memcached内存管理(3) ----------------assoc

assoc.{h,c}

这个文件的作用是通过item的hash值来对item进行存取和删除的操作，当然删除是指从hash表中删除，而不是真正的释放内存

一些重要的全局变量

typedef  unsigned long  int  ub4;   /* unsigned 4-byte quantities */
typedef  unsigned       char ub1;   /* unsigned 1-byte quantities */
/* how many powers of 2's worth of buckets we use */
static unsigned int hashpower = HASHPOWER_DEFAULT;             //决定hash表的大小  2^hashpower
#define hashsize(n) ((ub4)1<<(n))
#define hashmask(n) (hashsize(n)-1)

/* Main hash table. This is where we look except during expansion.   主hash表*/
static item** primary_hashtable = 0;
/*
 * Previous hash table. During expansion, we look here for keys that haven't
 * been moved over to the primary yet.
 */
static item** old_hashtable = 0;                    //当hash表在扩张时，保存主表信息，并用于存取操作，主表完成扩张时就释放占用的内存

/* Number of items in the hash table. */
static unsigned int hash_items = 0;

/* Flag: Are we in the middle of expanding now? 是否正在扩张*/ 
static bool expanding = false;

/*
 * During expansion we migrate values with bucket granularity; this is how
 * far we've gotten so far. Ranges from 0 .. hashsize(hashpower - 1) - 1.
 */
static unsigned int expand_bucket = 0;        //在扩张时，是以桶为粒度进行的，这是告诉我们扩张到哪个桶了。从0 到 hashsize(hashpower - 1) - 1

初始化hash表

void assoc_init(const int hashtable_init) {
    if (hashtable_init) {
        hashpower = hashtable_init;
    }
    primary_hashtable = calloc(hashsize(hashpower), sizeof(void *));
    if (! primary_hashtable) {
        fprintf(stderr, "Failed to init hashtable.\n");
        exit(EXIT_FAILURE);
    }
    STATS_LOCK();
    stats.hash_power_level = hashpower;
    stats.hash_bytes = hashsize(hashpower) * sizeof(void *);
    STATS_UNLOCK();
}

非常简单，就是给primary_hash表分配对应大小的内存，并置0

primary_hash 是item **，指向item *的指针， 就是说primary_hash[1]就是一个item *，指向一个item，其实这个表并不存放item，就是说充当item中的h_next指针。

查找操作

item *assoc_find(const char *key, const size_t nkey, const uint32_t hv) {
    item *it;
    unsigned int oldbucket;

    if (expanding &&
        (oldbucket = (hv & hashmask(hashpower - 1))) >= expand_bucket)
    {
        it = old_hashtable[oldbucket];
    } else {
        it = primary_hashtable[hv & hashmask(hashpower)];
    }

    item *ret = NULL;
    int depth = 0;
    while (it) {
        if ((nkey == it->nkey) && (memcmp(key, ITEM_key(it), nkey) == 0)) {
            ret = it;
            break;
        }
        it = it->h_next;
        ++depth;
    }
    MEMCACHED_ASSOC_FIND(key, nkey, depth);
    return ret;
}

参数 key 就是键值， nkey 键长， hv是经过hash操作得到的值

在thread.c中有调用

    uint32_t hv;
    hv = hash(key, nkey, 0);

可以看出hash值跟只跟key和nkey有关，跟具体data无关。

插入过程也很简单：

1.看桶的index，如果正在扩张，且index >= extend_bucket则选择old_hashtable，否则选择primary_hashtable。

2. 遍历对应的桶查找key相等的item，找到就就返回该item

这里有个depth变量，看需要在桶中找几次才能找到该key，如果depth太深表示在该bucket上有太多的item，hash表可能分配并不均匀。

插入操作

/* Note: this isn't an assoc_update.  The key must not already exist to call this */
int assoc_insert(item *it, const uint32_t hv) {
    unsigned int oldbucket;

//    assert(assoc_find(ITEM_key(it), it->nkey) == 0);  /* shouldn't have duplicately named things defined */

    if (expanding &&
        (oldbucket = (hv & hashmask(hashpower - 1))) >= expand_bucket)
    {
        it->h_next = old_hashtable[oldbucket];
        old_hashtable[oldbucket] = it;
    } else {
        it->h_next = primary_hashtable[hv & hashmask(hashpower)];
        primary_hashtable[hv & hashmask(hashpower)] = it;
    }

    hash_items++;
    if (! expanding && hash_items > (hashsize(hashpower) * 3) / 2) {
        assoc_expand();
    }

    MEMCACHED_ASSOC_INSERT(ITEM_key(it), it->nkey, hash_items);
    return 1;
}

操作跟find差不多，就是找到对应的桶后，将新的item作为bucket的第一个元素。

要注意的是当hash_item即hash表中的元素大于hashsize(hashpower)的1.5倍时，要扩张hash表了。

删除操作

void assoc_delete(const char *key, const size_t nkey, const uint32_t hv) {
    item **before = _hashitem_before(key, nkey, hv);

