projects/linux·blob v5.13 · refs

1241 lines · 29.5 KiB

1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Resizable, Scalable, Concurrent Hash Table
4 *
5 * Copyright (c) 2015 Herbert Xu <[email protected]>
6 * Copyright (c) 2014-2015 Thomas Graf <[email protected]>
7 * Copyright (c) 2008-2014 Patrick McHardy <[email protected]>
8 *
9 * Code partially derived from nft_hash
10 * Rewritten with rehash code from br_multicast plus single list
11 * pointer as suggested by Josh Triplett
12 */
13
14#include <linux/atomic.h>
15#include <linux/kernel.h>
16#include <linux/init.h>
17#include <linux/log2.h>
18#include <linux/sched.h>
19#include <linux/rculist.h>
20#include <linux/slab.h>
21#include <linux/vmalloc.h>
22#include <linux/mm.h>
23#include <linux/jhash.h>
24#include <linux/random.h>
25#include <linux/rhashtable.h>
26#include <linux/err.h>
27#include <linux/export.h>
28
29#define HASH_DEFAULT_SIZE 64UL
30#define HASH_MIN_SIZE 4U
31
32union nested_table {
33 union nested_table __rcu *table;
34 struct rhash_lock_head __rcu *bucket;
35};
36
37static u32 head_hashfn(struct rhashtable *ht,
38 const struct bucket_table *tbl,
39 const struct rhash_head *he)
40{
41 return rht_head_hashfn(ht, tbl, he, ht->p);
42}
43
44#ifdef CONFIG_PROVE_LOCKING
45#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
46
47int lockdep_rht_mutex_is_held(struct rhashtable *ht)
48{
49 return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
50}
51EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
52
53int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
54{
55 if (!debug_locks)
56 return 1;
57 if (unlikely(tbl->nest))
58 return 1;
59 return bit_spin_is_locked(0, (unsigned long *)&tbl->buckets[hash]);
60}
61EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
62#else
63#define ASSERT_RHT_MUTEX(HT)
64#endif
65
66static inline union nested_table *nested_table_top(
67 const struct bucket_table *tbl)
68{
69 /* The top-level bucket entry does not need RCU protection
70 * because it's set at the same time as tbl->nest.
71 */
72 return (void *)rcu_dereference_protected(tbl->buckets[0], 1);
73}
74
75static void nested_table_free(union nested_table *ntbl, unsigned int size)
76{
77 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
78 const unsigned int len = 1 << shift;
79 unsigned int i;
80
81 ntbl = rcu_dereference_protected(ntbl->table, 1);
82 if (!ntbl)
83 return;
84
85 if (size > len) {
86 size >>= shift;
87 for (i = 0; i < len; i++)
88 nested_table_free(ntbl + i, size);
89 }
90
91 kfree(ntbl);
92}
93
94static void nested_bucket_table_free(const struct bucket_table *tbl)
95{
96 unsigned int size = tbl->size >> tbl->nest;
97 unsigned int len = 1 << tbl->nest;
98 union nested_table *ntbl;
99 unsigned int i;
100
101 ntbl = nested_table_top(tbl);
102
103 for (i = 0; i < len; i++)
104 nested_table_free(ntbl + i, size);
105
106 kfree(ntbl);
107}
108
109static void bucket_table_free(const struct bucket_table *tbl)
110{
111 if (tbl->nest)
112 nested_bucket_table_free(tbl);
113
114 kvfree(tbl);
115}
116
117static void bucket_table_free_rcu(struct rcu_head *head)
118{
119 bucket_table_free(container_of(head, struct bucket_table, rcu));
120}
121
122static union nested_table *nested_table_alloc(struct rhashtable *ht,
123 union nested_table __rcu **prev,
124 bool leaf)
125{
126 union nested_table *ntbl;
127 int i;
128
129 ntbl = rcu_dereference(*prev);
130 if (ntbl)
131 return ntbl;
132
133 ntbl = kzalloc(PAGE_SIZE, GFP_ATOMIC);
134
135 if (ntbl && leaf) {
136 for (i = 0; i < PAGE_SIZE / sizeof(ntbl[0]); i++)
137 INIT_RHT_NULLS_HEAD(ntbl[i].bucket);
138 }
139
140 if (cmpxchg((union nested_table **)prev, NULL, ntbl) == NULL)
141 return ntbl;
142 /* Raced with another thread. */
143 kfree(ntbl);
144 return rcu_dereference(*prev);
145}
146
147static struct bucket_table *nested_bucket_table_alloc(struct rhashtable *ht,
148 size_t nbuckets,
149 gfp_t gfp)
150{
151 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
152 struct bucket_table *tbl;
153 size_t size;
154
155 if (nbuckets < (1 << (shift + 1)))
156 return NULL;
157
158 size = sizeof(*tbl) + sizeof(tbl->buckets[0]);
159
160 tbl = kzalloc(size, gfp);
161 if (!tbl)
162 return NULL;
163
164 if (!nested_table_alloc(ht, (union nested_table __rcu **)tbl->buckets,
165 false)) {
