~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/core/neighbour.c

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.77 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

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

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp