projects/linux·blob v4.13 · refs

1156 lines · 27.0 KiB

1/*
2 * Resizable, Scalable, Concurrent Hash Table
3 *
4 * Copyright (c) 2015 Herbert Xu <[email protected]>
5 * Copyright (c) 2014-2015 Thomas Graf <[email protected]>
6 * Copyright (c) 2008-2014 Patrick McHardy <[email protected]>
7 *
8 * Code partially derived from nft_hash
9 * Rewritten with rehash code from br_multicast plus single list
10 * pointer as suggested by Josh Triplett
11 *
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
15 */
16
17#include <linux/atomic.h>
18#include <linux/kernel.h>
19#include <linux/init.h>
20#include <linux/log2.h>
21#include <linux/sched.h>
22#include <linux/rculist.h>
23#include <linux/slab.h>
24#include <linux/vmalloc.h>
25#include <linux/mm.h>
26#include <linux/jhash.h>
27#include <linux/random.h>
28#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
36union nested_table {
37 union nested_table __rcu *table;
38 struct rhash_head __rcu *bucket;
39};
40
41static u32 head_hashfn(struct rhashtable *ht,
42 const struct bucket_table *tbl,
43 const struct rhash_head *he)
44{
45 return rht_head_hashfn(ht, tbl, he, ht->p);
46}
47
48#ifdef CONFIG_PROVE_LOCKING
49#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
50
51int lockdep_rht_mutex_is_held(struct rhashtable *ht)
52{
53 return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
54}
55EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
56
57int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
58{
59 spinlock_t *lock = rht_bucket_lock(tbl, hash);
60
61 return (debug_locks) ? lockdep_is_held(lock) : 1;
62}
63EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
64#else
65#define ASSERT_RHT_MUTEX(HT)
66#endif
67
68
69static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl,
70 gfp_t gfp)
71{
72 unsigned int i, size;
73#if defined(CONFIG_PROVE_LOCKING)
74 unsigned int nr_pcpus = 2;
75#else
76 unsigned int nr_pcpus = num_possible_cpus();
77#endif
78
79 nr_pcpus = min_t(unsigned int, nr_pcpus, 64UL);
80 size = roundup_pow_of_two(nr_pcpus * ht->p.locks_mul);
81
82 /* Never allocate more than 0.5 locks per bucket */
83 size = min_t(unsigned int, size, tbl->size >> 1);
84
85 if (tbl->nest)
86 size = min(size, 1U << tbl->nest);
87
88 if (sizeof(spinlock_t) != 0) {
89 if (gfpflags_allow_blocking(gfp))
90 tbl->locks = kvmalloc(size * sizeof(spinlock_t), gfp);
91 else
92 tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
93 gfp);
94 if (!tbl->locks)
95 return -ENOMEM;
96 for (i = 0; i < size; i++)
97 spin_lock_init(&tbl->locks[i]);
98 }
99 tbl->locks_mask = size - 1;
100
101 return 0;
102}
103
104static void nested_table_free(union nested_table *ntbl, unsigned int size)
105{
106 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
107 const unsigned int len = 1 << shift;
108 unsigned int i;
109
110 ntbl = rcu_dereference_raw(ntbl->table);
111 if (!ntbl)
112 return;
113
114 if (size > len) {
115 size >>= shift;
116 for (i = 0; i < len; i++)
117 nested_table_free(ntbl + i, size);
118 }
119
120 kfree(ntbl);
121}
122
123static void nested_bucket_table_free(const struct bucket_table *tbl)
124{
125 unsigned int size = tbl->size >> tbl->nest;
126 unsigned int len = 1 << tbl->nest;
127 union nested_table *ntbl;
128 unsigned int i;
129
130 ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);
131
132 for (i = 0; i < len; i++)
133 nested_table_free(ntbl + i, size);
134
135 kfree(ntbl);
136}
137
138static void bucket_table_free(const struct bucket_table *tbl)
139{
140 if (tbl->nest)
141 nested_bucket_table_free(tbl);
142
143 kvfree(tbl->locks);
144 kvfree(tbl);
145}
146
147static void bucket_table_free_rcu(struct rcu_head *head)
148{
149 bucket_table_free(container_of(head, struct bucket_table, rcu));
150}
151
152static union nested_table *nested_table_alloc(struct rhashtable *ht,
153 union nested_table __rcu **prev,
154 unsigned int shifted,
155 unsigned int nhash)
156{
157 union nested_table *ntbl;
158 int i;
159