166 kfree(tbl);
167 return NULL;
168 }
169
170 tbl->nest = (ilog2(nbuckets) - 1) % shift + 1;
171
172 return tbl;
173}
174
175static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
176 size_t nbuckets,
177 gfp_t gfp)
178{
179 struct bucket_table *tbl = NULL;
180 size_t size;
181 int i;
182 static struct lock_class_key __key;
183
184 tbl = kvzalloc(struct_size(tbl, buckets, nbuckets), gfp);
185
186 size = nbuckets;
187
188 if (tbl == NULL && (gfp & ~__GFP_NOFAIL) != GFP_KERNEL) {
189 tbl = nested_bucket_table_alloc(ht, nbuckets, gfp);
190 nbuckets = 0;
191 }
192
193 if (tbl == NULL)
194 return NULL;
195
196 lockdep_init_map(&tbl->dep_map, "rhashtable_bucket", &__key, 0);
197
198 tbl->size = size;
199
200 rcu_head_init(&tbl->rcu);
201 INIT_LIST_HEAD(&tbl->walkers);
202
203 tbl->hash_rnd = get_random_u32();
204
205 for (i = 0; i < nbuckets; i++)
206 INIT_RHT_NULLS_HEAD(tbl->buckets[i]);
207
208 return tbl;
209}
210
211static struct bucket_table *rhashtable_last_table(struct rhashtable *ht,
212 struct bucket_table *tbl)
213{
214 struct bucket_table *new_tbl;
215
216 do {
217 new_tbl = tbl;
218 tbl = rht_dereference_rcu(tbl->future_tbl, ht);
219 } while (tbl);
220
221 return new_tbl;
222}
223
224static int rhashtable_rehash_one(struct rhashtable *ht,
225 struct rhash_lock_head __rcu **bkt,
226 unsigned int old_hash)
227{
228 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
229 struct bucket_table *new_tbl = rhashtable_last_table(ht, old_tbl);
230 int err = -EAGAIN;
231 struct rhash_head *head, *next, *entry;
232 struct rhash_head __rcu **pprev = NULL;
233 unsigned int new_hash;
234
235 if (new_tbl->nest)
236 goto out;
237
238 err = -ENOENT;
239
240 rht_for_each_from(entry, rht_ptr(bkt, old_tbl, old_hash),
241 old_tbl, old_hash) {
242 err = 0;
243 next = rht_dereference_bucket(entry->next, old_tbl, old_hash);
244
245 if (rht_is_a_nulls(next))
246 break;
247
248 pprev = &entry->next;
249 }
250
251 if (err)
252 goto out;
253
254 new_hash = head_hashfn(ht, new_tbl, entry);
255
256 rht_lock_nested(new_tbl, &new_tbl->buckets[new_hash], SINGLE_DEPTH_NESTING);
257
258 head = rht_ptr(new_tbl->buckets + new_hash, new_tbl, new_hash);
259
260 RCU_INIT_POINTER(entry->next, head);
261
262 rht_assign_unlock(new_tbl, &new_tbl->buckets[new_hash], entry);
263
264 if (pprev)
265 rcu_assign_pointer(*pprev, next);
266 else
267 /* Need to preserved the bit lock. */
268 rht_assign_locked(bkt, next);
269
270out:
271 return err;
272}
273
274static int rhashtable_rehash_chain(struct rhashtable *ht,
275 unsigned int old_hash)
276{
277 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
278 struct rhash_lock_head __rcu **bkt = rht_bucket_var(old_tbl, old_hash);
279 int err;
280
281 if (!bkt)
282 return 0;
283 rht_lock(old_tbl, bkt);
284
285 while (!(err = rhashtable_rehash_one(ht, bkt, old_hash)))
286 ;
287
288 if (err == -ENOENT)
289 err = 0;
290 rht_unlock(old_tbl, bkt);
291
292 return err;
293}
294
295static int rhashtable_rehash_attach(struct rhashtable *ht,
296 struct bucket_table *old_tbl,
297 struct bucket_table *new_tbl)
298{
299 /* Make insertions go into the new, empty table right away. Deletions
300 * and lookups will be attempted in both tables until we synchronize.
301 * As cmpxchg() provides strong barriers, we do not need
302 * rcu_assign_pointer().
303 */
304
305 if (cmpxchg((struct bucket_table **)&old_tbl->future_tbl, NULL,
306 new_tbl) != NULL)
307 return -EEXIST;
308
309 return 0;
310}
311
312static int rhashtable_rehash_table(struct rhashtable *ht)
313{
314 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
315 struct bucket_table *new_tbl;
316 struct rhashtable_walker *walker;
317 unsigned int old_hash;
318 int err;
319
320 new_tbl = rht_dereference(old_tbl->future_tbl, ht);
321 if (!new_tbl)
322 return 0;
323
324 for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
325 err = rhashtable_rehash_chain(ht, old_hash);
326 if (err)
327 return err;
328 cond_resched();
329 }
330
331 /* Publish the new table pointer. */
332 rcu_assign_pointer(ht->tbl, new_tbl);
333
334 spin_lock(&ht->lock);
335 list_for_each_entry(walker, &old_tbl->walkers, list)
336 walker->tbl = NULL;
337
338 /* Wait for readers. All new readers will see the new
339 * table, and thus no references to the old table will
340 * remain.
