2014-08-02 17:47:44 +08:00
|
|
|
/*
|
|
|
|
* Resizable, Scalable, Concurrent Hash Table
|
|
|
|
*
|
2015-02-05 09:03:32 +08:00
|
|
|
* Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
|
2014-08-02 17:47:44 +08:00
|
|
|
* Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
|
|
|
|
*
|
|
|
|
* Based on the following paper:
|
|
|
|
* https://www.usenix.org/legacy/event/atc11/tech/final_files/Triplett.pdf
|
|
|
|
*
|
|
|
|
* Code partially derived from nft_hash
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/log2.h>
|
2015-02-26 23:20:34 +08:00
|
|
|
#include <linux/sched.h>
|
2014-08-02 17:47:44 +08:00
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/mm.h>
|
2014-12-10 23:33:11 +08:00
|
|
|
#include <linux/jhash.h>
|
2014-08-02 17:47:44 +08:00
|
|
|
#include <linux/random.h>
|
|
|
|
#include <linux/rhashtable.h>
|
2015-02-09 11:04:03 +08:00
|
|
|
#include <linux/err.h>
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
#define HASH_DEFAULT_SIZE 64UL
|
|
|
|
#define HASH_MIN_SIZE 4UL
|
2015-01-03 06:00:20 +08:00
|
|
|
#define BUCKET_LOCKS_PER_CPU 128UL
|
|
|
|
|
2015-01-03 06:00:21 +08:00
|
|
|
/* Base bits plus 1 bit for nulls marker */
|
|
|
|
#define HASH_RESERVED_SPACE (RHT_BASE_BITS + 1)
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
enum {
|
|
|
|
RHT_LOCK_NORMAL,
|
|
|
|
RHT_LOCK_NESTED,
|
|
|
|
};
|
|
|
|
|
|
|
|
/* The bucket lock is selected based on the hash and protects mutations
|
|
|
|
* on a group of hash buckets.
|
|
|
|
*
|
2015-02-05 09:03:32 +08:00
|
|
|
* A maximum of tbl->size/2 bucket locks is allocated. This ensures that
|
|
|
|
* a single lock always covers both buckets which may both contains
|
|
|
|
* entries which link to the same bucket of the old table during resizing.
|
|
|
|
* This allows to simplify the locking as locking the bucket in both
|
|
|
|
* tables during resize always guarantee protection.
|
|
|
|
*
|
2015-01-03 06:00:20 +08:00
|
|
|
* IMPORTANT: When holding the bucket lock of both the old and new table
|
|
|
|
* during expansions and shrinking, the old bucket lock must always be
|
|
|
|
* acquired first.
|
|
|
|
*/
|
|
|
|
static spinlock_t *bucket_lock(const struct bucket_table *tbl, u32 hash)
|
|
|
|
{
|
|
|
|
return &tbl->locks[hash & tbl->locks_mask];
|
|
|
|
}
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2014-08-13 22:38:30 +08:00
|
|
|
static void *rht_obj(const struct rhashtable *ht, const struct rhash_head *he)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
|
|
|
return (void *) he - ht->p.head_offset;
|
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:14 +08:00
|
|
|
static u32 rht_bucket_index(const struct bucket_table *tbl, u32 hash)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-03 06:00:14 +08:00
|
|
|
return hash & (tbl->size - 1);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:14 +08:00
|
|
|
static u32 obj_raw_hashfn(const struct rhashtable *ht, const void *ptr)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-03 06:00:14 +08:00
|
|
|
u32 hash;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:14 +08:00
|
|
|
if (unlikely(!ht->p.key_len))
|
|
|
|
hash = ht->p.obj_hashfn(ptr, ht->p.hash_rnd);
|
|
|
|
else
|
|
|
|
hash = ht->p.hashfn(ptr + ht->p.key_offset, ht->p.key_len,
|
|
|
|
ht->p.hash_rnd);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:21 +08:00
|
|
|
return hash >> HASH_RESERVED_SPACE;
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
static u32 key_hashfn(struct rhashtable *ht, const void *key, u32 len)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-02-05 09:03:31 +08:00
|
|
|
return ht->p.hashfn(key, len, ht->p.hash_rnd) >> HASH_RESERVED_SPACE;
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static u32 head_hashfn(const struct rhashtable *ht,
|
2015-01-03 06:00:14 +08:00
|
|
|
const struct bucket_table *tbl,
|
|
|
|
const struct rhash_head *he)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-03 06:00:14 +08:00
|
|
|
return rht_bucket_index(tbl, obj_raw_hashfn(ht, rht_obj(ht, he)));
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-02-05 09:03:34 +08:00
|
|
|
#ifdef CONFIG_PROVE_LOCKING
|
|
|
|
static void debug_dump_buckets(const struct rhashtable *ht,
|
|
|
|
const struct bucket_table *tbl)
|
|
|
|
{
|
|
|
|
struct rhash_head *he;
|
|
|
|
unsigned int i, hash;
|
|
|
|
|
|
|
|
for (i = 0; i < tbl->size; i++) {
|
|
|
|
pr_warn(" [Bucket %d] ", i);
|
|
|
|
rht_for_each_rcu(he, tbl, i) {
|
|
|
|
hash = head_hashfn(ht, tbl, he);
|
|
|
|
pr_cont("[hash = %#x, lock = %p] ",
|
|
|
|
hash, bucket_lock(tbl, hash));
|
|
|
|
}
|
|
|
|
pr_cont("\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static void debug_dump_table(struct rhashtable *ht,
|
|
|
|
const struct bucket_table *tbl,
|
|
|
|
unsigned int hash)
|
|
|
|
{
|
|
|
|
struct bucket_table *old_tbl, *future_tbl;
|
|
|
|
|
|
|
|
pr_emerg("BUG: lock for hash %#x in table %p not held\n",
|
|
|
|
hash, tbl);
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
future_tbl = rht_dereference_rcu(ht->future_tbl, ht);
|
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
if (future_tbl != old_tbl) {
|
|
|
|
pr_warn("Future table %p (size: %zd)\n",
|
|
|
|
future_tbl, future_tbl->size);
|
|
|
|
debug_dump_buckets(ht, future_tbl);
|
|
|
|
}
|
|
|
|
|
|
|
|
pr_warn("Table %p (size: %zd)\n", old_tbl, old_tbl->size);
|
|
|
|
debug_dump_buckets(ht, old_tbl);
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
|
|
|
|
#define ASSERT_BUCKET_LOCK(HT, TBL, HASH) \
|
|
|
|
do { \
|
|
|
|
if (unlikely(!lockdep_rht_bucket_is_held(TBL, HASH))) { \
|
|
|
|
debug_dump_table(HT, TBL, HASH); \
|
|
|
|
BUG(); \
|
|
|
|
} \
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
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)
|
|
|
|
{
|
|
|
|
spinlock_t *lock = bucket_lock(tbl, hash);
|
|
|
|
|
|
|
|
return (debug_locks) ? lockdep_is_held(lock) : 1;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
|
|
|
|
