OPERATING SYSTEM

INDEX

LESSON 1: INTRODUCTION TO OPERATING SYSTEM

LESSON 2: FILE SYSTEM – I
LESSON 3: FILE SYSTEM – II
LESSON 4: CPU SCHEDULING
LESSON 5: MEMORY MANAGEMENT – I
LESSON 6: MEMORY MANAGEMENT – II
LESSON 7: DISK SCHEDULING
LESSON 8: PROCESS MANAGEMENT
LESSON 9: DEADLOCKS
LESSON 10: CASE STUDY OF UNIX
LESSON 11: CASE STUDY OF MS-DOS
LESSON 12: CASE STUDY OF MS-WINDOWS NT

Lesson No. 1 Intro. to Operating System 1

Lesson Number: 1 Writer: Dr. Rakesh Kumar
Introduction to Operating System Vetter: Prof. Dharminder Kr.

1.0 OBJECTIVE
The objective of this lesson is to make the students familiar with the basics of
operating system. After studying this lesson they will be familiar with:
1. What is an operating system?
2. Important functions performed by an operating system.
3. Different types of operating systems.
1. 1 INTRODUCTION
Operating system (OS) is a program or set of programs, which acts as an
interface between a user of the computer & the computer hardware. The main
purpose of an OS is to provide an environment in which we can execute
programs. The main goals of the OS are (i) To make the computer system
convenient to use, (ii) To make the use of computer hardware in efficient way.
Operating System is system software, which may be viewed as collection of
software consisting of procedures for operating the computer & providing an
environment for execution of programs. It’s an interface between user &
computer. So an OS makes everything in the computer to work together
smoothly & efficiently.

Figure 1: The relationship between application & system software

Lesson No. 1 Intro. to Operating System 2

Basically, an OS has three main responsibilities: (a) Perform basic tasks such as
recognizing input from the keyboard, sending output to the display screen,
keeping track of files & directories on the disk, & controlling peripheral devices
such as disk drives & printers (b) Ensure that different programs & users running
at the same time do not interfere with each other; & (c) Provide a software
platform on top of which other programs can run. The OS is also responsible for
security, ensuring that unauthorized users do not access the system. Figure 1
illustrates the relationship between application software & system software.
The first two responsibilities address the need for managing the computer
hardware & the application programs that use the hardware. The third
responsibility focuses on providing an interface between application software &
hardware so that application software can be efficiently developed. Since the OS
is already responsible for managing the hardware, it should provide a
programming interface for application developers. As a user, we normally interact
with the OS through a set of commands. The commands are accepted &
executed by a part of the OS called the command processor or command line
interpreter.

Figure 2: The interface of various devices to an operating system

In order to understand operating systems we must understand the computer
hardware & the development of OS from beginning. Hardware means the

Lesson No. 1 Intro. to Operating System 3

physical machine & its electronic components including memory chips,
input/output devices, storage devices & the central processing unit. Software are
the programs written for these computer systems. Main memory is where the
data & instructions are stored to be processed. Input/Output devices are the
peripherals attached to the system, such as keyboard, printers, disk drives, CD
drives, magnetic tape drives, modem, monitor, etc. The central processing unit is
the brain of the computer system; it has circuitry to control the interpretation &
execution of instructions. It controls the operation of entire computer system. All
of the storage references, data manipulations & I/O operations are performed by
the CPU. The entire computer systems can be divided into four parts or
components (1) The hardware (2) The OS (3) The application programs &
system programs (4) The users.
The hardware provides the basic computing power. The system programs the
way in which these resources are used to solve the computing problems of the
users. There may be many different users trying to solve different problems. The
OS controls & coordinates the use of the hardware among the various users &
the application programs.

User User User User

Compiler Assembler Text Editor

Database

Application programs

Operating System

Computer Hardware

Figure 3. Basic components of a computer system

We can view an OS as a resource allocator. A computer system has many
resources, which are to be required to solve a computing problem. These

Lesson No. 1 Intro. to Operating System 4

resources are the CPU time, memory space, files storage space, input/output
devices & so on. The OS acts as a manager of all of these resources & allocates
them to the specific programs & users as needed by their tasks. Since there can
be many conflicting requests for the resources, the OS must decide which
requests are to be allocated resources to operate the computer system fairly &
efficiently.
An OS can also be viewed as a control program, used to control the various I/O
devices & the users programs. A control program controls the execution of the
user programs to prevent errors & improper use of the computer resources. It is
especially concerned with the operation & control of I/O devices. As stated above
the fundamental goal of computer system is to execute user programs & solve
user problems. For this goal computer hardware is constructed. But the bare
hardware is not easy to use & for this purpose application/system programs are
developed. These various programs require some common operations, such as
controlling/use of some input/output devices & the use of CPU time for execution.
The common functions of controlling & allocation of resources between different
users & application programs is brought together into one piece of software
called operating system. It is easy to define operating systems by what they do
rather than what they are. The primary goal of the operating systems is
convenience for the user to use the computer. Operating systems makes it easier
to compute. A secondary goal is efficient operation of the computer system. The
large computer systems are very expensive, & so it is desirable to make them as
efficient as possible. Operating systems thus makes the optimal use of computer
resources. In order to understand what operating systems are & what they do,
we have to study how they are developed. Operating systems & the computer
architecture have a great influence on each other. To facilitate the use of the
hardware operating systems were developed.
First, professional computer operators were used to operate the computer. The
programmers no longer operated the machine. As soon as one job was finished,
an operator could start the next one & if some errors came in the program, the
operator takes a dump of memory & registers, & from this the programmer have

