blob: b5344ef4c6846c4f9256c1d0d418f774284c8fcc [file] [log] [blame]
Thomas Graf7e1e7762014-08-02 11:47:44 +02001/*
2 * Resizable, Scalable, Concurrent Hash Table
3 *
Thomas Grafa5ec68e2015-02-05 02:03:32 +01004 * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
Thomas Graf7e1e7762014-08-02 11:47:44 +02005 * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
6 *
7 * Based on the following paper:
8 * https://www.usenix.org/legacy/event/atc11/tech/final_files/Triplett.pdf
9 *
10 * Code partially derived from nft_hash
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/kernel.h>
18#include <linux/init.h>
19#include <linux/log2.h>
Eric Dumazet5beb5c92015-02-26 07:20:34 -080020#include <linux/sched.h>
Thomas Graf7e1e7762014-08-02 11:47:44 +020021#include <linux/slab.h>
22#include <linux/vmalloc.h>
23#include <linux/mm.h>
Daniel Borkmann87545892014-12-10 16:33:11 +010024#include <linux/jhash.h>
Thomas Graf7e1e7762014-08-02 11:47:44 +020025#include <linux/random.h>
26#include <linux/rhashtable.h>
Stephen Rothwell61d7b092015-02-09 14:04:03 +110027#include <linux/err.h>
Thomas Graf7e1e7762014-08-02 11:47:44 +020028
29#define HASH_DEFAULT_SIZE 64UL
30#define HASH_MIN_SIZE 4UL
Thomas Graf97defe12015-01-02 23:00:20 +010031#define BUCKET_LOCKS_PER_CPU 128UL
32
Thomas Graff89bd6f2015-01-02 23:00:21 +010033/* Base bits plus 1 bit for nulls marker */
34#define HASH_RESERVED_SPACE (RHT_BASE_BITS + 1)
35
Thomas Graf97defe12015-01-02 23:00:20 +010036enum {
37 RHT_LOCK_NORMAL,
38 RHT_LOCK_NESTED,
Thomas Graf97defe12015-01-02 23:00:20 +010039};
40
41/* The bucket lock is selected based on the hash and protects mutations
42 * on a group of hash buckets.
43 *
Thomas Grafa5ec68e2015-02-05 02:03:32 +010044 * A maximum of tbl->size/2 bucket locks is allocated. This ensures that
45 * a single lock always covers both buckets which may both contains
46 * entries which link to the same bucket of the old table during resizing.
47 * This allows to simplify the locking as locking the bucket in both
48 * tables during resize always guarantee protection.
49 *
Thomas Graf97defe12015-01-02 23:00:20 +010050 * IMPORTANT: When holding the bucket lock of both the old and new table
51 * during expansions and shrinking, the old bucket lock must always be
52 * acquired first.
53 */
54static spinlock_t *bucket_lock(const struct bucket_table *tbl, u32 hash)
55{
56 return &tbl->locks[hash & tbl->locks_mask];
57}
Thomas Graf7e1e7762014-08-02 11:47:44 +020058
Thomas Grafc91eee52014-08-13 16:38:30 +020059static void *rht_obj(const struct rhashtable *ht, const struct rhash_head *he)
Thomas Graf7e1e7762014-08-02 11:47:44 +020060{
61 return (void *) he - ht->p.head_offset;
62}
Thomas Graf7e1e7762014-08-02 11:47:44 +020063
Thomas Graf8d24c0b2015-01-02 23:00:14 +010064static u32 rht_bucket_index(const struct bucket_table *tbl, u32 hash)
Thomas Graf7e1e7762014-08-02 11:47:44 +020065{
Thomas Graf8d24c0b2015-01-02 23:00:14 +010066 return hash & (tbl->size - 1);
Thomas Graf7e1e7762014-08-02 11:47:44 +020067}
68
Thomas Graf8d24c0b2015-01-02 23:00:14 +010069static u32 obj_raw_hashfn(const struct rhashtable *ht, const void *ptr)
70{
71 u32 hash;
72
73 if (unlikely(!ht->p.key_len))
74 hash = ht->p.obj_hashfn(ptr, ht->p.hash_rnd);
75 else
76 hash = ht->p.hashfn(ptr + ht->p.key_offset, ht->p.key_len,
77 ht->p.hash_rnd);
78
Thomas Graff89bd6f2015-01-02 23:00:21 +010079 return hash >> HASH_RESERVED_SPACE;
Thomas Graf8d24c0b2015-01-02 23:00:14 +010080}
81
Thomas Graf97defe12015-01-02 23:00:20 +010082static u32 key_hashfn(struct rhashtable *ht, const void *key, u32 len)
Thomas Graf7e1e7762014-08-02 11:47:44 +020083{
Thomas Grafc88455c2015-02-05 02:03:31 +010084 return ht->p.hashfn(key, len, ht->p.hash_rnd) >> HASH_RESERVED_SPACE;
Thomas Graf7e1e7762014-08-02 11:47:44 +020085}
Thomas Graf7e1e7762014-08-02 11:47:44 +020086
87static u32 head_hashfn(const struct rhashtable *ht,
Thomas Graf8d24c0b2015-01-02 23:00:14 +010088 const struct bucket_table *tbl,
89 const struct rhash_head *he)
Thomas Graf7e1e7762014-08-02 11:47:44 +020090{
Thomas Graf8d24c0b2015-01-02 23:00:14 +010091 return rht_bucket_index(tbl, obj_raw_hashfn(ht, rht_obj(ht, he)));
Thomas Graf7e1e7762014-08-02 11:47:44 +020092}
93
Thomas Grafa03eaec2015-02-05 02:03:34 +010094#ifdef CONFIG_PROVE_LOCKING
95static void debug_dump_buckets(const struct rhashtable *ht,
96 const struct bucket_table *tbl)
97{
98 struct rhash_head *he;
99 unsigned int i, hash;
100
101 for (i = 0; i < tbl->size; i++) {
102 pr_warn(" [Bucket %d] ", i);
103 rht_for_each_rcu(he, tbl, i) {
104 hash = head_hashfn(ht, tbl, he);
105 pr_cont("[hash = %#x, lock = %p] ",
106 hash, bucket_lock(tbl, hash));
107 }
108 pr_cont("\n");
109 }
110
111}
112
113static void debug_dump_table(struct rhashtable *ht,
114 const struct bucket_table *tbl,
115 unsigned int hash)
116{
117 struct bucket_table *old_tbl, *future_tbl;
118
119 pr_emerg("BUG: lock for hash %#x in table %p not held\n",
120 hash, tbl);
121
122 rcu_read_lock();
123 future_tbl = rht_dereference_rcu(ht->future_tbl, ht);
124 old_tbl = rht_dereference_rcu(ht->tbl, ht);
125 if (future_tbl != old_tbl) {
126 pr_warn("Future table %p (size: %zd)\n",
127 future_tbl, future_tbl->size);
128 debug_dump_buckets(ht, future_tbl);
129 }
130
131 pr_warn("Table %p (size: %zd)\n", old_tbl, old_tbl->size);
132 debug_dump_buckets(ht, old_tbl);
133
134 rcu_read_unlock();
135}
136
137#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
138#define ASSERT_BUCKET_LOCK(HT, TBL, HASH) \
139 do { \
140 if (unlikely(!lockdep_rht_bucket_is_held(TBL, HASH))) { \
141 debug_dump_table(HT, TBL, HASH); \
