Linux Audio

Check our new training course

Loading...
v3.1
 
   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#include <linux/slab.h>
 
  19#include <linux/types.h>
  20#include <linux/kernel.h>
  21#include <linux/module.h>
  22#include <linux/socket.h>
  23#include <linux/netdevice.h>
  24#include <linux/proc_fs.h>
  25#ifdef CONFIG_SYSCTL
  26#include <linux/sysctl.h>
  27#endif
  28#include <linux/times.h>
  29#include <net/net_namespace.h>
  30#include <net/neighbour.h>
 
  31#include <net/dst.h>
  32#include <net/sock.h>
  33#include <net/netevent.h>
  34#include <net/netlink.h>
  35#include <linux/rtnetlink.h>
  36#include <linux/random.h>
  37#include <linux/string.h>
  38#include <linux/log2.h>
 
 
  39
  40#define NEIGH_DEBUG 1
  41
  42#define NEIGH_PRINTK(x...) printk(x)
  43#define NEIGH_NOPRINTK(x...) do { ; } while(0)
  44#define NEIGH_PRINTK1 NEIGH_NOPRINTK
  45#define NEIGH_PRINTK2 NEIGH_NOPRINTK
  46
  47#if NEIGH_DEBUG >= 1
  48#undef NEIGH_PRINTK1
  49#define NEIGH_PRINTK1 NEIGH_PRINTK
  50#endif
  51#if NEIGH_DEBUG >= 2
  52#undef NEIGH_PRINTK2
  53#define NEIGH_PRINTK2 NEIGH_PRINTK
  54#endif
  55
  56#define PNEIGH_HASHMASK		0xF
  57
  58static void neigh_timer_handler(unsigned long arg);
  59static void __neigh_notify(struct neighbour *n, int type, int flags);
  60static void neigh_update_notify(struct neighbour *neigh);
  61static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev);
 
 
  62
  63static struct neigh_table *neigh_tables;
  64#ifdef CONFIG_PROC_FS
  65static const struct file_operations neigh_stat_seq_fops;
  66#endif
  67
  68/*
  69   Neighbour hash table buckets are protected with rwlock tbl->lock.
  70
  71   - All the scans/updates to hash buckets MUST be made under this lock.
  72   - NOTHING clever should be made under this lock: no callbacks
  73     to protocol backends, no attempts to send something to network.
  74     It will result in deadlocks, if backend/driver wants to use neighbour
  75     cache.
  76   - If the entry requires some non-trivial actions, increase
  77     its reference count and release table lock.
  78
  79   Neighbour entries are protected:
  80   - with reference count.
  81   - with rwlock neigh->lock
  82
  83   Reference count prevents destruction.
  84
  85   neigh->lock mainly serializes ll address data and its validity state.
  86   However, the same lock is used to protect another entry fields:
  87    - timer
  88    - resolution queue
  89
  90   Again, nothing clever shall be made under neigh->lock,
  91   the most complicated procedure, which we allow is dev->hard_header.
  92   It is supposed, that dev->hard_header is simplistic and does
  93   not make callbacks to neighbour tables.
  94
  95   The last lock is neigh_tbl_lock. It is pure SMP lock, protecting
  96   list of neighbour tables. This list is used only in process context,
  97 */
  98
  99static DEFINE_RWLOCK(neigh_tbl_lock);
 100
 101static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
 102{
 103	kfree_skb(skb);
 104	return -ENETDOWN;
 105}
 106
 107static void neigh_cleanup_and_release(struct neighbour *neigh)
 108{
 109	if (neigh->parms->neigh_cleanup)
 110		neigh->parms->neigh_cleanup(neigh);
 111
 112	__neigh_notify(neigh, RTM_DELNEIGH, 0);
 113	neigh_release(neigh);
 114}
 115
 116/*
 117 * It is random distribution in the interval (1/2)*base...(3/2)*base.
 118 * It corresponds to default IPv6 settings and is not overridable,
 119 * because it is really reasonable choice.
 120 */
 121
 122unsigned long neigh_rand_reach_time(unsigned long base)
 123{
 124	return base ? (net_random() % base) + (base >> 1) : 0;
 125}
 126EXPORT_SYMBOL(neigh_rand_reach_time);
 127
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 128
 129static int neigh_forced_gc(struct neigh_table *tbl)
 130{
 
 
 
 131	int shrunk = 0;
 132	int i;
 133	struct neigh_hash_table *nht;
 134
 135	NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
 136
 137	write_lock_bh(&tbl->lock);
 138	nht = rcu_dereference_protected(tbl->nht,
 139					lockdep_is_held(&tbl->lock));
 140	for (i = 0; i < (1 << nht->hash_shift); i++) {
 141		struct neighbour *n;
 142		struct neighbour __rcu **np;
 143
 144		np = &nht->hash_buckets[i];
 145		while ((n = rcu_dereference_protected(*np,
 146					lockdep_is_held(&tbl->lock))) != NULL) {
 147			/* Neighbour record may be discarded if:
 148			 * - nobody refers to it.
 149			 * - it is not permanent
 150			 */
 151			write_lock(&n->lock);
 152			if (atomic_read(&n->refcnt) == 1 &&
 153			    !(n->nud_state & NUD_PERMANENT)) {
 154				rcu_assign_pointer(*np,
 155					rcu_dereference_protected(n->next,
 156						  lockdep_is_held(&tbl->lock)));
 157				n->dead = 1;
 158				shrunk	= 1;
 159				write_unlock(&n->lock);
 160				neigh_cleanup_and_release(n);
 161				continue;
 162			}
 163			write_unlock(&n->lock);
 164			np = &n->next;
 
 
 
 
 165		}
 166	}
 167
 168	tbl->last_flush = jiffies;
 169
 170	write_unlock_bh(&tbl->lock);
 171
 172	return shrunk;
 173}
 174
 175static void neigh_add_timer(struct neighbour *n, unsigned long when)
 176{
 177	neigh_hold(n);
 178	if (unlikely(mod_timer(&n->timer, when))) {
 179		printk("NEIGH: BUG, double timer add, state is %x\n",
 180		       n->nud_state);
 181		dump_stack();
 182	}
 183}
 184
 185static int neigh_del_timer(struct neighbour *n)
 186{
 187	if ((n->nud_state & NUD_IN_TIMER) &&
 188	    del_timer(&n->timer)) {
 189		neigh_release(n);
 190		return 1;
 191	}
 192	return 0;
 193}
 194
 195static void pneigh_queue_purge(struct sk_buff_head *list)
 196{
 197	struct sk_buff *skb;
 198
 199	while ((skb = skb_dequeue(list)) != NULL) {
 200		dev_put(skb->dev);
 201		kfree_skb(skb);
 202	}
 203}
 204
 205static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev)
 
 206{
 207	int i;
 208	struct neigh_hash_table *nht;
 209
 210	nht = rcu_dereference_protected(tbl->nht,
 211					lockdep_is_held(&tbl->lock));
 212
 213	for (i = 0; i < (1 << nht->hash_shift); i++) {
 214		struct neighbour *n;
 215		struct neighbour __rcu **np = &nht->hash_buckets[i];
 216
 217		while ((n = rcu_dereference_protected(*np,
 218					lockdep_is_held(&tbl->lock))) != NULL) {
 219			if (dev && n->dev != dev) {
 220				np = &n->next;
 221				continue;
 222			}
 
 
 
 
 223			rcu_assign_pointer(*np,
 224				   rcu_dereference_protected(n->next,
 225						lockdep_is_held(&tbl->lock)));
 226			write_lock(&n->lock);
 227			neigh_del_timer(n);
 228			n->dead = 1;
 229
 230			if (atomic_read(&n->refcnt) != 1) {
 231				/* The most unpleasant situation.
 232				   We must destroy neighbour entry,
 233				   but someone still uses it.
 234
 235				   The destroy will be delayed until
 236				   the last user releases us, but
 237				   we must kill timers etc. and move
 238				   it to safe state.
 239				 */
 240				skb_queue_purge(&n->arp_queue);
 
 241				n->output = neigh_blackhole;
 242				if (n->nud_state & NUD_VALID)
 243					n->nud_state = NUD_NOARP;
 244				else
 245					n->nud_state = NUD_NONE;
 246				NEIGH_PRINTK2("neigh %p is stray.\n", n);
 247			}
 248			write_unlock(&n->lock);
 249			neigh_cleanup_and_release(n);
 250		}
 251	}
 252}
 253
 254void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
 255{
 256	write_lock_bh(&tbl->lock);
 257	neigh_flush_dev(tbl, dev);
 258	write_unlock_bh(&tbl->lock);
 259}
 260EXPORT_SYMBOL(neigh_changeaddr);
 261
 262int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
 
 263{
 264	write_lock_bh(&tbl->lock);
 265	neigh_flush_dev(tbl, dev);
 266	pneigh_ifdown(tbl, dev);
 267	write_unlock_bh(&tbl->lock);
 268
 269	del_timer_sync(&tbl->proxy_timer);
 270	pneigh_queue_purge(&tbl->proxy_queue);
 271	return 0;
 272}
 
 
 
 
 
 
 
 
 
 
 
 
 
 273EXPORT_SYMBOL(neigh_ifdown);
 274
 275static struct neighbour *neigh_alloc(struct neigh_table *tbl)
 
 
 276{
 277	struct neighbour *n = NULL;
 278	unsigned long now = jiffies;
 279	int entries;
 280
 281	entries = atomic_inc_return(&tbl->entries) - 1;
 
 
 
 282	if (entries >= tbl->gc_thresh3 ||
 283	    (entries >= tbl->gc_thresh2 &&
 284	     time_after(now, tbl->last_flush + 5 * HZ))) {
 285		if (!neigh_forced_gc(tbl) &&
 286		    entries >= tbl->gc_thresh3)
 
 
 
 287			goto out_entries;
 
 288	}
 289
 290	n = kmem_cache_zalloc(tbl->kmem_cachep, GFP_ATOMIC);
 
 291	if (!n)
 292		goto out_entries;
 293
 294	skb_queue_head_init(&n->arp_queue);
 295	rwlock_init(&n->lock);
 296	seqlock_init(&n->ha_lock);
 297	n->updated	  = n->used = now;
 298	n->nud_state	  = NUD_NONE;
 299	n->output	  = neigh_blackhole;
 300	seqlock_init(&n->hh.hh_lock);
 301	n->parms	  = neigh_parms_clone(&tbl->parms);
 302	setup_timer(&n->timer, neigh_timer_handler, (unsigned long)n);
 303
 304	NEIGH_CACHE_STAT_INC(tbl, allocs);
 305	n->tbl		  = tbl;
 306	atomic_set(&n->refcnt, 1);
 307	n->dead		  = 1;
 
 
 
 308out:
 309	return n;
 310
 311out_entries:
 312	atomic_dec(&tbl->entries);
 
 313	goto out;
 314}
 315
 
 
 
 
 
 316static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
 317{
 318	size_t size = (1 << shift) * sizeof(struct neighbour *);
 319	struct neigh_hash_table *ret;
 320	struct neighbour __rcu **buckets;
 
 321
 322	ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
 323	if (!ret)
 324		return NULL;
 325	if (size <= PAGE_SIZE)
 326		buckets = kzalloc(size, GFP_ATOMIC);
 327	else
 328		buckets = (struct neighbour __rcu **)
 329			  __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
 330					   get_order(size));
 
 
 331	if (!buckets) {
 332		kfree(ret);
 333		return NULL;
 334	}
 335	ret->hash_buckets = buckets;
 336	ret->hash_shift = shift;
 337	get_random_bytes(&ret->hash_rnd, sizeof(ret->hash_rnd));
 338	ret->hash_rnd |= 1;
 339	return ret;
 340}
 341
 342static void neigh_hash_free_rcu(struct rcu_head *head)
 343{
 344	struct neigh_hash_table *nht = container_of(head,
 345						    struct neigh_hash_table,
 346						    rcu);
 347	size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
 348	struct neighbour __rcu **buckets = nht->hash_buckets;
 349
 350	if (size <= PAGE_SIZE)
 351		kfree(buckets);
 352	else
 
 353		free_pages((unsigned long)buckets, get_order(size));
 
 354	kfree(nht);
 355}
 356
 357static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
 358						unsigned long new_shift)
 359{
 360	unsigned int i, hash;
 361	struct neigh_hash_table *new_nht, *old_nht;
 362
 363	NEIGH_CACHE_STAT_INC(tbl, hash_grows);
 364
 365	old_nht = rcu_dereference_protected(tbl->nht,
 366					    lockdep_is_held(&tbl->lock));
 367	new_nht = neigh_hash_alloc(new_shift);
 368	if (!new_nht)
 369		return old_nht;
 370
 371	for (i = 0; i < (1 << old_nht->hash_shift); i++) {
 372		struct neighbour *n, *next;
 373
 374		for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
 375						   lockdep_is_held(&tbl->lock));
 376		     n != NULL;
 377		     n = next) {
 378			hash = tbl->hash(n->primary_key, n->dev,
 379					 new_nht->hash_rnd);
 380
 381			hash >>= (32 - new_nht->hash_shift);
 382			next = rcu_dereference_protected(n->next,
 383						lockdep_is_held(&tbl->lock));
 384
 385			rcu_assign_pointer(n->next,
 386					   rcu_dereference_protected(
 387						new_nht->hash_buckets[hash],
 388						lockdep_is_held(&tbl->lock)));
 389			rcu_assign_pointer(new_nht->hash_buckets[hash], n);
 390		}
 391	}
 392
 393	rcu_assign_pointer(tbl->nht, new_nht);
 394	call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
 395	return new_nht;
 396}
 397
 398struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
 399			       struct net_device *dev)
 400{
 401	struct neighbour *n;
 402	int key_len = tbl->key_len;
 403	u32 hash_val;
 404	struct neigh_hash_table *nht;
 405
 406	NEIGH_CACHE_STAT_INC(tbl, lookups);
 407
 408	rcu_read_lock_bh();
 409	nht = rcu_dereference_bh(tbl->nht);
 410	hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
 411
 412	for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
 413	     n != NULL;
 414	     n = rcu_dereference_bh(n->next)) {
 415		if (dev == n->dev && !memcmp(n->primary_key, pkey, key_len)) {
 416			if (!atomic_inc_not_zero(&n->refcnt))
 417				n = NULL;
 418			NEIGH_CACHE_STAT_INC(tbl, hits);
 419			break;
 420		}
 421	}
 422
 423	rcu_read_unlock_bh();
 424	return n;
 425}
 426EXPORT_SYMBOL(neigh_lookup);
 427
 428struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
 429				     const void *pkey)
 430{
 431	struct neighbour *n;
 432	int key_len = tbl->key_len;
 433	u32 hash_val;
 434	struct neigh_hash_table *nht;
 435
 436	NEIGH_CACHE_STAT_INC(tbl, lookups);
 437
 438	rcu_read_lock_bh();
 439	nht = rcu_dereference_bh(tbl->nht);
 440	hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
 441
 442	for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
 443	     n != NULL;
 444	     n = rcu_dereference_bh(n->next)) {
 445		if (!memcmp(n->primary_key, pkey, key_len) &&
 446		    net_eq(dev_net(n->dev), net)) {
 447			if (!atomic_inc_not_zero(&n->refcnt))
 448				n = NULL;
 449			NEIGH_CACHE_STAT_INC(tbl, hits);
 450			break;
 451		}
 452	}
 453
 454	rcu_read_unlock_bh();
 455	return n;
 456}
 457EXPORT_SYMBOL(neigh_lookup_nodev);
 458
 459struct neighbour *neigh_create(struct neigh_table *tbl, const void *pkey,
 460			       struct net_device *dev)
 
 
 461{
 
 462	u32 hash_val;
 463	int key_len = tbl->key_len;
 464	int error;
 465	struct neighbour *n1, *rc, *n = neigh_alloc(tbl);
 466	struct neigh_hash_table *nht;
 467
 
 
 468	if (!n) {
 469		rc = ERR_PTR(-ENOBUFS);
 470		goto out;
 471	}
 472
 473	memcpy(n->primary_key, pkey, key_len);
 474	n->dev = dev;
 475	dev_hold(dev);
 476
 477	/* Protocol specific setup. */
 478	if (tbl->constructor &&	(error = tbl->constructor(n)) < 0) {
 479		rc = ERR_PTR(error);
 480		goto out_neigh_release;
 481	}
 482
 
 
 
 
 
 
 
 
 483	/* Device specific setup. */
 484	if (n->parms->neigh_setup &&
 485	    (error = n->parms->neigh_setup(n)) < 0) {
 486		rc = ERR_PTR(error);
 487		goto out_neigh_release;
 488	}
 489
 490	n->confirmed = jiffies - (n->parms->base_reachable_time << 1);
 491
 492	write_lock_bh(&tbl->lock);
 493	nht = rcu_dereference_protected(tbl->nht,
 494					lockdep_is_held(&tbl->lock));
 495
 496	if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
 497		nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
 498
 499	hash_val = tbl->hash(pkey, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
 500
 501	if (n->parms->dead) {
 502		rc = ERR_PTR(-EINVAL);
 503		goto out_tbl_unlock;
 504	}
 505
 506	for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
 507					    lockdep_is_held(&tbl->lock));
 508	     n1 != NULL;
 509	     n1 = rcu_dereference_protected(n1->next,
 510			lockdep_is_held(&tbl->lock))) {
 511		if (dev == n1->dev && !memcmp(n1->primary_key, pkey, key_len)) {
 512			neigh_hold(n1);
 
 513			rc = n1;
 514			goto out_tbl_unlock;
 515		}
 516	}
 517
 518	n->dead = 0;
 519	neigh_hold(n);
 
 
 
 
 520	rcu_assign_pointer(n->next,
 521			   rcu_dereference_protected(nht->hash_buckets[hash_val],
 522						     lockdep_is_held(&tbl->lock)));
 523	rcu_assign_pointer(nht->hash_buckets[hash_val], n);
 524	write_unlock_bh(&tbl->lock);
 525	NEIGH_PRINTK2("neigh %p is created.\n", n);
 526	rc = n;
 527out:
 528	return rc;
 529out_tbl_unlock:
 530	write_unlock_bh(&tbl->lock);
 531out_neigh_release:
 
 
 532	neigh_release(n);
 533	goto out;
 534}
 535EXPORT_SYMBOL(neigh_create);
 536
 537static u32 pneigh_hash(const void *pkey, int key_len)
 
 
 
 
 
 
 
 538{
 539	u32 hash_val = *(u32 *)(pkey + key_len - 4);
 540	hash_val ^= (hash_val >> 16);
 541	hash_val ^= hash_val >> 8;
 542	hash_val ^= hash_val >> 4;
 543	hash_val &= PNEIGH_HASHMASK;
 544	return hash_val;
 545}
 546
 547static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
 548					      struct net *net,
 549					      const void *pkey,
 550					      int key_len,
 551					      struct net_device *dev)
 552{
 553	while (n) {
 554		if (!memcmp(n->key, pkey, key_len) &&
 555		    net_eq(pneigh_net(n), net) &&
 556		    (n->dev == dev || !n->dev))
 557			return n;
 558		n = n->next;
 559	}
 560	return NULL;
 561}
 562
 563struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
 564		struct net *net, const void *pkey, struct net_device *dev)
 565{
 566	int key_len = tbl->key_len;
 567	u32 hash_val = pneigh_hash(pkey, key_len);
 568
 569	return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
 570				 net, pkey, key_len, dev);
 571}
 572EXPORT_SYMBOL_GPL(__pneigh_lookup);
 573
 574struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
 575				    struct net *net, const void *pkey,
 576				    struct net_device *dev, int creat)
 577{
 578	struct pneigh_entry *n;
 579	int key_len = tbl->key_len;
 580	u32 hash_val = pneigh_hash(pkey, key_len);
 581
 582	read_lock_bh(&tbl->lock);
 583	n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
 584			      net, pkey, key_len, dev);
 585	read_unlock_bh(&tbl->lock);
 586
 587	if (n || !creat)
 588		goto out;
 589
 590	ASSERT_RTNL();
 591
 592	n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
 593	if (!n)
 594		goto out;
 595
 596	write_pnet(&n->net, hold_net(net));
 
 597	memcpy(n->key, pkey, key_len);
 598	n->dev = dev;
 599	if (dev)
 600		dev_hold(dev);
 601
 602	if (tbl->pconstructor && tbl->pconstructor(n)) {
 603		if (dev)
 604			dev_put(dev);
 605		release_net(net);
 606		kfree(n);
 607		n = NULL;
 608		goto out;
 609	}
 610
 611	write_lock_bh(&tbl->lock);
 612	n->next = tbl->phash_buckets[hash_val];
 613	tbl->phash_buckets[hash_val] = n;
 614	write_unlock_bh(&tbl->lock);
 615out:
 616	return n;
 617}
 618EXPORT_SYMBOL(pneigh_lookup);
 619
 620
 621int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
 622		  struct net_device *dev)
 623{
 624	struct pneigh_entry *n, **np;
 625	int key_len = tbl->key_len;
 626	u32 hash_val = pneigh_hash(pkey, key_len);
 627
 628	write_lock_bh(&tbl->lock);
 629	for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
 630	     np = &n->next) {
 631		if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
 632		    net_eq(pneigh_net(n), net)) {
 633			*np = n->next;
 634			write_unlock_bh(&tbl->lock);
 635			if (tbl->pdestructor)
 636				tbl->pdestructor(n);
 637			if (n->dev)
 638				dev_put(n->dev);
 639			release_net(pneigh_net(n));
 640			kfree(n);
 641			return 0;
 642		}
 643	}
 644	write_unlock_bh(&tbl->lock);
 645	return -ENOENT;
 646}
 647
 648static int pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
 
 649{
 650	struct pneigh_entry *n, **np;
 651	u32 h;
 652
 653	for (h = 0; h <= PNEIGH_HASHMASK; h++) {
 654		np = &tbl->phash_buckets[h];
 655		while ((n = *np) != NULL) {
 656			if (!dev || n->dev == dev) {
 657				*np = n->next;
 658				if (tbl->pdestructor)
 659					tbl->pdestructor(n);
 660				if (n->dev)
 661					dev_put(n->dev);
 662				release_net(pneigh_net(n));
 663				kfree(n);
 664				continue;
 665			}
 666			np = &n->next;
 667		}
 668	}
 
 
 
 
 
 
 
 
 
 
 669	return -ENOENT;
 670}
 671
 672static void neigh_parms_destroy(struct neigh_parms *parms);
 673
 674static 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
 680static void neigh_destroy_rcu(struct rcu_head *head)
 681{
 682	struct neighbour *neigh = container_of(head, struct neighbour, rcu);
 683
 684	kmem_cache_free(neigh->tbl->kmem_cachep, neigh);
 685}
 686/*
 687 *	neighbour must already be out of the table;
 688 *
 689 */
 690void neigh_destroy(struct neighbour *neigh)
 691{
 
 
 692	NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
 693
 694	if (!neigh->dead) {
 695		printk(KERN_WARNING
 696		       "Destroying alive neighbour %p\n", neigh);
 697		dump_stack();
 698		return;
 699	}
 700
 701	if (neigh_del_timer(neigh))
 702		printk(KERN_WARNING "Impossible event.\n");
 703
 704	skb_queue_purge(&neigh->arp_queue);
 
 
 
 705
 706	dev_put(neigh->dev);
 
 
 
 707	neigh_parms_put(neigh->parms);
 708
 709	NEIGH_PRINTK2("neigh %p is destroyed.\n", neigh);
 710
 711	atomic_dec(&neigh->tbl->entries);
 712	call_rcu(&neigh->rcu, neigh_destroy_rcu);
 713}
 714EXPORT_SYMBOL(neigh_destroy);
 715
 716/* Neighbour state is suspicious;
 717   disable fast path.
 718
 719   Called with write_locked neigh.
 720 */
 721static void neigh_suspect(struct neighbour *neigh)
 722{
 723	NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
 724
 725	neigh->output = neigh->ops->output;
 726}
 727
 728/* Neighbour state is OK;
 729   enable fast path.
 730
 731   Called with write_locked neigh.
 732 */
 733static void neigh_connect(struct neighbour *neigh)
 734{
 735	NEIGH_PRINTK2("neigh %p is connected.\n", neigh);
 736
 737	neigh->output = neigh->ops->connected_output;
 738}
 739
 740static void neigh_periodic_work(struct work_struct *work)
 741{
 742	struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
 743	struct neighbour *n;
 744	struct neighbour __rcu **np;
 745	unsigned int i;
 746	struct neigh_hash_table *nht;
 747
 748	NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
 749
 750	write_lock_bh(&tbl->lock);
 751	nht = rcu_dereference_protected(tbl->nht,
 752					lockdep_is_held(&tbl->lock));
 753
 754	/*
 755	 *	periodically recompute ReachableTime from random function
 756	 */
 757
 758	if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
 759		struct neigh_parms *p;
 760		tbl->last_rand = jiffies;
 761		for (p = &tbl->parms; p; p = p->next)
 762			p->reachable_time =
 763				neigh_rand_reach_time(p->base_reachable_time);
 764	}
 765
 
 
 
 766	for (i = 0 ; i < (1 << nht->hash_shift); i++) {
 767		np = &nht->hash_buckets[i];
 768
 769		while ((n = rcu_dereference_protected(*np,
 770				lockdep_is_held(&tbl->lock))) != NULL) {
 771			unsigned int state;
 772
 773			write_lock(&n->lock);
 774
 775			state = n->nud_state;
 776			if (state & (NUD_PERMANENT | NUD_IN_TIMER)) {
 
 777				write_unlock(&n->lock);
 778				goto next_elt;
 779			}
 780
 781			if (time_before(n->used, n->confirmed))
 782				n->used = n->confirmed;
 783
 784			if (atomic_read(&n->refcnt) == 1 &&
 785			    (state == NUD_FAILED ||
 786			     time_after(jiffies, n->used + n->parms->gc_staletime))) {
 787				*np = n->next;
 788				n->dead = 1;
 789				write_unlock(&n->lock);
 790				neigh_cleanup_and_release(n);
 791				continue;
 792			}
 793			write_unlock(&n->lock);
 794
 795next_elt:
 796			np = &n->next;
 797		}
 798		/*
 799		 * It's fine to release lock here, even if hash table
 800		 * grows while we are preempted.
 801		 */
 802		write_unlock_bh(&tbl->lock);
 803		cond_resched();
 804		write_lock_bh(&tbl->lock);
 
 
 805	}
 806	/* Cycle through all hash buckets every base_reachable_time/2 ticks.
 807	 * ARP entry timeouts range from 1/2 base_reachable_time to 3/2
 808	 * base_reachable_time.
 
