net: ib_net pointer should depends on CONFIG_NET_NS
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / core / neighbour.c
CommitLineData
1da177e4
LT
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
1da177e4
LT
18#include <linux/types.h>
19#include <linux/kernel.h>
20#include <linux/module.h>
21#include <linux/socket.h>
1da177e4
LT
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>
457c4cbc 28#include <net/net_namespace.h>
1da177e4
LT
29#include <net/neighbour.h>
30#include <net/dst.h>
31#include <net/sock.h>
8d71740c 32#include <net/netevent.h>
a14a49d2 33#include <net/netlink.h>
1da177e4
LT
34#include <linux/rtnetlink.h>
35#include <linux/random.h>
543537bd 36#include <linux/string.h>
c3609d51 37#include <linux/log2.h>
1da177e4
LT
38
39#define NEIGH_DEBUG 1
40
41#define NEIGH_PRINTK(x...) printk(x)
42#define NEIGH_NOPRINTK(x...) do { ; } while(0)
43#define NEIGH_PRINTK0 NEIGH_PRINTK
44#define NEIGH_PRINTK1 NEIGH_NOPRINTK
45#define NEIGH_PRINTK2 NEIGH_NOPRINTK
46
47#if NEIGH_DEBUG >= 1
48#undef NEIGH_PRINTK1
49#define NEIGH_PRINTK1 NEIGH_PRINTK
50#endif
51#if NEIGH_DEBUG >= 2
52#undef NEIGH_PRINTK2
53#define NEIGH_PRINTK2 NEIGH_PRINTK
54#endif
55
56#define PNEIGH_HASHMASK 0xF
57
58static void neigh_timer_handler(unsigned long arg);
d961db35
TG
59static void __neigh_notify(struct neighbour *n, int type, int flags);
60static void neigh_update_notify(struct neighbour *neigh);
1da177e4 61static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
1da177e4
LT
62
63static struct neigh_table *neigh_tables;
45fc3b11 64#ifdef CONFIG_PROC_FS
9a32144e 65static const struct file_operations neigh_stat_seq_fops;
45fc3b11 66#endif
1da177e4
LT
67
68/*
69 Neighbour hash table buckets are protected with rwlock tbl->lock.
70
71 - All the scans/updates to hash buckets MUST be made under this lock.
72 - NOTHING clever should be made under this lock: no callbacks
73 to protocol backends, no attempts to send something to network.
74 It will result in deadlocks, if backend/driver wants to use neighbour
75 cache.
76 - If the entry requires some non-trivial actions, increase
77 its reference count and release table lock.
78
79 Neighbour entries are protected:
80 - with reference count.
81 - with rwlock neigh->lock
82
83 Reference count prevents destruction.
84
85 neigh->lock mainly serializes ll address data and its validity state.
86 However, the same lock is used to protect another entry fields:
87 - timer
88 - resolution queue
89
90 Again, nothing clever shall be made under neigh->lock,
91 the most complicated procedure, which we allow is dev->hard_header.
92 It is supposed, that dev->hard_header is simplistic and does
93 not make callbacks to neighbour tables.
94
95 The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
96 list of neighbour tables. This list is used only in process context,
97 */
98
99static DEFINE_RWLOCK(neigh_tbl_lock);
100
101static int neigh_blackhole(struct sk_buff *skb)
102{
103 kfree_skb(skb);
104 return -ENETDOWN;
105}
106
4f494554
TG
107static void neigh_cleanup_and_release(struct neighbour *neigh)
108{
109 if (neigh->parms->neigh_cleanup)
110 neigh->parms->neigh_cleanup(neigh);
111
d961db35 112 __neigh_notify(neigh, RTM_DELNEIGH, 0);
4f494554
TG
113 neigh_release(neigh);
114}
115
1da177e4
LT
116/*
117 * It is random distribution in the interval (1/2)*base...(3/2)*base.
118 * It corresponds to default IPv6 settings and is not overridable,
119 * because it is really reasonable choice.
120 */
121
122unsigned long neigh_rand_reach_time(unsigned long base)
123{
124 return (base ? (net_random() % base) + (base >> 1) : 0);
125}
0a204500 126EXPORT_SYMBOL(neigh_rand_reach_time);
1da177e4
LT
127
128
129static int neigh_forced_gc(struct neigh_table *tbl)
130{
131 int shrunk = 0;
132 int i;
133
134 NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
135
136 write_lock_bh(&tbl->lock);
137 for (i = 0; i <= tbl->hash_mask; i++) {
138 struct neighbour *n, **np;
139
140 np = &tbl->hash_buckets[i];
141 while ((n = *np) != NULL) {
142 /* Neighbour record may be discarded if:
143 * - nobody refers to it.
144 * - it is not permanent
145 */
146 write_lock(&n->lock);
147 if (atomic_read(&n->refcnt) == 1 &&
148 !(n->nud_state & NUD_PERMANENT)) {
149 *np = n->next;
150 n->dead = 1;
151 shrunk = 1;
152 write_unlock(&n->lock);
4f494554 153 neigh_cleanup_and_release(n);
1da177e4
LT
154 continue;
155 }
156 write_unlock(&n->lock);
157 np = &n->next;
158 }
159 }
160
161 tbl->last_flush = jiffies;
162
163 write_unlock_bh(&tbl->lock);
164
165 return shrunk;
166}
167
a43d8994
PE
168static void neigh_add_timer(struct neighbour *n, unsigned long when)
169{
170 neigh_hold(n);
171 if (unlikely(mod_timer(&n->timer, when))) {
172 printk("NEIGH: BUG, double timer add, state is %x\n",
173 n->nud_state);
174 dump_stack();
175 }
176}
177
1da177e4
LT
178static int neigh_del_timer(struct neighbour *n)
179{
180 if ((n->nud_state & NUD_IN_TIMER) &&
181 del_timer(&n->timer)) {
182 neigh_release(n);
183 return 1;
184 }
185 return 0;
186}
187
188static void pneigh_queue_purge(struct sk_buff_head *list)
189{
190 struct sk_buff *skb;
191
192 while ((skb = skb_dequeue(list)) != NULL) {
193 dev_put(skb->dev);
194 kfree_skb(skb);
195 }
196}
197
49636bb1 198static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
1da177e4
LT
199{
200 int i;
201
1da177e4
LT
202 for (i = 0; i <= tbl->hash_mask; i++) {
203 struct neighbour *n, **np = &tbl->hash_buckets[i];
204
205 while ((n = *np) != NULL) {
206 if (dev && n->dev != dev) {
207 np = &n->next;
208 continue;
209 }
210 *np = n->next;
211 write_lock(&n->lock);
212 neigh_del_timer(n);
213 n->dead = 1;
214
215 if (atomic_read(&n->refcnt) != 1) {
216 /* The most unpleasant situation.
217 We must destroy neighbour entry,
218 but someone still uses it.
219
220 The destroy will be delayed until
221 the last user releases us, but
222 we must kill timers etc. and move
223 it to safe state.
224 */
225 skb_queue_purge(&n->arp_queue);
226 n->output = neigh_blackhole;
227 if (n->nud_state & NUD_VALID)
228 n->nud_state = NUD_NOARP;
229 else
230 n->nud_state = NUD_NONE;
231 NEIGH_PRINTK2("neigh %p is stray.\n", n);
232 }
233 write_unlock(&n->lock);
4f494554 234 neigh_cleanup_and_release(n);
1da177e4
LT
235 }
236 }
49636bb1 237}
1da177e4 238
49636bb1
HX
239void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
240{
241 write_lock_bh(&tbl->lock);
242 neigh_flush_dev(tbl, dev);
243 write_unlock_bh(&tbl->lock);
244}
0a204500 245EXPORT_SYMBOL(neigh_changeaddr);
49636bb1
HX
246
247int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
248{
249 write_lock_bh(&tbl->lock);
250 neigh_flush_dev(tbl, dev);
1da177e4
LT
251 pneigh_ifdown(tbl, dev);
252 write_unlock_bh(&tbl->lock);
253
254 del_timer_sync(&tbl->proxy_timer);
255 pneigh_queue_purge(&tbl->proxy_queue);
256 return 0;
257}
0a204500 258EXPORT_SYMBOL(neigh_ifdown);
1da177e4
LT
259
260static struct neighbour *neigh_alloc(struct neigh_table *tbl)
261{
262 struct neighbour *n = NULL;
263 unsigned long now = jiffies;
264 int entries;
265
266 entries = atomic_inc_return(&tbl->entries) - 1;
267 if (entries >= tbl->gc_thresh3 ||
268 (entries >= tbl->gc_thresh2 &&
269 time_after(now, tbl->last_flush + 5 * HZ))) {
270 if (!neigh_forced_gc(tbl) &&
271 entries >= tbl->gc_thresh3)
272 goto out_entries;
273 }
274
c3762229 275 n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
1da177e4
LT
276 if (!n)
277 goto out_entries;
278
1da177e4
LT
279 skb_queue_head_init(&n->arp_queue);
280 rwlock_init(&n->lock);
281 n->updated = n->used = now;
282 n->nud_state = NUD_NONE;
283 n->output = neigh_blackhole;
284 n->parms = neigh_parms_clone(&tbl->parms);
b24b8a24 285 setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
1da177e4
LT
286
287 NEIGH_CACHE_STAT_INC(tbl, allocs);
288 n->tbl = tbl;
289 atomic_set(&n->refcnt, 1);
290 n->dead = 1;
291out:
292 return n;
293
294out_entries:
295 atomic_dec(&tbl->entries);
296 goto out;
297}
298
299static struct neighbour **neigh_hash_alloc(unsigned int entries)
300{
301 unsigned long size = entries * sizeof(struct neighbour *);
302 struct neighbour **ret;
303
304 if (size <= PAGE_SIZE) {
77d04bd9 305 ret = kzalloc(size, GFP_ATOMIC);
1da177e4
LT
306 } else {
307 ret = (struct neighbour **)
77d04bd9 308 __get_free_pages(GFP_ATOMIC|__GFP_ZERO, get_order(size));
1da177e4 309 }
1da177e4
LT
310 return ret;
311}
312
313static void neigh_hash_free(struct neighbour **hash, unsigned int entries)
314{
315 unsigned long size = entries * sizeof(struct neighbour *);
316
317 if (size <= PAGE_SIZE)
318 kfree(hash);
319 else
320 free_pages((unsigned long)hash, get_order(size));
321}
322
323static void neigh_hash_grow(struct neigh_table *tbl, unsigned long new_entries)
324{
325 struct neighbour **new_hash, **old_hash;
326 unsigned int i, new_hash_mask, old_entries;
327
328 NEIGH_CACHE_STAT_INC(tbl, hash_grows);
329
c3609d51 330 BUG_ON(!is_power_of_2(new_entries));
1da177e4
LT
331 new_hash = neigh_hash_alloc(new_entries);
332 if (!new_hash)
333 return;
334
335 old_entries = tbl->hash_mask + 1;
336 new_hash_mask = new_entries - 1;
337 old_hash = tbl->hash_buckets;
338
339 get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
340 for (i = 0; i < old_entries; i++) {
341 struct neighbour *n, *next;
342
343 for (n = old_hash[i]; n; n = next) {
344 unsigned int hash_val = tbl->hash(n->primary_key, n->dev);
345
346 hash_val &= new_hash_mask;
347 next = n->next;
348
349 n->next = new_hash[hash_val];
350 new_hash[hash_val] = n;
351 }
352 }
353 tbl->hash_buckets = new_hash;
354 tbl->hash_mask = new_hash_mask;
355
356 neigh_hash_free(old_hash, old_entries);
357}
358
359struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
360 struct net_device *dev)
361{
362 struct neighbour *n;
363 int key_len = tbl->key_len;
bc4bf5f3 364 u32 hash_val;
4ec93edb 365
1da177e4
LT
366 NEIGH_CACHE_STAT_INC(tbl, lookups);
367
368 read_lock_bh(&tbl->lock);
bc4bf5f3 369 hash_val = tbl->hash(pkey, dev);
c5e29460 370 for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
1da177e4
LT
371 if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
372 neigh_hold(n);
373 NEIGH_CACHE_STAT_INC(tbl, hits);
374 break;
375 }
376 }
377 read_unlock_bh(&tbl->lock);
378 return n;
379}
0a204500 380EXPORT_SYMBOL(neigh_lookup);
1da177e4 381
426b5303
EB
382struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
383 const void *pkey)
1da177e4
LT
384{
385 struct neighbour *n;
386 int key_len = tbl->key_len;
bc4bf5f3 387 u32 hash_val;
1da177e4
LT
388
389 NEIGH_CACHE_STAT_INC(tbl, lookups);
390
391 read_lock_bh(&tbl->lock);
bc4bf5f3 392 hash_val = tbl->hash(pkey, NULL);
c5e29460 393 for (n = tbl->hash_buckets[hash_val & tbl->hash_mask]; n; n = n->next) {
426b5303 394 if (!memcmp(n->primary_key, pkey, key_len) &&
878628fb 395 net_eq(dev_net(n->dev), net)) {
1da177e4
LT
396 neigh_hold(n);
397 NEIGH_CACHE_STAT_INC(tbl, hits);
398 break;
399 }
400 }
401 read_unlock_bh(&tbl->lock);
402 return n;
403}
0a204500 404EXPORT_SYMBOL(neigh_lookup_nodev);
1da177e4
LT
405
406struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
407 struct net_device *dev)
408{
409 u32 hash_val;
410 int key_len = tbl->key_len;
411 int error;
412 struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
413
414 if (!n) {
415 rc = ERR_PTR(-ENOBUFS);
416 goto out;
417 }
418
419 memcpy(n->primary_key, pkey, key_len);
420 n->dev = dev;
421 dev_hold(dev);
422
423 /* Protocol specific setup. */
424 if (tbl->constructor && (error = tbl->constructor(n)) < 0) {
425 rc = ERR_PTR(error);
426 goto out_neigh_release;
427 }
428
429 /* Device specific setup. */
430 if (n->parms->neigh_setup &&
431 (error = n->parms->neigh_setup(n)) < 0) {
432 rc = ERR_PTR(error);
433 goto out_neigh_release;
434 }
435
436 n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
437
438 write_lock_bh(&tbl->lock);
439
440 if (atomic_read(&tbl->entries) > (tbl->hash_mask + 1))
441 neigh_hash_grow(tbl, (tbl->hash_mask + 1) << 1);
442
443 hash_val = tbl->hash(pkey, dev) & tbl->hash_mask;
444
445 if (n->parms->dead) {
446 rc = ERR_PTR(-EINVAL);
447 goto out_tbl_unlock;
448 }
449
450 for (n1 = tbl->hash_buckets[hash_val]; n1; n1 = n1->next) {
451 if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
452 neigh_hold(n1);
453 rc = n1;
454 goto out_tbl_unlock;
455 }
456 }
457
458 n->next = tbl->hash_buckets[hash_val];
459 tbl->hash_buckets[hash_val] = n;
460 n->dead = 0;
461 neigh_hold(n);
462 write_unlock_bh(&tbl->lock);
463 NEIGH_PRINTK2("neigh %p is created.\n", n);
464 rc = n;
465out:
466 return rc;
467out_tbl_unlock:
468 write_unlock_bh(&tbl->lock);
469out_neigh_release:
470 neigh_release(n);
471 goto out;
472}
0a204500 473EXPORT_SYMBOL(neigh_create);
1da177e4 474
be01d655 475static u32 pneigh_hash(const void *pkey, int key_len)
fa86d322 476{
fa86d322 477 u32 hash_val = *(u32 *)(pkey + key_len - 4);
fa86d322
PE
478 hash_val ^= (hash_val >> 16);
479 hash_val ^= hash_val >> 8;
480 hash_val ^= hash_val >> 4;
481 hash_val &= PNEIGH_HASHMASK;
be01d655
YH
482 return hash_val;
483}
fa86d322 484
be01d655
YH
485static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
486 struct net *net,
487 const void *pkey,
488 int key_len,
489 struct net_device *dev)
490{
491 while (n) {
fa86d322 492 if (!memcmp(n->key, pkey, key_len) &&
be01d655 493 net_eq(pneigh_net(n), net) &&
fa86d322 494 (n->dev == dev || !n->dev))
be01d655
YH
495 return n;
496 n = n->next;
fa86d322 497 }
be01d655
YH
498 return NULL;
499}
fa86d322 500
be01d655
YH
501struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
502 struct net *net, const void *pkey, struct net_device *dev)
503{
504 int key_len = tbl->key_len;
505 u32 hash_val = pneigh_hash(pkey, key_len);
506
507 return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
508 net, pkey, key_len, dev);
fa86d322 509}
0a204500 510EXPORT_SYMBOL_GPL(__pneigh_lookup);
fa86d322 511
426b5303
EB
512struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
513 struct net *net, const void *pkey,
1da177e4
LT
514 struct net_device *dev, int creat)
515{
516 struct pneigh_entry *n;
517 int key_len = tbl->key_len;
be01d655 518 u32 hash_val = pneigh_hash(pkey, key_len);
1da177e4
LT
519
520 read_lock_bh(&tbl->lock);
be01d655
YH
521 n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
522 net, pkey, key_len, dev);
1da177e4 523 read_unlock_bh(&tbl->lock);
be01d655
YH
524
525 if (n || !creat)
1da177e4
LT
526 goto out;
527
4ae28944
PE
528 ASSERT_RTNL();
529
1da177e4
LT
530 n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
531 if (!n)
532 goto out;
533
57da52c1 534#ifdef CONFIG_NET_NS
426b5303 535 n->net = hold_net(net);
57da52c1 536#endif
1da177e4
LT
537 memcpy(n->key, pkey, key_len);
538 n->dev = dev;
539 if (dev)
540 dev_hold(dev);
541
542 if (tbl->pconstructor && tbl->pconstructor(n)) {
543 if (dev)
544 dev_put(dev);
da12f735 545 release_net(net);
1da177e4
LT
546 kfree(n);
547 n = NULL;
548 goto out;
549 }
550
551 write_lock_bh(&tbl->lock);
552 n->next = tbl->phash_buckets[hash_val];
553 tbl->phash_buckets[hash_val] = n;
554 write_unlock_bh(&tbl->lock);
555out:
556 return n;
557}
0a204500 558EXPORT_SYMBOL(pneigh_lookup);
1da177e4
LT
559
560
426b5303 561int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
1da177e4
LT
562 struct net_device *dev)
563{
564 struct pneigh_entry *n, **np;
565 int key_len = tbl->key_len;
be01d655 566 u32 hash_val = pneigh_hash(pkey, key_len);
1da177e4
LT
567
568 write_lock_bh(&tbl->lock);
569 for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
570 np = &n->next) {
426b5303 571 if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
878628fb 572 net_eq(pneigh_net(n), net)) {
1da177e4
LT
573 *np = n->next;
574 write_unlock_bh(&tbl->lock);
575 if (tbl->pdestructor)
576 tbl->pdestructor(n);
577 if (n->dev)
578 dev_put(n->dev);
57da52c1 579 release_net(pneigh_net(n));
1da177e4
LT
580 kfree(n);
581 return 0;
582 }
583 }
584 write_unlock_bh(&tbl->lock);
585 return -ENOENT;
586}
587
588static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
589{
590 struct pneigh_entry *n, **np;
591 u32 h;
592
593 for (h = 0; h <= PNEIGH_HASHMASK; h++) {
594 np = &tbl->phash_buckets[h];
595 while ((n = *np) != NULL) {
596 if (!dev || n->dev == dev) {
597 *np = n->next;
598 if (tbl->pdestructor)
599 tbl->pdestructor(n);
600 if (n->dev)
601 dev_put(n->dev);
57da52c1 602 release_net(pneigh_net(n));
1da177e4
LT
603 kfree(n);
604 continue;
605 }
606 np = &n->next;
607 }
608 }
609 return -ENOENT;
610}
611
06f0511d
DL
612static void neigh_parms_destroy(struct neigh_parms *parms);
613
614static inline void neigh_parms_put(struct neigh_parms *parms)
615{
616 if (atomic_dec_and_test(&parms->refcnt))
617 neigh_parms_destroy(parms);
618}
1da177e4
LT
619
620/*
621 * neighbour must already be out of the table;
622 *
623 */
624void neigh_destroy(struct neighbour *neigh)
625{
626 struct hh_cache *hh;
627
628 NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
629
630 if (!neigh->dead) {
631 printk(KERN_WARNING
632 "Destroying alive neighbour %p\n", neigh);
633 dump_stack();
634 return;
635 }
636
637 if (neigh_del_timer(neigh))
638 printk(KERN_WARNING "Impossible event.\n");
639
640 while ((hh = neigh->hh) != NULL) {
641 neigh->hh = hh->hh_next;
642 hh->hh_next = NULL;
3644f0ce
SH
643
644 write_seqlock_bh(&hh->hh_lock);
1da177e4 645 hh->hh_output = neigh_blackhole;
3644f0ce 646 write_sequnlock_bh(&hh->hh_lock);
1da177e4
LT
647 if (atomic_dec_and_test(&hh->hh_refcnt))
648 kfree(hh);
649 }
650
1da177e4
LT
651 skb_queue_purge(&neigh->arp_queue);
652
653 dev_put(neigh->dev);
654 neigh_parms_put(neigh->parms);
655
656 NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
657
658 atomic_dec(&neigh->tbl->entries);
659 kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
660}
0a204500 661EXPORT_SYMBOL(neigh_destroy);
1da177e4
LT
662
663/* Neighbour state is suspicious;
664 disable fast path.
665
666 Called with write_locked neigh.
667 */
668static void neigh_suspect(struct neighbour *neigh)
669{
670 struct hh_cache *hh;
671
672 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
673
674 neigh->output = neigh->ops->output;
675
676 for (hh = neigh->hh; hh; hh = hh->hh_next)
677 hh->hh_output = neigh->ops->output;
678}
679
680/* Neighbour state is OK;
681 enable fast path.
682
683 Called with write_locked neigh.
684 */
685static void neigh_connect(struct neighbour *neigh)
686{
687 struct hh_cache *hh;
688
689 NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
690
691 neigh->output = neigh->ops->connected_output;
692
693 for (hh = neigh->hh; hh; hh = hh->hh_next)
694 hh->hh_output = neigh->ops->hh_output;
695}
696
697static void neigh_periodic_timer(unsigned long arg)
698{
699 struct neigh_table *tbl = (struct neigh_table *)arg;
700 struct neighbour *n, **np;
701 unsigned long expire, now = jiffies;
702
703 NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
704
705 write_lock(&tbl->lock);
706
707 /*
708 * periodically recompute ReachableTime from random function
709 */
710
711 if (time_after(now, tbl->last_rand + 300 * HZ)) {
712 struct neigh_parms *p;
713 tbl->last_rand = now;
714 for (p = &tbl->parms; p; p = p->next)
715 p->reachable_time =
716 neigh_rand_reach_time(p->base_reachable_time);
717 }
718
719 np = &tbl->hash_buckets[tbl->hash_chain_gc];
720 tbl->hash_chain_gc = ((tbl->hash_chain_gc + 1) & tbl->hash_mask);
721
722 while ((n = *np) != NULL) {
723 unsigned int state;
724
725 write_lock(&n->lock);
726
727 state = n->nud_state;
728 if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
729 write_unlock(&n->lock);
730 goto next_elt;
731 }
732
733 if (time_before(n->used, n->confirmed))
734 n->used = n->confirmed;
735
736 if (atomic_read(&n->refcnt) == 1 &&
737 (state == NUD_FAILED ||
738 time_after(now, n->used + n->parms->gc_staletime))) {
739 *np = n->next;
740 n->dead = 1;
741 write_unlock(&n->lock);
4f494554 742 neigh_cleanup_and_release(n);
1da177e4
LT
743 continue;
744 }
745 write_unlock(&n->lock);
746
747next_elt:
748 np = &n->next;
749 }
750
4ec93edb
YH
751 /* Cycle through all hash buckets every base_reachable_time/2 ticks.
752 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
753 * base_reachable_time.
1da177e4
LT
754 */
755 expire = tbl->parms.base_reachable_time >> 1;
756 expire /= (tbl->hash_mask + 1);
757 if (!expire)
758 expire = 1;
759
f5a6e01c
AV
760 if (expire>HZ)
761 mod_timer(&tbl->gc_timer, round_jiffies(now + expire));
762 else
763 mod_timer(&tbl->gc_timer, now + expire);
1da177e4
LT
764
765 write_unlock(&tbl->lock);
766}
767
768static __inline__ int neigh_max_probes(struct neighbour *n)
769{
770 struct neigh_parms *p = n->parms;
771 return (n->nud_state & NUD_PROBE ?
