UNIX Shared memory and Semaphores in C

2019-06-09 00:29发布

问题:

I started a week ago understanding and working with semaphores and shared memory, and actually created this program; the problem is I can't find anything wrong with it. I've been looking at it for hours and everything seems correct. The code compiles and i can create the build but when I execute it nothing happens.

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/ipc.h> 
#include <sys/shm.h>
#include <stdio.h>
#include <sys/fcntl.h>
#include <semaphore.h>

#define MAXCHILDS 4
#define MAX_SIZE 10
#define MAX_WRITES 4

typedef struct{
    int m[MAX_SIZE][MAX_SIZE];
} matrix;

/*fork variables*/
pid_t child[MAXCHILDS];
/*semphores variables */
sem_t *empty, *full, * mutex;
/*share memory id*/
int shmid;
/*shared memory array pointer */
matrix * sh_mem;
/*pointer to matrix*/
int **p;

void init(){
      /*create pointer to matrix*/
      p = &sh_mem->m;
     /*semaphores unlink and creation */    
     sem_unlink("EMPTY");
     empty=sem_open("EMPTY",O_CREAT|O_EXCL,0700,MAX_WRITES);
     sem_unlink("FULL");
     full=sem_open("FULL",O_CREAT|O_EXCL,0700,0);
     sem_unlink("MUTEX");
     mutex=sem_open("MUTEX",O_CREAT|O_EXCL,0700,1);
    /*initialize shared memory */
    shmid = shmget(IPC_PRIVATE,sizeof(matrix),IPC_CREAT|0777);
    /*map shared memory*/
    sh_mem = (matrix*)shmat(shmid,NULL,0);
    if(sh_mem== (matrix*)(-1)){
        perror("shmat");
    }
}

void writer(int ** m){
    int i,k;
    for(i = 0;i<MAX_SIZE;i++){
        for(k= 0;k<MAX_SIZE;k++){
            m[i][k] = 0;
        }
    }
}

void reader(int **m){
    int i = 0;
    int k = 0;
    for(i = 0;i<MAX_SIZE;i++){
        for(k= 0;k<MAX_SIZE;k++){
            printf(m[i][k]);
        }
        printf("\n");
    }
}

void terminate() {
  sem_close(empty);
  sem_close(full);
  sem_close(mutex);
  sem_unlink("EMPTY");
  sem_unlink("FULL");
  sem_unlink("MUTEX");
  shmctl(shmid, IPC_RMID, NULL);
}

int main(int argc, char **argv)
{
    int i;
    init();

    for(i = 0;i<MAXCHILDS;i++){
        if ((child[i] = fork()) < 0) // error occured
        {
            perror("Fork Failed");
            exit(1);
        }
        if ((child[i] = fork())==0){
            writer(sh_mem->m);
            exit(0);
        }
    }
    /*father*/  
    reader(sh_mem->m);
    wait(NULL);

    terminate();

    return 0;
}

The children are supposed to write the the matrix in shared memory, and the father is supposed to read the shared memory array and the print the matrix. Can you help me with this? Thanks for the help ...

回答1:

The primary error here is that reader and writer take a different type of argument than you're passing to them, as gcc -Wall points out:

test.c: In function ‘main’:
test.c:92:13: warning: passing argument 1 of ‘writer’ from incompatible pointer type [enabled by default]
test.c:49:6: note: expected ‘int **’ but argument is of type ‘int (*)[10]’
test.c:97:5: warning: passing argument 1 of ‘reader’ from incompatible pointer type [enabled by default]
test.c:58:6: note: expected ‘int **’ but argument is of type ‘int (*)[10]’

As provided, the program segfaulted in the parent and every child. When I changed the parameter type of reader and writer from int **m to int m[MAX_SIZE][MAX_SIZE] (along with the fixes below), the program ran successfully, as far as I can tell.

There are a number of other errors:

  • You need to #include <sys/wait.h>.
  • The global int **p isn't used and its initialization has the same type error as the reader and writer functions did.
  • The printf call in reader needs a format string; I used "%d ".
  • As Jonathan Leffler pointed out, you need to call fork() only once each time through the loop in main.

All but the last of those were highlighted by compiler warnings.

In studying why this program was failing, I also used strace -f to identify which syscalls and processes were actually busted. The semaphore-related syscalls, for example, appear to be returning successfully--although as Jonathan pointed out, you should check their return values for errors, because failing as early as possible makes it much easier to debug problems.