341 * We do this inside the locked region so that
342 * rhashtable_walk_stop() can use rcu_head_after_call_rcu()
343 * to check if it should not re-link the table.
344 */
345 call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
346 spin_unlock(&ht->lock);
347
348 return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
349}
350
351static int rhashtable_rehash_alloc(struct rhashtable *ht,
352 struct bucket_table *old_tbl,
353 unsigned int size)
354{
355 struct bucket_table *new_tbl;
356 int err;
357
358 ASSERT_RHT_MUTEX(ht);
359
360 new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
361 if (new_tbl == NULL)
362 return -ENOMEM;
363
364 err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
365 if (err)
366 bucket_table_free(new_tbl);
367
368 return err;
369}
370
371/**
372 * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
373 * @ht: the hash table to shrink
374 *
375 * This function shrinks the hash table to fit, i.e., the smallest
376 * size would not cause it to expand right away automatically.
377 *
378 * The caller must ensure that no concurrent resizing occurs by holding
379 * ht->mutex.
380 *
381 * The caller must ensure that no concurrent table mutations take place.
382 * It is however valid to have concurrent lookups if they are RCU protected.
383 *
384 * It is valid to have concurrent insertions and deletions protected by per
385 * bucket locks or concurrent RCU protected lookups and traversals.
386 */
387static int rhashtable_shrink(struct rhashtable *ht)
388{
389 struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
390 unsigned int nelems = atomic_read(&ht->nelems);
391 unsigned int size = 0;
392
393 if (nelems)
394 size = roundup_pow_of_two(nelems * 3 / 2);
395 if (size < ht->p.min_size)
396 size = ht->p.min_size;
397
398 if (old_tbl->size <= size)
399 return 0;
400
401 if (rht_dereference(old_tbl->future_tbl, ht))
402 return -EEXIST;
403
404 return rhashtable_rehash_alloc(ht, old_tbl, size);
405}
406
407static void rht_deferred_worker(struct work_struct *work)
408{
409 struct rhashtable *ht;
410 struct bucket_table *tbl;
411 int err = 0;
412
413 ht = container_of(work, struct rhashtable, run_work);
414 mutex_lock(&ht->mutex);
415
416 tbl = rht_dereference(ht->tbl, ht);
417 tbl = rhashtable_last_table(ht, tbl);
418
419 if (rht_grow_above_75(ht, tbl))
420 err = rhashtable_rehash_alloc(ht, tbl, tbl->size * 2);
421 else if (ht->p.automatic_shrinking && rht_shrink_below_30(ht, tbl))
422 err = rhashtable_shrink(ht);
423 else if (tbl->nest)
424 err = rhashtable_rehash_alloc(ht, tbl, tbl->size);
425
426 if (!err || err == -EEXIST) {
427 int nerr;
428
429 nerr = rhashtable_rehash_table(ht);
430 err = err ?: nerr;
431 }
432
433 mutex_unlock(&ht->mutex);
434
435 if (err)
436 schedule_work(&ht->run_work);
437}
438
439static int rhashtable_insert_rehash(struct rhashtable *ht,
440 struct bucket_table *tbl)
441{
442 struct bucket_table *old_tbl;
443 struct bucket_table *new_tbl;
444 unsigned int size;
445 int err;
446
447 old_tbl = rht_dereference_rcu(ht->tbl, ht);
448
449 size = tbl->size;
450
451 err = -EBUSY;
452
453 if (rht_grow_above_75(ht, tbl))
454 size *= 2;
455 /* Do not schedule more than one rehash */
456 else if (old_tbl != tbl)
457 goto fail;
458
459 err = -ENOMEM;
460
461 new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC | __GFP_NOWARN);
462 if (new_tbl == NULL)
463 goto fail;
464
465 err = rhashtable_rehash_attach(ht, tbl, new_tbl);
466 if (err) {
467 bucket_table_free(new_tbl);
468 if (err == -EEXIST)
469 err = 0;
470 } else
471 schedule_work(&ht->run_work);
472
473 return err;
474
475fail:
476 /* Do not fail the insert if someone else did a rehash. */
477 if (likely(rcu_access_pointer(tbl->future_tbl)))
478 return 0;
479
480 /* Schedule async rehash to retry allocation in process context. */
481 if (err == -ENOMEM)
482 schedule_work(&ht->run_work);
483
484 return err;
485}
486
487static void *rhashtable_lookup_one(struct rhashtable *ht,
488 struct rhash_lock_head __rcu **bkt,
489 struct bucket_table *tbl, unsigned int hash,
490 const void *key, struct rhash_head *obj)
491{
492 struct rhashtable_compare_arg arg = {
493 .ht = ht,
494 .key = key,
495 };
496 struct rhash_head __rcu **pprev = NULL;
497 struct rhash_head *head;
498 int elasticity;
499
500 elasticity = RHT_ELASTICITY;
501 rht_for_each_from(head, rht_ptr(bkt, tbl, hash), tbl, hash) {
502 struct rhlist_head *list;
503 struct rhlist_head *plist;
504
505 elasticity--;
506 if (!key ||
507 (ht->p.obj_cmpfn ?