142 BUG(); \
143 } \
144 } while (0)
145
146int lockdep_rht_mutex_is_held(struct rhashtable *ht)
147{
148 return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
149}
150EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
151
152int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
153{
154 spinlock_t *lock = bucket_lock(tbl, hash);
155
156 return (debug_locks) ? lockdep_is_held(lock) : 1;
157}
158EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
159#else
160#define ASSERT_RHT_MUTEX(HT)
161#define ASSERT_BUCKET_LOCK(HT, TBL, HASH)
162#endif
163
164
Thomas Grafb8e19432015-01-02 23:00:17 +0100165static struct rhash_head __rcu **bucket_tail(struct bucket_table *tbl, u32 n)
166{
167 struct rhash_head __rcu **pprev;
168
169 for (pprev = &tbl->buckets[n];
Thomas Graff89bd6f2015-01-02 23:00:21 +0100170 !rht_is_a_nulls(rht_dereference_bucket(*pprev, tbl, n));
Thomas Grafb8e19432015-01-02 23:00:17 +0100171 pprev = &rht_dereference_bucket(*pprev, tbl, n)->next)
172 ;
173
174 return pprev;
175}
176
Thomas Graf97defe12015-01-02 23:00:20 +0100177static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl)
178{
179 unsigned int i, size;
180#if defined(CONFIG_PROVE_LOCKING)
181 unsigned int nr_pcpus = 2;
182#else
183 unsigned int nr_pcpus = num_possible_cpus();
184#endif
185
186 nr_pcpus = min_t(unsigned int, nr_pcpus, 32UL);
187 size = roundup_pow_of_two(nr_pcpus * ht->p.locks_mul);
188
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100189 /* Never allocate more than 0.5 locks per bucket */
190 size = min_t(unsigned int, size, tbl->size >> 1);
Thomas Graf97defe12015-01-02 23:00:20 +0100191
192 if (sizeof(spinlock_t) != 0) {
193#ifdef CONFIG_NUMA
194 if (size * sizeof(spinlock_t) > PAGE_SIZE)
195 tbl->locks = vmalloc(size * sizeof(spinlock_t));
196 else
197#endif
198 tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
199 GFP_KERNEL);
200 if (!tbl->locks)
201 return -ENOMEM;
202 for (i = 0; i < size; i++)
203 spin_lock_init(&tbl->locks[i]);
204 }
205 tbl->locks_mask = size - 1;
206
207 return 0;
208}
209
210static void bucket_table_free(const struct bucket_table *tbl)
211{
212 if (tbl)
213 kvfree(tbl->locks);
214
215 kvfree(tbl);
216}
217
218static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
219 size_t nbuckets)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200220{
Daniel Borkmanneb6d1ab2015-02-20 00:53:38 +0100221 struct bucket_table *tbl = NULL;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200222 size_t size;
Thomas Graff89bd6f2015-01-02 23:00:21 +0100223 int i;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200224
225 size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]);
Daniel Borkmanneb6d1ab2015-02-20 00:53:38 +0100226 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
227 tbl = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200228 if (tbl == NULL)
229 tbl = vzalloc(size);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200230 if (tbl == NULL)
231 return NULL;
232
233 tbl->size = nbuckets;
234
Thomas Graf97defe12015-01-02 23:00:20 +0100235 if (alloc_bucket_locks(ht, tbl) < 0) {
236 bucket_table_free(tbl);
237 return NULL;
238 }
Thomas Graf7e1e7762014-08-02 11:47:44 +0200239
Thomas Graff89bd6f2015-01-02 23:00:21 +0100240 for (i = 0; i < nbuckets; i++)
241 INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);
242
Thomas Graf97defe12015-01-02 23:00:20 +0100243 return tbl;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200244}
245
246/**
247 * rht_grow_above_75 - returns true if nelems > 0.75 * table-size
248 * @ht: hash table
249 * @new_size: new table size
250 */
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100251static bool rht_grow_above_75(const struct rhashtable *ht, size_t new_size)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200252{
253 /* Expand table when exceeding 75% load */
Ying Xuec0c09bf2015-01-07 13:41:56 +0800254 return atomic_read(&ht->nelems) > (new_size / 4 * 3) &&
Daniel Borkmann8331de72015-02-25 16:31:53 +0100255 (!ht->p.max_shift || atomic_read(&ht->shift) < ht->p.max_shift);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200256}
Thomas Graf7e1e7762014-08-02 11:47:44 +0200257
258/**
259 * rht_shrink_below_30 - returns true if nelems < 0.3 * table-size
260 * @ht: hash table
261 * @new_size: new table size
262 */
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100263static bool rht_shrink_below_30(const struct rhashtable *ht, size_t new_size)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200264{
265 /* Shrink table beneath 30% load */
Ying Xuec0c09bf2015-01-07 13:41:56 +0800266 return atomic_read(&ht->nelems) < (new_size * 3 / 10) &&
267 (atomic_read(&ht->shift) > ht->p.min_shift);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200268}
Thomas Graf7e1e7762014-08-02 11:47:44 +0200269
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100270static void lock_buckets(struct bucket_table *new_tbl,
271 struct bucket_table *old_tbl, unsigned int hash)
272 __acquires(old_bucket_lock)
273{
274 spin_lock_bh(bucket_lock(old_tbl, hash));
275 if (new_tbl != old_tbl)
276 spin_lock_bh_nested(bucket_lock(new_tbl, hash),
277 RHT_LOCK_NESTED);
278}
279
280static void unlock_buckets(struct bucket_table *new_tbl,
281 struct bucket_table *old_tbl, unsigned int hash)
282 __releases(old_bucket_lock)
283{
284 if (new_tbl != old_tbl)
285 spin_unlock_bh(bucket_lock(new_tbl, hash));
286 spin_unlock_bh(bucket_lock(old_tbl, hash));
287}
288
289/**
290 * Unlink entries on bucket which hash to different bucket.
291 *
292 * Returns true if no more work needs to be performed on the bucket.