772 p->ucast_probes :
773 p->ucast_probes + p->app_probes + p->mcast_probes);
774}
775
1da177e4
LT
776/* Called when a timer expires for a neighbour entry. */
777
778static void neigh_timer_handler(unsigned long arg)
779{
780 unsigned long now, next;
781 struct neighbour *neigh = (struct neighbour *)arg;
782 unsigned state;
783 int notify = 0;
784
785 write_lock(&neigh->lock);
786
787 state = neigh->nud_state;
788 now = jiffies;
789 next = now + HZ;
790
791 if (!(state & NUD_IN_TIMER)) {
792#ifndef CONFIG_SMP
793 printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
794#endif
795 goto out;
796 }
797
798 if (state & NUD_REACHABLE) {
4ec93edb 799 if (time_before_eq(now,
1da177e4
LT
800 neigh->confirmed + neigh->parms->reachable_time)) {
801 NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
802 next = neigh->confirmed + neigh->parms->reachable_time;
803 } else if (time_before_eq(now,
804 neigh->used + neigh->parms->delay_probe_time)) {
805 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
806 neigh->nud_state = NUD_DELAY;
955aaa2f 807 neigh->updated = jiffies;
1da177e4
LT
808 neigh_suspect(neigh);
809 next = now + neigh->parms->delay_probe_time;
810 } else {
811 NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
812 neigh->nud_state = NUD_STALE;
955aaa2f 813 neigh->updated = jiffies;
1da177e4 814 neigh_suspect(neigh);
8d71740c 815 notify = 1;
1da177e4
LT
816 }
817 } else if (state & NUD_DELAY) {
4ec93edb 818 if (time_before_eq(now,
1da177e4
LT
819 neigh->confirmed + neigh->parms->delay_probe_time)) {
820 NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
821 neigh->nud_state = NUD_REACHABLE;
955aaa2f 822 neigh->updated = jiffies;
1da177e4 823 neigh_connect(neigh);
8d71740c 824 notify = 1;
1da177e4
LT
825 next = neigh->confirmed + neigh->parms->reachable_time;
826 } else {
827 NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
828 neigh->nud_state = NUD_PROBE;
955aaa2f 829 neigh->updated = jiffies;
1da177e4
LT
830 atomic_set(&neigh->probes, 0);
831 next = now + neigh->parms->retrans_time;
832 }
833 } else {
834 /* NUD_PROBE|NUD_INCOMPLETE */
835 next = now + neigh->parms->retrans_time;
836 }
837
838 if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
839 atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
840 struct sk_buff *skb;
841
842 neigh->nud_state = NUD_FAILED;
955aaa2f 843 neigh->updated = jiffies;
1da177e4
LT
844 notify = 1;
845 NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
846 NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
847
848 /* It is very thin place. report_unreachable is very complicated
849 routine. Particularly, it can hit the same neighbour entry!
850
851 So that, we try to be accurate and avoid dead loop. --ANK
852 */
853 while (neigh->nud_state == NUD_FAILED &&
854 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
855 write_unlock(&neigh->lock);
856 neigh->ops->error_report(neigh, skb);
857 write_lock(&neigh->lock);
858 }
859 skb_queue_purge(&neigh->arp_queue);
860 }
861
862 if (neigh->nud_state & NUD_IN_TIMER) {
1da177e4
LT
863 if (time_before(next, jiffies + HZ/2))
864 next = jiffies + HZ/2;
6fb9974f
HX
865 if (!mod_timer(&neigh->timer, next))
866 neigh_hold(neigh);
1da177e4
LT
867 }
868 if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
869 struct sk_buff *skb = skb_peek(&neigh->arp_queue);
9ff56607
DM
870 /* keep skb alive even if arp_queue overflows */
871 if (skb)
7e36763b 872 skb = skb_copy(skb, GFP_ATOMIC);
9ff56607 873 write_unlock(&neigh->lock);
1da177e4
LT
874 neigh->ops->solicit(neigh, skb);
875 atomic_inc(&neigh->probes);
9ff56607
DM
876 if (skb)
877 kfree_skb(skb);
878 } else {
69cc64d8 879out:
9ff56607
DM
880 write_unlock(&neigh->lock);
881 }
d961db35 882
8d71740c 883 if (notify)
d961db35 884 neigh_update_notify(neigh);
1da177e4 885
1da177e4
LT
886 neigh_release(neigh);
887}
888
889int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
890{
891 int rc;
892 unsigned long now;
893
894 write_lock_bh(&neigh->lock);
895
896 rc = 0;
897 if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
898 goto out_unlock_bh;
899
900 now = jiffies;
4ec93edb 901
1da177e4
LT
902 if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
903 if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
904 atomic_set(&neigh->probes, neigh->parms->ucast_probes);
905 neigh->nud_state = NUD_INCOMPLETE;
955aaa2f 906 neigh->updated = jiffies;
667347f1 907 neigh_add_timer(neigh, now + 1);
1da177e4
LT
908 } else {
909 neigh->nud_state = NUD_FAILED;
955aaa2f 910 neigh->updated = jiffies;
1da177e4
LT
911 write_unlock_bh(&neigh->lock);
912
913 if (skb)
914 kfree_skb(skb);
915 return 1;
916 }
917 } else if (neigh->nud_state & NUD_STALE) {
918 NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
1da177e4 919 neigh->nud_state = NUD_DELAY;
955aaa2f 920 neigh->updated = jiffies;
667347f1
DM
921 neigh_add_timer(neigh,
922 jiffies + neigh->parms->delay_probe_time);
1da177e4
LT
923 }
924
925 if (neigh->nud_state == NUD_INCOMPLETE) {
926 if (skb) {
927 if (skb_queue_len(&neigh->arp_queue) >=
928 neigh->parms->queue_len) {
929 struct sk_buff *buff;
f72051b0 930 buff = __skb_dequeue(&neigh->arp_queue);
1da177e4 931 kfree_skb(buff);
9a6d276e 932 NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
1da177e4
LT
933 }
934 __skb_queue_tail(&neigh->arp_queue, skb);
935 }
936 rc = 1;
937 }
938out_unlock_bh:
939 write_unlock_bh(&neigh->lock);
940 return rc;
941}
0a204500 942EXPORT_SYMBOL(__neigh_event_send);
1da177e4 943
e92b43a3 944static void neigh_update_hhs(struct neighbour *neigh)
1da177e4
LT
945{
946 struct hh_cache *hh;
3b04ddde
SH
947 void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
948 = neigh->dev->header_ops->cache_update;
1da177e4
LT
949
950 if (update) {
951 for (hh = neigh->hh; hh; hh = hh->hh_next) {
3644f0ce 952 write_seqlock_bh(&hh->hh_lock);
1da177e4 953 update(hh, neigh->dev, neigh->ha);
3644f0ce 954 write_sequnlock_bh(&hh->hh_lock);
1da177e4
LT
955 }
956 }
957}
958
959
960
961/* Generic update routine.
962 -- lladdr is new lladdr or NULL, if it is not supplied.
963 -- new is new state.
964 -- flags
965 NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
966 if it is different.
967 NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
4ec93edb 968 lladdr instead of overriding it
1da177e4
LT
969 if it is different.
970 It also allows to retain current state
971 if lladdr is unchanged.
972 NEIGH_UPDATE_F_ADMIN means that the change is administrative.
973
4ec93edb 974 NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1da177e4
LT
975 NTF_ROUTER flag.
976 NEIGH_UPDATE_F_ISROUTER indicates if the neighbour is known as
977 a router.
978
979 Caller MUST hold reference count on the entry.
980 */
981
982int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
983 u32 flags)
984{
985 u8 old;
986 int err;
1da177e4 987 int notify = 0;
1da177e4
LT
988 struct net_device *dev;
989 int update_isrouter = 0;
990
991 write_lock_bh(&neigh->lock);
992
993 dev = neigh->dev;
994 old = neigh->nud_state;
995 err = -EPERM;
996
4ec93edb 997 if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1da177e4
LT
998 (old & (NUD_NOARP | NUD_PERMANENT)))
999 goto out;
1000
1001 if (!(new & NUD_VALID)) {
1002 neigh_del_timer(neigh);
1003 if (old & NUD_CONNECTED)
1004 neigh_suspect(neigh);
1005 neigh->nud_state = new;
1006 err = 0;
1da177e4 1007 notify = old & NUD_VALID;
1da177e4
LT
1008 goto out;
1009 }
1010
1011 /* Compare new lladdr with cached one */
1012 if (!dev->addr_len) {
1013 /* First case: device needs no address. */
1014 lladdr = neigh->ha;
1015 } else if (lladdr) {
1016 /* The second case: if something is already cached
1017 and a new address is proposed:
1018 - compare new & old
1019 - if they are different, check override flag
1020 */
4ec93edb 1021 if ((old & NUD_VALID) &&
1da177e4
LT
1022 !memcmp(lladdr, neigh->ha, dev->addr_len))
1023 lladdr = neigh->ha;
1024 } else {
1025 /* No address is supplied; if we know something,
1026 use it, otherwise discard the request.
1027 */
1028 err = -EINVAL;
1029 if (!(old & NUD_VALID))
1030 goto out;
1031 lladdr = neigh->ha;
1032 }
1033
1034 if (new & NUD_CONNECTED)
1035 neigh->confirmed = jiffies;
1036 neigh->updated = jiffies;
1037
1038 /* If entry was valid and address is not changed,
1039 do not change entry state, if new one is STALE.
1040 */
1041 err = 0;
1042 update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1043 if (old & NUD_VALID) {
1044 if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1045 update_isrouter = 0;
1046 if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1047 (old & NUD_CONNECTED)) {
1048 lladdr = neigh->ha;
1049 new = NUD_STALE;
1050 } else
1051 goto out;
1052 } else {
1053 if (lladdr == neigh->ha && new == NUD_STALE &&
1054 ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1055 (old & NUD_CONNECTED))
1056 )
1057 new = old;
1058 }
1059 }
1060
1061 if (new != old) {
1062 neigh_del_timer(neigh);
a43d8994 1063 if (new & NUD_IN_TIMER)
4ec93edb
YH
1064 neigh_add_timer(neigh, (jiffies +
1065 ((new & NUD_REACHABLE) ?
667347f1
DM
1066 neigh->parms->reachable_time :
1067 0)));
1da177e4
LT
1068 neigh->nud_state = new;
1069 }
1070
1071 if (lladdr != neigh->ha) {
1072 memcpy(&neigh->ha, lladdr, dev->addr_len);
1073 neigh_update_hhs(neigh);
1074 if (!(new & NUD_CONNECTED))
1075 neigh->confirmed = jiffies -
1076 (neigh->parms->base_reachable_time << 1);
1da177e4 1077 notify = 1;
1da177e4
LT
1078 }
1079 if (new == old)
1080 goto out;
1081 if (new & NUD_CONNECTED)
1082 neigh_connect(neigh);
1083 else
1084 neigh_suspect(neigh);
1085 if (!(old & NUD_VALID)) {
1086 struct sk_buff *skb;
1087
1088 /* Again: avoid dead loop if something went wrong */
1089
1090 while (neigh->nud_state & NUD_VALID &&
1091 (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1092 struct neighbour *n1 = neigh;
1093 write_unlock_bh(&neigh->lock);
1094 /* On shaper/eql skb->dst->neighbour != neigh :( */
1095 if (skb->dst && skb->dst->neighbour)
1096 n1 = skb->dst->neighbour;
1097 n1->output(skb);
1098 write_lock_bh(&neigh->lock);
1099 }
1100 skb_queue_purge(&neigh->arp_queue);
1101 }
1102out:
1103 if (update_isrouter) {
1104 neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1105 (neigh->flags | NTF_ROUTER) :
1106 (neigh->flags & ~NTF_ROUTER);
1107 }
1108 write_unlock_bh(&neigh->lock);
8d71740c
TT
1109
1110 if (notify)
d961db35
TG
1111 neigh_update_notify(neigh);
1112
1da177e4
LT
1113 return err;
1114}
0a204500 1115EXPORT_SYMBOL(neigh_update);
1da177e4
LT
1116
1117struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1118 u8 *lladdr, void *saddr,
1119 struct net_device *dev)
1120{
1121 struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1122 lladdr || !dev->addr_len);
1123 if (neigh)
4ec93edb 1124 neigh_update(neigh, lladdr, NUD_STALE,
1da177e4
LT
1125 NEIGH_UPDATE_F_OVERRIDE);
1126 return neigh;
1127}
0a204500 1128EXPORT_SYMBOL(neigh_event_ns);
1da177e4
LT
1129
1130static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst,
d77072ec 1131 __be16 protocol)
1da177e4
LT
1132{
1133 struct hh_cache *hh;
1134 struct net_device *dev = dst->dev;
1135
1136 for (hh = n->hh; hh; hh = hh->hh_next)
1137 if (hh->hh_type == protocol)
1138 break;
1139
77d04bd9 1140 if (!hh && (hh = kzalloc(sizeof(*hh), GFP_ATOMIC)) != NULL) {
3644f0ce 1141 seqlock_init(&hh->hh_lock);
1da177e4
LT
1142 hh->hh_type = protocol;
1143 atomic_set(&hh->hh_refcnt, 0);
1144 hh->hh_next = NULL;
3b04ddde
SH
1145
1146 if (dev->header_ops->cache(n, hh)) {
1da177e4
LT
1147 kfree(hh);
1148 hh = NULL;
1149 } else {
1150 atomic_inc(&hh->hh_refcnt);
1151 hh->hh_next = n->hh;
1152 n->hh = hh;
1153 if (n->nud_state & NUD_CONNECTED)
1154 hh->hh_output = n->ops->hh_output;
1155 else
1156 hh->hh_output = n->ops->output;
1157 }
1158 }
1159 if (hh) {
1160 atomic_inc(&hh->hh_refcnt);
1161 dst->hh = hh;
1162 }
1163}
1164
1165/* This function can be used in contexts, where only old dev_queue_xmit
1166 worked, f.e. if you want to override normal output path (eql, shaper),
1167 but resolution is not made yet.
