static void *assoc_maintenance_thread(void *arg) {
mutex_lock(&maintenance_lock);
while (do_run_maintenance_thread) {
int ii = 0;
/* There is only one expansion thread, so no need to global lock. */
for (ii = 0; ii < hash_bulk_move && expanding; ++ii) {
item *it, *next;
int bucket;
void *item_lock = NULL;
/* bucket = hv & hashmask(hashpower) =>the bucket of hash table
* is the lowest N bits of the hv, and the bucket of item_locks is
* also the lowest M bits of hv, and N is greater than M.
* So we can process expanding with only one item_lock. cool! */
if ((item_lock = item_trylock(expand_bucket))) {
for (it = old_hashtable[expand_bucket]; NULL != it; it = next) {
next = it->h_next;
bucket = hash(ITEM_key(it), it->nkey) & 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)) {
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");
}
} else {
usleep(10*1000);
}
if (item_lock) {
item_trylock_unlock(item_lock);
item_lock = NULL;
}
}
if (!expanding) {
/* We are done expanding.. just wait for next invocation */
started_expanding = false;
pthread_cond_wait(&maintenance_cond, &maintenance_lock);
/* assoc_expand() swaps out the hash table entirely, so we need
* all threads to not hold any references related to the hash
* table while this happens.
* This is instead of a more complex, possibly slower algorithm to
* allow dynamic hash table expansion without causing significant
* wait times.
*/
pause_threads(PAUSE_ALL_THREADS);
assoc_expand();
pause_threads(RESUME_ALL_THREADS);
}
}
return NULL;
}
/* 扩容线程函数，扩容策略如下:
* 扩容线程在main函数中创建，在assoc_insert后发现item数目大于哈希表容量1.5倍，唤醒扩容线程。
* 扩容线程先创建一个2倍容量的新哈希表，然后进行把数据从旧哈希表迁移到新哈希表。
* 迁移从旧表索引0开始，每次迁移一个桶(可以增加迁移粒度，但由于迁移需要加锁，可能导致work线程获取锁的等待时间增加)，
* 迁移完成后释放旧表。
*/
static void *assoc_maintenance_thread(void *arg) {
mutex_lock(&maintenance_lock);
while (do_run_maintenance_thread) {
int ii = 0;
/* There is only one expansion thread, so no need to global lock. */
//如果expanding为true才会进入循环体，所以迁移线程刚创建的时候，并不会进入循环体
for (ii = 0; ii < hash_bulk_move && expanding; ++ii) {
item *it, *next;
int bucket;
void *item_lock = NULL;
/* bucket = hv & hashmask(hashpower) =>the bucket of hash table
* is the lowest N bits of the hv, and the bucket of item_locks is
* also the lowest M bits of hv, and N is greater than M.
* So we can process expanding with only one item_lock. cool! */
//获取单个桶锁
if ((item_lock = item_trylock(expand_bucket))) {
//迁移一个桶中所有item
for (it = old_hashtable[expand_bucket]; NULL != it; it = next) {
next = it->h_next;
//重新计算哈希值
bucket = hash(ITEM_key(it), it->nkey) & hashmask(hashpower);
it->h_next = primary_hashtable[bucket];
primary_hashtable[bucket] = it;
}
old_hashtable[expand_bucket] = NULL;
expand_bucket++;
//迁移完成
if (expand_bucket == hashsize(hashpower - 1)) {
expanding = false; //将迁移标志设置0
free(old_hashtable); //释放旧表
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");
}
} else {
usleep(10*1000);
}
if (item_lock) {
item_trylock_unlock(item_lock);
item_lock = NULL;
}
}
if (!expanding) {
/* We are done expanding.. just wait for next invocation */
//不需要迁移，挂起迁移线程，直到worker线程插入数据后发现item数量已经到了1.5倍哈希表大小，
//此时调用worker线程调用assoc_start_expand函数，该函数会调用pthread_cond_signal
//唤醒迁移线程
started_expanding = false;
pthread_cond_wait(&maintenance_cond, &maintenance_lock);
/* assoc_expand() swaps out the hash table entirely, so we need
* all threads to not hold any references related to the hash
* table while this happens.
* This is instead of a more complex, possibly slower algorithm to
* allow dynamic hash table expansion without causing significant
* wait times.
