diff options
Diffstat (limited to 'lib/rhashtable.c')
-rw-r--r-- | lib/rhashtable.c | 1170 |
1 files changed, 751 insertions, 419 deletions
diff --git a/lib/rhashtable.c b/lib/rhashtable.c index 6c3c723e902b..9cc4c4a90d00 100644 --- a/lib/rhashtable.c +++ b/lib/rhashtable.c @@ -1,7 +1,7 @@ /* * Resizable, Scalable, Concurrent Hash Table * - * Copyright (c) 2014 Thomas Graf <tgraf@suug.ch> + * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch> * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net> * * Based on the following paper: @@ -23,94 +23,203 @@ #include <linux/jhash.h> #include <linux/random.h> #include <linux/rhashtable.h> +#include <linux/err.h> #define HASH_DEFAULT_SIZE 64UL #define HASH_MIN_SIZE 4UL +#define BUCKET_LOCKS_PER_CPU 128UL -#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT)) +/* Base bits plus 1 bit for nulls marker */ +#define HASH_RESERVED_SPACE (RHT_BASE_BITS + 1) -#ifdef CONFIG_PROVE_LOCKING -int lockdep_rht_mutex_is_held(const struct rhashtable *ht) +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. + * + * 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. + * + * 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 ht->p.mutex_is_held(ht->p.parent); + return &tbl->locks[hash & tbl->locks_mask]; } -EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held); -#endif static void *rht_obj(const struct rhashtable *ht, const struct rhash_head *he) { return (void *) he - ht->p.head_offset; } -static u32 __hashfn(const struct rhashtable *ht, const void *key, - u32 len, u32 hsize) +static u32 rht_bucket_index(const struct bucket_table *tbl, u32 hash) +{ + return hash & (tbl->size - 1); +} + +static u32 obj_raw_hashfn(const struct rhashtable *ht, const void *ptr) { - u32 h; + u32 hash; - h = ht->p.hashfn(key, len, ht->p.hash_rnd); + 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); - return h & (hsize - 1); + return hash >> HASH_RESERVED_SPACE; } -/** - * rhashtable_hashfn - compute hash for key of given length - * @ht: hash table to compute for - * @key: pointer to key - * @len: length of key - * - * Computes the hash value using the hash function provided in the 'hashfn' - * of struct rhashtable_params. The returned value is guaranteed to be - * smaller than the number of buckets in the hash table. - */ -u32 rhashtable_hashfn(const struct rhashtable *ht, const void *key, u32 len) +static u32 key_hashfn(struct rhashtable *ht, const void *key, u32 len) { - struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); + return ht->p.hashfn(key, len, ht->p.hash_rnd) >> HASH_RESERVED_SPACE; +} - return __hashfn(ht, key, len, tbl->size); +static u32 head_hashfn(const struct rhashtable *ht, + const struct bucket_table *tbl, + const struct rhash_head *he) +{ + return rht_bucket_index(tbl, obj_raw_hashfn(ht, rht_obj(ht, he))); } -EXPORT_SYMBOL_GPL(rhashtable_hashfn); -static u32 obj_hashfn(const struct rhashtable *ht, const void *ptr, u32 hsize) +#ifdef CONFIG_PROVE_LOCKING +static void debug_dump_buckets(const struct rhashtable *ht, + const struct bucket_table *tbl) { - if (unlikely(!ht->p.key_len)) { - u32 h; + struct rhash_head *he; + unsigned int i, hash; - h = ht->p.obj_hashfn(ptr, ht->p.hash_rnd); + 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); - return h & (hsize - 1); + 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); } - return __hashfn(ht, ptr + ht->p.key_offset, ht->p.key_len, hsize); + pr_warn("Table %p (size: %zd)\n", old_tbl, old_tbl->size); + debug_dump_buckets(ht, old_tbl); + + rcu_read_unlock(); } -/** - * rhashtable_obj_hashfn - compute hash for hashed object - * @ht: hash table to compute for - * @ptr: pointer to hashed object - * - * Computes the hash value using the hash function `hashfn` respectively - * 'obj_hashfn' depending on whether the hash table is set up to work with - * a fixed length key. The returned value is guaranteed to