TheRiver | blog

You have reached the world's edge, none but devils play past here

0%

linux-spinlock

Posted on Edited on
Symbols count in article: 6.3k Reading time ≈ 6 mins.

参考

Linux内核同步机制之（四）：spin lock

内核锁定技术

What exactly are “spin-locks”?

X86与ARM中的原子操作-原理及实现

原子操作对同步与互斥的意义

LINUX KERNEL SPINLOCK使用不当的后果

Linux 内核的排队自旋锁

概念

wikepedia

自旋锁是计算机科学用于多线程同步的一种锁，线程反复检查锁变量是否可用。由于线程在这一过程中保持执行，因此是一种忙等待。一旦获取了自旋锁，线程会一直保持该锁，直至显式释放自旋锁。

自旋锁避免了进程上下文的调度开销，因此对于线程只会阻塞很短时间的场合是有效的。因此操作系统的实现在很多地方往往用自旋锁。Windows操作系统提供的轻型读写锁（SRW Lock）内部就用了自旋锁。显然，单核CPU不适于使用自旋锁，这里的单核CPU指的是单核单线程的CPU，因为，在同一时间只有一个线程是处在运行状态，假设运行线程A发现无法获取锁，只能等待解锁，但因为A自身不挂起，所以那个持有锁的线程B没有办法进入运行状态，只能等到操作系统分给A的时间片用完，才能有机会被调度。这种情况下使用自旋锁的代价很高。

获取、释放自旋锁，实际上是读写自旋锁的存储内存或寄存器。因此这种读写操作必须是原子的。通常用test-and-set等原子操作来实现

上面说通常使用TSL指令实现，之前总结过TSL指令是同步的硬件实现方式，会锁住CPU的地址总线，所以是对多核CPU生效的，并且内部实现是忙等待机制。

数据结构

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
typedef struct spinlock {
	union {
		struct raw_spinlock rlock;

#ifdef CONFIG_DEBUG_LOCK_ALLOC
# define LOCK_PADSIZE (offsetof(struct raw_spinlock, dep_map))
		struct {
			u8 __padding[LOCK_PADSIZE];
			struct lockdep_map dep_map;
		};
#endif
	};
} spinlock_t;

typedef struct raw_spinlock {
	arch_spinlock_t raw_lock;
#ifdef CONFIG_GENERIC_LOCKBREAK
	unsigned int break_lock;
#endif
#ifdef CONFIG_DEBUG_SPINLOCK
	unsigned int magic, owner_cpu;
	void *owner;
#endif
#ifdef CONFIG_DEBUG_LOCK_ALLOC
	struct lockdep_map dep_map;
#endif
} raw_spinlock_t;

函数调用

spin_lock/spin_unlock — 禁止内核抢占

spin_lock_irq/spin_unlock_irq — 禁止内核抢占并屏蔽中断

spin_lock_irqsave/spin_unlock_irqrestore — 禁止内核抢占并屏蔽中断，事先保存中断屏蔽位并事后恢复原状

内核版本：
VERSION = 3
PATCHLEVEL = 10
SUBLEVEL = 1

spin_lock

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
static inline void spin_lock(spinlock_t *lock)
{
	raw_spin_lock(&lock->rlock);
}

#define raw_spin_lock(lock)	_raw_spin_lock(lock)

#ifndef CONFIG_INLINE_SPIN_LOCK
void __lockfunc _raw_spin_lock(raw_spinlock_t *lock)
{
	__raw_spin_lock(lock);
}

static inline void __raw_spin_lock(raw_spinlock_t *lock)
{
	unsigned long tmp;

	__asm__ __volatile__(
			"       movi    %0, 0\n"
			"       wsr     %0, scompare1\n"
			"1:     movi    %0, 1\n"
			"       s32c1i  %0, %1, 0\n"
			"       bnez    %0, 1b\n"
			: "=&a" (tmp)
			: "a" (&lock->slock)
			: "memory");
}

static inline void __raw_spin_lock(raw_spinlock_t *lock)
{
	preempt_disable();
	spin_acquire(&lock->dep_map, 0, 0, _RET_IP_);
	LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
}

spin_unlock

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
static inline void spin_unlock(spinlock_t *lock)
{
	raw_spin_unlock(&lock->rlock);
}

#define raw_spin_unlock(lock)		_raw_spin_unlock(lock)

#ifdef CONFIG_UNINLINE_SPIN_UNLOCK
void __lockfunc _raw_spin_unlock(raw_spinlock_t *lock)
{
	__raw_spin_unlock(lock);
}

static inline void __raw_spin_unlock(raw_spinlock_t *lock)
{
	unsigned long tmp;

	__asm__ __volatile__(
			"       movi    %0, 0\n"
			"       s32ri   %0, %1, 0\n"
			: "=&a" (tmp)
			: "a" (&lock->slock)
			: "memory");
}

static inline void __raw_spin_unlock(raw_spinlock_t *lock)
{
	spin_release(&lock->dep_map, 1, _RET_IP_);
	do_raw_spin_unlock(lock);
	preempt_enable();
}

spin_lock_irq

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
static inline void spin_lock_irq(spinlock_t *lock)
{
	raw_spin_lock_irq(&lock->rlock);
}