293 */
Thomas Grafa03eaec2015-02-05 02:03:34 +0100294static bool hashtable_chain_unzip(struct rhashtable *ht,
Thomas Graf7e1e7762014-08-02 11:47:44 +0200295 const struct bucket_table *new_tbl,
Thomas Graf97defe12015-01-02 23:00:20 +0100296 struct bucket_table *old_tbl,
297 size_t old_hash)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200298{
299 struct rhash_head *he, *p, *next;
Thomas Graf97defe12015-01-02 23:00:20 +0100300 unsigned int new_hash, new_hash2;
301
Thomas Grafa03eaec2015-02-05 02:03:34 +0100302 ASSERT_BUCKET_LOCK(ht, old_tbl, old_hash);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200303
304 /* Old bucket empty, no work needed. */
Thomas Graf97defe12015-01-02 23:00:20 +0100305 p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl,
306 old_hash);
Thomas Graff89bd6f2015-01-02 23:00:21 +0100307 if (rht_is_a_nulls(p))
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100308 return false;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200309
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100310 new_hash = head_hashfn(ht, new_tbl, p);
Thomas Grafa03eaec2015-02-05 02:03:34 +0100311 ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash);
Thomas Graf97defe12015-01-02 23:00:20 +0100312
Thomas Graf7e1e7762014-08-02 11:47:44 +0200313 /* Advance the old bucket pointer one or more times until it
314 * reaches a node that doesn't hash to the same bucket as the
315 * previous node p. Call the previous node p;
316 */
Thomas Graf97defe12015-01-02 23:00:20 +0100317 rht_for_each_continue(he, p->next, old_tbl, old_hash) {
318 new_hash2 = head_hashfn(ht, new_tbl, he);
Thomas Grafa03eaec2015-02-05 02:03:34 +0100319 ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash2);
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100320
Thomas Graf97defe12015-01-02 23:00:20 +0100321 if (new_hash != new_hash2)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200322 break;
323 p = he;
324 }
Thomas Graf97defe12015-01-02 23:00:20 +0100325 rcu_assign_pointer(old_tbl->buckets[old_hash], p->next);
326
Thomas Graf7e1e7762014-08-02 11:47:44 +0200327 /* Find the subsequent node which does hash to the same
328 * bucket as node P, or NULL if no such node exists.
329 */
Thomas Graff89bd6f2015-01-02 23:00:21 +0100330 INIT_RHT_NULLS_HEAD(next, ht, old_hash);
331 if (!rht_is_a_nulls(he)) {
Thomas Graf97defe12015-01-02 23:00:20 +0100332 rht_for_each_continue(he, he->next, old_tbl, old_hash) {
333 if (head_hashfn(ht, new_tbl, he) == new_hash) {
Thomas Graf7e1e7762014-08-02 11:47:44 +0200334 next = he;
335 break;
336 }
337 }
338 }
339
340 /* Set p's next pointer to that subsequent node pointer,
341 * bypassing the nodes which do not hash to p's bucket
342 */
Thomas Graf97defe12015-01-02 23:00:20 +0100343 rcu_assign_pointer(p->next, next);
344
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100345 p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl,
346 old_hash);
347
348 return !rht_is_a_nulls(p);
Thomas Graf97defe12015-01-02 23:00:20 +0100349}
350
Thomas Graf7cd10db2015-02-05 02:03:35 +0100351static void link_old_to_new(struct rhashtable *ht, struct bucket_table *new_tbl,
Thomas Graf97defe12015-01-02 23:00:20 +0100352 unsigned int new_hash, struct rhash_head *entry)
353{
Thomas Graf7cd10db2015-02-05 02:03:35 +0100354 ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash);
355
Thomas Graf97defe12015-01-02 23:00:20 +0100356 rcu_assign_pointer(*bucket_tail(new_tbl, new_hash), entry);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200357}
358
359/**
360 * rhashtable_expand - Expand hash table while allowing concurrent lookups
361 * @ht: the hash table to expand
Thomas Graf7e1e7762014-08-02 11:47:44 +0200362 *
363 * A secondary bucket array is allocated and the hash entries are migrated
364 * while keeping them on both lists until the end of the RCU grace period.
365 *
366 * This function may only be called in a context where it is safe to call
367 * synchronize_rcu(), e.g. not within a rcu_read_lock() section.
368 *
Thomas Graf97defe12015-01-02 23:00:20 +0100369 * The caller must ensure that no concurrent resizing occurs by holding
370 * ht->mutex.
371 *
372 * It is valid to have concurrent insertions and deletions protected by per
373 * bucket locks or concurrent RCU protected lookups and traversals.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200374 */
Thomas Graf6eba8222014-11-13 13:45:46 +0100375int rhashtable_expand(struct rhashtable *ht)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200376{
377 struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
378 struct rhash_head *he;
Thomas Graf97defe12015-01-02 23:00:20 +0100379 unsigned int new_hash, old_hash;
380 bool complete = false;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200381
382 ASSERT_RHT_MUTEX(ht);
383
Thomas Graf97defe12015-01-02 23:00:20 +0100384 new_tbl = bucket_table_alloc(ht, old_tbl->size * 2);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200385 if (new_tbl == NULL)
386 return -ENOMEM;
387
Ying Xuec0c09bf2015-01-07 13:41:56 +0800388 atomic_inc(&ht->shift);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200389
Thomas Graf97defe12015-01-02 23:00:20 +0100390 /* Make insertions go into the new, empty table right away. Deletions
391 * and lookups will be attempted in both tables until we synchronize.
392 * The synchronize_rcu() guarantees for the new table to be picked up
393 * so no new additions go into the old table while we relink.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200394 */
Thomas Graf97defe12015-01-02 23:00:20 +0100395 rcu_assign_pointer(ht->future_tbl, new_tbl);
396 synchronize_rcu();
397
398 /* For each new bucket, search the corresponding old bucket for the
399 * first entry that hashes to the new bucket, and link the end of
400 * newly formed bucket chain (containing entries added to future
401 * table) to that entry. Since all the entries which will end up in
402 * the new bucket appear in the same old bucket, this constructs an
403 * entirely valid new hash table, but with multiple buckets
404 * "zipped" together into a single imprecise chain.
405 */
406 for (new_hash = 0; new_hash < new_tbl->size; new_hash++) {
407 old_hash = rht_bucket_index(old_tbl, new_hash);
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100408 lock_buckets(new_tbl, old_tbl, new_hash);
Thomas Graf97defe12015-01-02 23:00:20 +0100409 rht_for_each(he, old_tbl, old_hash) {
410 if (head_hashfn(ht, new_tbl, he) == new_hash) {
Thomas Graf7cd10db2015-02-05 02:03:35 +0100411 link_old_to_new(ht, new_tbl, new_hash, he);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200412 break;
413 }
414 }
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100415 unlock_buckets(new_tbl, old_tbl, new_hash);
Eric Dumazet5beb5c92015-02-26 07:20:34 -0800416 cond_resched();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200417 }
418
Thomas Graf7e1e7762014-08-02 11:47:44 +0200419 /* Unzip interleaved hash chains */
Thomas Graf97defe12015-01-02 23:00:20 +0100420 while (!complete && !ht->being_destroyed) {
Thomas Graf7e1e7762014-08-02 11:47:44 +0200421 /* Wait for readers. All new readers will see the new
422 * table, and thus no references to the old table will
423 * remain.
424 */
425 synchronize_rcu();
426
427 /* For each bucket in the old table (each of which
428 * contains items from multiple buckets of the new
429 * table): ...
