postgres 源码解析51 LWLock轻量锁--2

news2024/10/2 6:43:13

本篇将着重讲解LWLock涉及的主要API工作流程与实现原理，相关基础知识见回顾：postgres 源码解析50 LWLock轻量锁–1

API介绍

函数API	功能
CreateLWLocks	分配LWLocks所需的内存并进行初始化
LWLockNewTrancheId	分配新的Tranche ID,供用户使用Extension模块中的LWLocks
LWLockAcquire	获取指定的LWLock
LWLockConditionalAcquire	获取LWLock,与LWLockAcquire不同的是未获取到不等待直接返回
LWLockAttemptLock	尝试获取LWLock
LWLockRelease	释放已申请的LWLock
LWLockWakeup	唤醒等待队列中LWLock

LWLock的主要操作

1 LWLock的空间分配（CreateLWLocks）

该函数为LWLocks分配内存空间并初始化，同时注册扩展NamedTranche。
执行流程：
1）首先计算出LWLocks和NamedTranche所占用的内存空间,该过程定义在LWLockShmemSize函数中；
2）定义并初始化全局LWLocks数组首地址MainLWLockArray；
3）调用 InitializeLWLocks函数初始化所有LWLocks, 主要是设置lock->state和lock->tranched_id字段信息；
4）注册扩展LWLocks,即建立Tranche ID与Tranche Name之间的联系。

/*
 * Allocate shmem space for the main LWLock array and all tranches and
 * initialize it.  We also register extension LWLock tranches here.
 */
void
CreateLWLocks(void)
{
	StaticAssertStmt(LW_VAL_EXCLUSIVE > (uint32) MAX_BACKENDS,
					 "MAX_BACKENDS too big for lwlock.c");

	StaticAssertStmt(sizeof(LWLock) <= LWLOCK_PADDED_SIZE,
					 "Miscalculated LWLock padding");

	if (!IsUnderPostmaster)
	{
		Size		spaceLocks = LWLockShmemSize();
		int		   *LWLockCounter;
		char	   *ptr;

		/* Allocate space */
		ptr = (char *) ShmemAlloc(spaceLocks);

		/* Leave room for dynamic allocation of tranches */
		ptr += sizeof(int);

		/* Ensure desired alignment of LWLock array */
		ptr += LWLOCK_PADDED_SIZE - ((uintptr_t) ptr) % LWLOCK_PADDED_SIZE;

		MainLWLockArray = (LWLockPadded *) ptr;

		/*
		 * Initialize the dynamic-allocation counter for tranches, which is
		 * stored just before the first LWLock.
		 */
		LWLockCounter = (int *) ((char *) MainLWLockArray - sizeof(int));
		*LWLockCounter = LWTRANCHE_FIRST_USER_DEFINED;

		/* Initialize all LWLocks */
		InitializeLWLocks();
	}

	/* Register named extension LWLock tranches in the current process. */
	for (int i = 0; i < NamedLWLockTrancheRequests; i++)
		LWLockRegisterTranche(NamedLWLockTrancheArray[i].trancheId,
							  NamedLWLockTrancheArray[i].trancheName);
}

2 LWLock Tranche ID的分配（LWLockNewTrancheId）

该函数实现了LWLock的分配，目的是向用户提供使用Extension模块中的轻量锁。
1 首先从共享内存中获取动态分配计数器LWLockCounter的地址；
2 申请spin_t自旋锁，该锁用于保护LWLock的分配过程；
3 自增分配计数器的值，释放锁；
4 最终返回该值作为新分配LWLock的全局标识Tranche ID;

/*
 * Allocate a new tranche ID.
 */
int
LWLockNewTrancheId(void)
{
	int			result;
	int		   *LWLockCounter;

	LWLockCounter = (int *) ((char *) MainLWLockArray - sizeof(int));
	SpinLockAcquire(ShmemLock);
	result = (*LWLockCounter)++;
	SpinLockRelease(ShmemLock);

	return result;
}

3 LWLock锁的获取（LWLockAcquire）

1 安全性检查，pg规定每个进程最多持有200个轻量锁，若超过报错退出；
2 开中断，调用LWLockAttemptLock函数尝试获取轻量锁（第一次），该函数返回值有两种，一种是该锁处于非空闲状态导致未获取到，则需要等待，另一种是成功获取到锁无需等待；
3 根据2结果进行后续处理。如果不等待，则表明成功获取锁跳出循环，反之进入步骤4；
4 调用LWLockQueueSelf函数将该进程加入到锁的等待队列；
5 再次调用LWLockAttemptLock（第二次）函数尝试获取轻量锁，如果无需等待，则成功获取锁，并将自身进程从锁的等待队列中移除跳出循环；反之在锁队列中等待直至被其他进程唤醒，回到步骤2。
6 更新本进程持有的锁数目和锁模式。