160 ntbl = rcu_dereference(*prev);
161 if (ntbl)
162 return ntbl;
163
164 ntbl = kzalloc(PAGE_SIZE, GFP_ATOMIC);
165
166 if (ntbl && shifted) {
167 for (i = 0; i < PAGE_SIZE / sizeof(ntbl[0].bucket); i++)
168 INIT_RHT_NULLS_HEAD(ntbl[i].bucket, ht,
169 (i << shifted) | nhash);
170 }
171
172 rcu_assign_pointer(*prev, ntbl);
173
174 return ntbl;
175}
176
177static struct bucket_table *nested_bucket_table_alloc(struct rhashtable *ht,
178 size_t nbuckets,
179 gfp_t gfp)
180{
181 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
182 struct bucket_table *tbl;
183 size_t size;
184
185 if (nbuckets < (1 << (shift + 1)))
186 return NULL;
187
188 size = sizeof(*tbl) + sizeof(tbl->buckets[0]);
189
190 tbl = kzalloc(size, gfp);
191 if (!tbl)
192 return NULL;
193
194 if (!nested_table_alloc(ht, (union nested_table __rcu **)tbl->buckets,
195 0, 0)) {
196 kfree(tbl);
197 return NULL;
198 }
199
200 tbl->nest = (ilog2(nbuckets) - 1) % shift + 1;
201
202 return tbl;
203}
204
205static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
206 size_t nbuckets,
207 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 else
217 tbl = kvzalloc(size, gfp);
218
219 size = nbuckets;
220
221 if (tbl == NULL && gfp != GFP_KERNEL) {
222 tbl = nested_bucket_table_alloc(ht, nbuckets, gfp);
223 nbuckets = 0;
224 }
225 if (tbl == NULL)
226 return NULL;
227
228 tbl->size = size;
229
230 if (alloc_bucket_locks(ht, tbl, gfp) < 0) {
231 bucket_table_free(tbl);
232 return NULL;
233 }
234
235 INIT_LIST_HEAD(&tbl->walkers);
236
237 tbl->hash_rnd = get_random_u32();
238
239 for (i = 0; i < nbuckets; i++)
240 INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);
241
242 return tbl;
243}
244
245static struct bucket_table *rhashtable_last_table(struct rhashtable *ht,
246 struct bucket_table *tbl)
247{
248 struct bucket_table *new_tbl;
249
250 do {
251 new_tbl = tbl;
252 tbl = rht_dereference_rcu(tbl->future_tbl, ht);
253 } while (tbl);
254
255 return new_tbl;
256}
257
258static int rhashtable_rehash_one(struct rhashtable *ht, unsigned int old_hash)
259{
260 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
261 struct bucket_table *new_tbl = rhashtable_last_table(ht,
262 rht_dereference_rcu(old_tbl->future_tbl, ht));
263 struct rhash_head __rcu **pprev = rht_bucket_var(old_tbl, old_hash);
264 int err = -EAGAIN;
265 struct rhash_head *head, *next, *entry;
266 spinlock_t *new_bucket_lock;
267 unsigned int new_hash;
268
269 if (new_tbl->nest)
270 goto out;
271
272 err = -ENOENT;
273
274 rht_for_each(entry, old_tbl, old_hash) {
275 err = 0;
276 next = rht_dereference_bucket(entry->next, old_tbl, old_hash);
277
278 if (rht_is_a_nulls(next))
279 break;
280
281 pprev = &entry->next;
282 }
283
284 if (err)
285 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 head = rht_dereference_bucket(new_tbl->buckets[new_hash],
293 new_tbl, new_hash);
294
295 RCU_INIT_POINTER(entry->next, head);
296
297 rcu_assign_pointer(new_tbl->buckets[new_hash], entry);
298 spin_unlock(new_bucket_lock);
299
300 rcu_assign_pointer(*pprev, next);
301
302out:
303 return err;
304}
305
306static int rhashtable_rehash_chain(struct rhashtable *ht,
307 unsigned int old_hash)
308{
309 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
310 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 if (err == -ENOENT) {
320 old_tbl->rehash++;
321 err = 0;
322 }
323 spin_unlock_bh(old_bucket_lock);
324
325 return err;
326}
327
328static int rhashtable_rehash_attach(struct rhashtable *ht,
329 struct bucket_table *old_tbl,
330 struct bucket_table *new_tbl)
331{
332 /* Protect future_tbl using the first bucket lock. */
333 spin_lock_bh(old_tbl->locks);
334
335 /* Did somebody beat us to it? */
336 if (rcu_access_pointer(old_tbl->future_tbl)) {
337 spin_unlock_bh(old_tbl->locks);
338 return -EEXIST;
339 }
340
341 /* Make insertions go into the new, empty table right away. Deletions
342 * and lookups will be attempted in both tables until we synchronize.
