1. Deadlock ICS 240: Operating Systems
Instructor: William McDaniel Albritton
Information and Computer Sciences Department at Leeward Community College
Original slides by Silberschatz, Galvin, and Gagne from Operating System Concepts with Java, 7th Edition with some modifications
Also includes material by Dr. Susan Vrbsky from the Computer Science Department at the University of Alabama
4/27/2017
4/27/2017
4/27/2017 1 1

2. Deadlock
Deadlock is a set of processes permanently blocked, waiting for an event that will never occur
Blocked waiting for resources that are held by another process in the set and will never become available
Deadlock is slightly different than starvation
With starvation, process blocked on resources that become available, but never acquired
4/27/2017

3. Deadlock Deadlock theory very well developed
Many algorithms can be used to prevent or deal with deadlocks
However, few current operating systems use these deadlock prevention algorithms due to high overhead
4/27/2017

4. Deadlock
Modern operating systems have increased possibility of deadlock, because of
Large number of concurrent processes
Multithreaded programs
Sharing of multiple resources in a system
Long-lived file and database servers
Because of these trends, operating systems in the near future will probably incorporate more algorithms that prevent deadlock
4/27/2017

5. The Deadlock Problem
A set of blocked processes each holding a resource and waiting to acquire a resource held by another process in the set
Example
System has 2 disk drives
Process #1 and Process #2 each hold one disk drive and each needs another one
4/27/2017

6. The Deadlock Problem
A set of blocked processes each holding a resource and waiting to acquire a resource held by another process in the set
Example
Semaphores A and B, initialized to 1
Semaphore A = new Semaphore(1);
Semaphore B = new Semaphore(1);
Process #1 Process #2 A.acquire(); B.acquire() B.acquire(); A.acquire()
4/27/2017

7. Bridge Crossing Example
Traffic only in one direction on the bridge
Each section of a bridge can be viewed as a resource
If a deadlock occurs, it can be resolved if one car backs up
Preempt resources and rollback
Several cars may have to be backed up if a deadlock occurs
Starvation is also possible in this example
4/27/2017

8. Bridge Crossing Example
Diagram
4/27/2017

9. System Model Resource types R1, R2, . . ., Rm
For example, CPU, memory, and I/O devices such as printers
Each resource type Ri may have one (1) or more instances
For example, a system with 3 CPUs has 3 instances of a CPU resource
A system with 10 printers has 10 instances of a printer resource
4/27/2017

10. System Model Each process utilizes a resource as follows
Process requests resource
If the request cannot be processed immediately, the process must wait until it can acquire the resource
For example, open a file
Process uses resource
For example, write to a file
Process releases resource
For example, close a file
4/27/2017

11. System Model
“Request, use, and release” may sound simple, but this is supported by a complex system
A system table is needed for the resources
Keep track of state of resource: free or allocated
Keep track of the process allocated to each resource
Each resource needs a queue for processes waiting on the resource
Multithreaded applications will be accessing the same resources, so easy to have deadlock
4/27/2017

12. Deadlock Characterization
Deadlock can arise if these four (4) conditions hold simultaneously
Mutual exclusion
Hold and wait
No preemption
Circular wait
4/27/2017

13. Deadlock Characterization
Mutual exclusion
Resources are not sharable
Only one process at a time can use a resource
If another process requests the resources, this process is delayed until the resource is released
Hold and wait
Process holds resources allocated to them while waiting to acquire additional resources
A process holding at least one resource is waiting to acquire additional resources held by other processes
4/27/2017

14. Deadlock Characterization
No preemption
The system cannot forcibly take a resource from a process
A resource can be released only voluntarily by the process holding it, after that process has completed its task
4/27/2017

15. Deadlock Characterization
Circular wait
There exists a set of processed {P0...Pn} such that
P0 is waiting for a resource held by P1
P1 is waiting for a resource held by P2
. . .
Pn-1 is waiting for a resource held by Pn
Pn is waiting for a resource held by P0
4/27/2017

16. Resource-Allocation Graph
A set of vertices (nodes, circles) V and a set of edges (arrows, lines) E
V is partitioned into two types
P = {P1, P2, …, Pn}, the set consisting of all the processes in the system
R = {R1, R2, …, Rm}, the set consisting of all resource types in the system
“Process #i requests Resouce #j” is represented by the directed edge Pi  Rj
“Resource #j is assigned to Process #i” is represented by the directed edge Rj  Pi
4/27/2017

