xref: /openbmc/linux/lib/rhashtable.c (revision 051d4420)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Resizable, Scalable, Concurrent Hash Table
4  *
5  * Copyright (c) 2015 Herbert Xu <herbert@gondor.apana.org.au>
6  * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
7  * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
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 
32 union nested_table {
33 	union nested_table __rcu *table;
34 	struct rhash_lock_head __rcu *bucket;
35 };
36 
37 static 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 
47 int lockdep_rht_mutex_is_held(struct rhashtable *ht)
48 {
49 	return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
50 }
51 EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
52 
53 int 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 }
61 EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
62 #else
63 #define ASSERT_RHT_MUTEX(HT)
64 #endif
65 
66 static 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 
75 static 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 
94 static 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 
109 static 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 
117 static void bucket_table_free_rcu(struct rcu_head *head)
118 {
119 	bucket_table_free(container_of(head, struct bucket_table, rcu));
120 }
121 
122 static 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 
147 static 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 
175 static 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 
211 static 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 
224 static 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 	unsigned long flags;
235 
236 	if (new_tbl->nest)
237 		goto out;
238 
239 	err = -ENOENT;
240 
241 	rht_for_each_from(entry, rht_ptr(bkt, old_tbl, old_hash),
242 			  old_tbl, old_hash) {
243 		err = 0;
244 		next = rht_dereference_bucket(entry->next, old_tbl, old_hash);
245 
246 		if (rht_is_a_nulls(next))
247 			break;
248 
249 		pprev = &entry->next;
250 	}
251 
252 	if (err)
253 		goto out;
254 
255 	new_hash = head_hashfn(ht, new_tbl, entry);
256 
257 	flags = rht_lock_nested(new_tbl, &new_tbl->buckets[new_hash],
258 				SINGLE_DEPTH_NESTING);
259 
260 	head = rht_ptr(new_tbl->buckets + new_hash, new_tbl, new_hash);
261 
262 	RCU_INIT_POINTER(entry->next, head);
263 
264 	rht_assign_unlock(new_tbl, &new_tbl->buckets[new_hash], entry, flags);
265 
266 	if (pprev)
267 		rcu_assign_pointer(*pprev, next);
268 	else
269 		/* Need to preserved the bit lock. */
270 		rht_assign_locked(bkt, next);
271 
272 out:
273 	return err;
274 }
275 
276 static int rhashtable_rehash_chain(struct rhashtable *ht,
277 				    unsigned int old_hash)
278 {
279 	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
280 	struct rhash_lock_head __rcu **bkt = rht_bucket_var(old_tbl, old_hash);
281 	unsigned long flags;
282 	int err;
283 
284 	if (!bkt)
285 		return 0;
286 	flags = rht_lock(old_tbl, bkt);
287 
288 	while (!(err = rhashtable_rehash_one(ht, bkt, old_hash)))
289 		;
290 
291 	if (err == -ENOENT)
292 		err = 0;
293 	rht_unlock(old_tbl, bkt, flags);
294 
295 	return err;
296 }
297 
298 static int rhashtable_rehash_attach(struct rhashtable *ht,
299 				    struct bucket_table *old_tbl,
300 				    struct bucket_table *new_tbl)
301 {
302 	/* Make insertions go into the new, empty table right away. Deletions
303 	 * and lookups will be attempted in both tables until we synchronize.
304 	 * As cmpxchg() provides strong barriers, we do not need
305 	 * rcu_assign_pointer().
306 	 */
307 
308 	if (cmpxchg((struct bucket_table **)&old_tbl->future_tbl, NULL,
309 		    new_tbl) != NULL)
310 		return -EEXIST;
311 
312 	return 0;
313 }
314 
315 static int rhashtable_rehash_table(struct rhashtable *ht)
316 {
317 	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
318 	struct bucket_table *new_tbl;
319 	struct rhashtable_walker *walker;
320 	unsigned int old_hash;
321 	int err;
322 
323 	new_tbl = rht_dereference(old_tbl->future_tbl, ht);
324 	if (!new_tbl)
325 		return 0;
326 
327 	for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
328 		err = rhashtable_rehash_chain(ht, old_hash);
329 		if (err)
330 			return err;
331 		cond_resched();
332 	}
333 
334 	/* Publish the new table pointer. */
335 	rcu_assign_pointer(ht->tbl, new_tbl);
336 
337 	spin_lock(&ht->lock);
338 	list_for_each_entry(walker, &old_tbl->walkers, list)
339 		walker->tbl = NULL;
340 
341 	/* Wait for readers. All new readers will see the new
342 	 * table, and thus no references to the old table will
343 	 * remain.