508 ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
509 rhashtable_compare(&arg, rht_obj(ht, head)))) {
510 pprev = &head->next;
511 continue;
512 }
513
514 if (!ht->rhlist)
515 return rht_obj(ht, head);
516
517 list = container_of(obj, struct rhlist_head, rhead);
518 plist = container_of(head, struct rhlist_head, rhead);
519
520 RCU_INIT_POINTER(list->next, plist);
521 head = rht_dereference_bucket(head->next, tbl, hash);
522 RCU_INIT_POINTER(list->rhead.next, head);
523 if (pprev)
524 rcu_assign_pointer(*pprev, obj);
525 else
526 /* Need to preserve the bit lock */
527 rht_assign_locked(bkt, obj);
528
529 return NULL;
530 }
531
532 if (elasticity <= 0)
533 return ERR_PTR(-EAGAIN);
534
535 return ERR_PTR(-ENOENT);
536}
537
538static struct bucket_table *rhashtable_insert_one(
539 struct rhashtable *ht, struct rhash_lock_head __rcu **bkt,
540 struct bucket_table *tbl, unsigned int hash, struct rhash_head *obj,
541 void *data)
542{
543 struct bucket_table *new_tbl;
544 struct rhash_head *head;
545
546 if (!IS_ERR_OR_NULL(data))
547 return ERR_PTR(-EEXIST);
548
549 if (PTR_ERR(data) != -EAGAIN && PTR_ERR(data) != -ENOENT)
550 return ERR_CAST(data);
551
552 new_tbl = rht_dereference_rcu(tbl->future_tbl, ht);
553 if (new_tbl)
554 return new_tbl;
555
556 if (PTR_ERR(data) != -ENOENT)
557 return ERR_CAST(data);
558
559 if (unlikely(rht_grow_above_max(ht, tbl)))
560 return ERR_PTR(-E2BIG);
561
562 if (unlikely(rht_grow_above_100(ht, tbl)))
563 return ERR_PTR(-EAGAIN);
564
565 head = rht_ptr(bkt, tbl, hash);
566
567 RCU_INIT_POINTER(obj->next, head);
568 if (ht->rhlist) {
569 struct rhlist_head *list;
570
571 list = container_of(obj, struct rhlist_head, rhead);
572 RCU_INIT_POINTER(list->next, NULL);
573 }
574
575 /* bkt is always the head of the list, so it holds
576 * the lock, which we need to preserve
577 */
578 rht_assign_locked(bkt, obj);
579
580 atomic_inc(&ht->nelems);
581 if (rht_grow_above_75(ht, tbl))
582 schedule_work(&ht->run_work);
583
584 return NULL;
585}
586
587static void *rhashtable_try_insert(struct rhashtable *ht, const void *key,
588 struct rhash_head *obj)
589{
590 struct bucket_table *new_tbl;
591 struct bucket_table *tbl;
592 struct rhash_lock_head __rcu **bkt;
593 unsigned int hash;
594 void *data;
595
596 new_tbl = rcu_dereference(ht->tbl);
597
598 do {
599 tbl = new_tbl;
600 hash = rht_head_hashfn(ht, tbl, obj, ht->p);
601 if (rcu_access_pointer(tbl->future_tbl))
602 /* Failure is OK */
603 bkt = rht_bucket_var(tbl, hash);
604 else
605 bkt = rht_bucket_insert(ht, tbl, hash);
606 if (bkt == NULL) {
607 new_tbl = rht_dereference_rcu(tbl->future_tbl, ht);
608 data = ERR_PTR(-EAGAIN);
609 } else {
610 rht_lock(tbl, bkt);
611 data = rhashtable_lookup_one(ht, bkt, tbl,
612 hash, key, obj);
613 new_tbl = rhashtable_insert_one(ht, bkt, tbl,
614 hash, obj, data);
615 if (PTR_ERR(new_tbl) != -EEXIST)
616 data = ERR_CAST(new_tbl);
617
618 rht_unlock(tbl, bkt);
619 }
620 } while (!IS_ERR_OR_NULL(new_tbl));
621
622 if (PTR_ERR(data) == -EAGAIN)
623 data = ERR_PTR(rhashtable_insert_rehash(ht, tbl) ?:
624 -EAGAIN);
625
626 return data;
627}
628
629void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
630 struct rhash_head *obj)
631{
632 void *data;
633
634 do {
635 rcu_read_lock();
636 data = rhashtable_try_insert(ht, key, obj);
637 rcu_read_unlock();
638 } while (PTR_ERR(data) == -EAGAIN);
639
640 return data;
641}
642EXPORT_SYMBOL_GPL(rhashtable_insert_slow);
643
644/**
645 * rhashtable_walk_enter - Initialise an iterator
646 * @ht: Table to walk over
647 * @iter: Hash table Iterator
648 *
649 * This function prepares a hash table walk.
