Galvin Edition 9 Exercise 5 Question 20 (Page No. 246-247)

Question

Consider the code example for allocating and releasing processes shown below:

$#define MAX PROCESSES 255
int number of processes = 0;
/* the implementation of fork() calls this function */
int allocate process() {
int new pid;
if (number of processes == MAX PROCESSES)
return -1;
else {
/* allocate necessary process resources */
++number of processes;
return new pid;
}
}
/* the implementation of exit() calls this function */
void release process() {
/* release process resources */
--number of processes;
}$

a. Identify the race condition(s).
b. Assume you have a mutex lock named mutex with the operations $acquire()$ and $release()$. Indicate where the locking needs to be placed to prevent the race condition(s).
c. Could we replace the integer variable int number of processes = 0 with the atomic integer atomic t number of processes = 0 to prevent the race condition(s)?

Comments

a) Race condition would come when any process tries to run statements

$\textbf{ ++number of processes;  --number of processes;  }$

In assemble language this will be implemented in 3 different steps

// For increment
1. Read temp, number of processes
2. ADD temp, 1
3. Store number of processes, temp


// For decrement
1. Read temp, number of processes
2. SUB temp, 1
3. Store number of processes, temp

More than one process can execute the above code in interleaved manner, resulting in race condition

 

b) To prevent race condition

acquire() lock will be used when process tries to perform $\textbf{ ++number of processes;}$

similarly release() locak will be used when process tries to perform $\textbf{ --number of processes;}$

 

c) Yes, atomic integer variable can be used to prevent race condition where at a time only one process can increment or decrement the variable number of process.