash_table *nht;
2334
2335 rcu_read_lock_bh();
2336 nht = rcu_dereference_bh(tbl->nht);
2337
2338 read_lock(&tbl->lock); /* avoid resizes */
2339 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2340 struct neighbour *n;
2341
2342 for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2343 n != NULL;
2344 n = rcu_dereference_bh(n->next))
2345 cb(n, cookie);
2346 }
2347 read_unlock(&tbl->lock);
2348 rcu_read_unlock_bh();
2349}
2350EXPORT_SYMBOL(neigh_for_each);
2351
2352/* The tbl->lock must be held as a writer and BH disabled. */
2353void __neigh_for_each_release(struct neigh_table *tbl,
2354 int (*cb)(struct neighbour *))
2355{
2356 int chain;
2357 struct neigh_hash_table *nht;
2358
2359 nht = rcu_dereference_protected(tbl->nht,
2360 lockdep_is_held(&tbl->lock));
2361 for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2362 struct neighbour *n;
2363 struct neighbour __rcu **np;
2364
2365 np = &nht->hash_buckets[chain];
2366 while ((n = rcu_dereference_protected(*np,
2367 lockdep_is_held(&tbl->lock))) != NULL) {
2368 int release;
2369
2370 write_lock(&n->lock);
2371 release = cb(n);
2372 if (release) {
2373 rcu_assign_pointer(*np,
2374 rcu_dereference_protected(n->next,
2375 lockdep_is_held(&tbl->lock)));
2376 n->dead = 1;
2377 } else
2378 np = &n->next;
2379 write_unlock(&n->lock);
2380 if (release)
2381 neigh_cleanup_and_release(n);
2382 }
2383 }
2384}
2385EXPORT_SYMBOL(__neigh_for_each_release);
2386
2387#ifdef CONFIG_PROC_FS
2388
2389static struct neighbour *neigh_get_first(struct seq_file *seq)
2390{
2391 struct neigh_seq_state *state = seq->private;
2392 struct net *net = seq_file_net(seq);
2393 struct neigh_hash_table *nht = state->nht;
2394 struct neighbour *n = NULL;
2395 int bucket = state->bucket;
2396
2397 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2398 for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2399 n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2400
2401 while (n) {
2402 if (!net_eq(dev_net(n->dev), net))
2403 goto next;
2404 if (state->neigh_sub_iter) {
2405 loff_t fakep = 0;
2406 void *v;
2407
2408 v = state->neigh_sub_iter(state, n, &fakep);
2409 if (!v)
2410 goto next;
2411 }
2412 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2413 break;
2414 if (n->nud_state & ~NUD_NOARP)
2415 break;
2416next:
2417 n = rcu_dereference_bh(n->next);
2418 }
2419
2420 if (n)
2421 break;
2422 }
2423 state->bucket = bucket;
2424
2425 return n;
2426}
2427
2428static struct neighbour *neigh_get_next(struct seq_file *seq,
2429 struct neighbour *n,
2430 loff_t *pos)
2431{
2432 struct neigh_seq_state *state = seq->private;
2433 struct net *net = seq_file_net(seq);
2434 struct neigh_hash_table *nht = state->nht;
2435
2436 if (state->neigh_sub_iter) {
2437 void *v = state->neigh_sub_iter(state, n, pos);
2438 if (v)
2439 return n;
2440 }
2441 n = rcu_dereference_bh(n->next);
2442
2443 while (1) {
2444 while (n) {
2445 if (!net_eq(dev_net(n->dev), net))
2446 goto next;
2447 if (state->neigh_sub_iter) {
2448 void *v = state->neigh_sub_iter(state, n, pos);
2449 if (v)
2450 return n;
2451 goto next;
2452 }