650 *
651 * Note that if you restart a walk after rhashtable_walk_stop you
652 * may see the same object twice. Also, you may miss objects if
653 * there are removals in between rhashtable_walk_stop and the next
654 * call to rhashtable_walk_start.
655 *
656 * For a completely stable walk you should construct your own data
657 * structure outside the hash table.
658 *
659 * This function may be called from any process context, including
660 * non-preemptable context, but cannot be called from softirq or
661 * hardirq context.
662 *
663 * You must call rhashtable_walk_exit after this function returns.
664 */
665void rhashtable_walk_enter(struct rhashtable *ht, struct rhashtable_iter *iter)
666{
667 iter->ht = ht;
668 iter->p = NULL;
669 iter->slot = 0;
670 iter->skip = 0;
671 iter->end_of_table = 0;
672
673 spin_lock(&ht->lock);
674 iter->walker.tbl =
675 rcu_dereference_protected(ht->tbl, lockdep_is_held(&ht->lock));
676 list_add(&iter->walker.list, &iter->walker.tbl->walkers);
677 spin_unlock(&ht->lock);
678}
679EXPORT_SYMBOL_GPL(rhashtable_walk_enter);
680
681/**
682 * rhashtable_walk_exit - Free an iterator
683 * @iter: Hash table Iterator
684 *
685 * This function frees resources allocated by rhashtable_walk_enter.
686 */
687void rhashtable_walk_exit(struct rhashtable_iter *iter)
688{
689 spin_lock(&iter->ht->lock);
690 if (iter->walker.tbl)
691 list_del(&iter->walker.list);
692 spin_unlock(&iter->ht->lock);
693}
694EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
695
696/**
697 * rhashtable_walk_start_check - Start a hash table walk
698 * @iter: Hash table iterator
699 *
700 * Start a hash table walk at the current iterator position. Note that we take
701 * the RCU lock in all cases including when we return an error. So you must
702 * always call rhashtable_walk_stop to clean up.
703 *
704 * Returns zero if successful.
705 *
706 * Returns -EAGAIN if resize event occured. Note that the iterator
707 * will rewind back to the beginning and you may use it immediately
708 * by calling rhashtable_walk_next.
709 *
710 * rhashtable_walk_start is defined as an inline variant that returns
711 * void. This is preferred in cases where the caller would ignore
712 * resize events and always continue.
713 */
714int rhashtable_walk_start_check(struct rhashtable_iter *iter)
715 __acquires(RCU)
716{
717 struct rhashtable *ht = iter->ht;
718 bool rhlist = ht->rhlist;
719
720 rcu_read_lock();
721
722 spin_lock(&ht->lock);
723 if (iter->walker.tbl)
724 list_del(&iter->walker.list);
725 spin_unlock(&ht->lock);
726
727 if (iter->end_of_table)
728 return 0;
729 if (!iter->walker.tbl) {
730 iter->walker.tbl = rht_dereference_rcu(ht->tbl, ht);
731 iter->slot = 0;
732 iter->skip = 0;
733 return -EAGAIN;
734 }
735
736 if (iter->p && !rhlist) {
737 /*
738 * We need to validate that 'p' is still in the table, and
739 * if so, update 'skip'
740 */
741 struct rhash_head *p;
742 int skip = 0;
743 rht_for_each_rcu(p, iter->walker.tbl, iter->slot) {
744 skip++;
745 if (p == iter->p) {
746 iter->skip = skip;
747 goto found;
748 }
749 }
750 iter->p = NULL;
751 } else if (iter->p && rhlist) {
752 /* Need to validate that 'list' is still in the table, and
753 * if so, update 'skip' and 'p'.
754 */
755 struct rhash_head *p;
756 struct rhlist_head *list;
757 int skip = 0;
758 rht_for_each_rcu(p, iter->walker.tbl, iter->slot) {
759 for (list = container_of(p, struct rhlist_head, rhead);
760 list;
761 list = rcu_dereference(list->next)) {
762 skip++;
763 if (list == iter->list) {
764 iter->p = p;
765 iter->skip = skip;
766 goto found;
767 }
768 }
769 }
770 iter->p = NULL;
771 }
772found:
773 return 0;
774}
775EXPORT_SYMBOL_GPL(rhashtable_walk_start_check);
776
777/**
778 * __rhashtable_walk_find_next - Find the next element in a table (or the first
779 * one in case of a new walk).
780 *
781 * @iter: Hash table iterator
782 *
783 * Returns the found object or NULL when the end of the table is reached.
784 *
785 * Returns -EAGAIN if resize event occurred.
