rhashtable.c 27.1 KB
Newer Older
1 2 3
/*
 * Resizable, Scalable, Concurrent Hash Table
 *
4
 * Copyright (c) 2015 Herbert Xu <herbert@gondor.apana.org.au>
5
 * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
6 7 8
 * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
 *
 * Code partially derived from nft_hash
9 10
 * Rewritten with rehash code from br_multicast plus single list
 * pointer as suggested by Josh Triplett
11 12 13 14 15 16
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

17
#include <linux/atomic.h>
18 19 20
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/log2.h>
21
#include <linux/sched.h>
22
#include <linux/rculist.h>
23 24 25
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <linux/mm.h>
26
#include <linux/jhash.h>
27 28
#include <linux/random.h>
#include <linux/rhashtable.h>
29
#include <linux/err.h>
30
#include <linux/export.h>
31 32

#define HASH_DEFAULT_SIZE	64UL
33
#define HASH_MIN_SIZE		4U
34
#define BUCKET_LOCKS_PER_CPU	32UL
35

36 37 38 39 40
union nested_table {
	union nested_table __rcu *table;
	struct rhash_head __rcu *bucket;
};

41
static u32 head_hashfn(struct rhashtable *ht,
42 43
		       const struct bucket_table *tbl,
		       const struct rhash_head *he)
44
{
45
	return rht_head_hashfn(ht, tbl, he, ht->p);
46 47
}

48 49 50 51 52 53 54 55 56 57 58
#ifdef CONFIG_PROVE_LOCKING
#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))

int lockdep_rht_mutex_is_held(struct rhashtable *ht)
{
	return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
}
EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);

int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
{
59
	spinlock_t *lock = rht_bucket_lock(tbl, hash);
60 61 62 63 64 65 66 67 68

	return (debug_locks) ? lockdep_is_held(lock) : 1;
}
EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
#else
#define ASSERT_RHT_MUTEX(HT)
#endif


69 70
static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl,
			      gfp_t gfp)
71 72 73 74 75 76 77 78
{
	unsigned int i, size;
#if defined(CONFIG_PROVE_LOCKING)
	unsigned int nr_pcpus = 2;
#else
	unsigned int nr_pcpus = num_possible_cpus();
#endif

79
	nr_pcpus = min_t(unsigned int, nr_pcpus, 64UL);
80 81
	size = roundup_pow_of_two(nr_pcpus * ht->p.locks_mul);

82 83
	/* Never allocate more than 0.5 locks per bucket */
	size = min_t(unsigned int, size, tbl->size >> 1);
84

85 86 87
	if (tbl->nest)
		size = min(size, 1U << tbl->nest);

88
	if (sizeof(spinlock_t) != 0) {
89 90 91
		if (gfpflags_allow_blocking(gfp))
			tbl->locks = kvmalloc(size * sizeof(spinlock_t), gfp);
		else
92 93
			tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
						   gfp);
94 95 96 97 98 99 100 101 102 103
		if (!tbl->locks)
			return -ENOMEM;
		for (i = 0; i < size; i++)
			spin_lock_init(&tbl->locks[i]);
	}
	tbl->locks_mask = size - 1;

	return 0;
}

104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
static void nested_table_free(union nested_table *ntbl, unsigned int size)
{
	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
	const unsigned int len = 1 << shift;
	unsigned int i;

	ntbl = rcu_dereference_raw(ntbl->table);
	if (!ntbl)
		return;

	if (size > len) {
		size >>= shift;
		for (i = 0; i < len; i++)
			nested_table_free(ntbl + i, size);
	}

	kfree(ntbl);
}

static void nested_bucket_table_free(const struct bucket_table *tbl)
{
	unsigned int size = tbl->size >> tbl->nest;
	unsigned int len = 1 << tbl->nest;
	union nested_table *ntbl;
	unsigned int i;

	ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);

	for (i = 0; i < len; i++)
		nested_table_free(ntbl + i, size);

	kfree(ntbl);
}

138 139
static void bucket_table_free(const struct bucket_table *tbl)
{
140 141 142
	if (tbl->nest)
		nested_bucket_table_free(tbl);

143
	kvfree(tbl->locks);
144 145 146
	kvfree(tbl);
}

147 148 149 150 151
static void bucket_table_free_rcu(struct rcu_head *head)
{
	bucket_table_free(container_of(head, struct bucket_table, rcu));
}