2453 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2454 break;
2455
2456 if (n->nud_state & ~NUD_NOARP)
2457 break;
2458next:
2459 n = rcu_dereference_bh(n->next);
2460 }
2461
2462 if (n)
2463 break;
2464
2465 if (++state->bucket >= (1 << nht->hash_shift))
2466 break;
2467
2468 n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2469 }
2470
2471 if (n && pos)
2472 --(*pos);
2473 return n;
2474}
2475
2476static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2477{
2478 struct neighbour *n = neigh_get_first(seq);
2479
2480 if (n) {
2481 --(*pos);
2482 while (*pos) {
2483 n = neigh_get_next(seq, n, pos);
2484 if (!n)
2485 break;
2486 }
2487 }
2488 return *pos ? NULL : n;
2489}
2490
2491static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2492{
2493 struct neigh_seq_state *state = seq->private;
2494 struct net *net = seq_file_net(seq);
2495 struct neigh_table *tbl = state->tbl;
2496 struct pneigh_entry *pn = NULL;
2497 int bucket = state->bucket;
2498
2499 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2500 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2501 pn = tbl->phash_buckets[bucket];
2502 while (pn && !net_eq(pneigh_net(pn), net))
2503 pn = pn->next;
2504 if (pn)
2505 break;
2506 }
2507 state->bucket = bucket;
2508
2509 return pn;
2510}
2511
2512static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2513 struct pneigh_entry *pn,
2514 loff_t *pos)
2515{
2516 struct neigh_seq_state *state = seq->private;
2517 struct net *net = seq_file_net(seq);
2518 struct neigh_table *tbl = state->tbl;
2519
2520 do {
2521 pn = pn->next;
2522 } while (pn && !net_eq(pneigh_net(pn), net));
2523
2524 while (!pn) {
2525 if (++state->bucket > PNEIGH_HASHMASK)
2526 break;
2527 pn = tbl->phash_buckets[state->bucket];
2528 while (pn && !net_eq(pneigh_net(pn), net))
2529 pn = pn->next;
2530 if (pn)
2531 break;
2532 }
2533
2534 if (pn && pos)
2535 --(*pos);
2536
2537 return pn;
2538}
2539
2540static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2541{
2542 struct pneigh_entry *pn = pneigh_get_first(seq);
2543
2544 if (pn) {
2545 --(*pos);
2546 while (*pos) {
2547 pn = pneigh_get_next(seq, pn, pos);
2548 if (!pn)
2549 break;
2550 }
2551 }
2552 return *pos ? NULL : pn;
2553}
2554
2555static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2556{
2557 struct neigh_seq_state *state = seq->private;
2558 void *rc;
2559 loff_t idxpos = *pos;
2560
2561 rc = neigh_get_idx(seq, &idxpos);
2562 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2563 rc = pneigh_get_idx(seq, &idxpos);
2564
2565 return rc;
2566}
2567
2568void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
2569 __acquires(rcu_bh)
2570{
2571 struct neigh_seq_state *state = seq->private;
2572
2573 state->tbl = tbl;
2574 state->bucket = 0;
2575 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2576
2577 rcu_read_lock_bh();
2578 state->nht = rcu_dereference_bh(tbl->nht);
2579
2580 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2581}
2582EXPORT_SYMBOL(neigh_seq_start);
2583
2584void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2585{
2586 struct neigh_seq_state *state;
2587 void *rc;
2588
2589 if (v == SEQ_START_TOKEN) {