#else
|
|
|
|
#define ASSERT_RHT_MUTEX(HT)
|
|
|
|
#define ASSERT_BUCKET_LOCK(HT, TBL, HASH)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
2015-01-03 06:00:17 +08:00
|
|
|
static struct rhash_head __rcu **bucket_tail(struct bucket_table *tbl, u32 n)
|
|
|
|
{
|
|
|
|
struct rhash_head __rcu **pprev;
|
|
|
|
|
|
|
|
for (pprev = &tbl->buckets[n];
|
2015-01-03 06:00:21 +08:00
|
|
|
!rht_is_a_nulls(rht_dereference_bucket(*pprev, tbl, n));
|
2015-01-03 06:00:17 +08:00
|
|
|
pprev = &rht_dereference_bucket(*pprev, tbl, n)->next)
|
|
|
|
;
|
|
|
|
|
|
|
|
return pprev;
|
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl)
|
|
|
|
{
|
|
|
|
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 09:03:32 +08:00
|
|
|
/* Never allocate more than 0.5 locks per bucket */
|
|
|
|
size = min_t(unsigned int, size, tbl->size >> 1);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
|
|
|
if (sizeof(spinlock_t) != 0) {
|
|
|
|
#ifdef CONFIG_NUMA
|
|
|
|
if (size * sizeof(spinlock_t) > PAGE_SIZE)
|
|
|
|
tbl->locks = vmalloc(size * sizeof(spinlock_t));
|
|
|
|
else
|
|
|
|
#endif
|
|
|
|
tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
|
|
|
|
GFP_KERNEL);
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
|
|
|
|
size_t nbuckets)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-02-20 07:53:38 +08:00
|
|
|
struct bucket_table *tbl = NULL;
|
2014-08-02 17:47:44 +08:00
|
|
|
size_t size;
|
2015-01-03 06:00:21 +08:00
|
|
|
int i;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]);
|
2015-02-20 07:53:38 +08:00
|
|
|
if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
|
|
|
|
tbl = kzalloc(size, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
|
2014-08-02 17:47:44 +08:00
|
|
|
if (tbl == NULL)
|
|
|
|
tbl = vzalloc(size);
|
|
|
|
if (tbl == NULL)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
tbl->size = nbuckets;
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
if (alloc_bucket_locks(ht, tbl) < 0) {
|
|
|
|
bucket_table_free(tbl);
|
|
|
|
return NULL;
|
|
|
|
}
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:21 +08:00
|
|
|
for (i = 0; i < nbuckets; i++)
|
|
|
|
INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
return tbl;
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rht_grow_above_75 - returns true if nelems > 0.75 * table-size
|
|
|
|
* @ht: hash table
|
|
|
|
* @new_size: new table size
|
|
|
|
*/
|
2015-02-25 23:31:54 +08:00
|
|
|
static bool rht_grow_above_75(const struct rhashtable *ht, size_t new_size)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
|
|
|
/* Expand table when exceeding 75% load */
|
2015-01-07 13:41:56 +08:00
|
|
|
return atomic_read(&ht->nelems) > (new_size / 4 * 3) &&
|
rhashtable: unconditionally grow when max_shift is not specified
While commit c0c09bfdc415 ("rhashtable: avoid unnecessary wakeup for
worker queue") rightfully moved part of the decision making of
whether we should expand or shrink from the expand/shrink functions
themselves into insert/delete functions in order to avoid unnecessary
worker wake-ups, it however introduced a regression by doing so.
Before that change, if no max_shift was specified (= 0) on rhashtable
initialization, rhashtable_expand() would just grow unconditionally
and lets the available memory be the limiting factor. After that
change, if no max_shift was specified, there would be _no_ expansion
step at all.
Given that netlink and tipc have a max_shift specified, it was not
visible there, but Josh Hunt reported that if nft that starts out
with a default element hint of 3 if not otherwise provided, would
slow i.e. inserts down trememdously as it cannot grow larger to
relax table occupancy.
Given that the test case verifies shrinks/expands manually, we also
must remove pointer to the helper functions to explicitly avoid
parallel resizing on insertions/deletions. test_bucket_stats() and
test_rht_lookup() could also be wrapped around rhashtable mutex to
explicitly synchronize a walk from resizing, but I think that defeats
the actual test case which intended to have explicit test steps,
i.e. 1) inserts, 2) expands, 3) shrinks, 4) deletions, with object
verification after each stage.
Reported-by: Josh Hunt <johunt@akamai.com>
Fixes: c0c09bfdc415 ("rhashtable: avoid unnecessary wakeup for worker queue")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Ying Xue <ying.xue@windriver.com>
Cc: Josh Hunt <johunt@akamai.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-02-25 23:31:53 +08:00
|
|
|
(!ht->p.max_shift || atomic_read(&ht->shift) < ht->p.max_shift);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rht_shrink_below_30 - returns true if nelems < 0.3 * table-size
|
|
|
|
* @ht: hash table
|
|
|
|
* @new_size: new table size
|
|
|
|
*/
|
2015-02-25 23:31:54 +08:00
|
|
|
static bool rht_shrink_below_30(const struct rhashtable *ht, size_t new_size)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
|
|
|
/* Shrink table beneath 30% load */
|
2015-01-07 13:41:56 +08:00
|
|
|
return atomic_read(&ht->nelems) < (new_size * 3 / 10) &&
|
|
|
|
(atomic_read(&ht->shift) > ht->p.min_shift);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
static void lock_buckets(struct bucket_table *new_tbl,
|
|
|
|
struct bucket_table *old_tbl, unsigned int hash)
|
|
|
|
__acquires(old_bucket_lock)
|
|
|
|
{
|
|
|
|
spin_lock_bh(bucket_lock(old_tbl, hash));
|
|
|
|
if (new_tbl != old_tbl)
|
|
|
|
spin_lock_bh_nested(bucket_lock(new_tbl, hash),
|
|
|
|
RHT_LOCK_NESTED);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unlock_buckets(struct bucket_table *new_tbl,
|
|
|
|
struct bucket_table *old_tbl, unsigned int hash)
|
|
|
|
__releases(old_bucket_lock)
|
|
|
|
{
|
|
|
|
if (new_tbl != old_tbl)
|
|
|
|
spin_unlock_bh(bucket_lock(new_tbl, hash));
|
|
|
|
spin_unlock_bh(bucket_lock(old_tbl, hash));
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Unlink entries on bucket which hash to different bucket.