343 */
344 rcu_assign_pointer(old_tbl->future_tbl, new_tbl);
345
346 spin_unlock_bh(old_tbl->locks);
347
348 return 0;
349}
350
351static int rhashtable_rehash_table(struct rhashtable *ht)
352{
353 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
354 struct bucket_table *new_tbl;
355 struct rhashtable_walker *walker;
356 unsigned int old_hash;
357 int err;
358
359 new_tbl = rht_dereference(old_tbl->future_tbl, ht);
360 if (!new_tbl)
361 return 0;
362
363 for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
364 err = rhashtable_rehash_chain(ht, old_hash);
365 if (err)
366 return err;
367 }
368
369 /* Publish the new table pointer. */
370 rcu_assign_pointer(ht->tbl, new_tbl);
371
372 spin_lock(&ht->lock);
373 list_for_each_entry(walker, &old_tbl->walkers, list)
374 walker->tbl = NULL;
375 spin_unlock(&ht->lock);
376
377 /* Wait for readers. All new readers will see the new
378 * table, and thus no references to the old table will
379 * remain.
380 */
381 call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
382
383 return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
384}
385
386static int rhashtable_rehash_alloc(struct rhashtable *ht,
387 struct bucket_table *old_tbl,
388 unsigned int size)
389{
390 struct bucket_table *new_tbl;
391 int err;
392
393 ASSERT_RHT_MUTEX(ht);
394
395 new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
396 if (new_tbl == NULL)
397 return -ENOMEM;
398
399 err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
400 if (err)
401 bucket_table_free(new_tbl);
402
403 return err;
404}
405
406/**
407 * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
408 * @ht: the hash table to shrink
409 *
410 * This function shrinks the hash table to fit, i.e., the smallest
411 * size would not cause it to expand right away automatically.
412 *
413 * The caller must ensure that no concurrent resizing occurs by holding
414 * ht->mutex.
415 *
416 * The caller must ensure that no concurrent table mutations take place.
417 * It is however valid to have concurrent lookups if they are RCU protected.
418 *
419 * It is valid to have concurrent insertions and deletions protected by per
420 * bucket locks or concurrent RCU protected lookups and traversals.
421 */
422static int rhashtable_shrink(struct rhashtable *ht)
423{
424 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
425 unsigned int nelems = atomic_read(&ht->nelems);
426 unsigned int size = 0;
427
428 if (nelems)
429 size = roundup_pow_of_two(nelems * 3 / 2);
430 if (size < ht->p.min_size)
431 size = ht->p.min_size;
432
433 if (old_tbl->size <= size)
434 return 0;
435
436 if (rht_dereference(old_tbl->future_tbl, ht))
437 return -EEXIST;
438
439 return rhashtable_rehash_alloc(ht, old_tbl, size);
440}
441
442static void rht_deferred_worker(struct work_struct *work)
443{
444 struct rhashtable *ht;
445 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 err = rhashtable_shrink(ht);
458 else if (tbl->nest)
459 err = rhashtable_rehash_alloc(ht, tbl, tbl->size);
460
461 if (!err)
462 err = rhashtable_rehash_table(ht);
463
464 mutex_unlock(&ht->mutex);
465
466 if (err)
467 schedule_work(&ht->run_work);
468}
469
470static int rhashtable_insert_rehash(struct rhashtable *ht,
471 struct bucket_table *tbl)
472{
473 struct bucket_table *old_tbl;
474 struct bucket_table *new_tbl;
475 unsigned int size;
476 int err;
477
478 old_tbl = rht_dereference_rcu(ht->tbl, ht);
479
480 size = tbl->size;
481
482 err = -EBUSY;
483
484 if (rht_grow_above_75(ht, tbl))
485 size *= 2;
486 /* Do not schedule more than one rehash */
