Operating Systems {week 10}

1. Rensselaer Polytechnic Institute CSCI-4210 – Operating Systems David Goldschmidt, Ph.D. Operating Systems{week 10}

2. A need for synchronization (i) • Without synchronization amongst processes (and threads), results are unpredictable how dovariables x and y become corrupted?

3. A need for synchronization (ii) • Processes compete for resources • Once obtained, the resource isfully dedicated to a process • Often, mutual exclusion is required • No other process is allowed access to the resource • Processes cooperate with other processes • Shared resources • Specific ordering or sequencing of events all of this applies to threads, too!

4. Semaphores (i) • A semaphore is a synchronization mechanism • Semaphore S is a special integer variable • OS provides two atomic operations on S: • wait(S) or P(S): • wait for a resource to become available • signal(S) or V(S): • signal that we’re done using a resource

5. Semaphores (ii) • The wait(S) operation decrementssemaphore S only when S is available • wait(S) { while ( S <= 0 ) { /** no-op **/ ; } S--; } this will block indefinitely in a busy wait

6. Semaphores (iii) • The signal(S) operation incrementssemaphore S to release a resource • signal(S) { S++; } • wait(S);// CRITICAL// SECTIONsignal(S); to protect critical section

7. Binary semaphores • A binary semaphore providesmutually exclusive access toa shared resource • Initialize semaphore S to 1 • Use wait(S) and signal(S) • Possible values of S are 0 and 1

8. Counting semaphores • A counting semaphore controls accessto a finite number of resources: • e.g. open files, network connections, shared buffers, etc. // n instances of a finite resource semaphore S = n Write pseudocode for theproducer-consumer problem usingsemaphores to synchronize accessto the shared buffer of size N

9. Starvation • A process faces starvation when it is forced to wait indefinitely for shared resource X as other processes use that shared resource X • Also known as indefinite blocking

10. Deadlock • A system enters a deadlockstate when multiple processes are unable to obtain a lock on all necessary resources • After acquiringa resource, aprocess holds thatresource indefinitely semaphore S, Q // P0 ... wait(S) wait(Q) ... signal(Q) signal(S) ... // P1 ... wait(Q) wait(S) ... signal(S) signal(Q) ... Deadlock!

11. Conditions for deadlock • Deadlock requires four conditions: • Mutual exclusion • Hold and wait • No preemption • Circular wait • i.e. a cycle!

12. Resource allocation graph • A resource allocation graph is a directed graph showing processes and resources

13. Deadlock!

14. Deadlock?

15. Rice Dining philosophers problem (i) • Five philosophers at a table • Each philosopher thinks or eats • To eat, a philosopher must pickup the closest two chopsticks • A philosopher may only pickup one chopstick at a time • Represents allocating shared resources to competing and cooperating processes

16. Rice Dining philosophers problem (ii) • Potential solution: • Can deadlock occur? semaphore fork[] = { 1, 1, 1, 1, 1 }; // philosopheri while ( true ) { think(); wait( fork[i] ); wait( fork[(i+1)%5] ); eat(); signal( fork[(i+1)%5] ); signal( fork[i] ); }

17. Handling deadlocks (i) • Allow the system to enter a deadlock state, then recover by: • Terminating one or all deadlocked processes • Rollback deadlocked processes to a safe checkpointed state • Or....

18. Handling deadlocks (ii) • Guarantee that the system will neverenter a deadlocked state • Deadlock prevention ensures that at least one of the four necessary conditions is never met • Deadlock avoidance allows a system to change state by allocating resource(s) only when it is certain deadlock will not occur as a result