|
|
|
|
*
|
|
|
|
* Returns true if no more work needs to be performed on the bucket.
|
|
|
|
*/
|
2015-02-05 09:03:34 +08:00
|
|
|
static bool hashtable_chain_unzip(struct rhashtable *ht,
|
2014-08-02 17:47:44 +08:00
|
|
|
const struct bucket_table *new_tbl,
|
2015-01-03 06:00:20 +08:00
|
|
|
struct bucket_table *old_tbl,
|
|
|
|
size_t old_hash)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
|
|
|
struct rhash_head *he, *p, *next;
|
2015-01-03 06:00:20 +08:00
|
|
|
unsigned int new_hash, new_hash2;
|
|
|
|
|
2015-02-05 09:03:34 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, old_tbl, old_hash);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
/* Old bucket empty, no work needed. */
|
2015-01-03 06:00:20 +08:00
|
|
|
p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl,
|
|
|
|
old_hash);
|
2015-01-03 06:00:21 +08:00
|
|
|
if (rht_is_a_nulls(p))
|
2015-02-05 09:03:32 +08:00
|
|
|
return false;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
new_hash = head_hashfn(ht, new_tbl, p);
|
2015-02-05 09:03:34 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
/* Advance the old bucket pointer one or more times until it
|
|
|
|
* reaches a node that doesn't hash to the same bucket as the
|
|
|
|
* previous node p. Call the previous node p;
|
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
rht_for_each_continue(he, p->next, old_tbl, old_hash) {
|
|
|
|
new_hash2 = head_hashfn(ht, new_tbl, he);
|
2015-02-05 09:03:34 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash2);
|
2015-02-05 09:03:32 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
if (new_hash != new_hash2)
|
2014-08-02 17:47:44 +08:00
|
|
|
break;
|
|
|
|
p = he;
|
|
|
|
}
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(old_tbl->buckets[old_hash], p->next);
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
/* Find the subsequent node which does hash to the same
|
|
|
|
* bucket as node P, or NULL if no such node exists.
|
|
|
|
*/
|
2015-01-03 06:00:21 +08:00
|
|
|
INIT_RHT_NULLS_HEAD(next, ht, old_hash);
|
|
|
|
if (!rht_is_a_nulls(he)) {
|
2015-01-03 06:00:20 +08:00
|
|
|
rht_for_each_continue(he, he->next, old_tbl, old_hash) {
|
|
|
|
if (head_hashfn(ht, new_tbl, he) == new_hash) {
|
2014-08-02 17:47:44 +08:00
|
|
|
next = he;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set p's next pointer to that subsequent node pointer,
|
|
|
|
* bypassing the nodes which do not hash to p's bucket
|
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(p->next, next);
|
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl,
|
|
|
|
old_hash);
|
|
|
|
|
|
|
|
return !rht_is_a_nulls(p);
|
2015-01-03 06:00:20 +08:00
|
|
|
}
|
|
|
|
|
2015-02-05 09:03:35 +08:00
|
|
|
static void link_old_to_new(struct rhashtable *ht, struct bucket_table *new_tbl,
|
2015-01-03 06:00:20 +08:00
|
|
|
unsigned int new_hash, struct rhash_head *entry)
|
|
|
|
{
|
2015-02-05 09:03:35 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash);
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(*bucket_tail(new_tbl, new_hash), entry);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_expand - Expand hash table while allowing concurrent lookups
|
|
|
|
* @ht: the hash table to expand
|
|
|
|
*
|
|
|
|
* A secondary bucket array is allocated and the hash entries are migrated
|
|
|
|
* while keeping them on both lists until the end of the RCU grace period.
|
|
|
|
*
|
|
|
|
* 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 06:00:20 +08:00
|
|
|
* 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 17:47:44 +08:00
|
|
|
*/
|
2014-11-13 20:45:46 +08:00
|
|
|
int rhashtable_expand(struct rhashtable *ht)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
|
|
|
struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
struct rhash_head *he;
|
2015-01-03 06:00:20 +08:00
|
|
|
unsigned int new_hash, old_hash;
|
|
|
|
bool complete = false;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
ASSERT_RHT_MUTEX(ht);
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
new_tbl = bucket_table_alloc(ht, old_tbl->size * 2);
|
2014-08-02 17:47:44 +08:00
|
|
|
if (new_tbl == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-01-07 13:41:56 +08:00
|
|
|
atomic_inc(&ht->shift);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
/* Make insertions go into the new, empty table right away. Deletions
|
|
|
|
* and lookups will be attempted in both tables until we synchronize.
|
|
|
|
* The synchronize_rcu() guarantees for the new table to be picked up
|
|
|
|
* so no new additions go into the old table while we relink.
|
|
|
|
*/
|
|
|
|
rcu_assign_pointer(ht->future_tbl, new_tbl);
|
|
|
|
synchronize_rcu();
|
|
|
|
|
|
|
|
/* For each new bucket, search the corresponding old bucket for the
|
|
|
|
* first entry that hashes to the new bucket, and link the end of
|
|
|
|
* newly formed bucket chain (containing entries added to future
|
|
|
|
* table) to that entry. Since all the entries which will end up in
|
|
|
|
* the new bucket appear in the same old bucket, this constructs an
|
|
|
|
* entirely valid new hash table, but with multiple buckets
|
|
|
|
* "zipped" together into a single imprecise chain.