be smaller than - * the number of buckets in the hash table. - */ -u32 rhashtable_obj_hashfn(const struct rhashtable *ht, void *ptr) +#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) { - struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); + return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1; +} +EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held); - return obj_hashfn(ht, ptr, tbl->size); +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(rhashtable_obj_hashfn); +EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held); +#else +#define ASSERT_RHT_MUTEX(HT) +#define ASSERT_BUCKET_LOCK(HT, TBL, HASH) +#endif -static u32 head_hashfn(const struct rhashtable *ht, - const struct rhash_head *he, u32 hsize) + +static struct rhash_head __rcu **bucket_tail(struct bucket_table *tbl, u32 n) { - return obj_hashfn(ht, rht_obj(ht, he), hsize); + struct rhash_head __rcu **pprev; + + for (pprev = &tbl->buckets[n]; + !rht_is_a_nulls(rht_dereference_bucket(*pprev, tbl, n)); + pprev = &rht_dereference_bucket(*pprev, tbl, n)->next) + ; + + return pprev; } -static struct bucket_table *bucket_table_alloc(size_t nbuckets) +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); + + /* Never allocate more than 0.5 locks per bucket */ + size = min_t(unsigned int, size, tbl->size >> 1); + + 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) { struct bucket_table *tbl; size_t size; + int i; size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]); tbl = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); @@ -122,12 +231,15 @@ static struct bucket_table *bucket_table_alloc(size_t nbuckets) tbl->size = nbuckets; - return tbl; -} + if (alloc_bucket_locks(ht, tbl) < 0) { + bucket_table_free(tbl); + return NULL; + } -static void bucket_table_free(const struct bucket_table *tbl) -{ - kvfree(tbl); + for (i = 0; i < nbuckets; i++) + INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i); + + return tbl; } /** @@ -138,7 +250,8 @@ static void bucket_table_free(const struct bucket_table *tbl) bool rht_grow_above_75(const struct rhashtable *ht, size_t new_size) { /* Expand table when exceeding 75% load */ - return ht->nelems > (new_size / 4 * 3); + return atomic_read(&ht->nelems) > (new_size / 4 * 3) && + (ht->p.max_shift && atomic_read(&ht->shift) < ht->p.max_shift); } EXPORT_SYMBOL_GPL(rht_grow_above_75); @@ -150,41 +263,75 @@ EXPORT_SYMBOL_GPL(rht_grow_above_75); bool rht_shrink_below_30(const struct rhashtable *ht, size_t new_size) { /* Shrink table beneath 30% load */ - return ht->nelems < (new_size * 3 / 10); + return atomic_read(&ht->nelems) < (new_size * 3 / 10) && + (atomic_read(&ht->shift) > ht->p.min_shift); } EXPORT_SYMBOL_GPL(rht_shrink_below_30); -static void hashtable_chain_unzip(const struct rhashtable *ht, +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. + */ +static bool hashtable_chain_unzip(struct rhashtable *ht, const struct bucket_table *new_tbl, - struct bucket_table *old_tbl, size_t n) + struct bucket_table *old_tbl, + size_t old_hash) { struct rhash_head *he, *p, *next; - unsigned int h; + unsigned int new_hash, new_hash2; + + ASSERT_BUCKET_LOCK(ht, old_tbl, old_hash); /* Old bucket empty, no work needed. */ - p = rht_dereference(old_tbl->buckets[n], ht); - if (!p) - return; + p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl, + old_hash); + if (rht_is_a_nulls(p)) + return false; + + new_hash = head_hashfn(ht, new_tbl, p); + ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash); /* 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; */ - h = head_hashfn(ht, p, new_tbl->size); - rht_for_each(he, p->next, ht) { - if (head_hashfn(ht, he, new_tbl->size) != h) + rht_for_each_continue(he, p->next, old_tbl, old_hash) { + new_hash2 = head_hashfn(ht, new_tbl, he); + ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash2); + + if (new_hash != new_hash2) break; p = he; } - RCU_INIT_POINTER(old_tbl->buckets[n], p->next); + rcu_assign_pointer(old_tbl->buckets[old_hash], p->next); /* Find the subsequent node which does hash to the same * bucket as node P, or NULL if no such node exists. */ - next = NULL; - if (he) { - rht_for_each(he, he->next, ht) { - if (head_hashfn(ht, he, new_tbl->size) == h) { + INIT_RHT_NULLS_HEAD(next, ht, old_hash); + if (!rht_is_a_nulls(he)) { + rht_for_each_continue(he, he->next, old_tbl, old_hash) { + if (head_hashfn(ht, new_tbl, he) == new_hash) { next = he; break; } @@ -194,7 +341,20 @@ static void hashtable_chain_unzip(const struct rhashtable *ht, /* Set p's next pointer to that subsequent node pointer, * bypassing the nodes which do not hash to p's bucket */ - RCU_INIT_POINTER(p->next, next); + rcu_assign_pointer(p->next, next); + + p = rht_dereference_bucket(old_tbl->buckets[old_hash], old_tbl, + old_hash); + + return !rht_is_a_nulls(p); +} + +static void link_old_to_new(struct rhashtable *ht, struct bucket_table *new_tbl, + unsigned int new_hash, struct rhash_head *entry) +{ + ASSERT_BUCKET_LOCK(ht, new_tbl, new_hash); + + rcu_assign_pointer(*bucket_tail(new_tbl, new_hash), entry); } /** @@ -207,53 +367,57 @@ static void hashtable_chain_unzip(const struct rhashtable *ht, * 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. * - * The caller must ensure that no concurrent table mutations take place. - * It is however valid to have concurrent lookups if they are RCU protected. + * 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. */ int rhashtable_expand(struct rhashtable *ht) { struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht); struct rhash_head *he; - unsigned int i, h; - bool complete; + unsigned int new_hash, old_hash; + bool complete = false; ASSERT_RHT_MUTEX(ht); - if (ht->p.max_shift && ht->shift >= ht->p.max_shift) - return 0; - - new_tbl = bucket_table_alloc(old_tbl->size * 2); + new_tbl = bucket_table_alloc(ht, old_tbl->size * 2); if (new_tbl == NULL) return -ENOMEM; - ht->shift++; + atomic_inc(&ht->shift); + + /* 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 new bucket 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. + /* 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. */ - for (i = 0; i < new_tbl->size; i++) { - h = i & (old_tbl->size - 1); - rht_for_each(he, old_tbl->buckets[h], ht) { - if (head_hashfn(ht, he, new_tbl->size) == i) { - RCU_INIT_POINTER(new_tbl->buckets[i], he); + for (new_hash = 0; new_hash < new_tbl->size; new_hash++) { + old_hash = rht_bucket_index(old_tbl, new_hash); + lock_buckets(new_tbl, old_tbl, new_hash); + rht_for_each(he, old_tbl, old_hash) { + if (head_hashfn(ht, new_tbl, he) == new_hash) { + link_old_to_new(ht, new_tbl, new_hash, he); break; } } + unlock_buckets(new_tbl, old_tbl, new_hash); } - /* Publish the new table pointer. Lookups may now traverse - * the new table, but they will not benefit from any - * additional efficiency until later steps unzip the buckets. - */ - rcu_assign_pointer(ht->tbl, new_tbl); - /* Unzip interleaved hash chains */ - do { + while (!complete && !ht->being_destroyed) { /* Wait for readers. All new readers will see the new * table, and thus no references to the old table will * remain. @@ -265,12 +429,19 @@ int rhashtable_expand(struct rhashtable *ht) * table): ... */ complete = true; - for (i = 0; i < old_tbl->size; i++) { - hashtable_chain_unzip(ht, new_tbl, old_tbl, i); - if (old_tbl->buckets[i] != NULL) + for (old_hash = 0; old_hash < old_tbl->size; old_hash++) { + lock_buckets(new_tbl, old_tbl, old_hash); + + if (hashtable_chain_unzip(ht, new_tbl, old_tbl, + old_hash)) complete = false; + + unlock_buckets(new_tbl, old_tbl, old_hash); } - } while (!complete); + } + + rcu_assign_pointer(ht->tbl, new_tbl); + synchronize_rcu(); bucket_table_free(old_tbl); return 0; @@ -284,45 +455,51 @@ EXPORT_SYMBOL_GPL(rhashtable_expand); * 