152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
static union nested_table *nested_table_alloc(struct rhashtable *ht,
					      union nested_table __rcu **prev,
					      unsigned int shifted,
					      unsigned int nhash)
{
	union nested_table *ntbl;
	int i;

	ntbl = rcu_dereference(*prev);
	if (ntbl)
		return ntbl;

	ntbl = kzalloc(PAGE_SIZE, GFP_ATOMIC);

	if (ntbl && shifted) {
		for (i = 0; i < PAGE_SIZE / sizeof(ntbl[0].bucket); i++)
			INIT_RHT_NULLS_HEAD(ntbl[i].bucket, ht,
					    (i << shifted) | nhash);
	}

	rcu_assign_pointer(*prev, ntbl);

	return ntbl;
}

static struct bucket_table *nested_bucket_table_alloc(struct rhashtable *ht,
						      size_t nbuckets,
						      gfp_t gfp)
{
	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
	struct bucket_table *tbl;
	size_t size;

	if (nbuckets < (1 << (shift + 1)))
		return NULL;

	size = sizeof(*tbl) + sizeof(tbl->buckets[0]);

	tbl = kzalloc(size, gfp);
	if (!tbl)
		return NULL;

	if (!nested_table_alloc(ht, (union nested_table __rcu **)tbl->buckets,
				0, 0)) {
		kfree(tbl);
		return NULL;
	}

	tbl->nest = (ilog2(nbuckets) - 1) % shift + 1;

	return tbl;
}

205
static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
206 207
					       size_t nbuckets,
					       gfp_t gfp)
208
{
209
	struct bucket_table *tbl = NULL;
210
	size_t size;
211
	int i;
212 213

	size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]);
214
	if (gfp != GFP_KERNEL)
215
		tbl = kzalloc(size, gfp | __GFP_NOWARN | __GFP_NORETRY);
216 217
	else
		tbl = kvzalloc(size, gfp);
218 219 220 221 222 223 224

	size = nbuckets;

	if (tbl == NULL && gfp != GFP_KERNEL) {
		tbl = nested_bucket_table_alloc(ht, nbuckets, gfp);
		nbuckets = 0;
	}
225 226 227
	if (tbl == NULL)
		return NULL;

228
	tbl->size = size;
229

230
	if (alloc_bucket_locks(ht, tbl, gfp) < 0) {
231 232 233
		bucket_table_free(tbl);
		return NULL;
	}
234

235 236
	INIT_LIST_HEAD(&tbl->walkers);

237
	tbl->hash_rnd = get_random_u32();
238

239 240 241
	for (i = 0; i < nbuckets; i++)
		INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);

242
	return tbl;
243 244
}

245 246 247 248 249 250 251 252 253 254 255 256 257
static struct bucket_table *rhashtable_last_table(struct rhashtable *ht,
						  struct bucket_table *tbl)
{
	struct bucket_table *new_tbl;

	do {
		new_tbl = tbl;
		tbl = rht_dereference_rcu(tbl->future_tbl, ht);
	} while (tbl);

	return new_tbl;
}

258
static int rhashtable_rehash_one(struct rhashtable *ht, unsigned int old_hash)
259
{
260
	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
261 262
	struct bucket_table *new_tbl = rhashtable_last_table(ht,
		rht_dereference_rcu(old_tbl->future_tbl, ht));
263 264
	struct rhash_head __rcu **pprev = rht_bucket_var(old_tbl, old_hash);
	int err = -EAGAIN;
265 266
	struct rhash_head *head, *next, *entry;
	spinlock_t *new_bucket_lock;
267
	unsigned int new_hash;
268

269 270 271 272 273
	if (new_tbl->nest)
		goto out;

	err = -ENOENT;

274 275 276 277 278 279
	rht_for_each(entry, old_tbl, old_hash) {
		err = 0;
		next = rht_dereference_bucket(entry->next, old_tbl, old_hash);

		if (rht_is_a_nulls(next))
			break;
280

281 282
		pprev = &entry->next;
	}
283

284 285
	if (err)
		goto out;
286

287
	new_hash = head_hashfn(ht, new_tbl, entry);
288

289
	new_bucket_lock = rht_bucket_lock(new_tbl, new_hash);
290

291
	spin_lock_nested(new_bucket_lock, SINGLE_DEPTH_NESTING);
292 293
	head = rht_dereference_bucket(new_tbl->buckets[new_hash],
				      new_tbl, new_hash);
294

295
	RCU_INIT_POINTER(entry->next, head);
296

297 298
	rcu_assign_pointer(new_tbl->buckets[new_hash], entry);
	spin_unlock(new_bucket_lock);
299

300
	rcu_assign_pointer(*pprev, next);
301

302 303 304
out:
	return err;
}
305

306
static int rhashtable_rehash_chain(struct rhashtable *ht,
307
				    unsigned int old_hash)
308 309 310
{
	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
	spinlock_t *old_bucket_lock;
311
	int err;
312

