Deadlocks: Understanding and Prevention in Computer Science, Study notes of Computer Science

Download Deadlocks: Understanding and Prevention in Computer Science and more Study notes Computer Science in PDF only on Docsity!

Chapter 3

Chapter 3: Deadlocks

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Overview

„^

Resources

„^

Why do deadlocks occur?

„^

Dealing with deadlocks^ „ Ignoring them: ostrich algorithm „ Detecting & recovering from deadlock „ Avoiding deadlock „ Preventing deadlock

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

When do deadlocks happen?

„^

Suppose^ „^

Process 1 holds resource Aand requests resource B „^

Process 2 holds B andrequests A „^

Both can be blocked, withneither able to proceed „^

Deadlocks occur when …^ „^

Processes are grantedexclusive access to devices orsoftware constructs(resources) „^

Each deadlocked processneeds a resource held byanother deadlocked process

A

B

B

A

Process 1

Process 2

DEADLOCK!

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Using resources

„^

Sequence of events required to use a resource^ „ Request the resource „ Use the resource „ Release the resource

„^

Can

’t use the resource if request is denied „^

Requesting process has options^ „

Block and wait for resource „ Continue (if possible) without it: may be able to use an alternateresource „ Process fails with error code „^

Some of these may be able to prevent deadlock

…

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Four conditions for deadlock

„^

Mutual exclusion^ „ Each resource is assigned to at most one process

„^

Hold and wait^ „ A process holding resources can request more resources

„^

No preemption^ „ Previously granted resources cannot be forcibly takenaway

„^

Circular wait^ „ There must be a circular chain of 2 or more processeswhere each is waiting for a resource held by the nextmember of the chain

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Resource allocation graphs

„^

Resource allocationmodeled by directed graphs „^

Example 1:^ „^

Resource R assigned toprocess A „^

Example 2:^ „^

Process B is requesting /waiting for resource S „^

Example 3:^ „^

Process C holds T, waitingfor U „^

Process D holds U, waitingfor T „^

C and D are in deadlock!

A R

B S

T U

D

C

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Getting into deadlock

A^

B^

C

Acquire RAcquire SRelease RRelease S

Acquire SAcquire TRelease SRelease T

Acquire TAcquire RRelease TRelease R

A R

B S

C T

Acquire R

A R

B S

C T

Acquire S

A R

B S

C T

Acquire T

A R

B S

C T

Acquire S

A R

B S

C T

Acquire T

A R

B S

C T

Deadlock! Acquire R

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Not getting into deadlock…

„^

Many situations

may

result in deadlock (but don

’t

have to)^ „ In previous example, A could release R before C requestsR, resulting in no deadlock „ Can we always get out of it this way?

„^

Find ways to:^ „ Detect deadlock and reverse it „ Stop it from happening in the first place

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Detecting deadlocks using graphs

„^

Process holdings and requests in the table and in the graph(they

’re equivalent)

„^

Graph contains a cycle => deadlock!^ „^

Easy to pick out by looking at it (in this case) „^

Need to mechanically detect deadlock „^

Not all processes are deadlocked (A, C, F not in deadlock)

R^

A S F W

C

U

V

G

S

W

F

V

T

E

S,T

U

D

S

C

T

B

S

R

A

Wants Holds

Process

E

D G

B T

V

U

Chapter 3

14

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Deadlock detection algorithm

„^

General idea: try to findcycles in the resourceallocation graph „^

Algorithm: depth-firstsearch at each node^ „^

Mark arcs as they

’re

traversed „^

Build list of visited nodes „^

If node to be added is alreadyon the list, a cycle exists! „^

Cycle == deadlock

Foreach node N i

n the gr

aph {

SetL = em pt

y li st

unm ar

k al

lar

cs

Traver

se (

N ,L)

} Ifno deadl

ock r

epor

ted by now ,t

her

e isn’

t

any def

ine Tr

aver

se (

C ,L) {

IfC i

n L,r

epor

tdeadl

ock!

Add C t

o L Foreach unm ar

ked ar

c fr

om C {

M ar

k the ar

c

SetA = ar

c dest

inat

ion

/*N O TE:L i

s a

localvar

iabl

e */

Traver

se (

A,L)

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Deadlock detection algorithm

1 0 3 2

Avail

D C B A

D C B A

Process

D C B A

Process Hold Want

cur

rent

=avai

l;

for(

j= 0;j< N ;j

for(

k=0;k<N ;k++){ if(f

inished[

k])

cont

inue; if(w ant

[k]< cur

rent

fini shed[

k]= 1; cur

rent+= hol

d[k]

break;} if(k==N ){print

f“D eadl

ock!

\n”;

//fi

nished[

k]==0 m eans pr

ocess i

s in

//the deadl

ock

break;} } Note: want[j],hold[j],current,avail are arrays!

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Recovering from deadlock

„^

Recovery through preemption^ „^

Take a resource from some other process „^

Depends on nature of the resource and the process „^

Recovery through rollback^ „^

Checkpoint a process periodically „^

Use this saved state to restart the process if it is found deadlocked „^

May present a problem if the process affects lots of

“external

”^ things

„^

Recovery through killing processes^ „^

Crudest but simplest way to break a deadlock: kill one of theprocesses in the deadlock cycle „^

Other processes can get its resources „^

Preferably, choose a process that can be rerun from the beginning^ „

Pick one that hasn

’t run too far already

Chapter 3

19

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Safe and unsafe states

Free: 3

7 2 C

4 2 B

9 3 A

Max Has

Free: 1

7 2 C

4 4 B

9 3 A

Max Has

Free: 5

7 2 C

0 B

9 3 A

Max Has

Free: 0

7 7 C

0 B

9 3 A

Max Has

Free: 7

0 C

0 B

9 3 A

Max Has

Demonstration that the first state is safe

Free: 3

7 2 C

4 2 B

9 3 A

Max Has

Free: 2

7 2 C

4 2 B

9 4 A

Max Has

Free: 0

7 2 C

4 4 B

9 4 A

Max Has

Free: 4

7 2 C

0 B

9 4 A

Max Has

Demonstration that the second state is unsafe

Chapter 3

CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)

Banker's Algorithm for a single resource

4 0 C^ Free: 10

7 0 D

5 0 B

6 0 A

Max Has

4 2 C

Free: 2

7 4 D

5 1 B

6 1 A

Max Has

4 2 C

Free: 1

7 4 D

5 2 B

6 1 A

Max Has

„^

Bankers

’^ algorithm: before granting a request, ensure that a

sequence exists that will allow all processes to complete^ „^

Use previous methods to find such a sequence „^

If a sequence exists, allow the requests „^

If there

’s no such sequence, deny the request

„^

Can be slow: must be done on each request! Any sequence finishes

C,B,A,D finishes

Deadlock (unsafe state)