使用两个缓冲区和两个互斥锁进行线程同步:C

时间:2011-04-07 16:53:28

标签: c multithreading synchronization deadlock mutex

我遇到了典型的 Producer& amp;消费者问题,我有一个主线程将要运行的生产者函数和一个多线程调用的消费者函数;从一个缓冲区中取出物品并使用安全的互斥锁将它们放入另一个缓冲区中。我想我需要两个互斥锁来管理两个缓冲区,但我正在一个终极循环中运行。

CODE:

    int add_rule_input(rule_t* rule, rule_node_t* list) {
    int i, error;   
str_node_t* sptr;
rule_node_t* rptr;


if(error = pthread_mutex_lock(&mut_access)){
        return error;
}

error = pthread_cond_wait(&buffer_empty, &mut_access);

//first check to see if dependencies are in the output queue    
for(sptr = rule->deps; sptr != NULL; sptr = sptr->next){
        dep_count++; 
pthread_mutex_lock(&mut_output);
for(i = 0; i < ArraySize; i++){

    if(outputQ[i] != NULL){

if(strcmp(sptr->str, outputQ[i]) == 0){
    array_count++;
                break; // go the next rule in the output q
        }else{
            //means the first element in our array did not have the current
                continue;       
            }   
        }else{
error = pthread_cond_wait(&buffer_empty, &mut_output);
            break;
}
    }
}       
pthread_mutex_unlock(&mut_output);

inputQ[bufin] = rule->target;//the element wherever the current place is
printf("buffer got %s buffin = %d\n\n", inputQ[bufin], bufin);
bufin = (bufin + 1);
totalitems++; 
pthread_cond_signal(&buffer_full);

return pthread_mutex_unlock(&mut_access);

}

我正在访问其他输出缓冲区的消费者功能

static void *consumer(void *arg){

rule_node_t* lptr = (rule_node_t*)arg;
str_node_t* dptr;   
 int error, i, j;
int test1 = 0;
//grab lock to read the input queue
    if(error = pthread_mutex_lock(&mut_access))
    return error;


if(error){

        pthread_mutex_unlock(&mut_access);
        return error;
    }


 // loop through all our rules
    while(lptr != NULL){

// loop through each rules dependencies to compare with item in the input queue
    for(dptr = lptr->rule->deps; dptr != NULL; dptr = dptr->next){
    // now loop through our input q if we get the lock  
        for(j = 0; j > ArraySize; j++){

              if(inputQ[j] != NULL){

        if(strcmp(dptr->str, inputQ[j]) == 0){
            fake_exec(lptr->rule); // if we get here there is a rule that needs to be executed
            pthread_mutex_lock(&mut_output);
            //update the output queue and release the lock

         if(outputQ[bufout] == NULL){
    outputQ[bufout]= lptr->rule->target;
    bufout = (bufout + 1);
    printf("bufout has %s\n", outputQ[bufout]);

    }
    pthread_mutex_unlock(&mut_output); 
}                           
    }   
    }
error = pthread_cond_wait(&buffer_full, &mut_access);   
}
    lptr = lptr->next;
}
 pthread_cond_signal(&buffer_empty);
pthread_mutex_unlock(&mut_access);  


 }

问题: 1}我试图首先抓住我的第一个缓冲区(inputq)并向其添加项目,如果我到达没有更多项目的点,那么我想释放这个锁,以便我的消费者可以从缓冲区并将它们放在outputq中,由于某种原因它似乎主线程不等待并释放锁定?

1 个答案:

答案 0 :(得分:1)

我在这里看到的问题很少。第一个是生成器部分,用于(sptr = rule-&gt; deps; sptr!= NULL; sptr = sptr-&gt; next)循环。在这个循环中你可以锁定mut_output互斥锁,因此可以多次锁定(如果它是递归互斥锁),但是当循环结束时只锁定一次。

另一个问题是pthread_cond_wait(&amp; buffer_empty,&amp; mut_output);.让我们想象一下,制片人的这种等待是如此。当它被推出时,mutex mut_access被生产者锁定,现在当消费者被执行时它尝试获取mut_access,但它不能因为它已经锁定所以消费者等待,当它发出信号buffer_empty条件变量来解锁生产者时更新到达部分。 可能在这个pthread_cond_wait中你想传递mut_access而不是mut_output。