读写锁实际是一种特殊的自旋锁，它把对共享资源的访问者划分成读者和写者，读者只对共享资源进行读访问，写者则需要对共享资源进行写操作。这种锁相对于自旋锁而言，能提高并发性，因为在多处理器系统中，它允许同时有多个读者来访问共享资源，最大可能的读者数为实际的逻辑CPU数。写者是排他性的，一个读写锁同时只能有一个写者或多个读者（与CPU数相关），但不能同时既有读者又有写者。
数据结构
rwlock_t
1
2
3
4
5
6
7
8
9
10
11
12
13
| typedef struct {
arch_rwlock_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
} rwlock_t;
|
file: include/linux/rwlock_types.h
arch_rwlock_t
MIPS
1
2
3
| typedef struct {
volatile unsigned int lock;
} arch_rwlock_t;
|
file: arch/mips/include/asm/spinlock_types.h
通用接口API
1
2
3
4
5
6
7
8
9
| #define __ARCH_RW_LOCK_UNLOCKED { 0 }
#define __RW_LOCK_UNLOCKED(lockname) \
(rwlock_t) { .raw_lock = __ARCH_RW_LOCK_UNLOCKED, \
RW_DEP_MAP_INIT(lockname) }
# define rwlock_init(lock) \
do { *(lock) = __RW_LOCK_UNLOCKED(lock); } while (0)
|
初始化rwlock_t->raw_lock->lock=0
1
2
| #define write_lock(lock) _raw_write_lock(lock)
#define read_lock(lock) _raw_read_lock(lock)
|
file: include/linux/rwlock.h
实现
读写锁包括读取锁和写入锁，多个读线程可以同时访问共享数据；写线程必须等待所有读线程都释放锁以后，才能取得锁；同样的，读线程必须等待写线程释放锁后，才能取得锁；
也就是说读写锁要确保的是如下互斥关系：可以同时读，但是读-写，写-写都是互斥的；
1
2
3
4
5
6
| read_lock
|->_raw_read_lock
|->__raw_read_lock
|->preempt_disable();
|->rwlock_acquire_read(&lock->dep_map, 0, 0, _RET_IP_);
|->LOCK_CONTENDED(lock, do_raw_read_trylock, do_raw_read_lock);
|
在读锁上锁时与spinlock的流程基本相同，都会关闭内核抢占，因此读写锁中也不能睡眠
1
2
| # define do_raw_read_trylock(rwlock) arch_read_trylock(&(rwlock)->raw_lock)
# define do_raw_read_lock(rwlock) do {__acquire(lock); arch_read_lock(&(rwlock)->raw_lock); } while (0)
|
file: include/linux/rwlock.h
MIPS
不同架构的实现：
arch_read_lock
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
| static inline void arch_read_lock(arch_rwlock_t *rw)
{
unsigned int tmp;
do {
__asm__ __volatile__(
"1: ll %1, %2 # arch_read_lock \n"
" bltz %1, 1b \n"
" addu %1, 1 \n"
"2: sc %1, %0 \n"
: "=m" (rw->lock), "=&r" (tmp)
: "m" (rw->lock)
: "memory");
} while (unlikely(!tmp));
smp_llsc_mb();
}
|
bltz $s,offset <==> if($s< 0) jump(offset « 2); 小于0，跳转
内嵌汇编源码：
1
2
3
4
5
6
| c2020000 ll v0,0(s0)
0440fffe bltz v0,8021c8f8 <do_raw_read_lock+0x38>
00000000 nop
24420001 addiu v0,v0,1
e2020000 sc v0,0(s0)
1040fffa beqz v0,8021c8f8 <do_raw_read_lock+0x38>
|
上读锁是通过原子操作对rwlock_t->raw_lock->lock += 1
arch_read_unlock
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
| static inline void arch_read_unlock(arch_rwlock_t *rw)
{
unsigned int tmp;
smp_mb__before_llsc();
do {
__asm__ __volatile__(
"1: ll %1, %2 # arch_read_unlock \n"
" sub %1, 1 \n"
" sc %1, %0 \n"
: "=m" (rw->lock), "=&r" (tmp)
: "m" (rw->lock)
: "memory");
} while (unlikely(!tmp));
}
|
解读锁是通过原子操作对rwlock_t->raw_lock->lock -= 1
arch_write_lock
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
| static inline void arch_write_lock(arch_rwlock_t *rw)
{
unsigned int tmp;
do {
__asm__ __volatile__(
"1: ll %1, %2 # arch_write_lock \n"
" bnez %1, 1b \n"
" lui %1, 0x8000 \n"
"2: sc %1, %0 \n"
: "=m" (rw->lock), "=&r" (tmp)
: "m" (rw->lock)
: "memory");
} while (unlikely(!tmp));
smp_llsc_mb();
}
|
lui $t,imm <==> $t=(imm « 16)
将立即数左移16位，低16位补零
内嵌汇编反汇编源码：
1
2
3
4
5
6
| c2020000 ll v0,0(s0)
1440fffe bnez v0,8021ca08 <do_raw_write_lock+0x70>
00000000 nop
3c028000 lui v0,0x8000
e2020000 sc v0,0(s0)
1040fffa beqz v0,8021ca08 <do_raw_write_lock+0x70>
|
arch_write_unlock
1
2
3
4
5
6
7
8
9
10
11
| static inline void arch_write_unlock(arch_rwlock_t *rw)
{
smp_mb();
__asm__ __volatile__(
" # arch_write_unlock \n"
" sw $0, %0 \n"
: "=m" (rw->lock)
: "m" (rw->lock)
: "memory");
}
|
Q&A
读锁怎么实现可以存在多个读者进行处理？？
读写锁与SMP多核之间的线程处理关系？？
参考
- Linux下写者优先的读写锁的设计
- 读/写自旋锁