487 else if (old_tbl != tbl)
488 goto fail;
489
490 err = -ENOMEM;
491
492 new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC);
493 if (new_tbl == NULL)
494 goto fail;
495
496 err = rhashtable_rehash_attach(ht, tbl, new_tbl);
497 if (err) {
498 bucket_table_free(new_tbl);
499 if (err == -EEXIST)
500 err = 0;
501 } else
502 schedule_work(&ht->run_work);
503
504 return err;
505
506fail:
507 /* Do not fail the insert if someone else did a rehash. */
508 if (likely(rcu_dereference_raw(tbl->future_tbl)))
509 return 0;
510
511 /* Schedule async rehash to retry allocation in process context. */
512 if (err == -ENOMEM)
513 schedule_work(&ht->run_work);
514
515 return err;
516}
517
518static void *rhashtable_lookup_one(struct rhashtable *ht,
519 struct bucket_table *tbl, unsigned int hash,
520 const void *key, struct rhash_head *obj)
521{
522 struct rhashtable_compare_arg arg = {
523 .ht = ht,
524 .key = key,
525 };
526 struct rhash_head __rcu **pprev;
527 struct rhash_head *head;
528 int elasticity;
529
530 elasticity = RHT_ELASTICITY;
531 pprev = rht_bucket_var(tbl, hash);
532 rht_for_each_continue(head, *pprev, tbl, hash) {
533 struct rhlist_head *list;
534 struct rhlist_head *plist;
535
536 elasticity--;
537 if (!key ||
538 (ht->p.obj_cmpfn ?
539 ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
540 rhashtable_compare(&arg, rht_obj(ht, head))))
541 continue;
542
543 if (!ht->rhlist)
544 return rht_obj(ht, head);
545
546 list = container_of(obj, struct rhlist_head, rhead);
547 plist = container_of(head, struct rhlist_head, rhead);
548
549 RCU_INIT_POINTER(list->next, plist);
550 head = rht_dereference_bucket(head->next, tbl, hash);
551 RCU_INIT_POINTER(list->rhead.next, head);
552 rcu_assign_pointer(*pprev, obj);
553
554 return NULL;
555 }
556
557 if (elasticity <= 0)
558 return ERR_PTR(-EAGAIN);
559
560 return ERR_PTR(-ENOENT);
561}
562
563static struct bucket_table *rhashtable_insert_one(struct rhashtable *ht,
564 struct bucket_table *tbl,
565 unsigned int hash,
566 struct rhash_head *obj,
567 void *data)
568{
569 struct rhash_head __rcu **pprev;
570 struct bucket_table *new_tbl;
571 struct rhash_head *head;
572
573 if (!IS_ERR_OR_NULL(data))
574 return ERR_PTR(-EEXIST);
575
576 if (PTR_ERR(data) != -EAGAIN && PTR_ERR(data) != -ENOENT)
577 return ERR_CAST(data);
578
579 new_tbl = rcu_dereference(tbl->future_tbl);
580 if (new_tbl)
581 return new_tbl;
582
583 if (PTR_ERR(data) != -ENOENT)
584 return ERR_CAST(data);
585
586 if (unlikely(rht_grow_above_max(ht, tbl)))
587 return ERR_PTR(-E2BIG);
588
589 if (unlikely(rht_grow_above_100(ht, tbl)))
590 return ERR_PTR(-EAGAIN);
591
592 pprev = rht_bucket_insert(ht, tbl, hash);
593 if (!pprev)
594 return ERR_PTR(-ENOMEM);
595
596 head = rht_dereference_bucket(*pprev, tbl, hash);
597
598 RCU_INIT_POINTER(obj->next, head);
599 if (ht->rhlist) {
600 struct rhlist_head *list;
601
602 list = container_of(obj, struct rhlist_head, rhead);
603 RCU_INIT_POINTER(list->next, NULL);
604 }
605
606 rcu_assign_pointer(*pprev, obj);
607
608 atomic_inc(&ht->nelems);
609 if (rht_grow_above_75(ht, tbl))
610 schedule_work(&ht->run_work);
611
612 return NULL;
613}
614
615static void *rhashtable_try_insert(struct rhashtable *ht, const void *key,
616 struct rhash_head *obj)
617{
618 struct bucket_table *new_tbl;
619 struct bucket_table *tbl;
620 unsigned int hash;
621 spinlock_t *lock;
622 void *data;
623
624 tbl = rcu_dereference(ht->tbl);
625
626 /* All insertions must grab the oldest table containing
627 * the hashed bucket that is yet to be rehashed.