313
	old_bucket_lock = rht_bucket_lock(old_tbl, old_hash);
314

315
	spin_lock_bh(old_bucket_lock);
316
	while (!(err = rhashtable_rehash_one(ht, old_hash)))
317
		;
318 319 320 321 322

	if (err == -ENOENT) {
		old_tbl->rehash++;
		err = 0;
	}
323
	spin_unlock_bh(old_bucket_lock);
324 325

	return err;
326 327
}

328 329 330
static int rhashtable_rehash_attach(struct rhashtable *ht,
				    struct bucket_table *old_tbl,
				    struct bucket_table *new_tbl)
331
{
332 333 334 335 336 337 338 339
	/* Protect future_tbl using the first bucket lock. */
	spin_lock_bh(old_tbl->locks);

	/* Did somebody beat us to it? */
	if (rcu_access_pointer(old_tbl->future_tbl)) {
		spin_unlock_bh(old_tbl->locks);
		return -EEXIST;
	}
340

341 342 343
	/* Make insertions go into the new, empty table right away. Deletions
	 * and lookups will be attempted in both tables until we synchronize.
	 */
344
	rcu_assign_pointer(old_tbl->future_tbl, new_tbl);
345

346 347 348 349 350 351 352 353 354 355
	spin_unlock_bh(old_tbl->locks);

	return 0;
}

static int rhashtable_rehash_table(struct rhashtable *ht)
{
	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
	struct bucket_table *new_tbl;
	struct rhashtable_walker *walker;
356
	unsigned int old_hash;
357
	int err;
358 359 360 361 362

	new_tbl = rht_dereference(old_tbl->future_tbl, ht);
	if (!new_tbl)
		return 0;

363 364 365 366 367
	for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
		err = rhashtable_rehash_chain(ht, old_hash);
		if (err)
			return err;
	}
368 369 370 371

	/* Publish the new table pointer. */
	rcu_assign_pointer(ht->tbl, new_tbl);

372
	spin_lock(&ht->lock);
373 374
	list_for_each_entry(walker, &old_tbl->walkers, list)
		walker->tbl = NULL;
375
	spin_unlock(&ht->lock);
376

377 378 379 380
	/* Wait for readers. All new readers will see the new
	 * table, and thus no references to the old table will
	 * remain.
	 */
381
	call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
382 383

	return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
384 385
}

386 387 388
static int rhashtable_rehash_alloc(struct rhashtable *ht,
				   struct bucket_table *old_tbl,
				   unsigned int size)
389
{
390
	struct bucket_table *new_tbl;
391
	int err;
392 393 394

	ASSERT_RHT_MUTEX(ht);

395
	new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
396 397 398
	if (new_tbl == NULL)
		return -ENOMEM;

399 400 401 402 403
	err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
	if (err)
		bucket_table_free(new_tbl);

	return err;
404 405 406 407 408 409
}

/**
 * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
 * @ht:		the hash table to shrink
 *
H
Herbert Xu 已提交
410 411
 * This function shrinks the hash table to fit, i.e., the smallest
 * size would not cause it to expand right away automatically.
412
 *
413 414 415
 * The caller must ensure that no concurrent resizing occurs by holding
 * ht->mutex.
 *
416 417
 * The caller must ensure that no concurrent table mutations take place.
 * It is however valid to have concurrent lookups if they are RCU protected.
418 419 420
 *
 * It is valid to have concurrent insertions and deletions protected by per
 * bucket locks or concurrent RCU protected lookups and traversals.
421
 */
422
static int rhashtable_shrink(struct rhashtable *ht)
423
{
424
	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
425 426
	unsigned int nelems = atomic_read(&ht->nelems);
	unsigned int size = 0;
427

428 429
	if (nelems)
		size = roundup_pow_of_two(nelems * 3 / 2);
H
Herbert Xu 已提交
430 431 432 433 434 435
	if (size < ht->p.min_size)
		size = ht->p.min_size;

	if (old_tbl->size <= size)
		return 0;

436 437 438
	if (rht_dereference(old_tbl->future_tbl, ht))
		return -EEXIST;

439
	return rhashtable_rehash_alloc(ht, old_tbl, size);
440 441
}

442 443 444 445
static void rht_deferred_worker(struct work_struct *work)
{
	struct rhashtable *ht;
	struct bucket_table *tbl;
446
	int err = 0;
447

448
	ht = container_of(work, struct rhashtable, run_work);
449
	mutex_lock(&ht->mutex);
450

451
	tbl = rht_dereference(ht->tbl, ht);
452
	tbl = rhashtable_last_table(ht, tbl);
453