1168 */
1169
1170int neigh_compat_output(struct sk_buff *skb)
1171{
1172 struct net_device *dev = skb->dev;
1173
bbe735e4 1174 __skb_pull(skb, skb_network_offset(skb));
1da177e4 1175
0c4e8581
SH
1176 if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1177 skb->len) < 0 &&
3b04ddde 1178 dev->header_ops->rebuild(skb))
1da177e4
LT
1179 return 0;
1180
1181 return dev_queue_xmit(skb);
1182}
0a204500 1183EXPORT_SYMBOL(neigh_compat_output);
1da177e4
LT
1184
1185/* Slow and careful. */
1186
1187int neigh_resolve_output(struct sk_buff *skb)
1188{
1189 struct dst_entry *dst = skb->dst;
1190 struct neighbour *neigh;
1191 int rc = 0;
1192
1193 if (!dst || !(neigh = dst->neighbour))
1194 goto discard;
1195
bbe735e4 1196 __skb_pull(skb, skb_network_offset(skb));
1da177e4
LT
1197
1198 if (!neigh_event_send(neigh, skb)) {
1199 int err;
1200 struct net_device *dev = neigh->dev;
3b04ddde 1201 if (dev->header_ops->cache && !dst->hh) {
1da177e4
LT
1202 write_lock_bh(&neigh->lock);
1203 if (!dst->hh)
1204 neigh_hh_init(neigh, dst, dst->ops->protocol);
0c4e8581
SH
1205 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1206 neigh->ha, NULL, skb->len);
1da177e4
LT
1207 write_unlock_bh(&neigh->lock);
1208 } else {
1209 read_lock_bh(&neigh->lock);
0c4e8581
SH
1210 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1211 neigh->ha, NULL, skb->len);
1da177e4
LT
1212 read_unlock_bh(&neigh->lock);
1213 }
1214 if (err >= 0)
1215 rc = neigh->ops->queue_xmit(skb);
1216 else
1217 goto out_kfree_skb;
1218 }
1219out:
1220 return rc;
1221discard:
1222 NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1223 dst, dst ? dst->neighbour : NULL);
1224out_kfree_skb:
1225 rc = -EINVAL;
1226 kfree_skb(skb);
1227 goto out;
1228}
0a204500 1229EXPORT_SYMBOL(neigh_resolve_output);
1da177e4
LT
1230
1231/* As fast as possible without hh cache */
1232
1233int neigh_connected_output(struct sk_buff *skb)
1234{
1235 int err;
1236 struct dst_entry *dst = skb->dst;
1237 struct neighbour *neigh = dst->neighbour;
1238 struct net_device *dev = neigh->dev;
1239
bbe735e4 1240 __skb_pull(skb, skb_network_offset(skb));
1da177e4
LT
1241
1242 read_lock_bh(&neigh->lock);
0c4e8581
SH
1243 err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1244 neigh->ha, NULL, skb->len);
1da177e4
LT
1245 read_unlock_bh(&neigh->lock);
1246 if (err >= 0)
1247 err = neigh->ops->queue_xmit(skb);
1248 else {
1249 err = -EINVAL;
1250 kfree_skb(skb);
1251 }
1252 return err;
1253}
0a204500 1254EXPORT_SYMBOL(neigh_connected_output);
1da177e4
LT
1255
1256static void neigh_proxy_process(unsigned long arg)
1257{
1258 struct neigh_table *tbl = (struct neigh_table *)arg;
1259 long sched_next = 0;
1260 unsigned long now = jiffies;
f72051b0 1261 struct sk_buff *skb, *n;
1da177e4
LT
1262
1263 spin_lock(&tbl->proxy_queue.lock);
1264
f72051b0
DM
1265 skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1266 long tdif = NEIGH_CB(skb)->sched_next - now;
1da177e4 1267
1da177e4 1268 if (tdif <= 0) {
f72051b0
DM
1269 struct net_device *dev = skb->dev;
1270 __skb_unlink(skb, &tbl->proxy_queue);
1da177e4 1271 if (tbl->proxy_redo && netif_running(dev))
f72051b0 1272 tbl->proxy_redo(skb);
1da177e4 1273 else
f72051b0 1274 kfree_skb(skb);
1da177e4
LT
1275
1276 dev_put(dev);
1277 } else if (!sched_next || tdif < sched_next)
1278 sched_next = tdif;
1279 }
1280 del_timer(&tbl->proxy_timer);
1281 if (sched_next)
1282 mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1283 spin_unlock(&tbl->proxy_queue.lock);
1284}
1285
1286void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1287 struct sk_buff *skb)
1288{
1289 unsigned long now = jiffies;
1290 unsigned long sched_next = now + (net_random() % p->proxy_delay);
1291
1292 if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1293 kfree_skb(skb);
1294 return;
1295 }
a61bbcf2
PM
1296
1297 NEIGH_CB(skb)->sched_next = sched_next;
1298 NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1da177e4
LT
1299
1300 spin_lock(&tbl->proxy_queue.lock);
1301 if (del_timer(&tbl->proxy_timer)) {
1302 if (time_before(tbl->proxy_timer.expires, sched_next))
1303 sched_next = tbl->proxy_timer.expires;
1304 }
1305 dst_release(skb->dst);
1306 skb->dst = NULL;
1307 dev_hold(skb->dev);
1308 __skb_queue_tail(&tbl->proxy_queue, skb);
1309 mod_timer(&tbl->proxy_timer, sched_next);
1310 spin_unlock(&tbl->proxy_queue.lock);
1311}
0a204500 1312EXPORT_SYMBOL(pneigh_enqueue);
1da177e4 1313
426b5303
EB
1314static inline struct neigh_parms *lookup_neigh_params(struct neigh_table *tbl,
1315 struct net *net, int ifindex)
1316{
1317 struct neigh_parms *p;
1318
1319 for (p = &tbl->parms; p; p = p->next) {
878628fb 1320 if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
426b5303
EB
1321 (!p->dev && !ifindex))
1322 return p;
1323 }
1324
1325 return NULL;
1326}
1da177e4
LT
1327
1328struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1329 struct neigh_table *tbl)
1330{
426b5303
EB
1331 struct neigh_parms *p, *ref;
1332 struct net *net;
1333
c346dca1 1334 net = dev_net(dev);
426b5303
EB
1335 ref = lookup_neigh_params(tbl, net, 0);
1336 if (!ref)
1337 return NULL;
1da177e4 1338
426b5303 1339 p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1da177e4 1340 if (p) {
1da177e4
LT
1341 p->tbl = tbl;
1342 atomic_set(&p->refcnt, 1);
1da177e4
LT
1343 p->reachable_time =
1344 neigh_rand_reach_time(p->base_reachable_time);
c7fb64db 1345
486b51d3
DL
1346 if (dev->neigh_setup && dev->neigh_setup(dev, p)) {
1347 kfree(p);
1348 return NULL;
1da177e4 1349 }
486b51d3
DL
1350
1351 dev_hold(dev);
1352 p->dev = dev;
57da52c1 1353#ifdef CONFIG_NET_NS
426b5303 1354 p->net = hold_net(net);
57da52c1 1355#endif
1da177e4
LT
1356 p->sysctl_table = NULL;
1357 write_lock_bh(&tbl->lock);
1358 p->next = tbl->parms.next;
1359 tbl->parms.next = p;
1360 write_unlock_bh(&tbl->lock);
1361 }
1362 return p;
1363}
0a204500 1364EXPORT_SYMBOL(neigh_parms_alloc);
1da177e4
LT
1365
1366static void neigh_rcu_free_parms(struct rcu_head *head)
1367{
1368 struct neigh_parms *parms =
1369 container_of(head, struct neigh_parms, rcu_head);
1370
1371 neigh_parms_put(parms);
1372}
1373
1374void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1375{
1376 struct neigh_parms **p;
1377
1378 if (!parms || parms == &tbl->parms)
1379 return;
1380 write_lock_bh(&tbl->lock);
1381 for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1382 if (*p == parms) {
1383 *p = parms->next;
1384 parms->dead = 1;
1385 write_unlock_bh(&tbl->lock);
cecbb639
DM
1386 if (parms->dev)
1387 dev_put(parms->dev);
1da177e4
LT
1388 call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1389 return;
1390 }
1391 }
1392 write_unlock_bh(&tbl->lock);
1393 NEIGH_PRINTK1("neigh_parms_release: not found\n");
1394}
0a204500 1395EXPORT_SYMBOL(neigh_parms_release);
1da177e4 1396
06f0511d 1397static void neigh_parms_destroy(struct neigh_parms *parms)
1da177e4 1398{
57da52c1 1399 release_net(neigh_parms_net(parms));
1da177e4
LT
1400 kfree(parms);
1401}
1402
c2ecba71
PE
1403static struct lock_class_key neigh_table_proxy_queue_class;
1404
bd89efc5 1405void neigh_table_init_no_netlink(struct neigh_table *tbl)
1da177e4
LT
1406{
1407 unsigned long now = jiffies;
1408 unsigned long phsize;
1409
57da52c1 1410#ifdef CONFIG_NET_NS
426b5303 1411 tbl->parms.net = &init_net;
57da52c1 1412#endif
1da177e4 1413 atomic_set(&tbl->parms.refcnt, 1);
1da177e4
LT
1414 tbl->parms.reachable_time =
1415 neigh_rand_reach_time(tbl->parms.base_reachable_time);
1416
1417 if (!tbl->kmem_cachep)
e5d679f3
AD
1418 tbl->kmem_cachep =
1419 kmem_cache_create(tbl->id, tbl->entry_size, 0,
1420 SLAB_HWCACHE_ALIGN|SLAB_PANIC,
20c2df83 1421 NULL);
1da177e4
LT
1422 tbl->stats = alloc_percpu(struct neigh_statistics);
1423 if (!tbl->stats)
1424 panic("cannot create neighbour cache statistics");
4ec93edb 1425
1da177e4 1426#ifdef CONFIG_PROC_FS
9b739ba5
AD
1427 if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1428 &neigh_stat_seq_fops, tbl))
1da177e4 1429 panic("cannot create neighbour proc dir entry");
1da177e4
LT
1430#endif
1431
1432 tbl->hash_mask = 1;
1433 tbl->hash_buckets = neigh_hash_alloc(tbl->hash_mask + 1);
1434
1435 phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
77d04bd9 1436 tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1da177e4
LT
1437
1438 if (!tbl->hash_buckets || !tbl->phash_buckets)
1439 panic("cannot allocate neighbour cache hashes");
1440
1da177e4
LT
1441 get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
1442
1443 rwlock_init(&tbl->lock);
b24b8a24 1444 setup_timer(&tbl->gc_timer, neigh_periodic_timer, (unsigned long)tbl);
1da177e4
LT
1445 tbl->gc_timer.expires = now + 1;
1446 add_timer(&tbl->gc_timer);
1447
b24b8a24 1448 setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
c2ecba71
PE
1449 skb_queue_head_init_class(&tbl->proxy_queue,
1450 &neigh_table_proxy_queue_class);
1da177e4
LT
1451
1452 tbl->last_flush = now;
1453 tbl->last_rand = now + tbl->parms.reachable_time * 20;
bd89efc5 1454}
0a204500 1455EXPORT_SYMBOL(neigh_table_init_no_netlink);
bd89efc5
SK
1456
1457void neigh_table_init(struct neigh_table *tbl)
1458{
1459 struct neigh_table *tmp;
1460
1461 neigh_table_init_no_netlink(tbl);
1da177e4 1462 write_lock(&neigh_tbl_lock);
bd89efc5
SK
1463 for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1464 if (tmp->family == tbl->family)
1465 break;
1466 }
1da177e4
LT
1467 tbl->next = neigh_tables;
1468 neigh_tables = tbl;
1469 write_unlock(&neigh_tbl_lock);
bd89efc5
SK
1470
1471 if (unlikely(tmp)) {
1472 printk(KERN_ERR "NEIGH: Registering multiple tables for "
1473 "family %d\n", tbl->family);
1474 dump_stack();
1475 }
1da177e4 1476}
0a204500 1477EXPORT_SYMBOL(neigh_table_init);
1da177e4
LT
1478
1479int neigh_table_clear(struct neigh_table *tbl)
1480{
1481 struct neigh_table **tp;
1482
1483 /* It is not clean... Fix it to unload IPv6 module safely */
1484 del_timer_sync(&tbl->gc_timer);
1485 del_timer_sync(&tbl->proxy_timer);
1486 pneigh_queue_purge(&tbl->proxy_queue);
1487 neigh_ifdown(tbl, NULL);
1488 if (atomic_read(&tbl->entries))
1489 printk(KERN_CRIT "neighbour leakage\n");
1490 write_lock(&neigh_tbl_lock);
1491 for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1492 if (*tp == tbl) {
1493 *tp = tbl->next;
1494 break;
1495 }
1496 }
1497 write_unlock(&neigh_tbl_lock);
1498
1499 neigh_hash_free(tbl->hash_buckets, tbl->hash_mask + 1);
1500 tbl->hash_buckets = NULL;
1501
1502 kfree(tbl->phash_buckets);
1503 tbl->phash_buckets = NULL;
1504
3f192b5c
AD
1505 remove_proc_entry(tbl->id, init_net.proc_net_stat);
1506
3fcde74b
KK
1507 free_percpu(tbl->stats);
1508 tbl->stats = NULL;
1509
bfb85c9f
RD
1510 kmem_cache_destroy(tbl->kmem_cachep);
1511 tbl->kmem_cachep = NULL;
1512
1da177e4
LT
1513 return 0;
1514}
0a204500 1515EXPORT_SYMBOL(neigh_table_clear);
1da177e4 1516
c8822a4e 1517static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1da177e4 1518{
3b1e0a65 1519 struct net *net = sock_net(skb->sk);
a14a49d2
TG
1520 struct ndmsg *ndm;
1521 struct nlattr *dst_attr;
1da177e4
LT
1522 struct neigh_table *tbl;
1523 struct net_device *dev = NULL;
a14a49d2 1524 int err = -EINVAL;
1da177e4 1525
a14a49d2 1526 if (nlmsg_len(nlh) < sizeof(*ndm))
1da177e4
LT
1527 goto out;
1528
a14a49d2
TG
1529 dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1530 if (dst_attr == NULL)
1531 goto out;
1532
1533 ndm = nlmsg_data(nlh);
1534 if (ndm->ndm_ifindex) {
881d966b 1535 dev = dev_get_by_index(net, ndm->ndm_ifindex);