628 */
629 for (;;) {
630 hash = rht_head_hashfn(ht, tbl, obj, ht->p);
631 lock = rht_bucket_lock(tbl, hash);
632 spin_lock_bh(lock);
633
634 if (tbl->rehash <= hash)
635 break;
636
637 spin_unlock_bh(lock);
638 tbl = rcu_dereference(tbl->future_tbl);
639 }
640
641 data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
642 new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
643 if (PTR_ERR(new_tbl) != -EEXIST)
644 data = ERR_CAST(new_tbl);
645
646 while (!IS_ERR_OR_NULL(new_tbl)) {
647 tbl = new_tbl;
648 hash = rht_head_hashfn(ht, tbl, obj, ht->p);
649 spin_lock_nested(rht_bucket_lock(tbl, hash),
650 SINGLE_DEPTH_NESTING);
651
652 data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
653 new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
654 if (PTR_ERR(new_tbl) != -EEXIST)
655 data = ERR_CAST(new_tbl);
656
657 spin_unlock(rht_bucket_lock(tbl, hash));
658 }
659
660 spin_unlock_bh(lock);
661
662 if (PTR_ERR(data) == -EAGAIN)
663 data = ERR_PTR(rhashtable_insert_rehash(ht, tbl) ?:
664 -EAGAIN);
665
666 return data;
667}
668
669void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
670 struct rhash_head *obj)
671{
672 void *data;
673
674 do {
675 rcu_read_lock();
676 data = rhashtable_try_insert(ht, key, obj);
677 rcu_read_unlock();
678 } while (PTR_ERR(data) == -EAGAIN);
679
680 return data;
681}
682EXPORT_SYMBOL_GPL(rhashtable_insert_slow);
683
684/**
685 * rhashtable_walk_enter - Initialise an iterator
686 * @ht: Table to walk over
687 * @iter: Hash table Iterator
688 *
689 * This function prepares a hash table walk.
690 *
691 * Note that if you restart a walk after rhashtable_walk_stop you
692 * may see the same object twice. Also, you may miss objects if
693 * there are removals in between rhashtable_walk_stop and the next
694 * call to rhashtable_walk_start.
695 *
696 * For a completely stable walk you should construct your own data
697 * structure outside the hash table.
698 *
699 * This function may sleep so you must not call it from interrupt
700 * context or with spin locks held.
701 *
702 * You must call rhashtable_walk_exit after this function returns.
703 */
704void rhashtable_walk_enter(struct rhashtable *ht, struct rhashtable_iter *iter)
705{
706 iter->ht = ht;
707 iter->p = NULL;
708 iter->slot = 0;
709 iter->skip = 0;
710
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}
717EXPORT_SYMBOL_GPL(rhashtable_walk_enter);
718
719/**
720 * rhashtable_walk_exit - Free an iterator
721 * @iter: Hash table Iterator
722 *
723 * This function frees resources allocated by rhashtable_walk_init.
724 */
725void rhashtable_walk_exit(struct rhashtable_iter *iter)
726{
727 spin_lock(&iter->ht->lock);
728 if (iter->walker.tbl)
729 list_del(&iter->walker.list);
730 spin_unlock(&iter->ht->lock);
731}
732EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
733
734/**
735 * rhashtable_walk_start - Start a hash table walk
736 * @iter: Hash table iterator
737 *
738 * Start a hash table walk. Note that we take the RCU lock in all
739 * cases including when we return an error. So you must always call
740 * rhashtable_walk_stop to clean up.
741 *
742 * Returns zero if successful.
743 *
744 * Returns -EAGAIN if resize event occured. Note that the iterator
745 * will rewind back to the beginning and you may use it immediately
746 * by calling rhashtable_walk_next.
