openvswitch/lib/hmap.c

/*
 * Copyright (c) 2008, 2009, 2010, 2012, 2013 Nicira, Inc.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <config.h>
#include "hmap.h"
#include <stdint.h>
#include <string.h>
#include "coverage.h"
#include "random.h"
#include "util.h"
#include "vlog.h"

VLOG_DEFINE_THIS_MODULE(hmap);

COVERAGE_DEFINE(hmap_pathological);
COVERAGE_DEFINE(hmap_expand);
COVERAGE_DEFINE(hmap_shrink);
COVERAGE_DEFINE(hmap_reserve);

/* Initializes 'hmap' as an empty hash table. */
void
hmap_init(struct hmap *hmap)
{
    hmap->buckets = &hmap->one;
    hmap->one = NULL;
    hmap->mask = 0;
    hmap->n = 0;
}

/* Frees memory reserved by 'hmap'.  It is the client's responsibility to free
 * the nodes themselves, if necessary. */
void
hmap_destroy(struct hmap *hmap)
{
    if (hmap && hmap->buckets != &hmap->one) {
        free(hmap->buckets);
    }
}

/* Removes all node from 'hmap', leaving it ready to accept more nodes.  Does
 * not free memory allocated for 'hmap'.
 *
 * This function is appropriate when 'hmap' will soon have about as many
 * elements as it before.  If 'hmap' will likely have fewer elements than
 * before, use hmap_destroy() followed by hmap_clear() to save memory and
 * iteration time. */
void
hmap_clear(struct hmap *hmap)
{
    if (hmap->n > 0) {
        hmap->n = 0;
        memset(hmap->buckets, 0, (hmap->mask + 1) * sizeof *hmap->buckets);
    }
}

/* Exchanges hash maps 'a' and 'b'. */
void
hmap_swap(struct hmap *a, struct hmap *b)
{
    struct hmap tmp = *a;
    *a = *b;
    *b = tmp;
    hmap_moved(a);
    hmap_moved(b);
}

/* Adjusts 'hmap' to compensate for having moved position in memory (e.g. due
 * to realloc()). */
void
hmap_moved(struct hmap *hmap)
{
    if (!hmap->mask) {
        hmap->buckets = &hmap->one;
    }
}

static void
resize(struct hmap *hmap, size_t new_mask, const char *where)
{
    struct hmap tmp;
    size_t i;

    ovs_assert(is_pow2(new_mask + 1));

    hmap_init(&tmp);
    if (new_mask) {
        tmp.buckets = xmalloc(sizeof *tmp.buckets * (new_mask + 1));
        tmp.mask = new_mask;
        for (i = 0; i <= tmp.mask; i++) {
            tmp.buckets[i] = NULL;
        }
    }
    for (i = 0; i <= hmap->mask; i++) {
        struct hmap_node *node, *next;
        int count = 0;
        for (node = hmap->buckets[i]; node; node = next) {
            next = node->next;
            hmap_insert_fast(&tmp, node, node->hash);
            count++;
        }
        if (count > 5) {
            static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
            COVERAGE_INC(hmap_pathological);
            VLOG_DBG_RL(&rl, "%s: %d nodes in bucket (%"PRIuSIZE" nodes, %"PRIuSIZE" buckets)",
                        where, count, hmap->n, hmap->mask + 1);
        }
    }
    hmap_swap(hmap, &tmp);
    hmap_destroy(&tmp);
}

static size_t
calc_mask(size_t capacity)
{
    size_t mask = capacity / 2;
    mask |= mask >> 1;
    mask |= mask >> 2;
    mask |= mask >> 4;
    mask |= mask >> 8;
    mask |= mask >> 16;
#if SIZE_MAX > UINT32_MAX
    mask |= mask >> 32;
#endif

    /* If we need to dynamically allocate buckets we might as well allocate at
     * least 4 of them. */
    mask |= (mask & 1) << 1;

    return mask;
}

/* Expands 'hmap', if necessary, to optimize the performance of searches.
 *
 * ('where' is used in debug logging.  Commonly one would use hmap_expand() to
 * automatically provide the caller's source file and line number for
 * 'where'.) */
void
hmap_expand_at(struct hmap *hmap, const char *where)
{
    size_t new_mask = calc_mask(hmap->n);
    if (new_mask > hmap->mask) {
        COVERAGE_INC(hmap_expand);
        resize(hmap, new_mask, where);
    }
}