786 */
787static void *__rhashtable_walk_find_next(struct rhashtable_iter *iter)
788{
789 struct bucket_table *tbl = iter->walker.tbl;
790 struct rhlist_head *list = iter->list;
791 struct rhashtable *ht = iter->ht;
792 struct rhash_head *p = iter->p;
793 bool rhlist = ht->rhlist;
794
795 if (!tbl)
796 return NULL;
797
798 for (; iter->slot < tbl->size; iter->slot++) {
799 int skip = iter->skip;
800
801 rht_for_each_rcu(p, tbl, iter->slot) {
802 if (rhlist) {
803 list = container_of(p, struct rhlist_head,
804 rhead);
805 do {
806 if (!skip)
807 goto next;
808 skip--;
809 list = rcu_dereference(list->next);
810 } while (list);
811
812 continue;
813 }
814 if (!skip)
815 break;
816 skip--;
817 }
818
819next:
820 if (!rht_is_a_nulls(p)) {
821 iter->skip++;
822 iter->p = p;
823 iter->list = list;
824 return rht_obj(ht, rhlist ? &list->rhead : p);
825 }
826
827 iter->skip = 0;
828 }
829
830 iter->p = NULL;
831
832 /* Ensure we see any new tables. */
833 smp_rmb();
834
835 iter->walker.tbl = rht_dereference_rcu(tbl->future_tbl, ht);
836 if (iter->walker.tbl) {
837 iter->slot = 0;
838 iter->skip = 0;
839 return ERR_PTR(-EAGAIN);
840 } else {
841 iter->end_of_table = true;
842 }
843
844 return NULL;
845}
846
847/**
848 * rhashtable_walk_next - Return the next object and advance the iterator
849 * @iter: Hash table iterator
850 *
851 * Note that you must call rhashtable_walk_stop when you are finished
852 * with the walk.
853 *
854 * Returns the next object or NULL when the end of the table is reached.
855 *
856 * Returns -EAGAIN if resize event occurred. Note that the iterator
857 * will rewind back to the beginning and you may continue to use it.
858 */
859void *rhashtable_walk_next(struct rhashtable_iter *iter)
860{
861 struct rhlist_head *list = iter->list;
862 struct rhashtable *ht = iter->ht;
863 struct rhash_head *p = iter->p;
864 bool rhlist = ht->rhlist;
865
866 if (p) {
867 if (!rhlist || !(list = rcu_dereference(list->next))) {
868 p = rcu_dereference(p->next);
869 list = container_of(p, struct rhlist_head, rhead);
870 }
871 if (!rht_is_a_nulls(p)) {
872 iter->skip++;
873 iter->p = p;
874 iter->list = list;
875 return rht_obj(ht, rhlist ? &list->rhead : p);
876 }
877
878 /* At the end of this slot, switch to next one and then find
879 * next entry from that point.
880 */
881 iter->skip = 0;
882 iter->slot++;
883 }
884
885 return __rhashtable_walk_find_next(iter);
886}
887EXPORT_SYMBOL_GPL(rhashtable_walk_next);
888
889/**
890 * rhashtable_walk_peek - Return the next object but don't advance the iterator
891 * @iter: Hash table iterator
892 *
893 * Returns the next object or NULL when the end of the table is reached.
894 *
895 * Returns -EAGAIN if resize event occurred. Note that the iterator
896 * will rewind back to the beginning and you may continue to use it.
897 */
898void *rhashtable_walk_peek(struct rhashtable_iter *iter)
899{
900 struct rhlist_head *list = iter->list;
901 struct rhashtable *ht = iter->ht;
902 struct rhash_head *p = iter->p;
903
904 if (p)
905 return rht_obj(ht, ht->rhlist ? &list->rhead : p);
906
907 /* No object found in current iter, find next one in the table. */
908
909 if (iter->skip) {
910 /* A nonzero skip value points to the next entry in the table
911 * beyond that last one that was found. Decrement skip so
912 * we find the current value. __rhashtable_walk_find_next
913 * will restore the original value of skip assuming that
914 * the table hasn't changed.
915 */
916 iter->skip--;
917 }
918
919 return __rhashtable_walk_find_next(iter);
920}
921EXPORT_SYMBOL_GPL(rhashtable_walk_peek);
922
923/**
924 * rhashtable_walk_stop - Finish a hash table walk
925 * @iter: Hash table iterator
926 *
927 * Finish a hash table walk. Does not reset the iterator to the start of the
928 * hash table.