747 */
748int rhashtable_walk_start(struct rhashtable_iter *iter)
749 __acquires(RCU)
750{
751 struct rhashtable *ht = iter->ht;
752
753 rcu_read_lock();
754
755 spin_lock(&ht->lock);
756 if (iter->walker.tbl)
757 list_del(&iter->walker.list);
758 spin_unlock(&ht->lock);
759
760 if (!iter->walker.tbl) {
761 iter->walker.tbl = rht_dereference_rcu(ht->tbl, ht);
762 return -EAGAIN;
763 }
764
765 return 0;
766}
767EXPORT_SYMBOL_GPL(rhashtable_walk_start);
768
769/**
770 * rhashtable_walk_next - Return the next object and advance the iterator
771 * @iter: Hash table iterator
772 *
773 * Note that you must call rhashtable_walk_stop when you are finished
774 * with the walk.
775 *
776 * Returns the next object or NULL when the end of the table is reached.
777 *
778 * Returns -EAGAIN if resize event occured. Note that the iterator
779 * will rewind back to the beginning and you may continue to use it.
780 */
781void *rhashtable_walk_next(struct rhashtable_iter *iter)
782{
783 struct bucket_table *tbl = iter->walker.tbl;
784 struct rhlist_head *list = iter->list;
785 struct rhashtable *ht = iter->ht;
786 struct rhash_head *p = iter->p;
787 bool rhlist = ht->rhlist;
788
789 if (p) {
790 if (!rhlist || !(list = rcu_dereference(list->next))) {
791 p = rcu_dereference(p->next);
792 list = container_of(p, struct rhlist_head, rhead);
793 }
794 goto next;
795 }
796
797 for (; iter->slot < tbl->size; iter->slot++) {
798 int skip = iter->skip;
799
800 rht_for_each_rcu(p, tbl, iter->slot) {
801 if (rhlist) {
802 list = container_of(p, struct rhlist_head,
803 rhead);
804 do {
805 if (!skip)
806 goto next;
807 skip--;
808 list = rcu_dereference(list->next);
809 } while (list);
810
811 continue;
812 }
813 if (!skip)
814 break;
815 skip--;
816 }
817
818next:
819 if (!rht_is_a_nulls(p)) {
820 iter->skip++;
821 iter->p = p;
822 iter->list = list;
823 return rht_obj(ht, rhlist ? &list->rhead : p);
824 }
825
826 iter->skip = 0;
827 }
828
829 iter->p = NULL;
830
831 /* Ensure we see any new tables. */
832 smp_rmb();
833
834 iter->walker.tbl = rht_dereference_rcu(tbl->future_tbl, ht);
835 if (iter->walker.tbl) {
836 iter->slot = 0;
837 iter->skip = 0;
838 return ERR_PTR(-EAGAIN);
839 }
840
841 return NULL;
842}
843EXPORT_SYMBOL_GPL(rhashtable_walk_next);
844
845/**
846 * rhashtable_walk_stop - Finish a hash table walk
847 * @iter: Hash table iterator
848 *
849 * Finish a hash table walk.
850 */
851void rhashtable_walk_stop(struct rhashtable_iter *iter)
852 __releases(RCU)
853{
854 struct rhashtable *ht;
855 struct bucket_table *tbl = iter->walker.tbl;
856
857 if (!tbl)
858 goto out;
859
860 ht = iter->ht;
861
862 spin_lock(&ht->lock);
863 if (tbl->rehash < tbl->size)
864 list_add(&iter->walker.list, &tbl->walkers);
865 else
866 iter->walker.tbl = NULL;
867 spin_unlock(&ht->lock);
868
869 iter->p = NULL;
870
871out:
872 rcu_read_unlock();
873}
874EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
875
876static size_t rounded_hashtable_size(const struct rhashtable_params *params)
877{
878 return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
879 (unsigned long)params->min_size);
880}
881
882static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
883{
884 return jhash2(key, length, seed);
885}
886
887/**
888 * rhashtable_init - initialize a new hash table
889 * @ht: hash table to be initialized
890 * @params: configuration parameters
891 *
892 * Initializes a new hash table based on the provided configuration
893 * parameters. A table can be configured either with a variable or
894 * fixed length key:
895 *
896 * Configuration Example 1: Fixed length keys
897 * struct test_obj {
898 * int key;
899 * void * my_member;
900 * struct rhash_head node;
901 * };
902 *
903 * struct rhashtable_params params = {
904 * .head_offset = offsetof(struct test_obj, node),
905 * .key_offset = offsetof(struct test_obj, key),
906 * .key_len = sizeof(int),
907 * .hashfn = jhash,
908 * .nulls_base = (1U << RHT_BASE_SHIFT),
909 * };
910 *
911 * Configuration Example 2: Variable length keys
912 * struct test_obj {
913 * [...]