2590 rc = neigh_get_first(seq);
2591 goto out;
2592 }
2593
2594 state = seq->private;
2595 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2596 rc = neigh_get_next(seq, v, NULL);
2597 if (rc)
2598 goto out;
2599 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2600 rc = pneigh_get_first(seq);
2601 } else {
2602 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2603 rc = pneigh_get_next(seq, v, NULL);
2604 }
2605out:
2606 ++(*pos);
2607 return rc;
2608}
2609EXPORT_SYMBOL(neigh_seq_next);
2610
2611void neigh_seq_stop(struct seq_file *seq, void *v)
2612 __releases(rcu_bh)
2613{
2614 rcu_read_unlock_bh();
2615}
2616EXPORT_SYMBOL(neigh_seq_stop);
2617
2618/* statistics via seq_file */
2619
2620static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2621{
2622 struct neigh_table *tbl = seq->private;
2623 int cpu;
2624
2625 if (*pos == 0)
2626 return SEQ_START_TOKEN;
2627
2628 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2629 if (!cpu_possible(cpu))
2630 continue;
2631 *pos = cpu+1;
2632 return per_cpu_ptr(tbl->stats, cpu);
2633 }
2634 return NULL;
2635}
2636
2637static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2638{
2639 struct neigh_table *tbl = seq->private;
2640 int cpu;
2641
2642 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2643 if (!cpu_possible(cpu))
2644 continue;
2645 *pos = cpu+1;
2646 return per_cpu_ptr(tbl->stats, cpu);
2647 }
2648 return NULL;
2649}
2650
2651static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2652{
2653
2654}
2655
2656static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2657{
2658 struct neigh_table *tbl = seq->private;
2659 struct neigh_statistics *st = v;
2660
2661 if (v == SEQ_START_TOKEN) {
2662 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");
2663 return 0;
2664 }
2665
2666 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
2667 "%08lx %08lx %08lx %08lx %08lx\n",
2668 atomic_read(&tbl->entries),
2669
2670 st->allocs,
2671 st->destroys,
2672 st->hash_grows,
2673
2674 st->lookups,
2675 st->hits,
2676
2677 st->res_failed,
2678
2679 st->rcv_probes_mcast,
2680 st->rcv_probes_ucast,
2681
2682 st->periodic_gc_runs,
2683 st->forced_gc_runs,
2684 st->unres_discards
2685 );
2686
2687 return 0;
2688}
2689
2690static const struct seq_operations neigh_stat_seq_ops = {
2691 .start = neigh_stat_seq_start,
2692 .next = neigh_stat_seq_next,
2693 .stop = neigh_stat_seq_stop,
2694 .show = neigh_stat_seq_show,
2695};
2696
2697static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2698{
2699 int ret = seq_open(file, &neigh_stat_seq_ops);
2700
2701 if (!ret) {
2702 struct seq_file *sf = file->private_data;
2703 sf->private = PDE(inode)->data;
2704 }
2705 return ret;
2706};
2707
2708static const struct file_operations neigh_stat_seq_fops = {
2709 .owner = THIS_MODULE,
2710 .open = neigh_stat_seq_open,
2711 .read = seq_read,
2712 .llseek = seq_lseek,
2713 .release = seq_release,
2714};
2715
2716#endif /* CONFIG_PROC_FS */
2717
2718static inline size_t neigh_nlmsg_size(void)
2719{
2720 return NLMSG_ALIGN(sizeof(struct ndmsg))
