2014-08-02 13:47:44 +04:00
|
|
|
/*
|
|
|
|
* Resizable, Scalable, Concurrent Hash Table
|
|
|
|
*
|
2015-03-20 13:57:00 +03:00
|
|
|
* Copyright (c) 2015 Herbert Xu <herbert@gondor.apana.org.au>
|
2015-02-05 04:03:32 +03:00
|
|
|
* Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
|
2014-08-02 13:47:44 +04:00
|
|
|
* Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
|
|
|
|
*
|
|
|
|
* Code partially derived from nft_hash
|
2015-03-20 13:57:00 +03:00
|
|
|
* Rewritten with rehash code from br_multicast plus single list
|
|
|
|
* pointer as suggested by Josh Triplett
|
2014-08-02 13:47:44 +04:00
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License version 2 as
|
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*/
|
|
|
|
|
2015-05-15 06:30:47 +03:00
|
|
|
#include <linux/atomic.h>
|
2014-08-02 13:47:44 +04:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/log2.h>
|
2015-02-26 18:20:34 +03:00
|
|
|
#include <linux/sched.h>
|
2014-08-02 13:47:44 +04:00
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/mm.h>
|
2014-12-10 18:33:11 +03:00
|
|
|
#include <linux/jhash.h>
|
2014-08-02 13:47:44 +04:00
|
|
|
#include <linux/random.h>
|
|
|
|
#include <linux/rhashtable.h>
|
2015-02-09 06:04:03 +03:00
|
|
|
#include <linux/err.h>
|
2015-06-06 23:07:23 +03:00
|
|
|
#include <linux/export.h>
|
2014-08-02 13:47:44 +04:00
|
|
|
|
|
|
|
#define HASH_DEFAULT_SIZE 64UL
|
2015-03-18 12:01:16 +03:00
|
|
|
#define HASH_MIN_SIZE 4U
|
2015-01-03 01:00:20 +03:00
|
|
|
#define BUCKET_LOCKS_PER_CPU 128UL
|
|
|
|
|
2015-03-10 01:27:55 +03:00
|
|
|
static u32 head_hashfn(struct rhashtable *ht,
|
2015-01-03 01:00:14 +03:00
|
|
|
const struct bucket_table *tbl,
|
|
|
|
const struct rhash_head *he)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
2015-03-20 13:57:00 +03:00
|
|
|
return rht_head_hashfn(ht, tbl, he, ht->p);
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
2015-02-05 04:03:34 +03:00
|
|
|
#ifdef CONFIG_PROVE_LOCKING
|
|
|
|
#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
|
|
|
|
|
|
|
|
int lockdep_rht_mutex_is_held(struct rhashtable *ht)
|
|
|
|
{
|
|
|
|
return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
|
|
|
|
|
|
|
|
int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
|
|
|
|
{
|
2015-03-20 13:57:00 +03:00
|
|
|
spinlock_t *lock = rht_bucket_lock(tbl, hash);
|
2015-02-05 04:03:34 +03:00
|
|
|
|
|
|
|
return (debug_locks) ? lockdep_is_held(lock) : 1;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
|
|
|
|
#else
|
|
|
|
#define ASSERT_RHT_MUTEX(HT)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
2015-03-23 16:50:27 +03:00
|
|
|
static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl,
|
|
|
|
gfp_t gfp)
|
2015-01-03 01:00:20 +03:00
|
|
|
{
|
|
|
|
unsigned int i, size;
|
|
|
|
#if defined(CONFIG_PROVE_LOCKING)
|
|
|
|
unsigned int nr_pcpus = 2;
|
|
|
|
#else
|
|
|
|
unsigned int nr_pcpus = num_possible_cpus();
|
|
|
|
#endif
|
|
|
|
|
|
|
|
nr_pcpus = min_t(unsigned int, nr_pcpus, 32UL);
|
|
|
|
size = roundup_pow_of_two(nr_pcpus * ht->p.locks_mul);
|
|
|
|
|
2015-02-05 04:03:32 +03:00
|
|
|
/* Never allocate more than 0.5 locks per bucket */
|
|
|
|
size = min_t(unsigned int, size, tbl->size >> 1);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
|
|
|
if (sizeof(spinlock_t) != 0) {
|
|
|
|
#ifdef CONFIG_NUMA
|
2015-03-23 16:50:27 +03:00
|
|
|
if (size * sizeof(spinlock_t) > PAGE_SIZE &&
|
|
|
|
gfp == GFP_KERNEL)
|
2015-01-03 01:00:20 +03:00
|
|
|
tbl->locks = vmalloc(size * sizeof(spinlock_t));
|
|
|
|
else
|
|
|
|
#endif
|
|
|
|
tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
|
2015-03-23 16:50:27 +03:00
|
|
|
gfp);
|
2015-01-03 01:00:20 +03:00
|
|
|
if (!tbl->locks)
|
|
|
|
return -ENOMEM;
|
|
|
|
for (i = 0; i < size; i++)
|
|
|
|
spin_lock_init(&tbl->locks[i]);
|
|
|
|
}
|
|
|
|
tbl->locks_mask = size - 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void bucket_table_free(const struct bucket_table *tbl)
|
|
|
|
{
|
|
|
|
if (tbl)
|
|
|
|
kvfree(tbl->locks);
|
|
|
|
|
|
|
|
kvfree(tbl);
|
|
|
|
}
|
|
|
|
|
2015-03-14 05:57:23 +03:00
|
|
|
static void bucket_table_free_rcu(struct rcu_head *head)
|
|
|
|
{
|
|
|
|
bucket_table_free(container_of(head, struct bucket_table, rcu));
|
|
|
|
}
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
|
2015-03-23 16:50:27 +03:00
|
|
|
size_t nbuckets,
|
|
|
|
gfp_t gfp)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
2015-02-20 02:53:38 +03:00
|
|
|
struct bucket_table *tbl = NULL;
|
2014-08-02 13:47:44 +04:00
|
|
|
size_t size;
|
2015-01-03 01:00:21 +03:00
|
|
|
int i;
|
2014-08-02 13:47:44 +04:00
|
|
|
|
|
|
|
size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]);
|
2015-03-23 16:50:27 +03:00
|
|
|
if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER) ||