914 * struct rhash_head node;
915 * };
916 *
917 * u32 my_hash_fn(const void *data, u32 len, u32 seed)
918 * {
919 * struct test_obj *obj = data;
920 *
921 * return [... hash ...];
922 * }
923 *
924 * struct rhashtable_params params = {
925 * .head_offset = offsetof(struct test_obj, node),
926 * .hashfn = jhash,
927 * .obj_hashfn = my_hash_fn,
928 * };
929 */
930int rhashtable_init(struct rhashtable *ht,
931 const struct rhashtable_params *params)
932{
933 struct bucket_table *tbl;
934 size_t size;
935
936 size = HASH_DEFAULT_SIZE;
937
938 if ((!params->key_len && !params->obj_hashfn) ||
939 (params->obj_hashfn && !params->obj_cmpfn))
940 return -EINVAL;
941
942 if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
943 return -EINVAL;
944
945 memset(ht, 0, sizeof(*ht));
946 mutex_init(&ht->mutex);
947 spin_lock_init(&ht->lock);
948 memcpy(&ht->p, params, sizeof(*params));
949
950 if (params->min_size)
951 ht->p.min_size = roundup_pow_of_two(params->min_size);
952
953 /* Cap total entries at 2^31 to avoid nelems overflow. */
954 ht->max_elems = 1u << 31;
955
956 if (params->max_size) {
957 ht->p.max_size = rounddown_pow_of_two(params->max_size);
958 if (ht->p.max_size < ht->max_elems / 2)
959 ht->max_elems = ht->p.max_size * 2;
960 }
961
962 ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
963
964 if (params->nelem_hint)
965 size = rounded_hashtable_size(&ht->p);
966
967 if (params->locks_mul)
968 ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
969 else
970 ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;
971
972 ht->key_len = ht->p.key_len;
973 if (!params->hashfn) {
974 ht->p.hashfn = jhash;
975
976 if (!(ht->key_len & (sizeof(u32) - 1))) {
977 ht->key_len /= sizeof(u32);
978 ht->p.hashfn = rhashtable_jhash2;
979 }
980 }
981
982 tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
983 if (tbl == NULL)
984 return -ENOMEM;
985
986 atomic_set(&ht->nelems, 0);
987
988 RCU_INIT_POINTER(ht->tbl, tbl);
989
990 INIT_WORK(&ht->run_work, rht_deferred_worker);
991
992 return 0;
993}
994EXPORT_SYMBOL_GPL(rhashtable_init);
995
996/**
997 * rhltable_init - initialize a new hash list table
998 * @hlt: hash list table to be initialized
999 * @params: configuration parameters
1000 *
1001 * Initializes a new hash list table.
1002 *
1003 * See documentation for rhashtable_init.
1004 */
1005int rhltable_init(struct rhltable *hlt, const struct rhashtable_params *params)
1006{
1007 int err;
1008
1009 /* No rhlist NULLs marking for now. */
1010 if (params->nulls_base)
1011 return -EINVAL;
1012
1013 err = rhashtable_init(&hlt->ht, params);
1014 hlt->ht.rhlist = true;
1015 return err;
1016}
1017EXPORT_SYMBOL_GPL(rhltable_init);
1018
1019static void rhashtable_free_one(struct rhashtable *ht, struct rhash_head *obj,
1020 void (*free_fn)(void *ptr, void *arg),
1021 void *arg)
1022{
1023 struct rhlist_head *list;
1024
1025 if (!ht->rhlist) {
1026 free_fn(rht_obj(ht, obj), arg);
1027 return;
1028 }
1029
1030 list = container_of(obj, struct rhlist_head, rhead);
1031 do {
1032 obj = &list->rhead;
1033 list = rht_dereference(list->next, ht);
1034 free_fn(rht_obj(ht, obj), arg);
1035 } while (list);
1036}
1037
1038/**
1039 * rhashtable_free_and_destroy - free elements and destroy hash table
1040 * @ht: the hash table to destroy
1041 * @free_fn: callback to release resources of element
1042 * @arg: pointer passed to free_fn
1043 *
1044 * Stops an eventual async resize. If defined, invokes free_fn for each
1045 * element to releasal resources. Please note that RCU protected
1046 * readers may still be accessing the elements. Releasing of resources
1047 * must occur in a compatible manner. Then frees the bucket array.