454
	if (rht_grow_above_75(ht, tbl))
455
		err = rhashtable_rehash_alloc(ht, tbl, tbl->size * 2);
456
	else if (ht->p.automatic_shrinking && rht_shrink_below_30(ht, tbl))
457 458 459
		err = rhashtable_shrink(ht);
	else if (tbl->nest)
		err = rhashtable_rehash_alloc(ht, tbl, tbl->size);
460

461 462
	if (!err)
		err = rhashtable_rehash_table(ht);
463

464
	mutex_unlock(&ht->mutex);
465 466 467

	if (err)
		schedule_work(&ht->run_work);
468 469
}

470 471
static int rhashtable_insert_rehash(struct rhashtable *ht,
				    struct bucket_table *tbl)
472 473 474 475 476 477 478 479 480 481
{
	struct bucket_table *old_tbl;
	struct bucket_table *new_tbl;
	unsigned int size;
	int err;

	old_tbl = rht_dereference_rcu(ht->tbl, ht);

	size = tbl->size;

482 483
	err = -EBUSY;

484 485
	if (rht_grow_above_75(ht, tbl))
		size *= 2;
486 487
	/* Do not schedule more than one rehash */
	else if (old_tbl != tbl)
488 489 490
		goto fail;

	err = -ENOMEM;
491 492

	new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC);
493 494
	if (new_tbl == NULL)
		goto fail;
495 496 497 498 499 500 501 502 503 504

	err = rhashtable_rehash_attach(ht, tbl, new_tbl);
	if (err) {
		bucket_table_free(new_tbl);
		if (err == -EEXIST)
			err = 0;
	} else
		schedule_work(&ht->run_work);

	return err;
505 506 507 508 509 510 511 512 513 514 515

fail:
	/* Do not fail the insert if someone else did a rehash. */
	if (likely(rcu_dereference_raw(tbl->future_tbl)))
		return 0;

	/* Schedule async rehash to retry allocation in process context. */
	if (err == -ENOMEM)
		schedule_work(&ht->run_work);

	return err;
516 517
}

518 519 520
static void *rhashtable_lookup_one(struct rhashtable *ht,
				   struct bucket_table *tbl, unsigned int hash,
				   const void *key, struct rhash_head *obj)
521
{
522 523 524 525 526
	struct rhashtable_compare_arg arg = {
		.ht = ht,
		.key = key,
	};
	struct rhash_head __rcu **pprev;
527
	struct rhash_head *head;
528
	int elasticity;
529

530
	elasticity = RHT_ELASTICITY;
531 532
	pprev = rht_bucket_var(tbl, hash);
	rht_for_each_continue(head, *pprev, tbl, hash) {
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
		struct rhlist_head *list;
		struct rhlist_head *plist;

		elasticity--;
		if (!key ||
		    (ht->p.obj_cmpfn ?
		     ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
		     rhashtable_compare(&arg, rht_obj(ht, head))))
			continue;

		if (!ht->rhlist)
			return rht_obj(ht, head);

		list = container_of(obj, struct rhlist_head, rhead);
		plist = container_of(head, struct rhlist_head, rhead);

		RCU_INIT_POINTER(list->next, plist);
		head = rht_dereference_bucket(head->next, tbl, hash);
		RCU_INIT_POINTER(list->rhead.next, head);
		rcu_assign_pointer(*pprev, obj);

		return NULL;
555
	}
556

557 558 559 560 561 562 563 564 565 566 567 568
	if (elasticity <= 0)
		return ERR_PTR(-EAGAIN);

	return ERR_PTR(-ENOENT);
}

static struct bucket_table *rhashtable_insert_one(struct rhashtable *ht,
						  struct bucket_table *tbl,
						  unsigned int hash,
						  struct rhash_head *obj,
						  void *data)
{
569
	struct rhash_head __rcu **pprev;
570 571 572 573 574
	struct bucket_table *new_tbl;
	struct rhash_head *head;

	if (!IS_ERR_OR_NULL(data))
		return ERR_PTR(-EEXIST);
575

576 577
	if (PTR_ERR(data) != -EAGAIN && PTR_ERR(data) != -ENOENT)
		return ERR_CAST(data);
578