929 */
930void rhashtable_walk_stop(struct rhashtable_iter *iter)
931 __releases(RCU)
932{
933 struct rhashtable *ht;
934 struct bucket_table *tbl = iter->walker.tbl;
935
936 if (!tbl)
937 goto out;
938
939 ht = iter->ht;
940
941 spin_lock(&ht->lock);
942 if (rcu_head_after_call_rcu(&tbl->rcu, bucket_table_free_rcu))
943 /* This bucket table is being freed, don't re-link it. */
944 iter->walker.tbl = NULL;
945 else
946 list_add(&iter->walker.list, &tbl->walkers);
947 spin_unlock(&ht->lock);
948
949out:
950 rcu_read_unlock();
951}
952EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
953
954static size_t rounded_hashtable_size(const struct rhashtable_params *params)
955{
956 size_t retsize;
957
958 if (params->nelem_hint)
959 retsize = max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
960 (unsigned long)params->min_size);
961 else
962 retsize = max(HASH_DEFAULT_SIZE,
963 (unsigned long)params->min_size);
964
965 return retsize;
966}
967
968static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
969{
970 return jhash2(key, length, seed);
971}
972
973/**
974 * rhashtable_init - initialize a new hash table
975 * @ht: hash table to be initialized
976 * @params: configuration parameters
977 *
978 * Initializes a new hash table based on the provided configuration
979 * parameters. A table can be configured either with a variable or
980 * fixed length key:
981 *
982 * Configuration Example 1: Fixed length keys
983 * struct test_obj {
984 * int key;
985 * void * my_member;
986 * struct rhash_head node;
987 * };
988 *
989 * struct rhashtable_params params = {
990 * .head_offset = offsetof(struct test_obj, node),
991 * .key_offset = offsetof(struct test_obj, key),
992 * .key_len = sizeof(int),
993 * .hashfn = jhash,
994 * };
995 *
996 * Configuration Example 2: Variable length keys
997 * struct test_obj {
998 * [...]
999 * struct rhash_head node;
1000 * };
1001 *
1002 * u32 my_hash_fn(const void *data, u32 len, u32 seed)
1003 * {
1004 * struct test_obj *obj = data;
1005 *
1006 * return [... hash ...];
1007 * }
1008 *
1009 * struct rhashtable_params params = {
1010 * .head_offset = offsetof(struct test_obj, node),
1011 * .hashfn = jhash,
1012 * .obj_hashfn = my_hash_fn,
1013 * };
1014 */
1015int rhashtable_init(struct rhashtable *ht,
1016 const struct rhashtable_params *params)
1017{
1018 struct bucket_table *tbl;
1019 size_t size;
1020
1021 if ((!params->key_len && !params->obj_hashfn) ||
1022 (params->obj_hashfn && !params->obj_cmpfn))
1023 return -EINVAL;
1024
1025 memset(ht, 0, sizeof(*ht));
1026 mutex_init(&ht->mutex);
1027 spin_lock_init(&ht->lock);
1028 memcpy(&ht->p, params, sizeof(*params));
1029
1030 if (params->min_size)
1031 ht->p.min_size = roundup_pow_of_two(params->min_size);
1032
1033 /* Cap total entries at 2^31 to avoid nelems overflow. */
1034 ht->max_elems = 1u << 31;
1035
1036 if (params->max_size) {
1037 ht->p.max_size = rounddown_pow_of_two(params->max_size);
1038 if (ht->p.max_size < ht->max_elems / 2)
1039 ht->max_elems = ht->p.max_size * 2;
1040 }
1041
1042 ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
1043
1044 size = rounded_hashtable_size(&ht->p);
1045
1046 ht->key_len = ht->p.key_len;
1047 if (!params->hashfn) {
1048 ht->p.hashfn = jhash;
1049
1050 if (!(ht->key_len & (sizeof(u32) - 1))) {
1051 ht->key_len /= sizeof(u32);
1052 ht->p.hashfn = rhashtable_jhash2;
1053 }
1054 }
1055
1056 /*
1057 * This is api initialization and thus we need to guarantee the
1058 * initial rhashtable allocation. Upon failure, retry with the
1059 * smallest possible size with __GFP_NOFAIL semantics.
1060 */
1061 tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
1062 if (unlikely(tbl == NULL)) {
1063 size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
1064 tbl = bucket_table_alloc(ht, size, GFP_KERNEL | __GFP_NOFAIL);
1065 }
1066
1067 atomic_set(&ht->nelems, 0);
1068
1069 RCU_INIT_POINTER(ht->tbl, tbl);
1070
1071 INIT_WORK(&ht->run_work, rht_deferred_worker);
1072
1073 return 0;
1074}
1075EXPORT_SYMBOL_GPL(rhashtable_init);
1076
1077/**
1078 * rhltable_init - initialize a new hash list table
1079 * @hlt: hash list table to be initialized
1080 * @params: configuration parameters
1081 *
1082 * Initializes a new hash list table.
1083 *
1084 * See documentation for rhashtable_init.
1085 */
1086int rhltable_init(struct rhltable *hlt, const struct rhashtable_params *params)
1087{
1088 int err;
1089
1090 err = rhashtable_init(&hlt->ht, params);
1091 hlt->ht.rhlist = true;
1092 return err;
1093}
1094EXPORT_SYMBOL_GPL(rhltable_init);
1095
1096static void rhashtable_free_one(struct rhashtable *ht, struct rhash_head *obj,
1097 void (*free_fn)(void *ptr, void *arg),
1098 void *arg)
1099{
1100 struct rhlist_head *list;
1101
1102 if (!ht->rhlist) {
1103 free_fn(rht_obj(ht, obj), arg);
1104 return;
1105 }
1106
1107 list = container_of(obj, struct rhlist_head, rhead);
1108 do {
1109 obj = &list->rhead;
1110 list = rht_dereference(list->next, ht);
1111 free_fn(rht_obj(ht, obj), arg);
1112 } while (list);
1113}
1114
1115/**
1116 * rhashtable_free_and_destroy - free elements and destroy hash table
1117 * @ht: the hash table to destroy
1118 * @free_fn: callback to release resources of element
1119 * @arg: pointer passed to free_fn
1120 *
1121 * Stops an eventual async resize. If defined, invokes free_fn for each
1122 * element to releasal resources. Please note that RCU protected
1123 * readers may still be accessing the elements. Releasing of resources
1124 * must occur in a compatible manner. Then frees the bucket array.