17. Resource-Allocation Graph
Process
Resource Type with 4 instances
Pi requests instance of Rj
Pi is holding an instance of Rj Pi Rj Pi Rj
4/27/2017

18. Example Graph #1 Resource Allocation Graph without a cycle
Processes: P1, P2, P3
Resource Types: R1, R2, R3, R4
Resource Instances
R1 & R3: one instance
R2: two instances
R4: three instances
4/27/2017

19. Example Graph #1 Process states Process P1 Process P2
Waiting for an instance of Resource Type R1
Acquired an instance of Resource Type R2
Process P2
Acquired an instance of Resource Type R1
Waiting for an instance of Resource Type R3
4/27/2017

20. Example Graph #1 Process states
Process P3
Acquired an instance of Resource Type R3
Since this graph has no cycles, no deadlock exists
If no cycles, then no deadlock
4/27/2017

21. Example Graph #2 Resource Allocation Graph With Deadlock
Cycles in this system
P1  R1  P2  R3 P3  R2  P1
P2  R3  P3  R2 P2
4/27/2017

22. Example Graph #2 Processes P1, P2, and P3 are deadlocked
P2 is waiting on R3
But R3 is held by P3
P3 is waiting on R2
But R2 is held by P1 and P2
P1 is waiting on R1
But R1 is held by P2
4/27/2017

23. Example Graph #3 Graph With A Cycle But No Deadlock
Cycles in this system
P1  R1  P3  R2 P1
Have one cycle, but no deadlock
P4 may release R2
Then, R2 can be allocated to P3
This breaks the cycle
4/27/2017

24. Basic Facts If graph contains no cycles If graph contains a cycle
No deadlock
If graph contains a cycle
If only one instance per resource type, then deadlock
If several instances per resource type, possibility of deadlock
In summary, if no cycles, then no deadlock; however, if cycles exist, then the system may or may not have deadlock
4/27/2017

25. Solutions to Deadlock Four ways to deal with deadlock problem
Prevention
Remove all possibility of deadlock by denying at least one of the four necessary conditions for deadlock
Detection and Recovery
Allow deadlock to occur, detect it when it does
Occasionally examine state of system, checking for deadlock
Clear deadlock when it occurs
4/27/2017

26. Solutions to Deadlock Four ways to deal with deadlock problem
Avoidance
Sidestep deadlock if it is imminent
Grant only those requests which cannot result in a state of deadlock
Ignore the problem
Pretend that deadlock never occurs
Used by most operating systems, including UNIX and Windows
4/27/2017

27. Deadlock Prevention
Design system so deadlock can not occur by disallowing one of the necessary conditions
Mutual Exclusion
Mutual exclusion is not required for sharable resources
However, we cannot disallow mutual exclusion for resources that cannot be shared
So we cannot prevent deadlocks by disallowing mutual exclusion, as mutual exclusion is required to manage non-sharable resources
4/27/2017

28. Deadlock Prevention
Design system so deadlock can not occur by disallowing one of the necessary conditions
Hold and Wait
Must guarantee that whenever a process requests a resource, it does not hold any other resources
Two possible protocols
Require process to request and be allocated all its resources before it begins execution
Allow process to request resources only when the process has none
4/27/2017

29. Deadlock Prevention Hold and Wait
Example: a process that copies a file from DVD to disk, sorts the file, and prints the file
Protocol #1: If the process is allocated all resources at once, then it has to hold the printer for the whole time period, even though it only needs the printer at the end. This is wasteful of resources.
Protocol #2: The process only acquires the DVD and disk, and then releases both. Then the process acquires the disk and printer, and then releases both. This assumes that the data is still on the file.
4/27/2017 29

30. Deadlock Prevention Hold and Wait Problems
A given process may not know all the resources that are needed
Wasteful of resources
Starvation is possible
4/27/2017

31. Deadlock Prevention
Design system so deadlock can not occur by disallowing one of the necessary conditions
No preemption
Allow system to take a process’s resources
Two possible protocols
If process requests a resource that it can not immediately have, system takes all resources currently allocated to the process and process must re-request all resources plus new one
If process requests a resource that it cannot immediately have, see if another waiting process has the resource and take it
4/27/2017

32. Deadlock Prevention No preemption Problems
Not all resources can be preempted
For example, a printer cannot be preempted
Preemption and resumption of resources is difficult, costly saving states, etc.
4/27/2017