579 580 581 582 583 584 585 586 587 588 589 590
	new_tbl = rcu_dereference(tbl->future_tbl);
	if (new_tbl)
		return new_tbl;

	if (PTR_ERR(data) != -ENOENT)
		return ERR_CAST(data);

	if (unlikely(rht_grow_above_max(ht, tbl)))
		return ERR_PTR(-E2BIG);

	if (unlikely(rht_grow_above_100(ht, tbl)))
		return ERR_PTR(-EAGAIN);
591

592 593 594 595 596
	pprev = rht_bucket_insert(ht, tbl, hash);
	if (!pprev)
		return ERR_PTR(-ENOMEM);

	head = rht_dereference_bucket(*pprev, tbl, hash);
597 598

	RCU_INIT_POINTER(obj->next, head);
599 600 601 602 603 604
	if (ht->rhlist) {
		struct rhlist_head *list;

		list = container_of(obj, struct rhlist_head, rhead);
		RCU_INIT_POINTER(list->next, NULL);
	}
605

606
	rcu_assign_pointer(*pprev, obj);
607 608

	atomic_inc(&ht->nelems);
609 610
	if (rht_grow_above_75(ht, tbl))
		schedule_work(&ht->run_work);
611

612 613
	return NULL;
}
614

615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680
static void *rhashtable_try_insert(struct rhashtable *ht, const void *key,
				   struct rhash_head *obj)
{
	struct bucket_table *new_tbl;
	struct bucket_table *tbl;
	unsigned int hash;
	spinlock_t *lock;
	void *data;

	tbl = rcu_dereference(ht->tbl);

	/* All insertions must grab the oldest table containing
	 * the hashed bucket that is yet to be rehashed.
	 */
	for (;;) {
		hash = rht_head_hashfn(ht, tbl, obj, ht->p);
		lock = rht_bucket_lock(tbl, hash);
		spin_lock_bh(lock);

		if (tbl->rehash <= hash)
			break;

		spin_unlock_bh(lock);
		tbl = rcu_dereference(tbl->future_tbl);
	}

	data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
	new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
	if (PTR_ERR(new_tbl) != -EEXIST)
		data = ERR_CAST(new_tbl);

	while (!IS_ERR_OR_NULL(new_tbl)) {
		tbl = new_tbl;
		hash = rht_head_hashfn(ht, tbl, obj, ht->p);
		spin_lock_nested(rht_bucket_lock(tbl, hash),
				 SINGLE_DEPTH_NESTING);

		data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
		new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
		if (PTR_ERR(new_tbl) != -EEXIST)
			data = ERR_CAST(new_tbl);

		spin_unlock(rht_bucket_lock(tbl, hash));
	}

	spin_unlock_bh(lock);

	if (PTR_ERR(data) == -EAGAIN)
		data = ERR_PTR(rhashtable_insert_rehash(ht, tbl) ?:
			       -EAGAIN);

	return data;
}

void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
			     struct rhash_head *obj)
{
	void *data;

	do {
		rcu_read_lock();
		data = rhashtable_try_insert(ht, key, obj);
		rcu_read_unlock();
	} while (PTR_ERR(data) == -EAGAIN);

	return data;
681 682 683
}
EXPORT_SYMBOL_GPL(rhashtable_insert_slow);

684
/**
685
 * rhashtable_walk_enter - Initialise an iterator
686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701
 * @ht:		Table to walk over
 * @iter:	Hash table Iterator
 *
 * This function prepares a hash table walk.
 *
 * Note that if you restart a walk after rhashtable_walk_stop you
 * may see the same object twice.  Also, you may miss objects if
 * there are removals in between rhashtable_walk_stop and the next
 * call to rhashtable_walk_start.
 *
 * For a completely stable walk you should construct your own data
 * structure outside the hash table.
 *
 * This function may sleep so you must not call it from interrupt
 * context or with spin locks held.
 *
702
 * You must call rhashtable_walk_exit after this function returns.
703
 */
704
void rhashtable_walk_enter(struct rhashtable *ht, struct rhashtable_iter *iter)
705 706 707 708 709 710
{
	iter->ht = ht;
	iter->p = NULL;
	iter->slot = 0;
	iter->skip = 0;

711
	spin_lock(&ht->lock);
712
	iter->walker.tbl =
713
		rcu_dereference_protected(ht->tbl, lockdep_is_held(&ht->lock));
714
	list_add(&iter->walker.list, &iter->walker.tbl->walkers);
715
	spin_unlock(&ht->lock);
716
}
717
EXPORT_SYMBOL_GPL(rhashtable_walk_enter);
718 719 720 721 722 723 724 725 726

/**
 * rhashtable_walk_exit - Free an iterator
 * @iter:	Hash table Iterator
 *
 * This function frees resources allocated by rhashtable_walk_init.
 */
void rhashtable_walk_exit(struct rhashtable_iter *iter)
{
727
	spin_lock(&iter->ht->lock);
728 729
	if (iter->walker.tbl)
		list_del(&iter->walker.list);
730
	spin_unlock(&iter->ht->lock);
731 732 733 734 735 736 737
}
EXPORT_SYMBOL_GPL(rhashtable_walk_exit);