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
for (new_hash = 0; new_hash < new_tbl->size; new_hash++) {
|
|
|
|
old_hash = rht_bucket_index(old_tbl, new_hash);
|
2015-02-05 09:03:32 +08:00
|
|
|
lock_buckets(new_tbl, old_tbl, new_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
rht_for_each(he, old_tbl, old_hash) {
|
|
|
|
if (head_hashfn(ht, new_tbl, he) == new_hash) {
|
2015-02-05 09:03:35 +08:00
|
|
|
link_old_to_new(ht, new_tbl, new_hash, he);
|
2014-08-02 17:47:44 +08:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(new_tbl, old_tbl, new_hash);
|
2015-02-26 23:20:34 +08:00
|
|
|
cond_resched();
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Unzip interleaved hash chains */
|
2015-01-03 06:00:20 +08:00
|
|
|
while (!complete && !ht->being_destroyed) {
|
2014-08-02 17:47:44 +08:00
|
|
|
/* Wait for readers. All new readers will see the new
|
|
|
|
* table, and thus no references to the old table will
|
|
|
|
* remain.
|
|
|
|
*/
|
|
|
|
synchronize_rcu();
|
|
|
|
|
|
|
|
/* For each bucket in the old table (each of which
|
|
|
|
* contains items from multiple buckets of the new
|
|
|
|
* table): ...
|
|
|
|
*/
|
|
|
|
complete = true;
|
2015-01-03 06:00:20 +08:00
|
|
|
for (old_hash = 0; old_hash < old_tbl->size; old_hash++) {
|
2015-02-05 09:03:32 +08:00
|
|
|
lock_buckets(new_tbl, old_tbl, old_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
if (hashtable_chain_unzip(ht, new_tbl, old_tbl,
|
|
|
|
old_hash))
|
2014-08-02 17:47:44 +08:00
|
|
|
complete = false;
|
2015-01-03 06:00:20 +08:00
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(new_tbl, old_tbl, old_hash);
|
2015-02-26 23:20:34 +08:00
|
|
|
cond_resched();
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
2015-01-03 06:00:20 +08:00
|
|
|
}
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-02-05 09:03:36 +08:00
|
|
|
rcu_assign_pointer(ht->tbl, new_tbl);
|
2015-02-05 09:03:33 +08:00
|
|
|
synchronize_rcu();
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
bucket_table_free(old_tbl);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_expand);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_shrink - Shrink hash table while allowing concurrent lookups
|
|
|
|
* @ht: the hash table to shrink
|
|
|
|
*
|
|
|
|
* 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 06:00:20 +08:00
|
|
|
* The caller must ensure that no concurrent resizing occurs by holding
|
|
|
|
* ht->mutex.
|
|
|
|
*
|
2014-08-02 17:47:44 +08:00
|
|
|
* 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 06:00:20 +08:00
|
|
|
*
|
|
|
|
* It is valid to have concurrent insertions and deletions protected by per
|
|
|
|
* bucket locks or concurrent RCU protected lookups and traversals.
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2014-11-13 20:45:46 +08:00
|
|
|
int rhashtable_shrink(struct rhashtable *ht)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-03 06:00:20 +08:00
|
|
|
struct bucket_table *new_tbl, *tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
unsigned int new_hash;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
ASSERT_RHT_MUTEX(ht);
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
new_tbl = bucket_table_alloc(ht, tbl->size / 2);
|
|
|
|
if (new_tbl == NULL)
|
2014-08-02 17:47:44 +08:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(ht->future_tbl, new_tbl);
|
|
|
|
synchronize_rcu();
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
/* Link the first entry in the old bucket to the end of the
|
|
|
|
* bucket in the new table. As entries are concurrently being
|
|
|
|
* added to the new table, lock down the new bucket. As we
|
|
|
|
* always divide the size in half when shrinking, each bucket
|
|
|
|
* in the new table maps to exactly two buckets in the old
|
|
|
|
* table.
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
for (new_hash = 0; new_hash < new_tbl->size; new_hash++) {
|
2015-02-05 09:03:32 +08:00
|
|
|
lock_buckets(new_tbl, tbl, new_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
|
|
|
rcu_assign_pointer(*bucket_tail(new_tbl, new_hash),
|
|
|
|
tbl->buckets[new_hash]);
|
2015-02-05 09:03:35 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, tbl, new_hash + new_tbl->size);
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(*bucket_tail(new_tbl, new_hash),
|
|
|
|
tbl->buckets[new_hash + new_tbl->size]);
|
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(new_tbl, tbl, new_hash);
|
2015-02-26 23:20:34 +08:00
|
|
|
cond_resched();
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Publish the new, valid hash table */
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_assign_pointer(ht->tbl, new_tbl);
|
2015-01-07 13:41:56 +08:00
|
|
|
atomic_dec(&ht->shift);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
/* Wait for readers. No new readers will have references to the
|
|
|
|
* old hash table.
|
|
|
|
*/
|
|
|
|
synchronize_rcu();
|
|
|
|
|
|
|
|
bucket_table_free(tbl);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_shrink);
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
static void rht_deferred_worker(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct rhashtable *ht;
|
|
|
|
struct bucket_table *tbl;
|
2015-02-04 04:33:23 +08:00
|
|
|
struct rhashtable_walker *walker;
|
2015-01-03 06:00:20 +08:00
|
|
|
|
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 11:13:09 +08:00
|
|
|
ht = container_of(work, struct rhashtable, run_work);
|
2015-01-03 06:00:20 +08:00
|
|
|
mutex_lock(&ht->mutex);
|
2015-02-04 04:33:22 +08:00
|
|
|
if (ht->being_destroyed)
|
|
|
|
goto unlock;
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
tbl = rht_dereference(ht->tbl, ht);
|
|
|
|
|
2015-02-04 04:33:23 +08:00
|
|
|
list_for_each_entry(walker, &ht->walkers, list)
|
|
|
|
walker->resize = true;
|
|
|
|
|
2015-02-25 23:31:54 +08:00
|
|
|
if (rht_grow_above_75(ht, tbl->size))
|
2015-01-03 06:00:20 +08:00
|
|
|
rhashtable_expand(ht);
|
2015-02-25 23:31:54 +08:00
|
|
|
else if (rht_shrink_below_30(ht, tbl->size))
|
2015-01-03 06:00:20 +08:00
|
|
|
rhashtable_shrink(ht);
|
2015-02-04 04:33:22 +08:00
|
|
|
unlock:
|
2015-01-03 06:00:20 +08:00
|
|
|
mutex_unlock(&ht->mutex);
|
|
|
|
}
|
|
|
|
|
2015-01-07 13:41:54 +08:00
|
|
|
static void __rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj,
|
2015-02-25 23:31:54 +08:00
|
|
|
struct bucket_table *tbl,
|
|
|
|
const struct bucket_table *old_tbl, u32 hash)
|
2015-01-07 13:41:54 +08:00
|
|
|
{
|
2015-02-25 23:31:54 +08:00
|
|
|
bool no_resize_running = tbl == old_tbl;
|
2015-02-07 00:08:43 +08:00
|
|
|
struct rhash_head *head;
|
|
|
|
|
|
|
|
hash = rht_bucket_index(tbl, hash);
|
|
|
|
head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash);
|
2015-01-07 13:41:54 +08:00
|
|
|
|
2015-02-05 09:03:35 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, tbl, hash);
|
|
|
|
|
2015-01-07 13:41:54 +08:00
|
|
|
if (rht_is_a_nulls(head))
|
|
|
|
INIT_RHT_NULLS_HEAD(obj->next, ht, hash);
|
|
|
|
else
|
|
|
|
RCU_INIT_POINTER(obj->next, head);
|
|
|
|
|
|
|
|
rcu_assign_pointer(tbl->buckets[hash], obj);
|
|
|
|
|
|
|
|
atomic_inc(&ht->nelems);
|
2015-02-25 23:31:54 +08:00
|
|
|
if (no_resize_running && rht_grow_above_75(ht, tbl->size))
|
|
|
|
schedule_work(&ht->run_work);
|
2015-01-07 13:41:54 +08:00
|
|
|
}
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
/**
|
2015-01-07 13:41:54 +08:00
|
|
|
* rhashtable_insert - insert object into hash table
|
2014-08-02 17:47:44 +08:00
|
|
|
* @ht: hash table
|
|
|
|
* @obj: pointer to hash head inside object
|
|
|
|
*
|
2015-01-03 06:00:20 +08:00
|
|
|
* Will take a per bucket spinlock to protect against mutual mutations
|
|
|
|
* on the same bucket. Multiple insertions may occur in parallel unless
|
|
|
|
* they map to the same bucket lock.
|
2014-08-02 17:47:44 +08:00
|
|
|
*
|
2015-01-03 06:00:20 +08:00
|
|
|
* It is safe to call this function from atomic context.
|
|
|
|
*
|
|
|
|
* Will trigger an automatic deferred table resizing if the size grows
|
|
|
|
* beyond the watermark indicated by grow_decision() which can be passed
|
|
|
|
* to rhashtable_init().
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2014-11-13 20:45:46 +08:00
|
|
|
void rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-02-05 09:03:32 +08:00
|
|
|
struct bucket_table *tbl, *old_tbl;
|
2015-01-03 06:00:20 +08:00
|
|
|
unsigned hash;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_lock();
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
tbl = rht_dereference_rcu(ht->future_tbl, ht);
|
2015-02-05 09:03:32 +08:00
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
2015-02-07 00:08:43 +08:00
|
|
|
hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
|
2015-01-03 06:00:20 +08:00
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
lock_buckets(tbl, old_tbl, hash);
|
2015-02-25 23:31:54 +08:00
|
|
|
__rhashtable_insert(ht, obj, tbl, old_tbl, hash);
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(tbl, old_tbl, hash);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_unlock();
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_insert);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_remove - remove object from hash table
|
|
|
|
* @ht: hash table
|
|
|
|
* @obj: pointer to hash head inside object
|
|
|
|
*
|
|
|
|
* Since the hash chain is single linked, the removal operation needs to
|
|
|
|
* walk the bucket chain upon removal. The removal operation is thus
|
|
|
|
* considerable slow if the hash table is not correctly sized.
|
|
|
|
*
|
2015-01-07 13:41:54 +08:00
|
|
|
* Will automatically shrink the table via rhashtable_expand() if the
|
2014-08-02 17:47:44 +08:00
|
|
|
* shrink_decision function specified at rhashtable_init() returns true.
|
|
|
|
*
|
|
|
|
* The caller must ensure that no concurrent table mutations occur. It is
|
|
|
|
* however valid to have concurrent lookups if they are RCU protected.
|
|
|
|
*/
|
2014-11-13 20:45:46 +08:00
|
|
|
bool rhashtable_remove(struct rhashtable *ht, struct rhash_head *obj)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-02-05 09:03:32 +08:00
|
|
|
struct bucket_table *tbl, *new_tbl, *old_tbl;
|
2014-08-02 17:47:44 +08:00
|
|
|
struct rhash_head __rcu **pprev;
|
2015-02-05 09:03:36 +08:00
|
|
|
struct rhash_head *he, *he2;
|
2015-02-05 09:03:32 +08:00
|
|
|
unsigned int hash, new_hash;
|
2015-01-21 19:54:01 +08:00
|
|
|
bool ret = false;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_lock();
|
2015-02-07 00:08:43 +08:00
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
tbl = new_tbl = rht_dereference_rcu(ht->future_tbl, ht);
|
2015-02-05 09:03:36 +08:00
|
|
|
new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
lock_buckets(new_tbl, old_tbl, new_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
restart:
|
2015-02-05 09:03:32 +08:00
|
|
|
hash = rht_bucket_index(tbl, new_hash);
|
2015-01-03 06:00:20 +08:00
|
|
|
pprev = &tbl->buckets[hash];
|
|
|
|
rht_for_each(he, tbl, hash) {
|
2014-08-02 17:47:44 +08:00
|
|
|
if (he != obj) {
|
|
|
|
pprev = &he->next;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2015-02-05 09:03:35 +08:00
|
|
|
ASSERT_BUCKET_LOCK(ht, tbl, hash);
|
2015-01-03 06:00:18 +08:00
|
|
|
|
2015-02-07 00:08:43 +08:00
|
|
|
if (old_tbl->size > new_tbl->size && tbl == old_tbl &&
|
|
|
|
!rht_is_a_nulls(obj->next) &&
|
|
|
|
head_hashfn(ht, tbl, obj->next) != hash) {
|
|
|
|
rcu_assign_pointer(*pprev, (struct rhash_head *) rht_marker(ht, hash));
|
|
|
|
} else if (unlikely(old_tbl->size < new_tbl->size && tbl == new_tbl)) {
|
|
|
|
rht_for_each_continue(he2, obj->next, tbl, hash) {
|
2015-02-05 09:03:36 +08:00
|
|
|
if (head_hashfn(ht, tbl, he2) == hash) {
|
|
|
|
rcu_assign_pointer(*pprev, he2);
|
|
|
|
goto found;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-02-07 00:08:43 +08:00
|
|
|
rcu_assign_pointer(*pprev, (struct rhash_head *) rht_marker(ht, hash));
|
2015-02-05 09:03:36 +08:00
|
|
|
} else {
|
|
|
|
rcu_assign_pointer(*pprev, obj->next);
|
|
|
|
}
|
|
|
|
|
|
|
|
found:
|
2015-01-21 19:54:01 +08:00
|
|
|
ret = true;
|
|
|
|
break;
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-01-21 19:54:01 +08:00
|
|
|
/* The entry may be linked in either 'tbl', 'future_tbl', or both.