344 	 * We do this inside the locked region so that
345 	 * rhashtable_walk_stop() can use rcu_head_after_call_rcu()
346 	 * to check if it should not re-link the table.
347 	 */
348 	call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
349 	spin_unlock(&ht->lock);
350 
351 	return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
352 }
353 
354 static int rhashtable_rehash_alloc(struct rhashtable *ht,
355 				   struct bucket_table *old_tbl,
356 				   unsigned int size)
357 {
358 	struct bucket_table *new_tbl;
359 	int err;
360 
361 	ASSERT_RHT_MUTEX(ht);
362 
363 	new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
364 	if (new_tbl == NULL)
365 		return -ENOMEM;
366 
367 	err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
368 	if (err)
369 		bucket_table_free(new_tbl);
370 
371 	return err;
372 }
373 
374 /**
375  * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
376  * @ht:		the hash table to shrink
377  *
378  * This function shrinks the hash table to fit, i.e., the smallest
379  * size would not cause it to expand right away automatically.
380  *
381  * The caller must ensure that no concurrent resizing occurs by holding
382  * ht->mutex.
383  *
384  * The caller must ensure that no concurrent table mutations take place.
385  * It is however valid to have concurrent lookups if they are RCU protected.
386  *
387  * It is valid to have concurrent insertions and deletions protected by per
388  * bucket locks or concurrent RCU protected lookups and traversals.
389  */
390 static int rhashtable_shrink(struct rhashtable *ht)
391 {
392 	struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
393 	unsigned int nelems = atomic_read(&ht->nelems);
394 	unsigned int size = 0;
395 
396 	if (nelems)
397 		size = roundup_pow_of_two(nelems * 3 / 2);
398 	if (size < ht->p.min_size)
399 		size = ht->p.min_size;
400 
401 	if (old_tbl->size <= size)
402 		return 0;
403 
404 	if (rht_dereference(old_tbl->future_tbl, ht))
405 		return -EEXIST;
406 
407 	return rhashtable_rehash_alloc(ht, old_tbl, size);
408 }
409 
410 static void rht_deferred_worker(struct work_struct *work)
411 {
412 	struct rhashtable *ht;
413 	struct bucket_table *tbl;
414 	int err = 0;
415 
416 	ht = container_of(work, struct rhashtable, run_work);
417 	mutex_lock(&ht->mutex);
418 
419 	tbl = rht_dereference(ht->tbl, ht);
420 	tbl = rhashtable_last_table(ht, tbl);
421 
422 	if (rht_grow_above_75(ht, tbl))
423 		err = rhashtable_rehash_alloc(ht, tbl, tbl->size * 2);
424 	else if (ht->p.automatic_shrinking && rht_shrink_below_30(ht, tbl))
425 		err = rhashtable_shrink(ht);
426 	else if (tbl->nest)
427 		err = rhashtable_rehash_alloc(ht, tbl, tbl->size);
428 
429 	if (!err || err == -EEXIST) {
430 		int nerr;
431 
432 		nerr = rhashtable_rehash_table(ht);
433 		err = err ?: nerr;
434 	}
435 
436 	mutex_unlock(&ht->mutex);
437 
438 	if (err)
439 		schedule_work(&ht->run_work);
440 }
441 
442 static int rhashtable_insert_rehash(struct rhashtable *ht,
443 				    struct bucket_table *tbl)
444 {
445 	struct bucket_table *old_tbl;
446 	struct bucket_table *new_tbl;
447 	unsigned int size;
448 	int err;
449 
450 	old_tbl = rht_dereference_rcu(ht->tbl, ht);
451 
452 	size = tbl->size;
453 
454 	err = -EBUSY;
455 
456 	if (rht_grow_above_75(ht, tbl))
457 		size *= 2;
458 	/* Do not schedule more than one rehash */
459 	else if (old_tbl != tbl)
460 		goto fail;
461 
462 	err = -ENOMEM;
463 
464 	new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC | __GFP_NOWARN);
465 	if (new_tbl == NULL)
466 		goto fail;
467 
468 	err = rhashtable_rehash_attach(ht, tbl, new_tbl);
469 	if (err) {
470 		bucket_table_free(new_tbl);
471 		if (err == -EEXIST)
472 			err = 0;
473 	} else
474 		schedule_work(&ht->run_work);
475 
476 	return err;
477 
478 fail:
479 	/* Do not fail the insert if someone else did a rehash. */
480 	if (likely(rcu_access_pointer(tbl->future_tbl)))
481 		return 0;
482 
483 	/* Schedule async rehash to retry allocation in process context. */
484 	if (err == -ENOMEM)
485 		schedule_work(&ht->run_work);
486 
487 	return err;
488 }
489 
490 static void *rhashtable_lookup_one(struct rhashtable *ht,
491 				   struct rhash_lock_head __rcu **bkt,
492 				   struct bucket_table *tbl, unsigned int hash,
493 				   const void *key, struct rhash_head *obj)
494 {
495 	struct rhashtable_compare_arg arg = {
496 		.ht = ht,
497 		.key = key,
498 	};
499 	struct rhash_head __rcu **pprev = NULL;
500 	struct rhash_head *head;
501 	int elasticity;
502 
503 	elasticity = RHT_ELASTICITY;
504 	rht_for_each_from(head, rht_ptr(bkt, tbl, hash), tbl, hash) {
505 		struct rhlist_head *list;
506 		struct rhlist_head *plist;
507 
508 		elasticity--;
509 		if (!key ||
510 		    (ht->p.obj_cmpfn ?