430 */
431 complete = true;
Thomas Graf97defe12015-01-02 23:00:20 +0100432 for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100433 lock_buckets(new_tbl, old_tbl, old_hash);
Thomas Graff89bd6f2015-01-02 23:00:21 +0100434
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100435 if (hashtable_chain_unzip(ht, new_tbl, old_tbl,
436 old_hash))
Thomas Graf7e1e7762014-08-02 11:47:44 +0200437 complete = false;
Thomas Graf97defe12015-01-02 23:00:20 +0100438
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100439 unlock_buckets(new_tbl, old_tbl, old_hash);
Eric Dumazet5beb5c92015-02-26 07:20:34 -0800440 cond_resched();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200441 }
Thomas Graf97defe12015-01-02 23:00:20 +0100442 }
Thomas Graf7e1e7762014-08-02 11:47:44 +0200443
Thomas Grafcf52d522015-02-05 02:03:36 +0100444 rcu_assign_pointer(ht->tbl, new_tbl);
Thomas Graf2af4b522015-02-05 02:03:33 +0100445 synchronize_rcu();
446
Thomas Graf7e1e7762014-08-02 11:47:44 +0200447 bucket_table_free(old_tbl);
448 return 0;
449}
450EXPORT_SYMBOL_GPL(rhashtable_expand);
451
452/**
453 * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
454 * @ht: the hash table to shrink
Thomas Graf7e1e7762014-08-02 11:47:44 +0200455 *
456 * This function may only be called in a context where it is safe to call
457 * synchronize_rcu(), e.g. not within a rcu_read_lock() section.
458 *
Thomas Graf97defe12015-01-02 23:00:20 +0100459 * The caller must ensure that no concurrent resizing occurs by holding
460 * ht->mutex.
461 *
Thomas Graf7e1e7762014-08-02 11:47:44 +0200462 * The caller must ensure that no concurrent table mutations take place.
463 * It is however valid to have concurrent lookups if they are RCU protected.
Thomas Graf97defe12015-01-02 23:00:20 +0100464 *
465 * It is valid to have concurrent insertions and deletions protected by per
466 * bucket locks or concurrent RCU protected lookups and traversals.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200467 */
Thomas Graf6eba8222014-11-13 13:45:46 +0100468int rhashtable_shrink(struct rhashtable *ht)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200469{
Thomas Graf97defe12015-01-02 23:00:20 +0100470 struct bucket_table *new_tbl, *tbl = rht_dereference(ht->tbl, ht);
Thomas Graf97defe12015-01-02 23:00:20 +0100471 unsigned int new_hash;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200472
473 ASSERT_RHT_MUTEX(ht);
474
Thomas Graf97defe12015-01-02 23:00:20 +0100475 new_tbl = bucket_table_alloc(ht, tbl->size / 2);
476 if (new_tbl == NULL)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200477 return -ENOMEM;
478
Thomas Graf97defe12015-01-02 23:00:20 +0100479 rcu_assign_pointer(ht->future_tbl, new_tbl);
480 synchronize_rcu();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200481
Thomas Graf97defe12015-01-02 23:00:20 +0100482 /* Link the first entry in the old bucket to the end of the
483 * bucket in the new table. As entries are concurrently being
484 * added to the new table, lock down the new bucket. As we
485 * always divide the size in half when shrinking, each bucket
486 * in the new table maps to exactly two buckets in the old
487 * table.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200488 */
Thomas Graf97defe12015-01-02 23:00:20 +0100489 for (new_hash = 0; new_hash < new_tbl->size; new_hash++) {
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100490 lock_buckets(new_tbl, tbl, new_hash);
Thomas Graf97defe12015-01-02 23:00:20 +0100491
492 rcu_assign_pointer(*bucket_tail(new_tbl, new_hash),
493 tbl->buckets[new_hash]);
Thomas Graf7cd10db2015-02-05 02:03:35 +0100494 ASSERT_BUCKET_LOCK(ht, tbl, new_hash + new_tbl->size);
Thomas Graf97defe12015-01-02 23:00:20 +0100495 rcu_assign_pointer(*bucket_tail(new_tbl, new_hash),
496 tbl->buckets[new_hash + new_tbl->size]);
497
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100498 unlock_buckets(new_tbl, tbl, new_hash);
Eric Dumazet5beb5c92015-02-26 07:20:34 -0800499 cond_resched();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200500 }
501
502 /* Publish the new, valid hash table */
Thomas Graf97defe12015-01-02 23:00:20 +0100503 rcu_assign_pointer(ht->tbl, new_tbl);
Ying Xuec0c09bf2015-01-07 13:41:56 +0800504 atomic_dec(&ht->shift);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200505
506 /* Wait for readers. No new readers will have references to the
507 * old hash table.
508 */
509 synchronize_rcu();
510
511 bucket_table_free(tbl);
512
513 return 0;
514}
515EXPORT_SYMBOL_GPL(rhashtable_shrink);
516
Thomas Graf97defe12015-01-02 23:00:20 +0100517static void rht_deferred_worker(struct work_struct *work)
518{
519 struct rhashtable *ht;
520 struct bucket_table *tbl;
Herbert Xuf2dba9c2015-02-04 07:33:23 +1100521 struct rhashtable_walker *walker;
Thomas Graf97defe12015-01-02 23:00:20 +0100522
Ying Xue57699a42015-01-16 11:13:09 +0800523 ht = container_of(work, struct rhashtable, run_work);
Thomas Graf97defe12015-01-02 23:00:20 +0100524 mutex_lock(&ht->mutex);
Herbert Xu28134a52015-02-04 07:33:22 +1100525 if (ht->being_destroyed)
526 goto unlock;
527
Thomas Graf97defe12015-01-02 23:00:20 +0100528 tbl = rht_dereference(ht->tbl, ht);
529
Herbert Xuf2dba9c2015-02-04 07:33:23 +1100530 list_for_each_entry(walker, &ht->walkers, list)
531 walker->resize = true;
532
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100533 if (rht_grow_above_75(ht, tbl->size))
Thomas Graf97defe12015-01-02 23:00:20 +0100534 rhashtable_expand(ht);
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100535 else if (rht_shrink_below_30(ht, tbl->size))
Thomas Graf97defe12015-01-02 23:00:20 +0100536 rhashtable_shrink(ht);
Herbert Xu28134a52015-02-04 07:33:22 +1100537unlock:
Thomas Graf97defe12015-01-02 23:00:20 +0100538 mutex_unlock(&ht->mutex);
539}
540
Ying Xuedb304852015-01-07 13:41:54 +0800541static void __rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj,
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100542 struct bucket_table *tbl,
543 const struct bucket_table *old_tbl, u32 hash)
Ying Xuedb304852015-01-07 13:41:54 +0800544{
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100545 bool no_resize_running = tbl == old_tbl;
Thomas Graf020219a2015-02-06 16:08:43 +0000546 struct rhash_head *head;
547
548 hash = rht_bucket_index(tbl, hash);
549 head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash);
Ying Xuedb304852015-01-07 13:41:54 +0800550
Thomas Graf7cd10db2015-02-05 02:03:35 +0100551 ASSERT_BUCKET_LOCK(ht, tbl, hash);
552
Ying Xuedb304852015-01-07 13:41:54 +0800553 if (rht_is_a_nulls(head))
554 INIT_RHT_NULLS_HEAD(obj->next, ht, hash);
555 else
556 RCU_INIT_POINTER(obj->next, head);
557
558 rcu_assign_pointer(tbl->buckets[hash], obj);
559
560 atomic_inc(&ht->nelems);
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100561 if (no_resize_running && rht_grow_above_75(ht, tbl->size))
562 schedule_work(&ht->run_work);
Ying Xuedb304852015-01-07 13:41:54 +0800563}
564
Thomas Graf7e1e7762014-08-02 11:47:44 +0200565/**
Ying Xuedb304852015-01-07 13:41:54 +0800566 * rhashtable_insert - insert object into hash table
Thomas Graf7e1e7762014-08-02 11:47:44 +0200567 * @ht: hash table
568 * @obj: pointer to hash head inside object
Thomas Graf7e1e7762014-08-02 11:47:44 +0200569 *
Thomas Graf97defe12015-01-02 23:00:20 +0100570 * Will take a per bucket spinlock to protect against mutual mutations
571 * on the same bucket. Multiple insertions may occur in parallel unless
572 * they map to the same bucket lock.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200573 *
Thomas Graf97defe12015-01-02 23:00:20 +0100574 * It is safe to call this function from atomic context.
