Deadlock Detection & Recovery Strategies

Explanation

The OS allows deadlock to occur, periodically invokes detection, and breaks the deadlock via preemption or aborts.

Key points

• Multiple instance detection uses Available, Allocation, and Request matrices
• Recovery Option 1: Process termination (abort all or abort one-by-one by lowest cost/priority)
• Recovery Option 2: Resource preemption with checkpoint rollback; must guard against starvation

Example

Code examples

// Deadlock Recovery: Select Victim Process with Minimum Penalty Cost
int selectVictimProcess(const vector<int>& cpuTimeSpent, const vector<int>& priority) {
    int victim = -1, minScore = 1e9;
    for (size_t i = 0; i < cpuTimeSpent.size(); i++) {
        int penalty = cpuTimeSpent[i] + priority[i] * 10;
        if (penalty < minScore) { minScore = penalty; victim = i; }
    }
    return victim;
}

Exam questions

• Describe the methods used for recovering from a deadlock situation.
• What is starvation in deadlock preemption, and how can the aging technique prevent it?