|
|
|
|
gfp != GFP_KERNEL)
|
|
|
|
tbl = kzalloc(size, gfp | __GFP_NOWARN | __GFP_NORETRY);
|
|
|
|
if (tbl == NULL && gfp == GFP_KERNEL)
|
2014-08-02 13:47:44 +04:00
|
|
|
tbl = vzalloc(size);
|
|
|
|
if (tbl == NULL)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
tbl->size = nbuckets;
|
|
|
|
|
2015-03-23 16:50:27 +03:00
|
|
|
if (alloc_bucket_locks(ht, tbl, gfp) < 0) {
|
2015-01-03 01:00:20 +03:00
|
|
|
bucket_table_free(tbl);
|
|
|
|
return NULL;
|
|
|
|
}
|
2014-08-02 13:47:44 +04:00
|
|
|
|
2015-03-14 05:57:20 +03:00
|
|
|
INIT_LIST_HEAD(&tbl->walkers);
|
|
|
|
|
2015-03-14 05:57:22 +03:00
|
|
|
get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
|
|
|
|
|
2015-01-03 01:00:21 +03:00
|
|
|
for (i = 0; i < nbuckets; i++)
|
|
|
|
INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
return tbl;
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
static struct bucket_table *rhashtable_last_table(struct rhashtable *ht,
|
|
|
|
struct bucket_table *tbl)
|
|
|
|
{
|
|
|
|
struct bucket_table *new_tbl;
|
|
|
|
|
|
|
|
do {
|
|
|
|
new_tbl = tbl;
|
|
|
|
tbl = rht_dereference_rcu(tbl->future_tbl, ht);
|
|
|
|
} while (tbl);
|
|
|
|
|
|
|
|
return new_tbl;
|
|
|
|
}
|
|
|
|
|
2015-03-24 16:18:17 +03:00
|
|
|
static int rhashtable_rehash_one(struct rhashtable *ht, unsigned int old_hash)
|
2015-02-05 04:03:32 +03:00
|
|
|
{
|
2015-03-11 01:43:48 +03:00
|
|
|
struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
|
2015-03-23 16:50:26 +03:00
|
|
|
struct bucket_table *new_tbl = rhashtable_last_table(ht,
|
|
|
|
rht_dereference_rcu(old_tbl->future_tbl, ht));
|
2015-03-11 01:43:48 +03:00
|
|
|
struct rhash_head __rcu **pprev = &old_tbl->buckets[old_hash];
|
|
|
|
int err = -ENOENT;
|
|
|
|
struct rhash_head *head, *next, *entry;
|
|
|
|
spinlock_t *new_bucket_lock;
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int new_hash;
|
2015-03-11 01:43:48 +03:00
|
|
|
|
|
|
|
rht_for_each(entry, old_tbl, old_hash) {
|
|
|
|
err = 0;
|
|
|
|
next = rht_dereference_bucket(entry->next, old_tbl, old_hash);
|
|
|
|
|
|
|
|
if (rht_is_a_nulls(next))
|
|
|
|
break;
|
2015-02-05 04:03:32 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
pprev = &entry->next;
|
|
|
|
}
|
2015-02-05 04:03:32 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
if (err)
|
|
|
|
goto out;
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
new_hash = head_hashfn(ht, new_tbl, entry);
|
2014-08-02 13:47:44 +04:00
|
|
|
|
2015-03-20 13:57:00 +03:00
|
|
|
new_bucket_lock = rht_bucket_lock(new_tbl, new_hash);
|
2014-08-02 13:47:44 +04:00
|
|
|
|
2015-03-14 05:57:21 +03:00
|
|
|
spin_lock_nested(new_bucket_lock, SINGLE_DEPTH_NESTING);
|
2015-03-11 01:43:48 +03:00
|
|
|
head = rht_dereference_bucket(new_tbl->buckets[new_hash],
|
|
|
|
new_tbl, new_hash);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-09-22 11:51:52 +03:00
|
|
|
RCU_INIT_POINTER(entry->next, head);
|
2015-02-05 04:03:32 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
rcu_assign_pointer(new_tbl->buckets[new_hash], entry);
|
|
|
|
spin_unlock(new_bucket_lock);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
rcu_assign_pointer(*pprev, next);
|
2014-08-02 13:47:44 +04:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
out:
|
|
|
|
return err;
|
|
|
|
}
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-03-24 16:18:17 +03:00
|
|
|
static void rhashtable_rehash_chain(struct rhashtable *ht,
|
|
|
|
unsigned int old_hash)
|
2015-03-11 01:43:48 +03:00
|
|
|
{
|
|
|
|
struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
spinlock_t *old_bucket_lock;
|
|
|
|
|
2015-03-20 13:57:00 +03:00
|
|
|
old_bucket_lock = rht_bucket_lock(old_tbl, old_hash);
|
2015-02-05 04:03:32 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
spin_lock_bh(old_bucket_lock);
|
|
|
|
while (!rhashtable_rehash_one(ht, old_hash))
|
|
|
|
;
|
2015-03-14 05:57:24 +03:00
|
|
|
old_tbl->rehash++;
|
2015-03-11 01:43:48 +03:00
|
|
|
spin_unlock_bh(old_bucket_lock);
|
2015-01-03 01:00:20 +03:00
|
|
|
}
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
static int rhashtable_rehash_attach(struct rhashtable *ht,
|
|
|
|
struct bucket_table *old_tbl,
|
|
|
|
struct bucket_table *new_tbl)
|
2015-01-03 01:00:20 +03:00
|
|
|
{
|
2015-03-23 16:50:26 +03:00
|
|
|
/* Protect future_tbl using the first bucket lock. */
|
|
|
|
spin_lock_bh(old_tbl->locks);
|
|
|
|
|
|
|
|
/* Did somebody beat us to it? */
|
|
|
|
if (rcu_access_pointer(old_tbl->future_tbl)) {
|
|
|
|
spin_unlock_bh(old_tbl->locks);
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
2015-02-05 04:03:35 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
/* Make insertions go into the new, empty table right away. Deletions
|
|
|
|
* and lookups will be attempted in both tables until we synchronize.