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. * + * The caller must ensure that no concurrent resizing occurs by holding + * ht->mutex. + * * The caller must ensure that no concurrent table mutations take place. * It is however valid to have concurrent lookups if they are RCU protected. + * + * It is valid to have concurrent insertions and deletions protected by per + * bucket locks or concurrent RCU protected lookups and traversals. */ int rhashtable_shrink(struct rhashtable *ht) { - struct bucket_table *ntbl, *tbl = rht_dereference(ht->tbl, ht); - struct rhash_head __rcu **pprev; - unsigned int i; + struct bucket_table *new_tbl, *tbl = rht_dereference(ht->tbl, ht); + unsigned int new_hash; ASSERT_RHT_MUTEX(ht); - if (ht->shift <= ht->p.min_shift) - return 0; - - ntbl = bucket_table_alloc(tbl->size / 2); - if (ntbl == NULL) + new_tbl = bucket_table_alloc(ht, tbl->size / 2); + if (new_tbl == NULL) return -ENOMEM; - ht->shift--; + rcu_assign_pointer(ht->future_tbl, new_tbl); + synchronize_rcu(); - /* Link each bucket in the new table to the first bucket - * in the old table that contains entries which will hash - * to the new bucket. + /* 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. */ - for (i = 0; i < ntbl->size; i++) { - ntbl->buckets[i] = tbl->buckets[i]; + for (new_hash = 0; new_hash < new_tbl->size; new_hash++) { + lock_buckets(new_tbl, tbl, new_hash); - /* Link each bucket in the new table to the first bucket - * in the old table that contains entries which will hash - * to the new bucket. - */ - for (pprev = &ntbl->buckets[i]; *pprev != NULL; - pprev = &rht_dereference(*pprev, ht)->next) - ; - RCU_INIT_POINTER(*pprev, tbl->buckets[i + ntbl->size]); + rcu_assign_pointer(*bucket_tail(new_tbl, new_hash), + tbl->buckets[new_hash]); + ASSERT_BUCKET_LOCK(ht, tbl, new_hash + new_tbl->size); + rcu_assign_pointer(*bucket_tail(new_tbl, new_hash), + tbl->buckets[new_hash + new_tbl->size]); + + unlock_buckets(new_tbl, tbl, new_hash); } /* Publish the new, valid hash table */ - rcu_assign_pointer(ht->tbl, ntbl); + rcu_assign_pointer(ht->tbl, new_tbl); + atomic_dec(&ht->shift); /* Wait for readers. No new readers will have references to the * old hash table. @@ -335,59 +512,99 @@ int rhashtable_shrink(struct rhashtable *ht) } EXPORT_SYMBOL_GPL(rhashtable_shrink); -/** - * rhashtable_insert - insert object into hash hash table - * @ht: hash table - * @obj: pointer to hash head inside object - * - * Will automatically grow the table via rhashtable_expand() if the the - * grow_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. - */ -void rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj) +static void rht_deferred_worker(struct work_struct *work) { - struct bucket_table *tbl = rht_dereference(ht->tbl, ht); - u32 hash; + struct rhashtable *ht; + struct bucket_table *tbl; + struct rhashtable_walker *walker; - ASSERT_RHT_MUTEX(ht); + ht = container_of(work, struct rhashtable, run_work); + mutex_lock(&ht->mutex); + if (ht->being_destroyed) + goto unlock; - hash = head_hashfn(ht, obj, tbl->size); - RCU_INIT_POINTER(obj->next, tbl->buckets[hash]); - rcu_assign_pointer(tbl->buckets[hash], obj); - ht->nelems++; + tbl = rht_dereference(ht->tbl, ht); + + list_for_each_entry(walker, &ht->walkers, list) + walker->resize = true; if (ht->p.grow_decision && ht->p.grow_decision(ht, tbl->size)) rhashtable_expand(ht); + else if (ht->p.shrink_decision && ht->p.shrink_decision(ht, tbl->size)) + rhashtable_shrink(ht); + +unlock: + mutex_unlock(&ht->mutex); +} + +static void rhashtable_wakeup_worker(struct rhashtable *ht) +{ + struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); + struct bucket_table *new_tbl = rht_dereference_rcu(ht->future_tbl, ht); + size_t size = tbl->size; + + /* Only adjust the table if no resizing is currently in progress. */ + if (tbl == new_tbl && + ((ht->p.grow_decision && ht->p.grow_decision(ht, size)) || + (ht->p.shrink_decision && ht->p.shrink_decision(ht, size)))) + schedule_work(&ht->run_work); +} + +static void __rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj, + struct bucket_table *tbl, u32 hash) +{ + struct rhash_head *head; + + hash = rht_bucket_index(tbl, hash); + head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash); + + ASSERT_BUCKET_LOCK(ht, tbl, hash); + + 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); + + rhashtable_wakeup_worker(ht); } -EXPORT_SYMBOL_GPL(rhashtable_insert); /** - * rhashtable_remove_pprev - remove object from hash table given previous element + * rhashtable_insert - insert object into hash table * @ht: hash table * @obj: pointer to hash head inside object - * @pprev: pointer to previous element * - * Identical to rhashtable_remove() but caller is alreayd aware of the element - * in front of the element to be deleted. This is in particular useful for - * deletion when combined with walking or lookup. + * 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. + * + * 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(). */ -void rhashtable_remove_pprev(struct rhashtable *ht, struct rhash_head *obj, - struct rhash_head __rcu **pprev) +void rhashtable_insert(struct rhashtable *ht, struct rhash_head *obj) { - struct bucket_table *tbl = rht_dereference(ht->tbl, ht); + struct bucket_table *tbl, *old_tbl; + unsigned hash; - ASSERT_RHT_MUTEX(ht); + rcu_read_lock(); - RCU_INIT_POINTER(*pprev, obj->next); - ht->nelems--; + tbl = rht_dereference_rcu(ht->future_tbl, ht); + old_tbl = rht_dereference_rcu(ht->tbl, ht); + hash = obj_raw_hashfn(ht, rht_obj(ht, obj)); - if (ht->p.shrink_decision && - ht->p.shrink_decision(ht, tbl->size)) - rhashtable_shrink(ht); + lock_buckets(tbl, old_tbl, hash); + __rhashtable_insert(ht, obj, tbl, hash); + unlock_buckets(tbl, old_tbl, hash); + + rcu_read_unlock(); } -EXPORT_SYMBOL_GPL(rhashtable_remove_pprev); +EXPORT_SYMBOL_GPL(rhashtable_insert); /** * rhashtable_remove - remove object from hash table @@ -398,7 +615,7 @@ EXPORT_SYMBOL_GPL(rhashtable_remove_pprev); * walk the bucket chain upon removal. The removal operation is thus * considerable slow if the hash table is not correctly sized. * - * Will automatically shrink the table via rhashtable_expand() if the the + * Will automatically shrink the table via rhashtable_expand() if the * shrink_decision function specified at rhashtable_init() returns true. * * The caller must ensure that no concurrent table mutations occur. It is @@ -406,30 +623,87 @@ EXPORT_SYMBOL_GPL(rhashtable_remove_pprev); */ bool rhashtable_remove(struct rhashtable *ht, struct rhash_head *obj) { - struct bucket_table *tbl = rht_dereference(ht->tbl, ht); + struct bucket_table *tbl, *new_tbl, *old_tbl; struct rhash_head __rcu **pprev; - struct rhash_head *he; - u32 h; + struct rhash_head *he, *he2; + unsigned int hash, new_hash; + bool ret = false; - ASSERT_RHT_MUTEX(ht); - - h = head_hashfn(ht, obj, tbl->size); - - pprev = &tbl->buckets[h]; - rht_for_each(he, tbl->buckets[h], ht) { + rcu_read_lock(); + old_tbl = rht_dereference_rcu(ht->tbl, ht); + tbl = new_tbl = rht_dereference_rcu(ht->future_tbl, ht); + new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj)); + + lock_buckets(new_tbl, old_tbl, new_hash); +restart: + hash = rht_bucket_index(tbl, new_hash); + pprev = &tbl->buckets[hash]; + rht_for_each(he, tbl, hash) { if (he != obj) { pprev = &he->next; continue; } - rhashtable_remove_pprev(ht, he, pprev); - return true; + ASSERT_BUCKET_LOCK(ht, tbl, hash); + + 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) { + if (head_hashfn(ht, tbl, he2) == hash) { + rcu_assign_pointer(*pprev, he2); + goto found; + } + } + + rcu_assign_pointer(*pprev, (struct rhash_head *) rht_marker(ht, hash)); + } else { + rcu_assign_pointer(*pprev, obj->next); + } + +found: + ret = true; + break; + } + + /* 