 809	 */
 810	schedule_delayed_work(&tbl->gc_work,
 811			      tbl->parms.base_reachable_time >> 1);
 812	write_unlock_bh(&tbl->lock);
 813}
 814
 815static __inline__ int neigh_max_probes(struct neighbour *n)
 816{
 817	struct neigh_parms *p = n->parms;
 818	return (n->nud_state & NUD_PROBE) ?
 819		p->ucast_probes :
 820		p->ucast_probes + p->app_probes + p->mcast_probes;
 821}
 822
 823static void neigh_invalidate(struct neighbour *neigh)
 824	__releases(neigh->lock)
 825	__acquires(neigh->lock)
 826{
 827	struct sk_buff *skb;
 828
 829	NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
 830	NEIGH_PRINTK2("neigh %p is failed.\n", neigh);
 831	neigh->updated = jiffies;
 832
 833	/* It is very thin place. report_unreachable is very complicated
 834	   routine. Particularly, it can hit the same neighbour entry!
 835
 836	   So that, we try to be accurate and avoid dead loop. --ANK
 837	 */
 838	while (neigh->nud_state == NUD_FAILED &&
 839	       (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
 840		write_unlock(&neigh->lock);
 841		neigh->ops->error_report(neigh, skb);
 842		write_lock(&neigh->lock);
 843	}
 844	skb_queue_purge(&neigh->arp_queue);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 845}
 846
 847/* Called when a timer expires for a neighbour entry. */
 848
 849static void neigh_timer_handler(unsigned long arg)
 850{
 851	unsigned long now, next;
 852	struct neighbour *neigh = (struct neighbour *)arg;
 853	unsigned state;
 854	int notify = 0;
 855
 856	write_lock(&neigh->lock);
 857
 858	state = neigh->nud_state;
 859	now = jiffies;
 860	next = now + HZ;
 861
 862	if (!(state & NUD_IN_TIMER)) {
 863#ifndef CONFIG_SMP
 864		printk(KERN_WARNING "neigh: timer & !nud_in_timer\n");
 865#endif
 866		goto out;
 867	}
 868
 869	if (state & NUD_REACHABLE) {
 870		if (time_before_eq(now,
 871				   neigh->confirmed + neigh->parms->reachable_time)) {
 872			NEIGH_PRINTK2("neigh %p is still alive.\n", neigh);
 873			next = neigh->confirmed + neigh->parms->reachable_time;
 874		} else if (time_before_eq(now,
 875					  neigh->used + neigh->parms->delay_probe_time)) {
 876			NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
 
 877			neigh->nud_state = NUD_DELAY;
 878			neigh->updated = jiffies;
 879			neigh_suspect(neigh);
 880			next = now + neigh->parms->delay_probe_time;
 881		} else {
 882			NEIGH_PRINTK2("neigh %p is suspected.\n", neigh);
 883			neigh->nud_state = NUD_STALE;
 884			neigh->updated = jiffies;
 885			neigh_suspect(neigh);
 886			notify = 1;
 887		}
 888	} else if (state & NUD_DELAY) {
 889		if (time_before_eq(now,
 890				   neigh->confirmed + neigh->parms->delay_probe_time)) {
 891			NEIGH_PRINTK2("neigh %p is now reachable.\n", neigh);
 
 892			neigh->nud_state = NUD_REACHABLE;
 893			neigh->updated = jiffies;
 894			neigh_connect(neigh);
 895			notify = 1;
 896			next = neigh->confirmed + neigh->parms->reachable_time;
 897		} else {
 898			NEIGH_PRINTK2("neigh %p is probed.\n", neigh);
 899			neigh->nud_state = NUD_PROBE;
 900			neigh->updated = jiffies;
 901			atomic_set(&neigh->probes, 0);
 902			next = now + neigh->parms->retrans_time;
 
 
 903		}
 904	} else {
 905		/* NUD_PROBE|NUD_INCOMPLETE */
 906		next = now + neigh->parms->retrans_time;
 907	}
 908
 909	if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
 910	    atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
 911		neigh->nud_state = NUD_FAILED;
 912		notify = 1;
 913		neigh_invalidate(neigh);
 
 914	}
 915
 916	if (neigh->nud_state & NUD_IN_TIMER) {
 917		if (time_before(next, jiffies + HZ/2))
 918			next = jiffies + HZ/2;
 919		if (!mod_timer(&neigh->timer, next))
 920			neigh_hold(neigh);
 921	}
 922	if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
 923		struct sk_buff *skb = skb_peek(&neigh->arp_queue);
 924		/* keep skb alive even if arp_queue overflows */
 925		if (skb)
 926			skb = skb_copy(skb, GFP_ATOMIC);
 927		write_unlock(&neigh->lock);
 928		neigh->ops->solicit(neigh, skb);
 929		atomic_inc(&neigh->probes);
 930		kfree_skb(skb);
 931	} else {
 932out:
 933		write_unlock(&neigh->lock);
 934	}
 935
 936	if (notify)
 937		neigh_update_notify(neigh);
 
 
 938
 939	neigh_release(neigh);
 940}
 941
 942int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
 943{
 944	int rc;
 945	unsigned long now;
 946
 947	write_lock_bh(&neigh->lock);
 948
 949	rc = 0;
 950	if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
 951		goto out_unlock_bh;
 952
 953	now = jiffies;
 954
 955	if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
 956		if (neigh->parms->mcast_probes + neigh->parms->app_probes) {
 957			atomic_set(&neigh->probes, neigh->parms->ucast_probes);
 
 
 
 
 
 958			neigh->nud_state     = NUD_INCOMPLETE;
 959			neigh->updated = jiffies;
 960			neigh_add_timer(neigh, now + 1);
 
 
 
 961		} else {
 962			neigh->nud_state = NUD_FAILED;
 963			neigh->updated = jiffies;
 964			write_unlock_bh(&neigh->lock);
 965
 966			kfree_skb(skb);
 967			return 1;
 968		}
 969	} else if (neigh->nud_state & NUD_STALE) {
 970		NEIGH_PRINTK2("neigh %p is delayed.\n", neigh);
 
 971		neigh->nud_state = NUD_DELAY;
 972		neigh->updated = jiffies;
 973		neigh_add_timer(neigh,
 974				jiffies + neigh->parms->delay_probe_time);
 975	}
 976
 977	if (neigh->nud_state == NUD_INCOMPLETE) {
 978		if (skb) {
 979			if (skb_queue_len(&neigh->arp_queue) >=
 980			    neigh->parms->queue_len) {
 981				struct sk_buff *buff;
 
 982				buff = __skb_dequeue(&neigh->arp_queue);
 
 
 
 983				kfree_skb(buff);
 984				NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
 985			}
 986			skb_dst_force(skb);
 987			__skb_queue_tail(&neigh->arp_queue, skb);
 
 988		}
 989		rc = 1;
 990	}
 991out_unlock_bh:
 992	write_unlock_bh(&neigh->lock);
 
 
 
 
 
 993	return rc;
 
 
 
 
 
 
 
 
 994}
 995EXPORT_SYMBOL(__neigh_event_send);
 996
 997static void neigh_update_hhs(struct neighbour *neigh)
 998{
 999	struct hh_cache *hh;
1000	void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1001		= NULL;
1002
1003	if (neigh->dev->header_ops)
1004		update = neigh->dev->header_ops->cache_update;
1005
1006	if (update) {
1007		hh = &neigh->hh;
1008		if (hh->hh_len) {
1009			write_seqlock_bh(&hh->hh_lock);
1010			update(hh, neigh->dev, neigh->ha);
1011			write_sequnlock_bh(&hh->hh_lock);
1012		}
1013	}
1014}
1015
1016
1017
1018/* Generic update routine.
1019   -- lladdr is new lladdr or NULL, if it is not supplied.
1020   -- new    is new state.
1021   -- flags
1022	NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1023				if it is different.
1024	NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1025				lladdr instead of overriding it
1026				if it is different.
1027				It also allows to retain current state
1028				if lladdr is unchanged.
1029	NEIGH_UPDATE_F_ADMIN	means that the change is administrative.
1030
1031	NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1032				NTF_ROUTER flag.
1033	NEIGH_UPDATE_F_ISROUTER	indicates if the neighbour is known as
1034				a router.
1035
1036   Caller MUST hold reference count on the entry.
1037 */
1038
1039int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1040		 u32 flags)
 
1041{
 
1042	u8 old;
1043	int err;
1044	int notify = 0;
1045	struct net_device *dev;
1046	int update_isrouter = 0;
1047
 
 
1048	write_lock_bh(&neigh->lock);
1049
1050	dev    = neigh->dev;
1051	old    = neigh->nud_state;
1052	err    = -EPERM;
1053
 
 
 
 
 
1054	if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1055	    (old & (NUD_NOARP | NUD_PERMANENT)))
1056		goto out;
1057
 
 
1058	if (!(new & NUD_VALID)) {
1059		neigh_del_timer(neigh);
1060		if (old & NUD_CONNECTED)
1061			neigh_suspect(neigh);
1062		neigh->nud_state = new;
1063		err = 0;
1064		notify = old & NUD_VALID;
1065		if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1066		    (new & NUD_FAILED)) {
1067			neigh_invalidate(neigh);
1068			notify = 1;
1069		}
1070		goto out;
1071	}
1072
1073	/* Compare new lladdr with cached one */
1074	if (!dev->addr_len) {
1075		/* First case: device needs no address. */
1076		lladdr = neigh->ha;
1077	} else if (lladdr) {
1078		/* The second case: if something is already cached
1079		   and a new address is proposed:
1080		   - compare new & old
1081		   - if they are different, check override flag
1082		 */
1083		if ((old & NUD_VALID) &&
1084		    !memcmp(lladdr, neigh->ha, dev->addr_len))
1085			lladdr = neigh->ha;
1086	} else {
1087		/* No address is supplied; if we know something,
1088		   use it, otherwise discard the request.
1089		 */
1090		err = -EINVAL;
1091		if (!(old & NUD_VALID))
 
1092			goto out;
 
1093		lladdr = neigh->ha;
1094	}
1095
 
 
 
1096	if (new & NUD_CONNECTED)
1097		neigh->confirmed = jiffies;
1098	neigh->updated = jiffies;
1099
1100	/* If entry was valid and address is not changed,
1101	   do not change entry state, if new one is STALE.
1102	 */
1103	err = 0;
1104	update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1105	if (old & NUD_VALID) {
1106		if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1107			update_isrouter = 0;
1108			if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1109			    (old & NUD_CONNECTED)) {
1110				lladdr = neigh->ha;
1111				new = NUD_STALE;
1112			} else
1113				goto out;
1114		} else {
1115			if (lladdr == neigh->ha && new == NUD_STALE &&
1116			    ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) ||
1117			     (old & NUD_CONNECTED))
1118			    )
1119				new = old;
1120		}
1121	}
1122
 
 
 
 
 
 
 
1123	if (new != old) {
1124		neigh_del_timer(neigh);
 
 
1125		if (new & NUD_IN_TIMER)
1126			neigh_add_timer(neigh, (jiffies +
1127						((new & NUD_REACHABLE) ?
1128						 neigh->parms->reachable_time :
1129						 0)));
1130		neigh->nud_state = new;
 
1131	}
1132
1133	if (lladdr != neigh->ha) {
1134		write_seqlock(&neigh->ha_lock);
1135		memcpy(&neigh->ha, lladdr, dev->addr_len);
1136		write_sequnlock(&neigh->ha_lock);
1137		neigh_update_hhs(neigh);
1138		if (!(new & NUD_CONNECTED))
1139			neigh->confirmed = jiffies -
1140				      (neigh->parms->base_reachable_time << 1);
1141		notify = 1;
1142	}
1143	if (new == old)
1144		goto out;
1145	if (new & NUD_CONNECTED)
1146		neigh_connect(neigh);
1147	else
1148		neigh_suspect(neigh);
1149	if (!(old & NUD_VALID)) {
1150		struct sk_buff *skb;
1151
1152		/* Again: avoid dead loop if something went wrong */
1153
1154		while (neigh->nud_state & NUD_VALID &&
1155		       (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1156			struct dst_entry *dst = skb_dst(skb);
1157			struct neighbour *n2, *n1 = neigh;
1158			write_unlock_bh(&neigh->lock);
1159			/* On shaper/eql skb->dst->neighbour != neigh :( */
1160			if (dst && (n2 = dst_get_neighbour(dst)) != NULL)
1161				n1 = n2;
 
 
 
 
 
 
 
 
 
 
 
 
 
1162			n1->output(n1, skb);
 
 
 
 
1163			write_lock_bh(&neigh->lock);
1164		}
1165		skb_queue_purge(&neigh->arp_queue);
 
1166	}
1167out:
1168	if (update_isrouter) {
1169		neigh->flags = (flags & NEIGH_UPDATE_F_ISROUTER) ?
1170			(neigh->flags | NTF_ROUTER) :
1171			(neigh->flags & ~NTF_ROUTER);
1172	}
1173	write_unlock_bh(&neigh->lock);
1174
 
 
 
1175	if (notify)
1176		neigh_update_notify(neigh);
 
 
1177
1178	return err;
1179}
 
 
 
 
 
 
1180EXPORT_SYMBOL(neigh_update);
1181
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1182struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1183				 u8 *lladdr, void *saddr,
1184				 struct net_device *dev)
1185{
1186	struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1187						 lladdr || !dev->addr_len);
1188	if (neigh)
1189		neigh_update(neigh, lladdr, NUD_STALE,
1190			     NEIGH_UPDATE_F_OVERRIDE);
1191	return neigh;
1192}
1193EXPORT_SYMBOL(neigh_event_ns);
1194
1195/* called with read_lock_bh(&n->lock); */
1196static void neigh_hh_init(struct neighbour *n, struct dst_entry *dst)
1197{
1198	struct net_device *dev = dst->dev;
1199	__be16 prot = dst->ops->protocol;
1200	struct hh_cache	*hh = &n->hh;
1201
1202	write_lock_bh(&n->lock);
1203
1204	/* Only one thread can come in here and initialize the
1205	 * hh_cache entry.
1206	 */
1207	if (!hh->hh_len)
1208		dev->header_ops->cache(n, hh, prot);
1209
1210	write_unlock_bh(&n->lock);
1211}
1212
1213/* This function can be used in contexts, where only old dev_queue_xmit
1214 * worked, f.e. if you want to override normal output path (eql, shaper),
1215 * but resolution is not made yet.
1216 */
1217
1218int neigh_compat_output(struct neighbour *neigh, struct sk_buff *skb)
1219{
1220	struct net_device *dev = skb->dev;
1221
1222	__skb_pull(skb, skb_network_offset(skb));
1223
1224	if (dev_hard_header(skb, dev, ntohs(skb->protocol), NULL, NULL,
1225			    skb->len) < 0 &&
1226	    dev->header_ops->rebuild(skb))
1227		return 0;
1228
1229	return dev_queue_xmit(skb);
1230}
1231EXPORT_SYMBOL(neigh_compat_output);
1232
1233/* Slow and careful. */
1234
1235int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1236{
1237	struct dst_entry *dst = skb_dst(skb);
1238	int rc = 0;
1239
1240	if (!dst)
1241		goto discard;
1242
1243	__skb_pull(skb, skb_network_offset(skb));
1244
1245	if (!neigh_event_send(neigh, skb)) {
1246		int err;
1247		struct net_device *dev = neigh->dev;
1248		unsigned int seq;
1249
1250		if (dev->header_ops->cache && !neigh->hh.hh_len)
1251			neigh_hh_init(neigh, dst);
1252
1253		do {
 
1254			seq = read_seqbegin(&neigh->ha_lock);
1255			err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1256					      neigh->ha, NULL, skb->len);
1257		} while (read_seqretry(&neigh->ha_lock, seq));
1258
1259		if (err >= 0)
1260			rc = dev_queue_xmit(skb);
1261		else
1262			goto out_kfree_skb;
1263	}
1264out:
1265	return rc;
1266discard:
1267	NEIGH_PRINTK1("neigh_resolve_output: dst=%p neigh=%p\n",
1268		      dst, neigh);
1269out_kfree_skb:
1270	rc = -EINVAL;
1271	kfree_skb(skb);
1272	goto out;
1273}
1274EXPORT_SYMBOL(neigh_resolve_output);
1275
1276/* As fast as possible without hh cache */
1277
1278int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1279{
1280	struct net_device *dev = neigh->dev;
1281	unsigned int seq;
1282	int err;
1283
1284	__skb_pull(skb, skb_network_offset(skb));
1285
1286	do {
 
1287		seq = read_seqbegin(&neigh->ha_lock);
1288		err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1289				      neigh->ha, NULL, skb->len);
1290	} while (read_seqretry(&neigh->ha_lock, seq));
1291
1292	if (err >= 0)
1293		err = dev_queue_xmit(skb);
1294	else {
1295		err = -EINVAL;
1296		kfree_skb(skb);
1297	}
1298	return err;
1299}
1300EXPORT_SYMBOL(neigh_connected_output);
1301
1302int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1303{
1304	return dev_queue_xmit(skb);
1305}
1306EXPORT_SYMBOL(neigh_direct_output);
1307
1308static void neigh_proxy_process(unsigned long arg)
1309{
1310	struct neigh_table *tbl = (struct neigh_table *)arg;
1311	long sched_next = 0;
1312	unsigned long now = jiffies;
1313	struct sk_buff *skb, *n;
1314
1315	spin_lock(&tbl->proxy_queue.lock);
1316
1317	skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1318		long tdif = NEIGH_CB(skb)->sched_next - now;
1319
1320		if (tdif <= 0) {
1321			struct net_device *dev = skb->dev;
1322
1323			__skb_unlink(skb, &tbl->proxy_queue);
1324			if (tbl->proxy_redo && netif_running(dev)) {
1325				rcu_read_lock();
1326				tbl->proxy_redo(skb);
1327				rcu_read_unlock();
1328			} else {
1329				kfree_skb(skb);
1330			}
1331
1332			dev_put(dev);
1333		} else if (!sched_next || tdif < sched_next)
1334			sched_next = tdif;
1335	}
1336	del_timer(&tbl->proxy_timer);
1337	if (sched_next)
1338		mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1339	spin_unlock(&tbl->proxy_queue.lock);
1340}
1341
1342void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1343		    struct sk_buff *skb)
1344{
1345	unsigned long now = jiffies;
1346	unsigned long sched_next = now + (net_random() % p->proxy_delay);
1347
1348	if (tbl->proxy_queue.qlen > p->proxy_qlen) {
1349		kfree_skb(skb);
1350		return;
1351	}
1352
1353	NEIGH_CB(skb)->sched_next = sched_next;
1354	NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1355
1356	spin_lock(&tbl->proxy_queue.lock);
1357	if (del_timer(&tbl->proxy_timer)) {
1358		if (time_before(tbl->proxy_timer.expires, sched_next))
1359			sched_next = tbl->proxy_timer.expires;
1360	}
1361	skb_dst_drop(skb);
1362	dev_hold(skb->dev);
1363	__skb_queue_tail(&tbl->proxy_queue, skb);
1364	mod_timer(&tbl->proxy_timer, sched_next);
1365	spin_unlock(&tbl->proxy_queue.lock);
1366}
1367EXPORT_SYMBOL(pneigh_enqueue);
1368
1369static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1370						      struct net *net, int ifindex)
1371{
1372	struct neigh_parms *p;
1373
1374	for (p = &tbl->parms; p; p = p->next) {
1375		if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1376		    (!p->dev && !ifindex))
1377			return p;
1378	}
1379
1380	return NULL;
1381}
1382
1383struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1384				      struct neigh_table *tbl)
1385{
1386	struct neigh_parms *p, *ref;
1387	struct net *net = dev_net(dev);
1388	const struct net_device_ops *ops = dev->netdev_ops;
1389
1390	ref = lookup_neigh_parms(tbl, net, 0);
1391	if (!ref)
1392		return NULL;
1393
1394	p = kmemdup(ref, sizeof(*p), GFP_KERNEL);
1395	if (p) {
1396		p->tbl		  = tbl;
1397		atomic_set(&p->refcnt, 1);
1398		p->reachable_time =
1399				neigh_rand_reach_time(p->base_reachable_time);
 
 
 
 
1400
1401		if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
 
1402			kfree(p);
1403			return NULL;
1404		}
1405
1406		dev_hold(dev);
1407		p->dev = dev;
1408		write_pnet(&p->net, hold_net(net));
1409		p->sysctl_table = NULL;
1410		write_lock_bh(&tbl->lock);
1411		p->next		= tbl->parms.next;
1412		tbl->parms.next = p;
1413		write_unlock_bh(&tbl->lock);
 
 
1414	}
1415	return p;
1416}
1417EXPORT_SYMBOL(neigh_parms_alloc);
1418
1419static void neigh_rcu_free_parms(struct rcu_head *head)
1420{
1421	struct neigh_parms *parms =
1422		container_of(head, struct neigh_parms, rcu_head);
1423
1424	neigh_parms_put(parms);
1425}
1426
1427void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1428{
1429	struct neigh_parms **p;
1430
1431	if (!parms || parms == &tbl->parms)
1432		return;
1433	write_lock_bh(&tbl->lock);
1434	for (p = &tbl->parms.next; *p; p = &(*p)->next) {
1435		if (*p == parms) {
1436			*p = parms->next;
1437			parms->dead = 1;
1438			write_unlock_bh(&tbl->lock);
1439			if (parms->dev)
1440				dev_put(parms->dev);
1441			call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1442			return;
1443		}
1444	}
1445	write_unlock_bh(&tbl->lock);
1446	NEIGH_PRINTK1("neigh_parms_release: not found\n");
 
 
1447}
1448EXPORT_SYMBOL(neigh_parms_release);
1449
1450static void neigh_parms_destroy(struct neigh_parms *parms)
1451{
1452	release_net(neigh_parms_net(parms));
1453	kfree(parms);
1454}
1455
1456static struct lock_class_key neigh_table_proxy_queue_class;
1457
1458void neigh_table_init_no_netlink(struct neigh_table *tbl)
 
 
1459{
1460	unsigned long now = jiffies;
1461	unsigned long phsize;
1462
 
 
 
1463	write_pnet(&tbl->parms.net, &init_net);
1464	atomic_set(&tbl->parms.refcnt, 1);
1465	tbl->parms.reachable_time =
1466			  neigh_rand_reach_time(tbl->parms.base_reachable_time);
1467
1468	if (!tbl->kmem_cachep)
1469		tbl->kmem_cachep =
1470			kmem_cache_create(tbl->id, tbl->entry_size, 0,
1471					  SLAB_HWCACHE_ALIGN|SLAB_PANIC,
1472					  NULL);
1473	tbl->stats = alloc_percpu(struct neigh_statistics);
1474	if (!tbl->stats)
1475		panic("cannot create neighbour cache statistics");
1476
1477#ifdef CONFIG_PROC_FS
1478	if (!proc_create_data(tbl->id, 0, init_net.proc_net_stat,
1479			      &neigh_stat_seq_fops, tbl))
1480		panic("cannot create neighbour proc dir entry");
1481#endif
1482
1483	RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1484
1485	phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1486	tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1487
1488	if (!tbl->nht || !tbl->phash_buckets)
1489		panic("cannot allocate neighbour cache hashes");
1490
 
 
 
 
 
 
1491	rwlock_init(&tbl->lock);
1492	INIT_DELAYED_WORK_DEFERRABLE(&tbl->gc_work, neigh_periodic_work);
1493	schedule_delayed_work(&tbl->gc_work, tbl->parms.reachable_time);
1494	setup_timer(&tbl->proxy_timer, neigh_proxy_process, (unsigned long)tbl);
 
1495	skb_queue_head_init_class(&tbl->proxy_queue,
1496			&neigh_table_proxy_queue_class);
1497
1498	tbl->last_flush = now;
1499	tbl->last_rand	= now + tbl->parms.reachable_time * 20;
1500}
1501EXPORT_SYMBOL(neigh_table_init_no_netlink);
1502
1503void neigh_table_init(struct neigh_table *tbl)
1504{
1505	struct neigh_table *tmp;
1506
1507	neigh_table_init_no_netlink(tbl);
1508	write_lock(&neigh_tbl_lock);
1509	for (tmp = neigh_tables; tmp; tmp = tmp->next) {
1510		if (tmp->family == tbl->family)
1511			break;
1512	}
1513	tbl->next	= neigh_tables;
1514	neigh_tables	= tbl;
1515	write_unlock(&neigh_tbl_lock);
1516
1517	if (unlikely(tmp)) {
1518		printk(KERN_ERR "NEIGH: Registering multiple tables for "
1519		       "family %d\n", tbl->family);
1520		dump_stack();
1521	}
1522}
1523EXPORT_SYMBOL(neigh_table_init);
1524
1525int neigh_table_clear(struct neigh_table *tbl)
1526{
1527	struct neigh_table **tp;
1528
1529	/* It is not clean... Fix it to unload IPv6 module safely */
1530	cancel_delayed_work_sync(&tbl->gc_work);
1531	del_timer_sync(&tbl->proxy_timer);
1532	pneigh_queue_purge(&tbl->proxy_queue);
1533	neigh_ifdown(tbl, NULL);
1534	if (atomic_read(&tbl->entries))
1535		printk(KERN_CRIT "neighbour leakage\n");
1536	write_lock(&neigh_tbl_lock);
1537	for (tp = &neigh_tables; *tp; tp = &(*tp)->next) {
1538		if (*tp == tbl) {
1539			*tp = tbl->next;
1540			break;
1541		}
1542	}
1543	write_unlock(&neigh_tbl_lock);
1544
1545	call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1546		 neigh_hash_free_rcu);
1547	tbl->nht = NULL;
1548
1549	kfree(tbl->phash_buckets);
1550	tbl->phash_buckets = NULL;
1551
1552	remove_proc_entry(tbl->id, init_net.proc_net_stat);
1553
1554	free_percpu(tbl->stats);
1555	tbl->stats = NULL;
1556
1557	kmem_cache_destroy(tbl->kmem_cachep);
1558	tbl->kmem_cachep = NULL;
1559
1560	return 0;
1561}
1562EXPORT_SYMBOL(neigh_table_clear);
1563
1564static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1565{
1566	struct net *net = sock_net(skb->sk);
1567	struct ndmsg *ndm;
1568	struct nlattr *dst_attr;
1569	struct neigh_table *tbl;
 
1570	struct net_device *dev = NULL;
1571	int err = -EINVAL;
1572
1573	ASSERT_RTNL();
1574	if (nlmsg_len(nlh) < sizeof(*ndm))
1575		goto out;
1576
1577	dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1578	if (dst_attr == NULL)
 
1579		goto out;
 
1580
1581	ndm = nlmsg_data(nlh);
1582	if (ndm->ndm_ifindex) {
1583		dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1584		if (dev == NULL) {
1585			err = -ENODEV;
1586			goto out;
1587		}
1588	}
1589
1590	read_lock(&neigh_tbl_lock);
1591	for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1592		struct neighbour *neigh;
1593
1594		if (tbl->family != ndm->ndm_family)
1595			continue;
1596		read_unlock(&neigh_tbl_lock);
1597
1598		if (nla_len(dst_attr) < tbl->key_len)
1599			goto out;
1600
1601		if (ndm->ndm_flags & NTF_PROXY) {
1602			err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1603			goto out;
1604		}
1605
1606		if (dev == NULL)
1607			goto out;
 
 
1608
1609		neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1610		if (neigh == NULL) {
1611			err = -ENOENT;
1612			goto out;
1613		}
1614
1615		err = neigh_update(neigh, NULL, NUD_FAILED,
1616				   NEIGH_UPDATE_F_OVERRIDE |
1617				   NEIGH_UPDATE_F_ADMIN);
1618		neigh_release(neigh);
1619		goto out;
1620	}
1621	read_unlock(&neigh_tbl_lock);
1622	err = -EAFNOSUPPORT;
 
 
 
 
 
 
1623
1624out:
1625	return err;
1626}
1627
1628static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 
1629{
 
 
1630	struct net *net = sock_net(skb->sk);
1631	struct ndmsg *ndm;
1632	struct nlattr *tb[NDA_MAX+1];
1633	struct neigh_table *tbl;
1634	struct net_device *dev = NULL;
 
 
 
1635	int err;
1636
1637	ASSERT_RTNL();
1638	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
 
1639	if (err < 0)
1640		goto out;
1641
1642	err = -EINVAL;
1643	if (tb[NDA_DST] == NULL)
 
1644		goto out;
 
1645
1646	ndm = nlmsg_data(nlh);
1647	if (ndm->ndm_ifindex) {
1648		dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1649		if (dev == NULL) {
1650			err = -ENODEV;
1651			goto out;
1652		}
1653
1654		if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len)
 
1655			goto out;
 
1656	}
1657
1658	read_lock(&neigh_tbl_lock);
1659	for (tbl = neigh_tables; tbl; tbl = tbl->next) {
1660		int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE;
1661		struct neighbour *neigh;
1662		void *dst, *lladdr;
1663
1664		if (tbl->family != ndm->ndm_family)
1665			continue;
1666		read_unlock(&neigh_tbl_lock);
 
1667
1668		if (nla_len(tb[NDA_DST]) < tbl->key_len)
1669			goto out;
1670		dst = nla_data(tb[NDA_DST]);
1671		lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
1672
1673		if (ndm->ndm_flags & NTF_PROXY) {
1674			struct pneigh_entry *pn;
1675
1676			err = -ENOBUFS;
1677			pn = pneigh_lookup(tbl, net, dst, dev, 1);
1678			if (pn) {
1679				pn->flags = ndm->ndm_flags;
1680				err = 0;
1681			}
1682			goto out;
 
 
 
1683		}
 
 
1684
1685		if (dev == NULL)
1686			goto out;
 
 
1687
1688		neigh = neigh_lookup(tbl, dst, dev);
1689		if (neigh == NULL) {
1690			if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1691				err = -ENOENT;
1692				goto out;
1693			}
1694
1695			neigh = __neigh_lookup_errno(tbl, dst, dev);
1696			if (IS_ERR(neigh)) {
1697				err = PTR_ERR(neigh);
1698				goto out;
1699			}
1700		} else {
1701			if (nlh->nlmsg_flags & NLM_F_EXCL) {
1702				err = -EEXIST;
1703				neigh_release(neigh);
1704				goto out;
1705			}
1706
1707			if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1708				flags &= ~NEIGH_UPDATE_F_OVERRIDE;
 
1709		}
1710
1711		if (ndm->ndm_flags & NTF_USE) {
1712			neigh_event_send(neigh, NULL);
1713			err = 0;
1714		} else
1715			err = neigh_update(neigh, lladdr, ndm->ndm_state, flags);
1716		neigh_release(neigh);
1717		goto out;
 
 
 
 
 
 
 
 
 
 
1718	}
1719
1720	read_unlock(&neigh_tbl_lock);
1721	err = -EAFNOSUPPORT;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1722out:
1723	return err;
1724}
1725
1726static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1727{
1728	struct nlattr *nest;
1729
1730	nest = nla_nest_start(skb, NDTA_PARMS);
1731	if (nest == NULL)
1732		return -ENOBUFS;
1733
1734	if (parms->dev)
1735		NLA_PUT_U32(skb, NDTPA_IFINDEX, parms->dev->ifindex);
1736
1737	NLA_PUT_U32(skb, NDTPA_REFCNT, atomic_read(&parms->refcnt));
1738	NLA_PUT_U32(skb, NDTPA_QUEUE_LEN, parms->queue_len);
1739	NLA_PUT_U32(skb, NDTPA_PROXY_QLEN, parms->proxy_qlen);
1740	NLA_PUT_U32(skb, NDTPA_APP_PROBES, parms->app_probes);
1741	NLA_PUT_U32(skb, NDTPA_UCAST_PROBES, parms->ucast_probes);
1742	NLA_PUT_U32(skb, NDTPA_MCAST_PROBES, parms->mcast_probes);
1743	NLA_PUT_MSECS(skb, NDTPA_REACHABLE_TIME, parms->reachable_time);
1744	NLA_PUT_MSECS(skb, NDTPA_BASE_REACHABLE_TIME,
1745		      parms->base_reachable_time);
1746	NLA_PUT_MSECS(skb, NDTPA_GC_STALETIME, parms->gc_staletime);
1747	NLA_PUT_MSECS(skb, NDTPA_DELAY_PROBE_TIME, parms->delay_probe_time);
1748	NLA_PUT_MSECS(skb, NDTPA_RETRANS_TIME, parms->retrans_time);
1749	NLA_PUT_MSECS(skb, NDTPA_ANYCAST_DELAY, parms->anycast_delay);
1750	NLA_PUT_MSECS(skb, NDTPA_PROXY_DELAY, parms->proxy_delay);
1751	NLA_PUT_MSECS(skb, NDTPA_LOCKTIME, parms->locktime);
1752
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1753	return nla_nest_end(skb, nest);
1754
1755nla_put_failure:
1756	nla_nest_cancel(skb, nest);
1757	return -EMSGSIZE;
1758}
1759
1760static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
1761			      u32 pid, u32 seq, int type, int flags)
1762{
1763	struct nlmsghdr *nlh;
1764	struct ndtmsg *ndtmsg;
1765
1766	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1767	if (nlh == NULL)
1768		return -EMSGSIZE;
1769
1770	ndtmsg = nlmsg_data(nlh);
1771
1772	read_lock_bh(&tbl->lock);
1773	ndtmsg->ndtm_family = tbl->family;
1774	ndtmsg->ndtm_pad1   = 0;
1775	ndtmsg->ndtm_pad2   = 0;
1776
1777	NLA_PUT_STRING(skb, NDTA_NAME, tbl->id);
1778	NLA_PUT_MSECS(skb, NDTA_GC_INTERVAL, tbl->gc_interval);
1779	NLA_PUT_U32(skb, NDTA_THRESH1, tbl->gc_thresh1);
1780	NLA_PUT_U32(skb, NDTA_THRESH2, tbl->gc_thresh2);
1781	NLA_PUT_U32(skb, NDTA_THRESH3, tbl->gc_thresh3);
1782
1783	{
1784		unsigned long now = jiffies;
1785		unsigned int flush_delta = now - tbl->last_flush;
1786		unsigned int rand_delta = now - tbl->last_rand;
1787		struct neigh_hash_table *nht;
1788		struct ndt_config ndc = {
1789			.ndtc_key_len		= tbl->key_len,
1790			.ndtc_entry_size	= tbl->entry_size,
1791			.ndtc_entries		= atomic_read(&tbl->entries),
1792			.ndtc_last_flush	= jiffies_to_msecs(flush_delta),
1793			.ndtc_last_rand		= jiffies_to_msecs(rand_delta),
1794			.ndtc_proxy_qlen	= tbl->proxy_queue.qlen,
1795		};
1796
1797		rcu_read_lock_bh();
1798		nht = rcu_dereference_bh(tbl->nht);
1799		ndc.ndtc_hash_rnd = nht->hash_rnd;
1800		ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
1801		rcu_read_unlock_bh();
1802
1803		NLA_PUT(skb, NDTA_CONFIG, sizeof(ndc), &ndc);
 
