Split View: [운영체제] 08. 교착 상태: 예방, 회피, 탐지, 복구

[Everyday Analogy of Deadlock]

Two cars meeting on a narrow bridge from opposite sides:
  <--- Car A      Car B --->
        [Bridge]
Both wait for the other to back off -> Deadlock

In a system:
Thread 1: lock(A) -> attempts lock(B) (waits)
Thread 2: lock(B) -> attempts lock(A) (waits)

// Code example that causes deadlock
#include <pthread.h>

pthread_mutex_t mutex_a = PTHREAD_MUTEX_INITIALIZER;
pthread_mutex_t mutex_b = PTHREAD_MUTEX_INITIALIZER;

void *thread1(void *arg) {
    pthread_mutex_lock(&mutex_a);    // Acquire A
    sleep(1);                         // Timing issue
    pthread_mutex_lock(&mutex_b);    // Wait for B (deadlock!)
    // ...
    pthread_mutex_unlock(&mutex_b);
    pthread_mutex_unlock(&mutex_a);
    return NULL;
}

void *thread2(void *arg) {
    pthread_mutex_lock(&mutex_b);    // Acquire B
    sleep(1);                         // Timing issue
    pthread_mutex_lock(&mutex_a);    // Wait for A (deadlock!)
    // ...
    pthread_mutex_unlock(&mutex_a);
    pthread_mutex_unlock(&mutex_b);
    return NULL;
}

[Resource-Allocation Graph Elements]

Process: Circle (O)
Resource type: Dots inside a rectangle (instances)
Request edge: Pi -> Rj  (process requests resource)
Assignment edge: Rj -> Pi  (resource assigned to process)

[Resource-Allocation Graph with Deadlock]

    P1 --request--> R1 (1) --assigned--> P2
    ^                                     |
    |                                     |
  assigned                             request
    |                                     |
    R2 (1) <--request-- P3 <--assigned-- R3 (1)

P1 requests R1 (held by P2)
P2 requests R3 (held by P3)
P3 requests R2 (held by P1)
-> Cycle exists -> Deadlock!

[Resource-Allocation Graph without Deadlock - Multiple Instances]

    P1 --request--> R1 (2)  Instance1 --assigned--> P2
                             Instance2 --assigned--> P3

When P2 or P3 releases R1, P1 can proceed
-> Cycle may exist but no deadlock
   (when resource has 2+ instances)

Method A: Request all needed resources at once before execution
  Pros: Simple
  Cons: Low resource utilization, possible starvation

Method B: Release all held resources before requesting new ones
  Pros: Flexible
  Cons: Complex implementation

Process P holds resource A and requests B:
  If B is unavailable:
    Forcibly release A (preempt)
    Put P in waiting state for both A and B
    Restart when both A and B are available

// Circular wait prevention: Resource ordering protocol
// Resource numbers: mutex_a = 1, mutex_b = 2, mutex_c = 3

// Correct usage: Always acquire in ascending order
void safe_function() {
    pthread_mutex_lock(&mutex_a);    // Number 1
    pthread_mutex_lock(&mutex_b);    // Number 2
    pthread_mutex_lock(&mutex_c);    // Number 3

    // Critical section

    pthread_mutex_unlock(&mutex_c);
    pthread_mutex_unlock(&mutex_b);
    pthread_mutex_unlock(&mutex_a);
}

// Incorrect usage: Requesting in reverse order risks deadlock
void unsafe_function() {
    pthread_mutex_lock(&mutex_c);    // Number 3
    pthread_mutex_lock(&mutex_a);    // Number 1 (order violation!)
}

[Relationship between Safe State and Deadlock]

+-------------------------------------------+
|              All States                    |
|  +----------------------------------+     |
|  |          Safe States              |     |
|  |                                  |     |
|  |                                  |     |
|  +----------------------------------+     |
|            Unsafe States                  |
|       (includes deadlock region)          |
|                 +------+                  |
|                 |Dead- |                  |
|                 |lock  |                  |
|                 +------+                  |
+-------------------------------------------+

Safe state -> No deadlock (guaranteed)
Unsafe state -> Deadlock possible (not certain)

Available[m]:     Number of available instances per resource type
Max[n][m]:        Maximum demand per process
Allocation[n][m]: Current allocation per process
Need[n][m]:       Remaining need per process (Max - Allocation)

[Safety Check Algorithm]

1. Work = copy of Available
   Finish[i] = false (for all i)

2. Find i such that:
   Finish[i] == false AND Need[i] <= Work

3. If found:
   Work = Work + Allocation[i]
   Finish[i] = true
   Go back to step 2

4. If not found:
   If all Finish[i] are true -> Safe state
   Otherwise -> Unsafe state

[Banker's Algorithm Example]

Resources: A(10), B(5), C(7)

        Allocation    Max        Need       Available
        A  B  C      A  B  C    A  B  C    A  B  C
P0      0  1  0      7  5  3    7  4  3    3  3  2
P1      2  0  0      3  2  2    1  2  2
P2      3  0  2      9  0  2    6  0  0
P3      2  1  1      2  2  2    0  1  1
P4      0  0  2      4  3  3    4  3  1

Finding safe sequence:
1. Work = [3,3,2]
   P1's Need[1,2,2] <= [3,3,2]? Yes!
   Work = [3,3,2] + [2,0,0] = [5,3,2], Finish[P1] = true