a14a49d2
TG
1536 if (dev == NULL) {
1537 err = -ENODEV;
1538 goto out;
1539 }
1540 }
1541
1da177e4
LT
1542 read_lock(&neigh_tbl_lock);
1543 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
a14a49d2 1544 struct neighbour *neigh;
1da177e4
LT
1545
1546 if (tbl->family != ndm->ndm_family)
1547 continue;
1548 read_unlock(&neigh_tbl_lock);
1549
a14a49d2 1550 if (nla_len(dst_attr) < tbl->key_len)
1da177e4
LT
1551 goto out_dev_put;
1552
1553 if (ndm->ndm_flags & NTF_PROXY) {
426b5303 1554 err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1da177e4
LT
1555 goto out_dev_put;
1556 }
1557
a14a49d2
TG
1558 if (dev == NULL)
1559 goto out_dev_put;
1da177e4 1560
a14a49d2
TG
1561 neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1562 if (neigh == NULL) {
1563 err = -ENOENT;
1564 goto out_dev_put;
1da177e4 1565 }
a14a49d2
TG
1566
1567 err = neigh_update(neigh, NULL, NUD_FAILED,
1568 NEIGH_UPDATE_F_OVERRIDE |
1569 NEIGH_UPDATE_F_ADMIN);
1570 neigh_release(neigh);
1da177e4
LT
1571 goto out_dev_put;
1572 }
1573 read_unlock(&neigh_tbl_lock);
a14a49d2
TG
1574 err = -EAFNOSUPPORT;
1575
1da177e4
LT
1576out_dev_put:
1577 if (dev)
1578 dev_put(dev);
1579out:
1580 return err;
1581}
1582
c8822a4e 1583static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1da177e4 1584{
3b1e0a65 1585 struct net *net = sock_net(skb->sk);
5208debd
TG
1586 struct ndmsg *ndm;
1587 struct nlattr *tb[NDA_MAX+1];
1da177e4
LT
1588 struct neigh_table *tbl;
1589 struct net_device *dev = NULL;
5208debd 1590 int err;
1da177e4 1591
5208debd
TG
1592 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
1593 if (err < 0)
1da177e4
LT
1594 goto out;
1595
5208debd
TG
1596 err = -EINVAL;
1597 if (tb[NDA_DST] == NULL)
1598 goto out;
1599
1600 ndm = nlmsg_data(nlh);
1601 if (ndm->ndm_ifindex) {
881d966b 1602 dev = dev_get_by_index(net, ndm->ndm_ifindex);
5208debd
TG
1603 if (dev == NULL) {
1604 err = -ENODEV;
1605 goto out;
1606 }
1607
1608 if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
1609 goto out_dev_put;
1610 }
1611
1da177e4
LT
1612 read_lock(&neigh_tbl_lock);
1613 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
5208debd
TG
1614 int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1615 struct neighbour *neigh;
1616 void *dst, *lladdr;
1da177e4
LT
1617
1618 if (tbl->family != ndm->ndm_family)
1619 continue;
1620 read_unlock(&neigh_tbl_lock);
1621
5208debd 1622 if (nla_len(tb[NDA_DST]) < tbl->key_len)
1da177e4 1623 goto out_dev_put;
5208debd
TG
1624 dst = nla_data(tb[NDA_DST]);
1625 lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1da177e4
LT
1626
1627 if (ndm->ndm_flags & NTF_PROXY) {
62dd9318
VN
1628 struct pneigh_entry *pn;
1629
1630 err = -ENOBUFS;
426b5303 1631 pn = pneigh_lookup(tbl, net, dst, dev, 1);
62dd9318
VN
1632 if (pn) {
1633 pn->flags = ndm->ndm_flags;
1634 err = 0;
1635 }
1da177e4
LT
1636 goto out_dev_put;
1637 }
1638
5208debd 1639 if (dev == NULL)
1da177e4 1640 goto out_dev_put;
5208debd
TG
1641
1642 neigh = neigh_lookup(tbl, dst, dev);
1643 if (neigh == NULL) {
1644 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1645 err = -ENOENT;
1646 goto out_dev_put;
1647 }
4ec93edb 1648
5208debd
TG
1649 neigh = __neigh_lookup_errno(tbl, dst, dev);
1650 if (IS_ERR(neigh)) {
1651 err = PTR_ERR(neigh);
1da177e4
LT
1652 goto out_dev_put;
1653 }
1da177e4 1654 } else {
5208debd
TG
1655 if (nlh->nlmsg_flags & NLM_F_EXCL) {
1656 err = -EEXIST;
1657 neigh_release(neigh);
1da177e4
LT
1658 goto out_dev_put;
1659 }
1da177e4 1660
5208debd
TG
1661 if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1662 flags &= ~NEIGH_UPDATE_F_OVERRIDE;
1663 }
1da177e4 1664
5208debd
TG
1665 err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1666 neigh_release(neigh);
1da177e4
LT
1667 goto out_dev_put;
1668 }
1669
1670 read_unlock(&neigh_tbl_lock);
5208debd
TG
1671 err = -EAFNOSUPPORT;
1672
1da177e4
LT
1673out_dev_put:
1674 if (dev)
1675 dev_put(dev);
1676out:
1677 return err;
1678}
1679
c7fb64db
TG
1680static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1681{
ca860fb3
TG
1682 struct nlattr *nest;
1683
1684 nest = nla_nest_start(skb, NDTA_PARMS);
1685 if (nest == NULL)
1686 return -ENOBUFS;
c7fb64db
TG
1687
1688 if (parms->dev)
ca860fb3
TG
1689 NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1690
1691 NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1692 NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1693 NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1694 NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1695 NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1696 NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1697 NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1698 NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
c7fb64db 1699 parms->base_reachable_time);
ca860fb3
TG
1700 NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1701 NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1702 NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1703 NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1704 NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1705 NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
c7fb64db 1706
ca860fb3 1707 return nla_nest_end(skb, nest);
c7fb64db 1708
ca860fb3 1709nla_put_failure:
bc3ed28c
TG
1710 nla_nest_cancel(skb, nest);
1711 return -EMSGSIZE;
c7fb64db
TG
1712}
1713
ca860fb3
TG
1714static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1715 u32 pid, u32 seq, int type, int flags)
c7fb64db
TG
1716{
1717 struct nlmsghdr *nlh;
1718 struct ndtmsg *ndtmsg;
1719
ca860fb3
TG
1720 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1721 if (nlh == NULL)
26932566 1722 return -EMSGSIZE;
c7fb64db 1723
ca860fb3 1724 ndtmsg = nlmsg_data(nlh);
c7fb64db
TG
1725
1726 read_lock_bh(&tbl->lock);
1727 ndtmsg->ndtm_family = tbl->family;
9ef1d4c7
PM
1728 ndtmsg->ndtm_pad1 = 0;
1729 ndtmsg->ndtm_pad2 = 0;
c7fb64db 1730
ca860fb3
TG
1731 NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1732 NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1733 NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1734 NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1735 NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
c7fb64db
TG
1736
1737 {
1738 unsigned long now = jiffies;
1739 unsigned int flush_delta = now - tbl->last_flush;
1740 unsigned int rand_delta = now - tbl->last_rand;
1741
1742 struct ndt_config ndc = {
1743 .ndtc_key_len = tbl->key_len,
1744 .ndtc_entry_size = tbl->entry_size,
1745 .ndtc_entries = atomic_read(&tbl->entries),
1746 .ndtc_last_flush = jiffies_to_msecs(flush_delta),
1747 .ndtc_last_rand = jiffies_to_msecs(rand_delta),
1748 .ndtc_hash_rnd = tbl->hash_rnd,
1749 .ndtc_hash_mask = tbl->hash_mask,
1750 .ndtc_hash_chain_gc = tbl->hash_chain_gc,
1751 .ndtc_proxy_qlen = tbl->proxy_queue.qlen,
1752 };
1753
ca860fb3 1754 NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
c7fb64db
TG
1755 }
1756
1757 {
1758 int cpu;
1759 struct ndt_stats ndst;
1760
1761 memset(&ndst, 0, sizeof(ndst));
1762
6f912042 1763 for_each_possible_cpu(cpu) {
c7fb64db
TG
1764 struct neigh_statistics *st;
1765
c7fb64db
TG
1766 st = per_cpu_ptr(tbl->stats, cpu);
1767 ndst.ndts_allocs += st->allocs;
1768 ndst.ndts_destroys += st->destroys;
1769 ndst.ndts_hash_grows += st->hash_grows;
1770 ndst.ndts_res_failed += st->res_failed;
1771 ndst.ndts_lookups += st->lookups;
1772 ndst.ndts_hits += st->hits;
1773 ndst.ndts_rcv_probes_mcast += st->rcv_probes_mcast;
1774 ndst.ndts_rcv_probes_ucast += st->rcv_probes_ucast;
1775 ndst.ndts_periodic_gc_runs += st->periodic_gc_runs;
1776 ndst.ndts_forced_gc_runs += st->forced_gc_runs;
1777 }
1778
ca860fb3 1779 NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
c7fb64db
TG
1780 }
1781
1782 BUG_ON(tbl->parms.dev);
1783 if (neightbl_fill_parms(skb, &tbl->parms) < 0)
ca860fb3 1784 goto nla_put_failure;
c7fb64db
TG
1785
1786 read_unlock_bh(&tbl->lock);
ca860fb3 1787 return nlmsg_end(skb, nlh);
c7fb64db 1788
ca860fb3 1789nla_put_failure:
c7fb64db 1790 read_unlock_bh(&tbl->lock);
26932566
PM
1791 nlmsg_cancel(skb, nlh);
1792 return -EMSGSIZE;
c7fb64db
TG
1793}
1794
ca860fb3
TG
1795static int neightbl_fill_param_info(struct sk_buff *skb,
1796 struct neigh_table *tbl,
c7fb64db 1797 struct neigh_parms *parms,
ca860fb3
TG
1798 u32 pid, u32 seq, int type,
1799 unsigned int flags)
c7fb64db
TG
1800{
1801 struct ndtmsg *ndtmsg;
1802 struct nlmsghdr *nlh;
1803
ca860fb3
TG
1804 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1805 if (nlh == NULL)
26932566 1806 return -EMSGSIZE;
c7fb64db 1807
ca860fb3 1808 ndtmsg = nlmsg_data(nlh);
c7fb64db
TG
1809
1810 read_lock_bh(&tbl->lock);
1811 ndtmsg->ndtm_family = tbl->family;
9ef1d4c7
PM
1812 ndtmsg->ndtm_pad1 = 0;
1813 ndtmsg->ndtm_pad2 = 0;
c7fb64db 1814
ca860fb3
TG
1815 if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1816 neightbl_fill_parms(skb, parms) < 0)
1817 goto errout;
c7fb64db
TG
1818
1819 read_unlock_bh(&tbl->lock);
ca860fb3
TG
1820 return nlmsg_end(skb, nlh);
1821errout:
c7fb64db 1822 read_unlock_bh(&tbl->lock);
26932566
PM
1823 nlmsg_cancel(skb, nlh);
1824 return -EMSGSIZE;
c7fb64db 1825}
4ec93edb 1826
ef7c79ed 1827static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
6b3f8674
TG
1828 [NDTA_NAME] = { .type = NLA_STRING },
1829 [NDTA_THRESH1] = { .type = NLA_U32 },
1830 [NDTA_THRESH2] = { .type = NLA_U32 },
1831 [NDTA_THRESH3] = { .type = NLA_U32 },
1832 [NDTA_GC_INTERVAL] = { .type = NLA_U64 },
1833 [NDTA_PARMS] = { .type = NLA_NESTED },
1834};
1835
ef7c79ed 1836static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
6b3f8674
TG
1837 [NDTPA_IFINDEX] = { .type = NLA_U32 },
1838 [NDTPA_QUEUE_LEN] = { .type = NLA_U32 },
1839 [NDTPA_PROXY_QLEN] = { .type = NLA_U32 },
1840 [NDTPA_APP_PROBES] = { .type = NLA_U32 },
1841 [NDTPA_UCAST_PROBES] = { .type = NLA_U32 },
1842 [NDTPA_MCAST_PROBES] = { .type = NLA_U32 },
1843 [NDTPA_BASE_REACHABLE_TIME] = { .type = NLA_U64 },
1844 [NDTPA_GC_STALETIME] = { .type = NLA_U64 },
1845 [NDTPA_DELAY_PROBE_TIME] = { .type = NLA_U64 },
1846 [NDTPA_RETRANS_TIME] = { .type = NLA_U64 },
1847 [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 },
1848 [NDTPA_PROXY_DELAY] = { .type = NLA_U64 },
1849 [NDTPA_LOCKTIME] = { .type = NLA_U64 },
1850};
1851
c8822a4e 1852static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
c7fb64db 1853{
3b1e0a65 1854 struct net *net = sock_net(skb->sk);
c7fb64db 1855 struct neigh_table *tbl;
6b3f8674
TG
1856 struct ndtmsg *ndtmsg;
1857 struct nlattr *tb[NDTA_MAX+1];
1858 int err;
c7fb64db 1859
6b3f8674
TG
1860 err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1861 nl_neightbl_policy);
1862 if (err < 0)
1863 goto errout;
c7fb64db 1864
6b3f8674
TG
1865 if (tb[NDTA_NAME] == NULL) {
1866 err = -EINVAL;
1867 goto errout;
1868 }
1869
1870 ndtmsg = nlmsg_data(nlh);
c7fb64db
TG
1871 read_lock(&neigh_tbl_lock);
1872 for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1873 if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1874 continue;
1875
6b3f8674 1876 if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
c7fb64db
TG
1877 break;
1878 }
1879
1880 if (tbl == NULL) {
1881 err = -ENOENT;
6b3f8674 1882 goto errout_locked;
c7fb64db
TG
1883 }
1884
4ec93edb 1885 /*
c7fb64db
TG
1886 * We acquire tbl->lock to be nice to the periodic timers and
1887 * make sure they always see a consistent set of values.
1888 */
1889 write_lock_bh(&tbl->lock);
1890
6b3f8674
TG
1891 if (tb[NDTA_PARMS]) {
1892 struct nlattr *tbp[NDTPA_MAX+1];
c7fb64db 1893 struct neigh_parms *p;
6b3f8674 1894 int i, ifindex = 0;
c7fb64db 1895
6b3f8674
TG
1896 err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1897 nl_ntbl_parm_policy);
1898 if (err < 0)
1899 goto errout_tbl_lock;
c7fb64db 1900
6b3f8674
TG
1901 if (tbp[NDTPA_IFINDEX])
1902 ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
c7fb64db 1903
426b5303 1904 p = lookup_neigh_params(tbl, net, ifindex);
c7fb64db
TG
1905 if (p == NULL) {
1906 err = -ENOENT;
6b3f8674 1907 goto errout_tbl_lock;
c7fb64db 1908 }
c7fb64db 1909
6b3f8674
TG
1910 for (i = 1; i <= NDTPA_MAX; i++) {
1911 if (tbp[i] == NULL)
1912 continue;
c7fb64db 1913
6b3f8674
TG
1914 switch (i) {
1915 case NDTPA_QUEUE_LEN:
1916 p->queue_len = nla_get_u32(tbp[i]);
1917 break;
1918 case NDTPA_PROXY_QLEN:
1919 p->proxy_qlen = nla_get_u32(tbp[i]);
1920 break;
1921 case NDTPA_APP_PROBES:
1922 p->app_probes = nla_get_u32(tbp[i]);
1923 break;
1924 case NDTPA_UCAST_PROBES:
1925 p->ucast_probes = nla_get_u32(tbp[i]);
1926 break;
1927 case NDTPA_MCAST_PROBES:
1928 p->mcast_probes = nla_get_u32(tbp[i]);
1929 break;
1930 case NDTPA_BASE_REACHABLE_TIME:
1931 p->base_reachable_time = nla_get_msecs(tbp[i]);
1932 break;
1933 case NDTPA_GC_STALETIME:
1934 p->gc_staletime = nla_get_msecs(tbp[i]);
1935 break;
1936 case NDTPA_DELAY_PROBE_TIME:
1937 p->delay_probe_time = nla_get_msecs(tbp[i]);
1938 break;
1939 case NDTPA_RETRANS_TIME:
1940 p->retrans_time = nla_get_msecs(tbp[i]);
1941 break;
1942 case NDTPA_ANYCAST_DELAY:
1943 p->anycast_delay = nla_get_msecs(tbp[i]);
1944 break;
1945 case NDTPA_PROXY_DELAY:
1946 p->proxy_delay = nla_get_msecs(tbp[i]);
1947 break;
1948 case NDTPA_LOCKTIME:
1949 p->locktime = nla_get_msecs(tbp[i]);
1950 break;
1951 }
1952 }
1953 }
c7fb64db 1954
6b3f8674
TG
1955 if (tb[NDTA_THRESH1])
1956 tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
c7fb64db 1957
6b3f8674
TG
1958 if (tb[NDTA_THRESH2])
1959 tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
c7fb64db 1960
6b3f8674
TG
1961 if (tb[NDTA_THRESH3])
1962 tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
c7fb64db 1963
6b3f8674
TG
1964 if (tb[NDTA_GC_INTERVAL])
1965 tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
c7fb64db
TG
1966
1967 err = 0;
1968
6b3f8674 1969errout_tbl_lock:
c7fb64db 1970 write_unlock_bh(&tbl->lock);
6b3f8674 1971errout_locked:
c7fb64db 1972 read_unlock(&neigh_tbl_lock);
6b3f8674 1973errout:
c7fb64db
TG
1974 return err;
1975}
1976
c8822a4e 1977static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
c7fb64db 1978{
3b1e0a65 1979 struct net *net = sock_net(skb->sk);
ca860fb3
TG
1980 int family, tidx, nidx = 0;
1981 int tbl_skip = cb->args[0];
1982 int neigh_skip = cb->args[1];
c7fb64db
TG
1983 struct neigh_table *tbl;
1984
ca860fb3 1985 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
c7fb64db
TG
1986
1987 read_lock(&neigh_tbl_lock);
ca860fb3 1988 for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
c7fb64db
TG
1989 struct neigh_parms *p;
1990
ca860fb3 1991 if (tidx < tbl_skip || (family && tbl->family != family))
c7fb64db
TG
1992 continue;
1993
ca860fb3
TG
1994 if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
1995 cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
1996 NLM_F_MULTI) <= 0)
c7fb64db
TG
1997 break;
1998
426b5303 1999 for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
878628fb 2000 if (!net_eq(neigh_parms_net(p), net))
426b5303
EB
2001 continue;
2002
2003 if (nidx++ < neigh_skip)
c7fb64db
TG
2004 continue;
2005
ca860fb3
TG
2006 if (neightbl_fill_param_info(skb, tbl, p,
2007 NETLINK_CB(cb->skb).pid,
2008 cb->nlh->nlmsg_seq,
2009 RTM_NEWNEIGHTBL,
2010 NLM_F_MULTI) <= 0)
c7fb64db
TG
2011 goto out;
2012 }
2013
ca860fb3 2014 neigh_skip = 0;
c7fb64db
TG
2015 }
2016out:
2017 read_unlock(&neigh_tbl_lock);
ca860fb3
TG
2018 cb->args[0] = tidx;
2019 cb->args[1] = nidx;
c7fb64db
TG
2020
2021 return skb->len;
2022}
1da177e4 2023
8b8aec50
TG
2024static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2025 u32 pid, u32 seq, int type, unsigned int flags)
1da177e4
LT
2026{
2027 unsigned long now = jiffies;
1da177e4 2028 struct nda_cacheinfo ci;
8b8aec50
TG
2029 struct nlmsghdr *nlh;
2030 struct ndmsg *ndm;
2031
2032 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2033 if (nlh == NULL)
26932566 2034 return -EMSGSIZE;
1da177e4 2035
8b8aec50
TG
2036 ndm = nlmsg_data(nlh);
2037 ndm->ndm_family = neigh->ops->family;
9ef1d4c7
PM
2038 ndm->ndm_pad1 = 0;
2039 ndm->ndm_pad2 = 0;
8b8aec50
TG
2040 ndm->ndm_flags = neigh->flags;
2041 ndm->ndm_type = neigh->type;
2042 ndm->ndm_ifindex = neigh->dev->ifindex;
1da177e4 2043
8b8aec50
TG
2044 NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
2045
2046 read_lock_bh(&neigh->lock);
2047 ndm->ndm_state = neigh->nud_state;
2048 if ((neigh->nud_state & NUD_VALID) &&
2049 nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, neigh->ha) < 0) {
2050 read_unlock_bh(&neigh->lock);
2051 goto nla_put_failure;
2052 }
2053
b9f5f52c
SH
2054 ci.ndm_used = jiffies_to_clock_t(now - neigh->used);
2055 ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2056 ci.ndm_updated = jiffies_to_clock_t(now - neigh->updated);
8b8aec50
TG
2057 ci.ndm_refcnt = atomic_read(&neigh->refcnt) - 1;
2058 read_unlock_bh(&neigh->lock);
2059
2060 NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2061 NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
2062
2063 return nlmsg_end(skb, nlh);
2064
2065nla_put_failure:
26932566
PM
2066 nlmsg_cancel(skb, nlh);
2067 return -EMSGSIZE;
1da177e4
LT
2068}
2069
d961db35
TG
2070static void neigh_update_notify(struct neighbour *neigh)
2071{
2072 call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2073 __neigh_notify(neigh, RTM_NEWNEIGH, 0);
2074}
1da177e4
LT
2075
2076static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2077 struct netlink_callback *cb)
2078{
3b1e0a65 2079 struct net * net = sock_net(skb->sk);
1da177e4
LT
2080 struct neighbour *n;
2081 int rc, h, s_h = cb->args[1];
2082 int idx, s_idx = idx = cb->args[2];
2083
c5e29460 2084 read_lock_bh(&tbl->lock);
1da177e4
LT
2085 for (h = 0; h <= tbl->hash_mask; h++) {
2086 if (h < s_h)
2087 continue;
2088 if (h > s_h)
2089 s_idx = 0;
426b5303
EB
2090 for (n = tbl->hash_buckets[h], idx = 0; n; n = n->next) {
2091 int lidx;
c346dca1 2092 if (dev_net(n->dev) != net)
426b5303
EB
2093 continue;
2094 lidx = idx++;
2095 if (lidx < s_idx)
1da177e4
LT
2096 continue;
2097 if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2098 cb->nlh->nlmsg_seq,
b6544c0b
JHS
2099 RTM_NEWNEIGH,
2100 NLM_F_MULTI) <= 0) {
1da177e4
LT
2101 read_unlock_bh(&tbl->lock);
2102 rc = -1;
2103 goto out;
2104 }
2105 }
1da177e4 2106 }
c5e29460 2107 read_unlock_bh(&tbl->lock);
1da177e4
LT
2108 rc = skb->len;
2109out:
2110 cb->args[1] = h;
2111 cb->args[2] = idx;
2112 return rc;
2113}
2114
c8822a4e 2115static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
1da177e4
LT
2116{
2117 struct neigh_table *tbl;
2118 int t, family, s_t;
2119
2120 read_lock(&neigh_tbl_lock);
8b8aec50 2121 family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
1da177e4
LT
2122 s_t = cb->args[0];
2123
2124 for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
2125 if (t < s_t || (family && tbl->family != family))
2126 continue;
2127 if (t > s_t)
2128 memset(&cb->args[1], 0, sizeof(cb->args) -
2129 sizeof(cb->args[0]));
2130 if (neigh_dump_table(tbl, skb, cb) < 0)
2131 break;
2132 }
2133 read_unlock(&neigh_tbl_lock);
2134
2135 cb->args[0] = t;
2136 return skb->len;
2137}
2138
2139void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2140{
2141 int chain;
2142
2143 read_lock_bh(&tbl->lock);
2144 for (chain = 0; chain <= tbl->hash_mask; chain++) {
2145 struct neighbour *n;
2146
2147 for (n = tbl->hash_buckets[chain]; n; n = n->next)
2148 cb(n, cookie);
2149 }
2150 read_unlock_bh(&tbl->lock);
2151}
2152EXPORT_SYMBOL(neigh_for_each);
2153
2154/* The tbl->lock must be held as a writer and BH disabled. */
2155void __neigh_for_each_release(struct neigh_table *tbl,
2156 int (*cb)(struct neighbour *))
2157{
2158 int chain;
2159
2160 for (chain = 0; chain <= tbl->hash_mask; chain++) {
2161 struct neighbour *n, **np;
2162
2163 np = &tbl->hash_buckets[chain];
2164 while ((n = *np) != NULL) {
2165 int release;
2166
2167 write_lock(&n->lock);
2168 release = cb(n);
2169 if (release) {
2170 *np = n->next;
2171 n->dead = 1;
2172 } else
2173 np = &n->next;
2174 write_unlock(&n->lock);
4f494554
TG
2175 if (release)
2176 neigh_cleanup_and_release(n);
1da177e4
LT
2177 }
2178 }
2179}
2180EXPORT_SYMBOL(__neigh_for_each_release);
2181
2182#ifdef CONFIG_PROC_FS
2183
2184static struct neighbour *neigh_get_first(struct seq_file *seq)
2185{
2186 struct neigh_seq_state *state = seq->private;
1218854a 2187 struct net *net = seq_file_net(seq);
1da177e4
LT
2188 struct neigh_table *tbl = state->tbl;
2189 struct neighbour *n = NULL;
2190 int bucket = state->bucket;
2191
2192 state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2193 for (bucket = 0; bucket <= tbl->hash_mask; bucket++) {