1048 *
1049 * This function will eventually sleep to wait for an async resize
1050 * to complete. The caller is responsible that no further write operations
1051 * occurs in parallel.
1052 */
1053void rhashtable_free_and_destroy(struct rhashtable *ht,
1054 void (*free_fn)(void *ptr, void *arg),
1055 void *arg)
1056{
1057 struct bucket_table *tbl;
1058 unsigned int i;
1059
1060 cancel_work_sync(&ht->run_work);
1061
1062 mutex_lock(&ht->mutex);
1063 tbl = rht_dereference(ht->tbl, ht);
1064 if (free_fn) {
1065 for (i = 0; i < tbl->size; i++) {
1066 struct rhash_head *pos, *next;
1067
1068 for (pos = rht_dereference(*rht_bucket(tbl, i), ht),
1069 next = !rht_is_a_nulls(pos) ?
1070 rht_dereference(pos->next, ht) : NULL;
1071 !rht_is_a_nulls(pos);
1072 pos = next,
1073 next = !rht_is_a_nulls(pos) ?
1074 rht_dereference(pos->next, ht) : NULL)
1075 rhashtable_free_one(ht, pos, free_fn, arg);
1076 }
1077 }
1078
1079 bucket_table_free(tbl);
1080 mutex_unlock(&ht->mutex);
1081}
1082EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
1083
1084void rhashtable_destroy(struct rhashtable *ht)
1085{
1086 return rhashtable_free_and_destroy(ht, NULL, NULL);
1087}
1088EXPORT_SYMBOL_GPL(rhashtable_destroy);
1089
1090struct rhash_head __rcu **rht_bucket_nested(const struct bucket_table *tbl,
1091 unsigned int hash)
1092{
1093 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1094 static struct rhash_head __rcu *rhnull =
1095 (struct rhash_head __rcu *)NULLS_MARKER(0);
1096 unsigned int index = hash & ((1 << tbl->nest) - 1);
1097 unsigned int size = tbl->size >> tbl->nest;
1098 unsigned int subhash = hash;
1099 union nested_table *ntbl;
1100
1101 ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);
1102 ntbl = rht_dereference_bucket_rcu(ntbl[index].table, tbl, hash);
1103 subhash >>= tbl->nest;
1104
1105 while (ntbl && size > (1 << shift)) {
1106 index = subhash & ((1 << shift) - 1);
1107 ntbl = rht_dereference_bucket_rcu(ntbl[index].table,
1108 tbl, hash);
1109 size >>= shift;
1110 subhash >>= shift;
1111 }
1112
1113 if (!ntbl)
1114 return &rhnull;
1115
1116 return &ntbl[subhash].bucket;
1117
1118}
1119EXPORT_SYMBOL_GPL(rht_bucket_nested);
1120
1121struct rhash_head __rcu **rht_bucket_nested_insert(struct rhashtable *ht,
1122 struct bucket_table *tbl,
1123 unsigned int hash)
1124{
1125 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1126 unsigned int index = hash & ((1 << tbl->nest) - 1);
1127 unsigned int size = tbl->size >> tbl->nest;
1128 union nested_table *ntbl;
1129 unsigned int shifted;
1130 unsigned int nhash;
1131
1132 ntbl = (union nested_table *)rcu_dereference_raw(tbl->buckets[0]);
1133 hash >>= tbl->nest;
1134 nhash = index;
1135 shifted = tbl->nest;
1136 ntbl = nested_table_alloc(ht, &ntbl[index].table,
1137 size <= (1 << shift) ? shifted : 0, nhash);
1138
1139 while (ntbl && size > (1 << shift)) {
1140 index = hash & ((1 << shift) - 1);
1141 size >>= shift;
1142 hash >>= shift;
1143 nhash |= index << shifted;
1144 shifted += shift;
1145 ntbl = nested_table_alloc(ht, &ntbl[index].table,
1146 size <= (1 << shift) ? shifted : 0,
1147 nhash);
1148 }
1149
1150 if (!ntbl)
1151 return NULL;
1152
1153 return &ntbl[hash].bucket;
1154
1155}
1156EXPORT_SYMBOL_GPL(rht_bucket_nested_insert);