/* Shrinks 'hmap', if necessary, to optimize the performance of iteration.
 *
 * ('where' is used in debug logging.  Commonly one would use hmap_shrink() to
 * automatically provide the caller's source file and line number for
 * 'where'.) */
void
hmap_shrink_at(struct hmap *hmap, const char *where)
{
    size_t new_mask = calc_mask(hmap->n);
    if (new_mask < hmap->mask) {
        COVERAGE_INC(hmap_shrink);
        resize(hmap, new_mask, where);
    }
}

/* Expands 'hmap', if necessary, to optimize the performance of searches when
 * it has up to 'n' elements.  (But iteration will be slow in a hash map whose
 * allocated capacity is much higher than its current number of nodes.)
 *
 * ('where' is used in debug logging.  Commonly one would use hmap_reserve() to
 * automatically provide the caller's source file and line number for
 * 'where'.) */
void
hmap_reserve_at(struct hmap *hmap, size_t n, const char *where)
{
    size_t new_mask = calc_mask(n);
    if (new_mask > hmap->mask) {
        COVERAGE_INC(hmap_reserve);
        resize(hmap, new_mask, where);
    }
}

/* Adjusts 'hmap' to compensate for 'old_node' having moved position in memory
 * to 'node' (e.g. due to realloc()). */
void
hmap_node_moved(struct hmap *hmap,
                struct hmap_node *old_node, struct hmap_node *node)
{
    struct hmap_node **bucket = &hmap->buckets[node->hash & hmap->mask];
    while (*bucket != old_node) {
        bucket = &(*bucket)->next;
    }
    *bucket = node;
}

/* Chooses and returns a randomly selected node from 'hmap', which must not be
 * empty.
 *
 * I wouldn't depend on this algorithm to be fair, since I haven't analyzed it.
 * But it does at least ensure that any node in 'hmap' can be chosen. */
struct hmap_node *
hmap_random_node(const struct hmap *hmap)
{
    struct hmap_node *bucket, *node;
    size_t n, i;

    /* Choose a random non-empty bucket. */
    for (i = random_uint32(); ; i++) {
        bucket = hmap->buckets[i & hmap->mask];
        if (bucket) {
            break;
        }
    }

    /* Count nodes in bucket. */
    n = 0;
    for (node = bucket; node; node = node->next) {
        n++;
    }

    /* Choose random node from bucket. */
    i = random_range(n);
    for (node = bucket; i-- > 0; node = node->next) {
        continue;
    }
    return node;
}

/* Returns the next node in 'hmap' in hash order, or NULL if no nodes remain in
 * 'hmap'.  Uses '*bucketp' and '*offsetp' to determine where to begin
 * iteration, and stores new values to pass on the next iteration into them
 * before returning.
 *
 * It's better to use plain HMAP_FOR_EACH and related functions, since they are
 * faster and better at dealing with hmaps that change during iteration.
 *
 * Before beginning iteration, store 0 into '*bucketp' and '*offsetp'.
 */
struct hmap_node *
hmap_at_position(const struct hmap *hmap,
                 uint32_t *bucketp, uint32_t *offsetp)
{
    size_t offset;
    size_t b_idx;

    offset = *offsetp;
    for (b_idx = *bucketp; b_idx <= hmap->mask; b_idx++) {
        struct hmap_node *node;
        size_t n_idx;

        for (n_idx = 0, node = hmap->buckets[b_idx]; node != NULL;
             n_idx++, node = node->next) {
            if (n_idx == offset) {
                if (node->next) {
                    *bucketp = node->hash & hmap->mask;
                    *offsetp = offset + 1;
                } else {
                    *bucketp = (node->hash & hmap->mask) + 1;
                    *offsetp = 0;
                }
                return node;
            }
        }
        offset = 0;
    }

    *bucketp = 0;
    *offsetp = 0;
    return NULL;
}

/* Returns true if 'node' is in 'hmap', false otherwise. */
bool
hmap_contains(const struct hmap *hmap, const struct hmap_node *node)
{
    struct hmap_node *p;

    for (p = hmap_first_in_bucket(hmap, node->hash); p; p = p->next) {
        if (p == node) {
            return true;
        }
    }

    return false;
}