511 		     ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
512 		     rhashtable_compare(&arg, rht_obj(ht, head)))) {
513 			pprev = &head->next;
514 			continue;
515 		}
516 
517 		if (!ht->rhlist)
518 			return rht_obj(ht, head);
519 
520 		list = container_of(obj, struct rhlist_head, rhead);
521 		plist = container_of(head, struct rhlist_head, rhead);
522 
523 		RCU_INIT_POINTER(list->next, plist);
524 		head = rht_dereference_bucket(head->next, tbl, hash);
525 		RCU_INIT_POINTER(list->rhead.next, head);
526 		if (pprev)
527 			rcu_assign_pointer(*pprev, obj);
528 		else
529 			/* Need to preserve the bit lock */
530 			rht_assign_locked(bkt, obj);
531 
532 		return NULL;
533 	}
534 
535 	if (elasticity <= 0)
536 		return ERR_PTR(-EAGAIN);
537 
538 	return ERR_PTR(-ENOENT);
539 }
540 
541 static struct bucket_table *rhashtable_insert_one(
542 	struct rhashtable *ht, struct rhash_lock_head __rcu **bkt,
543 	struct bucket_table *tbl, unsigned int hash, struct rhash_head *obj,
544 	void *data)
545 {
546 	struct bucket_table *new_tbl;
547 	struct rhash_head *head;
548 
549 	if (!IS_ERR_OR_NULL(data))
550 		return ERR_PTR(-EEXIST);
551 
552 	if (PTR_ERR(data) != -EAGAIN && PTR_ERR(data) != -ENOENT)
553 		return ERR_CAST(data);
554 
555 	new_tbl = rht_dereference_rcu(tbl->future_tbl, ht);
556 	if (new_tbl)
557 		return new_tbl;
558 
559 	if (PTR_ERR(data) != -ENOENT)
560 		return ERR_CAST(data);
561 
562 	if (unlikely(rht_grow_above_max(ht, tbl)))
563 		return ERR_PTR(-E2BIG);
564 
565 	if (unlikely(rht_grow_above_100(ht, tbl)))
566 		return ERR_PTR(-EAGAIN);
567 
568 	head = rht_ptr(bkt, tbl, hash);
569 
570 	RCU_INIT_POINTER(obj->next, head);
571 	if (ht->rhlist) {
572 		struct rhlist_head *list;
573 
574 		list = container_of(obj, struct rhlist_head, rhead);
575 		RCU_INIT_POINTER(list->next, NULL);
576 	}
577 
578 	/* bkt is always the head of the list, so it holds
579 	 * the lock, which we need to preserve
580 	 */
581 	rht_assign_locked(bkt, obj);
582 
583 	atomic_inc(&ht->nelems);
584 	if (rht_grow_above_75(ht, tbl))
585 		schedule_work(&ht->run_work);
586 
587 	return NULL;
588 }
589 
590 static void *rhashtable_try_insert(struct rhashtable *ht, const void *key,
591 				   struct rhash_head *obj)
592 {
593 	struct bucket_table *new_tbl;
594 	struct bucket_table *tbl;
595 	struct rhash_lock_head __rcu **bkt;
596 	unsigned long flags;
597 	unsigned int hash;
598 	void *data;
599 
600 	new_tbl = rcu_dereference(ht->tbl);
601 
602 	do {
603 		tbl = new_tbl;
604 		hash = rht_head_hashfn(ht, tbl, obj, ht->p);
605 		if (rcu_access_pointer(tbl->future_tbl))
606 			/* Failure is OK */
607 			bkt = rht_bucket_var(tbl, hash);
608 		else
609 			bkt = rht_bucket_insert(ht, tbl, hash);
610 		if (bkt == NULL) {
611 			new_tbl = rht_dereference_rcu(tbl->future_tbl, ht);
612 			data = ERR_PTR(-EAGAIN);
613 		} else {
614 			flags = rht_lock(tbl, bkt);
615 			data = rhashtable_lookup_one(ht, bkt, tbl,
616 						     hash, key, obj);
617 			new_tbl = rhashtable_insert_one(ht, bkt, tbl,
618 							hash, obj, data);
619 			if (PTR_ERR(new_tbl) != -EEXIST)