|
|
|
|
* 'future_tbl' only exists for a short period of time during
|
|
|
|
* resizing. Thus traversing both is fine and the added cost is
|
|
|
|
* very rare.
|
|
|
|
*/
|
2015-02-07 00:08:43 +08:00
|
|
|
if (tbl != old_tbl) {
|
|
|
|
tbl = old_tbl;
|
2015-01-03 06:00:20 +08:00
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(new_tbl, old_tbl, new_hash);
|
2015-01-21 19:54:01 +08:00
|
|
|
|
|
|
|
if (ret) {
|
2015-02-25 23:31:54 +08:00
|
|
|
bool no_resize_running = new_tbl == old_tbl;
|
|
|
|
|
2015-01-21 19:54:01 +08:00
|
|
|
atomic_dec(&ht->nelems);
|
2015-02-25 23:31:54 +08:00
|
|
|
if (no_resize_running && rht_shrink_below_30(ht, new_tbl->size))
|
|
|
|
schedule_work(&ht->run_work);
|
2015-01-21 19:54:01 +08:00
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
2015-01-21 19:54:01 +08:00
|
|
|
return ret;
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_remove);
|
|
|
|
|
2015-01-07 13:41:52 +08:00
|
|
|
struct rhashtable_compare_arg {
|
|
|
|
struct rhashtable *ht;
|
|
|
|
const void *key;
|
|
|
|
};
|
|
|
|
|
|
|
|
static bool rhashtable_compare(void *ptr, void *arg)
|
|
|
|
{
|
|
|
|
struct rhashtable_compare_arg *x = arg;
|
|
|
|
struct rhashtable *ht = x->ht;
|
|
|
|
|
|
|
|
return !memcmp(ptr + ht->p.key_offset, x->key, ht->p.key_len);
|
|
|
|
}
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
/**
|
|
|
|
* rhashtable_lookup - lookup key in hash table
|
|
|
|
* @ht: hash table
|
|
|
|
* @key: pointer to key
|
|
|
|
*
|
|
|
|
* Computes the hash value for the key and traverses the bucket chain looking
|
|
|
|
* for a entry with an identical key. The first matching entry is returned.
|
|
|
|
*
|
|
|
|
* This lookup function may only be used for fixed key hash table (key_len
|
2015-01-07 13:41:54 +08:00
|
|
|
* parameter set). It will BUG() if used inappropriately.
|
2014-08-02 17:47:44 +08:00
|
|
|
*
|
2015-01-03 06:00:20 +08:00
|
|
|
* Lookups may occur in parallel with hashtable mutations and resizing.
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
void *rhashtable_lookup(struct rhashtable *ht, const void *key)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-07 13:41:52 +08:00
|
|
|
struct rhashtable_compare_arg arg = {
|
|
|
|
.ht = ht,
|
|
|
|
.key = key,
|
|
|
|
};
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
BUG_ON(!ht->p.key_len);
|
|
|
|
|
2015-01-07 13:41:52 +08:00
|
|
|
return rhashtable_lookup_compare(ht, key, &rhashtable_compare, &arg);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_lookup);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_lookup_compare - search hash table with compare function
|
|
|
|
* @ht: hash table
|
2015-01-03 06:00:14 +08:00
|
|
|
* @key: the pointer to the key
|
2014-08-02 17:47:44 +08:00
|
|
|
* @compare: compare function, must return true on match
|
|
|
|
* @arg: argument passed on to compare function
|
|
|
|
*
|
|
|
|
* Traverses the bucket chain behind the provided hash value and calls the
|
|
|
|
* specified compare function for each entry.
|
|
|
|
*
|
2015-01-03 06:00:20 +08:00
|
|
|
* Lookups may occur in parallel with hashtable mutations and resizing.
|
2014-08-02 17:47:44 +08:00
|
|
|
*
|
|
|
|
* Returns the first entry on which the compare function returned true.
|
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
void *rhashtable_lookup_compare(struct rhashtable *ht, const void *key,
|
2014-08-02 17:47:44 +08:00
|
|
|
bool (*compare)(void *, void *), void *arg)
|
|
|
|
{
|
2015-01-03 06:00:20 +08:00
|
|
|
const struct bucket_table *tbl, *old_tbl;
|
2014-08-02 17:47:44 +08:00
|
|
|
struct rhash_head *he;
|
2015-01-03 06:00:14 +08:00
|
|
|
u32 hash;
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
tbl = rht_dereference_rcu(ht->future_tbl, ht);
|
2015-01-03 06:00:14 +08:00
|
|
|
hash = key_hashfn(ht, key, ht->p.key_len);
|
2015-01-03 06:00:20 +08:00
|
|
|
restart:
|
|
|
|
rht_for_each_rcu(he, tbl, rht_bucket_index(tbl, hash)) {
|
2014-08-02 17:47:44 +08:00
|
|
|
if (!compare(rht_obj(ht, he), arg))
|
|
|
|
continue;
|
2015-01-03 06:00:20 +08:00
|
|
|
rcu_read_unlock();
|
2015-01-03 06:00:15 +08:00
|
|
|
return rht_obj(ht, he);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
if (unlikely(tbl != old_tbl)) {
|
|
|
|
tbl = old_tbl;
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_lookup_compare);
|
|
|
|
|
2015-01-07 13:41:54 +08:00
|
|
|
/**
|
|
|
|
* rhashtable_lookup_insert - lookup and insert object into hash table
|
|
|
|
* @ht: hash table
|
|
|
|
* @obj: pointer to hash head inside object
|
|
|
|
*
|
|
|
|
* Locks down the bucket chain in both the old and new table if a resize
|
|
|
|
* is in progress to ensure that writers can't remove from the old table
|
|
|
|
* and can't insert to the new table during the atomic operation of search
|
|
|
|
* and insertion. Searches for duplicates in both the old and new table if
|
|
|
|
* a resize is in progress.