注：从源代码可以看出，有两次相邻的LWLockAttemptLock调用，推测其原因为LWLock锁保护的临界区较小，锁的持有时间较短，可能在将自身加入锁等待队列期间，其他进程已释放该锁，在一定程度上能够减小等待时间（一种优化）。
在这里插入图片描述

4 LWLockConditionalAcquire

轻量锁获取的另一种接口函数为LWLockConditionalAcquire，该函数与LWLockAcquire的不同点在于获取不到锁时不等待直接返回。

5 LWLockAttemptLock

该函数的功能是尝试获取LWLock，其本质是通过一系列原子操作来实现指定模式下锁的获取。
1 首先调用 pg_atomic_read_u32读取该锁的状态信息old_state；
2 进入死循环
1）根据指定的锁模式判断锁是否处于空闲lock_free并更新desired_state（该变量初始值=old_state，更新操作是加上LW_VAL_EXCLUSIVE/LW_VAL_SHRAED）；
2）如果成功调用pg_atomic_compare_exchange_u32函数实现上述old_stat与desired_state交换，则进一步根据lock_free标识判断：若lock_free = true, 则标记获取该LWLock并结束;反之，获取不到该锁并结束；
3）若未成功调用pg_atomic_compare_exchange_u32函数，则表明该锁已被其他人获取，需要再次循环重试。

/*
 * Internal function that tries to atomically acquire the lwlock in the passed
 * in mode.
 *
 * This function will not block waiting for a lock to become free - that's the
 * callers job.
 *
 * Returns true if the lock isn't free and we need to wait.
 */
static bool
LWLockAttemptLock(LWLock *lock, LWLockMode mode)
{
	uint32		old_state;

	AssertArg(mode == LW_EXCLUSIVE || mode == LW_SHARED);

	/*
	 * Read once outside the loop, later iterations will get the newer value
	 * via compare & exchange.
	 */
	old_state = pg_atomic_read_u32(&lock->state);

	/* loop until we've determined whether we could acquire the lock or not */
	while (true)
	{
		uint32		desired_state;
		bool		lock_free;

		desired_state = old_state;

		if (mode == LW_EXCLUSIVE)
		{
			lock_free = (old_state & LW_LOCK_MASK) == 0;
			if (lock_free)
				desired_state += LW_VAL_EXCLUSIVE;
		}
		else
		{
			lock_free = (old_state & LW_VAL_EXCLUSIVE) == 0;
			if (lock_free)
				desired_state += LW_VAL_SHARED;
		}

		/*
		 * Attempt to swap in the state we are expecting. If we didn't see
		 * lock to be free, that's just the old value. If we saw it as free,
		 * we'll attempt to mark it acquired. The reason that we always swap
		 * in the value is that this doubles as a memory barrier. We could try
		 * to be smarter and only swap in values if we saw the lock as free,
		 * but benchmark haven't shown it as beneficial so far.
		 *
		 * Retry if the value changed since we last looked at it.
		 */
		if (pg_atomic_compare_exchange_u32(&lock->state,
										   &old_state, desired_state))
		{
			if (lock_free)
			{
				/* Great! Got the lock. */
				if (mode == LW_EXCLUSIVE)
					lock->owner = MyProc;
				return false;
			}
			else
				return true;	/* somebody else has the lock */
		}
	}
	pg_unreachable();
}

LWLock锁的释放（LWLockRelease）

LWLock锁的释放定义在LWLockRelease函数中，实现流程如：
1 首先检查该LWLock是否持有，如果不持有则报错退出；
2 将锁从进程自身的持锁队列中移除，更新持锁队列；
3 根据锁模式，调用pg_atomic_sub_fetch_u32函数清除锁标记表明锁释放；
4 如果有其他进程等待该锁，则需根据某种调度策略唤醒等待进程；