620 				data = ERR_CAST(new_tbl);
621 
622 			rht_unlock(tbl, bkt, flags);
623 		}
624 	} while (!IS_ERR_OR_NULL(new_tbl));
625 
626 	if (PTR_ERR(data) == -EAGAIN)
627 		data = ERR_PTR(rhashtable_insert_rehash(ht, tbl) ?:
628 			       -EAGAIN);
629 
630 	return data;
631 }
632 
633 void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
634 			     struct rhash_head *obj)
635 {
636 	void *data;
637 
638 	do {
639 		rcu_read_lock();
640 		data = rhashtable_try_insert(ht, key, obj);
641 		rcu_read_unlock();
642 	} while (PTR_ERR(data) == -EAGAIN);
643 
644 	return data;
645 }
646 EXPORT_SYMBOL_GPL(rhashtable_insert_slow);
647 
648 /**
649  * rhashtable_walk_enter - Initialise an iterator
650  * @ht:		Table to walk over
651  * @iter:	Hash table Iterator
652  *
653  * This function prepares a hash table walk.
654  *
655  * Note that if you restart a walk after rhashtable_walk_stop you
656  * may see the same object twice.  Also, you may miss objects if
657  * there are removals in between rhashtable_walk_stop and the next
658  * call to rhashtable_walk_start.
659  *
660  * For a completely stable walk you should construct your own data
661  * structure outside the hash table.
662  *
663  * This function may be called from any process context, including
664  * non-preemptable context, but cannot be called from softirq or
665  * hardirq context.
666  *
667  * You must call rhashtable_walk_exit after this function returns.
668  */
669 void rhashtable_walk_enter(struct rhashtable *ht, struct rhashtable_iter *iter)
670 {
671 	iter->ht = ht;
672 	iter->p = NULL;
673 	iter->slot = 0;
674 	iter->skip = 0;
675 	iter->end_of_table = 0;
676 
677 	spin_lock(&ht->lock);
678 	iter->walker.tbl =
679 		rcu_dereference_protected(ht->tbl, lockdep_is_held(&ht->lock));
680 	list_add(&iter->walker.list, &iter->walker.tbl->walkers);
681 	spin_unlock(&ht->lock);
682 }
683 EXPORT_SYMBOL_GPL(rhashtable_walk_enter);
684 
685 /**
686  * rhashtable_walk_exit - Free an iterator
687  * @iter:	Hash table Iterator
688  *
689  * This function frees resources allocated by rhashtable_walk_enter.
690  */
691 void rhashtable_walk_exit(struct rhashtable_iter *iter)
692 {
693 	spin_lock(&iter->ht->lock);
694 	if (iter->walker.tbl)
695 		list_del(&iter->walker.list);
696 	spin_unlock(&iter->ht->lock);
697 }
698 EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
699 
700 /**
701  * rhashtable_walk_start_check - Start a hash table walk
702  * @iter:	Hash table iterator
703  *
704  * Start a hash table walk at the current iterator position.  Note that we take
705  * the RCU lock in all cases including when we return an error.  So you must
706  * always call rhashtable_walk_stop to clean up.
707  *
708  * Returns zero if successful.
709  *
710  * Returns -EAGAIN if resize event occurred.  Note that the iterator
711  * will rewind back to the beginning and you may use it immediately
712  * by calling rhashtable_walk_next.
713  *
714  * rhashtable_walk_start is defined as an inline variant that returns
715  * void. This is preferred in cases where the caller would ignore
716  * resize events and always continue.