/**
 * rhashtable_walk_start - Start a hash table walk
 * @iter:	Hash table iterator
 *
738 739 740
 * Start a hash table walk at the current iterator position.  Note that we take
 * the RCU lock in all cases including when we return an error.  So you must
 * always call rhashtable_walk_stop to clean up.
741 742 743 744 745 746 747 748
 *
 * Returns zero if successful.
 *
 * Returns -EAGAIN if resize event occured.  Note that the iterator
 * will rewind back to the beginning and you may use it immediately
 * by calling rhashtable_walk_next.
 */
int rhashtable_walk_start(struct rhashtable_iter *iter)
749
	__acquires(RCU)
750
{
751 752
	struct rhashtable *ht = iter->ht;

753
	rcu_read_lock();
754

755
	spin_lock(&ht->lock);
756 757
	if (iter->walker.tbl)
		list_del(&iter->walker.list);
758
	spin_unlock(&ht->lock);
759

760 761
	if (!iter->walker.tbl) {
		iter->walker.tbl = rht_dereference_rcu(ht->tbl, ht);
762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
		return -EAGAIN;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(rhashtable_walk_start);

/**
 * rhashtable_walk_next - Return the next object and advance the iterator
 * @iter:	Hash table iterator
 *
 * Note that you must call rhashtable_walk_stop when you are finished
 * with the walk.
 *
 * Returns the next object or NULL when the end of the table is reached.
 *
 * Returns -EAGAIN if resize event occured.  Note that the iterator
 * will rewind back to the beginning and you may continue to use it.
 */
void *rhashtable_walk_next(struct rhashtable_iter *iter)
{
783
	struct bucket_table *tbl = iter->walker.tbl;
784
	struct rhlist_head *list = iter->list;
785 786
	struct rhashtable *ht = iter->ht;
	struct rhash_head *p = iter->p;
787
	bool rhlist = ht->rhlist;
788 789

	if (p) {
790 791 792 793
		if (!rhlist || !(list = rcu_dereference(list->next))) {
			p = rcu_dereference(p->next);
			list = container_of(p, struct rhlist_head, rhead);
		}
794 795 796 797 798 799 800
		goto next;
	}

	for (; iter->slot < tbl->size; iter->slot++) {
		int skip = iter->skip;

		rht_for_each_rcu(p, tbl, iter->slot) {
801 802 803 804 805 806 807 808 809 810 811 812
			if (rhlist) {
				list = container_of(p, struct rhlist_head,
						    rhead);
				do {
					if (!skip)
						goto next;
					skip--;
					list = rcu_dereference(list->next);
				} while (list);

				continue;
			}
813 814 815 816 817 818 819 820 821
			if (!skip)
				break;
			skip--;
		}

next:
		if (!rht_is_a_nulls(p)) {
			iter->skip++;
			iter->p = p;
822 823
			iter->list = list;
			return rht_obj(ht, rhlist ? &list->rhead : p);
824 825 826 827 828
		}

		iter->skip = 0;
	}

829 830
	iter->p = NULL;

831 832 833
	/* Ensure we see any new tables. */
	smp_rmb();

834 835
	iter->walker.tbl = rht_dereference_rcu(tbl->future_tbl, ht);
	if (iter->walker.tbl) {
836 837 838 839 840
		iter->slot = 0;
		iter->skip = 0;
		return ERR_PTR(-EAGAIN);
	}

841
	return NULL;
842 843 844 845 846 847 848
}
EXPORT_SYMBOL_GPL(rhashtable_walk_next);

/**
 * rhashtable_walk_stop - Finish a hash table walk
 * @iter:	Hash table iterator
 *
849 850
 * Finish a hash table walk.  Does not reset the iterator to the start of the
 * hash table.
851 852
 */
void rhashtable_walk_stop(struct rhashtable_iter *iter)
853
	__releases(RCU)
854
{
855
	struct rhashtable *ht;
856
	struct bucket_table *tbl = iter->walker.tbl;
857 858

	if (!tbl)
859
		goto out;
860 861 862

	ht = iter->ht;

863
	spin_lock(&ht->lock);
864
	if (tbl->rehash < tbl->size)
865
		list_add(&iter->walker.list, &tbl->walkers);
866
	else
867
		iter->walker.tbl = NULL;
868
	spin_unlock(&ht->lock);
869

870
	iter->p = NULL;
871 872 873

out:
	rcu_read_unlock();
874 875 876
}
EXPORT_SYMBOL_GPL(rhashtable_walk_stop);

877
static size_t rounded_hashtable_size(const struct rhashtable_params *params)
878
{
879
	return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
880
		   (unsigned long)params->min_size);
881 882
}