|
|
|
|
*/
|
2015-03-14 05:57:25 +03:00
|
|
|
rcu_assign_pointer(old_tbl->future_tbl, new_tbl);
|
2015-03-11 01:43:48 +03:00
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
spin_unlock_bh(old_tbl->locks);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int rhashtable_rehash_table(struct rhashtable *ht)
|
|
|
|
{
|
|
|
|
struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
struct bucket_table *new_tbl;
|
|
|
|
struct rhashtable_walker *walker;
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int old_hash;
|
2015-03-23 16:50:26 +03:00
|
|
|
|
|
|
|
new_tbl = rht_dereference(old_tbl->future_tbl, ht);
|
|
|
|
if (!new_tbl)
|
|
|
|
return 0;
|
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
for (old_hash = 0; old_hash < old_tbl->size; old_hash++)
|
|
|
|
rhashtable_rehash_chain(ht, old_hash);
|
|
|
|
|
|
|
|
/* Publish the new table pointer. */
|
|
|
|
rcu_assign_pointer(ht->tbl, new_tbl);
|
|
|
|
|
2015-03-24 01:53:17 +03:00
|
|
|
spin_lock(&ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
list_for_each_entry(walker, &old_tbl->walkers, list)
|
|
|
|
walker->tbl = NULL;
|
2015-03-24 01:53:17 +03:00
|
|
|
spin_unlock(&ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
|
2015-03-11 01:43:48 +03:00
|
|
|
/* Wait for readers. All new readers will see the new
|
|
|
|
* table, and thus no references to the old table will
|
|
|
|
* remain.
|
|
|
|
*/
|
2015-03-14 05:57:23 +03:00
|
|
|
call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
|
2015-03-23 16:50:26 +03:00
|
|
|
|
|
|
|
return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_expand - Expand hash table while allowing concurrent lookups
|
|
|
|
* @ht: the hash table to expand
|
|
|
|
*
|
2015-03-11 01:43:48 +03:00
|
|
|
* A secondary bucket array is allocated and the hash entries are migrated.
|
2014-08-02 13:47:44 +04:00
|
|
|
*
|
|
|
|
* This function may only be called in a context where it is safe to call
|
|
|
|
* synchronize_rcu(), e.g. not within a rcu_read_lock() section.
|
|
|
|
*
|
2015-01-03 01:00:20 +03:00
|
|
|
* The caller must ensure that no concurrent resizing occurs by holding
|
|
|
|
* ht->mutex.
|
|
|
|
*
|
|
|
|
* It is valid to have concurrent insertions and deletions protected by per
|
|
|
|
* bucket locks or concurrent RCU protected lookups and traversals.
|
2014-08-02 13:47:44 +04:00
|
|
|
*/
|
2015-03-23 16:50:26 +03:00
|
|
|
static int rhashtable_expand(struct rhashtable *ht)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
|
|
|
struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
|
2015-03-23 16:50:26 +03:00
|
|
|
int err;
|
2014-08-02 13:47:44 +04:00
|
|
|
|
|
|
|
ASSERT_RHT_MUTEX(ht);
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
old_tbl = rhashtable_last_table(ht, old_tbl);
|
|
|
|
|
2015-03-23 16:50:27 +03:00
|
|
|
new_tbl = bucket_table_alloc(ht, old_tbl->size * 2, GFP_KERNEL);
|
2014-08-02 13:47:44 +04:00
|
|
|
if (new_tbl == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
|
|
|
|
if (err)
|
|
|
|
bucket_table_free(new_tbl);
|
|
|
|
|
|
|
|
return err;
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_shrink - Shrink hash table while allowing concurrent lookups
|
|
|
|
* @ht: the hash table to shrink
|
|
|
|
*
|
2015-03-23 16:50:25 +03:00
|
|
|
* This function shrinks the hash table to fit, i.e., the smallest
|
|
|
|
* size would not cause it to expand right away automatically.
|
2014-08-02 13:47:44 +04:00
|
|
|
*
|
2015-01-03 01:00:20 +03:00
|
|
|
* The caller must ensure that no concurrent resizing occurs by holding
|
|
|
|
* ht->mutex.
|
|
|
|
*
|
2014-08-02 13:47:44 +04:00
|
|
|
* The caller must ensure that no concurrent table mutations take place.
|
|
|
|
* It is however valid to have concurrent lookups if they are RCU protected.
|
2015-01-03 01:00:20 +03:00
|
|
|
*
|
|
|
|
* It is valid to have concurrent insertions and deletions protected by per
|
|
|
|
* bucket locks or concurrent RCU protected lookups and traversals.