2721 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2722 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2723 + nla_total_size(sizeof(struct nda_cacheinfo))
2724 + nla_total_size(4); /* NDA_PROBES */
2725}
2726
2727static void __neigh_notify(struct neighbour *n, int type, int flags)
2728{
2729 struct net *net = dev_net(n->dev);
2730 struct sk_buff *skb;
2731 int err = -ENOBUFS;
2732
2733 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2734 if (skb == NULL)
2735 goto errout;
2736
2737 err = neigh_fill_info(skb, n, 0, 0, type, flags);
2738 if (err < 0) {
2739 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2740 WARN_ON(err == -EMSGSIZE);
2741 kfree_skb(skb);
2742 goto errout;
2743 }
2744 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2745 return;
2746errout:
2747 if (err < 0)
2748 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2749}
2750
2751#ifdef CONFIG_ARPD
2752void neigh_app_ns(struct neighbour *n)
2753{
2754 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2755}
2756EXPORT_SYMBOL(neigh_app_ns);
2757#endif /* CONFIG_ARPD */
2758
2759#ifdef CONFIG_SYSCTL
2760
2761static int proc_unres_qlen(ctl_table *ctl, int write, void __user *buffer,
2762 size_t *lenp, loff_t *ppos)
2763{
2764 int size, ret;
2765 ctl_table tmp = *ctl;
2766
2767 tmp.data = &size;
2768 size = DIV_ROUND_UP(*(int *)ctl->data, SKB_TRUESIZE(ETH_FRAME_LEN));
2769 ret = proc_dointvec(&tmp, write, buffer, lenp, ppos);
2770 if (write && !ret)
2771 *(int *)ctl->data = size * SKB_TRUESIZE(ETH_FRAME_LEN);
2772 return ret;
2773}
2774
2775enum {
2776 NEIGH_VAR_MCAST_PROBE,
2777 NEIGH_VAR_UCAST_PROBE,
2778 NEIGH_VAR_APP_PROBE,
2779 NEIGH_VAR_RETRANS_TIME,
2780 NEIGH_VAR_BASE_REACHABLE_TIME,
2781 NEIGH_VAR_DELAY_PROBE_TIME,
2782 NEIGH_VAR_GC_STALETIME,
2783 NEIGH_VAR_QUEUE_LEN,
2784 NEIGH_VAR_QUEUE_LEN_BYTES,
2785 NEIGH_VAR_PROXY_QLEN,
2786 NEIGH_VAR_ANYCAST_DELAY,
2787 NEIGH_VAR_PROXY_DELAY,
2788 NEIGH_VAR_LOCKTIME,
2789 NEIGH_VAR_RETRANS_TIME_MS,
2790 NEIGH_VAR_BASE_REACHABLE_TIME_MS,
2791 NEIGH_VAR_GC_INTERVAL,
2792 NEIGH_VAR_GC_THRESH1,
2793 NEIGH_VAR_GC_THRESH2,
2794 NEIGH_VAR_GC_THRESH3,
2795 NEIGH_VAR_MAX
2796};
2797
2798static struct neigh_sysctl_table {
2799 struct ctl_table_header *sysctl_header;
2800 struct ctl_table neigh_vars[NEIGH_VAR_MAX + 1];
2801} neigh_sysctl_template __read_mostly = {
2802 .neigh_vars = {
2803 [NEIGH_VAR_MCAST_PROBE] = {
2804 .procname = "mcast_solicit",
2805 .maxlen = sizeof(int),
2806 .mode = 0644,
2807 .proc_handler = proc_dointvec,
2808 },
2809 [NEIGH_VAR_UCAST_PROBE] = {
2810 .procname = "ucast_solicit",
2811 .maxlen = sizeof(int),
2812 .mode = 0644,
2813 .proc_handler = proc_dointvec,
2814 },
2815 [NEIGH_VAR_APP_PROBE] = {
2816 .procname = "app_solicit",
2817 .maxlen = sizeof(int),
2818 .mode = 0644,
2819 .proc_handler = proc_dointvec,
2820 },
2821 [NEIGH_VAR_RETRANS_TIME] = {
2822 .procname = "retrans_time",
2823 .maxlen = sizeof(int),
2824 .mode = 0644,
2825 .proc_handler = proc_dointvec_userhz_jiffies,
2826 },
2827 [NEIGH_VAR_BASE_REACHABLE_TIME] = {
2828 .procname = "base_reachable_time",
2829 .maxlen = sizeof(int),
2830 .mode = 0644,
2831 .proc_handler = proc_dointvec_jiffies,
2832 },