|
|
|
|
*
|
|
|
|
* This lookup function may only be used for fixed key hash table (key_len
|
|
|
|
* parameter set). It will BUG() if used inappropriately.
|
|
|
|
*
|
|
|
|
* It is safe to call this function from atomic context.
|
|
|
|
*
|
|
|
|
* Will trigger an automatic deferred table resizing if the size grows
|
|
|
|
* beyond the watermark indicated by grow_decision() which can be passed
|
|
|
|
* to rhashtable_init().
|
|
|
|
*/
|
|
|
|
bool rhashtable_lookup_insert(struct rhashtable *ht, struct rhash_head *obj)
|
2015-01-12 14:52:22 +08:00
|
|
|
{
|
|
|
|
struct rhashtable_compare_arg arg = {
|
|
|
|
.ht = ht,
|
|
|
|
.key = rht_obj(ht, obj) + ht->p.key_offset,
|
|
|
|
};
|
|
|
|
|
|
|
|
BUG_ON(!ht->p.key_len);
|
|
|
|
|
|
|
|
return rhashtable_lookup_compare_insert(ht, obj, &rhashtable_compare,
|
|
|
|
&arg);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_lookup_insert);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_lookup_compare_insert - search and insert object to hash table
|
|
|
|
* with compare function
|
|
|
|
* @ht: hash table
|
|
|
|
* @obj: pointer to hash head inside object
|
|
|
|
* @compare: compare function, must return true on match
|
|
|
|
* @arg: argument passed on to compare function
|
|
|
|
*
|
|
|
|
* Locks down the bucket chain in both the old and new table if a resize
|
|
|
|
* is in progress to ensure that writers can't remove from the old table
|
|
|
|
* and can't insert to the new table during the atomic operation of search
|
|
|
|
* and insertion. Searches for duplicates in both the old and new table if
|
|
|
|
* a resize is in progress.
|
|
|
|
*
|
|
|
|
* Lookups may occur in parallel with hashtable mutations and resizing.
|
|
|
|
*
|
|
|
|
* Will trigger an automatic deferred table resizing if the size grows
|
|
|
|
* beyond the watermark indicated by grow_decision() which can be passed
|
|
|
|
* to rhashtable_init().
|
|
|
|
*/
|
|
|
|
bool rhashtable_lookup_compare_insert(struct rhashtable *ht,
|
|
|
|
struct rhash_head *obj,
|
|
|
|
bool (*compare)(void *, void *),
|
|
|
|
void *arg)
|
2015-01-07 13:41:54 +08:00
|
|
|
{
|
|
|
|
struct bucket_table *new_tbl, *old_tbl;
|
2015-02-05 09:03:32 +08:00
|
|
|
u32 new_hash;
|
2015-01-07 13:41:54 +08:00
|
|
|
bool success = true;
|
|
|
|
|
|
|
|
BUG_ON(!ht->p.key_len);
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
old_tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
new_tbl = rht_dereference_rcu(ht->future_tbl, ht);
|
2015-02-07 00:08:43 +08:00
|
|
|
new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj));
|
2015-02-05 09:03:32 +08:00
|
|
|
|
|
|
|
lock_buckets(new_tbl, old_tbl, new_hash);
|
2015-01-07 13:41:54 +08:00
|
|
|
|
2015-01-12 14:52:22 +08:00
|
|
|
if (rhashtable_lookup_compare(ht, rht_obj(ht, obj) + ht->p.key_offset,
|
|
|
|
compare, arg)) {
|
2015-01-07 13:41:54 +08:00
|
|
|
success = false;
|
|
|
|
goto exit;
|
|
|
|
}
|
|
|
|
|
2015-02-25 23:31:54 +08:00
|
|
|
__rhashtable_insert(ht, obj, new_tbl, old_tbl, new_hash);
|
2015-01-07 13:41:54 +08:00
|
|
|
|
|
|
|
exit:
|
2015-02-05 09:03:32 +08:00
|
|
|
unlock_buckets(new_tbl, old_tbl, new_hash);
|
2015-01-07 13:41:54 +08:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return success;
|
|
|
|
}
|
2015-01-12 14:52:22 +08:00
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_lookup_compare_insert);
|
2015-01-07 13:41:54 +08:00
|
|
|
|
2015-02-04 04:33:23 +08:00
|
|
|
/**
|
|
|
|
* rhashtable_walk_init - Initialise an iterator
|
|
|
|
* @ht: Table to walk over
|
|
|
|
* @iter: Hash table Iterator
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
int rhashtable_walk_init(struct rhashtable *ht, struct rhashtable_iter *iter)
|
|
|
|
{
|
|
|
|
iter->ht = ht;
|
|
|
|
iter->p = NULL;
|
|
|
|
iter->slot = 0;
|
|
|
|
iter->skip = 0;
|
|
|
|
|
|
|
|
iter->walker = kmalloc(sizeof(*iter->walker), GFP_KERNEL);
|
|
|
|
if (!iter->walker)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-02-23 17:35:06 +08:00
|
|
|
INIT_LIST_HEAD(&iter->walker->list);
|
|
|
|
iter->walker->resize = false;
|
|
|
|
|
2015-02-04 04:33:23 +08:00
|
|
|
mutex_lock(&ht->mutex);
|
|
|
|
list_add(&iter->walker->list, &ht->walkers);
|
|
|
|
mutex_unlock(&ht->mutex);
|
|
|
|
|
|
|
|
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)
|
|
|
|
{
|
|
|
|
mutex_lock(&iter->ht->mutex);
|
|
|
|
list_del(&iter->walker->list);
|
|
|
|
mutex_unlock(&iter->ht->mutex);
|
|
|
|
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)
|
|
|
|
{
|
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
if (iter->walker->resize) {
|
|
|
|
iter->slot = 0;
|
|
|
|
iter->skip = 0;
|
|
|
|
iter->walker->resize = false;
|
|
|
|
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)
|
|
|
|
{
|
|
|
|
const struct bucket_table *tbl;
|
|
|
|
struct rhashtable *ht = iter->ht;
|
|
|
|
struct rhash_head *p = iter->p;
|
|
|
|
void *obj = NULL;
|
|
|
|
|
|
|
|
tbl = rht_dereference_rcu(ht->tbl, ht);
|
|
|
|
|
|
|
|
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;
|
|
|
|
obj = rht_obj(ht, p);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
iter->skip = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
iter->p = NULL;
|
|
|
|
|
|
|
|
out:
|
|
|
|
if (iter->walker->resize) {
|
|
|
|
iter->p = NULL;
|
|
|
|
iter->slot = 0;
|
|
|
|
iter->skip = 0;
|
|
|
|
iter->walker->resize = false;
|
|
|
|
return ERR_PTR(-EAGAIN);
|
|
|
|
}
|
|
|
|
|
|
|
|
return obj;
|
|
|
|
}
|
|
|
|
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)
|
|
|
|
{
|
|
|
|
rcu_read_unlock();
|
|
|
|
iter->p = NULL;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
|
|
|
|
|
2014-09-03 09:22:36 +08:00
|
|
|
static size_t rounded_hashtable_size(struct rhashtable_params *params)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2014-09-03 09:22:36 +08:00
|
|
|
return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
|
|
|
|
1UL << params->min_shift);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* 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 23:33:11 +08:00
|
|
|
* .hashfn = jhash,
|
2015-01-03 06:00:21 +08:00
|
|
|
* .nulls_base = (1U << RHT_BASE_SHIFT),
|
2014-08-02 17:47:44 +08:00
|
|
|
* };
|
|
|
|
*
|
|
|
|
* Configuration Example 2: Variable length keys
|
|
|
|
* struct test_obj {
|
|
|
|
* [...]