|
2014-08-02 13:47:44 +04:00
|
|
|
*/
|
2015-03-23 16:50:26 +03:00
|
|
|
static int rhashtable_shrink(struct rhashtable *ht)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
2015-03-12 17:28:40 +03:00
|
|
|
struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int size;
|
2015-03-23 16:50:26 +03:00
|
|
|
int err;
|
2014-08-02 13:47:44 +04:00
|
|
|
|
|
|
|
ASSERT_RHT_MUTEX(ht);
|
|
|
|
|
2015-03-24 16:18:17 +03:00
|
|
|
size = roundup_pow_of_two(atomic_read(&ht->nelems) * 3 / 2);
|
2015-03-23 16:50:25 +03:00
|
|
|
if (size < ht->p.min_size)
|
|
|
|
size = ht->p.min_size;
|
|
|
|
|
|
|
|
if (old_tbl->size <= size)
|
|
|
|
return 0;
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
if (rht_dereference(old_tbl->future_tbl, ht))
|
|
|
|
return -EEXIST;
|
|
|
|
|
2015-03-23 16:50:27 +03:00
|
|
|
new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
|
2015-01-03 01:00:20 +03:00
|
|
|
if (new_tbl == NULL)
|
2014-08-02 13:47:44 +04:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
|
|
|
|
if (err)
|
|
|
|
bucket_table_free(new_tbl);
|
|
|
|
|
|
|
|
return err;
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
static void rht_deferred_worker(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct rhashtable *ht;
|
|
|
|
struct bucket_table *tbl;
|
2015-03-23 16:50:26 +03:00
|
|
|
int err = 0;
|
2015-01-03 01:00:20 +03:00
|
|
|
|
rhashtable: Fix race in rhashtable_destroy() and use regular work_struct
When we put our declared work task in the global workqueue with
schedule_delayed_work(), its delay parameter is always zero.
Therefore, we should define a regular work in rhashtable structure
instead of a delayed work.
By the way, we add a condition to check whether resizing functions
are NULL before cancelling the work, avoiding to cancel an
uninitialized work.
Lastly, while we wait for all work items we submitted before to run
to completion with cancel_delayed_work(), ht->mutex has been taken in
rhashtable_destroy(). Moreover, cancel_delayed_work() doesn't return
until all work items are accomplished, and when work items are
scheduled, the work's function - rht_deferred_worker() will be called.
However, as rht_deferred_worker() also needs to acquire the lock,
deadlock might happen at the moment as the lock is already held before.
So if the cancel work function is moved out of the lock covered scope,
this will avoid the deadlock.
Fixes: 97defe1 ("rhashtable: Per bucket locks & deferred expansion/shrinking")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Cc: Thomas Graf <tgraf@suug.ch>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 06:13:09 +03:00
|
|
|
ht = container_of(work, struct rhashtable, run_work);
|
2015-01-03 01:00:20 +03:00
|
|
|
mutex_lock(&ht->mutex);
|
2015-02-03 23:33:22 +03:00
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
tbl = rht_dereference(ht->tbl, ht);
|
2015-03-23 16:50:26 +03:00
|
|
|
tbl = rhashtable_last_table(ht, tbl);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-03-12 17:28:40 +03:00
|
|
|
if (rht_grow_above_75(ht, tbl))
|
2015-01-03 01:00:20 +03:00
|
|
|
rhashtable_expand(ht);
|
2015-03-24 23:42:19 +03:00
|
|
|
else if (ht->p.automatic_shrinking && rht_shrink_below_30(ht, tbl))
|
2015-01-03 01:00:20 +03:00
|
|
|
rhashtable_shrink(ht);
|
2015-03-23 16:50:26 +03:00
|
|
|
|
|
|
|
err = rhashtable_rehash_table(ht);
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
mutex_unlock(&ht->mutex);
|
2015-03-23 16:50:26 +03:00
|
|
|
|
|
|
|
if (err)
|
|
|
|
schedule_work(&ht->run_work);
|
2015-01-03 01:00:20 +03:00
|
|
|
}
|
|
|
|
|
2015-03-23 16:50:28 +03:00
|
|
|
static bool rhashtable_check_elasticity(struct rhashtable *ht,
|
|
|
|
struct bucket_table *tbl,
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int hash)
|
2015-03-23 16:50:28 +03:00
|
|
|
{
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int elasticity = ht->elasticity;
|
2015-03-23 16:50:28 +03:00
|
|
|
struct rhash_head *head;
|
|
|
|
|
|
|
|
rht_for_each(head, tbl, hash)
|
|
|
|
if (!--elasticity)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2015-12-03 15:41:29 +03:00
|
|
|
int rhashtable_insert_rehash(struct rhashtable *ht,
|
|
|
|
struct bucket_table *tbl)
|
2015-03-23 16:50:28 +03:00
|
|
|
{
|
|
|
|
struct bucket_table *old_tbl;
|
|
|
|
struct bucket_table *new_tbl;
|
|
|
|
unsigned int size;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
|
|
|
|
size = tbl->size;
|
|
|
|
|
2015-12-03 15:41:29 +03:00
|
|
|
err = -EBUSY;
|
|
|
|
|
2015-03-23 16:50:28 +03:00
|
|
|
if (rht_grow_above_75(ht, tbl))
|
|
|
|
size *= 2;
|
2015-04-22 10:41:46 +03:00
|
|
|
/* Do not schedule more than one rehash */
|
|
|
|
else if (old_tbl != tbl)
|
2015-12-03 15:41:29 +03:00
|
|
|
goto fail;
|
|
|
|
|
|
|
|
err = -ENOMEM;
|
2015-03-23 16:50:28 +03:00
|
|
|
|
|
|
|
new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC);
|
2015-12-03 15:41:29 +03:00
|
|
|
if (new_tbl == NULL)
|
|
|
|
goto fail;
|
2015-03-23 16:50:28 +03:00
|
|
|
|
|
|
|
err = rhashtable_rehash_attach(ht, tbl, new_tbl);
|
|
|
|
if (err) {
|
|
|
|
bucket_table_free(new_tbl);
|
|
|
|