1804	}
1805
1806	{
1807		int cpu;
1808		struct ndt_stats ndst;
1809
1810		memset(&ndst, 0, sizeof(ndst));
1811
1812		for_each_possible_cpu(cpu) {
1813			struct neigh_statistics	*st;
1814
1815			st = per_cpu_ptr(tbl->stats, cpu);
1816			ndst.ndts_allocs		+= st->allocs;
1817			ndst.ndts_destroys		+= st->destroys;
1818			ndst.ndts_hash_grows		+= st->hash_grows;
1819			ndst.ndts_res_failed		+= st->res_failed;
1820			ndst.ndts_lookups		+= st->lookups;
1821			ndst.ndts_hits			+= st->hits;
1822			ndst.ndts_rcv_probes_mcast	+= st->rcv_probes_mcast;
1823			ndst.ndts_rcv_probes_ucast	+= st->rcv_probes_ucast;
1824			ndst.ndts_periodic_gc_runs	+= st->periodic_gc_runs;
1825			ndst.ndts_forced_gc_runs	+= st->forced_gc_runs;
 
1826		}
1827
1828		NLA_PUT(skb, NDTA_STATS, sizeof(ndst), &ndst);
 
 
1829	}
1830
1831	BUG_ON(tbl->parms.dev);
1832	if (neightbl_fill_parms(skb, &tbl->parms) < 0)
1833		goto nla_put_failure;
1834
1835	read_unlock_bh(&tbl->lock);
1836	return nlmsg_end(skb, nlh);
 
1837
1838nla_put_failure:
1839	read_unlock_bh(&tbl->lock);
1840	nlmsg_cancel(skb, nlh);
1841	return -EMSGSIZE;
1842}
1843
1844static int neightbl_fill_param_info(struct sk_buff *skb,
1845				    struct neigh_table *tbl,
1846				    struct neigh_parms *parms,
1847				    u32 pid, u32 seq, int type,
1848				    unsigned int flags)
1849{
1850	struct ndtmsg *ndtmsg;
1851	struct nlmsghdr *nlh;
1852
1853	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
1854	if (nlh == NULL)
1855		return -EMSGSIZE;
1856
1857	ndtmsg = nlmsg_data(nlh);
1858
1859	read_lock_bh(&tbl->lock);
1860	ndtmsg->ndtm_family = tbl->family;
1861	ndtmsg->ndtm_pad1   = 0;
1862	ndtmsg->ndtm_pad2   = 0;
1863
1864	if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
1865	    neightbl_fill_parms(skb, parms) < 0)
1866		goto errout;
1867
1868	read_unlock_bh(&tbl->lock);
1869	return nlmsg_end(skb, nlh);
 
1870errout:
1871	read_unlock_bh(&tbl->lock);
1872	nlmsg_cancel(skb, nlh);
1873	return -EMSGSIZE;
1874}
1875
1876static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
1877	[NDTA_NAME]		= { .type = NLA_STRING },
1878	[NDTA_THRESH1]		= { .type = NLA_U32 },
1879	[NDTA_THRESH2]		= { .type = NLA_U32 },
1880	[NDTA_THRESH3]		= { .type = NLA_U32 },
1881	[NDTA_GC_INTERVAL]	= { .type = NLA_U64 },
1882	[NDTA_PARMS]		= { .type = NLA_NESTED },
1883};
1884
1885static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
1886	[NDTPA_IFINDEX]			= { .type = NLA_U32 },
1887	[NDTPA_QUEUE_LEN]		= { .type = NLA_U32 },
1888	[NDTPA_PROXY_QLEN]		= { .type = NLA_U32 },
1889	[NDTPA_APP_PROBES]		= { .type = NLA_U32 },
1890	[NDTPA_UCAST_PROBES]		= { .type = NLA_U32 },
1891	[NDTPA_MCAST_PROBES]		= { .type = NLA_U32 },
 
1892	[NDTPA_BASE_REACHABLE_TIME]	= { .type = NLA_U64 },
1893	[NDTPA_GC_STALETIME]		= { .type = NLA_U64 },
1894	[NDTPA_DELAY_PROBE_TIME]	= { .type = NLA_U64 },
1895	[NDTPA_RETRANS_TIME]		= { .type = NLA_U64 },
1896	[NDTPA_ANYCAST_DELAY]		= { .type = NLA_U64 },
1897	[NDTPA_PROXY_DELAY]		= { .type = NLA_U64 },
1898	[NDTPA_LOCKTIME]		= { .type = NLA_U64 },
1899};
1900
1901static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 
1902{
1903	struct net *net = sock_net(skb->sk);
1904	struct neigh_table *tbl;
1905	struct ndtmsg *ndtmsg;
1906	struct nlattr *tb[NDTA_MAX+1];
1907	int err;
 
1908
1909	err = nlmsg_parse(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
1910			  nl_neightbl_policy);
1911	if (err < 0)
1912		goto errout;
1913
1914	if (tb[NDTA_NAME] == NULL) {
1915		err = -EINVAL;
1916		goto errout;
1917	}
1918
1919	ndtmsg = nlmsg_data(nlh);
1920	read_lock(&neigh_tbl_lock);
1921	for (tbl = neigh_tables; tbl; tbl = tbl->next) {
 
 
 
1922		if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
1923			continue;
1924
1925		if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0)
1926			break;
 
1927	}
1928
1929	if (tbl == NULL) {
1930		err = -ENOENT;
1931		goto errout_locked;
1932	}
1933
1934	/*
1935	 * We acquire tbl->lock to be nice to the periodic timers and
1936	 * make sure they always see a consistent set of values.
1937	 */
1938	write_lock_bh(&tbl->lock);
1939
1940	if (tb[NDTA_PARMS]) {
1941		struct nlattr *tbp[NDTPA_MAX+1];
1942		struct neigh_parms *p;
1943		int i, ifindex = 0;
1944
1945		err = nla_parse_nested(tbp, NDTPA_MAX, tb[NDTA_PARMS],
1946				       nl_ntbl_parm_policy);
 
1947		if (err < 0)
1948			goto errout_tbl_lock;
1949
1950		if (tbp[NDTPA_IFINDEX])
1951			ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
1952
1953		p = lookup_neigh_parms(tbl, net, ifindex);
1954		if (p == NULL) {
1955			err = -ENOENT;
1956			goto errout_tbl_lock;
1957		}
1958
1959		for (i = 1; i <= NDTPA_MAX; i++) {
1960			if (tbp[i] == NULL)
1961				continue;
1962
1963			switch (i) {
1964			case NDTPA_QUEUE_LEN:
1965				p->queue_len = nla_get_u32(tbp[i]);
 
 
 
 
 
 
1966				break;
1967			case NDTPA_PROXY_QLEN:
1968				p->proxy_qlen = nla_get_u32(tbp[i]);
 
1969				break;
1970			case NDTPA_APP_PROBES:
1971				p->app_probes = nla_get_u32(tbp[i]);
 
1972				break;
1973			case NDTPA_UCAST_PROBES:
1974				p->ucast_probes = nla_get_u32(tbp[i]);
 
1975				break;
1976			case NDTPA_MCAST_PROBES:
1977				p->mcast_probes = nla_get_u32(tbp[i]);
 
 
 
 
 
1978				break;
1979			case NDTPA_BASE_REACHABLE_TIME:
1980				p->base_reachable_time = nla_get_msecs(tbp[i]);
 
 
 
 
 
 
 
1981				break;
1982			case NDTPA_GC_STALETIME:
1983				p->gc_staletime = nla_get_msecs(tbp[i]);
 
1984				break;
1985			case NDTPA_DELAY_PROBE_TIME:
1986				p->delay_probe_time = nla_get_msecs(tbp[i]);
 
 
1987				break;
1988			case NDTPA_RETRANS_TIME:
1989				p->retrans_time = nla_get_msecs(tbp[i]);
 
1990				break;
1991			case NDTPA_ANYCAST_DELAY:
1992				p->anycast_delay = nla_get_msecs(tbp[i]);
 
1993				break;
1994			case NDTPA_PROXY_DELAY:
1995				p->proxy_delay = nla_get_msecs(tbp[i]);
 
1996				break;
1997			case NDTPA_LOCKTIME:
1998				p->locktime = nla_get_msecs(tbp[i]);
 
1999				break;
2000			}
2001		}
2002	}
2003
 
 
 
 
 
 
2004	if (tb[NDTA_THRESH1])
2005		tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2006
2007	if (tb[NDTA_THRESH2])
2008		tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2009
2010	if (tb[NDTA_THRESH3])
2011		tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2012
2013	if (tb[NDTA_GC_INTERVAL])
2014		tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2015
2016	err = 0;
2017
2018errout_tbl_lock:
2019	write_unlock_bh(&tbl->lock);
2020errout_locked:
2021	read_unlock(&neigh_tbl_lock);
2022errout:
2023	return err;
2024}
2025
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2026static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2027{
 
2028	struct net *net = sock_net(skb->sk);
2029	int family, tidx, nidx = 0;
2030	int tbl_skip = cb->args[0];
2031	int neigh_skip = cb->args[1];
2032	struct neigh_table *tbl;
2033
2034	family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
 
2035
2036	read_lock(&neigh_tbl_lock);
2037	for (tbl = neigh_tables, tidx = 0; tbl; tbl = tbl->next, tidx++) {
 
 
 
 
 
2038		struct neigh_parms *p;
2039
 
 
 
 
2040		if (tidx < tbl_skip || (family && tbl->family != family))
2041			continue;
2042
2043		if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).pid,
2044				       cb->nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2045				       NLM_F_MULTI) <= 0)
2046			break;
2047
2048		for (nidx = 0, p = tbl->parms.next; p; p = p->next) {
 
 
2049			if (!net_eq(neigh_parms_net(p), net))
2050				continue;
2051
2052			if (nidx < neigh_skip)
2053				goto next;
2054
2055			if (neightbl_fill_param_info(skb, tbl, p,
2056						     NETLINK_CB(cb->skb).pid,
2057						     cb->nlh->nlmsg_seq,
2058						     RTM_NEWNEIGHTBL,
2059						     NLM_F_MULTI) <= 0)
2060				goto out;
2061		next:
2062			nidx++;
2063		}
2064
2065		neigh_skip = 0;
2066	}
2067out:
2068	read_unlock(&neigh_tbl_lock);
2069	cb->args[0] = tidx;
2070	cb->args[1] = nidx;
2071
2072	return skb->len;
2073}
2074
2075static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2076			   u32 pid, u32 seq, int type, unsigned int flags)
2077{
2078	unsigned long now = jiffies;
2079	struct nda_cacheinfo ci;
2080	struct nlmsghdr *nlh;
2081	struct ndmsg *ndm;
2082
2083	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2084	if (nlh == NULL)
2085		return -EMSGSIZE;
2086
2087	ndm = nlmsg_data(nlh);
2088	ndm->ndm_family	 = neigh->ops->family;
2089	ndm->ndm_pad1    = 0;
2090	ndm->ndm_pad2    = 0;
2091	ndm->ndm_flags	 = neigh->flags;
2092	ndm->ndm_type	 = neigh->type;
2093	ndm->ndm_ifindex = neigh->dev->ifindex;
2094
2095	NLA_PUT(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key);
 
2096
2097	read_lock_bh(&neigh->lock);
2098	ndm->ndm_state	 = neigh->nud_state;
2099	if (neigh->nud_state & NUD_VALID) {
2100		char haddr[MAX_ADDR_LEN];
2101
2102		neigh_ha_snapshot(haddr, neigh, neigh->dev);
2103		if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2104			read_unlock_bh(&neigh->lock);
2105			goto nla_put_failure;
2106		}
2107	}
2108
2109	ci.ndm_used	 = jiffies_to_clock_t(now - neigh->used);
2110	ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2111	ci.ndm_updated	 = jiffies_to_clock_t(now - neigh->updated);
2112	ci.ndm_refcnt	 = atomic_read(&neigh->refcnt) - 1;
2113	read_unlock_bh(&neigh->lock);
2114
2115	NLA_PUT_U32(skb, NDA_PROBES, atomic_read(&neigh->probes));
2116	NLA_PUT(skb, NDA_CACHEINFO, sizeof(ci), &ci);
 
 
 
 
2117
2118	return nlmsg_end(skb, nlh);
 
2119
2120nla_put_failure:
2121	nlmsg_cancel(skb, nlh);
2122	return -EMSGSIZE;
2123}
2124
2125static void neigh_update_notify(struct neighbour *neigh)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2126{
2127	call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2128	__neigh_notify(neigh, RTM_NEWNEIGH, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2129}
2130
 
 
 
 
 
 
 
 
 
 
 
 
 
2131static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2132			    struct netlink_callback *cb)
 
2133{
2134	struct net *net = sock_net(skb->sk);
2135	struct neighbour *n;
2136	int rc, h, s_h = cb->args[1];
2137	int idx, s_idx = idx = cb->args[2];
2138	struct neigh_hash_table *nht;
 
 
 
 
2139
2140	rcu_read_lock_bh();
2141	nht = rcu_dereference_bh(tbl->nht);
2142
2143	for (h = 0; h < (1 << nht->hash_shift); h++) {
2144		if (h < s_h)
2145			continue;
2146		if (h > s_h)
2147			s_idx = 0;
2148		for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2149		     n != NULL;
2150		     n = rcu_dereference_bh(n->next)) {
2151			if (!net_eq(dev_net(n->dev), net))
2152				continue;
2153			if (idx < s_idx)
 
2154				goto next;
2155			if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).pid,
2156					    cb->nlh->nlmsg_seq,
2157					    RTM_NEWNEIGH,
2158					    NLM_F_MULTI) <= 0) {
2159				rc = -1;
2160				goto out;
2161			}
2162next:
2163			idx++;
2164		}
2165	}
2166	rc = skb->len;
2167out:
2168	rcu_read_unlock_bh();
2169	cb->args[1] = h;
2170	cb->args[2] = idx;
2171	return rc;
2172}
2173
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2174static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2175{
 
 
2176	struct neigh_table *tbl;
2177	int t, family, s_t;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2178
2179	read_lock(&neigh_tbl_lock);
2180	family = ((struct rtgenmsg *) nlmsg_data(cb->nlh))->rtgen_family;
2181	s_t = cb->args[0];
2182
2183	for (tbl = neigh_tables, t = 0; tbl; tbl = tbl->next, t++) {
 
 
 
 
2184		if (t < s_t || (family && tbl->family != family))
2185			continue;
2186		if (t > s_t)
2187			memset(&cb->args[1], 0, sizeof(cb->args) -
2188						sizeof(cb->args[0]));
2189		if (neigh_dump_table(tbl, skb, cb) < 0)
 
 
 
 
2190			break;
2191	}
2192	read_unlock(&neigh_tbl_lock);
2193
2194	cb->args[0] = t;
2195	return skb->len;
2196}
2197
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2198void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2199{
2200	int chain;
2201	struct neigh_hash_table *nht;
2202
2203	rcu_read_lock_bh();
2204	nht = rcu_dereference_bh(tbl->nht);
2205
2206	read_lock(&tbl->lock); /* avoid resizes */
2207	for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2208		struct neighbour *n;
2209
2210		for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2211		     n != NULL;
2212		     n = rcu_dereference_bh(n->next))
2213			cb(n, cookie);
2214	}
2215	read_unlock(&tbl->lock);
2216	rcu_read_unlock_bh();
2217}
2218EXPORT_SYMBOL(neigh_for_each);
2219
2220/* The tbl->lock must be held as a writer and BH disabled. */
2221void __neigh_for_each_release(struct neigh_table *tbl,
2222			      int (*cb)(struct neighbour *))
2223{
2224	int chain;
2225	struct neigh_hash_table *nht;
2226
2227	nht = rcu_dereference_protected(tbl->nht,
2228					lockdep_is_held(&tbl->lock));
2229	for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2230		struct neighbour *n;
2231		struct neighbour __rcu **np;
2232
2233		np = &nht->hash_buckets[chain];
2234		while ((n = rcu_dereference_protected(*np,
2235					lockdep_is_held(&tbl->lock))) != NULL) {
2236			int release;
2237
2238			write_lock(&n->lock);
2239			release = cb(n);
2240			if (release) {
2241				rcu_assign_pointer(*np,
2242					rcu_dereference_protected(n->next,
2243						lockdep_is_held(&tbl->lock)));
2244				n->dead = 1;
2245			} else
2246				np = &n->next;
2247			write_unlock(&n->lock);
2248			if (release)
2249				neigh_cleanup_and_release(n);
2250		}
2251	}
2252}
2253EXPORT_SYMBOL(__neigh_for_each_release);
2254
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2255#ifdef CONFIG_PROC_FS
2256
2257static struct neighbour *neigh_get_first(struct seq_file *seq)
2258{
2259	struct neigh_seq_state *state = seq->private;
2260	struct net *net = seq_file_net(seq);
2261	struct neigh_hash_table *nht = state->nht;
2262	struct neighbour *n = NULL;
2263	int bucket = state->bucket;
2264
2265	state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
2266	for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
2267		n = rcu_dereference_bh(nht->hash_buckets[bucket]);
2268
2269		while (n) {
2270			if (!net_eq(dev_net(n->dev), net))
2271				goto next;
2272			if (state->neigh_sub_iter) {
2273				loff_t fakep = 0;
2274				void *v;
2275
2276				v = state->neigh_sub_iter(state, n, &fakep);
2277				if (!v)
2278					goto next;
2279			}
2280			if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2281				break;
2282			if (n->nud_state & ~NUD_NOARP)
2283				break;
2284next:
2285			n = rcu_dereference_bh(n->next);
2286		}
2287
2288		if (n)
2289			break;
2290	}
2291	state->bucket = bucket;
2292
2293	return n;
2294}
2295
2296static struct neighbour *neigh_get_next(struct seq_file *seq,
2297					struct neighbour *n,
2298					loff_t *pos)
2299{
2300	struct neigh_seq_state *state = seq->private;
2301	struct net *net = seq_file_net(seq);
2302	struct neigh_hash_table *nht = state->nht;
2303
2304	if (state->neigh_sub_iter) {
2305		void *v = state->neigh_sub_iter(state, n, pos);
2306		if (v)
2307			return n;
2308	}
2309	n = rcu_dereference_bh(n->next);
2310
2311	while (1) {
2312		while (n) {
2313			if (!net_eq(dev_net(n->dev), net))
2314				goto next;
2315			if (state->neigh_sub_iter) {
2316				void *v = state->neigh_sub_iter(state, n, pos);
2317				if (v)
2318					return n;
2319				goto next;
2320			}
2321			if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
2322				break;
2323
2324			if (n->nud_state & ~NUD_NOARP)
2325				break;
2326next:
2327			n = rcu_dereference_bh(n->next);
2328		}
2329
2330		if (n)
2331			break;
2332
2333		if (++state->bucket >= (1 << nht->hash_shift))
2334			break;
2335
2336		n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
2337	}
2338
2339	if (n && pos)
2340		--(*pos);
2341	return n;
2342}
2343
2344static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
2345{
2346	struct neighbour *n = neigh_get_first(seq);
2347
2348	if (n) {
2349		--(*pos);
2350		while (*pos) {
2351			n = neigh_get_next(seq, n, pos);
2352			if (!n)
2353				break;
2354		}
2355	}
2356	return *pos ? NULL : n;
2357}
2358
2359static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
2360{
2361	struct neigh_seq_state *state = seq->private;
2362	struct net *net = seq_file_net(seq);
2363	struct neigh_table *tbl = state->tbl;
2364	struct pneigh_entry *pn = NULL;
2365	int bucket = state->bucket;
2366
2367	state->flags |= NEIGH_SEQ_IS_PNEIGH;
2368	for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
2369		pn = tbl->phash_buckets[bucket];
2370		while (pn && !net_eq(pneigh_net(pn), net))
2371			pn = pn->next;
2372		if (pn)
2373			break;
2374	}
2375	state->bucket = bucket;
2376
2377	return pn;
2378}
2379
2380static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
2381					    struct pneigh_entry *pn,
2382					    loff_t *pos)
2383{
2384	struct neigh_seq_state *state = seq->private;
2385	struct net *net = seq_file_net(seq);
2386	struct neigh_table *tbl = state->tbl;
2387
2388	pn = pn->next;
 
 
 
2389	while (!pn) {
2390		if (++state->bucket > PNEIGH_HASHMASK)
2391			break;
2392		pn = tbl->phash_buckets[state->bucket];
2393		while (pn && !net_eq(pneigh_net(pn), net))
2394			pn = pn->next;
2395		if (pn)
2396			break;
2397	}
2398
2399	if (pn && pos)
2400		--(*pos);
2401
2402	return pn;
2403}
2404
2405static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
2406{
2407	struct pneigh_entry *pn = pneigh_get_first(seq);
2408
2409	if (pn) {
2410		--(*pos);
2411		while (*pos) {
2412			pn = pneigh_get_next(seq, pn, pos);
2413			if (!pn)
2414				break;
2415		}
2416	}
2417	return *pos ? NULL : pn;
2418}
2419
2420static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
2421{
2422	struct neigh_seq_state *state = seq->private;
2423	void *rc;
2424	loff_t idxpos = *pos;
2425
2426	rc = neigh_get_idx(seq, &idxpos);
2427	if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2428		rc = pneigh_get_idx(seq, &idxpos);
2429
2430	return rc;
2431}
2432
2433void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
 
2434	__acquires(rcu_bh)
2435{
2436	struct neigh_seq_state *state = seq->private;
2437
2438	state->tbl = tbl;
2439	state->bucket = 0;
2440	state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
2441
2442	rcu_read_lock_bh();
2443	state->nht = rcu_dereference_bh(tbl->nht);
 
2444
2445	return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
2446}
2447EXPORT_SYMBOL(neigh_seq_start);
2448
2449void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2450{
2451	struct neigh_seq_state *state;
2452	void *rc;
2453
2454	if (v == SEQ_START_TOKEN) {
2455		rc = neigh_get_first(seq);
2456		goto out;
2457	}
2458
2459	state = seq->private;
2460	if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
2461		rc = neigh_get_next(seq, v, NULL);
2462		if (rc)
2463			goto out;
2464		if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
2465			rc = pneigh_get_first(seq);
2466	} else {
2467		BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
2468		rc = pneigh_get_next(seq, v, NULL);
2469	}
2470out:
2471	++(*pos);
2472	return rc;
2473}
2474EXPORT_SYMBOL(neigh_seq_next);
2475
2476void neigh_seq_stop(struct seq_file *seq, void *v)
 
2477	__releases(rcu_bh)
2478{
 
 
 
 
2479	rcu_read_unlock_bh();
2480}
2481EXPORT_SYMBOL(neigh_seq_stop);
2482
2483/* statistics via seq_file */
2484
2485static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
2486{
2487	struct neigh_table *tbl = seq->private;
2488	int cpu;
2489
2490	if (*pos == 0)
2491		return SEQ_START_TOKEN;
2492
2493	for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
2494		if (!cpu_possible(cpu))
2495			continue;
2496		*pos = cpu+1;
2497		return per_cpu_ptr(tbl->stats, cpu);
2498	}
2499	return NULL;
2500}
2501
2502static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2503{
2504	struct neigh_table *tbl = seq->private;
2505	int cpu;
2506
2507	for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
2508		if (!cpu_possible(cpu))
2509			continue;
2510		*pos = cpu+1;
2511		return per_cpu_ptr(tbl->stats, cpu);
2512	}
 
2513	return NULL;
2514}
2515
2516static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
2517{
2518
2519}
2520
2521static int neigh_stat_seq_show(struct seq_file *seq, void *v)
2522{
2523	struct neigh_table *tbl = seq->private;
2524	struct neigh_statistics *st = v;
2525
2526	if (v == SEQ_START_TOKEN) {
2527		seq_printf(seq, "entries  allocs destroys hash_grows  lookups hits  res_failed  rcv_probes_mcast rcv_probes_ucast  periodic_gc_runs forced_gc_runs unresolved_discards\n");
2528		return 0;
2529	}
2530
2531	seq_printf(seq, "%08x  %08lx %08lx %08lx  %08lx %08lx  %08lx  "
2532			"%08lx %08lx  %08lx %08lx %08lx\n",
2533		   atomic_read(&tbl->entries),
2534
2535		   st->allocs,
2536		   st->destroys,
2537		   st->hash_grows,
2538
2539		   st->lookups,
2540		   st->hits,
2541
2542		   st->res_failed,
2543
2544		   st->rcv_probes_mcast,
2545		   st->rcv_probes_ucast,
2546
2547		   st->periodic_gc_runs,
2548		   st->forced_gc_runs,
2549		   st->unres_discards
 
2550		   );
2551
2552	return 0;
2553}
2554
2555static const struct seq_operations neigh_stat_seq_ops = {
2556	.start	= neigh_stat_seq_start,
2557	.next	= neigh_stat_seq_next,
2558	.stop	= neigh_stat_seq_stop,
2559	.show	= neigh_stat_seq_show,
2560};
2561
2562static int neigh_stat_seq_open(struct inode *inode, struct file *file)
2563{
2564	int ret = seq_open(file, &neigh_stat_seq_ops);
2565
2566	if (!ret) {
2567		struct seq_file *sf = file->private_data;
2568		sf->private = PDE(inode)->data;
2569	}
2570	return ret;
2571};
2572
2573static const struct file_operations neigh_stat_seq_fops = {
2574	.owner	 = THIS_MODULE,
2575	.open 	 = neigh_stat_seq_open,
2576	.read	 = seq_read,
2577	.llseek	 = seq_lseek,
2578	.release = seq_release,
2579};
2580
2581#endif /* CONFIG_PROC_FS */
2582
2583static inline size_t neigh_nlmsg_size(void)
2584{
2585	return NLMSG_ALIGN(sizeof(struct ndmsg))
2586	       + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2587	       + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2588	       + nla_total_size(sizeof(struct nda_cacheinfo))
2589	       + nla_total_size(4); /* NDA_PROBES */
2590}
2591
2592static void __neigh_notify(struct neighbour *n, int type, int flags)
2593{
2594	struct net *net = dev_net(n->dev);
2595	struct sk_buff *skb;
2596	int err = -ENOBUFS;
2597
2598	skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
2599	if (skb == NULL)
2600		goto errout;
2601
2602	err = neigh_fill_info(skb, n, 0, 0, type, flags);
2603	if (err < 0) {
2604		/* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
2605		WARN_ON(err == -EMSGSIZE);
2606		kfree_skb(skb);
2607		goto errout;
2608	}
2609	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2610	return;
2611errout:
2612	if (err < 0)
2613		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2614}
2615
2616#ifdef CONFIG_ARPD
2617void neigh_app_ns(struct neighbour *n)
2618{
2619	__neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST);
2620}
2621EXPORT_SYMBOL(neigh_app_ns);
2622#endif /* CONFIG_ARPD */
2623
2624#ifdef CONFIG_SYSCTL
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2625
2626#define NEIGH_VARS_MAX 19
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2627
2628static struct neigh_sysctl_table {
2629	struct ctl_table_header *sysctl_header;
2630	struct ctl_table neigh_vars[NEIGH_VARS_MAX];
2631	char *dev_name;
2632} neigh_sysctl_template __read_mostly = {
2633	.neigh_vars = {
2634		{
2635			.procname	= "mcast_solicit",
2636			.maxlen		= sizeof(int),
2637			.mode		= 0644,
2638			.proc_handler	= proc_dointvec,
2639		},
2640		{
2641			.procname	= "ucast_solicit",
2642			.maxlen		= sizeof(int),
2643			.mode		= 0644,
2644			.proc_handler	= proc_dointvec,
2645		},
2646		{
2647			.procname	= "app_solicit",
2648			.maxlen		= sizeof(int),
2649			.mode		= 0644,
2650			.proc_handler	= proc_dointvec,
2651		},
2652		{
2653			.procname	= "retrans_time",
2654			.maxlen		= sizeof(int),
2655			.mode		= 0644,
2656			.proc_handler	= proc_dointvec_userhz_jiffies,
2657		},
2658		{
2659			.procname	= "base_reachable_time",
2660			.maxlen		= sizeof(int),
2661			.mode		= 0644,
2662			.proc_handler	= proc_dointvec_jiffies,
2663		},
2664		{
2665			.procname	= "delay_first_probe_time",
2666			.maxlen		= sizeof(int),
2667			.mode		= 0644,
2668			.proc_handler	= proc_dointvec_jiffies,
2669		},
2670		{
2671			.procname	= "gc_stale_time",
2672			.maxlen		= sizeof(int),
2673			.mode		= 0644,
2674			.proc_handler	= proc_dointvec_jiffies,
2675		},
2676		{
2677			.procname	= "unres_qlen",
2678			.maxlen		= sizeof(int),
2679			.mode		= 0644,
2680			.proc_handler	= proc_dointvec,
2681		},
2682		{
2683			.procname	= "proxy_qlen",
2684			.maxlen		= sizeof(int),
2685			.mode		= 0644,
2686			.proc_handler	= proc_dointvec,
2687		},
2688		{
2689			.procname	= "anycast_delay",
2690			.maxlen		= sizeof(int),
2691			.mode		= 0644,
2692			.proc_handler	= proc_dointvec_userhz_jiffies,
2693		},
2694		{
2695			.procname	= "proxy_delay",
2696			.maxlen		= sizeof(int),
2697			.mode		= 0644,
2698			.proc_handler	= proc_dointvec_userhz_jiffies,
2699		},
2700		{
2701			.procname	= "locktime",
2702			.maxlen		= sizeof(int),
2703			.mode		= 0644,
2704			.proc_handler	= proc_dointvec_userhz_jiffies,
2705		},
2706		{
2707			.procname	= "retrans_time_ms",
2708			.maxlen		= sizeof(int),
2709			.mode		= 0644,
2710			.proc_handler	= proc_dointvec_ms_jiffies,
2711		},
2712		{
2713			.procname	= "base_reachable_time_ms",
2714			.maxlen		= sizeof(int),
2715			.mode		= 0644,
2716			.proc_handler	= proc_dointvec_ms_jiffies,
2717		},
2718		{
2719			.procname	= "gc_interval",
2720			.maxlen		= sizeof(int),
2721			.mode		= 0644,
2722			.proc_handler	= proc_dointvec_jiffies,
2723		},
2724		{
2725			.procname	= "gc_thresh1",
2726			.maxlen		= sizeof(int),
2727			.mode		= 0644,
2728			.proc_handler	= proc_dointvec,
 