883 884 885 886 887
static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
{
	return jhash2(key, length, seed);
}

888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
/**
 * rhashtable_init - initialize a new hash table
 * @ht:		hash table to be initialized
 * @params:	configuration parameters
 *
 * Initializes a new hash table based on the provided configuration
 * parameters. A table can be configured either with a variable or
 * fixed length key:
 *
 * Configuration Example 1: Fixed length keys
 * struct test_obj {
 *	int			key;
 *	void *			my_member;
 *	struct rhash_head	node;
 * };
 *
 * struct rhashtable_params params = {
 *	.head_offset = offsetof(struct test_obj, node),
 *	.key_offset = offsetof(struct test_obj, key),
 *	.key_len = sizeof(int),
908
 *	.hashfn = jhash,
909
 *	.nulls_base = (1U << RHT_BASE_SHIFT),
910 911 912 913 914 915 916 917
 * };
 *
 * Configuration Example 2: Variable length keys
 * struct test_obj {
 *	[...]
 *	struct rhash_head	node;
 * };
 *
918
 * u32 my_hash_fn(const void *data, u32 len, u32 seed)
919 920 921 922 923 924 925 926
 * {
 *	struct test_obj *obj = data;
 *
 *	return [... hash ...];
 * }
 *
 * struct rhashtable_params params = {
 *	.head_offset = offsetof(struct test_obj, node),
927
 *	.hashfn = jhash,
928 929 930
 *	.obj_hashfn = my_hash_fn,
 * };
 */
931 932
int rhashtable_init(struct rhashtable *ht,
		    const struct rhashtable_params *params)
933 934 935 936 937 938
{
	struct bucket_table *tbl;
	size_t size;

	size = HASH_DEFAULT_SIZE;

939
	if ((!params->key_len && !params->obj_hashfn) ||
940
	    (params->obj_hashfn && !params->obj_cmpfn))
941 942
		return -EINVAL;

943 944 945
	if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
		return -EINVAL;

946 947
	memset(ht, 0, sizeof(*ht));
	mutex_init(&ht->mutex);
948
	spin_lock_init(&ht->lock);
949 950
	memcpy(&ht->p, params, sizeof(*params));

951 952 953
	if (params->min_size)
		ht->p.min_size = roundup_pow_of_two(params->min_size);

954 955
	/* Cap total entries at 2^31 to avoid nelems overflow. */
	ht->max_elems = 1u << 31;
956 957 958 959 960 961

	if (params->max_size) {
		ht->p.max_size = rounddown_pow_of_two(params->max_size);
		if (ht->p.max_size < ht->max_elems / 2)
			ht->max_elems = ht->p.max_size * 2;
	}
962

963
	ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
964

965 966 967
	if (params->nelem_hint)
		size = rounded_hashtable_size(&ht->p);

968 969 970 971 972
	if (params->locks_mul)
		ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
	else
		ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;

973 974 975 976 977 978 979 980 981 982
	ht->key_len = ht->p.key_len;
	if (!params->hashfn) {
		ht->p.hashfn = jhash;

		if (!(ht->key_len & (sizeof(u32) - 1))) {
			ht->key_len /= sizeof(u32);
			ht->p.hashfn = rhashtable_jhash2;
		}
	}

983
	tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
984 985 986
	if (tbl == NULL)
		return -ENOMEM;

987
	atomic_set(&ht->nelems, 0);
988

989 990
	RCU_INIT_POINTER(ht->tbl, tbl);

991
	INIT_WORK(&ht->run_work, rht_deferred_worker);
992

993 994 995 996
	return 0;
}
EXPORT_SYMBOL_GPL(rhashtable_init);

997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
/**
 * rhltable_init - initialize a new hash list table
 * @hlt:	hash list table to be initialized
 * @params:	configuration parameters
 *
 * Initializes a new hash list table.
 *
 * See documentation for rhashtable_init.
 */
int rhltable_init(struct rhltable *hlt, const struct rhashtable_params *params)
{
	int err;

	/* No rhlist NULLs marking for now. */
	if (params->nulls_base)
		return -EINVAL;

	err = rhashtable_init(&hlt->ht, params);
	hlt->ht.rhlist = true;
	return err;
}
EXPORT_SYMBOL_GPL(rhltable_init);

static void rhashtable_free_one(struct rhashtable *ht, struct rhash_head *obj,
				void (*free_fn)(void *ptr, void *arg),
				void *arg)
{
	struct rhlist_head *list;

	if (!ht->rhlist) {
		free_fn(rht_obj(ht, obj), arg);
		return;
	}

	list = container_of(obj, struct rhlist_head, rhead);
	do {
		obj = &list->rhead;
		list = rht_dereference(list->next, ht);
		free_fn(rht_obj(ht, obj), arg);
	} while (list);
}