#define raw_spin_lock_irq(lock)		_raw_spin_lock_irq(lock)

#ifndef CONFIG_INLINE_SPIN_LOCK_IRQ
void __lockfunc _raw_spin_lock_irq(raw_spinlock_t *lock)
{
	__raw_spin_lock_irq(lock);
}

static inline void __raw_spin_lock_irq(raw_spinlock_t *lock)
{
	local_irq_disable();
	preempt_disable();
	spin_acquire(&lock->dep_map, 0, 0, _RET_IP_);
	LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
}

spin_unlock_irq

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
static inline void spin_unlock_irq(spinlock_t *lock)
{
	raw_spin_unlock_irq(&lock->rlock);
}

#define raw_spin_unlock_irq(lock)	_raw_spin_unlock_irq(lock)

#ifndef CONFIG_INLINE_SPIN_UNLOCK_IRQ
void __lockfunc _raw_spin_unlock_irq(raw_spinlock_t *lock)
{
	__raw_spin_unlock_irq(lock);
}

static inline void __raw_spin_unlock_irq(raw_spinlock_t *lock)
{
	spin_release(&lock->dep_map, 1, _RET_IP_);
	do_raw_spin_unlock(lock);
	local_irq_enable();
	preempt_enable();
}

spin_lock_irqsave

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
#define spin_lock_irqsave(lock, flags)				\
do {								\
	raw_spin_lock_irqsave(spinlock_check(lock), flags);	\
} while (0)

//#if
#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)

#define raw_spin_lock_irqsave(lock, flags)			\
	do {						\
		typecheck(unsigned long, flags);	\
		flags = _raw_spin_lock_irqsave(lock);	\
	} while (0)

//#else
#else
#define raw_spin_lock_irqsave(lock, flags)		\
	do {						\
		typecheck(unsigned long, flags);	\
		_raw_spin_lock_irqsave(lock, flags);	\
	} while (0)
//#endif

#define _raw_spin_lock_irqsave(lock, flags)	__LOCK_IRQSAVE(lock, flags)

#ifdef CONFIG_INLINE_SPIN_LOCK_IRQSAVE
#define _raw_spin_lock_irqsave(lock) __raw_spin_lock_irqsave(lock)
#endif

//#if
#if !defined(CONFIG_GENERIC_LOCKBREAK) || defined(CONFIG_DEBUG_LOCK_ALLOC)
static inline unsigned long __raw_spin_lock_irqsave(raw_spinlock_t *lock)
{
	unsigned long flags;

	local_irq_save(flags);
	preempt_disable();
	spin_acquire(&lock->dep_map, 0, 0, _RET_IP_);
	/*
	 * On lockdep we dont want the hand-coded irq-enable of
	 * do_raw_spin_lock_flags() code, because lockdep assumes
	 * that interrupts are not re-enabled during lock-acquire:
	 */
#ifdef CONFIG_LOCKDEP
	LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
#else
	do_raw_spin_lock_flags(lock, &flags);
#endif
	return flags;
}  
...
#endif

spin_unlock_irqrestore

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
static inline void spin_unlock_irqrestore(spinlock_t *lock, unsigned long flags)
{
	raw_spin_unlock_irqrestore(&lock->rlock, flags);
}

#define raw_spin_unlock_irqrestore(lock, flags)		\
	do {							\
		typecheck(unsigned long, flags);		\
		_raw_spin_unlock_irqrestore(lock, flags);	\
	} while (0)

#define _raw_spin_unlock_irqrestore(lock, flags) \
					__UNLOCK_IRQRESTORE(lock, flags)

#ifndef CONFIG_INLINE_SPIN_UNLOCK_IRQRESTORE
void __lockfunc _raw_spin_unlock_irqrestore(raw_spinlock_t *lock, unsigned long flags)
{
	__raw_spin_unlock_irqrestore(lock, flags);
}

static inline void __raw_spin_unlock_irqrestore(raw_spinlock_t *lock,
					    unsigned long flags)
{
	spin_release(&lock->dep_map, 1, _RET_IP_);
	do_raw_spin_unlock(lock);
	local_irq_restore(flags);
	preempt_enable();
}

总结

  • spinlock比较底层，跟踪代码有多种实现不好界定，目前使用的比较少，就不继续深究了
  • spinlock可以禁用内核抢占，也可以禁用中断
  • spinlock底层通常用TSL实现
  • spinlock对多核cpu有效(TSL)
  • uni core不适用spinlock,需要等待时间片用尽然后等待其他进程释放锁，代价高
  • spinlock是busy waiting,没有进程切换的代价，但是占用cpu,所以加锁时间不宜过长

ending

80546350_p0.jpg

----------- ending -----------