 
2729		},
2730		{
2731			.procname	= "gc_thresh2",
2732			.maxlen		= sizeof(int),
2733			.mode		= 0644,
2734			.proc_handler	= proc_dointvec,
 
 
2735		},
2736		{
2737			.procname	= "gc_thresh3",
2738			.maxlen		= sizeof(int),
2739			.mode		= 0644,
2740			.proc_handler	= proc_dointvec,
 
 
2741		},
2742		{},
2743	},
2744};
2745
2746int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
2747			  char *p_name, proc_handler *handler)
2748{
 
2749	struct neigh_sysctl_table *t;
2750	const char *dev_name_source = NULL;
2751
2752#define NEIGH_CTL_PATH_ROOT	0
2753#define NEIGH_CTL_PATH_PROTO	1
2754#define NEIGH_CTL_PATH_NEIGH	2
2755#define NEIGH_CTL_PATH_DEV	3
2756
2757	struct ctl_path neigh_path[] = {
2758		{ .procname = "net",	 },
2759		{ .procname = "proto",	 },
2760		{ .procname = "neigh",	 },
2761		{ .procname = "default", },
2762		{ },
2763	};
2764
2765	t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
2766	if (!t)
2767		goto err;
2768
2769	t->neigh_vars[0].data  = &p->mcast_probes;
2770	t->neigh_vars[1].data  = &p->ucast_probes;
2771	t->neigh_vars[2].data  = &p->app_probes;
2772	t->neigh_vars[3].data  = &p->retrans_time;
2773	t->neigh_vars[4].data  = &p->base_reachable_time;
2774	t->neigh_vars[5].data  = &p->delay_probe_time;
2775	t->neigh_vars[6].data  = &p->gc_staletime;
2776	t->neigh_vars[7].data  = &p->queue_len;
2777	t->neigh_vars[8].data  = &p->proxy_qlen;
2778	t->neigh_vars[9].data  = &p->anycast_delay;
2779	t->neigh_vars[10].data = &p->proxy_delay;
2780	t->neigh_vars[11].data = &p->locktime;
2781	t->neigh_vars[12].data  = &p->retrans_time;
2782	t->neigh_vars[13].data  = &p->base_reachable_time;
2783
2784	if (dev) {
2785		dev_name_source = dev->name;
2786		/* Terminate the table early */
2787		memset(&t->neigh_vars[14], 0, sizeof(t->neigh_vars[14]));
 
2788	} else {
2789		dev_name_source = neigh_path[NEIGH_CTL_PATH_DEV].procname;
2790		t->neigh_vars[14].data = (int *)(p + 1);
2791		t->neigh_vars[15].data = (int *)(p + 1) + 1;
2792		t->neigh_vars[16].data = (int *)(p + 1) + 2;
2793		t->neigh_vars[17].data = (int *)(p + 1) + 3;
 
2794	}
2795
2796
2797	if (handler) {
2798		/* RetransTime */
2799		t->neigh_vars[3].proc_handler = handler;
2800		t->neigh_vars[3].extra1 = dev;
2801		/* ReachableTime */
2802		t->neigh_vars[4].proc_handler = handler;
2803		t->neigh_vars[4].extra1 = dev;
2804		/* RetransTime (in milliseconds)*/
2805		t->neigh_vars[12].proc_handler = handler;
2806		t->neigh_vars[12].extra1 = dev;
 
 
 
 
 
 
 
 
 
 
 
2807		/* ReachableTime (in milliseconds) */
2808		t->neigh_vars[13].proc_handler = handler;
2809		t->neigh_vars[13].extra1 = dev;
2810	}
2811
2812	t->dev_name = kstrdup(dev_name_source, GFP_KERNEL);
2813	if (!t->dev_name)
2814		goto free;
2815
2816	neigh_path[NEIGH_CTL_PATH_DEV].procname = t->dev_name;
2817	neigh_path[NEIGH_CTL_PATH_PROTO].procname = p_name;
 
 
 
 
 
 
 
 
2818
 
 
2819	t->sysctl_header =
2820		register_net_sysctl_table(neigh_parms_net(p), neigh_path, t->neigh_vars);
2821	if (!t->sysctl_header)
2822		goto free_procname;
2823
2824	p->sysctl_table = t;
2825	return 0;
2826
2827free_procname:
2828	kfree(t->dev_name);
2829free:
2830	kfree(t);
2831err:
2832	return -ENOBUFS;
2833}
2834EXPORT_SYMBOL(neigh_sysctl_register);
2835
2836void neigh_sysctl_unregister(struct neigh_parms *p)
2837{
2838	if (p->sysctl_table) {
2839		struct neigh_sysctl_table *t = p->sysctl_table;
2840		p->sysctl_table = NULL;
2841		unregister_sysctl_table(t->sysctl_header);
2842		kfree(t->dev_name);
2843		kfree(t);
2844	}
2845}
2846EXPORT_SYMBOL(neigh_sysctl_unregister);
2847
2848#endif	/* CONFIG_SYSCTL */
2849
2850static int __init neigh_init(void)
2851{
2852	rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, NULL);
2853	rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, NULL);
2854	rtnl_register(PF_UNSPEC, RTM_GETNEIGH, NULL, neigh_dump_info, NULL);
2855
2856	rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
2857		      NULL);
2858	rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, NULL);
2859
2860	return 0;
2861}
2862
2863subsys_initcall(neigh_init);
2864
v5.14.15
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	Generic address resolution entity
   4 *
   5 *	Authors:
   6 *	Pedro Roque		<roque@di.fc.ul.pt>
   7 *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
   8 *
 
 
 
 
 
   9 *	Fixes:
  10 *	Vitaly E. Lavrov	releasing NULL neighbor in neigh_add.
  11 *	Harald Welte		Add neighbour cache statistics like rtstat
  12 */
  13
  14#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  15
  16#include <linux/slab.h>
  17#include <linux/kmemleak.h>
  18#include <linux/types.h>
  19#include <linux/kernel.h>
  20#include <linux/module.h>
  21#include <linux/socket.h>
  22#include <linux/netdevice.h>
  23#include <linux/proc_fs.h>
  24#ifdef CONFIG_SYSCTL
  25#include <linux/sysctl.h>
  26#endif
  27#include <linux/times.h>
  28#include <net/net_namespace.h>
  29#include <net/neighbour.h>
  30#include <net/arp.h>
  31#include <net/dst.h>
  32#include <net/sock.h>
  33#include <net/netevent.h>
  34#include <net/netlink.h>
  35#include <linux/rtnetlink.h>
  36#include <linux/random.h>
  37#include <linux/string.h>
  38#include <linux/log2.h>
  39#include <linux/inetdevice.h>
  40#include <net/addrconf.h>
  41
  42#include <trace/events/neigh.h>
  43
  44#define NEIGH_DEBUG 1
  45#define neigh_dbg(level, fmt, ...)		\
  46do {						\
  47	if (level <= NEIGH_DEBUG)		\
  48		pr_debug(fmt, ##__VA_ARGS__);	\
  49} while (0)
 
 
 
 
 
 
 
  50
  51#define PNEIGH_HASHMASK		0xF
  52
  53static void neigh_timer_handler(struct timer_list *t);
  54static void __neigh_notify(struct neighbour *n, int type, int flags,
  55			   u32 pid);
  56static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid);
  57static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
  58				    struct net_device *dev);
  59
 
  60#ifdef CONFIG_PROC_FS
  61static const struct seq_operations neigh_stat_seq_ops;
  62#endif
  63
  64/*
  65   Neighbour hash table buckets are protected with rwlock tbl->lock.
  66
  67   - All the scans/updates to hash buckets MUST be made under this lock.
  68   - NOTHING clever should be made under this lock: no callbacks
  69     to protocol backends, no attempts to send something to network.
  70     It will result in deadlocks, if backend/driver wants to use neighbour
  71     cache.
  72   - If the entry requires some non-trivial actions, increase
  73     its reference count and release table lock.
  74
  75   Neighbour entries are protected:
  76   - with reference count.
  77   - with rwlock neigh->lock
  78
  79   Reference count prevents destruction.
  80
  81   neigh->lock mainly serializes ll address data and its validity state.
  82   However, the same lock is used to protect another entry fields:
  83    - timer
  84    - resolution queue
  85
  86   Again, nothing clever shall be made under neigh->lock,
  87   the most complicated procedure, which we allow is dev->hard_header.
  88   It is supposed, that dev->hard_header is simplistic and does
  89   not make callbacks to neighbour tables.
 
 
 
  90 */
  91
 
 
  92static int neigh_blackhole(struct neighbour *neigh, struct sk_buff *skb)
  93{
  94	kfree_skb(skb);
  95	return -ENETDOWN;
  96}
  97
  98static void neigh_cleanup_and_release(struct neighbour *neigh)
  99{
 100	trace_neigh_cleanup_and_release(neigh, 0);
 101	__neigh_notify(neigh, RTM_DELNEIGH, 0, 0);
 102	call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
 
 103	neigh_release(neigh);
 104}
 105
 106/*
 107 * It is random distribution in the interval (1/2)*base...(3/2)*base.
 108 * It corresponds to default IPv6 settings and is not overridable,
 109 * because it is really reasonable choice.
 110 */
 111
 112unsigned long neigh_rand_reach_time(unsigned long base)
 113{
 114	return base ? (prandom_u32() % base) + (base >> 1) : 0;
 115}
 116EXPORT_SYMBOL(neigh_rand_reach_time);
 117
 118static void neigh_mark_dead(struct neighbour *n)
 119{
 120	n->dead = 1;
 121	if (!list_empty(&n->gc_list)) {
 122		list_del_init(&n->gc_list);
 123		atomic_dec(&n->tbl->gc_entries);
 124	}
 125}
 126
 127static void neigh_update_gc_list(struct neighbour *n)
 128{
 129	bool on_gc_list, exempt_from_gc;
 130
 131	write_lock_bh(&n->tbl->lock);
 132	write_lock(&n->lock);
 133
 134	if (n->dead)
 135		goto out;
 136
 137	/* remove from the gc list if new state is permanent or if neighbor
 138	 * is externally learned; otherwise entry should be on the gc list
 139	 */
 140	exempt_from_gc = n->nud_state & NUD_PERMANENT ||
 141			 n->flags & NTF_EXT_LEARNED;
 142	on_gc_list = !list_empty(&n->gc_list);
 143
 144	if (exempt_from_gc && on_gc_list) {
 145		list_del_init(&n->gc_list);
 146		atomic_dec(&n->tbl->gc_entries);
 147	} else if (!exempt_from_gc && !on_gc_list) {
 148		/* add entries to the tail; cleaning removes from the front */
 149		list_add_tail(&n->gc_list, &n->tbl->gc_list);
 150		atomic_inc(&n->tbl->gc_entries);
 151	}
 152
 153out:
 154	write_unlock(&n->lock);
 155	write_unlock_bh(&n->tbl->lock);
 156}
 157
 158static bool neigh_update_ext_learned(struct neighbour *neigh, u32 flags,
 159				     int *notify)
 160{
 161	bool rc = false;
 162	u8 ndm_flags;
 163
 164	if (!(flags & NEIGH_UPDATE_F_ADMIN))
 165		return rc;
 166
 167	ndm_flags = (flags & NEIGH_UPDATE_F_EXT_LEARNED) ? NTF_EXT_LEARNED : 0;
 168	if ((neigh->flags ^ ndm_flags) & NTF_EXT_LEARNED) {
 169		if (ndm_flags & NTF_EXT_LEARNED)
 170			neigh->flags |= NTF_EXT_LEARNED;
 171		else
 172			neigh->flags &= ~NTF_EXT_LEARNED;
 173		rc = true;
 174		*notify = 1;
 175	}
 176
 177	return rc;
 178}
 179
 180static bool neigh_del(struct neighbour *n, struct neighbour __rcu **np,
 181		      struct neigh_table *tbl)
 182{
 183	bool retval = false;
 184
 185	write_lock(&n->lock);
 186	if (refcount_read(&n->refcnt) == 1) {
 187		struct neighbour *neigh;
 188
 189		neigh = rcu_dereference_protected(n->next,
 190						  lockdep_is_held(&tbl->lock));
 191		rcu_assign_pointer(*np, neigh);
 192		neigh_mark_dead(n);
 193		retval = true;
 194	}
 195	write_unlock(&n->lock);
 196	if (retval)
 197		neigh_cleanup_and_release(n);
 198	return retval;
 199}
 200
 201bool neigh_remove_one(struct neighbour *ndel, struct neigh_table *tbl)
 202{
 203	struct neigh_hash_table *nht;
 204	void *pkey = ndel->primary_key;
 205	u32 hash_val;
 206	struct neighbour *n;
 207	struct neighbour __rcu **np;
 208
 209	nht = rcu_dereference_protected(tbl->nht,
 210					lockdep_is_held(&tbl->lock));
 211	hash_val = tbl->hash(pkey, ndel->dev, nht->hash_rnd);
 212	hash_val = hash_val >> (32 - nht->hash_shift);
 213
 214	np = &nht->hash_buckets[hash_val];
 215	while ((n = rcu_dereference_protected(*np,
 216					      lockdep_is_held(&tbl->lock)))) {
 217		if (n == ndel)
 218			return neigh_del(n, np, tbl);
 219		np = &n->next;
 220	}
 221	return false;
 222}
 223
 224static int neigh_forced_gc(struct neigh_table *tbl)
 225{
 226	int max_clean = atomic_read(&tbl->gc_entries) - tbl->gc_thresh2;
 227	unsigned long tref = jiffies - 5 * HZ;
 228	struct neighbour *n, *tmp;
 229	int shrunk = 0;
 
 
 230
 231	NEIGH_CACHE_STAT_INC(tbl, forced_gc_runs);
 232
 233	write_lock_bh(&tbl->lock);
 
 
 
 
 
 234
 235	list_for_each_entry_safe(n, tmp, &tbl->gc_list, gc_list) {
 236		if (refcount_read(&n->refcnt) == 1) {
 237			bool remove = false;
 238
 
 
 
 239			write_lock(&n->lock);
 240			if ((n->nud_state == NUD_FAILED) ||
 241			    (n->nud_state == NUD_NOARP) ||
 242			    (tbl->is_multicast &&
 243			     tbl->is_multicast(n->primary_key)) ||
 244			    time_after(tref, n->updated))
 245				remove = true;
 
 
 
 
 
 246			write_unlock(&n->lock);
 247
 248			if (remove && neigh_remove_one(n, tbl))
 249				shrunk++;
 250			if (shrunk >= max_clean)
 251				break;
 252		}
 253	}
 254
 255	tbl->last_flush = jiffies;
 256
 257	write_unlock_bh(&tbl->lock);
 258
 259	return shrunk;
 260}
 261
 262static void neigh_add_timer(struct neighbour *n, unsigned long when)
 263{
 264	neigh_hold(n);
 265	if (unlikely(mod_timer(&n->timer, when))) {
 266		printk("NEIGH: BUG, double timer add, state is %x\n",
 267		       n->nud_state);
 268		dump_stack();
 269	}
 270}
 271
 272static int neigh_del_timer(struct neighbour *n)
 273{
 274	if ((n->nud_state & NUD_IN_TIMER) &&
 275	    del_timer(&n->timer)) {
 276		neigh_release(n);
 277		return 1;
 278	}
 279	return 0;
 280}
 281
 282static void pneigh_queue_purge(struct sk_buff_head *list)
 283{
 284	struct sk_buff *skb;
 285
 286	while ((skb = skb_dequeue(list)) != NULL) {
 287		dev_put(skb->dev);
 288		kfree_skb(skb);
 289	}
 290}
 291
 292static void neigh_flush_dev(struct neigh_table *tbl, struct net_device *dev,
 293			    bool skip_perm)
 294{
 295	int i;
 296	struct neigh_hash_table *nht;
 297
 298	nht = rcu_dereference_protected(tbl->nht,
 299					lockdep_is_held(&tbl->lock));
 300
 301	for (i = 0; i < (1 << nht->hash_shift); i++) {
 302		struct neighbour *n;
 303		struct neighbour __rcu **np = &nht->hash_buckets[i];
 304
 305		while ((n = rcu_dereference_protected(*np,
 306					lockdep_is_held(&tbl->lock))) != NULL) {
 307			if (dev && n->dev != dev) {
 308				np = &n->next;
 309				continue;
 310			}
 311			if (skip_perm && n->nud_state & NUD_PERMANENT) {
 312				np = &n->next;
 313				continue;
 314			}
 315			rcu_assign_pointer(*np,
 316				   rcu_dereference_protected(n->next,
 317						lockdep_is_held(&tbl->lock)));
 318			write_lock(&n->lock);
 319			neigh_del_timer(n);
 320			neigh_mark_dead(n);
 321			if (refcount_read(&n->refcnt) != 1) {
 
 322				/* The most unpleasant situation.
 323				   We must destroy neighbour entry,
 324				   but someone still uses it.
 325
 326				   The destroy will be delayed until
 327				   the last user releases us, but
 328				   we must kill timers etc. and move
 329				   it to safe state.
 330				 */
 331				__skb_queue_purge(&n->arp_queue);
 332				n->arp_queue_len_bytes = 0;
 333				n->output = neigh_blackhole;
 334				if (n->nud_state & NUD_VALID)
 335					n->nud_state = NUD_NOARP;
 336				else
 337					n->nud_state = NUD_NONE;
 338				neigh_dbg(2, "neigh %p is stray\n", n);
 339			}
 340			write_unlock(&n->lock);
 341			neigh_cleanup_and_release(n);
 342		}
 343	}
 344}
 345
 346void neigh_changeaddr(struct neigh_table *tbl, struct net_device *dev)
 347{
 348	write_lock_bh(&tbl->lock);
 349	neigh_flush_dev(tbl, dev, false);
 350	write_unlock_bh(&tbl->lock);
 351}
 352EXPORT_SYMBOL(neigh_changeaddr);
 353
 354static int __neigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
 355			  bool skip_perm)
 356{
 357	write_lock_bh(&tbl->lock);
 358	neigh_flush_dev(tbl, dev, skip_perm);
 359	pneigh_ifdown_and_unlock(tbl, dev);
 
 360
 361	del_timer_sync(&tbl->proxy_timer);
 362	pneigh_queue_purge(&tbl->proxy_queue);
 363	return 0;
 364}
 365
 366int neigh_carrier_down(struct neigh_table *tbl, struct net_device *dev)
 367{
 368	__neigh_ifdown(tbl, dev, true);
 369	return 0;
 370}
 371EXPORT_SYMBOL(neigh_carrier_down);
 372
 373int neigh_ifdown(struct neigh_table *tbl, struct net_device *dev)
 374{
 375	__neigh_ifdown(tbl, dev, false);
 376	return 0;
 377}
 378EXPORT_SYMBOL(neigh_ifdown);
 379
 380static struct neighbour *neigh_alloc(struct neigh_table *tbl,
 381				     struct net_device *dev,
 382				     bool exempt_from_gc)
 383{
 384	struct neighbour *n = NULL;
 385	unsigned long now = jiffies;
 386	int entries;
 387
 388	if (exempt_from_gc)
 389		goto do_alloc;
 390
 391	entries = atomic_inc_return(&tbl->gc_entries) - 1;
 392	if (entries >= tbl->gc_thresh3 ||
 393	    (entries >= tbl->gc_thresh2 &&
 394	     time_after(now, tbl->last_flush + 5 * HZ))) {
 395		if (!neigh_forced_gc(tbl) &&
 396		    entries >= tbl->gc_thresh3) {
 397			net_info_ratelimited("%s: neighbor table overflow!\n",
 398					     tbl->id);
 399			NEIGH_CACHE_STAT_INC(tbl, table_fulls);
 400			goto out_entries;
 401		}
 402	}
 403
 404do_alloc:
 405	n = kzalloc(tbl->entry_size + dev->neigh_priv_len, GFP_ATOMIC);
 406	if (!n)
 407		goto out_entries;
 408
 409	__skb_queue_head_init(&n->arp_queue);
 410	rwlock_init(&n->lock);
 411	seqlock_init(&n->ha_lock);
 412	n->updated	  = n->used = now;
 413	n->nud_state	  = NUD_NONE;
 414	n->output	  = neigh_blackhole;
 415	seqlock_init(&n->hh.hh_lock);
 416	n->parms	  = neigh_parms_clone(&tbl->parms);
 417	timer_setup(&n->timer, neigh_timer_handler, 0);
 418
 419	NEIGH_CACHE_STAT_INC(tbl, allocs);
 420	n->tbl		  = tbl;
 421	refcount_set(&n->refcnt, 1);
 422	n->dead		  = 1;
 423	INIT_LIST_HEAD(&n->gc_list);
 424
 425	atomic_inc(&tbl->entries);
 426out:
 427	return n;
 428
 429out_entries:
 430	if (!exempt_from_gc)
 431		atomic_dec(&tbl->gc_entries);
 432	goto out;
 433}
 434
 435static void neigh_get_hash_rnd(u32 *x)
 436{
 437	*x = get_random_u32() | 1;
 438}
 439
 440static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
 441{
 442	size_t size = (1 << shift) * sizeof(struct neighbour *);
 443	struct neigh_hash_table *ret;
 444	struct neighbour __rcu **buckets;
 445	int i;
 446
 447	ret = kmalloc(sizeof(*ret), GFP_ATOMIC);
 448	if (!ret)
 449		return NULL;
 450	if (size <= PAGE_SIZE) {
 451		buckets = kzalloc(size, GFP_ATOMIC);
 452	} else {
 453		buckets = (struct neighbour __rcu **)
 454			  __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
 455					   get_order(size));
 456		kmemleak_alloc(buckets, size, 1, GFP_ATOMIC);
 457	}
 458	if (!buckets) {
 459		kfree(ret);
 460		return NULL;
 461	}
 462	ret->hash_buckets = buckets;
 463	ret->hash_shift = shift;
 464	for (i = 0; i < NEIGH_NUM_HASH_RND; i++)
 465		neigh_get_hash_rnd(&ret->hash_rnd[i]);
 466	return ret;
 467}
 468
 469static void neigh_hash_free_rcu(struct rcu_head *head)
 470{
 471	struct neigh_hash_table *nht = container_of(head,
 472						    struct neigh_hash_table,
 473						    rcu);
 474	size_t size = (1 << nht->hash_shift) * sizeof(struct neighbour *);
 475	struct neighbour __rcu **buckets = nht->hash_buckets;
 476
 477	if (size <= PAGE_SIZE) {
 478		kfree(buckets);
 479	} else {
 480		kmemleak_free(buckets);
 481		free_pages((unsigned long)buckets, get_order(size));
 482	}
 483	kfree(nht);
 484}
 485
 486static struct neigh_hash_table *neigh_hash_grow(struct neigh_table *tbl,
 487						unsigned long new_shift)
 488{
 489	unsigned int i, hash;
 490	struct neigh_hash_table *new_nht, *old_nht;
 491
 492	NEIGH_CACHE_STAT_INC(tbl, hash_grows);
 493
 494	old_nht = rcu_dereference_protected(tbl->nht,
 495					    lockdep_is_held(&tbl->lock));
 496	new_nht = neigh_hash_alloc(new_shift);
 497	if (!new_nht)
 498		return old_nht;
 499
 500	for (i = 0; i < (1 << old_nht->hash_shift); i++) {
 501		struct neighbour *n, *next;
 502
 503		for (n = rcu_dereference_protected(old_nht->hash_buckets[i],
 504						   lockdep_is_held(&tbl->lock));
 505		     n != NULL;
 506		     n = next) {
 507			hash = tbl->hash(n->primary_key, n->dev,
 508					 new_nht->hash_rnd);
 509
 510			hash >>= (32 - new_nht->hash_shift);
 511			next = rcu_dereference_protected(n->next,
 512						lockdep_is_held(&tbl->lock));
 513
 514			rcu_assign_pointer(n->next,
 515					   rcu_dereference_protected(
 516						new_nht->hash_buckets[hash],
 517						lockdep_is_held(&tbl->lock)));
 518			rcu_assign_pointer(new_nht->hash_buckets[hash], n);
 519		}
 520	}
 521
 522	rcu_assign_pointer(tbl->nht, new_nht);
 523	call_rcu(&old_nht->rcu, neigh_hash_free_rcu);
 524	return new_nht;
 525}
 526
 527struct neighbour *neigh_lookup(struct neigh_table *tbl, const void *pkey,
 528			       struct net_device *dev)
 529{
 530	struct neighbour *n;
 
 
 
 531
 532	NEIGH_CACHE_STAT_INC(tbl, lookups);
 533
 534	rcu_read_lock_bh();
 535	n = __neigh_lookup_noref(tbl, pkey, dev);
 536	if (n) {
 537		if (!refcount_inc_not_zero(&n->refcnt))
 538			n = NULL;
 539		NEIGH_CACHE_STAT_INC(tbl, hits);
 
 
 
 
 
 
 
 540	}
 541
 542	rcu_read_unlock_bh();
 543	return n;
 544}
 545EXPORT_SYMBOL(neigh_lookup);
 546
 547struct neighbour *neigh_lookup_nodev(struct neigh_table *tbl, struct net *net,
 548				     const void *pkey)
 549{
 550	struct neighbour *n;
 551	unsigned int key_len = tbl->key_len;
 552	u32 hash_val;
 553	struct neigh_hash_table *nht;
 554
 555	NEIGH_CACHE_STAT_INC(tbl, lookups);
 556
 557	rcu_read_lock_bh();
 558	nht = rcu_dereference_bh(tbl->nht);
 559	hash_val = tbl->hash(pkey, NULL, nht->hash_rnd) >> (32 - nht->hash_shift);
 560
 561	for (n = rcu_dereference_bh(nht->hash_buckets[hash_val]);
 562	     n != NULL;
 563	     n = rcu_dereference_bh(n->next)) {
 564		if (!memcmp(n->primary_key, pkey, key_len) &&
 565		    net_eq(dev_net(n->dev), net)) {
 566			if (!refcount_inc_not_zero(&n->refcnt))
 567				n = NULL;
 568			NEIGH_CACHE_STAT_INC(tbl, hits);
 569			break;
 570		}
 571	}
 572
 573	rcu_read_unlock_bh();
 574	return n;
 575}
 576EXPORT_SYMBOL(neigh_lookup_nodev);
 577
 578static struct neighbour *___neigh_create(struct neigh_table *tbl,
 579					 const void *pkey,
 580					 struct net_device *dev,
 581					 bool exempt_from_gc, bool want_ref)
 582{
 583	struct neighbour *n1, *rc, *n = neigh_alloc(tbl, dev, exempt_from_gc);
 584	u32 hash_val;
 585	unsigned int key_len = tbl->key_len;
 586	int error;
 