575 *
576 * Will trigger an automatic deferred table resizing if the size grows
577 * beyond the watermark indicated by grow_decision() which can be passed
578 * to rhashtable_init().
Thomas Graf7e1e7762014-08-02 11:47:44 +0200579 */
Thomas Graf6eba8222014-11-13 13:45:46 +0100580void rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200581{
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100582 struct bucket_table *tbl, *old_tbl;
Thomas Graf97defe12015-01-02 23:00:20 +0100583 unsigned hash;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200584
Thomas Graf97defe12015-01-02 23:00:20 +0100585 rcu_read_lock();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200586
Thomas Graf97defe12015-01-02 23:00:20 +0100587 tbl = rht_dereference_rcu(ht->future_tbl, ht);
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100588 old_tbl = rht_dereference_rcu(ht->tbl, ht);
Thomas Graf020219a2015-02-06 16:08:43 +0000589 hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
Thomas Graf97defe12015-01-02 23:00:20 +0100590
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100591 lock_buckets(tbl, old_tbl, hash);
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100592 __rhashtable_insert(ht, obj, tbl, old_tbl, hash);
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100593 unlock_buckets(tbl, old_tbl, hash);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200594
Thomas Graf97defe12015-01-02 23:00:20 +0100595 rcu_read_unlock();
Thomas Graf7e1e7762014-08-02 11:47:44 +0200596}
597EXPORT_SYMBOL_GPL(rhashtable_insert);
598
599/**
Thomas Graf7e1e7762014-08-02 11:47:44 +0200600 * rhashtable_remove - remove object from hash table
601 * @ht: hash table
602 * @obj: pointer to hash head inside object
Thomas Graf7e1e7762014-08-02 11:47:44 +0200603 *
604 * Since the hash chain is single linked, the removal operation needs to
605 * walk the bucket chain upon removal. The removal operation is thus
606 * considerable slow if the hash table is not correctly sized.
607 *
Ying Xuedb304852015-01-07 13:41:54 +0800608 * Will automatically shrink the table via rhashtable_expand() if the
Thomas Graf7e1e7762014-08-02 11:47:44 +0200609 * shrink_decision function specified at rhashtable_init() returns true.
610 *
611 * The caller must ensure that no concurrent table mutations occur. It is
612 * however valid to have concurrent lookups if they are RCU protected.
613 */
Thomas Graf6eba8222014-11-13 13:45:46 +0100614bool rhashtable_remove(struct rhashtable *ht, struct rhash_head *obj)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200615{
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100616 struct bucket_table *tbl, *new_tbl, *old_tbl;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200617 struct rhash_head __rcu **pprev;
Thomas Grafcf52d522015-02-05 02:03:36 +0100618 struct rhash_head *he, *he2;
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100619 unsigned int hash, new_hash;
Thomas Graffe6a0432015-01-21 11:54:01 +0000620 bool ret = false;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200621
Thomas Graf97defe12015-01-02 23:00:20 +0100622 rcu_read_lock();
Thomas Graf020219a2015-02-06 16:08:43 +0000623 old_tbl = rht_dereference_rcu(ht->tbl, ht);
624 tbl = new_tbl = rht_dereference_rcu(ht->future_tbl, ht);
Thomas Grafcf52d522015-02-05 02:03:36 +0100625 new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
Thomas Graf7e1e7762014-08-02 11:47:44 +0200626
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100627 lock_buckets(new_tbl, old_tbl, new_hash);
Thomas Graf97defe12015-01-02 23:00:20 +0100628restart:
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100629 hash = rht_bucket_index(tbl, new_hash);
Thomas Graf97defe12015-01-02 23:00:20 +0100630 pprev = &tbl->buckets[hash];
631 rht_for_each(he, tbl, hash) {
Thomas Graf7e1e7762014-08-02 11:47:44 +0200632 if (he != obj) {
633 pprev = &he->next;
634 continue;
635 }
636
Thomas Graf7cd10db2015-02-05 02:03:35 +0100637 ASSERT_BUCKET_LOCK(ht, tbl, hash);
Thomas Graf897362e2015-01-02 23:00:18 +0100638
Thomas Graf020219a2015-02-06 16:08:43 +0000639 if (old_tbl->size > new_tbl->size && tbl == old_tbl &&
640 !rht_is_a_nulls(obj->next) &&
641 head_hashfn(ht, tbl, obj->next) != hash) {
642 rcu_assign_pointer(*pprev, (struct rhash_head *) rht_marker(ht, hash));
643 } else if (unlikely(old_tbl->size < new_tbl->size && tbl == new_tbl)) {
644 rht_for_each_continue(he2, obj->next, tbl, hash) {
Thomas Grafcf52d522015-02-05 02:03:36 +0100645 if (head_hashfn(ht, tbl, he2) == hash) {
646 rcu_assign_pointer(*pprev, he2);
647 goto found;
648 }
649 }
650
Thomas Graf020219a2015-02-06 16:08:43 +0000651 rcu_assign_pointer(*pprev, (struct rhash_head *) rht_marker(ht, hash));
Thomas Grafcf52d522015-02-05 02:03:36 +0100652 } else {
653 rcu_assign_pointer(*pprev, obj->next);
654 }
655
656found:
Thomas Graffe6a0432015-01-21 11:54:01 +0000657 ret = true;
658 break;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200659 }
660
Thomas Graffe6a0432015-01-21 11:54:01 +0000661 /* The entry may be linked in either 'tbl', 'future_tbl', or both.
662 * 'future_tbl' only exists for a short period of time during
663 * resizing. Thus traversing both is fine and the added cost is
664 * very rare.