33. Deadlock Prevention
Design system so deadlock can not occur by disallowing one of the necessary conditions
Circular wait
Give an order to all resource types
type 1, type 2, type3.....type n
If hold resource of type i, can only request from classes greater than i
Must request resources in an increasing order of enumeration
Assign more expensive resources higher numbers
4/27/2017 33

34. Deadlock Prevention Circular wait Problems
Poor resource utilization (resources requested before needed)
Degradation of concurrency
Process may request max needed for each class although not always required
Preventing circular wait has less overhead than preventing the other 3 conditions and is a common preventive technique
4/27/2017 34

35. Deadlock Avoidance
Requires that the system has some additional information up front
Simplest and most useful model requires that each process declare the maximum number of resources of each type that it may need
The deadlock-avoidance algorithm dynamically examines the resource-allocation state to ensure that there can never be a circular-wait condition
Resource-allocation state is defined by the number of available and allocated resources, and the maximum demands of the processes
4/27/2017

36. Safe State Deadlock avoidance means to keep the system in a safe state
Therefore the system must ensure that there is never a circular wait
Safe state
There exists a sequence of processes (P1,P2,...Px) such that each process gets the maximum resources it could possibly want
In other words, a cycle does NOT have the possibility of occurring
4/27/2017

37. Basic Facts If a system is in safe state
No deadlocks
If a system is in unsafe state
Possibility of deadlock
Avoidance
Ensure that a system will never enter an unsafe state
4/27/2017

38. Avoidance Algorithms Single instance of each resource type
Use a resource-allocation graph
Multiple instances of each resource type
Use the banker’s algorithm
4/27/2017

39. Resource-Allocation Graph Scheme
Claim edge Pi  Rj indicated that process Pj may request resource Rj
A claim edge is represented by a dashed line
Resources in the system must be claimed in advance
In other words, before Pi starts executing, all its claim edges must already appear in the resource-allocation graph
4/27/2017

40. Resource-Allocation Graph Scheme
Changing from one edge to another
Claim edge converts to request edge when a process requests a resource
Request edge converted to an assignment edge when the resource is allocated to the process
When a resource is released by a process, assignment edge reconverts to a claim edge
4/27/2017 40

41. Resource-Allocation Graph Algorithm
Suppose that process Pi requests a resource Rj
The request can be granted only if converting the request edge to an assignment edge does not result in the formation of a cycle in the resource allocation graph
4/27/2017

42. Resource-Allocation Graph
Processes
P1, P2
Resources
R1, R2
Claim edge
P1  R2
P2  R2
Request edge
P2  R1
Assignment edge
R1  P1
4/27/2017

43. Unsafe State In Resource-Allocation Graph
Claim edge
P1  R2
Request edge
P2  R1
Assignment edge
R1  P1
R2  P2
4/27/2017

44. Unsafe State In Resource-Allocation Graph
R2 should not be allocated to P2, because this will create a cycle
A cycle indicates that the system is in an unsafe state
If P1 requests R2, and P2 requests P1, then a deadlock will occur
4/27/2017 44

45. Banker’s Algorithm
The Banker’s Algorithm is used when there are multiple instances of a resource type
Originally created by Dijkstra in 1968
The name was chosen because we can use this algorithm in a banking system to make sure that the bank never distributes its cash, so that it can no longer satisfy the needs of all its customers
4/27/2017

46. Banker’s Algorithm 3 main parts to the Banker’s Algorithm
When a new process enters the system, it must give the maximum number of resource instances that it needs
The system uses this information to check for a safe state before allocating a request
If not a safe state, then a process must wait until it can get all its resources
4/27/2017 46

47. Data Structures for Banker’s Algorithm
Let n = number of processes, and m = number of resources types.
Available: Vector of length m. If available [j] = k, there are k instances of resource type Rj available.
Max: n x m matrix. If Max [i,j] = k, then process Pi may request at most k instances of resource type Rj.
Allocation: n x m matrix. If Allocation[i,j] = k then Pi is currently allocated k instances of Rj.
Need: n x m matrix. If Need[i,j] = k, then Pi may need k more instances of Rj to complete its task.
Need [i,j] = Max[i,j] – Allocation [i,j].
4/27/2017

48. Safety Algorithm
1. Let Work and Finish be vectors of length m and n, respectively. Initialize: Work = Available Finish [i] = false for i = 0, 1, …, n Find and i such that both: (a) Finish [i] = false (b) Needi  Work If no such i exists, go to step Work = Work + Allocationi Finish[i] = true go to step 2 4. If Finish [i] == true for all i, then the system is in a safe state.
4/27/2017