 587	struct neigh_hash_table *nht;
 588
 589	trace_neigh_create(tbl, dev, pkey, n, exempt_from_gc);
 590
 591	if (!n) {
 592		rc = ERR_PTR(-ENOBUFS);
 593		goto out;
 594	}
 595
 596	memcpy(n->primary_key, pkey, key_len);
 597	n->dev = dev;
 598	dev_hold(dev);
 599
 600	/* Protocol specific setup. */
 601	if (tbl->constructor &&	(error = tbl->constructor(n)) < 0) {
 602		rc = ERR_PTR(error);
 603		goto out_neigh_release;
 604	}
 605
 606	if (dev->netdev_ops->ndo_neigh_construct) {
 607		error = dev->netdev_ops->ndo_neigh_construct(dev, n);
 608		if (error < 0) {
 609			rc = ERR_PTR(error);
 610			goto out_neigh_release;
 611		}
 612	}
 613
 614	/* Device specific setup. */
 615	if (n->parms->neigh_setup &&
 616	    (error = n->parms->neigh_setup(n)) < 0) {
 617		rc = ERR_PTR(error);
 618		goto out_neigh_release;
 619	}
 620
 621	n->confirmed = jiffies - (NEIGH_VAR(n->parms, BASE_REACHABLE_TIME) << 1);
 622
 623	write_lock_bh(&tbl->lock);
 624	nht = rcu_dereference_protected(tbl->nht,
 625					lockdep_is_held(&tbl->lock));
 626
 627	if (atomic_read(&tbl->entries) > (1 << nht->hash_shift))
 628		nht = neigh_hash_grow(tbl, nht->hash_shift + 1);
 629
 630	hash_val = tbl->hash(n->primary_key, dev, nht->hash_rnd) >> (32 - nht->hash_shift);
 631
 632	if (n->parms->dead) {
 633		rc = ERR_PTR(-EINVAL);
 634		goto out_tbl_unlock;
 635	}
 636
 637	for (n1 = rcu_dereference_protected(nht->hash_buckets[hash_val],
 638					    lockdep_is_held(&tbl->lock));
 639	     n1 != NULL;
 640	     n1 = rcu_dereference_protected(n1->next,
 641			lockdep_is_held(&tbl->lock))) {
 642		if (dev == n1->dev && !memcmp(n1->primary_key, n->primary_key, key_len)) {
 643			if (want_ref)
 644				neigh_hold(n1);
 645			rc = n1;
 646			goto out_tbl_unlock;
 647		}
 648	}
 649
 650	n->dead = 0;
 651	if (!exempt_from_gc)
 652		list_add_tail(&n->gc_list, &n->tbl->gc_list);
 653
 654	if (want_ref)
 655		neigh_hold(n);
 656	rcu_assign_pointer(n->next,
 657			   rcu_dereference_protected(nht->hash_buckets[hash_val],
 658						     lockdep_is_held(&tbl->lock)));
 659	rcu_assign_pointer(nht->hash_buckets[hash_val], n);
 660	write_unlock_bh(&tbl->lock);
 661	neigh_dbg(2, "neigh %p is created\n", n);
 662	rc = n;
 663out:
 664	return rc;
 665out_tbl_unlock:
 666	write_unlock_bh(&tbl->lock);
 667out_neigh_release:
 668	if (!exempt_from_gc)
 669		atomic_dec(&tbl->gc_entries);
 670	neigh_release(n);
 671	goto out;
 672}
 
 673
 674struct neighbour *__neigh_create(struct neigh_table *tbl, const void *pkey,
 675				 struct net_device *dev, bool want_ref)
 676{
 677	return ___neigh_create(tbl, pkey, dev, false, want_ref);
 678}
 679EXPORT_SYMBOL(__neigh_create);
 680
 681static u32 pneigh_hash(const void *pkey, unsigned int key_len)
 682{
 683	u32 hash_val = *(u32 *)(pkey + key_len - 4);
 684	hash_val ^= (hash_val >> 16);
 685	hash_val ^= hash_val >> 8;
 686	hash_val ^= hash_val >> 4;
 687	hash_val &= PNEIGH_HASHMASK;
 688	return hash_val;
 689}
 690
 691static struct pneigh_entry *__pneigh_lookup_1(struct pneigh_entry *n,
 692					      struct net *net,
 693					      const void *pkey,
 694					      unsigned int key_len,
 695					      struct net_device *dev)
 696{
 697	while (n) {
 698		if (!memcmp(n->key, pkey, key_len) &&
 699		    net_eq(pneigh_net(n), net) &&
 700		    (n->dev == dev || !n->dev))
 701			return n;
 702		n = n->next;
 703	}
 704	return NULL;
 705}
 706
 707struct pneigh_entry *__pneigh_lookup(struct neigh_table *tbl,
 708		struct net *net, const void *pkey, struct net_device *dev)
 709{
 710	unsigned int key_len = tbl->key_len;
 711	u32 hash_val = pneigh_hash(pkey, key_len);
 712
 713	return __pneigh_lookup_1(tbl->phash_buckets[hash_val],
 714				 net, pkey, key_len, dev);
 715}
 716EXPORT_SYMBOL_GPL(__pneigh_lookup);
 717
 718struct pneigh_entry * pneigh_lookup(struct neigh_table *tbl,
 719				    struct net *net, const void *pkey,
 720				    struct net_device *dev, int creat)
 721{
 722	struct pneigh_entry *n;
 723	unsigned int key_len = tbl->key_len;
 724	u32 hash_val = pneigh_hash(pkey, key_len);
 725
 726	read_lock_bh(&tbl->lock);
 727	n = __pneigh_lookup_1(tbl->phash_buckets[hash_val],
 728			      net, pkey, key_len, dev);
 729	read_unlock_bh(&tbl->lock);
 730
 731	if (n || !creat)
 732		goto out;
 733
 734	ASSERT_RTNL();
 735
 736	n = kmalloc(sizeof(*n) + key_len, GFP_KERNEL);
 737	if (!n)
 738		goto out;
 739
 740	n->protocol = 0;
 741	write_pnet(&n->net, net);
 742	memcpy(n->key, pkey, key_len);
 743	n->dev = dev;
 744	if (dev)
 745		dev_hold(dev);
 746
 747	if (tbl->pconstructor && tbl->pconstructor(n)) {
 748		if (dev)
 749			dev_put(dev);
 
 750		kfree(n);
 751		n = NULL;
 752		goto out;
 753	}
 754
 755	write_lock_bh(&tbl->lock);
 756	n->next = tbl->phash_buckets[hash_val];
 757	tbl->phash_buckets[hash_val] = n;
 758	write_unlock_bh(&tbl->lock);
 759out:
 760	return n;
 761}
 762EXPORT_SYMBOL(pneigh_lookup);
 763
 764
 765int pneigh_delete(struct neigh_table *tbl, struct net *net, const void *pkey,
 766		  struct net_device *dev)
 767{
 768	struct pneigh_entry *n, **np;
 769	unsigned int key_len = tbl->key_len;
 770	u32 hash_val = pneigh_hash(pkey, key_len);
 771
 772	write_lock_bh(&tbl->lock);
 773	for (np = &tbl->phash_buckets[hash_val]; (n = *np) != NULL;
 774	     np = &n->next) {
 775		if (!memcmp(n->key, pkey, key_len) && n->dev == dev &&
 776		    net_eq(pneigh_net(n), net)) {
 777			*np = n->next;
 778			write_unlock_bh(&tbl->lock);
 779			if (tbl->pdestructor)
 780				tbl->pdestructor(n);
 781			if (n->dev)
 782				dev_put(n->dev);
 
 783			kfree(n);
 784			return 0;
 785		}
 786	}
 787	write_unlock_bh(&tbl->lock);
 788	return -ENOENT;
 789}
 790
 791static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
 792				    struct net_device *dev)
 793{
 794	struct pneigh_entry *n, **np, *freelist = NULL;
 795	u32 h;
 796
 797	for (h = 0; h <= PNEIGH_HASHMASK; h++) {
 798		np = &tbl->phash_buckets[h];
 799		while ((n = *np) != NULL) {
 800			if (!dev || n->dev == dev) {
 801				*np = n->next;
 802				n->next = freelist;
 803				freelist = n;
 
 
 
 
 804				continue;
 805			}
 806			np = &n->next;
 807		}
 808	}
 809	write_unlock_bh(&tbl->lock);
 810	while ((n = freelist)) {
 811		freelist = n->next;
 812		n->next = NULL;
 813		if (tbl->pdestructor)
 814			tbl->pdestructor(n);
 815		if (n->dev)
 816			dev_put(n->dev);
 817		kfree(n);
 818	}
 819	return -ENOENT;
 820}
 821
 822static void neigh_parms_destroy(struct neigh_parms *parms);
 823
 824static inline void neigh_parms_put(struct neigh_parms *parms)
 825{
 826	if (refcount_dec_and_test(&parms->refcnt))
 827		neigh_parms_destroy(parms);
 828}
 829
 
 
 
 
 
 
 830/*
 831 *	neighbour must already be out of the table;
 832 *
 833 */
 834void neigh_destroy(struct neighbour *neigh)
 835{
 836	struct net_device *dev = neigh->dev;
 837
 838	NEIGH_CACHE_STAT_INC(neigh->tbl, destroys);
 839
 840	if (!neigh->dead) {
 841		pr_warn("Destroying alive neighbour %p\n", neigh);
 
 842		dump_stack();
 843		return;
 844	}
 845
 846	if (neigh_del_timer(neigh))
 847		pr_warn("Impossible event\n");
 848
 849	write_lock_bh(&neigh->lock);
 850	__skb_queue_purge(&neigh->arp_queue);
 851	write_unlock_bh(&neigh->lock);
 852	neigh->arp_queue_len_bytes = 0;
 853
 854	if (dev->netdev_ops->ndo_neigh_destroy)
 855		dev->netdev_ops->ndo_neigh_destroy(dev, neigh);
 856
 857	dev_put(dev);
 858	neigh_parms_put(neigh->parms);
 859
 860	neigh_dbg(2, "neigh %p is destroyed\n", neigh);
 861
 862	atomic_dec(&neigh->tbl->entries);
 863	kfree_rcu(neigh, rcu);
 864}
 865EXPORT_SYMBOL(neigh_destroy);
 866
 867/* Neighbour state is suspicious;
 868   disable fast path.
 869
 870   Called with write_locked neigh.
 871 */
 872static void neigh_suspect(struct neighbour *neigh)
 873{
 874	neigh_dbg(2, "neigh %p is suspected\n", neigh);
 875
 876	neigh->output = neigh->ops->output;
 877}
 878
 879/* Neighbour state is OK;
 880   enable fast path.
 881
 882   Called with write_locked neigh.
 883 */
 884static void neigh_connect(struct neighbour *neigh)
 885{
 886	neigh_dbg(2, "neigh %p is connected\n", neigh);
 887
 888	neigh->output = neigh->ops->connected_output;
 889}
 890
 891static void neigh_periodic_work(struct work_struct *work)
 892{
 893	struct neigh_table *tbl = container_of(work, struct neigh_table, gc_work.work);
 894	struct neighbour *n;
 895	struct neighbour __rcu **np;
 896	unsigned int i;
 897	struct neigh_hash_table *nht;
 898
 899	NEIGH_CACHE_STAT_INC(tbl, periodic_gc_runs);
 900
 901	write_lock_bh(&tbl->lock);
 902	nht = rcu_dereference_protected(tbl->nht,
 903					lockdep_is_held(&tbl->lock));
 904
 905	/*
 906	 *	periodically recompute ReachableTime from random function
 907	 */
 908
 909	if (time_after(jiffies, tbl->last_rand + 300 * HZ)) {
 910		struct neigh_parms *p;
 911		tbl->last_rand = jiffies;
 912		list_for_each_entry(p, &tbl->parms_list, list)
 913			p->reachable_time =
 914				neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
 915	}
 916
 917	if (atomic_read(&tbl->entries) < tbl->gc_thresh1)
 918		goto out;
 919
 920	for (i = 0 ; i < (1 << nht->hash_shift); i++) {
 921		np = &nht->hash_buckets[i];
 922
 923		while ((n = rcu_dereference_protected(*np,
 924				lockdep_is_held(&tbl->lock))) != NULL) {
 925			unsigned int state;
 926
 927			write_lock(&n->lock);
 928
 929			state = n->nud_state;
 930			if ((state & (NUD_PERMANENT | NUD_IN_TIMER)) ||
 931			    (n->flags & NTF_EXT_LEARNED)) {
 932				write_unlock(&n->lock);
 933				goto next_elt;
 934			}
 935
 936			if (time_before(n->used, n->confirmed))
 937				n->used = n->confirmed;
 938
 939			if (refcount_read(&n->refcnt) == 1 &&
 940			    (state == NUD_FAILED ||
 941			     time_after(jiffies, n->used + NEIGH_VAR(n->parms, GC_STALETIME)))) {
 942				*np = n->next;
 943				neigh_mark_dead(n);
 944				write_unlock(&n->lock);
 945				neigh_cleanup_and_release(n);
 946				continue;
 947			}
 948			write_unlock(&n->lock);
 949
 950next_elt:
 951			np = &n->next;
 952		}
 953		/*
 954		 * It's fine to release lock here, even if hash table
 955		 * grows while we are preempted.
 956		 */
 957		write_unlock_bh(&tbl->lock);
 958		cond_resched();
 959		write_lock_bh(&tbl->lock);
 960		nht = rcu_dereference_protected(tbl->nht,
 961						lockdep_is_held(&tbl->lock));
 962	}
 963out:
 964	/* Cycle through all hash buckets every BASE_REACHABLE_TIME/2 ticks.
 965	 * ARP entry timeouts range from 1/2 BASE_REACHABLE_TIME to 3/2
 966	 * BASE_REACHABLE_TIME.
 967	 */
 968	queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
 969			      NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME) >> 1);
 970	write_unlock_bh(&tbl->lock);
 971}
 972
 973static __inline__ int neigh_max_probes(struct neighbour *n)
 974{
 975	struct neigh_parms *p = n->parms;
 976	return NEIGH_VAR(p, UCAST_PROBES) + NEIGH_VAR(p, APP_PROBES) +
 977	       (n->nud_state & NUD_PROBE ? NEIGH_VAR(p, MCAST_REPROBES) :
 978	        NEIGH_VAR(p, MCAST_PROBES));
 979}
 980
 981static void neigh_invalidate(struct neighbour *neigh)
 982	__releases(neigh->lock)
 983	__acquires(neigh->lock)
 984{
 985	struct sk_buff *skb;
 986
 987	NEIGH_CACHE_STAT_INC(neigh->tbl, res_failed);
 988	neigh_dbg(2, "neigh %p is failed\n", neigh);
 989	neigh->updated = jiffies;
 990
 991	/* It is very thin place. report_unreachable is very complicated
 992	   routine. Particularly, it can hit the same neighbour entry!
 993
 994	   So that, we try to be accurate and avoid dead loop. --ANK
 995	 */
 996	while (neigh->nud_state == NUD_FAILED &&
 997	       (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
 998		write_unlock(&neigh->lock);
 999		neigh->ops->error_report(neigh, skb);
1000		write_lock(&neigh->lock);
1001	}
1002	__skb_queue_purge(&neigh->arp_queue);
1003	neigh->arp_queue_len_bytes = 0;
1004}
1005
1006static void neigh_probe(struct neighbour *neigh)
1007	__releases(neigh->lock)
1008{
1009	struct sk_buff *skb = skb_peek_tail(&neigh->arp_queue);
1010	/* keep skb alive even if arp_queue overflows */
1011	if (skb)
1012		skb = skb_clone(skb, GFP_ATOMIC);
1013	write_unlock(&neigh->lock);
1014	if (neigh->ops->solicit)
1015		neigh->ops->solicit(neigh, skb);
1016	atomic_inc(&neigh->probes);
1017	consume_skb(skb);
1018}
1019
1020/* Called when a timer expires for a neighbour entry. */
1021
1022static void neigh_timer_handler(struct timer_list *t)
1023{
1024	unsigned long now, next;
1025	struct neighbour *neigh = from_timer(neigh, t, timer);
1026	unsigned int state;
1027	int notify = 0;
1028
1029	write_lock(&neigh->lock);
1030
1031	state = neigh->nud_state;
1032	now = jiffies;
1033	next = now + HZ;
1034
1035	if (!(state & NUD_IN_TIMER))
 
 
 
1036		goto out;
 
1037
1038	if (state & NUD_REACHABLE) {
1039		if (time_before_eq(now,
1040				   neigh->confirmed + neigh->parms->reachable_time)) {
1041			neigh_dbg(2, "neigh %p is still alive\n", neigh);
1042			next = neigh->confirmed + neigh->parms->reachable_time;
1043		} else if (time_before_eq(now,
1044					  neigh->used +
1045					  NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
1046			neigh_dbg(2, "neigh %p is delayed\n", neigh);
1047			neigh->nud_state = NUD_DELAY;
1048			neigh->updated = jiffies;
1049			neigh_suspect(neigh);
1050			next = now + NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME);
1051		} else {
1052			neigh_dbg(2, "neigh %p is suspected\n", neigh);
1053			neigh->nud_state = NUD_STALE;
1054			neigh->updated = jiffies;
1055			neigh_suspect(neigh);
1056			notify = 1;
1057		}
1058	} else if (state & NUD_DELAY) {
1059		if (time_before_eq(now,
1060				   neigh->confirmed +
1061				   NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME))) {
1062			neigh_dbg(2, "neigh %p is now reachable\n", neigh);
1063			neigh->nud_state = NUD_REACHABLE;
1064			neigh->updated = jiffies;
1065			neigh_connect(neigh);
1066			notify = 1;
1067			next = neigh->confirmed + neigh->parms->reachable_time;
1068		} else {
1069			neigh_dbg(2, "neigh %p is probed\n", neigh);
1070			neigh->nud_state = NUD_PROBE;
1071			neigh->updated = jiffies;
1072			atomic_set(&neigh->probes, 0);
1073			notify = 1;
1074			next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1075					 HZ/100);
1076		}
1077	} else {
1078		/* NUD_PROBE|NUD_INCOMPLETE */
1079		next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME), HZ/100);
1080	}
1081
1082	if ((neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) &&
1083	    atomic_read(&neigh->probes) >= neigh_max_probes(neigh)) {
1084		neigh->nud_state = NUD_FAILED;
1085		notify = 1;
1086		neigh_invalidate(neigh);
1087		goto out;
1088	}
1089
1090	if (neigh->nud_state & NUD_IN_TIMER) {
1091		if (time_before(next, jiffies + HZ/100))
1092			next = jiffies + HZ/100;
1093		if (!mod_timer(&neigh->timer, next))
1094			neigh_hold(neigh);
1095	}
1096	if (neigh->nud_state & (NUD_INCOMPLETE | NUD_PROBE)) {
1097		neigh_probe(neigh);
 
 
 
 
 
 
 
1098	} else {
1099out:
1100		write_unlock(&neigh->lock);
1101	}
1102
1103	if (notify)
1104		neigh_update_notify(neigh, 0);
1105
1106	trace_neigh_timer_handler(neigh, 0);
1107
1108	neigh_release(neigh);
1109}
1110
1111int __neigh_event_send(struct neighbour *neigh, struct sk_buff *skb)
1112{
1113	int rc;
1114	bool immediate_probe = false;
1115
1116	write_lock_bh(&neigh->lock);
1117
1118	rc = 0;
1119	if (neigh->nud_state & (NUD_CONNECTED | NUD_DELAY | NUD_PROBE))
1120		goto out_unlock_bh;
1121	if (neigh->dead)
1122		goto out_dead;
1123
1124	if (!(neigh->nud_state & (NUD_STALE | NUD_INCOMPLETE))) {
1125		if (NEIGH_VAR(neigh->parms, MCAST_PROBES) +
1126		    NEIGH_VAR(neigh->parms, APP_PROBES)) {
1127			unsigned long next, now = jiffies;
1128
1129			atomic_set(&neigh->probes,
1130				   NEIGH_VAR(neigh->parms, UCAST_PROBES));
1131			neigh_del_timer(neigh);
1132			neigh->nud_state     = NUD_INCOMPLETE;
1133			neigh->updated = now;
1134			next = now + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1135					 HZ/100);
1136			neigh_add_timer(neigh, next);
1137			immediate_probe = true;
1138		} else {
1139			neigh->nud_state = NUD_FAILED;
1140			neigh->updated = jiffies;
1141			write_unlock_bh(&neigh->lock);
1142
1143			kfree_skb(skb);
1144			return 1;
1145		}
1146	} else if (neigh->nud_state & NUD_STALE) {
1147		neigh_dbg(2, "neigh %p is delayed\n", neigh);
1148		neigh_del_timer(neigh);
1149		neigh->nud_state = NUD_DELAY;
1150		neigh->updated = jiffies;
1151		neigh_add_timer(neigh, jiffies +
1152				NEIGH_VAR(neigh->parms, DELAY_PROBE_TIME));
1153	}
1154
1155	if (neigh->nud_state == NUD_INCOMPLETE) {
1156		if (skb) {
1157			while (neigh->arp_queue_len_bytes + skb->truesize >
1158			       NEIGH_VAR(neigh->parms, QUEUE_LEN_BYTES)) {
1159				struct sk_buff *buff;
1160
1161				buff = __skb_dequeue(&neigh->arp_queue);
1162				if (!buff)
1163					break;
1164				neigh->arp_queue_len_bytes -= buff->truesize;
1165				kfree_skb(buff);
1166				NEIGH_CACHE_STAT_INC(neigh->tbl, unres_discards);
1167			}
1168			skb_dst_force(skb);
1169			__skb_queue_tail(&neigh->arp_queue, skb);
1170			neigh->arp_queue_len_bytes += skb->truesize;
1171		}
1172		rc = 1;
1173	}
1174out_unlock_bh:
1175	if (immediate_probe)
1176		neigh_probe(neigh);
1177	else
1178		write_unlock(&neigh->lock);
1179	local_bh_enable();
1180	trace_neigh_event_send_done(neigh, rc);
1181	return rc;
1182
1183out_dead:
1184	if (neigh->nud_state & NUD_STALE)
1185		goto out_unlock_bh;
1186	write_unlock_bh(&neigh->lock);
1187	kfree_skb(skb);
1188	trace_neigh_event_send_dead(neigh, 1);
1189	return 1;
1190}
1191EXPORT_SYMBOL(__neigh_event_send);
1192
1193static void neigh_update_hhs(struct neighbour *neigh)
1194{
1195	struct hh_cache *hh;
1196	void (*update)(struct hh_cache*, const struct net_device*, const unsigned char *)
1197		= NULL;
1198
1199	if (neigh->dev->header_ops)
1200		update = neigh->dev->header_ops->cache_update;
1201
1202	if (update) {
1203		hh = &neigh->hh;
1204		if (READ_ONCE(hh->hh_len)) {
1205			write_seqlock_bh(&hh->hh_lock);
1206			update(hh, neigh->dev, neigh->ha);
1207			write_sequnlock_bh(&hh->hh_lock);
1208		}
1209	}
1210}
1211
1212
1213
1214/* Generic update routine.
1215   -- lladdr is new lladdr or NULL, if it is not supplied.
1216   -- new    is new state.
1217   -- flags
1218	NEIGH_UPDATE_F_OVERRIDE allows to override existing lladdr,
1219				if it is different.
1220	NEIGH_UPDATE_F_WEAK_OVERRIDE will suspect existing "connected"
1221				lladdr instead of overriding it
1222				if it is different.
 
 
1223	NEIGH_UPDATE_F_ADMIN	means that the change is administrative.
1224
1225	NEIGH_UPDATE_F_OVERRIDE_ISROUTER allows to override existing
1226				NTF_ROUTER flag.
1227	NEIGH_UPDATE_F_ISROUTER	indicates if the neighbour is known as
1228				a router.
1229
1230   Caller MUST hold reference count on the entry.
1231 */
1232
1233static int __neigh_update(struct neighbour *neigh, const u8 *lladdr,
1234			  u8 new, u32 flags, u32 nlmsg_pid,
1235			  struct netlink_ext_ack *extack)
1236{
1237	bool ext_learn_change = false;
1238	u8 old;
1239	int err;
1240	int notify = 0;
1241	struct net_device *dev;
1242	int update_isrouter = 0;
1243
1244	trace_neigh_update(neigh, lladdr, new, flags, nlmsg_pid);
1245
1246	write_lock_bh(&neigh->lock);
1247
1248	dev    = neigh->dev;
1249	old    = neigh->nud_state;
1250	err    = -EPERM;
1251
1252	if (neigh->dead) {
1253		NL_SET_ERR_MSG(extack, "Neighbor entry is now dead");
1254		new = old;
1255		goto out;
1256	}
1257	if (!(flags & NEIGH_UPDATE_F_ADMIN) &&
1258	    (old & (NUD_NOARP | NUD_PERMANENT)))
1259		goto out;
1260
1261	ext_learn_change = neigh_update_ext_learned(neigh, flags, &notify);
1262
1263	if (!(new & NUD_VALID)) {
1264		neigh_del_timer(neigh);
1265		if (old & NUD_CONNECTED)
1266			neigh_suspect(neigh);
1267		neigh->nud_state = new;
1268		err = 0;
1269		notify = old & NUD_VALID;
1270		if ((old & (NUD_INCOMPLETE | NUD_PROBE)) &&
1271		    (new & NUD_FAILED)) {
1272			neigh_invalidate(neigh);
1273			notify = 1;
1274		}
1275		goto out;
1276	}
1277
1278	/* Compare new lladdr with cached one */
1279	if (!dev->addr_len) {
1280		/* First case: device needs no address. */
1281		lladdr = neigh->ha;
1282	} else if (lladdr) {
1283		/* The second case: if something is already cached
1284		   and a new address is proposed:
1285		   - compare new & old
1286		   - if they are different, check override flag
1287		 */
1288		if ((old & NUD_VALID) &&
1289		    !memcmp(lladdr, neigh->ha, dev->addr_len))
1290			lladdr = neigh->ha;
1291	} else {
1292		/* No address is supplied; if we know something,
1293		   use it, otherwise discard the request.
1294		 */
1295		err = -EINVAL;
1296		if (!(old & NUD_VALID)) {
1297			NL_SET_ERR_MSG(extack, "No link layer address given");
1298			goto out;
1299		}
1300		lladdr = neigh->ha;
1301	}
1302
1303	/* Update confirmed timestamp for neighbour entry after we
1304	 * received ARP packet even if it doesn't change IP to MAC binding.
1305	 */
1306	if (new & NUD_CONNECTED)
1307		neigh->confirmed = jiffies;
 
1308
1309	/* If entry was valid and address is not changed,
1310	   do not change entry state, if new one is STALE.
1311	 */
1312	err = 0;
1313	update_isrouter = flags & NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1314	if (old & NUD_VALID) {
1315		if (lladdr != neigh->ha && !(flags & NEIGH_UPDATE_F_OVERRIDE)) {
1316			update_isrouter = 0;
1317			if ((flags & NEIGH_UPDATE_F_WEAK_OVERRIDE) &&
1318			    (old & NUD_CONNECTED)) {
1319				lladdr = neigh->ha;
1320				new = NUD_STALE;
1321			} else
1322				goto out;
1323		} else {
1324			if (lladdr == neigh->ha && new == NUD_STALE &&
1325			    !(flags & NEIGH_UPDATE_F_ADMIN))
 
 
1326				new = old;
1327		}
1328	}
1329
1330	/* Update timestamp only once we know we will make a change to the
1331	 * neighbour entry. Otherwise we risk to move the locktime window with
1332	 * noop updates and ignore relevant ARP updates.
1333	 */
1334	if (new != old || lladdr != neigh->ha)
1335		neigh->updated = jiffies;
1336
1337	if (new != old) {
1338		neigh_del_timer(neigh);
1339		if (new & NUD_PROBE)
1340			atomic_set(&neigh->probes, 0);
1341		if (new & NUD_IN_TIMER)
1342			neigh_add_timer(neigh, (jiffies +
1343						((new & NUD_REACHABLE) ?
1344						 neigh->parms->reachable_time :
1345						 0)));
1346		neigh->nud_state = new;
1347		notify = 1;
1348	}
1349
1350	if (lladdr != neigh->ha) {
1351		write_seqlock(&neigh->ha_lock);
1352		memcpy(&neigh->ha, lladdr, dev->addr_len);
1353		write_sequnlock(&neigh->ha_lock);
1354		neigh_update_hhs(neigh);
1355		if (!(new & NUD_CONNECTED))
1356			neigh->confirmed = jiffies -
1357				      (NEIGH_VAR(neigh->parms, BASE_REACHABLE_TIME) << 1);
1358		notify = 1;
1359	}
1360	if (new == old)
1361		goto out;
1362	if (new & NUD_CONNECTED)
1363		neigh_connect(neigh);
1364	else
1365		neigh_suspect(neigh);
1366	if (!(old & NUD_VALID)) {
1367		struct sk_buff *skb;
1368
1369		/* Again: avoid dead loop if something went wrong */
1370
1371		while (neigh->nud_state & NUD_VALID &&
1372		       (skb = __skb_dequeue(&neigh->arp_queue)) != NULL) {
1373			struct dst_entry *dst = skb_dst(skb);
1374			struct neighbour *n2, *n1 = neigh;
1375			write_unlock_bh(&neigh->lock);
1376
1377			rcu_read_lock();
1378
1379			/* Why not just use 'neigh' as-is?  The problem is that
1380			 * things such as shaper, eql, and sch_teql can end up
1381			 * using alternative, different, neigh objects to output
1382			 * the packet in the output path.  So what we need to do
1383			 * here is re-lookup the top-level neigh in the path so
1384			 * we can reinject the packet there.
1385			 */
1386			n2 = NULL;
1387			if (dst && dst->obsolete != DST_OBSOLETE_DEAD) {
1388				n2 = dst_neigh_lookup_skb(dst, skb);
1389				if (n2)
1390					n1 = n2;
1391			}
1392			n1->output(n1, skb);
1393			if (n2)
1394				neigh_release(n2);
1395			rcu_read_unlock();
1396
1397			write_lock_bh(&neigh->lock);
1398		}
1399		__skb_queue_purge(&neigh->arp_queue);
1400		neigh->arp_queue_len_bytes = 0;
1401	}
1402out:
1403	if (update_isrouter)
1404		neigh_update_is_router(neigh, flags, &notify);
 