2833 [NEIGH_VAR_DELAY_PROBE_TIME] = {
2834 .procname = "delay_first_probe_time",
2835 .maxlen = sizeof(int),
2836 .mode = 0644,
2837 .proc_handler = proc_dointvec_jiffies,
2838 },
2839 [NEIGH_VAR_GC_STALETIME] = {
2840 .procname = "gc_stale_time",
2841 .maxlen = sizeof(int),
2842 .mode = 0644,
2843 .proc_handler = proc_dointvec_jiffies,
2844 },
2845 [NEIGH_VAR_QUEUE_LEN] = {
2846 .procname = "unres_qlen",
2847 .maxlen = sizeof(int),
2848 .mode = 0644,
2849 .proc_handler = proc_unres_qlen,
2850 },
2851 [NEIGH_VAR_QUEUE_LEN_BYTES] = {
2852 .procname = "unres_qlen_bytes",
2853 .maxlen = sizeof(int),
2854 .mode = 0644,
2855 .proc_handler = proc_dointvec,
2856 },
2857 [NEIGH_VAR_PROXY_QLEN] = {
2858 .procname = "proxy_qlen",
2859 .maxlen = sizeof(int),
2860 .mode = 0644,
2861 .proc_handler = proc_dointvec,
2862 },
2863 [NEIGH_VAR_ANYCAST_DELAY] = {
2864 .procname = "anycast_delay",
2865 .maxlen = sizeof(int),
2866 .mode = 0644,
2867 .proc_handler = proc_dointvec_userhz_jiffies,
2868 },
2869 [NEIGH_VAR_PROXY_DELAY] = {
2870 .procname = "proxy_delay",
2871 .maxlen = sizeof(int),
2872 .mode = 0644,
2873 .proc_handler = proc_dointvec_userhz_jiffies,
2874 },
2875 [NEIGH_VAR_LOCKTIME] = {
2876 .procname = "locktime",
2877 .maxlen = sizeof(int),
2878 .mode = 0644,
2879 .proc_handler = proc_dointvec_userhz_jiffies,
2880 },
2881 [NEIGH_VAR_RETRANS_TIME_MS] = {
2882 .procname = "retrans_time_ms",
2883 .maxlen = sizeof(int),
2884 .mode = 0644,
2885 .proc_handler = proc_dointvec_ms_jiffies,
2886 },
2887 [NEIGH_VAR_BASE_REACHABLE_TIME_MS] = {
2888 .procname = "base_reachable_time_ms",
2889 .maxlen = sizeof(int),
2890 .mode = 0644,
2891 .proc_handler = proc_dointvec_ms_jiffies,
2892 },
2893 [NEIGH_VAR_GC_INTERVAL] = {
2894 .procname = "gc_interval",
2895 .maxlen = sizeof(int),
2896 .mode = 0644,
2897 .proc_handler = proc_dointvec_jiffies,
2898 },
2899 [NEIGH_VAR_GC_THRESH1] = {
2900 .procname = "gc_thresh1",
2901 .maxlen = sizeof(int),
2902 .mode = 0644,
2903 .proc_handler = proc_dointvec,
2904 },
2905 [NEIGH_VAR_GC_THRESH2] = {
2906 .procname = "gc_thresh2",
2907 .maxlen = sizeof(int),
2908 .mode = 0644,
2909 .proc_handler = proc_dointvec,
2910 },
2911 [NEIGH_VAR_GC_THRESH3] = {
2912 .procname = "gc_thresh3",
2913 .maxlen = sizeof(int),
2914 .mode = 0644,
2915 .proc_handler = proc_dointvec,
2916 },
2917 {},
2918 },
2919};
2920
2921int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2922 char *p_name, proc_handler *handler)
2923{
2924 struct neigh_sysctl_table *t;
2925 const char *dev_name_source = NULL;
2926 char neigh_path[ sizeof("net//neigh/") + IFNAMSIZ + IFNAMSIZ ];
2927
2928 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2929 if (!t)
2930 goto err;
2931
2932 t->neigh_vars[NEIGH_VAR_MCAST_PROBE].data = &p->mcast_probes;
2933 t->neigh_vars[NEIGH_VAR_UCAST_PROBE].data = &p->ucast_probes;
2934 t->neigh_vars[NEIGH_VAR_APP_PROBE].data = &p->app_probes;
2935 t->neigh_vars[NEIGH_VAR_RETRANS_TIME].data = &p->retrans_time;
2936 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].data = &p->base_reachable_time;
2937 t->neigh_vars[NEIGH_VAR_DELAY_PROBE_TIME].data = &p->delay_probe_time;