2194 n = tbl->hash_buckets[bucket];
2195
2196 while (n) {
878628fb 2197 if (!net_eq(dev_net(n->dev), net))
426b5303 2198 goto next;
1da177e4
LT
2199 if (state->neigh_sub_iter) {
2200 loff_t fakep = 0;
2201 void *v;
2202
2203 v = state->neigh_sub_iter(state, n, &fakep);
2204 if (!v)
2205 goto next;
2206 }
2207 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2208 break;
2209 if (n->nud_state & ~NUD_NOARP)
2210 break;
2211 next:
2212 n = n->next;
2213 }
2214
2215 if (n)
2216 break;
2217 }
2218 state->bucket = bucket;
2219
2220 return n;
2221}
2222
2223static struct neighbour *neigh_get_next(struct seq_file *seq,
2224 struct neighbour *n,
2225 loff_t *pos)
2226{
2227 struct neigh_seq_state *state = seq->private;
1218854a 2228 struct net *net = seq_file_net(seq);
1da177e4
LT
2229 struct neigh_table *tbl = state->tbl;
2230
2231 if (state->neigh_sub_iter) {
2232 void *v = state->neigh_sub_iter(state, n, pos);
2233 if (v)
2234 return n;
2235 }
2236 n = n->next;
2237
2238 while (1) {
2239 while (n) {
878628fb 2240 if (!net_eq(dev_net(n->dev), net))
426b5303 2241 goto next;
1da177e4
LT
2242 if (state->neigh_sub_iter) {
2243 void *v = state->neigh_sub_iter(state, n, pos);
2244 if (v)
2245 return n;
2246 goto next;
2247 }
2248 if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2249 break;
2250
2251 if (n->nud_state & ~NUD_NOARP)
2252 break;
2253 next:
2254 n = n->next;
2255 }
2256
2257 if (n)
2258 break;
2259
2260 if (++state->bucket > tbl->hash_mask)
2261 break;
2262
2263 n = tbl->hash_buckets[state->bucket];
2264 }
2265
2266 if (n && pos)
2267 --(*pos);
2268 return n;
2269}
2270
2271static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2272{
2273 struct neighbour *n = neigh_get_first(seq);
2274
2275 if (n) {
745e2031 2276 --(*pos);
1da177e4
LT
2277 while (*pos) {
2278 n = neigh_get_next(seq, n, pos);
2279 if (!n)
2280 break;
2281 }
2282 }
2283 return *pos ? NULL : n;
2284}
2285
2286static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2287{
2288 struct neigh_seq_state *state = seq->private;
1218854a 2289 struct net *net = seq_file_net(seq);
1da177e4
LT
2290 struct neigh_table *tbl = state->tbl;
2291 struct pneigh_entry *pn = NULL;
2292 int bucket = state->bucket;
2293
2294 state->flags |= NEIGH_SEQ_IS_PNEIGH;
2295 for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2296 pn = tbl->phash_buckets[bucket];
878628fb 2297 while (pn && !net_eq(pneigh_net(pn), net))
426b5303 2298 pn = pn->next;
1da177e4
LT
2299 if (pn)
2300 break;
2301 }
2302 state->bucket = bucket;
2303
2304 return pn;
2305}
2306
2307static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2308 struct pneigh_entry *pn,
2309 loff_t *pos)
2310{
2311 struct neigh_seq_state *state = seq->private;
1218854a 2312 struct net *net = seq_file_net(seq);
1da177e4
LT
2313 struct neigh_table *tbl = state->tbl;
2314
2315 pn = pn->next;
2316 while (!pn) {
2317 if (++state->bucket > PNEIGH_HASHMASK)
2318 break;
2319 pn = tbl->phash_buckets[state->bucket];
878628fb 2320 while (pn && !net_eq(pneigh_net(pn), net))
426b5303 2321 pn = pn->next;
1da177e4
LT
2322 if (pn)
2323 break;
2324 }
2325
2326 if (pn && pos)
2327 --(*pos);
2328
2329 return pn;
2330}
2331
2332static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2333{
2334 struct pneigh_entry *pn = pneigh_get_first(seq);
2335
2336 if (pn) {
745e2031 2337 --(*pos);
1da177e4
LT
2338 while (*pos) {
2339 pn = pneigh_get_next(seq, pn, pos);
2340 if (!pn)
2341 break;
2342 }
2343 }
2344 return *pos ? NULL : pn;
2345}
2346
2347static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2348{
2349 struct neigh_seq_state *state = seq->private;
2350 void *rc;
745e2031 2351 loff_t idxpos = *pos;
1da177e4 2352
745e2031 2353 rc = neigh_get_idx(seq, &idxpos);
1da177e4 2354 if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
745e2031 2355 rc = pneigh_get_idx(seq, &idxpos);
1da177e4
LT
2356
2357 return rc;
2358}
2359
2360void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
9a429c49 2361 __acquires(tbl->lock)
1da177e4
LT
2362{
2363 struct neigh_seq_state *state = seq->private;
1da177e4
LT
2364
2365 state->tbl = tbl;
2366 state->bucket = 0;
2367 state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2368
2369 read_lock_bh(&tbl->lock);
2370
745e2031 2371 return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
1da177e4
LT
2372}
2373EXPORT_SYMBOL(neigh_seq_start);
2374
2375void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2376{
2377 struct neigh_seq_state *state;
2378 void *rc;
2379
2380 if (v == SEQ_START_TOKEN) {
bff69732 2381 rc = neigh_get_first(seq);
1da177e4
LT
2382 goto out;
2383 }
2384
2385 state = seq->private;
2386 if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2387 rc = neigh_get_next(seq, v, NULL);
2388 if (rc)
2389 goto out;
2390 if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2391 rc = pneigh_get_first(seq);
2392 } else {
2393 BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2394 rc = pneigh_get_next(seq, v, NULL);
2395 }
2396out:
2397 ++(*pos);
2398 return rc;
2399}
2400EXPORT_SYMBOL(neigh_seq_next);
2401
2402void neigh_seq_stop(struct seq_file *seq, void *v)
9a429c49 2403 __releases(tbl->lock)
1da177e4
LT
2404{
2405 struct neigh_seq_state *state = seq->private;
2406 struct neigh_table *tbl = state->tbl;
2407
2408 read_unlock_bh(&tbl->lock);
2409}
2410EXPORT_SYMBOL(neigh_seq_stop);
2411
2412/* statistics via seq_file */
2413
2414static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2415{
2416 struct proc_dir_entry *pde = seq->private;
2417 struct neigh_table *tbl = pde->data;
2418 int cpu;
2419
2420 if (*pos == 0)
2421 return SEQ_START_TOKEN;
4ec93edb 2422
1da177e4
LT
2423 for (cpu = *pos-1; cpu < NR_CPUS; ++cpu) {
2424 if (!cpu_possible(cpu))
2425 continue;
2426 *pos = cpu+1;
2427 return per_cpu_ptr(tbl->stats, cpu);
2428 }
2429 return NULL;
2430}
2431
2432static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2433{
2434 struct proc_dir_entry *pde = seq->private;
2435 struct neigh_table *tbl = pde->data;
2436 int cpu;
2437
2438 for (cpu = *pos; cpu < NR_CPUS; ++cpu) {
2439 if (!cpu_possible(cpu))
2440 continue;
2441 *pos = cpu+1;
2442 return per_cpu_ptr(tbl->stats, cpu);
2443 }
2444 return NULL;
2445}
2446
2447static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2448{
2449
2450}
2451
2452static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2453{
2454 struct proc_dir_entry *pde = seq->private;
2455 struct neigh_table *tbl = pde->data;
2456 struct neigh_statistics *st = v;
2457
2458 if (v == SEQ_START_TOKEN) {
9a6d276e 2459 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");
1da177e4
LT
2460 return 0;
2461 }
2462
2463 seq_printf(seq, "%08x %08lx %08lx %08lx %08lx %08lx %08lx "
9a6d276e 2464 "%08lx %08lx %08lx %08lx %08lx\n",
1da177e4
LT
2465 atomic_read(&tbl->entries),
2466
2467 st->allocs,
2468 st->destroys,
2469 st->hash_grows,
2470
2471 st->lookups,
2472 st->hits,
2473
2474 st->res_failed,
2475
2476 st->rcv_probes_mcast,
2477 st->rcv_probes_ucast,
2478
2479 st->periodic_gc_runs,
9a6d276e
NH
2480 st->forced_gc_runs,
2481 st->unres_discards
1da177e4
LT
2482 );
2483
2484 return 0;
2485}
2486
f690808e 2487static const struct seq_operations neigh_stat_seq_ops = {
1da177e4
LT
2488 .start = neigh_stat_seq_start,
2489 .next = neigh_stat_seq_next,
2490 .stop = neigh_stat_seq_stop,
2491 .show = neigh_stat_seq_show,
2492};
2493
2494static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2495{
2496 int ret = seq_open(file, &neigh_stat_seq_ops);
2497
2498 if (!ret) {
2499 struct seq_file *sf = file->private_data;
2500 sf->private = PDE(inode);
2501 }
2502 return ret;
2503};
2504
9a32144e 2505static const struct file_operations neigh_stat_seq_fops = {
1da177e4
LT
2506 .owner = THIS_MODULE,
2507 .open = neigh_stat_seq_open,
2508 .read = seq_read,
2509 .llseek = seq_lseek,
2510 .release = seq_release,
2511};
2512
2513#endif /* CONFIG_PROC_FS */
2514
339bf98f
TG
2515static inline size_t neigh_nlmsg_size(void)
2516{
2517 return NLMSG_ALIGN(sizeof(struct ndmsg))
2518 + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2519 + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2520 + nla_total_size(sizeof(struct nda_cacheinfo))
2521 + nla_total_size(4); /* NDA_PROBES */
2522}
2523
b8673311 2524static void __neigh_notify(struct neighbour *n, int type, int flags)
1da177e4 2525{
c346dca1 2526 struct net *net = dev_net(n->dev);
8b8aec50 2527 struct sk_buff *skb;
b8673311 2528 int err = -ENOBUFS;
1da177e4 2529
339bf98f 2530 skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
8b8aec50 2531 if (skb == NULL)
b8673311 2532 goto errout;
1da177e4 2533
b8673311 2534 err = neigh_fill_info(skb, n, 0, 0, type, flags);
26932566
PM
2535 if (err < 0) {
2536 /* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2537 WARN_ON(err == -EMSGSIZE);
2538 kfree_skb(skb);
2539 goto errout;
2540 }
426b5303 2541 err = rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
b8673311
TG
2542errout:
2543 if (err < 0)
426b5303 2544 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
1da177e4
LT
2545}
2546
d961db35 2547#ifdef CONFIG_ARPD
b8673311 2548void neigh_app_ns(struct neighbour *n)
1da177e4 2549{
b8673311
TG