    //before      指向前一个item 的h_next（如果有个话）
    // *before   h_next的值，就是该删除的item的地址

    if (*before) {
        item *nxt;           
        hash_items--;
        /* The DTrace probe cannot be triggered as the last instruction
         * due to possible tail-optimization by the compiler
         */
        MEMCACHED_ASSOC_DELETE(key, nkey, hash_items);
        nxt = (*before)->h_next;              //(*before)->h_next就是该删除的item的下一个item，也可能是null
        (*before)->h_next = 0;   /* probably pointless, but whatever. */
        *before = nxt;                                
        return;
    }
    /* Note:  we never actually get here.  the callers don't delete things
       they can't find. */
    assert(*before != 0);
}

就是单链表的删除操作，这里的具体过程如下

1. 找到指向 将要被删除的item 的指针， 即 item ** before，如果有2个以上元素的桶，其实就是指向上一个item的h_next，只有一个元素则是primary_hash[n] ，这里调用来_hashitem_before

2. 正常的删除单向链表操作。

可以看到这里只是单纯的从链表删除，没有任何的释放内存操作。

扩张hash表

/* grows the hashtable to the next power of 2. */
static void assoc_expand(void) {
    old_hashtable = primary_hashtable;

    primary_hashtable = calloc(hashsize(hashpower + 1), sizeof(void *));
    if (primary_hashtable) {
        if (settings.verbose > 1)
            fprintf(stderr, "Hash table expansion starting\n");
        hashpower++;
        expanding = true;
        expand_bucket = 0;
        STATS_LOCK();
        stats.hash_power_level = hashpower;
        stats.hash_bytes += hashsize(hashpower) * sizeof(void *);
        stats.hash_is_expanding = 1;
        STATS_UNLOCK();
        pthread_cond_signal(&maintenance_cond);
    } else {
        primary_hashtable = old_hashtable;
        /* Bad news, but we can keep running. */
    }
}

这里成功的话，只是将正在primary_hashtable交由old_hashtable处理，然后申请新的内存来存取新表，新表是旧表的2倍
最后通过

pthread_cond_signal(&maintenance_cond);

唤醒正在等待这个条件变量的线程。

static pthread_t maintenance_tid;

int start_assoc_maintenance_thread() {
    int ret;
    char *env = getenv("MEMCACHED_HASH_BULK_MOVE");
    if (env != NULL) {
        hash_bulk_move = atoi(env);
        if (hash_bulk_move == 0) {
            hash_bulk_move = DEFAULT_HASH_BULK_MOVE;
        }
    }
    if ((ret = pthread_create(&maintenance_tid, NULL,
                              assoc_maintenance_thread, NULL)) != 0) {
        fprintf(stderr, "Can't create thread: %s\n", strerror(ret));
        return -1;
    }
    return 0;
}

创建新线程，线程运行函数

static volatile int do_run_maintenance_thread = 1;

#define DEFAULT_HASH_BULK_MOVE 1
int hash_bulk_move = DEFAULT_HASH_BULK_MOVE;

static void *assoc_maintenance_thread(void *arg) {

    while (do_run_maintenance_thread) {
        int ii = 0;

        /* Lock the cache, and bulk move multiple buckets to the new
         * hash table. */
        mutex_lock(&cache_lock);

        for (ii = 0; ii < hash_bulk_move && expanding; ++ii) {
            item *it, *next;
            int bucket;

           //将旧表中的元素按新的hash方法   hashpower + 1了，放到新的hash表中
            for (it = old_hashtable[expand_bucket]; NULL != it; it = next) {
                next = it->h_next;

                bucket = hash(ITEM_key(it), it->nkey, 0) & hashmask(hashpower);
                it->h_next = primary_hashtable[bucket];
                primary_hashtable[bucket] = it;
            }

            old_hashtable[expand_bucket] = NULL;

            expand_bucket++;
            if (expand_bucket == hashsize(hashpower - 1)) {           //旧表中的东西全部搬完后，关闭expanding状态。释放old_hashtable
                expanding = false;
                free(old_hashtable);
                STATS_LOCK();
                stats.hash_bytes -= hashsize(hashpower - 1) * sizeof(void *);
                stats.hash_is_expanding = 0;
                STATS_UNLOCK();
                if (settings.verbose > 1)
                    fprintf(stderr, "Hash table expansion done\n");
            }
        }

        if (!expanding) {
            /* We are done expanding.. just wait for next invocation */
            //等待唤醒
           pthread_cond_wait(&maintenance_cond, &cache_lock);
        }

        mutex_unlock(&cache_lock);
    }
    return NULL;
}

大概就这样了。