2938 t->neigh_vars[NEIGH_VAR_GC_STALETIME].data = &p->gc_staletime;
2939 t->neigh_vars[NEIGH_VAR_QUEUE_LEN].data = &p->queue_len_bytes;
2940 t->neigh_vars[NEIGH_VAR_QUEUE_LEN_BYTES].data = &p->queue_len_bytes;
2941 t->neigh_vars[NEIGH_VAR_PROXY_QLEN].data = &p->proxy_qlen;
2942 t->neigh_vars[NEIGH_VAR_ANYCAST_DELAY].data = &p->anycast_delay;
2943 t->neigh_vars[NEIGH_VAR_PROXY_DELAY].data = &p->proxy_delay;
2944 t->neigh_vars[NEIGH_VAR_LOCKTIME].data = &p->locktime;
2945 t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].data = &p->retrans_time;
2946 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].data = &p->base_reachable_time;
2947
2948 if (dev) {
2949 dev_name_source = dev->name;
2950 /* Terminate the table early */
2951 memset(&t->neigh_vars[NEIGH_VAR_GC_INTERVAL], 0,
2952 sizeof(t->neigh_vars[NEIGH_VAR_GC_INTERVAL]));
2953 } else {
2954 dev_name_source = "default";
2955 t->neigh_vars[NEIGH_VAR_GC_INTERVAL].data = (int *)(p + 1);
2956 t->neigh_vars[NEIGH_VAR_GC_THRESH1].data = (int *)(p + 1) + 1;
2957 t->neigh_vars[NEIGH_VAR_GC_THRESH2].data = (int *)(p + 1) + 2;
2958 t->neigh_vars[NEIGH_VAR_GC_THRESH3].data = (int *)(p + 1) + 3;
2959 }
2960
2961
2962 if (handler) {
2963 /* RetransTime */
2964 t->neigh_vars[NEIGH_VAR_RETRANS_TIME].proc_handler = handler;
2965 t->neigh_vars[NEIGH_VAR_RETRANS_TIME].extra1 = dev;
2966 /* ReachableTime */
2967 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler = handler;
2968 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].extra1 = dev;
2969 /* RetransTime (in milliseconds)*/
2970 t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].proc_handler = handler;
2971 t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].extra1 = dev;
2972 /* ReachableTime (in milliseconds) */
2973 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler = handler;
2974 t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].extra1 = dev;
2975 }
2976
2977 snprintf(neigh_path, sizeof(neigh_path), "net/%s/neigh/%s",
2978 p_name, dev_name_source);
2979 t->sysctl_header =
2980 register_net_sysctl(neigh_parms_net(p), neigh_path, t->neigh_vars);
2981 if (!t->sysctl_header)
2982 goto free;
2983
2984 p->sysctl_table = t;
2985 return 0;
2986
2987free:
2988 kfree(t);
2989err:
2990 return -ENOBUFS;
2991}
2992EXPORT_SYMBOL(neigh_sysctl_register);
2993
2994void neigh_sysctl_unregister(struct neigh_parms *p)
2995{
2996 if (p->sysctl_table) {
2997 struct neigh_sysctl_table *t = p->sysctl_table;
2998 p->sysctl_table = NULL;
2999 unregister_net_sysctl_table(t->sysctl_header);
3000 kfree(t);
3001 }
3002}
3003EXPORT_SYMBOL(neigh_sysctl_unregister);
3004
3005#endif /* CONFIG_SYSCTL */
3006
3007static int __init neigh_init(void)
3008{
3009 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, NULL);
3010 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, NULL);
3011 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, NULL);
3012
3013 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
3014 NULL);
3015 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, NULL);
3016
3017 return 0;
3018}
3019
3020subsys_initcall(neigh_init);
3021