信号量和互斥锁(mutex)的区别:互斥锁只允许一个线程进入临界区,而信号量允许多个线程同时进入临界区。
不多做解释,要使用信号量同步,需要包含头文件semaphore.h。
主要用到的函数:
int sem_init(sem_t *sem, int pshared, unsigned int value);
,其中sem
是要初始化的信号量,pshared
表示此信号量是在进程间共享还是线程间共享,value是信号量的初始值。int sem_destroy(sem_t *sem);
,其中sem
是要销毁的信号量。只有用sem_init
初始化的信号量才能用sem_destroy
销毁。int sem_wait(sem_t *sem);
等待信号量,如果信号量的值大于0,将信号量的值减1,立即返回。如果信号量的值为0,则线程阻塞。相当于P操作。成功返回0,失败返回-1。int sem_post(sem_t *sem);
释放信号量,让信号量的值加1。相当于V操作。
下列的代码演示了如何用信号量同步,模拟一个窗口服务系统。
/* @purpose: 基于信号量的多线程同步,操作系统原理中的P,V操作
* @author: jollywing@foxmail.com
* @create: 2015-03-20 Fri
* */
#include <pthread.h>
#include <semaphore.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
/* @Scene: 某行业营业厅同时只能服务两个顾客。
* 有多个顾客到来,每个顾客如果发现服务窗口已满,就等待,
* 如果有可用的服务窗口,就接受服务。 */
/* 将信号量定义为全局变量,方便多个线程共享 */
sem_t sem;
/* 每个线程要运行的例程 */
void * get_service(void *thread_id)
{
/* 注意:立即保存thread_id的值,因为thread_id是对主线程中循环变量i的引用,它可能马上被修改 */
int customer_id = *((int *)thread_id);
if(sem_wait(&sem) == 0) {
usleep(100); /* service time: 100ms */
printf("customer %d receive service ...\n", customer_id);
sem_post(&sem);
}
}
#define CUSTOMER_NUM 10
int main(int argc, char *argv[])
{
/* 初始化信号量,初始值为2,表示有两个顾客可以同时接收服务 */
/* @prototype: int sem_init(sem_t *sem, int pshared, unsigned int value); */
/* pshared: if pshared == 0, the semaphore is shared among threads of a process
* otherwise the semaphore is shared between processes. */
sem_init(&sem, 0, 2);
/* 为每个顾客定义一个线程id, pthread_t 其实是unsigned long int */
pthread_t customers[CUSTOMER_NUM];
int i, ret;
/* 为每个顾客生成一个线程 */
for(i = 0; i < CUSTOMER_NUM; i++){
int customer_id = i;
ret = pthread_create(&customers[i], NULL, get_service, &customer_id);
if(ret != 0){
perror("pthread_create");
exit(1);
}
else {
printf("Customer %d arrived.\n", i);
}
usleep(10);
}
/* 等待所有顾客的线程结束 */
/* 注意:这地方不能再用i做循环变量,因为可能线程中正在访问i的值 */
int j;
for(j = 0; j < CUSTOMER_NUM; j++) {
pthread_join(customers[j], NULL);
}
/* Only a semaphore that has been initialized by sem_init(3)
* should be destroyed using sem_destroy().*/
sem_destroy(&sem);
return 0;
}
编译:gcc main.c -lpthread
。
运行结果(注意,每次运行都不相同):
Customer 0 arrived.
Customer 1 arrived.
customer 0 receive service ...
Customer 2 arrived.
customer 1 receive service ...
Customer 3 arrived.
customer 2 receive service ...
Customer 4 arrived.
customer 3 receive service ...
Customer 5 arrived.
customer 4 receive service ...
Customer 6 arrived.
customer 5 receive service ...
Customer 7 arrived.
customer 6 receive service ...
Customer 8 arrived.
customer 7 receive service ...
Customer 9 arrived.
customer 8 receive service ...
customer 9 receive service ...