1039
/**
1040
 * rhashtable_free_and_destroy - free elements and destroy hash table
1041
 * @ht:		the hash table to destroy
1042 1043
 * @free_fn:	callback to release resources of element
 * @arg:	pointer passed to free_fn
1044
 *
1045 1046 1047 1048 1049 1050 1051 1052
 * Stops an eventual async resize. If defined, invokes free_fn for each
 * element to releasal resources. Please note that RCU protected
 * readers may still be accessing the elements. Releasing of resources
 * must occur in a compatible manner. Then frees the bucket array.
 *
 * This function will eventually sleep to wait for an async resize
 * to complete. The caller is responsible that no further write operations
 * occurs in parallel.
1053
 */
1054 1055 1056
void rhashtable_free_and_destroy(struct rhashtable *ht,
				 void (*free_fn)(void *ptr, void *arg),
				 void *arg)
1057
{
1058
	struct bucket_table *tbl;
1059
	unsigned int i;
1060

1061
	cancel_work_sync(&ht->run_work);
1062

1063
	mutex_lock(&ht->mutex);
1064 1065 1066 1067 1068
	tbl = rht_dereference(ht->tbl, ht);
	if (free_fn) {
		for (i = 0; i < tbl->size; i++) {
			struct rhash_head *pos, *next;

1069
			for (pos = rht_dereference(*rht_bucket(tbl, i), ht),
1070 1071 1072 1073 1074 1075
			     next = !rht_is_a_nulls(pos) ?
					rht_dereference(pos->next, ht) : NULL;
			     !rht_is_a_nulls(pos);
			     pos = next,
			     next = !rht_is_a_nulls(pos) ?
					rht_dereference(pos->next, ht) : NULL)
1076
				rhashtable_free_one(ht, pos, free_fn, arg);
1077 1078 1079 1080
		}
	}

	bucket_table_free(tbl);
1081
	mutex_unlock(&ht->mutex);
1082
}
1083 1084 1085 1086 1087 1088
EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);

void rhashtable_destroy(struct rhashtable *ht)
{
	return rhashtable_free_and_destroy(ht, NULL, NULL);
}
1089
EXPORT_SYMBOL_GPL(rhashtable_destroy);
1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102

struct rhash_head __rcu **rht_bucket_nested(const struct bucket_table *tbl,
					    unsigned int hash)
{
	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
	static struct rhash_head __rcu *rhnull =
		(struct rhash_head __rcu *)NULLS_MARKER(0);
	unsigned int index = hash & ((1 << tbl->nest) - 1);
	unsigned int size = tbl->size >> tbl->nest;
	unsigned int subhash = hash;
	union nested_table *ntbl;

	ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);
1103
	ntbl = rht_dereference_bucket_rcu(ntbl[index].table, tbl, hash);
1104 1105 1106 1107
	subhash >>= tbl->nest;

	while (ntbl && size > (1 << shift)) {
		index = subhash & ((1 << shift) - 1);
1108 1109
		ntbl = rht_dereference_bucket_rcu(ntbl[index].table,
						  tbl, hash);
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157
		size >>= shift;
		subhash >>= shift;
	}

	if (!ntbl)
		return &rhnull;

	return &ntbl[subhash].bucket;

}
EXPORT_SYMBOL_GPL(rht_bucket_nested);

struct rhash_head __rcu **rht_bucket_nested_insert(struct rhashtable *ht,
						   struct bucket_table *tbl,
						   unsigned int hash)
{
	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
	unsigned int index = hash & ((1 << tbl->nest) - 1);
	unsigned int size = tbl->size >> tbl->nest;
	union nested_table *ntbl;
	unsigned int shifted;
	unsigned int nhash;

	ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);
	hash >>= tbl->nest;
	nhash = index;
	shifted = tbl->nest;
	ntbl = nested_table_alloc(ht, &ntbl[index].table,
				  size <= (1 << shift) ? shifted : 0, nhash);

	while (ntbl && size > (1 << shift)) {
		index = hash & ((1 << shift) - 1);
		size >>= shift;
		hash >>= shift;
		nhash |= index << shifted;
		shifted += shift;
		ntbl = nested_table_alloc(ht, &ntbl[index].table,
					  size <= (1 << shift) ? shifted : 0,
					  nhash);
	}

	if (!ntbl)
		return NULL;

	return &ntbl[hash].bucket;

}
EXPORT_SYMBOL_GPL(rht_bucket_nested_insert);
反馈
建议
客服 返回
顶部