使用LDREX和STREX在多处理器和共享内存系统中实现进程间通信

出于性能的考虑，请使对应的LDREX和STREX指令之间的指令数量最少。

举例：spin_lock源码（arm架构）

static inline void arch_spin_lock(arch_spinlock_t *lock)
{
	unsigned long tmp;
	u32 newval;
	arch_spinlock_t lockval;

	prefetchw(&lock->slock);   /* ldrex和strex都是arm架构的独占读写指令. */
	__asm__ __volatile__(      /* ldrex:从内存加载数据;strex:一定条件下向内存存储数据 */
"1:	ldrex	%0, [%3]\n"        /* lockval = lock->slock(lock是个union,占4字节。也是next(高16位)和owner(低16位)) 同时标记对lock->slock内存区域的独占访问 */
"	add	%1, %0, %4\n"          /* newval = next + (1 << 16);相当于next+1  */
"	strex	%2, %1, [%3]\n"    /* 如果lock->slock内存区域被标记独占访问,则将newval的值更新到lock->slock,且tmp=0。没被标记,则不执行,tmp=1 */                                       
"	teq	%2, #0\n"              /* 如果tmp不为0,则跳转到1处继续执行。确保next+1 */
"	bne	1b"
	: "=&r" (lockval), "=&r" (newval), "=&r" (tmp)
	: "r" (&lock->slock), "I" (1 << TICKET_SHIFT)
	: "cc");

	while (lockval.tickets.next != lockval.tickets.owner) {
		wfe();
		lockval.tickets.owner = ACCESS_ONCE(lock->tickets.owner);
	}

	smp_mb();
}