665 */
Thomas Graf020219a2015-02-06 16:08:43 +0000666 if (tbl != old_tbl) {
667 tbl = old_tbl;
Thomas Graf97defe12015-01-02 23:00:20 +0100668 goto restart;
669 }
670
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100671 unlock_buckets(new_tbl, old_tbl, new_hash);
Thomas Graffe6a0432015-01-21 11:54:01 +0000672
673 if (ret) {
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100674 bool no_resize_running = new_tbl == old_tbl;
675
Thomas Graffe6a0432015-01-21 11:54:01 +0000676 atomic_dec(&ht->nelems);
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100677 if (no_resize_running && rht_shrink_below_30(ht, new_tbl->size))
678 schedule_work(&ht->run_work);
Thomas Graffe6a0432015-01-21 11:54:01 +0000679 }
680
Thomas Graf97defe12015-01-02 23:00:20 +0100681 rcu_read_unlock();
682
Thomas Graffe6a0432015-01-21 11:54:01 +0000683 return ret;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200684}
685EXPORT_SYMBOL_GPL(rhashtable_remove);
686
Ying Xueefb975a62015-01-07 13:41:52 +0800687struct rhashtable_compare_arg {
688 struct rhashtable *ht;
689 const void *key;
690};
691
692static bool rhashtable_compare(void *ptr, void *arg)
693{
694 struct rhashtable_compare_arg *x = arg;
695 struct rhashtable *ht = x->ht;
696
697 return !memcmp(ptr + ht->p.key_offset, x->key, ht->p.key_len);
698}
699
Thomas Graf7e1e7762014-08-02 11:47:44 +0200700/**
701 * rhashtable_lookup - lookup key in hash table
702 * @ht: hash table
703 * @key: pointer to key
704 *
705 * Computes the hash value for the key and traverses the bucket chain looking
706 * for a entry with an identical key. The first matching entry is returned.
707 *
708 * This lookup function may only be used for fixed key hash table (key_len
Ying Xuedb304852015-01-07 13:41:54 +0800709 * parameter set). It will BUG() if used inappropriately.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200710 *
Thomas Graf97defe12015-01-02 23:00:20 +0100711 * Lookups may occur in parallel with hashtable mutations and resizing.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200712 */
Thomas Graf97defe12015-01-02 23:00:20 +0100713void *rhashtable_lookup(struct rhashtable *ht, const void *key)
Thomas Graf7e1e7762014-08-02 11:47:44 +0200714{
Ying Xueefb975a62015-01-07 13:41:52 +0800715 struct rhashtable_compare_arg arg = {
716 .ht = ht,
717 .key = key,
718 };
Thomas Graf7e1e7762014-08-02 11:47:44 +0200719
720 BUG_ON(!ht->p.key_len);
721
Ying Xueefb975a62015-01-07 13:41:52 +0800722 return rhashtable_lookup_compare(ht, key, &rhashtable_compare, &arg);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200723}
724EXPORT_SYMBOL_GPL(rhashtable_lookup);
725
726/**
727 * rhashtable_lookup_compare - search hash table with compare function
728 * @ht: hash table
Thomas Graf8d24c0b2015-01-02 23:00:14 +0100729 * @key: the pointer to the key
Thomas Graf7e1e7762014-08-02 11:47:44 +0200730 * @compare: compare function, must return true on match
731 * @arg: argument passed on to compare function
732 *
733 * Traverses the bucket chain behind the provided hash value and calls the
734 * specified compare function for each entry.
735 *
Thomas Graf97defe12015-01-02 23:00:20 +0100736 * Lookups may occur in parallel with hashtable mutations and resizing.
Thomas Graf7e1e7762014-08-02 11:47:44 +0200737 *
738 * Returns the first entry on which the compare function returned true.
739 */
Thomas Graf97defe12015-01-02 23:00:20 +0100740void *rhashtable_lookup_compare(struct rhashtable *ht, const void *key,
Thomas Graf7e1e7762014-08-02 11:47:44 +0200741 bool (*compare)(void *, void *), void *arg)
742{
Thomas Graf97defe12015-01-02 23:00:20 +0100743 const struct bucket_table *tbl, *old_tbl;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200744 struct rhash_head *he;
Thomas Graf8d24c0b2015-01-02 23:00:14 +0100745 u32 hash;
Thomas Graf7e1e7762014-08-02 11:47:44 +0200746
Thomas Graf97defe12015-01-02 23:00:20 +0100747 rcu_read_lock();
748
749 old_tbl = rht_dereference_rcu(ht->tbl, ht);
750 tbl = rht_dereference_rcu(ht->future_tbl, ht);
Thomas Graf8d24c0b2015-01-02 23:00:14 +0100751 hash = key_hashfn(ht, key, ht->p.key_len);
Thomas Graf97defe12015-01-02 23:00:20 +0100752restart:
753 rht_for_each_rcu(he, tbl, rht_bucket_index(tbl, hash)) {
Thomas Graf7e1e7762014-08-02 11:47:44 +0200754 if (!compare(rht_obj(ht, he), arg))
755 continue;
Thomas Graf97defe12015-01-02 23:00:20 +0100756 rcu_read_unlock();
Thomas Grafa4b18cd2015-01-02 23:00:15 +0100757 return rht_obj(ht, he);
Thomas Graf7e1e7762014-08-02 11:47:44 +0200758 }
759
Thomas Graf97defe12015-01-02 23:00:20 +0100760 if (unlikely(tbl != old_tbl)) {
761 tbl = old_tbl;
762 goto restart;
763 }
764 rcu_read_unlock();
765
Thomas Graf7e1e7762014-08-02 11:47:44 +0200766 return NULL;
767}
768EXPORT_SYMBOL_GPL(rhashtable_lookup_compare);
769
Ying Xuedb304852015-01-07 13:41:54 +0800770/**
771 * rhashtable_lookup_insert - lookup and insert object into hash table
772 * @ht: hash table
773 * @obj: pointer to hash head inside object
774 *
775 * Locks down the bucket chain in both the old and new table if a resize
776 * is in progress to ensure that writers can't remove from the old table
777 * and can't insert to the new table during the atomic operation of search
778 * and insertion. Searches for duplicates in both the old and new table if
779 * a resize is in progress.
780 *
781 * This lookup function may only be used for fixed key hash table (key_len
782 * parameter set). It will BUG() if used inappropriately.
783 *
784 * It is safe to call this function from atomic context.
785 *
786 * Will trigger an automatic deferred table resizing if the size grows
787 * beyond the watermark indicated by grow_decision() which can be passed
788 * to rhashtable_init().
789 */
790bool rhashtable_lookup_insert(struct rhashtable *ht, struct rhash_head *obj)
791{
Ying Xue7a868d12015-01-12 14:52:22 +0800792 struct rhashtable_compare_arg arg = {
793 .ht = ht,
794 .key = rht_obj(ht, obj) + ht->p.key_offset,
795 };
796
797 BUG_ON(!ht->p.key_len);
798
799 return rhashtable_lookup_compare_insert(ht, obj, &rhashtable_compare,
800 &arg);
801}
802EXPORT_SYMBOL_GPL(rhashtable_lookup_insert);
803
804/**
805 * rhashtable_lookup_compare_insert - search and insert object to hash table
806 * with compare function
807 * @ht: hash table
808 * @obj: pointer to hash head inside object
809 * @compare: compare function, must return true on match
810 * @arg: argument passed on to compare function
811 *
812 * Locks down the bucket chain in both the old and new table if a resize
813 * is in progress to ensure that writers can't remove from the old table
814 * and can't insert to the new table during the atomic operation of search
815 * and insertion. Searches for duplicates in both the old and new table if
816 * a resize is in progress.