2550 __neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2551}
0a204500 2552EXPORT_SYMBOL(neigh_app_ns);
1da177e4
LT
2553#endif /* CONFIG_ARPD */
2554
2555#ifdef CONFIG_SYSCTL
2556
2557static struct neigh_sysctl_table {
2558 struct ctl_table_header *sysctl_header;
c3bac5a7
PE
2559 struct ctl_table neigh_vars[__NET_NEIGH_MAX];
2560 char *dev_name;
ab32ea5d 2561} neigh_sysctl_template __read_mostly = {
1da177e4
LT
2562 .neigh_vars = {
2563 {
2564 .ctl_name = NET_NEIGH_MCAST_SOLICIT,
2565 .procname = "mcast_solicit",
2566 .maxlen = sizeof(int),
2567 .mode = 0644,
6d9f239a 2568 .proc_handler = proc_dointvec,
1da177e4
LT
2569 },
2570 {
2571 .ctl_name = NET_NEIGH_UCAST_SOLICIT,
2572 .procname = "ucast_solicit",
2573 .maxlen = sizeof(int),
2574 .mode = 0644,
6d9f239a 2575 .proc_handler = proc_dointvec,
1da177e4
LT
2576 },
2577 {
2578 .ctl_name = NET_NEIGH_APP_SOLICIT,
2579 .procname = "app_solicit",
2580 .maxlen = sizeof(int),
2581 .mode = 0644,
6d9f239a 2582 .proc_handler = proc_dointvec,
1da177e4
LT
2583 },
2584 {
1da177e4
LT
2585 .procname = "retrans_time",
2586 .maxlen = sizeof(int),
2587 .mode = 0644,
6d9f239a 2588 .proc_handler = proc_dointvec_userhz_jiffies,
1da177e4
LT
2589 },
2590 {
2591 .ctl_name = NET_NEIGH_REACHABLE_TIME,
2592 .procname = "base_reachable_time",
2593 .maxlen = sizeof(int),
2594 .mode = 0644,
6d9f239a
AD
2595 .proc_handler = proc_dointvec_jiffies,
2596 .strategy = sysctl_jiffies,
1da177e4
LT
2597 },
2598 {
2599 .ctl_name = NET_NEIGH_DELAY_PROBE_TIME,
2600 .procname = "delay_first_probe_time",
2601 .maxlen = sizeof(int),
2602 .mode = 0644,
6d9f239a
AD
2603 .proc_handler = proc_dointvec_jiffies,
2604 .strategy = sysctl_jiffies,
1da177e4
LT
2605 },
2606 {
2607 .ctl_name = NET_NEIGH_GC_STALE_TIME,
2608 .procname = "gc_stale_time",
2609 .maxlen = sizeof(int),
2610 .mode = 0644,
6d9f239a
AD
2611 .proc_handler = proc_dointvec_jiffies,
2612 .strategy = sysctl_jiffies,
1da177e4
LT
2613 },
2614 {
2615 .ctl_name = NET_NEIGH_UNRES_QLEN,
2616 .procname = "unres_qlen",
2617 .maxlen = sizeof(int),
2618 .mode = 0644,
6d9f239a 2619 .proc_handler = proc_dointvec,
1da177e4
LT
2620 },
2621 {
2622 .ctl_name = NET_NEIGH_PROXY_QLEN,
2623 .procname = "proxy_qlen",
2624 .maxlen = sizeof(int),
2625 .mode = 0644,
6d9f239a 2626 .proc_handler = proc_dointvec,
1da177e4
LT
2627 },
2628 {
1da177e4
LT
2629 .procname = "anycast_delay",
2630 .maxlen = sizeof(int),
2631 .mode = 0644,
6d9f239a 2632 .proc_handler = proc_dointvec_userhz_jiffies,
1da177e4
LT
2633 },
2634 {
1da177e4
LT
2635 .procname = "proxy_delay",
2636 .maxlen = sizeof(int),
2637 .mode = 0644,
6d9f239a 2638 .proc_handler = proc_dointvec_userhz_jiffies,
1da177e4
LT
2639 },
2640 {
1da177e4
LT
2641 .procname = "locktime",
2642 .maxlen = sizeof(int),
2643 .mode = 0644,
6d9f239a 2644 .proc_handler = proc_dointvec_userhz_jiffies,
1da177e4 2645 },
d12af679
EB
2646 {
2647 .ctl_name = NET_NEIGH_RETRANS_TIME_MS,
2648 .procname = "retrans_time_ms",
2649 .maxlen = sizeof(int),
2650 .mode = 0644,
6d9f239a
AD
2651 .proc_handler = proc_dointvec_ms_jiffies,
2652 .strategy = sysctl_ms_jiffies,
d12af679
EB
2653 },
2654 {
2655 .ctl_name = NET_NEIGH_REACHABLE_TIME_MS,
2656 .procname = "base_reachable_time_ms",
2657 .maxlen = sizeof(int),
2658 .mode = 0644,
6d9f239a
AD
2659 .proc_handler = proc_dointvec_ms_jiffies,
2660 .strategy = sysctl_ms_jiffies,
d12af679 2661 },
1da177e4
LT
2662 {
2663 .ctl_name = NET_NEIGH_GC_INTERVAL,
2664 .procname = "gc_interval",
2665 .maxlen = sizeof(int),
2666 .mode = 0644,
6d9f239a
AD
2667 .proc_handler = proc_dointvec_jiffies,
2668 .strategy = sysctl_jiffies,
1da177e4
LT
2669 },
2670 {
2671 .ctl_name = NET_NEIGH_GC_THRESH1,
2672 .procname = "gc_thresh1",
2673 .maxlen = sizeof(int),
2674 .mode = 0644,
6d9f239a 2675 .proc_handler = proc_dointvec,
1da177e4
LT
2676 },
2677 {
2678 .ctl_name = NET_NEIGH_GC_THRESH2,
2679 .procname = "gc_thresh2",
2680 .maxlen = sizeof(int),
2681 .mode = 0644,
6d9f239a 2682 .proc_handler = proc_dointvec,
1da177e4
LT
2683 },
2684 {
2685 .ctl_name = NET_NEIGH_GC_THRESH3,
2686 .procname = "gc_thresh3",
2687 .maxlen = sizeof(int),
2688 .mode = 0644,
6d9f239a 2689 .proc_handler = proc_dointvec,
1da177e4 2690 },
c3bac5a7 2691 {},
1da177e4
LT
2692 },
2693};
2694
2695int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
4ec93edb 2696 int p_id, int pdev_id, char *p_name,
1da177e4
LT
2697 proc_handler *handler, ctl_handler *strategy)
2698{
3c607bbb 2699 struct neigh_sysctl_table *t;
1da177e4 2700 const char *dev_name_source = NULL;
c3bac5a7
PE
2701
2702#define NEIGH_CTL_PATH_ROOT 0
2703#define NEIGH_CTL_PATH_PROTO 1
2704#define NEIGH_CTL_PATH_NEIGH 2
2705#define NEIGH_CTL_PATH_DEV 3
2706
2707 struct ctl_path neigh_path[] = {
2708 { .procname = "net", .ctl_name = CTL_NET, },
2709 { .procname = "proto", .ctl_name = 0, },
2710 { .procname = "neigh", .ctl_name = 0, },
2711 { .procname = "default", .ctl_name = NET_PROTO_CONF_DEFAULT, },
2712 { },
2713 };
1da177e4 2714
3c607bbb 2715 t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
1da177e4 2716 if (!t)
3c607bbb
PE
2717 goto err;
2718
1da177e4
LT
2719 t->neigh_vars[0].data = &p->mcast_probes;
2720 t->neigh_vars[1].data = &p->ucast_probes;
2721 t->neigh_vars[2].data = &p->app_probes;
2722 t->neigh_vars[3].data = &p->retrans_time;
2723 t->neigh_vars[4].data = &p->base_reachable_time;
2724 t->neigh_vars[5].data = &p->delay_probe_time;
2725 t->neigh_vars[6].data = &p->gc_staletime;
2726 t->neigh_vars[7].data = &p->queue_len;
2727 t->neigh_vars[8].data = &p->proxy_qlen;
2728 t->neigh_vars[9].data = &p->anycast_delay;
2729 t->neigh_vars[10].data = &p->proxy_delay;
2730 t->neigh_vars[11].data = &p->locktime;
d12af679
EB
2731 t->neigh_vars[12].data = &p->retrans_time;
2732 t->neigh_vars[13].data = &p->base_reachable_time;
1da177e4
LT
2733
2734 if (dev) {
2735 dev_name_source = dev->name;
c3bac5a7 2736 neigh_path[NEIGH_CTL_PATH_DEV].ctl_name = dev->ifindex;
d12af679
EB
2737 /* Terminate the table early */
2738 memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
1da177e4 2739 } else {
c3bac5a7 2740 dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
d12af679
EB
2741 t->neigh_vars[14].data = (int *)(p + 1);
2742 t->neigh_vars[15].data = (int *)(p + 1) + 1;
2743 t->neigh_vars[16].data = (int *)(p + 1) + 2;
2744 t->neigh_vars[17].data = (int *)(p + 1) + 3;
1da177e4
LT
2745 }
2746
1da177e4
LT
2747
2748 if (handler || strategy) {
2749 /* RetransTime */
2750 t->neigh_vars[3].proc_handler = handler;
2751 t->neigh_vars[3].strategy = strategy;
2752 t->neigh_vars[3].extra1 = dev;
d12af679
EB
2753 if (!strategy)
2754 t->neigh_vars[3].ctl_name = CTL_UNNUMBERED;
1da177e4
LT
2755 /* ReachableTime */
2756 t->neigh_vars[4].proc_handler = handler;
2757 t->neigh_vars[4].strategy = strategy;
2758 t->neigh_vars[4].extra1 = dev;
d12af679
EB
2759 if (!strategy)
2760 t->neigh_vars[4].ctl_name = CTL_UNNUMBERED;
1da177e4 2761 /* RetransTime (in milliseconds)*/
d12af679
EB
2762 t->neigh_vars[12].proc_handler = handler;
2763 t->neigh_vars[12].strategy = strategy;
2764 t->neigh_vars[12].extra1 = dev;
2765 if (!strategy)
2766 t->neigh_vars[12].ctl_name = CTL_UNNUMBERED;
1da177e4 2767 /* ReachableTime (in milliseconds) */
d12af679
EB
2768 t->neigh_vars[13].proc_handler = handler;
2769 t->neigh_vars[13].strategy = strategy;
2770 t->neigh_vars[13].extra1 = dev;
2771 if (!strategy)
2772 t->neigh_vars[13].ctl_name = CTL_UNNUMBERED;
1da177e4
LT
2773 }
2774
c3bac5a7
PE
2775 t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2776 if (!t->dev_name)
1da177e4 2777 goto free;
1da177e4 2778
c3bac5a7
PE
2779 neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2780 neigh_path[NEIGH_CTL_PATH_NEIGH].ctl_name = pdev_id;
2781 neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
2782 neigh_path[NEIGH_CTL_PATH_PROTO].ctl_name = p_id;
1da177e4 2783
4ab438fc 2784 t->sysctl_header =
57da52c1 2785 register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
3c607bbb 2786 if (!t->sysctl_header)
1da177e4 2787 goto free_procname;
3c607bbb 2788
1da177e4
LT
2789 p->sysctl_table = t;
2790 return 0;
2791
3c607bbb 2792free_procname:
c3bac5a7 2793 kfree(t->dev_name);
3c607bbb 2794free:
1da177e4 2795 kfree(t);
3c607bbb
PE
2796err:
2797 return -ENOBUFS;
1da177e4 2798}
0a204500 2799EXPORT_SYMBOL(neigh_sysctl_register);
1da177e4
LT
2800
2801void neigh_sysctl_unregister(struct neigh_parms *p)
2802{
2803 if (p->sysctl_table) {
2804 struct neigh_sysctl_table *t = p->sysctl_table;
2805 p->sysctl_table = NULL;
2806 unregister_sysctl_table(t->sysctl_header);
c3bac5a7 2807 kfree(t->dev_name);
1da177e4
LT
2808 kfree(t);
2809 }
2810}
0a204500 2811EXPORT_SYMBOL(neigh_sysctl_unregister);
1da177e4
LT
2812
2813#endif /* CONFIG_SYSCTL */
2814
c8822a4e
TG
2815static int __init neigh_init(void)
2816{
2817 rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL);
2818 rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL);
2819 rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info);
2820
2821 rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info);
2822 rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL);
2823
2824 return 0;
2825}
2826
2827subsys_initcall(neigh_init);
2828