mutex never fails

[oxpac]

---

Hi,

I have tried in many ways to get the

printf("lock failed in 1\n"); or
printf("lock failed in 2\n");

atleast once, but i never could do it.

Plese help me in achieving that.

#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <unistd.h>

pthread_mutex_t mutex= PTHREAD_MUTEX_INITIALIZER;

void *fun1(void* );
void *fun2(void* );

int arg = 20;

int main(  void ) {
        pthread_t tid1,tid2;
        pthread_attr_t attr;
        pthread_attr_init(&attr);
        printf(" arg = %d\n", arg);

        pthread_create(&tid1, NULL, fun1,(void*)1);
        pthread_join(tid1,NULL);
        printf(" 1arg = %d\n", arg);
        pthread_create(&tid2, NULL, fun2,(void*)2);
        pthread_join(tid2,NULL);
        printf(" 2arg = %d\n", arg);
        printf("end\n");
        pthread_exit(NULL);

return 0;
}

void * fun1( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        if (-1 != pthread_mutex_trylock(&mutex))
        {
        printf("locked in 1\n");
        if (arg == 20)
                arg = 10;
        sleep(15);
        pthread_mutex_unlock(&mutex);

        } else {
        printf("lock failed in 1\n");
        }
        pthread_exit(NULL);
}

void* fun2( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        if (-1 != pthread_mutex_trylock(&mutex))
        {
        printf("locked in 2\n");
        if (arg == 10)
                arg = 30;
        pthread_mutex_unlock(&mutex);
        } else {
        printf("lock failed in 2\n");
        }
        pthread_exit(NULL);
}

[Rubberman]

You aren't testing the lock correctly. Here is some information from the man page for the pthread_mutex_lock() functions:

RETURN VALUE
       If  successful,  the pthread_mutex_lock() and pthread_mutex_unlock() functions shall return zero; otherwise, an
       error number shall be returned to indicate the error.

       The pthread_mutex_trylock() function shall return zero if a lock on the mutex object  referenced  by  mutex  is
       acquired. Otherwise, an error number is returned to indicate the error.

IE, you need to try this instead:

void * fun1( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        if (0 == pthread_mutex_trylock(&mutex))
        {
        printf("locked in 1\n");
        if (arg == 20)
                arg = 10;
        sleep(15);
        pthread_mutex_unlock(&mutex);

        } else {
        printf("lock failed in 1\n");
        }
        pthread_exit(NULL);
}

[oxpac]

You aren't testing the lock correctly. Here is some information from the man page for the pthread_mutex_lock() functions:

RETURN VALUE
       If  successful,  the pthread_mutex_lock() and pthread_mutex_unlock() functions shall return zero; otherwise, an
       error number shall be returned to indicate the error.

       The pthread_mutex_trylock() function shall return zero if a lock on the mutex object  referenced  by  mutex  is
       acquired. Otherwise, an error number is returned to indicate the error.

IE, you need to try this instead:

void * fun1( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        if (0 == pthread_mutex_trylock(&mutex))
        {
        printf("locked in 1\n");
        if (arg == 20)
                arg = 10;
        sleep(15);
        pthread_mutex_unlock(&mutex);

        } else {
        printf("lock failed in 1\n");
        }
        pthread_exit(NULL);
}

I tried by changing as you suggested , but the result is same.


arg = 20
thrd 1 schded
locked in 1
1arg = 10
thrd 2 schded
locked in 2
2arg = 30
end


i think it is some thing to do with the scheduling.
i dont know what kind of scheduling is used in my system
once i printed the scheduling policy it gave me "OTHER".

so i guess the system is executing one thread after the other.

can somebody point out where exactly the real problem is?

Thanks in advance.

[Rubberman]

The wrong logic in the fun1() and fun2() code notwithstanding, your problem is likely a scheduling one as you surmise. Try a short sleep before you try to take the lock in fun2() (shorter than the 15 second sleep in fun1()) so you know that fun1() has established and is holding the lock.

[oxpac]

       
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <unistd.h>

pthread_mutex_t mutex= PTHREAD_MUTEX_INITIALIZER;

void *fun1(void* );
void *fun2(void* );

int arg = 20;

int main(  void ) {
        pthread_t tid1,tid2;
        pthread_attr_t attr;
        pthread_attr_init(&attr);
        printf(" arg = %d\n", arg);
        pthread_create(&tid1, NULL, fun1,(void*)1);
        pthread_join(tid1,NULL);
        printf(" 1arg = %d\n", arg);
        pthread_create(&tid2, NULL, fun2,(void*)2);
        pthread_join(tid2,NULL);
        printf(" 2arg = %d\n", arg);
        printf("end\n");
        pthread_exit(NULL);

return 0;
}

void * fun1( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        if (0 == pthread_mutex_trylock(&mutex))
        {
        printf("locked in 1\n");
        if (arg == 20)
                arg = 10;
        sleep(15);
        pthread_mutex_unlock(&mutex);

        } else {
        printf("lock failed in 1\n");
        }
        pthread_exit(NULL);
}

void* fun2( void *p)
{
        int i;
        printf(" thrd %ld schded \n",(long) p);
        sleep(2);
        if (0 == pthread_mutex_trylock(&mutex))
        {
        printf("locked in 2\n");
        if (arg == 10)
                arg = 30;
        pthread_mutex_unlock(&mutex);
        } else {
        printf("lock failed in 2\n");
        }
        pthread_exit(NULL);
}

i modified , but the result is same...

dont get whats wrong exaclty..

can i change the scheduling policy, will it effect the system in any other wayy.

[Rubberman]

Well, it seems to be working as though pthread_mutex_lock() was called instead of trylock() - at least that's how it behaves on my RHEL 6 system. It may be that in our systems, the trylock() version is not yet implemented. There is no notice of that in the man pages, so I will have to do some more study.

[oxpac]

But we are not getting any undefined symbol errors here.
how to check a particular system call is implemented in the system or not?

[Rubberman]

I haven't looked at the source yet since it was late last night when I made my previous reply (and I am engaged this evening as well), but I don't think it's a matter of implementation as much a matter of how you are setting up your environment and using the function. It may be that on the 2.6.24 kernel, it is not using the sys_msgctl() function, or are setting it up differently than you are. A full listing of your kernel module source code, and how your user-side functions are accessing it, would be helpful in my analysis.

[oxpac]

I haven't looked at the source yet since it was late last night when I made my previous reply (and I am engaged this evening as well), but I don't think it's a matter of implementation as much a matter of how you are setting up your environment and using the function. It may be that on the 2.6.24 kernel, it is not using the sys_msgctl() function, or are setting it up differently than you are. A full listing of your kernel module source code, and how your user-side functions are accessing it, would be helpful in my analysis.

this is my system info.

skylab@skylab:~$ uname -r
3.0.0-22-generic-pae
skylab@skylab:~$ uname -v
#36-Ubuntu SMP Tue Jun 12 17:31:49 UTC 2012

[Rubberman]

I haven't looked at the source yet since it was late last night when I made my previous reply (and I am engaged this evening as well), but I don't think it's a matter of implementation as much a matter of how you are setting up your environment and using the function. It may be that on the 2.6.24 kernel, it is not using the sys_msgctl() function, or are setting it up differently than you are. A full listing of your kernel module source code, and how your user-side functions are accessing it, would be helpful in my analysis.

Oops. I meant to post this reply on another thread... Doh! Sorry...