817 *
818 * Lookups may occur in parallel with hashtable mutations and resizing.
819 *
820 * Will trigger an automatic deferred table resizing if the size grows
821 * beyond the watermark indicated by grow_decision() which can be passed
822 * to rhashtable_init().
823 */
824bool rhashtable_lookup_compare_insert(struct rhashtable *ht,
825 struct rhash_head *obj,
826 bool (*compare)(void *, void *),
827 void *arg)
828{
Ying Xuedb304852015-01-07 13:41:54 +0800829 struct bucket_table *new_tbl, *old_tbl;
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100830 u32 new_hash;
Ying Xuedb304852015-01-07 13:41:54 +0800831 bool success = true;
832
833 BUG_ON(!ht->p.key_len);
834
835 rcu_read_lock();
Ying Xuedb304852015-01-07 13:41:54 +0800836 old_tbl = rht_dereference_rcu(ht->tbl, ht);
Ying Xuedb304852015-01-07 13:41:54 +0800837 new_tbl = rht_dereference_rcu(ht->future_tbl, ht);
Thomas Graf020219a2015-02-06 16:08:43 +0000838 new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100839
840 lock_buckets(new_tbl, old_tbl, new_hash);
Ying Xuedb304852015-01-07 13:41:54 +0800841
Ying Xue7a868d12015-01-12 14:52:22 +0800842 if (rhashtable_lookup_compare(ht, rht_obj(ht, obj) + ht->p.key_offset,
843 compare, arg)) {
Ying Xuedb304852015-01-07 13:41:54 +0800844 success = false;
845 goto exit;
846 }
847
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +0100848 __rhashtable_insert(ht, obj, new_tbl, old_tbl, new_hash);
Ying Xuedb304852015-01-07 13:41:54 +0800849
850exit:
Thomas Grafa5ec68e2015-02-05 02:03:32 +0100851 unlock_buckets(new_tbl, old_tbl, new_hash);
Ying Xuedb304852015-01-07 13:41:54 +0800852 rcu_read_unlock();
853
854 return success;
855}
Ying Xue7a868d12015-01-12 14:52:22 +0800856EXPORT_SYMBOL_GPL(rhashtable_lookup_compare_insert);
Ying Xuedb304852015-01-07 13:41:54 +0800857
Herbert Xuf2dba9c2015-02-04 07:33:23 +1100858/**
859 * rhashtable_walk_init - Initialise an iterator
860 * @ht: Table to walk over
861 * @iter: Hash table Iterator
862 *
863 * This function prepares a hash table walk.
864 *
865 * Note that if you restart a walk after rhashtable_walk_stop you
866 * may see the same object twice. Also, you may miss objects if
867 * there are removals in between rhashtable_walk_stop and the next
868 * call to rhashtable_walk_start.
869 *
870 * For a completely stable walk you should construct your own data
871 * structure outside the hash table.
872 *
873 * This function may sleep so you must not call it from interrupt
874 * context or with spin locks held.
875 *
876 * You must call rhashtable_walk_exit if this function returns
877 * successfully.
878 */
879int rhashtable_walk_init(struct rhashtable *ht, struct rhashtable_iter *iter)
880{
881 iter->ht = ht;
882 iter->p = NULL;
883 iter->slot = 0;
884 iter->skip = 0;
885
886 iter->walker = kmalloc(sizeof(*iter->walker), GFP_KERNEL);
887 if (!iter->walker)
888 return -ENOMEM;
889
Sasha Levin71bb0012015-02-23 04:35:06 -0500890 INIT_LIST_HEAD(&iter->walker->list);
891 iter->walker->resize = false;
892
Herbert Xuf2dba9c2015-02-04 07:33:23 +1100893 mutex_lock(&ht->mutex);
894 list_add(&iter->walker->list, &ht->walkers);
895 mutex_unlock(&ht->mutex);
896
897 return 0;
898}
899EXPORT_SYMBOL_GPL(rhashtable_walk_init);
900
901/**
902 * rhashtable_walk_exit - Free an iterator
903 * @iter: Hash table Iterator
904 *
905 * This function frees resources allocated by rhashtable_walk_init.
906 */
907void rhashtable_walk_exit(struct rhashtable_iter *iter)
908{
909 mutex_lock(&iter->ht->mutex);
910 list_del(&iter->walker->list);
911 mutex_unlock(&iter->ht->mutex);
912 kfree(iter->walker);
913}
914EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
915
916/**
917 * rhashtable_walk_start - Start a hash table walk
918 * @iter: Hash table iterator
919 *
920 * Start a hash table walk. Note that we take the RCU lock in all
921 * cases including when we return an error. So you must always call
922 * rhashtable_walk_stop to clean up.
923 *
924 * Returns zero if successful.
925 *
926 * Returns -EAGAIN if resize event occured. Note that the iterator
927 * will rewind back to the beginning and you may use it immediately
928 * by calling rhashtable_walk_next.
929 */
930int rhashtable_walk_start(struct rhashtable_iter *iter)
931{
932 rcu_read_lock();
933
934 if (iter->walker->resize) {
935 iter->slot = 0;
936 iter->skip = 0;
937 iter->walker->resize = false;
938 return -EAGAIN;
939 }
940
941 return 0;
942}
943EXPORT_SYMBOL_GPL(rhashtable_walk_start);
944
945/**
946 * rhashtable_walk_next - Return the next object and advance the iterator
947 * @iter: Hash table iterator
948 *
949 * Note that you must call rhashtable_walk_stop when you are finished
950 * with the walk.
951 *
952 * Returns the next object or NULL when the end of the table is reached.
953 *
954 * Returns -EAGAIN if resize event occured. Note that the iterator
955 * will rewind back to the beginning and you may continue to use it.
956 */
957void *rhashtable_walk_next(struct rhashtable_iter *iter)
958{
959 const struct bucket_table *tbl;
960 struct rhashtable *ht = iter->ht;
961 struct rhash_head *p = iter->p;
962 void *obj = NULL;
963
964 tbl = rht_dereference_rcu(ht->tbl, ht);
965
966 if (p) {
967 p = rht_dereference_bucket_rcu(p->next, tbl, iter->slot);
968 goto next;
969 }
970
971 for (; iter->slot < tbl->size; iter->slot++) {
972 int skip = iter->skip;
973
974 rht_for_each_rcu(p, tbl, iter->slot) {
975 if (!skip)
976 break;
977 skip--;
978 }
979
980next:
981 if (!rht_is_a_nulls(p)) {
982 iter->skip++;
983 iter->p = p;
984 obj = rht_obj(ht, p);
985 goto out;
986 }
987
988 iter->skip = 0;
989 }
990
991 iter->p = NULL;
992
993out:
994 if (iter->walker->resize) {
995 iter->p = NULL;
996 iter->slot = 0;
997 iter->skip = 0;
998 iter->walker->resize = false;
999 return ERR_PTR(-EAGAIN);
1000 }
1001
1002 return obj;
1003}
1004EXPORT_SYMBOL_GPL(rhashtable_walk_next);
1005
1006/**
1007 * rhashtable_walk_stop - Finish a hash table walk
1008 * @iter: Hash table iterator
1009 *
1010 * Finish a hash table walk.