if (err == -EEXIST)
|
|
|
|
err = 0;
|
|
|
|
} else
|
|
|
|
schedule_work(&ht->run_work);
|
|
|
|
|
|
|
|
return err;
|
2015-12-03 15:41:29 +03:00
|
|
|
|
|
|
|
fail:
|
|
|
|
/* Do not fail the insert if someone else did a rehash. */
|
|
|
|
if (likely(rcu_dereference_raw(tbl->future_tbl)))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Schedule async rehash to retry allocation in process context. */
|
|
|
|
if (err == -ENOMEM)
|
|
|
|
schedule_work(&ht->run_work);
|
|
|
|
|
|
|
|
return err;
|
2015-03-23 16:50:28 +03:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_insert_rehash);
|
|
|
|
|
2015-12-03 15:41:29 +03:00
|
|
|
struct bucket_table *rhashtable_insert_slow(struct rhashtable *ht,
|
|
|
|
const void *key,
|
|
|
|
struct rhash_head *obj,
|
|
|
|
struct bucket_table *tbl)
|
2015-03-20 13:57:00 +03:00
|
|
|
{
|
|
|
|
struct rhash_head *head;
|
2015-03-24 16:18:17 +03:00
|
|
|
unsigned int hash;
|
2015-03-23 16:50:28 +03:00
|
|
|
int err;
|
2015-03-20 13:57:00 +03:00
|
|
|
|
2015-03-23 16:50:26 +03:00
|
|
|
tbl = rhashtable_last_table(ht, tbl);
|
2015-03-20 13:57:00 +03:00
|
|
|
hash = head_hashfn(ht, tbl, obj);
|
|
|
|
spin_lock_nested(rht_bucket_lock(tbl, hash), SINGLE_DEPTH_NESTING);
|
|
|
|
|
2015-03-23 16:50:28 +03:00
|
|
|
err = -EEXIST;
|
2015-03-20 13:57:00 +03:00
|
|
|
if (key && rhashtable_lookup_fast(ht, key, ht->p))
|
|
|
|
goto exit;
|
|
|
|
|
2015-05-15 06:30:47 +03:00
|
|
|
err = -E2BIG;
|
|
|
|
if (unlikely(rht_grow_above_max(ht, tbl)))
|
|
|
|
goto exit;
|
|
|
|
|
2015-03-23 16:50:28 +03:00
|
|
|
err = -EAGAIN;
|
|
|
|
if (rhashtable_check_elasticity(ht, tbl, hash) ||
|
|
|
|
rht_grow_above_100(ht, tbl))
|
|
|
|
goto exit;
|
|
|
|
|
2015-03-20 13:57:00 +03:00
|
|
|
err = 0;
|
|
|
|
|
|
|
|
head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash);
|
|
|
|
|
|
|
|
RCU_INIT_POINTER(obj->next, head);
|
|
|
|
|
|
|
|
rcu_assign_pointer(tbl->buckets[hash], obj);
|
|
|
|
|
|
|
|
atomic_inc(&ht->nelems);
|
|
|
|
|
|
|
|
exit:
|
|
|
|
spin_unlock(rht_bucket_lock(tbl, hash));
|
|
|
|
|
2015-12-03 15:41:29 +03:00
|
|
|
if (err == 0)
|
|
|
|
return NULL;
|
|
|
|
else if (err == -EAGAIN)
|
|
|
|
return tbl;
|
|
|
|
else
|
|
|
|
return ERR_PTR(err);
|
2015-03-20 13:57:00 +03:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_insert_slow);
|
|
|
|
|
2015-02-03 23:33:23 +03:00
|
|
|
/**
|
|
|
|
* rhashtable_walk_init - Initialise an iterator
|
|
|
|
* @ht: Table to walk over
|
|
|
|
* @iter: Hash table Iterator
|
2016-03-02 18:09:19 +03:00
|
|
|
* @gfp: GFP flags for allocations
|
2015-02-03 23:33:23 +03:00
|
|
|
*
|
|
|
|
* This function prepares a hash table walk.
|
|
|
|
*
|
|
|
|
* Note that if you restart a walk after rhashtable_walk_stop you
|
|
|
|
* may see the same object twice. Also, you may miss objects if
|
|
|
|
* there are removals in between rhashtable_walk_stop and the next
|
|
|
|
* call to rhashtable_walk_start.
|
|
|
|
*
|
|
|
|
* For a completely stable walk you should construct your own data
|
|
|
|
* structure outside the hash table.
|
|
|
|
*
|
|
|
|
* This function may sleep so you must not call it from interrupt
|
|
|
|
* context or with spin locks held.
|
|
|
|
*
|
|
|
|
* You must call rhashtable_walk_exit if this function returns
|
|
|
|
* successfully.
|
|
|
|
*/
|
2016-03-02 18:09:19 +03:00
|
|
|
int rhashtable_walk_init(struct rhashtable *ht, struct rhashtable_iter *iter,
|
|
|
|
gfp_t gfp)
|
2015-02-03 23:33:23 +03:00
|
|
|
{
|
|
|
|
iter->ht = ht;
|
|
|
|
iter->p = NULL;
|
|
|
|
iter->slot = 0;
|
|
|
|
iter->skip = 0;
|
|
|
|
|
2016-03-02 18:09:19 +03:00
|
|
|
iter->walker = kmalloc(sizeof(*iter->walker), gfp);
|
2015-02-03 23:33:23 +03:00
|
|
|
if (!iter->walker)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_lock(&ht->lock);
|
2015-12-19 05:45:28 +03:00
|
|
|
iter->walker->tbl =
|
|
|
|
rcu_dereference_protected(ht->tbl, lockdep_is_held(&ht->lock));
|
2015-03-14 05:57:20 +03:00
|
|
|
list_add(&iter->walker->list, &iter->walker->tbl->walkers);
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_unlock(&ht->lock);
|
2015-02-03 23:33:23 +03:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_init);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_walk_exit - Free an iterator
|
|
|
|
* @iter: Hash table Iterator
|
|
|
|
*
|
|
|
|
* This function frees resources allocated by rhashtable_walk_init.
|
|
|
|
*/
|
|
|
|
void rhashtable_walk_exit(struct rhashtable_iter *iter)
|
|
|
|
{
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_lock(&iter->ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
if (iter->walker->tbl)
|
|
|
|
list_del(&iter->walker->list);
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_unlock(&iter->ht->lock);
|
2015-02-03 23:33:23 +03:00
|
|
|
kfree(iter->walker);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_walk_start - Start a hash table walk
|
|
|
|
* @iter: Hash table iterator
|
|
|
|
*
|
|
|
|
* Start a hash table walk. Note that we take the RCU lock in all
|
|
|
|
* cases including when we return an error. So you must always call
|
|
|
|
* rhashtable_walk_stop to clean up.