 
 
1405	write_unlock_bh(&neigh->lock);
1406
1407	if (((new ^ old) & NUD_PERMANENT) || ext_learn_change)
1408		neigh_update_gc_list(neigh);
1409
1410	if (notify)
1411		neigh_update_notify(neigh, nlmsg_pid);
1412
1413	trace_neigh_update_done(neigh, err);
1414
1415	return err;
1416}
1417
1418int neigh_update(struct neighbour *neigh, const u8 *lladdr, u8 new,
1419		 u32 flags, u32 nlmsg_pid)
1420{
1421	return __neigh_update(neigh, lladdr, new, flags, nlmsg_pid, NULL);
1422}
1423EXPORT_SYMBOL(neigh_update);
1424
1425/* Update the neigh to listen temporarily for probe responses, even if it is
1426 * in a NUD_FAILED state. The caller has to hold neigh->lock for writing.
1427 */
1428void __neigh_set_probe_once(struct neighbour *neigh)
1429{
1430	if (neigh->dead)
1431		return;
1432	neigh->updated = jiffies;
1433	if (!(neigh->nud_state & NUD_FAILED))
1434		return;
1435	neigh->nud_state = NUD_INCOMPLETE;
1436	atomic_set(&neigh->probes, neigh_max_probes(neigh));
1437	neigh_add_timer(neigh,
1438			jiffies + max(NEIGH_VAR(neigh->parms, RETRANS_TIME),
1439				      HZ/100));
1440}
1441EXPORT_SYMBOL(__neigh_set_probe_once);
1442
1443struct neighbour *neigh_event_ns(struct neigh_table *tbl,
1444				 u8 *lladdr, void *saddr,
1445				 struct net_device *dev)
1446{
1447	struct neighbour *neigh = __neigh_lookup(tbl, saddr, dev,
1448						 lladdr || !dev->addr_len);
1449	if (neigh)
1450		neigh_update(neigh, lladdr, NUD_STALE,
1451			     NEIGH_UPDATE_F_OVERRIDE, 0);
1452	return neigh;
1453}
1454EXPORT_SYMBOL(neigh_event_ns);
1455
1456/* called with read_lock_bh(&n->lock); */
1457static void neigh_hh_init(struct neighbour *n)
1458{
1459	struct net_device *dev = n->dev;
1460	__be16 prot = n->tbl->protocol;
1461	struct hh_cache	*hh = &n->hh;
1462
1463	write_lock_bh(&n->lock);
1464
1465	/* Only one thread can come in here and initialize the
1466	 * hh_cache entry.
1467	 */
1468	if (!hh->hh_len)
1469		dev->header_ops->cache(n, hh, prot);
1470
1471	write_unlock_bh(&n->lock);
1472}
1473
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1474/* Slow and careful. */
1475
1476int neigh_resolve_output(struct neighbour *neigh, struct sk_buff *skb)
1477{
 
1478	int rc = 0;
1479
 
 
 
 
 
1480	if (!neigh_event_send(neigh, skb)) {
1481		int err;
1482		struct net_device *dev = neigh->dev;
1483		unsigned int seq;
1484
1485		if (dev->header_ops->cache && !READ_ONCE(neigh->hh.hh_len))
1486			neigh_hh_init(neigh);
1487
1488		do {
1489			__skb_pull(skb, skb_network_offset(skb));
1490			seq = read_seqbegin(&neigh->ha_lock);
1491			err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1492					      neigh->ha, NULL, skb->len);
1493		} while (read_seqretry(&neigh->ha_lock, seq));
1494
1495		if (err >= 0)
1496			rc = dev_queue_xmit(skb);
1497		else
1498			goto out_kfree_skb;
1499	}
1500out:
1501	return rc;
 
 
 
1502out_kfree_skb:
1503	rc = -EINVAL;
1504	kfree_skb(skb);
1505	goto out;
1506}
1507EXPORT_SYMBOL(neigh_resolve_output);
1508
1509/* As fast as possible without hh cache */
1510
1511int neigh_connected_output(struct neighbour *neigh, struct sk_buff *skb)
1512{
1513	struct net_device *dev = neigh->dev;
1514	unsigned int seq;
1515	int err;
1516
 
 
1517	do {
1518		__skb_pull(skb, skb_network_offset(skb));
1519		seq = read_seqbegin(&neigh->ha_lock);
1520		err = dev_hard_header(skb, dev, ntohs(skb->protocol),
1521				      neigh->ha, NULL, skb->len);
1522	} while (read_seqretry(&neigh->ha_lock, seq));
1523
1524	if (err >= 0)
1525		err = dev_queue_xmit(skb);
1526	else {
1527		err = -EINVAL;
1528		kfree_skb(skb);
1529	}
1530	return err;
1531}
1532EXPORT_SYMBOL(neigh_connected_output);
1533
1534int neigh_direct_output(struct neighbour *neigh, struct sk_buff *skb)
1535{
1536	return dev_queue_xmit(skb);
1537}
1538EXPORT_SYMBOL(neigh_direct_output);
1539
1540static void neigh_proxy_process(struct timer_list *t)
1541{
1542	struct neigh_table *tbl = from_timer(tbl, t, proxy_timer);
1543	long sched_next = 0;
1544	unsigned long now = jiffies;
1545	struct sk_buff *skb, *n;
1546
1547	spin_lock(&tbl->proxy_queue.lock);
1548
1549	skb_queue_walk_safe(&tbl->proxy_queue, skb, n) {
1550		long tdif = NEIGH_CB(skb)->sched_next - now;
1551
1552		if (tdif <= 0) {
1553			struct net_device *dev = skb->dev;
1554
1555			__skb_unlink(skb, &tbl->proxy_queue);
1556			if (tbl->proxy_redo && netif_running(dev)) {
1557				rcu_read_lock();
1558				tbl->proxy_redo(skb);
1559				rcu_read_unlock();
1560			} else {
1561				kfree_skb(skb);
1562			}
1563
1564			dev_put(dev);
1565		} else if (!sched_next || tdif < sched_next)
1566			sched_next = tdif;
1567	}
1568	del_timer(&tbl->proxy_timer);
1569	if (sched_next)
1570		mod_timer(&tbl->proxy_timer, jiffies + sched_next);
1571	spin_unlock(&tbl->proxy_queue.lock);
1572}
1573
1574void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
1575		    struct sk_buff *skb)
1576{
1577	unsigned long sched_next = jiffies +
1578			prandom_u32_max(NEIGH_VAR(p, PROXY_DELAY));
1579
1580	if (tbl->proxy_queue.qlen > NEIGH_VAR(p, PROXY_QLEN)) {
1581		kfree_skb(skb);
1582		return;
1583	}
1584
1585	NEIGH_CB(skb)->sched_next = sched_next;
1586	NEIGH_CB(skb)->flags |= LOCALLY_ENQUEUED;
1587
1588	spin_lock(&tbl->proxy_queue.lock);
1589	if (del_timer(&tbl->proxy_timer)) {
1590		if (time_before(tbl->proxy_timer.expires, sched_next))
1591			sched_next = tbl->proxy_timer.expires;
1592	}
1593	skb_dst_drop(skb);
1594	dev_hold(skb->dev);
1595	__skb_queue_tail(&tbl->proxy_queue, skb);
1596	mod_timer(&tbl->proxy_timer, sched_next);
1597	spin_unlock(&tbl->proxy_queue.lock);
1598}
1599EXPORT_SYMBOL(pneigh_enqueue);
1600
1601static inline struct neigh_parms *lookup_neigh_parms(struct neigh_table *tbl,
1602						      struct net *net, int ifindex)
1603{
1604	struct neigh_parms *p;
1605
1606	list_for_each_entry(p, &tbl->parms_list, list) {
1607		if ((p->dev && p->dev->ifindex == ifindex && net_eq(neigh_parms_net(p), net)) ||
1608		    (!p->dev && !ifindex && net_eq(net, &init_net)))
1609			return p;
1610	}
1611
1612	return NULL;
1613}
1614
1615struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
1616				      struct neigh_table *tbl)
1617{
1618	struct neigh_parms *p;
1619	struct net *net = dev_net(dev);
1620	const struct net_device_ops *ops = dev->netdev_ops;
1621
1622	p = kmemdup(&tbl->parms, sizeof(*p), GFP_KERNEL);
 
 
 
 
1623	if (p) {
1624		p->tbl		  = tbl;
1625		refcount_set(&p->refcnt, 1);
1626		p->reachable_time =
1627				neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
1628		dev_hold(dev);
1629		p->dev = dev;
1630		write_pnet(&p->net, net);
1631		p->sysctl_table = NULL;
1632
1633		if (ops->ndo_neigh_setup && ops->ndo_neigh_setup(dev, p)) {
1634			dev_put(dev);
1635			kfree(p);
1636			return NULL;
1637		}
1638
 
 
 
 
1639		write_lock_bh(&tbl->lock);
1640		list_add(&p->list, &tbl->parms.list);
 
1641		write_unlock_bh(&tbl->lock);
1642
1643		neigh_parms_data_state_cleanall(p);
1644	}
1645	return p;
1646}
1647EXPORT_SYMBOL(neigh_parms_alloc);
1648
1649static void neigh_rcu_free_parms(struct rcu_head *head)
1650{
1651	struct neigh_parms *parms =
1652		container_of(head, struct neigh_parms, rcu_head);
1653
1654	neigh_parms_put(parms);
1655}
1656
1657void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
1658{
 
 
1659	if (!parms || parms == &tbl->parms)
1660		return;
1661	write_lock_bh(&tbl->lock);
1662	list_del(&parms->list);
1663	parms->dead = 1;
 
 
 
 
 
 
 
 
 
1664	write_unlock_bh(&tbl->lock);
1665	if (parms->dev)
1666		dev_put(parms->dev);
1667	call_rcu(&parms->rcu_head, neigh_rcu_free_parms);
1668}
1669EXPORT_SYMBOL(neigh_parms_release);
1670
1671static void neigh_parms_destroy(struct neigh_parms *parms)
1672{
 
1673	kfree(parms);
1674}
1675
1676static struct lock_class_key neigh_table_proxy_queue_class;
1677
1678static struct neigh_table *neigh_tables[NEIGH_NR_TABLES] __read_mostly;
1679
1680void neigh_table_init(int index, struct neigh_table *tbl)
1681{
1682	unsigned long now = jiffies;
1683	unsigned long phsize;
1684
1685	INIT_LIST_HEAD(&tbl->parms_list);
1686	INIT_LIST_HEAD(&tbl->gc_list);
1687	list_add(&tbl->parms.list, &tbl->parms_list);
1688	write_pnet(&tbl->parms.net, &init_net);
1689	refcount_set(&tbl->parms.refcnt, 1);
1690	tbl->parms.reachable_time =
1691			  neigh_rand_reach_time(NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME));
1692
 
 
 
 
 
1693	tbl->stats = alloc_percpu(struct neigh_statistics);
1694	if (!tbl->stats)
1695		panic("cannot create neighbour cache statistics");
1696
1697#ifdef CONFIG_PROC_FS
1698	if (!proc_create_seq_data(tbl->id, 0, init_net.proc_net_stat,
1699			      &neigh_stat_seq_ops, tbl))
1700		panic("cannot create neighbour proc dir entry");
1701#endif
1702
1703	RCU_INIT_POINTER(tbl->nht, neigh_hash_alloc(3));
1704
1705	phsize = (PNEIGH_HASHMASK + 1) * sizeof(struct pneigh_entry *);
1706	tbl->phash_buckets = kzalloc(phsize, GFP_KERNEL);
1707
1708	if (!tbl->nht || !tbl->phash_buckets)
1709		panic("cannot allocate neighbour cache hashes");
1710
1711	if (!tbl->entry_size)
1712		tbl->entry_size = ALIGN(offsetof(struct neighbour, primary_key) +
1713					tbl->key_len, NEIGH_PRIV_ALIGN);
1714	else
1715		WARN_ON(tbl->entry_size % NEIGH_PRIV_ALIGN);
1716
1717	rwlock_init(&tbl->lock);
1718	INIT_DEFERRABLE_WORK(&tbl->gc_work, neigh_periodic_work);
1719	queue_delayed_work(system_power_efficient_wq, &tbl->gc_work,
1720			tbl->parms.reachable_time);
1721	timer_setup(&tbl->proxy_timer, neigh_proxy_process, 0);
1722	skb_queue_head_init_class(&tbl->proxy_queue,
1723			&neigh_table_proxy_queue_class);
1724
1725	tbl->last_flush = now;
1726	tbl->last_rand	= now + tbl->parms.reachable_time * 20;
 
 
 
 
 
 
1727
1728	neigh_tables[index] = tbl;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1729}
1730EXPORT_SYMBOL(neigh_table_init);
1731
1732int neigh_table_clear(int index, struct neigh_table *tbl)
1733{
1734	neigh_tables[index] = NULL;
 
1735	/* It is not clean... Fix it to unload IPv6 module safely */
1736	cancel_delayed_work_sync(&tbl->gc_work);
1737	del_timer_sync(&tbl->proxy_timer);
1738	pneigh_queue_purge(&tbl->proxy_queue);
1739	neigh_ifdown(tbl, NULL);
1740	if (atomic_read(&tbl->entries))
1741		pr_crit("neighbour leakage\n");
 
 
 
 
 
 
 
 
1742
1743	call_rcu(&rcu_dereference_protected(tbl->nht, 1)->rcu,
1744		 neigh_hash_free_rcu);
1745	tbl->nht = NULL;
1746
1747	kfree(tbl->phash_buckets);
1748	tbl->phash_buckets = NULL;
1749
1750	remove_proc_entry(tbl->id, init_net.proc_net_stat);
1751
1752	free_percpu(tbl->stats);
1753	tbl->stats = NULL;
1754
 
 
 
1755	return 0;
1756}
1757EXPORT_SYMBOL(neigh_table_clear);
1758
1759static struct neigh_table *neigh_find_table(int family)
1760{
1761	struct neigh_table *tbl = NULL;
1762
1763	switch (family) {
1764	case AF_INET:
1765		tbl = neigh_tables[NEIGH_ARP_TABLE];
1766		break;
1767	case AF_INET6:
1768		tbl = neigh_tables[NEIGH_ND_TABLE];
1769		break;
1770	case AF_DECnet:
1771		tbl = neigh_tables[NEIGH_DN_TABLE];
1772		break;
1773	}
1774
1775	return tbl;
1776}
1777
1778const struct nla_policy nda_policy[NDA_MAX+1] = {
1779	[NDA_UNSPEC]		= { .strict_start_type = NDA_NH_ID },
1780	[NDA_DST]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1781	[NDA_LLADDR]		= { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1782	[NDA_CACHEINFO]		= { .len = sizeof(struct nda_cacheinfo) },
1783	[NDA_PROBES]		= { .type = NLA_U32 },
1784	[NDA_VLAN]		= { .type = NLA_U16 },
1785	[NDA_PORT]		= { .type = NLA_U16 },
1786	[NDA_VNI]		= { .type = NLA_U32 },
1787	[NDA_IFINDEX]		= { .type = NLA_U32 },
1788	[NDA_MASTER]		= { .type = NLA_U32 },
1789	[NDA_PROTOCOL]		= { .type = NLA_U8 },
1790	[NDA_NH_ID]		= { .type = NLA_U32 },
1791	[NDA_FDB_EXT_ATTRS]	= { .type = NLA_NESTED },
1792};
1793
1794static int neigh_delete(struct sk_buff *skb, struct nlmsghdr *nlh,
1795			struct netlink_ext_ack *extack)
1796{
1797	struct net *net = sock_net(skb->sk);
1798	struct ndmsg *ndm;
1799	struct nlattr *dst_attr;
1800	struct neigh_table *tbl;
1801	struct neighbour *neigh;
1802	struct net_device *dev = NULL;
1803	int err = -EINVAL;
1804
1805	ASSERT_RTNL();
1806	if (nlmsg_len(nlh) < sizeof(*ndm))
1807		goto out;
1808
1809	dst_attr = nlmsg_find_attr(nlh, sizeof(*ndm), NDA_DST);
1810	if (!dst_attr) {
1811		NL_SET_ERR_MSG(extack, "Network address not specified");
1812		goto out;
1813	}
1814
1815	ndm = nlmsg_data(nlh);
1816	if (ndm->ndm_ifindex) {
1817		dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1818		if (dev == NULL) {
1819			err = -ENODEV;
1820			goto out;
1821		}
1822	}
1823
1824	tbl = neigh_find_table(ndm->ndm_family);
1825	if (tbl == NULL)
1826		return -EAFNOSUPPORT;
1827
1828	if (nla_len(dst_attr) < (int)tbl->key_len) {
1829		NL_SET_ERR_MSG(extack, "Invalid network address");
1830		goto out;
1831	}
 
 
 
 
 
 
 
1832
1833	if (ndm->ndm_flags & NTF_PROXY) {
1834		err = pneigh_delete(tbl, net, nla_data(dst_attr), dev);
1835		goto out;
1836	}
1837
1838	if (dev == NULL)
1839		goto out;
 
 
 
1840
1841	neigh = neigh_lookup(tbl, nla_data(dst_attr), dev);
1842	if (neigh == NULL) {
1843		err = -ENOENT;
 
1844		goto out;
1845	}
1846
1847	err = __neigh_update(neigh, NULL, NUD_FAILED,
1848			     NEIGH_UPDATE_F_OVERRIDE | NEIGH_UPDATE_F_ADMIN,
1849			     NETLINK_CB(skb).portid, extack);
1850	write_lock_bh(&tbl->lock);
1851	neigh_release(neigh);
1852	neigh_remove_one(neigh, tbl);
1853	write_unlock_bh(&tbl->lock);
1854
1855out:
1856	return err;
1857}
1858
1859static int neigh_add(struct sk_buff *skb, struct nlmsghdr *nlh,
1860		     struct netlink_ext_ack *extack)
1861{
1862	int flags = NEIGH_UPDATE_F_ADMIN | NEIGH_UPDATE_F_OVERRIDE |
1863		NEIGH_UPDATE_F_OVERRIDE_ISROUTER;
1864	struct net *net = sock_net(skb->sk);
1865	struct ndmsg *ndm;
1866	struct nlattr *tb[NDA_MAX+1];
1867	struct neigh_table *tbl;
1868	struct net_device *dev = NULL;
1869	struct neighbour *neigh;
1870	void *dst, *lladdr;
1871	u8 protocol = 0;
1872	int err;
1873
1874	ASSERT_RTNL();
1875	err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
1876				     nda_policy, extack);
1877	if (err < 0)
1878		goto out;
1879
1880	err = -EINVAL;
1881	if (!tb[NDA_DST]) {
1882		NL_SET_ERR_MSG(extack, "Network address not specified");
1883		goto out;
1884	}
1885
1886	ndm = nlmsg_data(nlh);
1887	if (ndm->ndm_ifindex) {
1888		dev = __dev_get_by_index(net, ndm->ndm_ifindex);
1889		if (dev == NULL) {
1890			err = -ENODEV;
1891			goto out;
1892		}
1893
1894		if (tb[NDA_LLADDR] && nla_len(tb[NDA_LLADDR]) < dev->addr_len) {
1895			NL_SET_ERR_MSG(extack, "Invalid link address");
1896			goto out;
1897		}
1898	}
1899
1900	tbl = neigh_find_table(ndm->ndm_family);
1901	if (tbl == NULL)
1902		return -EAFNOSUPPORT;
 
 
1903
1904	if (nla_len(tb[NDA_DST]) < (int)tbl->key_len) {
1905		NL_SET_ERR_MSG(extack, "Invalid network address");
1906		goto out;
1907	}
1908
1909	dst = nla_data(tb[NDA_DST]);
1910	lladdr = tb[NDA_LLADDR] ? nla_data(tb[NDA_LLADDR]) : NULL;
 
 
1911
1912	if (tb[NDA_PROTOCOL])
1913		protocol = nla_get_u8(tb[NDA_PROTOCOL]);
1914
1915	if (ndm->ndm_flags & NTF_PROXY) {
1916		struct pneigh_entry *pn;
1917
1918		err = -ENOBUFS;
1919		pn = pneigh_lookup(tbl, net, dst, dev, 1);
1920		if (pn) {
1921			pn->flags = ndm->ndm_flags;
1922			if (protocol)
1923				pn->protocol = protocol;
1924			err = 0;
1925		}
1926		goto out;
1927	}
1928
1929	if (!dev) {
1930		NL_SET_ERR_MSG(extack, "Device not specified");
1931		goto out;
1932	}
1933
1934	if (tbl->allow_add && !tbl->allow_add(dev, extack)) {
1935		err = -EINVAL;
1936		goto out;
1937	}
 
 
1938
1939	neigh = neigh_lookup(tbl, dst, dev);
1940	if (neigh == NULL) {
1941		bool exempt_from_gc;
 
 
 
 
 
 
 
 
1942
1943		if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1944			err = -ENOENT;
1945			goto out;
1946		}
1947
1948		exempt_from_gc = ndm->ndm_state & NUD_PERMANENT ||
1949				 ndm->ndm_flags & NTF_EXT_LEARNED;
1950		neigh = ___neigh_create(tbl, dst, dev, exempt_from_gc, true);
1951		if (IS_ERR(neigh)) {
1952			err = PTR_ERR(neigh);
1953			goto out;
1954		}
1955	} else {
1956		if (nlh->nlmsg_flags & NLM_F_EXCL) {
1957			err = -EEXIST;
1958			neigh_release(neigh);
1959			goto out;
1960		}
1961
1962		if (!(nlh->nlmsg_flags & NLM_F_REPLACE))
1963			flags &= ~(NEIGH_UPDATE_F_OVERRIDE |
1964				   NEIGH_UPDATE_F_OVERRIDE_ISROUTER);
1965	}
1966
1967	if (protocol)
1968		neigh->protocol = protocol;
1969
1970	if (ndm->ndm_flags & NTF_EXT_LEARNED)
1971		flags |= NEIGH_UPDATE_F_EXT_LEARNED;
1972
1973	if (ndm->ndm_flags & NTF_ROUTER)
1974		flags |= NEIGH_UPDATE_F_ISROUTER;
1975
1976	if (ndm->ndm_flags & NTF_USE) {
1977		neigh_event_send(neigh, NULL);
1978		err = 0;
1979	} else
1980		err = __neigh_update(neigh, lladdr, ndm->ndm_state, flags,
1981				     NETLINK_CB(skb).portid, extack);
1982
1983	neigh_release(neigh);
1984
1985out:
1986	return err;
1987}
1988
1989static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
1990{
1991	struct nlattr *nest;
1992
1993	nest = nla_nest_start_noflag(skb, NDTA_PARMS);
1994	if (nest == NULL)
1995		return -ENOBUFS;
1996
1997	if ((parms->dev &&
1998	     nla_put_u32(skb, NDTPA_IFINDEX, parms->dev->ifindex)) ||
1999	    nla_put_u32(skb, NDTPA_REFCNT, refcount_read(&parms->refcnt)) ||
2000	    nla_put_u32(skb, NDTPA_QUEUE_LENBYTES,
2001			NEIGH_VAR(parms, QUEUE_LEN_BYTES)) ||
2002	    /* approximative value for deprecated QUEUE_LEN (in packets) */
2003	    nla_put_u32(skb, NDTPA_QUEUE_LEN,
2004			NEIGH_VAR(parms, QUEUE_LEN_BYTES) / SKB_TRUESIZE(ETH_FRAME_LEN)) ||
2005	    nla_put_u32(skb, NDTPA_PROXY_QLEN, NEIGH_VAR(parms, PROXY_QLEN)) ||
2006	    nla_put_u32(skb, NDTPA_APP_PROBES, NEIGH_VAR(parms, APP_PROBES)) ||
2007	    nla_put_u32(skb, NDTPA_UCAST_PROBES,
2008			NEIGH_VAR(parms, UCAST_PROBES)) ||
2009	    nla_put_u32(skb, NDTPA_MCAST_PROBES,
2010			NEIGH_VAR(parms, MCAST_PROBES)) ||
2011	    nla_put_u32(skb, NDTPA_MCAST_REPROBES,
2012			NEIGH_VAR(parms, MCAST_REPROBES)) ||
2013	    nla_put_msecs(skb, NDTPA_REACHABLE_TIME, parms->reachable_time,
2014			  NDTPA_PAD) ||
2015	    nla_put_msecs(skb, NDTPA_BASE_REACHABLE_TIME,
2016			  NEIGH_VAR(parms, BASE_REACHABLE_TIME), NDTPA_PAD) ||
2017	    nla_put_msecs(skb, NDTPA_GC_STALETIME,
2018			  NEIGH_VAR(parms, GC_STALETIME), NDTPA_PAD) ||
2019	    nla_put_msecs(skb, NDTPA_DELAY_PROBE_TIME,
2020			  NEIGH_VAR(parms, DELAY_PROBE_TIME), NDTPA_PAD) ||
2021	    nla_put_msecs(skb, NDTPA_RETRANS_TIME,
2022			  NEIGH_VAR(parms, RETRANS_TIME), NDTPA_PAD) ||
2023	    nla_put_msecs(skb, NDTPA_ANYCAST_DELAY,
2024			  NEIGH_VAR(parms, ANYCAST_DELAY), NDTPA_PAD) ||
2025	    nla_put_msecs(skb, NDTPA_PROXY_DELAY,
2026			  NEIGH_VAR(parms, PROXY_DELAY), NDTPA_PAD) ||
2027	    nla_put_msecs(skb, NDTPA_LOCKTIME,
2028			  NEIGH_VAR(parms, LOCKTIME), NDTPA_PAD))
2029		goto nla_put_failure;
2030	return nla_nest_end(skb, nest);
2031
2032nla_put_failure:
2033	nla_nest_cancel(skb, nest);
2034	return -EMSGSIZE;
2035}
2036
2037static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
2038			      u32 pid, u32 seq, int type, int flags)
2039{
2040	struct nlmsghdr *nlh;
2041	struct ndtmsg *ndtmsg;
2042
2043	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
2044	if (nlh == NULL)
2045		return -EMSGSIZE;
2046
2047	ndtmsg = nlmsg_data(nlh);
2048
2049	read_lock_bh(&tbl->lock);
2050	ndtmsg->ndtm_family = tbl->family;
2051	ndtmsg->ndtm_pad1   = 0;
2052	ndtmsg->ndtm_pad2   = 0;
2053
2054	if (nla_put_string(skb, NDTA_NAME, tbl->id) ||
2055	    nla_put_msecs(skb, NDTA_GC_INTERVAL, tbl->gc_interval, NDTA_PAD) ||
2056	    nla_put_u32(skb, NDTA_THRESH1, tbl->gc_thresh1) ||
2057	    nla_put_u32(skb, NDTA_THRESH2, tbl->gc_thresh2) ||
2058	    nla_put_u32(skb, NDTA_THRESH3, tbl->gc_thresh3))
2059		goto nla_put_failure;
2060	{
2061		unsigned long now = jiffies;
2062		long flush_delta = now - tbl->last_flush;
2063		long rand_delta = now - tbl->last_rand;
2064		struct neigh_hash_table *nht;
2065		struct ndt_config ndc = {
2066			.ndtc_key_len		= tbl->key_len,
2067			.ndtc_entry_size	= tbl->entry_size,
2068			.ndtc_entries		= atomic_read(&tbl->entries),
2069			.ndtc_last_flush	= jiffies_to_msecs(flush_delta),
2070			.ndtc_last_rand		= jiffies_to_msecs(rand_delta),
2071			.ndtc_proxy_qlen	= tbl->proxy_queue.qlen,
2072		};
2073
2074		rcu_read_lock_bh();
2075		nht = rcu_dereference_bh(tbl->nht);
2076		ndc.ndtc_hash_rnd = nht->hash_rnd[0];
2077		ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
2078		rcu_read_unlock_bh();
2079
2080		if (nla_put(skb, NDTA_CONFIG, sizeof(ndc), &ndc))
2081			goto nla_put_failure;
2082	}
2083
2084	{
2085		int cpu;
2086		struct ndt_stats ndst;
2087
2088		memset(&ndst, 0, sizeof(ndst));
2089
2090		for_each_possible_cpu(cpu) {
2091			struct neigh_statistics	*st;
2092
2093			st = per_cpu_ptr(tbl->stats, cpu);
2094			ndst.ndts_allocs		+= st->allocs;
2095			ndst.ndts_destroys		+= st->destroys;
2096			ndst.ndts_hash_grows		+= st->hash_grows;
2097			ndst.ndts_res_failed		+= st->res_failed;
2098			ndst.ndts_lookups		+= st->lookups;
2099			ndst.ndts_hits			+= st->hits;
2100			ndst.ndts_rcv_probes_mcast	+= st->rcv_probes_mcast;
2101			ndst.ndts_rcv_probes_ucast	+= st->rcv_probes_ucast;
2102			ndst.ndts_periodic_gc_runs	+= st->periodic_gc_runs;
2103			ndst.ndts_forced_gc_runs	+= st->forced_gc_runs;
2104			ndst.ndts_table_fulls		+= st->table_fulls;
2105		}
2106
2107		if (nla_put_64bit(skb, NDTA_STATS, sizeof(ndst), &ndst,
2108				  NDTA_PAD))
2109			goto nla_put_failure;
2110	}
2111
2112	BUG_ON(tbl->parms.dev);
2113	if (neightbl_fill_parms(skb, &tbl->parms) < 0)
2114		goto nla_put_failure;
2115
2116	read_unlock_bh(&tbl->lock);
2117	nlmsg_end(skb, nlh);
2118	return 0;
2119
2120nla_put_failure:
2121	read_unlock_bh(&tbl->lock);
2122	nlmsg_cancel(skb, nlh);
2123	return -EMSGSIZE;
2124}
2125
2126static int neightbl_fill_param_info(struct sk_buff *skb,
2127				    struct neigh_table *tbl,
2128				    struct neigh_parms *parms,
2129				    u32 pid, u32 seq, int type,
2130				    unsigned int flags)
2131{
2132	struct ndtmsg *ndtmsg;
2133	struct nlmsghdr *nlh;
2134
2135	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndtmsg), flags);
2136	if (nlh == NULL)
2137		return -EMSGSIZE;
2138
2139	ndtmsg = nlmsg_data(nlh);
2140
2141	read_lock_bh(&tbl->lock);
2142	ndtmsg->ndtm_family = tbl->family;
2143	ndtmsg->ndtm_pad1   = 0;
2144	ndtmsg->ndtm_pad2   = 0;
2145
2146	if (nla_put_string(skb, NDTA_NAME, tbl->id) < 0 ||
2147	    neightbl_fill_parms(skb, parms) < 0)
2148		goto errout;
2149
2150	read_unlock_bh(&tbl->lock);
2151	nlmsg_end(skb, nlh);
2152	return 0;
2153errout:
2154	read_unlock_bh(&tbl->lock);
2155	nlmsg_cancel(skb, nlh);
2156	return -EMSGSIZE;
2157}
2158
2159static const struct nla_policy nl_neightbl_policy[NDTA_MAX+1] = {
2160	[NDTA_NAME]		= { .type = NLA_STRING },
2161	[NDTA_THRESH1]		= { .type = NLA_U32 },
2162	[NDTA_THRESH2]		= { .type = NLA_U32 },
2163	[NDTA_THRESH3]		= { .type = NLA_U32 },
2164	[NDTA_GC_INTERVAL]	= { .type = NLA_U64 },
2165	[NDTA_PARMS]		= { .type = NLA_NESTED },
2166};
2167
2168static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = {
2169	[NDTPA_IFINDEX]			= { .type = NLA_U32 },
2170	[NDTPA_QUEUE_LEN]		= { .type = NLA_U32 },
2171	[NDTPA_PROXY_QLEN]		= { .type = NLA_U32 },
2172	[NDTPA_APP_PROBES]		= { .type = NLA_U32 },
2173	[NDTPA_UCAST_PROBES]		= { .type = NLA_U32 },
2174	[NDTPA_MCAST_PROBES]		= { .type = NLA_U32 },
2175	[NDTPA_MCAST_REPROBES]		= { .type = NLA_U32 },
2176	[NDTPA_BASE_REACHABLE_TIME]	= { .type = NLA_U64 },
2177	[NDTPA_GC_STALETIME]		= { .type = NLA_U64 },
2178	[NDTPA_DELAY_PROBE_TIME]	= { .type = NLA_U64 },
2179	[NDTPA_RETRANS_TIME]		= { .type = NLA_U64 },
2180	[NDTPA_ANYCAST_DELAY]		= { .type = NLA_U64 },
2181	[NDTPA_PROXY_DELAY]		= { .type = NLA_U64 },
2182	[NDTPA_LOCKTIME]		= { .type = NLA_U64 },
2183};
2184
2185static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
2186			struct netlink_ext_ack *extack)
2187{
2188	struct net *net = sock_net(skb->sk);
2189	struct neigh_table *tbl;
2190	struct ndtmsg *ndtmsg;
2191	struct nlattr *tb[NDTA_MAX+1];
2192	bool found = false;
2193	int err, tidx;
2194
2195	err = nlmsg_parse_deprecated(nlh, sizeof(*ndtmsg), tb, NDTA_MAX,
2196				     nl_neightbl_policy, extack);
2197	if (err < 0)
2198		goto errout;
2199
2200	if (tb[NDTA_NAME] == NULL) {
2201		err = -EINVAL;
2202		goto errout;
2203	}
2204
2205	ndtmsg = nlmsg_data(nlh);
2206
2207	for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2208		tbl = neigh_tables[tidx];
2209		if (!tbl)
2210			continue;
2211		if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
2212			continue;
2213		if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0) {
2214			found = true;
2215			break;
2216		}
2217	}
2218
2219	if (!found)
2220		return -ENOENT;
 