2. Work = [5,3,2]
   P3's Need[0,1,1] <= [5,3,2]? Yes!
   Work = [5,3,2] + [2,1,1] = [7,4,3], Finish[P3] = true

3. Work = [7,4,3]
   P4's Need[4,3,1] <= [7,4,3]? Yes!
   Work = [7,4,3] + [0,0,2] = [7,4,5], Finish[P4] = true

4. Work = [7,4,5]
   P2's Need[6,0,0] <= [7,4,5]? Yes!
   Work = [7,4,5] + [3,0,2] = [10,4,7], Finish[P2] = true

5. Work = [10,4,7]
   P0's Need[7,4,3] <= [10,4,7]? Yes!
   Finish[P0] = true

Safe sequence: P1 -> P3 -> P4 -> P2 -> P0
The system is currently in a safe state!

[When P1 additionally requests resources (1,0,2)]

1. Request[1] = [1,0,2] <= Need[1] = [1,2,2]? Yes (valid request)
2. Request[1] = [1,0,2] <= Available = [3,3,2]? Yes (resources exist)
3. Hypothetical allocation:
   Available = [3,3,2] - [1,0,2] = [2,3,0]
   Allocation[1] = [2,0,0] + [1,0,2] = [3,0,2]
   Need[1] = [1,2,2] - [1,0,2] = [0,2,0]

4. Run safety check:
   Check if safe sequence exists with new Available = [2,3,0]
   -> Allow request if safe, deny if unsafe

// Banker's Algorithm Java Implementation
public class BankersAlgorithm {
    private int[][] max;
    private int[][] allocation;
    private int[][] need;
    private int[] available;
    private int numProcesses;
    private int numResources;

    public BankersAlgorithm(int[][] max, int[][] alloc,
                             int[] avail) {
        this.max = max;
        this.allocation = alloc;
        this.available = avail;
        this.numProcesses = max.length;
        this.numResources = avail.length;
        this.need = new int[numProcesses][numResources];

        for (int i = 0; i < numProcesses; i++)
            for (int j = 0; j < numResources; j++)
                need[i][j] = max[i][j] - allocation[i][j];
    }

    public boolean isSafe() {
        int[] work = available.clone();
        boolean[] finish = new boolean[numProcesses];
        int count = 0;

        while (count < numProcesses) {
            boolean found = false;
            for (int i = 0; i < numProcesses; i++) {
                if (!finish[i] && canAllocate(need[i], work)) {
                    // Reclaim resources
                    for (int j = 0; j < numResources; j++)
                        work[j] += allocation[i][j];
                    finish[i] = true;
                    count++;
                    found = true;
                    System.out.println("P" + i + " can execute");
                }
            }
            if (!found) return false;
        }
        return true;
    }

    private boolean canAllocate(int[] need, int[] work) {
        for (int j = 0; j < numResources; j++)
            if (need[j] > work[j]) return false;
        return true;
    }
}

[Wait-For Graph]

Remove resource nodes from the resource-allocation graph
and show only direct dependencies between processes

P1 --> P2 --> P3
 ^            |
 |            v
 +---- P4 <--+

Cycle exists: P1->P2->P3->P4->P1 -> Deadlock!

[Deadlock Detection Algorithm]

1. Work = Available
   If Allocation[i] != 0 then Finish[i] = false
   If Allocation[i] == 0 then Finish[i] = true

2. Find i where Finish[i] == false AND Request[i] <= Work

3. If found: Work = Work + Allocation[i]
             Finish[i] = true
             Go to step 2

4. If Finish[i] == false for any process
   -> Those processes are in deadlock

[Resource Preemption Considerations]

1. Victim selection: Which process to take resources from?
   Select based on minimizing cost

2. Rollback: Roll back the victimized process to a safe state
   - Total rollback: Restart from the beginning
   - Partial rollback: Roll back just enough to break the deadlock

3. Starvation prevention: Ensure the same process is not always the victim
   Include rollback count in cost factors

// Deadlock prevention using trylock
int try_acquire_both(pthread_mutex_t *a, pthread_mutex_t *b) {
    while (1) {
        pthread_mutex_lock(a);

        if (pthread_mutex_trylock(b) == 0) {
            return 0;  // Success: both locks acquired
        }

        // If b cannot be obtained, release a and retry
        pthread_mutex_unlock(a);
        usleep(rand() % 1000);  // Wait briefly then retry
    }
}

// Deadlock prevention using Java's tryLock
import java.util.concurrent.locks.ReentrantLock;
import java.util.concurrent.TimeUnit;

public class DeadlockFree {
    private final ReentrantLock lockA = new ReentrantLock();
    private final ReentrantLock lockB = new ReentrantLock();

    public void safeMethod() throws InterruptedException {
        while (true) {
            boolean gotA = lockA.tryLock(100, TimeUnit.MILLISECONDS);
            boolean gotB = lockB.tryLock(100, TimeUnit.MILLISECONDS);

            if (gotA && gotB) {
                try {
                    // Both locks acquired - work safely
                    System.out.println("Both resources acquired!");
                    break;
                } finally {
                    lockA.unlock();
                    lockB.unlock();
                }
            }

            // If either failed, release held lock
            if (gotA) lockA.unlock();
            if (gotB) lockB.unlock();

            // Random wait then retry (prevent livelock)
            Thread.sleep((long)(Math.random() * 100));
        }
    }
}