717  */
718 int rhashtable_walk_start_check(struct rhashtable_iter *iter)
719 	__acquires(RCU)
720 {
721 	struct rhashtable *ht = iter->ht;
722 	bool rhlist = ht->rhlist;
723 
724 	rcu_read_lock();
725 
726 	spin_lock(&ht->lock);
727 	if (iter->walker.tbl)
728 		list_del(&iter->walker.list);
729 	spin_unlock(&ht->lock);
730 
731 	if (iter->end_of_table)
732 		return 0;
733 	if (!iter->walker.tbl) {
734 		iter->walker.tbl = rht_dereference_rcu(ht->tbl, ht);
735 		iter->slot = 0;
736 		iter->skip = 0;
737 		return -EAGAIN;
738 	}
739 
740 	if (iter->p && !rhlist) {
741 		/*
742 		 * We need to validate that 'p' is still in the table, and
743 		 * if so, update 'skip'
744 		 */
745 		struct rhash_head *p;
746 		int skip = 0;
747 		rht_for_each_rcu(p, iter->walker.tbl, iter->slot) {
748 			skip++;
749 			if (p == iter->p) {
750 				iter->skip = skip;
751 				goto found;
752 			}
753 		}
754 		iter->p = NULL;
755 	} else if (iter->p && rhlist) {
756 		/* Need to validate that 'list' is still in the table, and
757 		 * if so, update 'skip' and 'p'.
758 		 */
759 		struct rhash_head *p;
760 		struct rhlist_head *list;
761 		int skip = 0;
762 		rht_for_each_rcu(p, iter->walker.tbl, iter->slot) {
763 			for (list = container_of(p, struct rhlist_head, rhead);
764 			     list;
765 			     list = rcu_dereference(list->next)) {
766 				skip++;
767 				if (list == iter->list) {
768 					iter->p = p;
769 					iter->skip = skip;
770 					goto found;
771 				}
772 			}
773 		}
774 		iter->p = NULL;
775 	}
776 found:
777 	return 0;
778 }
779 EXPORT_SYMBOL_GPL(rhashtable_walk_start_check);
780 
781 /**
782  * __rhashtable_walk_find_next - Find the next element in a table (or the first
783  * one in case of a new walk).
784  *
785  * @iter:	Hash table iterator
786  *
787  * Returns the found object or NULL when the end of the table is reached.
788  *
789  * Returns -EAGAIN if resize event occurred.
790  */
791 static void *__rhashtable_walk_find_next(struct rhashtable_iter *iter)
792 {
793 	struct bucket_table *tbl = iter->walker.tbl;
794 	struct rhlist_head *list = iter->list;
795 	struct rhashtable *ht = iter->ht;
796 	struct rhash_head *p = iter->p;
797 	bool rhlist = ht->rhlist;
798 
799 	if (!tbl)
800 		return NULL;
801 
802 	for (; iter->slot < tbl->size; iter->slot++) {
803 		int skip = iter->skip;
804 
805 		rht_for_each_rcu(p, tbl, iter->slot) {
806 			if (rhlist) {
807 				list = container_of(p, struct rhlist_head,
808 						    rhead);
809 				do {
810 					if (!skip)
811 						goto next;
812 					skip--;
813 					list = rcu_dereference(list->next);
814 				} while (list);
815 
816 				continue;
817 			}
818 			if (!skip)
819 				break;
820 			skip--;
821 		}
822 
823 next:
824 		if (!rht_is_a_nulls(p)) {
825 			iter->skip++;
826 			iter->p = p;
827 			iter->list = list;
828 			return rht_obj(ht, rhlist ? &list->rhead : p);
829 		}
830 
831 		iter->skip = 0;
832 	}
833 
834 	iter->p = NULL;
835 
836 	/* Ensure we see any new tables. */
837 	smp_rmb();
838 
839 	iter->walker.tbl = rht_dereference_rcu(tbl->future_tbl, ht);
840 	if (iter->walker.tbl) {
841 		iter->slot = 0;
842 		iter->skip = 0;
843 		return ERR_PTR(-EAGAIN);
844 	} else {
845 		iter->end_of_table = true;
846 	}
847 
848 	return NULL;
849 }
850 
851 /**
852  * rhashtable_walk_next - Return the next object and advance the iterator
853  * @iter:	Hash table iterator
854  *
855  * Note that you must call rhashtable_walk_stop when you are finished
856  * with the walk.
857  *
858  * Returns the next object or NULL when the end of the table is reached.
859  *
860  * Returns -EAGAIN if resize event occurred.  Note that the iterator
861  * will rewind back to the beginning and you may continue to use it.
862  */
863 void *rhashtable_walk_next(struct rhashtable_iter *iter)
864 {
865 	struct rhlist_head *list = iter->list;
866 	struct rhashtable *ht = iter->ht;
867 	struct rhash_head *p = iter->p;
868 	bool rhlist = ht->rhlist;
869 
870 	if (p) {
871 		if (!rhlist || !(list = rcu_dereference(list->next))) {
872 			p = rcu_dereference(p->next);
873 			list = container_of(p, struct rhlist_head, rhead);
874 		}
875 		if (!rht_is_a_nulls(p)) {
876 			iter->skip++;
877 			iter->p = p;
878 			iter->list = list;
879 			return rht_obj(ht, rhlist ? &list->rhead : p);
880 		}
881 
882 		/* At the end of this slot, switch to next one and then find
883 		 * next entry from that point.