|
|
|
|
*
|
|
|
|
* Returns zero if successful.
|
|
|
|
*
|
|
|
|
* Returns -EAGAIN if resize event occured. Note that the iterator
|
|
|
|
* will rewind back to the beginning and you may use it immediately
|
|
|
|
* by calling rhashtable_walk_next.
|
|
|
|
*/
|
|
|
|
int rhashtable_walk_start(struct rhashtable_iter *iter)
|
2015-03-16 12:42:27 +03:00
|
|
|
__acquires(RCU)
|
2015-02-03 23:33:23 +03:00
|
|
|
{
|
2015-03-14 05:57:20 +03:00
|
|
|
struct rhashtable *ht = iter->ht;
|
|
|
|
|
2015-12-16 11:45:54 +03:00
|
|
|
rcu_read_lock();
|
2015-03-14 05:57:20 +03:00
|
|
|
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_lock(&ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
if (iter->walker->tbl)
|
|
|
|
list_del(&iter->walker->list);
|
2015-12-16 11:45:54 +03:00
|
|
|
spin_unlock(&ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
|
|
|
|
if (!iter->walker->tbl) {
|
|
|
|
iter->walker->tbl = rht_dereference_rcu(ht->tbl, ht);
|
2015-02-03 23:33:23 +03:00
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_start);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_walk_next - Return the next object and advance the iterator
|
|
|
|
* @iter: Hash table iterator
|
|
|
|
*
|
|
|
|
* Note that you must call rhashtable_walk_stop when you are finished
|
|
|
|
* with the walk.
|
|
|
|
*
|
|
|
|
* Returns the next object or NULL when the end of the table is reached.
|
|
|
|
*
|
|
|
|
* Returns -EAGAIN if resize event occured. Note that the iterator
|
|
|
|
* will rewind back to the beginning and you may continue to use it.
|
|
|
|
*/
|
|
|
|
void *rhashtable_walk_next(struct rhashtable_iter *iter)
|
|
|
|
{
|
2015-03-14 05:57:20 +03:00
|
|
|
struct bucket_table *tbl = iter->walker->tbl;
|
2015-02-03 23:33:23 +03:00
|
|
|
struct rhashtable *ht = iter->ht;
|
|
|
|
struct rhash_head *p = iter->p;
|
|
|
|
|
|
|
|
if (p) {
|
|
|
|
p = rht_dereference_bucket_rcu(p->next, tbl, iter->slot);
|
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (; iter->slot < tbl->size; iter->slot++) {
|
|
|
|
int skip = iter->skip;
|
|
|
|
|
|
|
|
rht_for_each_rcu(p, tbl, iter->slot) {
|
|
|
|
if (!skip)
|
|
|
|
break;
|
|
|
|
skip--;
|
|
|
|
}
|
|
|
|
|
|
|
|
next:
|
|
|
|
if (!rht_is_a_nulls(p)) {
|
|
|
|
iter->skip++;
|
|
|
|
iter->p = p;
|
2015-05-05 03:22:53 +03:00
|
|
|
return rht_obj(ht, p);
|
2015-02-03 23:33:23 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
iter->skip = 0;
|
|
|
|
}
|
|
|
|
|
2015-07-06 16:51:20 +03:00
|
|
|
iter->p = NULL;
|
|
|
|
|
2015-03-23 16:50:19 +03:00
|
|
|
/* Ensure we see any new tables. */
|
|
|
|
smp_rmb();
|
|
|
|
|
2015-03-14 05:57:25 +03:00
|
|
|
iter->walker->tbl = rht_dereference_rcu(tbl->future_tbl, ht);
|
|
|
|
if (iter->walker->tbl) {
|
2015-02-03 23:33:23 +03:00
|
|
|
iter->slot = 0;
|
|
|
|
iter->skip = 0;
|
|
|
|
return ERR_PTR(-EAGAIN);
|
|
|
|
}
|
|
|
|
|
2015-05-05 03:22:53 +03:00
|
|
|
return NULL;
|
2015-02-03 23:33:23 +03:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_next);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_walk_stop - Finish a hash table walk
|
|
|
|
* @iter: Hash table iterator
|
|
|
|
*
|
|
|
|
* Finish a hash table walk.
