[RTT例程练习] 1.6 线程优先级反转原理

优先级反转是不希望出现的现象。

简单来说，有三个线程A, B, C， 优先级A>B>C，A和C有共享资源S。A.B挂起，C运行且有S，A没有S所以一直挂起，B进入调度，调度完之后C再运行直到释放S之后，A才能运行。

可以看出，A虽然优先级最高，但实际运行却是最后一个运行，其实际的优先级反转了。

下面这段来自官网：

三个线程的优先级顺序是 thread2 > worker > thread1，首先thread2得到执行，它得到信号量，并且释放，然后延时等待，然后worker线程得到处理器控制权开始运行，它也进行了延时操作，然后，thread1拿到了控制权，并且它申请得到了信号量，接着进行了打印操作，在它打印结束进行延时操作时，由于worker的优先级高于thread1，worker重新获得了控制，由于它并不需要信号量来完成下面的操作，于是很顺利的它把自己的一大串打印任务都执行完成了，纵然thread2的优先级要高于它，但是奈何获取不到信号量，什么也干不了，只能被阻塞而干等，于是实验原理中提到的那一幕便发生了。worker执行结束后，执行权回到了握有信号量的thread1手中，当它完成自己的操作，并且释放信号量后，优先级最高的thread2才能继续执行。

对应程序：

#include <rtthread.h>

static rt_sem_t sem;
static rt_uint8_t t1_count, t2_count, worker_count;
static rt_thread_t t1, t2, worker;

static void thread1_entry(void *parameter)
{
rt_err_t result;

result = rt_sem_take(sem, RT_WAITING_FOREVER);

for (t1_count = 0; t1_count < 10; t1_count++)
{
rt_kprintf("thread1: got semaphore, count: %d\n", t1_count);
rt_thread_delay(RT_TICK_PER_SECOND);
}

rt_kprintf("thread1: release semaphore\n");
rt_sem_release(sem);
}

static void thread2_entry(void *parameter)
{
rt_err_t result;

while (1)
{
result = rt_sem_take(sem, RT_WAITING_FOREVER);
rt_kprintf("thread2: got semaphore\n");
if (result != RT_EOK)
return;
rt_kprintf("thread2: release semaphore\n");
rt_sem_release(sem);

rt_thread_delay(90);
result = rt_sem_take(sem, RT_WAITING_FOREVER);
t2_count++;
rt_kprintf("thread2: got semaphore, count: %d\n", t2_count);
}
}

static void worker_thread_entry(void *parameter)
{
rt_thread_delay(5);
for (worker_count = 0; worker_count < 10; worker_count++)
{
rt_kprintf("worker: count: %d\n", worker_count);
}
rt_thread_delay(RT_TICK_PER_SECOND);
}

int rt_application_init()
{
sem = rt_sem_create("sem", 1, RT_IPC_FLAG_PRIO);
if (sem ==RT_NULL)
{
return 0;
}

t1_count = t2_count = 0;

t1 = rt_thread_create("t1",
thread1_entry, RT_NULL,
512, 7, 10);
if (t1 != RT_NULL)
rt_thread_startup(t1);

t2 = rt_thread_create("t2",
thread2_entry, RT_NULL,
512, 5, 10);
if (t2 != RT_NULL)
rt_thread_startup(t2);

worker = rt_thread_create("worker",
worker_thread_entry, RT_NULL,
512, 6, 10);
if (worker != RT_NULL)
rt_thread_startup(worker);

return 0;
}
/*@}*/

输出结果为：

thread2: got semaphore
thread2: release semaphore
thread1: got semaphore, count: 0
worker: count: 0
worker: count: 1
worker: count: 2
worker: count: 3
worker: count: 4
worker: count: 5
worker: count: 6
worker: count: 7
worker: count: 8
worker: count: 9
thread1: got semaphore, count: 1
thread1: got semaphore, count: 2
thread1: got semaphore, count: 3
thread1: got semaphore, count: 4
thread1: got semaphore, count: 5
thread1: got semaphore, count: 6
thread1: got semaphore, count: 7
thread1: got semaphore, count: 8
thread1: got semaphore, count: 9
thread1: release semaphore
thread2: got semaphore, count: 1

可以看到thread2 被降到了 thread1 和 worker 之后。