|
|
|
|
* struct rhash_head node;
|
|
|
|
* };
|
|
|
|
*
|
|
|
|
* u32 my_hash_fn(const void *data, u32 seed)
|
|
|
|
* {
|
|
|
|
* struct test_obj *obj = data;
|
|
|
|
*
|
|
|
|
* return [... hash ...];
|
|
|
|
* }
|
|
|
|
*
|
|
|
|
* struct rhashtable_params params = {
|
|
|
|
* .head_offset = offsetof(struct test_obj, node),
|
2014-12-10 23:33:11 +08:00
|
|
|
* .hashfn = jhash,
|
2014-08-02 17:47:44 +08:00
|
|
|
* .obj_hashfn = my_hash_fn,
|
|
|
|
* };
|
|
|
|
*/
|
|
|
|
int rhashtable_init(struct rhashtable *ht, struct rhashtable_params *params)
|
|
|
|
{
|
|
|
|
struct bucket_table *tbl;
|
|
|
|
size_t size;
|
|
|
|
|
|
|
|
size = HASH_DEFAULT_SIZE;
|
|
|
|
|
|
|
|
if ((params->key_len && !params->hashfn) ||
|
|
|
|
(!params->key_len && !params->obj_hashfn))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-01-03 06:00:21 +08:00
|
|
|
if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2014-09-03 09:22:36 +08:00
|
|
|
params->min_shift = max_t(size_t, params->min_shift,
|
|
|
|
ilog2(HASH_MIN_SIZE));
|
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
if (params->nelem_hint)
|
2014-09-03 09:22:36 +08:00
|
|
|
size = rounded_hashtable_size(params);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
2015-01-03 06:00:20 +08:00
|
|
|
memset(ht, 0, sizeof(*ht));
|
|
|
|
mutex_init(&ht->mutex);
|
|
|
|
memcpy(&ht->p, params, sizeof(*params));
|
2015-02-04 04:33:23 +08:00
|
|
|
INIT_LIST_HEAD(&ht->walkers);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
|
|
|
if (params->locks_mul)
|
|
|
|
ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
|
|
|
|
else
|
|
|
|
ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;
|
|
|
|
|
|
|
|
tbl = bucket_table_alloc(ht, size);
|
2014-08-02 17:47:44 +08:00
|
|
|
if (tbl == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2015-01-07 13:41:57 +08:00
|
|
|
atomic_set(&ht->nelems, 0);
|
2015-01-07 13:41:56 +08:00
|
|
|
atomic_set(&ht->shift, ilog2(tbl->size));
|
2014-08-02 17:47:44 +08:00
|
|
|
RCU_INIT_POINTER(ht->tbl, tbl);
|
2015-01-03 06:00:20 +08:00
|
|
|
RCU_INIT_POINTER(ht->future_tbl, tbl);
|
2014-08-02 17:47:44 +08:00
|
|
|
|
|
|
|
if (!ht->p.hash_rnd)
|
|
|
|
get_random_bytes(&ht->p.hash_rnd, sizeof(ht->p.hash_rnd));
|
|
|
|
|
2015-02-25 23:31:54 +08:00
|
|
|
INIT_WORK(&ht->run_work, rht_deferred_worker);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
2014-08-02 17:47:44 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_init);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rhashtable_destroy - destroy hash table
|
|
|
|
* @ht: the hash table to destroy
|
|
|
|
*
|
2014-09-02 06:26:05 +08:00
|
|
|
* Frees the bucket array. This function is not rcu safe, therefore the caller
|
|
|
|
* has to make sure that no resizing may happen by unpublishing the hashtable
|
|
|
|
* and waiting for the quiescent cycle before releasing the bucket array.
|
2014-08-02 17:47:44 +08:00
|
|
|
*/
|
2015-01-03 06:00:20 +08:00
|
|
|
void rhashtable_destroy(struct rhashtable *ht)
|
2014-08-02 17:47:44 +08:00
|
|
|
{
|
2015-01-03 06:00:20 +08:00
|
|
|
ht->being_destroyed = true;
|
|
|
|
|
2015-02-25 23:31:54 +08:00
|
|
|
cancel_work_sync(&ht->run_work);
|
2015-01-03 06:00:20 +08:00
|
|
|
|
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 11:13:09 +08:00
|
|
|
mutex_lock(&ht->mutex);
|
2015-01-03 06:00:20 +08:00
|
|
|
bucket_table_free(rht_dereference(ht->tbl, ht));
|
|
|
|
mutex_unlock(&ht->mutex);
|
2014-08-02 17:47:44 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(rhashtable_destroy);
|