884 		 */
885 		iter->skip = 0;
886 		iter->slot++;
887 	}
888 
889 	return __rhashtable_walk_find_next(iter);
890 }
891 EXPORT_SYMBOL_GPL(rhashtable_walk_next);
892 
893 /**
894  * rhashtable_walk_peek - Return the next object but don't advance the iterator
895  * @iter:	Hash table iterator
896  *
897  * Returns the next object or NULL when the end of the table is reached.
898  *
899  * Returns -EAGAIN if resize event occurred.  Note that the iterator
900  * will rewind back to the beginning and you may continue to use it.
901  */
902 void *rhashtable_walk_peek(struct rhashtable_iter *iter)
903 {
904 	struct rhlist_head *list = iter->list;
905 	struct rhashtable *ht = iter->ht;
906 	struct rhash_head *p = iter->p;
907 
908 	if (p)
909 		return rht_obj(ht, ht->rhlist ? &list->rhead : p);
910 
911 	/* No object found in current iter, find next one in the table. */
912 
913 	if (iter->skip) {
914 		/* A nonzero skip value points to the next entry in the table
915 		 * beyond that last one that was found. Decrement skip so
916 		 * we find the current value. __rhashtable_walk_find_next
917 		 * will restore the original value of skip assuming that
918 		 * the table hasn't changed.
919 		 */
920 		iter->skip--;
921 	}
922 
923 	return __rhashtable_walk_find_next(iter);
924 }
925 EXPORT_SYMBOL_GPL(rhashtable_walk_peek);
926 
927 /**
928  * rhashtable_walk_stop - Finish a hash table walk
929  * @iter:	Hash table iterator
930  *
931  * Finish a hash table walk.  Does not reset the iterator to the start of the
932  * hash table.
933  */
934 void rhashtable_walk_stop(struct rhashtable_iter *iter)
935 	__releases(RCU)
936 {
937 	struct rhashtable *ht;
938 	struct bucket_table *tbl = iter->walker.tbl;
939 
940 	if (!tbl)
941 		goto out;
942 
943 	ht = iter->ht;
944 
945 	spin_lock(&ht->lock);
946 	if (rcu_head_after_call_rcu(&tbl->rcu, bucket_table_free_rcu))
947 		/* This bucket table is being freed, don't re-link it. */
948 		iter->walker.tbl = NULL;
949 	else
950 		list_add(&iter->walker.list, &tbl->walkers);
951 	spin_unlock(&ht->lock);
952 
953 out:
954 	rcu_read_unlock();
955 }
956 EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
957 
958 static size_t rounded_hashtable_size(const struct rhashtable_params *params)
959 {
960 	size_t retsize;
961 
962 	if (params->nelem_hint)
963 		retsize = max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
964 			      (unsigned long)params->min_size);
965 	else
966 		retsize = max(HASH_DEFAULT_SIZE,
967 			      (unsigned long)params->min_size);
968 
969 	return retsize;
970 }
971 
972 static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
973 {
974 	return jhash2(key, length, seed);
975 }
976 
977 /**
978  * rhashtable_init - initialize a new hash table
979  * @ht:		hash table to be initialized
980  * @params:	configuration parameters
981  *
982  * Initializes a new hash table based on the provided configuration
983  * parameters. A table can be configured either with a variable or
984  * fixed length key:
985  *
986  * Configuration Example 1: Fixed length keys
987  * struct test_obj {
988  *	int			key;
989  *	void *			my_member;
990  *	struct rhash_head	node;
991  * };
992  *
993  * struct rhashtable_params params = {
994  *	.head_offset = offsetof(struct test_obj, node),
995  *	.key_offset = offsetof(struct test_obj, key),
996  *	.key_len = sizeof(int),
997  *	.hashfn = jhash,
998  * };
999  *
1000  * Configuration Example 2: Variable length keys
1001  * struct test_obj {
1002  *	[...]