1011 */
1012void rhashtable_walk_stop(struct rhashtable_iter *iter)
1013{
1014 rcu_read_unlock();
1015 iter->p = NULL;
1016}
1017EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
1018
Ying Xue94000172014-09-03 09:22:36 +08001019static size_t rounded_hashtable_size(struct rhashtable_params *params)
Thomas Graf7e1e7762014-08-02 11:47:44 +02001020{
Ying Xue94000172014-09-03 09:22:36 +08001021 return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
1022 1UL << params->min_shift);
Thomas Graf7e1e7762014-08-02 11:47:44 +02001023}
1024
1025/**
1026 * rhashtable_init - initialize a new hash table
1027 * @ht: hash table to be initialized
1028 * @params: configuration parameters
1029 *
1030 * Initializes a new hash table based on the provided configuration
1031 * parameters. A table can be configured either with a variable or
1032 * fixed length key:
1033 *
1034 * Configuration Example 1: Fixed length keys
1035 * struct test_obj {
1036 * int key;
1037 * void * my_member;
1038 * struct rhash_head node;
1039 * };
1040 *
1041 * struct rhashtable_params params = {
1042 * .head_offset = offsetof(struct test_obj, node),
1043 * .key_offset = offsetof(struct test_obj, key),
1044 * .key_len = sizeof(int),
Daniel Borkmann87545892014-12-10 16:33:11 +01001045 * .hashfn = jhash,
Thomas Graff89bd6f2015-01-02 23:00:21 +01001046 * .nulls_base = (1U << RHT_BASE_SHIFT),
Thomas Graf7e1e7762014-08-02 11:47:44 +02001047 * };
1048 *
1049 * Configuration Example 2: Variable length keys
1050 * struct test_obj {
1051 * [...]
1052 * struct rhash_head node;
1053 * };
1054 *
1055 * u32 my_hash_fn(const void *data, u32 seed)
1056 * {
1057 * struct test_obj *obj = data;
1058 *
1059 * return [... hash ...];
1060 * }
1061 *
1062 * struct rhashtable_params params = {
1063 * .head_offset = offsetof(struct test_obj, node),
Daniel Borkmann87545892014-12-10 16:33:11 +01001064 * .hashfn = jhash,
Thomas Graf7e1e7762014-08-02 11:47:44 +02001065 * .obj_hashfn = my_hash_fn,
Thomas Graf7e1e7762014-08-02 11:47:44 +02001066 * };
1067 */
1068int rhashtable_init(struct rhashtable *ht, struct rhashtable_params *params)
1069{
1070 struct bucket_table *tbl;
1071 size_t size;
1072
1073 size = HASH_DEFAULT_SIZE;
1074
1075 if ((params->key_len && !params->hashfn) ||
1076 (!params->key_len && !params->obj_hashfn))
1077 return -EINVAL;
1078
Thomas Graff89bd6f2015-01-02 23:00:21 +01001079 if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
1080 return -EINVAL;
1081
Ying Xue94000172014-09-03 09:22:36 +08001082 params->min_shift = max_t(size_t, params->min_shift,
1083 ilog2(HASH_MIN_SIZE));
1084
Thomas Graf7e1e7762014-08-02 11:47:44 +02001085 if (params->nelem_hint)
Ying Xue94000172014-09-03 09:22:36 +08001086 size = rounded_hashtable_size(params);
Thomas Graf7e1e7762014-08-02 11:47:44 +02001087
Thomas Graf97defe12015-01-02 23:00:20 +01001088 memset(ht, 0, sizeof(*ht));
1089 mutex_init(&ht->mutex);
1090 memcpy(&ht->p, params, sizeof(*params));
Herbert Xuf2dba9c2015-02-04 07:33:23 +11001091 INIT_LIST_HEAD(&ht->walkers);
Thomas Graf97defe12015-01-02 23:00:20 +01001092
1093 if (params->locks_mul)
1094 ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
1095 else
1096 ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;
1097
1098 tbl = bucket_table_alloc(ht, size);
Thomas Graf7e1e7762014-08-02 11:47:44 +02001099 if (tbl == NULL)
1100 return -ENOMEM;
1101
Ying Xue545a1482015-01-07 13:41:57 +08001102 atomic_set(&ht->nelems, 0);
Ying Xuec0c09bf2015-01-07 13:41:56 +08001103 atomic_set(&ht->shift, ilog2(tbl->size));
Thomas Graf7e1e7762014-08-02 11:47:44 +02001104 RCU_INIT_POINTER(ht->tbl, tbl);
Thomas Graf97defe12015-01-02 23:00:20 +01001105 RCU_INIT_POINTER(ht->future_tbl, tbl);
Thomas Graf7e1e7762014-08-02 11:47:44 +02001106
1107 if (!ht->p.hash_rnd)
1108 get_random_bytes(&ht->p.hash_rnd, sizeof(ht->p.hash_rnd));
1109
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +01001110 INIT_WORK(&ht->run_work, rht_deferred_worker);
Thomas Graf97defe12015-01-02 23:00:20 +01001111
Thomas Graf7e1e7762014-08-02 11:47:44 +02001112 return 0;
1113}
1114EXPORT_SYMBOL_GPL(rhashtable_init);
1115
1116/**
1117 * rhashtable_destroy - destroy hash table
1118 * @ht: the hash table to destroy
1119 *
Pablo Neira Ayusoae82ddc2014-09-02 00:26:05 +02001120 * Frees the bucket array. This function is not rcu safe, therefore the caller
1121 * has to make sure that no resizing may happen by unpublishing the hashtable
1122 * and waiting for the quiescent cycle before releasing the bucket array.
Thomas Graf7e1e7762014-08-02 11:47:44 +02001123 */
Thomas Graf97defe12015-01-02 23:00:20 +01001124void rhashtable_destroy(struct rhashtable *ht)
Thomas Graf7e1e7762014-08-02 11:47:44 +02001125{
Thomas Graf97defe12015-01-02 23:00:20 +01001126 ht->being_destroyed = true;
1127
Daniel Borkmann4c4b52d2015-02-25 16:31:54 +01001128 cancel_work_sync(&ht->run_work);
Ying Xue57699a42015-01-16 11:13:09 +08001129
Thomas Graf97defe12015-01-02 23:00:20 +01001130 mutex_lock(&ht->mutex);
Thomas Graf97defe12015-01-02 23:00:20 +01001131 bucket_table_free(rht_dereference(ht->tbl, ht));
Thomas Graf97defe12015-01-02 23:00:20 +01001132 mutex_unlock(&ht->mutex);
Thomas Graf7e1e7762014-08-02 11:47:44 +02001133}
1134EXPORT_SYMBOL_GPL(rhashtable_destroy);