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. + */ + if (tbl != old_tbl) { + tbl = old_tbl; + goto restart; + } + + unlock_buckets(new_tbl, old_tbl, new_hash); + + if (ret) { + atomic_dec(&ht->nelems); + rhashtable_wakeup_worker(ht); } - return false; + rcu_read_unlock(); + + return ret; } EXPORT_SYMBOL_GPL(rhashtable_remove); +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); +} + /** * rhashtable_lookup - lookup key in hash table * @ht: hash table @@ -439,65 +713,313 @@ EXPORT_SYMBOL_GPL(rhashtable_remove); * 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 - * paramter set). It will BUG() if used inappropriately. + * parameter set). It will BUG() if used inappropriately. * - * Lookups may occur in parallel with hash mutations as long as the lookup is - * guarded by rcu_read_lock(). The caller must take care of this. + * Lookups may occur in parallel with hashtable mutations and resizing. */ -void *rhashtable_lookup(const struct rhashtable *ht, const void *key) +void *rhashtable_lookup(struct rhashtable *ht, const void *key) { - const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); - struct rhash_head *he; - u32 h; + struct rhashtable_compare_arg arg = { + .ht = ht, + .key = key, + }; BUG_ON(!ht->p.key_len); - h = __hashfn(ht, key, ht->p.key_len, tbl->size); - rht_for_each_rcu(he, tbl->buckets[h], ht) { - if (memcmp(rht_obj(ht, he) + ht->p.key_offset, key, - ht->p.key_len)) - continue; - return (void *) he - ht->p.head_offset; - } - - return NULL; + return rhashtable_lookup_compare(ht, key, &rhashtable_compare, &arg); } EXPORT_SYMBOL_GPL(rhashtable_lookup); /** * rhashtable_lookup_compare - search hash table with compare function * @ht: hash table - * @hash: hash value of desired entry + * @key: the pointer to the key * @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. * - * Lookups may occur in parallel with hash mutations as long as the lookup is - * guarded by rcu_read_lock(). The caller must take care of this. + * Lookups may occur in parallel with hashtable mutations and resizing. * * Returns the first entry on which the compare function returned true. */ -void *rhashtable_lookup_compare(const struct rhashtable *ht, u32 hash, +void *rhashtable_lookup_compare(struct rhashtable *ht, const void *key, bool (*compare)(void *, void *), void *arg) { - const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); + const struct bucket_table *tbl, *old_tbl; struct rhash_head *he; + u32 hash; - if (unlikely(hash >= tbl->size)) - return NULL; + rcu_read_lock(); - rht_for_each_rcu(he, tbl->buckets[hash], ht) { + old_tbl = rht_dereference_rcu(ht->tbl, ht); + tbl = rht_dereference_rcu(ht->future_tbl, ht); + hash = key_hashfn(ht, key, ht->p.key_len); +restart: + rht_for_each_rcu(he, tbl, rht_bucket_index(tbl, hash)) { if (!compare(rht_obj(ht, he), arg)) continue; - return (void *) he - ht->p.head_offset; + rcu_read_unlock(); + return rht_obj(ht, he); + } + + if (unlikely(tbl != old_tbl)) { + tbl = old_tbl; + goto restart; } + rcu_read_unlock(); return NULL; } EXPORT_SYMBOL_GPL(rhashtable_lookup_compare); +/** + * 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) +{ + 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) +{ + struct bucket_table *new_tbl, *old_tbl; + u32 new_hash; + 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); + new_hash = obj_raw_hashfn(ht, rht_obj(ht, obj)); + + lock_buckets(new_tbl, old_tbl, new_hash); + + if (rhashtable_lookup_compare(ht, rht_obj(ht, obj) + ht->p.key_offset, + compare, arg)) { + success = false; + goto exit; + } + + __rhashtable_insert(ht, obj, new_tbl, new_hash); + +exit: + unlock_buckets(new_tbl, old_tbl, new_hash); + rcu_read_unlock(); + + return success; +} +EXPORT_SYMBOL_GPL(rhashtable_lookup_compare_insert); + +/** + * 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; + + 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 |