 
2221
2222	/*
2223	 * We acquire tbl->lock to be nice to the periodic timers and
2224	 * make sure they always see a consistent set of values.
2225	 */
2226	write_lock_bh(&tbl->lock);
2227
2228	if (tb[NDTA_PARMS]) {
2229		struct nlattr *tbp[NDTPA_MAX+1];
2230		struct neigh_parms *p;
2231		int i, ifindex = 0;
2232
2233		err = nla_parse_nested_deprecated(tbp, NDTPA_MAX,
2234						  tb[NDTA_PARMS],
2235						  nl_ntbl_parm_policy, extack);
2236		if (err < 0)
2237			goto errout_tbl_lock;
2238
2239		if (tbp[NDTPA_IFINDEX])
2240			ifindex = nla_get_u32(tbp[NDTPA_IFINDEX]);
2241
2242		p = lookup_neigh_parms(tbl, net, ifindex);
2243		if (p == NULL) {
2244			err = -ENOENT;
2245			goto errout_tbl_lock;
2246		}
2247
2248		for (i = 1; i <= NDTPA_MAX; i++) {
2249			if (tbp[i] == NULL)
2250				continue;
2251
2252			switch (i) {
2253			case NDTPA_QUEUE_LEN:
2254				NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2255					      nla_get_u32(tbp[i]) *
2256					      SKB_TRUESIZE(ETH_FRAME_LEN));
2257				break;
2258			case NDTPA_QUEUE_LENBYTES:
2259				NEIGH_VAR_SET(p, QUEUE_LEN_BYTES,
2260					      nla_get_u32(tbp[i]));
2261				break;
2262			case NDTPA_PROXY_QLEN:
2263				NEIGH_VAR_SET(p, PROXY_QLEN,
2264					      nla_get_u32(tbp[i]));
2265				break;
2266			case NDTPA_APP_PROBES:
2267				NEIGH_VAR_SET(p, APP_PROBES,
2268					      nla_get_u32(tbp[i]));
2269				break;
2270			case NDTPA_UCAST_PROBES:
2271				NEIGH_VAR_SET(p, UCAST_PROBES,
2272					      nla_get_u32(tbp[i]));
2273				break;
2274			case NDTPA_MCAST_PROBES:
2275				NEIGH_VAR_SET(p, MCAST_PROBES,
2276					      nla_get_u32(tbp[i]));
2277				break;
2278			case NDTPA_MCAST_REPROBES:
2279				NEIGH_VAR_SET(p, MCAST_REPROBES,
2280					      nla_get_u32(tbp[i]));
2281				break;
2282			case NDTPA_BASE_REACHABLE_TIME:
2283				NEIGH_VAR_SET(p, BASE_REACHABLE_TIME,
2284					      nla_get_msecs(tbp[i]));
2285				/* update reachable_time as well, otherwise, the change will
2286				 * only be effective after the next time neigh_periodic_work
2287				 * decides to recompute it (can be multiple minutes)
2288				 */
2289				p->reachable_time =
2290					neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
2291				break;
2292			case NDTPA_GC_STALETIME:
2293				NEIGH_VAR_SET(p, GC_STALETIME,
2294					      nla_get_msecs(tbp[i]));
2295				break;
2296			case NDTPA_DELAY_PROBE_TIME:
2297				NEIGH_VAR_SET(p, DELAY_PROBE_TIME,
2298					      nla_get_msecs(tbp[i]));
2299				call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
2300				break;
2301			case NDTPA_RETRANS_TIME:
2302				NEIGH_VAR_SET(p, RETRANS_TIME,
2303					      nla_get_msecs(tbp[i]));
2304				break;
2305			case NDTPA_ANYCAST_DELAY:
2306				NEIGH_VAR_SET(p, ANYCAST_DELAY,
2307					      nla_get_msecs(tbp[i]));
2308				break;
2309			case NDTPA_PROXY_DELAY:
2310				NEIGH_VAR_SET(p, PROXY_DELAY,
2311					      nla_get_msecs(tbp[i]));
2312				break;
2313			case NDTPA_LOCKTIME:
2314				NEIGH_VAR_SET(p, LOCKTIME,
2315					      nla_get_msecs(tbp[i]));
2316				break;
2317			}
2318		}
2319	}
2320
2321	err = -ENOENT;
2322	if ((tb[NDTA_THRESH1] || tb[NDTA_THRESH2] ||
2323	     tb[NDTA_THRESH3] || tb[NDTA_GC_INTERVAL]) &&
2324	    !net_eq(net, &init_net))
2325		goto errout_tbl_lock;
2326
2327	if (tb[NDTA_THRESH1])
2328		tbl->gc_thresh1 = nla_get_u32(tb[NDTA_THRESH1]);
2329
2330	if (tb[NDTA_THRESH2])
2331		tbl->gc_thresh2 = nla_get_u32(tb[NDTA_THRESH2]);
2332
2333	if (tb[NDTA_THRESH3])
2334		tbl->gc_thresh3 = nla_get_u32(tb[NDTA_THRESH3]);
2335
2336	if (tb[NDTA_GC_INTERVAL])
2337		tbl->gc_interval = nla_get_msecs(tb[NDTA_GC_INTERVAL]);
2338
2339	err = 0;
2340
2341errout_tbl_lock:
2342	write_unlock_bh(&tbl->lock);
 
 
2343errout:
2344	return err;
2345}
2346
2347static int neightbl_valid_dump_info(const struct nlmsghdr *nlh,
2348				    struct netlink_ext_ack *extack)
2349{
2350	struct ndtmsg *ndtm;
2351
2352	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndtm))) {
2353		NL_SET_ERR_MSG(extack, "Invalid header for neighbor table dump request");
2354		return -EINVAL;
2355	}
2356
2357	ndtm = nlmsg_data(nlh);
2358	if (ndtm->ndtm_pad1  || ndtm->ndtm_pad2) {
2359		NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor table dump request");
2360		return -EINVAL;
2361	}
2362
2363	if (nlmsg_attrlen(nlh, sizeof(*ndtm))) {
2364		NL_SET_ERR_MSG(extack, "Invalid data after header in neighbor table dump request");
2365		return -EINVAL;
2366	}
2367
2368	return 0;
2369}
2370
2371static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2372{
2373	const struct nlmsghdr *nlh = cb->nlh;
2374	struct net *net = sock_net(skb->sk);
2375	int family, tidx, nidx = 0;
2376	int tbl_skip = cb->args[0];
2377	int neigh_skip = cb->args[1];
2378	struct neigh_table *tbl;
2379
2380	if (cb->strict_check) {
2381		int err = neightbl_valid_dump_info(nlh, cb->extack);
2382
2383		if (err < 0)
2384			return err;
2385	}
2386
2387	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2388
2389	for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
2390		struct neigh_parms *p;
2391
2392		tbl = neigh_tables[tidx];
2393		if (!tbl)
2394			continue;
2395
2396		if (tidx < tbl_skip || (family && tbl->family != family))
2397			continue;
2398
2399		if (neightbl_fill_info(skb, tbl, NETLINK_CB(cb->skb).portid,
2400				       nlh->nlmsg_seq, RTM_NEWNEIGHTBL,
2401				       NLM_F_MULTI) < 0)
2402			break;
2403
2404		nidx = 0;
2405		p = list_next_entry(&tbl->parms, list);
2406		list_for_each_entry_from(p, &tbl->parms_list, list) {
2407			if (!net_eq(neigh_parms_net(p), net))
2408				continue;
2409
2410			if (nidx < neigh_skip)
2411				goto next;
2412
2413			if (neightbl_fill_param_info(skb, tbl, p,
2414						     NETLINK_CB(cb->skb).portid,
2415						     nlh->nlmsg_seq,
2416						     RTM_NEWNEIGHTBL,
2417						     NLM_F_MULTI) < 0)
2418				goto out;
2419		next:
2420			nidx++;
2421		}
2422
2423		neigh_skip = 0;
2424	}
2425out:
 
2426	cb->args[0] = tidx;
2427	cb->args[1] = nidx;
2428
2429	return skb->len;
2430}
2431
2432static int neigh_fill_info(struct sk_buff *skb, struct neighbour *neigh,
2433			   u32 pid, u32 seq, int type, unsigned int flags)
2434{
2435	unsigned long now = jiffies;
2436	struct nda_cacheinfo ci;
2437	struct nlmsghdr *nlh;
2438	struct ndmsg *ndm;
2439
2440	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2441	if (nlh == NULL)
2442		return -EMSGSIZE;
2443
2444	ndm = nlmsg_data(nlh);
2445	ndm->ndm_family	 = neigh->ops->family;
2446	ndm->ndm_pad1    = 0;
2447	ndm->ndm_pad2    = 0;
2448	ndm->ndm_flags	 = neigh->flags;
2449	ndm->ndm_type	 = neigh->type;
2450	ndm->ndm_ifindex = neigh->dev->ifindex;
2451
2452	if (nla_put(skb, NDA_DST, neigh->tbl->key_len, neigh->primary_key))
2453		goto nla_put_failure;
2454
2455	read_lock_bh(&neigh->lock);
2456	ndm->ndm_state	 = neigh->nud_state;
2457	if (neigh->nud_state & NUD_VALID) {
2458		char haddr[MAX_ADDR_LEN];
2459
2460		neigh_ha_snapshot(haddr, neigh, neigh->dev);
2461		if (nla_put(skb, NDA_LLADDR, neigh->dev->addr_len, haddr) < 0) {
2462			read_unlock_bh(&neigh->lock);
2463			goto nla_put_failure;
2464		}
2465	}
2466
2467	ci.ndm_used	 = jiffies_to_clock_t(now - neigh->used);
2468	ci.ndm_confirmed = jiffies_to_clock_t(now - neigh->confirmed);
2469	ci.ndm_updated	 = jiffies_to_clock_t(now - neigh->updated);
2470	ci.ndm_refcnt	 = refcount_read(&neigh->refcnt) - 1;
2471	read_unlock_bh(&neigh->lock);
2472
2473	if (nla_put_u32(skb, NDA_PROBES, atomic_read(&neigh->probes)) ||
2474	    nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
2475		goto nla_put_failure;
2476
2477	if (neigh->protocol && nla_put_u8(skb, NDA_PROTOCOL, neigh->protocol))
2478		goto nla_put_failure;
2479
2480	nlmsg_end(skb, nlh);
2481	return 0;
2482
2483nla_put_failure:
2484	nlmsg_cancel(skb, nlh);
2485	return -EMSGSIZE;
2486}
2487
2488static int pneigh_fill_info(struct sk_buff *skb, struct pneigh_entry *pn,
2489			    u32 pid, u32 seq, int type, unsigned int flags,
2490			    struct neigh_table *tbl)
2491{
2492	struct nlmsghdr *nlh;
2493	struct ndmsg *ndm;
2494
2495	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), flags);
2496	if (nlh == NULL)
2497		return -EMSGSIZE;
2498
2499	ndm = nlmsg_data(nlh);
2500	ndm->ndm_family	 = tbl->family;
2501	ndm->ndm_pad1    = 0;
2502	ndm->ndm_pad2    = 0;
2503	ndm->ndm_flags	 = pn->flags | NTF_PROXY;
2504	ndm->ndm_type	 = RTN_UNICAST;
2505	ndm->ndm_ifindex = pn->dev ? pn->dev->ifindex : 0;
2506	ndm->ndm_state	 = NUD_NONE;
2507
2508	if (nla_put(skb, NDA_DST, tbl->key_len, pn->key))
2509		goto nla_put_failure;
2510
2511	if (pn->protocol && nla_put_u8(skb, NDA_PROTOCOL, pn->protocol))
2512		goto nla_put_failure;
2513
2514	nlmsg_end(skb, nlh);
2515	return 0;
2516
2517nla_put_failure:
2518	nlmsg_cancel(skb, nlh);
2519	return -EMSGSIZE;
2520}
2521
2522static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid)
2523{
2524	call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, neigh);
2525	__neigh_notify(neigh, RTM_NEWNEIGH, 0, nlmsg_pid);
2526}
2527
2528static bool neigh_master_filtered(struct net_device *dev, int master_idx)
2529{
2530	struct net_device *master;
2531
2532	if (!master_idx)
2533		return false;
2534
2535	master = dev ? netdev_master_upper_dev_get(dev) : NULL;
2536	if (!master || master->ifindex != master_idx)
2537		return true;
2538
2539	return false;
2540}
2541
2542static bool neigh_ifindex_filtered(struct net_device *dev, int filter_idx)
2543{
2544	if (filter_idx && (!dev || dev->ifindex != filter_idx))
2545		return true;
2546
2547	return false;
2548}
2549
2550struct neigh_dump_filter {
2551	int master_idx;
2552	int dev_idx;
2553};
2554
2555static int neigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2556			    struct netlink_callback *cb,
2557			    struct neigh_dump_filter *filter)
2558{
2559	struct net *net = sock_net(skb->sk);
2560	struct neighbour *n;
2561	int rc, h, s_h = cb->args[1];
2562	int idx, s_idx = idx = cb->args[2];
2563	struct neigh_hash_table *nht;
2564	unsigned int flags = NLM_F_MULTI;
2565
2566	if (filter->dev_idx || filter->master_idx)
2567		flags |= NLM_F_DUMP_FILTERED;
2568
2569	rcu_read_lock_bh();
2570	nht = rcu_dereference_bh(tbl->nht);
2571
2572	for (h = s_h; h < (1 << nht->hash_shift); h++) {
 
 
2573		if (h > s_h)
2574			s_idx = 0;
2575		for (n = rcu_dereference_bh(nht->hash_buckets[h]), idx = 0;
2576		     n != NULL;
2577		     n = rcu_dereference_bh(n->next)) {
2578			if (idx < s_idx || !net_eq(dev_net(n->dev), net))
2579				goto next;
2580			if (neigh_ifindex_filtered(n->dev, filter->dev_idx) ||
2581			    neigh_master_filtered(n->dev, filter->master_idx))
2582				goto next;
2583			if (neigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2584					    cb->nlh->nlmsg_seq,
2585					    RTM_NEWNEIGH,
2586					    flags) < 0) {
2587				rc = -1;
2588				goto out;
2589			}
2590next:
2591			idx++;
2592		}
2593	}
2594	rc = skb->len;
2595out:
2596	rcu_read_unlock_bh();
2597	cb->args[1] = h;
2598	cb->args[2] = idx;
2599	return rc;
2600}
2601
2602static int pneigh_dump_table(struct neigh_table *tbl, struct sk_buff *skb,
2603			     struct netlink_callback *cb,
2604			     struct neigh_dump_filter *filter)
2605{
2606	struct pneigh_entry *n;
2607	struct net *net = sock_net(skb->sk);
2608	int rc, h, s_h = cb->args[3];
2609	int idx, s_idx = idx = cb->args[4];
2610	unsigned int flags = NLM_F_MULTI;
2611
2612	if (filter->dev_idx || filter->master_idx)
2613		flags |= NLM_F_DUMP_FILTERED;
2614
2615	read_lock_bh(&tbl->lock);
2616
2617	for (h = s_h; h <= PNEIGH_HASHMASK; h++) {
2618		if (h > s_h)
2619			s_idx = 0;
2620		for (n = tbl->phash_buckets[h], idx = 0; n; n = n->next) {
2621			if (idx < s_idx || pneigh_net(n) != net)
2622				goto next;
2623			if (neigh_ifindex_filtered(n->dev, filter->dev_idx) ||
2624			    neigh_master_filtered(n->dev, filter->master_idx))
2625				goto next;
2626			if (pneigh_fill_info(skb, n, NETLINK_CB(cb->skb).portid,
2627					    cb->nlh->nlmsg_seq,
2628					    RTM_NEWNEIGH, flags, tbl) < 0) {
2629				read_unlock_bh(&tbl->lock);
2630				rc = -1;
2631				goto out;
2632			}
2633		next:
2634			idx++;
2635		}
2636	}
2637
2638	read_unlock_bh(&tbl->lock);
2639	rc = skb->len;
2640out:
2641	cb->args[3] = h;
2642	cb->args[4] = idx;
2643	return rc;
2644
2645}
2646
2647static int neigh_valid_dump_req(const struct nlmsghdr *nlh,
2648				bool strict_check,
2649				struct neigh_dump_filter *filter,
2650				struct netlink_ext_ack *extack)
2651{
2652	struct nlattr *tb[NDA_MAX + 1];
2653	int err, i;
2654
2655	if (strict_check) {
2656		struct ndmsg *ndm;
2657
2658		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
2659			NL_SET_ERR_MSG(extack, "Invalid header for neighbor dump request");
2660			return -EINVAL;
2661		}
2662
2663		ndm = nlmsg_data(nlh);
2664		if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_ifindex ||
2665		    ndm->ndm_state || ndm->ndm_type) {
2666			NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor dump request");
2667			return -EINVAL;
2668		}
2669
2670		if (ndm->ndm_flags & ~NTF_PROXY) {
2671			NL_SET_ERR_MSG(extack, "Invalid flags in header for neighbor dump request");
2672			return -EINVAL;
2673		}
2674
2675		err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg),
2676						    tb, NDA_MAX, nda_policy,
2677						    extack);
2678	} else {
2679		err = nlmsg_parse_deprecated(nlh, sizeof(struct ndmsg), tb,
2680					     NDA_MAX, nda_policy, extack);
2681	}
2682	if (err < 0)
2683		return err;
2684
2685	for (i = 0; i <= NDA_MAX; ++i) {
2686		if (!tb[i])
2687			continue;
2688
2689		/* all new attributes should require strict_check */
2690		switch (i) {
2691		case NDA_IFINDEX:
2692			filter->dev_idx = nla_get_u32(tb[i]);
2693			break;
2694		case NDA_MASTER:
2695			filter->master_idx = nla_get_u32(tb[i]);
2696			break;
2697		default:
2698			if (strict_check) {
2699				NL_SET_ERR_MSG(extack, "Unsupported attribute in neighbor dump request");
2700				return -EINVAL;
2701			}
2702		}
2703	}
2704
2705	return 0;
2706}
2707
2708static int neigh_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
2709{
2710	const struct nlmsghdr *nlh = cb->nlh;
2711	struct neigh_dump_filter filter = {};
2712	struct neigh_table *tbl;
2713	int t, family, s_t;
2714	int proxy = 0;
2715	int err;
2716
2717	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2718
2719	/* check for full ndmsg structure presence, family member is
2720	 * the same for both structures
2721	 */
2722	if (nlmsg_len(nlh) >= sizeof(struct ndmsg) &&
2723	    ((struct ndmsg *)nlmsg_data(nlh))->ndm_flags == NTF_PROXY)
2724		proxy = 1;
2725
2726	err = neigh_valid_dump_req(nlh, cb->strict_check, &filter, cb->extack);
2727	if (err < 0 && cb->strict_check)
2728		return err;
2729
 
 
2730	s_t = cb->args[0];
2731
2732	for (t = 0; t < NEIGH_NR_TABLES; t++) {
2733		tbl = neigh_tables[t];
2734
2735		if (!tbl)
2736			continue;
2737		if (t < s_t || (family && tbl->family != family))
2738			continue;
2739		if (t > s_t)
2740			memset(&cb->args[1], 0, sizeof(cb->args) -
2741						sizeof(cb->args[0]));
2742		if (proxy)
2743			err = pneigh_dump_table(tbl, skb, cb, &filter);
2744		else
2745			err = neigh_dump_table(tbl, skb, cb, &filter);
2746		if (err < 0)
2747			break;
2748	}
 
