[Mutex] Implement a mutex

[trylimits]

We should implement a mutex which block processes from accessing critical sections in parallel.

The implemented mutex has to interact with the scheduler or process manager to block the process which accesses a locked mutex. The interface should be as follows:

MutexLock(&myMutex);     // If myMutex is not locked atm the process may pass, otherwise it should be blocked
[...] Critical section

MutexUnlock(&myMutex); // Previously blocked processes should be unblocked, i.e. state ready

[trylimits]

I have implemented a simple mutex in src/processmanager/mutex.h/c

Usage as suggested in OP.

Test which was used to verify functionality:

void simpleProc1(int argc, char** argv)
{
    MutexLock(&mutex);
    volatile int j = 0;
    for (j = 0; j < 10; j++)
    {
        volatile int i;
        printf("Process 1\n");
        for(i = 0; i < 0x0200000; i++);
    }
    MutexUnlock(&mutex);
}
void simpleProc2(int argc, char** argv)
{
    MutexLock(&mutex);
    while(1)
    {
        volatile int i;
        printf("---------Process 2\n");
        for(i = 0; i < 0x0200000; i++);
    }
    MutexUnlock(&mutex);
}

int main()
{
     [...]
        ProcessManagerStartProcess("sp1", &simpleProc1);
    ProcessManagerStartProcess("sp2", &simpleProc2);
     [...]
}

Output:

dabt interrupt from pid=0, with fault state=22
dabt interrupt from pid=0, with fault state=5
Process 1
dabt interrupt from pid=1, with fault state=5
Process 1
Process 1
Process 1
Process 1
Process 1
Process 1
Process 1
Process 1
Process 1
---------Process 2
---------Process 2
---------Process 2
[...]

[trylimits]

With above implementation there is still one issue/question open:

What should happen if a process is force killed (e.g. because of NPE) while the process has locked a mutex. Should the mutex be unlocked? Otherwise all blocked processes will be stuck forever. @mpe5651 @Blackjack92

[ghost]

I think in this case we should unlock the mutex. otherwise the blocked processes are in a deadlock. I'll do some research for a possibly better solution.

[trylimits]

@mpe5651 I agree, but imo this is a little bit hard to implement. If a process gets force killed there has to be an instance which manages all used mutex (or mutexes? :/) and it's owner in order to be able to release the mutex. However, a mutex should be a "standalone" component and should not be registered somewhere else.

[wake-0]

@trylimits you are right. Which component contains the mutex at the moment?

[trylimits]

See implementation at src/processmanager/mutex.c. For the moment the mutex works properly so shouldn't we mark this as low priority?

[wake-0]

We should implement the suggested change from comment @mpe5651.

[trylimits]

According to coaching we can assume that the critical section's implementation is bug free. We should not investigate this case further, unless we introduce a ResourceManager in the future, which we probably don't do :)