49. Resource-Request Algorithm for Process Pi
Request = request vector for process Pi. If Requesti [j] = k then process Pi wants k instances of resource type Rj.
If Requesti  Needi go to step 2. Otherwise, raise error condition, since process has exceeded its maximum claim
If Requesti  Available, go to step 3. Otherwise Pi must wait, since resources are not available
Pretend to allocate requested resources to Pi by modifying the state as follows
Available = Available – Request;
Allocationi = Allocationi + Requesti;
Needi = Needi – Requesti;
If safe  the resources are allocated to Pi
If unsafe  Pi must wait, and the old resource-allocation state is restored
4/27/2017

50. Example of Banker’s Algorithm
5 processes P0 through P4
3 resource types
A (10 instances), B (5instances), and C (7 instances)
Snapshot at time T0
Allocation Max Available
A B C A B C A B C P P P P P
4/27/2017

51. Example of Banker’s Algorithm
The content of the matrix Need is defined to be Max – Allocation.
Need
A B C P P P P P
The system is in a safe state since the sequence < P1, P3, P4, P2, P0> satisfies safety criteria.
4/27/2017

52. Example: P1 Request (1,0,2) Check that Request  Available
that is, (1,0,2)  (3,3,2)  true.
Allocation Need Available
A B C A B C A B C P P P P P
Executing safety algorithm shows that sequence < P1, P3, P4, P0, P2> satisfies safety requirement
4/27/2017

53. Deadlock Detection & Recovery
If we do not use deadlock prevention or deadlock avoidance in our system, then we may have deadlock at some time
In this case, we need a deadlock detection and recovery scheme
Allow system to enter deadlock state
Detection algorithm
Recovery algorithm
4/27/2017

54. Detection Algorithm
This detection algorithm is used when there is a single instance of each resource type
Maintain wait-for graph
Nodes are processes
Pi  Pj if Pi is waiting for Pj
Periodically invoke an algorithm that searches for a cycle in the wait-for graph
If there is a cycle, there exists a deadlock
Only when there is single instance of each resource type
For multiple instance of each resource type, a more complicated algorithm is used
4/27/2017

55. Detection Algorithm Creation of wait-for graph
A wait-for graph is created by removing the resource nodes from the resource-allocation graph and collapsing the edges
4/27/2017

56. Detection Algorithm Resource-allocation graph
Corresponding wait-for graph
4/27/2017

57. Detection Algorithm
The algorithm used to detect a cycle in a graph (when there is a single instance of each resource type) requires an order of O(n2) operations, where n is the number of vertices in the graph
For multiple instances of resources, the algorithm requires an order of O(m x n2) operations to detect a cycle, where m is the number of instances of a resource
In both cases, it takes a long time to detect a cycle
4/27/2017 57

58. Detection Algorithm When to run the detection algorithm?
Considerations
Resources allocated to deadlocked processes are idle (resource waste)
As time progresses, number of processes in deadlocked set may grow
How often do we think a deadlock will occur?
How many processes will be affected when it happens?
4/27/2017 58

59. Detection Algorithm When to run the detection algorithm? Possibilities
Invoke if a request is not immediately granted (too much overhead)|
Fixed time interval, say every 20 seconds or so when CPU drops below a certain threshold , say 40%
Processes blocked
CPU available to run algorithm
4/27/2017 59

60. Recovery Algorithm
We have two possibilities to recover from a deadlock
Process termination
Resource preemption
4/27/2017

61. Process Termination Kill (terminate) processes
System reclaims all resources allocated to killed process
We have two choices
Abort all deadlocked processes
If all killed, will deadlock occur again? Deadlock due to a particular execution sequence, so unlikely to repeat exact sequence
Abort one process at a time until deadlock cycle is eliminated
This can be complex. What happens if kill producer? Must kill also consumer or send error messages to processes trying to consume
4/27/2017 61

62. Process Termination
When we kill one process at a time, how do we select a victim?
Process priority
Process execution completed (how long process has computed)
Number and type of resources held by process
Number and type of resources required to complete process
How many processes will require termination
4/27/2017 62

63. Resource Preemption
Preempt one or more resources from one or more of the deadlocked processes
What to do with preempted process?
Return and restart a process from a safe state checkpoint - a recording of a previous state
How to ensure no starvation?
Be sure resources are not preempted from the same process each time by including the number of times a process has rolledback in the cost factor (a priority function)
Rollback: return to some safe state, and restart process for that state
4/27/2017