1003  *	struct rhash_head	node;
1004  * };
1005  *
1006  * u32 my_hash_fn(const void *data, u32 len, u32 seed)
1007  * {
1008  *	struct test_obj *obj = data;
1009  *
1010  *	return [... hash ...];
1011  * }
1012  *
1013  * struct rhashtable_params params = {
1014  *	.head_offset = offsetof(struct test_obj, node),
1015  *	.hashfn = jhash,
1016  *	.obj_hashfn = my_hash_fn,
1017  * };
1018  */
1019 int rhashtable_init(struct rhashtable *ht,
1020 		    const struct rhashtable_params *params)
1021 {
1022 	struct bucket_table *tbl;
1023 	size_t size;
1024 
1025 	if ((!params->key_len && !params->obj_hashfn) ||
1026 	    (params->obj_hashfn && !params->obj_cmpfn))
1027 		return -EINVAL;
1028 
1029 	memset(ht, 0, sizeof(*ht));
1030 	mutex_init(&ht->mutex);
1031 	spin_lock_init(&ht->lock);
1032 	memcpy(&ht->p, params, sizeof(*params));
1033 
1034 	if (params->min_size)
1035 		ht->p.min_size = roundup_pow_of_two(params->min_size);
1036 
1037 	/* Cap total entries at 2^31 to avoid nelems overflow. */
1038 	ht->max_elems = 1u << 31;
1039 
1040 	if (params->max_size) {
1041 		ht->p.max_size = rounddown_pow_of_two(params->max_size);
1042 		if (ht->p.max_size < ht->max_elems / 2)
1043 			ht->max_elems = ht->p.max_size * 2;
1044 	}
1045 
1046 	ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
1047 
1048 	size = rounded_hashtable_size(&ht->p);
1049 
1050 	ht->key_len = ht->p.key_len;
1051 	if (!params->hashfn) {
1052 		ht->p.hashfn = jhash;
1053 
1054 		if (!(ht->key_len & (sizeof(u32) - 1))) {
1055 			ht->key_len /= sizeof(u32);
1056 			ht->p.hashfn = rhashtable_jhash2;
1057 		}
1058 	}
1059 
1060 	/*
1061 	 * This is api initialization and thus we need to guarantee the
1062 	 * initial rhashtable allocation. Upon failure, retry with the
1063 	 * smallest possible size with __GFP_NOFAIL semantics.
1064 	 */
1065 	tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
1066 	if (unlikely(tbl == NULL)) {
1067 		size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
1068 		tbl = bucket_table_alloc(ht, size, GFP_KERNEL | __GFP_NOFAIL);
1069 	}
1070 
1071 	atomic_set(&ht->nelems, 0);
1072 
1073 	RCU_INIT_POINTER(ht->tbl, tbl);
1074 
1075 	INIT_WORK(&ht->run_work, rht_deferred_worker);
1076 
1077 	return 0;
1078 }
1079 EXPORT_SYMBOL_GPL(rhashtable_init);
1080 
1081 /**
1082  * rhltable_init - initialize a new hash list table
1083  * @hlt:	hash list table to be initialized
1084  * @params:	configuration parameters
1085  *
1086  * Initializes a new hash list table.
1087  *
1088  * See documentation for rhashtable_init.
1089  */
1090 int rhltable_init(struct rhltable *hlt, const struct rhashtable_params *params)
1091 {
1092 	int err;
1093 
1094 	err = rhashtable_init(&hlt->ht, params);
1095 	hlt->ht.rhlist = true;
1096 	return err;
1097 }
1098 EXPORT_SYMBOL_GPL(rhltable_init);
1099 
1100 static void rhashtable_free_one(struct rhashtable *ht, struct rhash_head *obj,
1101 				void (*free_fn)(void *ptr, void *arg),
1102 				void *arg)
1103 {
1104 	struct rhlist_head *list;
1105 
1106 	if (!ht->rhlist) {
1107 		free_fn(rht_obj(ht, obj), arg);
1108 		return;
1109 	}
1110 
1111 	list = container_of(obj, struct rhlist_head, rhead);
1112 	do {
1113 		obj = &list->rhead;
1114 		list = rht_dereference(list->next, ht);
1115 		free_fn(rht_obj(ht, obj), arg);
1116 	} while (list);
1117 }
1118 
1119 /**
1120  * rhashtable_free_and_destroy - free elements and destroy hash table
1121  * @ht:		the hash table to destroy
1122  * @free_fn:	callback to release resources of element
1123  * @arg:	pointer passed to free_fn
1124  *
1125  * Stops an eventual async resize. If defined, invokes free_fn for each
1126  * element to releasal resources. Please note that RCU protected
1127  * readers may still be accessing the elements. Releasing of resources
1128  * must occur in a compatible manner. Then frees the bucket array.
1129  *
1130  * This function will eventually sleep to wait for an async resize
1131  * to complete. The caller is responsible that no further write operations
1132  * occurs in parallel.