*/
pause_threads(PAUSE_ALL_THREADS);
assoc_expand(); //申请更大的哈希表,并将expanding设置为true
pause_threads(RESUME_ALL_THREADS);
}
}
return NULL;
}
void *admin_thread(void *UnusedArg)
{
SetNameFunction("admin_thr");
while(1)
{
P(mutex_admin_condvar);
while(reload_exports == FALSE)
pthread_cond_wait(&(admin_condvar), &(mutex_admin_condvar));
reload_exports = FALSE;
V(mutex_admin_condvar);
if (rebuild_export_list() <= 0)
{
LogCrit(COMPONENT_MAIN, "Could not reload the exports list.");
continue;
}
if(pause_threads(PAUSE_RELOAD_EXPORTS) == PAUSE_EXIT)
{
LogDebug(COMPONENT_MAIN,
"Export reload interrupted by shutdown while pausing threads");
/* Be helpfull and exit
* (other termination will just blow us away, and that's ok...
*/
break;
}
/* Clear the id mapping cache for gss principals to uid/gid.
* The id mapping may have changed.
*/
#ifdef _HAVE_GSSAPI
#ifdef _USE_NFSIDMAP
uidgidmap_clear();
idmap_clear();
namemap_clear();
#endif /* _USE_NFSIDMAP */
#endif /* _HAVE_GSSAPI */
if (ChangeoverExports())
{
LogCrit(COMPONENT_MAIN, "ChangeoverExports failed.");
continue;
}
LogEvent(COMPONENT_MAIN,
"Exports reloaded and active");
/* wake_workers could return PAUSE_PAUSE, but we don't have to do
* anything special in that case.
*/
if(wake_threads(AWAKEN_RELOAD_EXPORTS) == PAUSE_EXIT)
{
LogDebug(COMPONENT_MAIN,
"Export reload interrupted by shutdown while waking threads");
/* Be helpfull and exit
* (other termination will just blow us away, and that's ok...
*/
break;
}
}
return NULL;
} /* admin_thread */
static void *assoc_maintenance_thread(void *arg) {
mutex_lock(&maintenance_lock);
//do_run_maintenance_thread是全局变量，初始值为1，在stop_assoc_maintenance_thread
//函数中会被赋值0，终止迁移线程
while (do_run_maintenance_thread) {
int ii = 0;
/* There is only one expansion thread, so no need to global lock. */
for (ii = 0; ii < hash_bulk_move && expanding; ++ii) {
item *it, *next;
int bucket;
void *item_lock = NULL;
/* bucket = hv & hashmask(hashpower) =>the bucket of hash table
* is the lowest N bits of the hv, and the bucket of item_locks is
* also the lowest M bits of hv, and N is greater than M.
* So we can process expanding with only one item_lock. cool! */
//hash_bulk_move用来控制每次迁移，移动多少个桶的item。默认是一个.
//如果expanding为true才会进入循环体，所以迁移线程刚创建的时候，并不会进入循环体
if ((item_lock = item_trylock(expand_bucket))) {
//在assoc_expand函数中expand_bucket会被赋值0
//遍历旧哈希表中由expand_bucket指明的桶,将该桶的所有item
//迁移到新哈希表中。
for (it = old_hashtable[expand_bucket]; NULL != it; it = next) {
next = it->h_next;
//重新计算新的哈希值,得到其在新哈希表的位置
bucket = hash(ITEM_key(it), it->nkey) & hashmask(hashpower);
//将这个item插入到新哈希表中
it->h_next = primary_hashtable[bucket];
primary_hashtable[bucket] = it;
}
//不需要清空旧桶。直接将冲突链的链头赋值为NULL即可
old_hashtable[expand_bucket] = NULL;
//迁移完一个桶，接着把expand_bucket指向下一个待迁移的桶
expand_bucket++;
if (expand_bucket == hashsize(hashpower - 1)) {//全部数据迁移完毕
expanding = false;//将扩展标志设置为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");
}
} else {
usleep(10*1000);
}
if (item_lock) {
//遍历完hash_bulk_move个桶的所有item后，就释放锁
item_trylock_unlock(item_lock);
item_lock = NULL;
}
}
if (!expanding) {//不需要迁移数据(了)
/* We are done expanding.. just wait for next invocation */
started_expanding = false;//重置
//挂起迁移线程，直到worker线程插入数据后发现item数量已经到了1.5倍哈希表大小，
//此时调用worker线程调用assoc_start_expand函数，该函数会调用pthread_cond_signal
//唤醒迁移线程
pthread_cond_wait(&maintenance_cond, &maintenance_lock);
/* assoc_expand() swaps out the hash table entirely, so we need
* all threads to not hold any references related to the hash
* table while this happens.
* This is instead of a more complex, possibly slower algorithm to
* allow dynamic hash table expansion without causing significant
* wait times.
*/
pause_threads(PAUSE_ALL_THREADS);
assoc_expand();//申请更大的哈希表,并将expanding设置为true
pause_threads(RESUME_ALL_THREADS);
}
}
return NULL;
}