1125 *
1126 * This function will eventually sleep to wait for an async resize
1127 * to complete. The caller is responsible that no further write operations
1128 * occurs in parallel.
1129 */
1130void rhashtable_free_and_destroy(struct rhashtable *ht,
1131 void (*free_fn)(void *ptr, void *arg),
1132 void *arg)
1133{
1134 struct bucket_table *tbl, *next_tbl;
1135 unsigned int i;
1136
1137 cancel_work_sync(&ht->run_work);
1138
1139 mutex_lock(&ht->mutex);
1140 tbl = rht_dereference(ht->tbl, ht);
1141restart:
1142 if (free_fn) {
1143 for (i = 0; i < tbl->size; i++) {
1144 struct rhash_head *pos, *next;
1145
1146 cond_resched();
1147 for (pos = rht_ptr_exclusive(rht_bucket(tbl, i)),
1148 next = !rht_is_a_nulls(pos) ?
1149 rht_dereference(pos->next, ht) : NULL;
1150 !rht_is_a_nulls(pos);
1151 pos = next,
1152 next = !rht_is_a_nulls(pos) ?
1153 rht_dereference(pos->next, ht) : NULL)
1154 rhashtable_free_one(ht, pos, free_fn, arg);
1155 }
1156 }
1157
1158 next_tbl = rht_dereference(tbl->future_tbl, ht);
1159 bucket_table_free(tbl);
1160 if (next_tbl) {
1161 tbl = next_tbl;
1162 goto restart;
1163 }
1164 mutex_unlock(&ht->mutex);
1165}
1166EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
1167
1168void rhashtable_destroy(struct rhashtable *ht)
1169{
1170 return rhashtable_free_and_destroy(ht, NULL, NULL);
1171}
1172EXPORT_SYMBOL_GPL(rhashtable_destroy);
1173
1174struct rhash_lock_head __rcu **__rht_bucket_nested(
1175 const struct bucket_table *tbl, unsigned int hash)
1176{
1177 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1178 unsigned int index = hash & ((1 << tbl->nest) - 1);
1179 unsigned int size = tbl->size >> tbl->nest;
1180 unsigned int subhash = hash;
1181 union nested_table *ntbl;
1182
1183 ntbl = nested_table_top(tbl);
1184 ntbl = rht_dereference_bucket_rcu(ntbl[index].table, tbl, hash);
1185 subhash >>= tbl->nest;
1186
1187 while (ntbl && size > (1 << shift)) {
1188 index = subhash & ((1 << shift) - 1);
1189 ntbl = rht_dereference_bucket_rcu(ntbl[index].table,
1190 tbl, hash);
1191 size >>= shift;
1192 subhash >>= shift;
1193 }
1194
1195 if (!ntbl)
1196 return NULL;
1197
1198 return &ntbl[subhash].bucket;
1199
1200}
1201EXPORT_SYMBOL_GPL(__rht_bucket_nested);
1202
1203struct rhash_lock_head __rcu **rht_bucket_nested(
1204 const struct bucket_table *tbl, unsigned int hash)
1205{
1206 static struct rhash_lock_head __rcu *rhnull;
1207
1208 if (!rhnull)
1209 INIT_RHT_NULLS_HEAD(rhnull);
1210 return __rht_bucket_nested(tbl, hash) ?: &rhnull;
1211}
1212EXPORT_SYMBOL_GPL(rht_bucket_nested);
1213
1214struct rhash_lock_head __rcu **rht_bucket_nested_insert(
1215 struct rhashtable *ht, struct bucket_table *tbl, unsigned int hash)
1216{
1217 const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1218 unsigned int index = hash & ((1 << tbl->nest) - 1);
1219 unsigned int size = tbl->size >> tbl->nest;
1220 union nested_table *ntbl;
1221
1222 ntbl = nested_table_top(tbl);
1223 hash >>= tbl->nest;
1224 ntbl = nested_table_alloc(ht, &ntbl[index].table,
1225 size <= (1 << shift));
1226
1227 while (ntbl && size > (1 << shift)) {
1228 index = hash & ((1 << shift) - 1);
1229 size >>= shift;
1230 hash >>= shift;
1231 ntbl = nested_table_alloc(ht, &ntbl[index].table,
1232 size <= (1 << shift));
1233 }
1234
1235 if (!ntbl)
1236 return NULL;
1237
1238 return &ntbl[hash].bucket;
1239
1240}
1241EXPORT_SYMBOL_GPL(rht_bucket_nested_insert);