1133  */
1134 void rhashtable_free_and_destroy(struct rhashtable *ht,
1135 				 void (*free_fn)(void *ptr, void *arg),
1136 				 void *arg)
1137 {
1138 	struct bucket_table *tbl, *next_tbl;
1139 	unsigned int i;
1140 
1141 	cancel_work_sync(&ht->run_work);
1142 
1143 	mutex_lock(&ht->mutex);
1144 	tbl = rht_dereference(ht->tbl, ht);
1145 restart:
1146 	if (free_fn) {
1147 		for (i = 0; i < tbl->size; i++) {
1148 			struct rhash_head *pos, *next;
1149 
1150 			cond_resched();
1151 			for (pos = rht_ptr_exclusive(rht_bucket(tbl, i)),
1152 			     next = !rht_is_a_nulls(pos) ?
1153 					rht_dereference(pos->next, ht) : NULL;
1154 			     !rht_is_a_nulls(pos);
1155 			     pos = next,
1156 			     next = !rht_is_a_nulls(pos) ?
1157 					rht_dereference(pos->next, ht) : NULL)
1158 				rhashtable_free_one(ht, pos, free_fn, arg);
1159 		}
1160 	}
1161 
1162 	next_tbl = rht_dereference(tbl->future_tbl, ht);
1163 	bucket_table_free(tbl);
1164 	if (next_tbl) {
1165 		tbl = next_tbl;
1166 		goto restart;
1167 	}
1168 	mutex_unlock(&ht->mutex);
1169 }
1170 EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
1171 
1172 void rhashtable_destroy(struct rhashtable *ht)
1173 {
1174 	return rhashtable_free_and_destroy(ht, NULL, NULL);
1175 }
1176 EXPORT_SYMBOL_GPL(rhashtable_destroy);
1177 
1178 struct rhash_lock_head __rcu **__rht_bucket_nested(
1179 	const struct bucket_table *tbl, unsigned int hash)
1180 {
1181 	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1182 	unsigned int index = hash & ((1 << tbl->nest) - 1);
1183 	unsigned int size = tbl->size >> tbl->nest;
1184 	unsigned int subhash = hash;
1185 	union nested_table *ntbl;
1186 
1187 	ntbl = nested_table_top(tbl);
1188 	ntbl = rht_dereference_bucket_rcu(ntbl[index].table, tbl, hash);
1189 	subhash >>= tbl->nest;
1190 
1191 	while (ntbl && size > (1 << shift)) {
1192 		index = subhash & ((1 << shift) - 1);
1193 		ntbl = rht_dereference_bucket_rcu(ntbl[index].table,
1194 						  tbl, hash);
1195 		size >>= shift;
1196 		subhash >>= shift;
1197 	}
1198 
1199 	if (!ntbl)
1200 		return NULL;
1201 
1202 	return &ntbl[subhash].bucket;
1203 
1204 }
1205 EXPORT_SYMBOL_GPL(__rht_bucket_nested);
1206 
1207 struct rhash_lock_head __rcu **rht_bucket_nested(
1208 	const struct bucket_table *tbl, unsigned int hash)
1209 {
1210 	static struct rhash_lock_head __rcu *rhnull;
1211 
1212 	if (!rhnull)
1213 		INIT_RHT_NULLS_HEAD(rhnull);
1214 	return __rht_bucket_nested(tbl, hash) ?: &rhnull;
1215 }
1216 EXPORT_SYMBOL_GPL(rht_bucket_nested);
1217 
1218 struct rhash_lock_head __rcu **rht_bucket_nested_insert(
1219 	struct rhashtable *ht, struct bucket_table *tbl, unsigned int hash)
1220 {
1221 	const unsigned int shift = PAGE_SHIFT - ilog2(sizeof(void *));
1222 	unsigned int index = hash & ((1 << tbl->nest) - 1);
1223 	unsigned int size = tbl->size >> tbl->nest;
1224 	union nested_table *ntbl;
1225 
1226 	ntbl = nested_table_top(tbl);
1227 	hash >>= tbl->nest;
1228 	ntbl = nested_table_alloc(ht, &ntbl[index].table,
1229 				  size <= (1 << shift));
1230 
1231 	while (ntbl && size > (1 << shift)) {
1232 		index = hash & ((1 << shift) - 1);
1233 		size >>= shift;
1234 		hash >>= shift;
1235 		ntbl = nested_table_alloc(ht, &ntbl[index].table,
1236 					  size <= (1 << shift));
1237 	}
1238 
1239 	if (!ntbl)
1240 		return NULL;
1241 
1242 	return &ntbl[hash].bucket;
1243 
1244 }
1245 EXPORT_SYMBOL_GPL(rht_bucket_nested_insert);
1246