|
|
|
|
*/
|
|
|
|
void rhashtable_walk_stop(struct rhashtable_iter *iter)
|
2015-03-16 12:42:27 +03:00
|
|
|
__releases(RCU)
|
2015-02-03 23:33:23 +03:00
|
|
|
{
|
2015-03-14 05:57:20 +03:00
|
|
|
struct rhashtable *ht;
|
|
|
|
struct bucket_table *tbl = iter->walker->tbl;
|
|
|
|
|
|
|
|
if (!tbl)
|
2015-03-15 13:12:04 +03:00
|
|
|
goto out;
|
2015-03-14 05:57:20 +03:00
|
|
|
|
|
|
|
ht = iter->ht;
|
|
|
|
|
2015-03-24 01:53:17 +03:00
|
|
|
spin_lock(&ht->lock);
|
2015-03-14 05:57:25 +03:00
|
|
|
if (tbl->rehash < tbl->size)
|
2015-03-14 05:57:20 +03:00
|
|
|
list_add(&iter->walker->list, &tbl->walkers);
|
|
|
|
else
|
|
|
|
iter->walker->tbl = NULL;
|
2015-03-24 01:53:17 +03:00
|
|
|
spin_unlock(&ht->lock);
|
2015-03-14 05:57:20 +03:00
|
|
|
|
2015-02-03 23:33:23 +03:00
|
|
|
iter->p = NULL;
|
2015-03-15 13:12:04 +03:00
|
|
|
|
|
|
|
out:
|
|
|
|
rcu_read_unlock();
|
2015-02-03 23:33:23 +03:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
|
|
|
|
|
2015-03-20 13:56:59 +03:00
|
|
|
static size_t rounded_hashtable_size(const struct rhashtable_params *params)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
2014-09-03 05:22:36 +04:00
|
|
|
return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
|
2015-03-18 12:01:21 +03:00
|
|
|
(unsigned long)params->min_size);
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
|
|
|
|
2015-03-23 16:50:21 +03:00
|
|
|
static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
|
|
|
|
{
|
|
|
|
return jhash2(key, length, seed);
|
|
|
|
}
|
|
|
|
|
2014-08-02 13:47:44 +04:00
|
|
|
/**
|
|
|
|
* rhashtable_init - initialize a new hash table
|
|
|
|
* @ht: hash table to be initialized
|
|
|
|
* @params: configuration parameters
|
|
|
|
*
|
|
|
|
* Initializes a new hash table based on the provided configuration
|
|
|
|
* parameters. A table can be configured either with a variable or
|
|
|
|
* fixed length key:
|
|
|
|
*
|
|
|
|
* Configuration Example 1: Fixed length keys
|
|
|
|
* struct test_obj {
|
|
|
|
* int key;
|
|
|
|
* void * my_member;
|
|
|
|
* struct rhash_head node;
|
|
|
|
* };
|
|
|
|
*
|
|
|
|
* struct rhashtable_params params = {
|
|
|
|
* .head_offset = offsetof(struct test_obj, node),
|
|
|
|
* .key_offset = offsetof(struct test_obj, key),
|
|
|
|
* .key_len = sizeof(int),
|
2014-12-10 18:33:11 +03:00
|
|
|
* .hashfn = jhash,
|
2015-01-03 01:00:21 +03:00
|
|
|
* .nulls_base = (1U << RHT_BASE_SHIFT),
|
2014-08-02 13:47:44 +04:00
|
|
|
* };
|
|
|
|
*
|
|
|
|
* Configuration Example 2: Variable length keys
|
|
|
|
* struct test_obj {
|
|
|
|
* [...]
|
|
|
|
* struct rhash_head node;
|
|
|
|
* };
|
|
|
|
*
|
2015-03-25 16:07:45 +03:00
|
|
|
* u32 my_hash_fn(const void *data, u32 len, u32 seed)
|
2014-08-02 13:47:44 +04:00
|
|
|
* {
|
|
|
|
* struct test_obj *obj = data;
|
|
|
|
*
|
|
|
|
* return [... hash ...];
|
|
|
|
* }
|
|
|
|
*
|
|
|
|
* struct rhashtable_params params = {
|
|
|
|
* .head_offset = offsetof(struct test_obj, node),
|
2014-12-10 18:33:11 +03:00
|
|
|
* .hashfn = jhash,
|
2014-08-02 13:47:44 +04:00
|
|
|
* .obj_hashfn = my_hash_fn,
|
|
|
|
* };
|
|
|
|
*/
|
2015-03-20 13:56:59 +03:00
|
|
|
int rhashtable_init(struct rhashtable *ht,
|
|
|
|
const struct rhashtable_params *params)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
|
|
|
struct bucket_table *tbl;
|
|
|
|
size_t size;
|
|
|
|
|
|
|
|
size = HASH_DEFAULT_SIZE;
|
|
|
|
|
2015-03-23 16:50:21 +03:00
|
|
|
if ((!params->key_len && !params->obj_hashfn) ||
|
2015-03-20 13:57:00 +03:00
|
|
|
(params->obj_hashfn && !params->obj_cmpfn))
|
2014-08-02 13:47:44 +04:00
|
|
|
return -EINVAL;
|
|
|
|
|
2015-01-03 01:00:21 +03:00
|
|
|
if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
memset(ht, 0, sizeof(*ht));
|
|
|
|
mutex_init(&ht->mutex);
|
2015-03-24 01:53:17 +03:00
|
|
|
spin_lock_init(&ht->lock);
|
2015-01-03 01:00:20 +03:00
|
|
|
memcpy(&ht->p, params, sizeof(*params));
|
|
|
|
|
2015-03-20 01:31:13 +03:00
|
|
|
if (params->min_size)
|
|
|
|
ht->p.min_size = roundup_pow_of_two(params->min_size);
|
|
|
|
|
|
|
|
if (params->max_size)
|
|
|
|
ht->p.max_size = rounddown_pow_of_two(params->max_size);
|
|
|
|
|
2015-05-15 06:30:47 +03:00
|
|
|
if (params->insecure_max_entries)
|
|
|
|
ht->p.insecure_max_entries =
|
|
|
|
rounddown_pow_of_two(params->insecure_max_entries);
|
|
|
|
else
|
|
|
|
ht->p.insecure_max_entries = ht->p.max_size * 2;
|
|
|
|
|
2015-03-20 13:56:59 +03:00
|
|
|
ht->p.min_size = max(ht->p.min_size, HASH_MIN_SIZE);
|
2015-03-20 01:31:13 +03:00
|
|
|
|
2015-12-16 13:13:14 +03:00
|
|
|
if (params->nelem_hint)
|
|
|
|
size = rounded_hashtable_size(&ht->p);
|
|
|
|
|
2015-03-24 05:37:30 +03:00
|
|
|
/* The maximum (not average) chain length grows with the
|
|
|
|
* size of the hash table, at a rate of (log N)/(log log N).
|
|
|
|
* The value of 16 is selected so that even if the hash
|
|
|
|
* table grew to 2^32 you would not expect the maximum
|
|
|
|
* chain length to exceed it unless we are under attack
|
|
|
|
* (or extremely unlucky).
|
|
|
|
*
|
|
|
|
* As this limit is only to detect attacks, we don't need
|
|
|
|
* to set it to a lower value as you'd need the chain
|
|
|
|
* length to vastly exceed 16 to have any real effect
|
|
|
|
* on the system.