2749
2750	cb->args[0] = t;
2751	return skb->len;
2752}
2753
2754static int neigh_valid_get_req(const struct nlmsghdr *nlh,
2755			       struct neigh_table **tbl,
2756			       void **dst, int *dev_idx, u8 *ndm_flags,
2757			       struct netlink_ext_ack *extack)
2758{
2759	struct nlattr *tb[NDA_MAX + 1];
2760	struct ndmsg *ndm;
2761	int err, i;
2762
2763	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
2764		NL_SET_ERR_MSG(extack, "Invalid header for neighbor get request");
2765		return -EINVAL;
2766	}
2767
2768	ndm = nlmsg_data(nlh);
2769	if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
2770	    ndm->ndm_type) {
2771		NL_SET_ERR_MSG(extack, "Invalid values in header for neighbor get request");
2772		return -EINVAL;
2773	}
2774
2775	if (ndm->ndm_flags & ~NTF_PROXY) {
2776		NL_SET_ERR_MSG(extack, "Invalid flags in header for neighbor get request");
2777		return -EINVAL;
2778	}
2779
2780	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
2781					    NDA_MAX, nda_policy, extack);
2782	if (err < 0)
2783		return err;
2784
2785	*ndm_flags = ndm->ndm_flags;
2786	*dev_idx = ndm->ndm_ifindex;
2787	*tbl = neigh_find_table(ndm->ndm_family);
2788	if (*tbl == NULL) {
2789		NL_SET_ERR_MSG(extack, "Unsupported family in header for neighbor get request");
2790		return -EAFNOSUPPORT;
2791	}
2792
2793	for (i = 0; i <= NDA_MAX; ++i) {
2794		if (!tb[i])
2795			continue;
2796
2797		switch (i) {
2798		case NDA_DST:
2799			if (nla_len(tb[i]) != (int)(*tbl)->key_len) {
2800				NL_SET_ERR_MSG(extack, "Invalid network address in neighbor get request");
2801				return -EINVAL;
2802			}
2803			*dst = nla_data(tb[i]);
2804			break;
2805		default:
2806			NL_SET_ERR_MSG(extack, "Unsupported attribute in neighbor get request");
2807			return -EINVAL;
2808		}
2809	}
2810
2811	return 0;
2812}
2813
2814static inline size_t neigh_nlmsg_size(void)
2815{
2816	return NLMSG_ALIGN(sizeof(struct ndmsg))
2817	       + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2818	       + nla_total_size(MAX_ADDR_LEN) /* NDA_LLADDR */
2819	       + nla_total_size(sizeof(struct nda_cacheinfo))
2820	       + nla_total_size(4)  /* NDA_PROBES */
2821	       + nla_total_size(1); /* NDA_PROTOCOL */
2822}
2823
2824static int neigh_get_reply(struct net *net, struct neighbour *neigh,
2825			   u32 pid, u32 seq)
2826{
2827	struct sk_buff *skb;
2828	int err = 0;
2829
2830	skb = nlmsg_new(neigh_nlmsg_size(), GFP_KERNEL);
2831	if (!skb)
2832		return -ENOBUFS;
2833
2834	err = neigh_fill_info(skb, neigh, pid, seq, RTM_NEWNEIGH, 0);
2835	if (err) {
2836		kfree_skb(skb);
2837		goto errout;
2838	}
2839
2840	err = rtnl_unicast(skb, net, pid);
2841errout:
2842	return err;
2843}
2844
2845static inline size_t pneigh_nlmsg_size(void)
2846{
2847	return NLMSG_ALIGN(sizeof(struct ndmsg))
2848	       + nla_total_size(MAX_ADDR_LEN) /* NDA_DST */
2849	       + nla_total_size(1); /* NDA_PROTOCOL */
2850}
2851
2852static int pneigh_get_reply(struct net *net, struct pneigh_entry *neigh,
2853			    u32 pid, u32 seq, struct neigh_table *tbl)
2854{
2855	struct sk_buff *skb;
2856	int err = 0;
2857
2858	skb = nlmsg_new(pneigh_nlmsg_size(), GFP_KERNEL);
2859	if (!skb)
2860		return -ENOBUFS;
2861
2862	err = pneigh_fill_info(skb, neigh, pid, seq, RTM_NEWNEIGH, 0, tbl);
2863	if (err) {
2864		kfree_skb(skb);
2865		goto errout;
2866	}
2867
2868	err = rtnl_unicast(skb, net, pid);
2869errout:
2870	return err;
2871}
2872
2873static int neigh_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
2874		     struct netlink_ext_ack *extack)
2875{
2876	struct net *net = sock_net(in_skb->sk);
2877	struct net_device *dev = NULL;
2878	struct neigh_table *tbl = NULL;
2879	struct neighbour *neigh;
2880	void *dst = NULL;
2881	u8 ndm_flags = 0;
2882	int dev_idx = 0;
2883	int err;
2884
2885	err = neigh_valid_get_req(nlh, &tbl, &dst, &dev_idx, &ndm_flags,
2886				  extack);
2887	if (err < 0)
2888		return err;
2889
2890	if (dev_idx) {
2891		dev = __dev_get_by_index(net, dev_idx);
2892		if (!dev) {
2893			NL_SET_ERR_MSG(extack, "Unknown device ifindex");
2894			return -ENODEV;
2895		}
2896	}
2897
2898	if (!dst) {
2899		NL_SET_ERR_MSG(extack, "Network address not specified");
2900		return -EINVAL;
2901	}
2902
2903	if (ndm_flags & NTF_PROXY) {
2904		struct pneigh_entry *pn;
2905
2906		pn = pneigh_lookup(tbl, net, dst, dev, 0);
2907		if (!pn) {
2908			NL_SET_ERR_MSG(extack, "Proxy neighbour entry not found");
2909			return -ENOENT;
2910		}
2911		return pneigh_get_reply(net, pn, NETLINK_CB(in_skb).portid,
2912					nlh->nlmsg_seq, tbl);
2913	}
2914
2915	if (!dev) {
2916		NL_SET_ERR_MSG(extack, "No device specified");
2917		return -EINVAL;
2918	}
2919
2920	neigh = neigh_lookup(tbl, dst, dev);
2921	if (!neigh) {
2922		NL_SET_ERR_MSG(extack, "Neighbour entry not found");
2923		return -ENOENT;
2924	}
2925
2926	err = neigh_get_reply(net, neigh, NETLINK_CB(in_skb).portid,
2927			      nlh->nlmsg_seq);
2928
2929	neigh_release(neigh);
2930
2931	return err;
2932}
2933
2934void neigh_for_each(struct neigh_table *tbl, void (*cb)(struct neighbour *, void *), void *cookie)
2935{
2936	int chain;
2937	struct neigh_hash_table *nht;
2938
2939	rcu_read_lock_bh();
2940	nht = rcu_dereference_bh(tbl->nht);
2941
2942	read_lock(&tbl->lock); /* avoid resizes */
2943	for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2944		struct neighbour *n;
2945
2946		for (n = rcu_dereference_bh(nht->hash_buckets[chain]);
2947		     n != NULL;
2948		     n = rcu_dereference_bh(n->next))
2949			cb(n, cookie);
2950	}
2951	read_unlock(&tbl->lock);
2952	rcu_read_unlock_bh();
2953}
2954EXPORT_SYMBOL(neigh_for_each);
2955
2956/* The tbl->lock must be held as a writer and BH disabled. */
2957void __neigh_for_each_release(struct neigh_table *tbl,
2958			      int (*cb)(struct neighbour *))
2959{
2960	int chain;
2961	struct neigh_hash_table *nht;
2962
2963	nht = rcu_dereference_protected(tbl->nht,
2964					lockdep_is_held(&tbl->lock));
2965	for (chain = 0; chain < (1 << nht->hash_shift); chain++) {
2966		struct neighbour *n;
2967		struct neighbour __rcu **np;
2968
2969		np = &nht->hash_buckets[chain];
2970		while ((n = rcu_dereference_protected(*np,
2971					lockdep_is_held(&tbl->lock))) != NULL) {
2972			int release;
2973
2974			write_lock(&n->lock);
2975			release = cb(n);
2976			if (release) {
2977				rcu_assign_pointer(*np,
2978					rcu_dereference_protected(n->next,
2979						lockdep_is_held(&tbl->lock)));
2980				neigh_mark_dead(n);
2981			} else
2982				np = &n->next;
2983			write_unlock(&n->lock);
2984			if (release)
2985				neigh_cleanup_and_release(n);
2986		}
2987	}
2988}
2989EXPORT_SYMBOL(__neigh_for_each_release);
2990
2991int neigh_xmit(int index, struct net_device *dev,
2992	       const void *addr, struct sk_buff *skb)
2993{
2994	int err = -EAFNOSUPPORT;
2995	if (likely(index < NEIGH_NR_TABLES)) {
2996		struct neigh_table *tbl;
2997		struct neighbour *neigh;
2998
2999		tbl = neigh_tables[index];
3000		if (!tbl)
3001			goto out;
3002		rcu_read_lock_bh();
3003		if (index == NEIGH_ARP_TABLE) {
3004			u32 key = *((u32 *)addr);
3005
3006			neigh = __ipv4_neigh_lookup_noref(dev, key);
3007		} else {
3008			neigh = __neigh_lookup_noref(tbl, addr, dev);
3009		}
3010		if (!neigh)
3011			neigh = __neigh_create(tbl, addr, dev, false);
3012		err = PTR_ERR(neigh);
3013		if (IS_ERR(neigh)) {
3014			rcu_read_unlock_bh();
3015			goto out_kfree_skb;
3016		}
3017		err = neigh->output(neigh, skb);
3018		rcu_read_unlock_bh();
3019	}
3020	else if (index == NEIGH_LINK_TABLE) {
3021		err = dev_hard_header(skb, dev, ntohs(skb->protocol),
3022				      addr, NULL, skb->len);
3023		if (err < 0)
3024			goto out_kfree_skb;
3025		err = dev_queue_xmit(skb);
3026	}
3027out:
3028	return err;
3029out_kfree_skb:
3030	kfree_skb(skb);
3031	goto out;
3032}
3033EXPORT_SYMBOL(neigh_xmit);
3034
3035#ifdef CONFIG_PROC_FS
3036
3037static struct neighbour *neigh_get_first(struct seq_file *seq)
3038{
3039	struct neigh_seq_state *state = seq->private;
3040	struct net *net = seq_file_net(seq);
3041	struct neigh_hash_table *nht = state->nht;
3042	struct neighbour *n = NULL;
3043	int bucket;
3044
3045	state->flags &= ~NEIGH_SEQ_IS_PNEIGH;
3046	for (bucket = 0; bucket < (1 << nht->hash_shift); bucket++) {
3047		n = rcu_dereference_bh(nht->hash_buckets[bucket]);
3048
3049		while (n) {
3050			if (!net_eq(dev_net(n->dev), net))
3051				goto next;
3052			if (state->neigh_sub_iter) {
3053				loff_t fakep = 0;
3054				void *v;
3055
3056				v = state->neigh_sub_iter(state, n, &fakep);
3057				if (!v)
3058					goto next;
3059			}
3060			if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
3061				break;
3062			if (n->nud_state & ~NUD_NOARP)
3063				break;
3064next:
3065			n = rcu_dereference_bh(n->next);
3066		}
3067
3068		if (n)
3069			break;
3070	}
3071	state->bucket = bucket;
3072
3073	return n;
3074}
3075
3076static struct neighbour *neigh_get_next(struct seq_file *seq,
3077					struct neighbour *n,
3078					loff_t *pos)
3079{
3080	struct neigh_seq_state *state = seq->private;
3081	struct net *net = seq_file_net(seq);
3082	struct neigh_hash_table *nht = state->nht;
3083
3084	if (state->neigh_sub_iter) {
3085		void *v = state->neigh_sub_iter(state, n, pos);
3086		if (v)
3087			return n;
3088	}
3089	n = rcu_dereference_bh(n->next);
3090
3091	while (1) {
3092		while (n) {
3093			if (!net_eq(dev_net(n->dev), net))
3094				goto next;
3095			if (state->neigh_sub_iter) {
3096				void *v = state->neigh_sub_iter(state, n, pos);
3097				if (v)
3098					return n;
3099				goto next;
3100			}
3101			if (!(state->flags & NEIGH_SEQ_SKIP_NOARP))
3102				break;
3103
3104			if (n->nud_state & ~NUD_NOARP)
3105				break;
3106next:
3107			n = rcu_dereference_bh(n->next);
3108		}
3109
3110		if (n)
3111			break;
3112
3113		if (++state->bucket >= (1 << nht->hash_shift))
3114			break;
3115
3116		n = rcu_dereference_bh(nht->hash_buckets[state->bucket]);
3117	}
3118
3119	if (n && pos)
3120		--(*pos);
3121	return n;
3122}
3123
3124static struct neighbour *neigh_get_idx(struct seq_file *seq, loff_t *pos)
3125{
3126	struct neighbour *n = neigh_get_first(seq);
3127
3128	if (n) {
3129		--(*pos);
3130		while (*pos) {
3131			n = neigh_get_next(seq, n, pos);
3132			if (!n)
3133				break;
3134		}
3135	}
3136	return *pos ? NULL : n;
3137}
3138
3139static struct pneigh_entry *pneigh_get_first(struct seq_file *seq)
3140{
3141	struct neigh_seq_state *state = seq->private;
3142	struct net *net = seq_file_net(seq);
3143	struct neigh_table *tbl = state->tbl;
3144	struct pneigh_entry *pn = NULL;
3145	int bucket;
3146
3147	state->flags |= NEIGH_SEQ_IS_PNEIGH;
3148	for (bucket = 0; bucket <= PNEIGH_HASHMASK; bucket++) {
3149		pn = tbl->phash_buckets[bucket];
3150		while (pn && !net_eq(pneigh_net(pn), net))
3151			pn = pn->next;
3152		if (pn)
3153			break;
3154	}
3155	state->bucket = bucket;
3156
3157	return pn;
3158}
3159
3160static struct pneigh_entry *pneigh_get_next(struct seq_file *seq,
3161					    struct pneigh_entry *pn,
3162					    loff_t *pos)
3163{
3164	struct neigh_seq_state *state = seq->private;
3165	struct net *net = seq_file_net(seq);
3166	struct neigh_table *tbl = state->tbl;
3167
3168	do {
3169		pn = pn->next;
3170	} while (pn && !net_eq(pneigh_net(pn), net));
3171
3172	while (!pn) {
3173		if (++state->bucket > PNEIGH_HASHMASK)
3174			break;
3175		pn = tbl->phash_buckets[state->bucket];
3176		while (pn && !net_eq(pneigh_net(pn), net))
3177			pn = pn->next;
3178		if (pn)
3179			break;
3180	}
3181
3182	if (pn && pos)
3183		--(*pos);
3184
3185	return pn;
3186}
3187
3188static struct pneigh_entry *pneigh_get_idx(struct seq_file *seq, loff_t *pos)
3189{
3190	struct pneigh_entry *pn = pneigh_get_first(seq);
3191
3192	if (pn) {
3193		--(*pos);
3194		while (*pos) {
3195			pn = pneigh_get_next(seq, pn, pos);
3196			if (!pn)
3197				break;
3198		}
3199	}
3200	return *pos ? NULL : pn;
3201}
3202
3203static void *neigh_get_idx_any(struct seq_file *seq, loff_t *pos)
3204{
3205	struct neigh_seq_state *state = seq->private;
3206	void *rc;
3207	loff_t idxpos = *pos;
3208
3209	rc = neigh_get_idx(seq, &idxpos);
3210	if (!rc && !(state->flags & NEIGH_SEQ_NEIGH_ONLY))
3211		rc = pneigh_get_idx(seq, &idxpos);
3212
3213	return rc;
3214}
3215
3216void *neigh_seq_start(struct seq_file *seq, loff_t *pos, struct neigh_table *tbl, unsigned int neigh_seq_flags)
3217	__acquires(tbl->lock)
3218	__acquires(rcu_bh)
3219{
3220	struct neigh_seq_state *state = seq->private;
3221
3222	state->tbl = tbl;
3223	state->bucket = 0;
3224	state->flags = (neigh_seq_flags & ~NEIGH_SEQ_IS_PNEIGH);
3225
3226	rcu_read_lock_bh();
3227	state->nht = rcu_dereference_bh(tbl->nht);
3228	read_lock(&tbl->lock);
3229
3230	return *pos ? neigh_get_idx_any(seq, pos) : SEQ_START_TOKEN;
3231}
3232EXPORT_SYMBOL(neigh_seq_start);
3233
3234void *neigh_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3235{
3236	struct neigh_seq_state *state;
3237	void *rc;
3238
3239	if (v == SEQ_START_TOKEN) {
3240		rc = neigh_get_first(seq);
3241		goto out;
3242	}
3243
3244	state = seq->private;
3245	if (!(state->flags & NEIGH_SEQ_IS_PNEIGH)) {
3246		rc = neigh_get_next(seq, v, NULL);
3247		if (rc)
3248			goto out;
3249		if (!(state->flags & NEIGH_SEQ_NEIGH_ONLY))
3250			rc = pneigh_get_first(seq);
3251	} else {
3252		BUG_ON(state->flags & NEIGH_SEQ_NEIGH_ONLY);
3253		rc = pneigh_get_next(seq, v, NULL);
3254	}
3255out:
3256	++(*pos);
3257	return rc;
3258}
3259EXPORT_SYMBOL(neigh_seq_next);
3260
3261void neigh_seq_stop(struct seq_file *seq, void *v)
3262	__releases(tbl->lock)
3263	__releases(rcu_bh)
3264{
3265	struct neigh_seq_state *state = seq->private;
3266	struct neigh_table *tbl = state->tbl;
3267
3268	read_unlock(&tbl->lock);
3269	rcu_read_unlock_bh();
3270}
3271EXPORT_SYMBOL(neigh_seq_stop);
3272
3273/* statistics via seq_file */
3274
3275static void *neigh_stat_seq_start(struct seq_file *seq, loff_t *pos)
3276{
3277	struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3278	int cpu;
3279
3280	if (*pos == 0)
3281		return SEQ_START_TOKEN;
3282
3283	for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
3284		if (!cpu_possible(cpu))
3285			continue;
3286		*pos = cpu+1;
3287		return per_cpu_ptr(tbl->stats, cpu);
3288	}
3289	return NULL;
3290}
3291
3292static void *neigh_stat_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3293{
3294	struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3295	int cpu;
3296
3297	for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
3298		if (!cpu_possible(cpu))
3299			continue;
3300		*pos = cpu+1;
3301		return per_cpu_ptr(tbl->stats, cpu);
3302	}
3303	(*pos)++;
3304	return NULL;
3305}
3306
3307static void neigh_stat_seq_stop(struct seq_file *seq, void *v)
3308{
3309
3310}
3311
3312static int neigh_stat_seq_show(struct seq_file *seq, void *v)
3313{
3314	struct neigh_table *tbl = PDE_DATA(file_inode(seq->file));
3315	struct neigh_statistics *st = v;
3316
3317	if (v == SEQ_START_TOKEN) {
3318		seq_printf(seq, "entries  allocs destroys hash_grows  lookups hits  res_failed  rcv_probes_mcast rcv_probes_ucast  periodic_gc_runs forced_gc_runs unresolved_discards table_fulls\n");
3319		return 0;
3320	}
3321
3322	seq_printf(seq, "%08x  %08lx %08lx %08lx  %08lx %08lx  %08lx  "
3323			"%08lx %08lx  %08lx %08lx %08lx %08lx\n",
3324		   atomic_read(&tbl->entries),
3325
3326		   st->allocs,
3327		   st->destroys,
3328		   st->hash_grows,
3329
3330		   st->lookups,
3331		   st->hits,
3332
3333		   st->res_failed,
3334
3335		   st->rcv_probes_mcast,
3336		   st->rcv_probes_ucast,
3337
3338		   st->periodic_gc_runs,
3339		   st->forced_gc_runs,
3340		   st->unres_discards,
3341		   st->table_fulls
3342		   );
3343
3344	return 0;
3345}
3346
3347static const struct seq_operations neigh_stat_seq_ops = {
3348	.start	= neigh_stat_seq_start,
3349	.next	= neigh_stat_seq_next,
3350	.stop	= neigh_stat_seq_stop,
3351	.show	= neigh_stat_seq_show,
3352};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3353#endif /* CONFIG_PROC_FS */
3354
3355static void __neigh_notify(struct neighbour *n, int type, int flags,
3356			   u32 pid)
 
 
 
 
 
 
 
 
3357{
3358	struct net *net = dev_net(n->dev);
3359	struct sk_buff *skb;
3360	int err = -ENOBUFS;
3361
3362	skb = nlmsg_new(neigh_nlmsg_size(), GFP_ATOMIC);
3363	if (skb == NULL)
3364		goto errout;
3365
3366	err = neigh_fill_info(skb, n, pid, 0, type, flags);
3367	if (err < 0) {
3368		/* -EMSGSIZE implies BUG in neigh_nlmsg_size() */
3369		WARN_ON(err == -EMSGSIZE);
3370		kfree_skb(skb);
3371		goto errout;
3372	}
3373	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3374	return;
3375errout:
3376	if (err < 0)
3377		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3378}
3379
 
3380void neigh_app_ns(struct neighbour *n)
3381{
3382	__neigh_notify(n, RTM_GETNEIGH, NLM_F_REQUEST, 0);
3383}
3384EXPORT_SYMBOL(neigh_app_ns);
 
3385
3386#ifdef CONFIG_SYSCTL
3387static int unres_qlen_max = INT_MAX / SKB_TRUESIZE(ETH_FRAME_LEN);
3388
3389static int proc_unres_qlen(struct ctl_table *ctl, int write,
3390			   void *buffer, size_t *lenp, loff_t *ppos)
3391{
3392	int size, ret;
3393	struct ctl_table tmp = *ctl;
3394
3395	tmp.extra1 = SYSCTL_ZERO;
3396	tmp.extra2 = &unres_qlen_max;
3397	tmp.data = &size;
3398
3399	size = *(int *)ctl->data / SKB_TRUESIZE(ETH_FRAME_LEN);
3400	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
3401
3402	if (write && !ret)
3403		*(int *)ctl->data = size * SKB_TRUESIZE(ETH_FRAME_LEN);
3404	return ret;
3405}
3406
3407static struct neigh_parms *neigh_get_dev_parms_rcu(struct net_device *dev,
3408						   int family)
3409{
3410	switch (family) {
3411	case AF_INET:
3412		return __in_dev_arp_parms_get_rcu(dev);
3413	case AF_INET6:
3414		return __in6_dev_nd_parms_get_rcu(dev);
3415	}
3416	return NULL;
3417}
3418
3419static void neigh_copy_dflt_parms(struct net *net, struct neigh_parms *p,
3420				  int index)
3421{
3422	struct net_device *dev;
3423	int family = neigh_parms_family(p);
3424
3425	rcu_read_lock();
3426	for_each_netdev_rcu(net, dev) {
3427		struct neigh_parms *dst_p =
3428				neigh_get_dev_parms_rcu(dev, family);
3429
3430		if (dst_p && !test_bit(index, dst_p->data_state))
3431			dst_p->data[index] = p->data[index];
3432	}
3433	rcu_read_unlock();
3434}
3435
3436static void neigh_proc_update(struct ctl_table *ctl, int write)
3437{
3438	struct net_device *dev = ctl->extra1;
3439	struct neigh_parms *p = ctl->extra2;
3440	struct net *net = neigh_parms_net(p);
3441	int index = (int *) ctl->data - p->data;
3442
3443	if (!write)
3444		return;
3445
3446	set_bit(index, p->data_state);
3447	if (index == NEIGH_VAR_DELAY_PROBE_TIME)
3448		call_netevent_notifiers(NETEVENT_DELAY_PROBE_TIME_UPDATE, p);
3449	if (!dev) /* NULL dev means this is default value */
3450		neigh_copy_dflt_parms(net, p, index);
3451}
3452
3453static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write,
3454					   void *buffer, size_t *lenp,
3455					   loff_t *ppos)
3456{
3457	struct ctl_table tmp = *ctl;
3458	int ret;
3459
3460	tmp.extra1 = SYSCTL_ZERO;
3461	tmp.extra2 = SYSCTL_INT_MAX;
3462
3463	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
3464	neigh_proc_update(ctl, write);
3465	return ret;
3466}
3467
3468int neigh_proc_dointvec(struct ctl_table *ctl, int write, void *buffer,
3469			size_t *lenp, loff_t *ppos)
3470{
3471	int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
3472
3473	neigh_proc_update(ctl, write);
3474	return ret;
3475}
3476EXPORT_SYMBOL(neigh_proc_dointvec);
3477
3478int neigh_proc_dointvec_jiffies(struct ctl_table *ctl, int write, void *buffer,
3479				size_t *lenp, loff_t *ppos)
3480{
3481	int ret = proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3482
3483	neigh_proc_update(ctl, write);
3484	return ret;
3485}
3486EXPORT_SYMBOL(neigh_proc_dointvec_jiffies);
3487
3488static int neigh_proc_dointvec_userhz_jiffies(struct ctl_table *ctl, int write,
3489					      void *buffer, size_t *lenp,
3490					      loff_t *ppos)
3491{
3492	int ret = proc_dointvec_userhz_jiffies(ctl, write, buffer, lenp, ppos);
3493
3494	neigh_proc_update(ctl, write);
3495	return ret;
3496}
3497
3498int neigh_proc_dointvec_ms_jiffies(struct ctl_table *ctl, int write,
3499				   void *buffer, size_t *lenp, loff_t *ppos)
3500{
3501	int ret = proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3502
3503	neigh_proc_update(ctl, write);
3504	return ret;
3505}
3506EXPORT_SYMBOL(neigh_proc_dointvec_ms_jiffies);
3507
3508static int neigh_proc_dointvec_unres_qlen(struct ctl_table *ctl, int write,
3509					  void *buffer, size_t *lenp,
3510					  loff_t *ppos)
3511{
3512	int ret = proc_unres_qlen(ctl, write, buffer, lenp, ppos);
3513
3514	neigh_proc_update(ctl, write);
3515	return ret;
3516}
3517
3518static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,
3519					  void *buffer, size_t *lenp,
3520					  loff_t *ppos)
3521{
3522	struct neigh_parms *p = ctl->extra2;
3523	int ret;
3524
3525	if (strcmp(ctl->procname, "base_reachable_time") == 0)
3526		ret = neigh_proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos);
3527	else if (strcmp(ctl->procname, "base_reachable_time_ms") == 0)
3528		ret = neigh_proc_dointvec_ms_jiffies(ctl, write, buffer, lenp, ppos);
3529	else
3530		ret = -1;
3531
3532	if (write && ret == 0) {
3533		/* update reachable_time as well, otherwise, the change will
3534		 * only be effective after the next time neigh_periodic_work
3535		 * decides to recompute it
3536		 */
3537		p->reachable_time =
3538			neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
3539	}
3540	return ret;
3541}
3542
3543#define NEIGH_PARMS_DATA_OFFSET(index)	\
3544	(&((struct neigh_parms *) 0)->data[index])
3545
3546#define NEIGH_SYSCTL_ENTRY(attr, data_attr, name, mval, proc) \
3547	[NEIGH_VAR_ ## attr] = { \
3548		.procname	= name, \
3549		.data		= NEIGH_PARMS_DATA_OFFSET(NEIGH_VAR_ ## data_attr), \
3550		.maxlen		= sizeof(int), \
3551		.mode		= mval, \
3552		.proc_handler	= proc, \
3553	}
3554
3555#define NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(attr, name) \
3556	NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_zero_intmax)
3557
3558#define NEIGH_SYSCTL_JIFFIES_ENTRY(attr, name) \
3559	NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_jiffies)
3560
3561#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(attr, name) \
3562	NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)
3563
3564#define NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(attr, data_attr, name) \
3565	NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
3566
3567#define NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(attr, data_attr, name) \
3568	NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_unres_qlen)
3569
3570static struct neigh_sysctl_table {
3571	struct ctl_table_header *sysctl_header;
3572	struct ctl_table neigh_vars[NEIGH_VAR_MAX + 1];
 
3573} neigh_sysctl_template __read_mostly = {
3574	.neigh_vars = {
3575		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_PROBES, "mcast_solicit"),
3576		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(UCAST_PROBES, "ucast_solicit"),
3577		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(APP_PROBES, "app_solicit"),
3578		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(MCAST_REPROBES, "mcast_resolicit"),
3579		NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(RETRANS_TIME, "retrans_time"),
3580		NEIGH_SYSCTL_JIFFIES_ENTRY(BASE_REACHABLE_TIME, "base_reachable_time"),
3581		NEIGH_SYSCTL_JIFFIES_ENTRY(DELAY_PROBE_TIME, "delay_first_probe_time"),
3582		NEIGH_SYSCTL_JIFFIES_ENTRY(GC_STALETIME, "gc_stale_time"),
3583		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(QUEUE_LEN_BYTES, "unres_qlen_bytes"),
3584		NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(PROXY_QLEN, "proxy_qlen"),
3585		NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(ANYCAST_DELAY, "anycast_delay"),
3586		NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(PROXY_DELAY, "proxy_delay"),
3587		NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY(LOCKTIME, "locktime"),
3588		NEIGH_SYSCTL_UNRES_QLEN_REUSED_ENTRY(QUEUE_LEN, QUEUE_LEN_BYTES, "unres_qlen"),
3589		NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(RETRANS_TIME_MS, RETRANS_TIME, "retrans_time_ms"),
3590		NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY(BASE_REACHABLE_TIME_MS, BASE_REACHABLE_TIME, "base_reachable_time_ms"),
3591		[NEIGH_VAR_GC_INTERVAL] = {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3592			.procname	= "gc_interval",
3593			.maxlen		= sizeof(int),
3594			.mode		= 0644,
3595			.proc_handler	= proc_dointvec_jiffies,
3596		},
3597		[NEIGH_VAR_GC_THRESH1] = {
3598			.procname	= "gc_thresh1",
3599			.maxlen		= sizeof(int),
3600			.mode		= 0644,
3601			.extra1		= SYSCTL_ZERO,
3602			.extra2		= SYSCTL_INT_MAX,
3603			.proc_handler	= proc_dointvec_minmax,
3604		},
3605		[NEIGH_VAR_GC_THRESH2] = {
3606			.procname	= "gc_thresh2",
3607			.maxlen		= sizeof(int),
3608			.mode		= 0644,
3609			.extra1		= SYSCTL_ZERO,
3610			.extra2		= SYSCTL_INT_MAX,
3611			.proc_handler	= proc_dointvec_minmax,
3612		},
3613		[NEIGH_VAR_GC_THRESH3] = {
3614			.procname	= "gc_thresh3",
3615			.maxlen		= sizeof(int),
3616			.mode		= 0644,
3617			.extra1		= SYSCTL_ZERO,
3618			.extra2		= SYSCTL_INT_MAX,
3619			.proc_handler	= proc_dointvec_minmax,
3620		},
3621		{},
3622	},
3623};
3624
3625int neigh_sysctl_register(struct net_device *dev, struct neigh_parms *p,
3626			  proc_handler *handler)
3627{
3628	int i;
3629	struct neigh_sysctl_table *t;
3630	const char *dev_name_source;
3631	char neigh_path[ sizeof("net//neigh/") + IFNAMSIZ + IFNAMSIZ ];
3632	char *p_name;
 
 
 
 
 
 
 
 
 
 
 
3633
3634	t = kmemdup(&neigh_sysctl_template, sizeof(*t), GFP_KERNEL);
3635	if (!t)
3636		goto err;
3637
3638	for (i = 0; i < NEIGH_VAR_GC_INTERVAL; i++) {
3639		t->neigh_vars[i].data += (long) p;
3640		t->neigh_vars[i].extra1 = dev;
3641		t->neigh_vars[i].extra2 = p;
3642	}
 
 
 
 
 
 
 
 
 
3643
3644	if (dev) {
3645		dev_name_source = dev->name;
3646		/* Terminate the table early */
3647		memset(&t->neigh_vars[NEIGH_VAR_GC_INTERVAL], 0,
3648		       sizeof(t->neigh_vars[NEIGH_VAR_GC_INTERVAL]));
3649	} else {
3650		struct neigh_table *tbl = p->tbl;
3651		dev_name_source = "default";
3652		t->neigh_vars[NEIGH_VAR_GC_INTERVAL].data = &tbl->gc_interval;
3653		t->neigh_vars[NEIGH_VAR_GC_THRESH1].data = &tbl->gc_thresh1;
3654		t->neigh_vars[NEIGH_VAR_GC_THRESH2].data = &tbl->gc_thresh2;
3655		t->neigh_vars[NEIGH_VAR_GC_THRESH3].data = &tbl->gc_thresh3;
3656	}
3657
 
3658	if (handler) {
3659		/* RetransTime */
3660		t->neigh_vars[NEIGH_VAR_RETRANS_TIME].proc_handler = handler;
 
3661		/* ReachableTime */
3662		t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler = handler;
 
3663		/* RetransTime (in milliseconds)*/
3664		t->neigh_vars[NEIGH_VAR_RETRANS_TIME_MS].proc_handler = handler;
3665		/* ReachableTime (in milliseconds) */
3666		t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler = handler;
3667	} else {
3668		/* Those handlers will update p->reachable_time after
3669		 * base_reachable_time(_ms) is set to ensure the new timer starts being
3670		 * applied after the next neighbour update instead of waiting for
3671		 * neigh_periodic_work to update its value (can be multiple minutes)
3672		 * So any handler that replaces them should do this as well
3673		 */
3674		/* ReachableTime */
3675		t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME].proc_handler =
3676			neigh_proc_base_reachable_time;
3677		/* ReachableTime (in milliseconds) */
3678		t->neigh_vars[NEIGH_VAR_BASE_REACHABLE_TIME_MS].proc_handler =
3679			neigh_proc_base_reachable_time;
3680	}
3681
3682	/* Don't export sysctls to unprivileged users */
3683	if (neigh_parms_net(p)->user_ns != &init_user_ns)
3684		t->neigh_vars[0].procname = NULL;
3685
3686	switch (neigh_parms_family(p)) {
3687	case AF_INET:
3688	      p_name = "ipv4";
3689	      break;
3690	case AF_INET6:
3691	      p_name = "ipv6";
3692	      break;
3693	default:
3694	      BUG();
3695	}
3696
3697	snprintf(neigh_path, sizeof(neigh_path), "net/%s/neigh/%s",
3698		p_name, dev_name_source);
3699	t->sysctl_header =
3700		register_net_sysctl(neigh_parms_net(p), neigh_path, t->neigh_vars);
3701	if (!t->sysctl_header)
3702		goto free;
3703
3704	p->sysctl_table = t;
3705	return 0;
3706
 
 
3707free:
3708	kfree(t);
3709err:
3710	return -ENOBUFS;
3711}
3712EXPORT_SYMBOL(neigh_sysctl_register);
3713
3714void neigh_sysctl_unregister(struct neigh_parms *p)
3715{
3716	if (p->sysctl_table) {
3717		struct neigh_sysctl_table *t = p->sysctl_table;
3718		p->sysctl_table = NULL;
3719		unregister_net_sysctl_table(t->sysctl_header);
 
3720		kfree(t);
3721	}
3722}
3723EXPORT_SYMBOL(neigh_sysctl_unregister);
3724
3725#endif	/* CONFIG_SYSCTL */
3726
3727static int __init neigh_init(void)
3728{
3729	rtnl_register(PF_UNSPEC, RTM_NEWNEIGH, neigh_add, NULL, 0);
3730	rtnl_register(PF_UNSPEC, RTM_DELNEIGH, neigh_delete, NULL, 0);
3731	rtnl_register(PF_UNSPEC, RTM_GETNEIGH, neigh_get, neigh_dump_info, 0);
3732
3733	rtnl_register(PF_UNSPEC, RTM_GETNEIGHTBL, NULL, neightbl_dump_info,
3734		      0);
3735	rtnl_register(PF_UNSPEC, RTM_SETNEIGHTBL, neightbl_set, NULL, 0);
3736
3737	return 0;
3738}
3739
3740subsys_initcall(neigh_init);