|
|
|
|
*/
|
2015-03-23 16:50:28 +03:00
|
|
|
if (!params->insecure_elasticity)
|
|
|
|
ht->elasticity = 16;
|
|
|
|
|
2015-01-03 01:00:20 +03:00
|
|
|
if (params->locks_mul)
|
|
|
|
ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
|
|
|
|
else
|
|
|
|
ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;
|
|
|
|
|
2015-03-23 16:50:21 +03:00
|
|
|
ht->key_len = ht->p.key_len;
|
|
|
|
if (!params->hashfn) {
|
|
|
|
ht->p.hashfn = jhash;
|
|
|
|
|
|
|
|
if (!(ht->key_len & (sizeof(u32) - 1))) {
|
|
|
|
ht->key_len /= sizeof(u32);
|
|
|
|
ht->p.hashfn = rhashtable_jhash2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-23 16:50:27 +03:00
|
|
|
tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
|
2014-08-02 13:47:44 +04:00
|
|
|
if (tbl == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-01-07 08:41:57 +03:00
|
|
|
atomic_set(&ht->nelems, 0);
|
2015-03-12 17:28:40 +03:00
|
|
|
|
2014-08-02 13:47:44 +04:00
|
|
|
RCU_INIT_POINTER(ht->tbl, tbl);
|
|
|
|
|
2015-02-25 18:31:54 +03:00
|
|
|
INIT_WORK(&ht->run_work, rht_deferred_worker);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2014-08-02 13:47:44 +04:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_init);
|
|
|
|
|
|
|
|
/**
|
2015-03-24 16:18:20 +03:00
|
|
|
* rhashtable_free_and_destroy - free elements and destroy hash table
|
2014-08-02 13:47:44 +04:00
|
|
|
* @ht: the hash table to destroy
|
2015-03-24 16:18:20 +03:00
|
|
|
* @free_fn: callback to release resources of element
|
|
|
|
* @arg: pointer passed to free_fn
|
2014-08-02 13:47:44 +04:00
|
|
|
*
|
2015-03-24 16:18:20 +03:00
|
|
|
* Stops an eventual async resize. If defined, invokes free_fn for each
|
|
|
|
* element to releasal resources. Please note that RCU protected
|
|
|
|
* readers may still be accessing the elements. Releasing of resources
|
|
|
|
* must occur in a compatible manner. Then frees the bucket array.
|
|
|
|
*
|
|
|
|
* This function will eventually sleep to wait for an async resize
|
|
|
|
* to complete. The caller is responsible that no further write operations
|
|
|
|
* occurs in parallel.
|
2014-08-02 13:47:44 +04:00
|
|
|
*/
|
2015-03-24 16:18:20 +03:00
|
|
|
void rhashtable_free_and_destroy(struct rhashtable *ht,
|
|
|
|
void (*free_fn)(void *ptr, void *arg),
|
|
|
|
void *arg)
|
2014-08-02 13:47:44 +04:00
|
|
|
{
|
2015-03-24 16:18:20 +03:00
|
|
|
const struct bucket_table *tbl;
|
|
|
|
unsigned int i;
|
2015-01-03 01:00:20 +03:00
|
|
|
|
2015-02-25 18:31:54 +03:00
|
|
|
cancel_work_sync(&ht->run_work);
|
2015-01-03 01:00:20 +03:00
|
|
|
|
rhashtable: Fix race in rhashtable_destroy() and use regular work_struct
When we put our declared work task in the global workqueue with
schedule_delayed_work(), its delay parameter is always zero.
Therefore, we should define a regular work in rhashtable structure
instead of a delayed work.
By the way, we add a condition to check whether resizing functions
are NULL before cancelling the work, avoiding to cancel an
uninitialized work.
Lastly, while we wait for all work items we submitted before to run
to completion with cancel_delayed_work(), ht->mutex has been taken in
rhashtable_destroy(). Moreover, cancel_delayed_work() doesn't return
until all work items are accomplished, and when work items are
scheduled, the work's function - rht_deferred_worker() will be called.
However, as rht_deferred_worker() also needs to acquire the lock,
deadlock might happen at the moment as the lock is already held before.
So if the cancel work function is moved out of the lock covered scope,
this will avoid the deadlock.
Fixes: 97defe1 ("rhashtable: Per bucket locks & deferred expansion/shrinking")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Cc: Thomas Graf <tgraf@suug.ch>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 06:13:09 +03:00
|
|
|
mutex_lock(&ht->mutex);
|
2015-03-24 16:18:20 +03:00
|
|
|
tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
if (free_fn) {
|
|
|
|
for (i = 0; i < tbl->size; i++) {
|
|
|
|
struct rhash_head *pos, *next;
|
|
|
|
|
|
|
|
for (pos = rht_dereference(tbl->buckets[i], ht),
|
|
|
|
next = !rht_is_a_nulls(pos) ?
|
|
|
|
rht_dereference(pos->next, ht) : NULL;
|
|
|
|
!rht_is_a_nulls(pos);
|
|
|
|
pos = next,
|
|
|
|
next = !rht_is_a_nulls(pos) ?
|
|
|
|
rht_dereference(pos->next, ht) : NULL)
|
|
|
|
free_fn(rht_obj(ht, pos), arg);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
bucket_table_free(tbl);
|
2015-01-03 01:00:20 +03:00
|
|
|
mutex_unlock(&ht->mutex);
|
2014-08-02 13:47:44 +04:00
|
|
|
}
|
2015-03-24 16:18:20 +03:00
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
|
|
|
|
|
|
|
|
void rhashtable_destroy(struct rhashtable *ht)
|
|
|
|
{
|
|
|
|
return rhashtable_free_and_destroy(ht, NULL, NULL);
|
|
|
|
}
|
2014-08-02 13:47:44 +04:00
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_destroy);
|