Lesson No. 1 Intro. to Operating System 5

to debug their programs. The second major solution to reduce the setup time was
to batch together jobs of similar needs & run through the computer as a group.
But there were still problems. For example, when a job stopped, the operator
would have to notice it by observing the console, determining why the program
stopped, takes a dump if necessary & start with the next job. To overcome this
idle time, automatic job sequencing was introduced. But even with batching
technique, the faster computers allowed expensive time lags between the CPU &
the I/O devices. Eventually several factors helped improve the performance of
CPU. First, the speed of I/O devices became faster. Second, to use more of the
available storage area in these devices, records were blocked before they were
retrieved. Third, to reduce the gap in speed between the I/O devices & the CPU,
an interface called the control unit was placed between them to perform the
function of buffering. A buffer is an interim storage area that works like this: as
the slow input device reads a record, the control unit places each character of the
record into the buffer. When the buffer is full, the entire record is transmitted to
the CPU. The process is just opposite to the output devices. Fourth, in addition to
buffering, an early form of spooling was developed by moving off-line the
operations of card reading, printing etc. SPOOL is an acronym that stands for the
simultaneous peripherals operations on-line. Foe example, incoming jobs would
be transferred from the card decks to tape/disks off-line. Then they would be
read into the CPU from the tape/disks at a speed much faster than the card
reader.

CPU
Card Line
Reader printer
On-line

Card Tape CPU Tape Line

reader drive drive printer

Off-line

Lesson No. 1 Intro. to Operating System 6

 Disk

Line
Card
printer
reader
CPU

SPOOLING

Figure 4: the on-line, off-line & spooling processes

Moreover, the range & extent of services provided by an OS depends on a
number of factors. Among other things, the needs & characteristics of the target
environmental that the OS is intended to support largely determine user- visible
functions of an operating system. For example, an OS intended for program
development in an interactive environment may have a quite different set of
system calls & commands than the OS designed for run-time support of a car
engine.
1.2 PRESENTATION OF CONTENTS
1.2.1 Operating System as a Resource Manager
1.2.1.1 Memory Management Functions
1.2.1.2 Processor / Process Management Functions
1.2.1.3 Device Management Functions
1.2.1.4 Information Management Functions
1.2.2 Extended Machine View of an Operating System
1.2.3 Hierarchical Structure of an Operating System
1.2.4 Evolution of Processing Trends
1.2.4.1 Serial Processing
1.2.4.2 Batch Processing

Lesson No. 1 Intro. to Operating System 7

 1.2.4.3 Multi Programming
1.2.5 Types Of Operating Systems
1.2.5.1 Batch Operating System
1.2.5.2 Multi Programming Operating System
1.2.5.3 Multitasking Operating System
1.2.5.4 Multi-user Operating System
1.2.5.5 Multithreading
1.2.5.6 Time Sharing System
1.2.5.7 Real Time Systems
1.2.5.8 Combination Operating Systems
1.2.5.9 Distributed Operating Systems
1.2.6 System Calls
1.2.1 OPERATING SYSTEM AS A RESOURCE MANAGER
The OS is a manager of system resources. A computer system has many
resources as stated above. Since there can be many conflicting requests for the
resources, the OS must decide which requests are to be allocated resources to
operate the computer system fairly & efficiently. Here we present a framework of
the study of OS based on the view that the OS is manager of resources. The OS
as a resources manager can be classified in to the following three popular views:
primary view, hierarchical view, & extended machine view. The primary view is
that the OS is a collection of programs designed to manage the system’s
resources, namely, memory, processors, peripheral devices, & information. It is
the function of OS to see that they are used efficiently & to resolve conflicts
arising from competition among the various users. The OS must keep track of
status of each resource; decide which process is to get the resource, allocate it,
& eventually reclaim it.
The major functions of each category of OS are.
1.2.1.1 Memory Management Functions
To execute a program, it must be mapped to absolute addresses & loaded into
memory. As the program executes, it accesses instructions & data from memory
by generating these absolute addresses. In multiprogramming environment,

Lesson No. 1 Intro. to Operating System 8

multiple programs are maintained in the memory simultaneously. The OS is
responsible for the following memory management functions:
¾ Keep track of which segment of memory is in use & by whom.
¾ Deciding which processes are to be loaded into memory when space
becomes available. In multiprogramming environment it decides which
process gets the available memory, when it gets it, where does it get it, & how
much.
¾ Allocation or de-allocation the contents of memory when the process request
for it otherwise reclaim the memory when the process does not require it or
has been terminated.
1.2.1.2 Processor/Process Management Functions
A process is an instance of a program in execution. While a program is just a
passive entity, process is an active entity performing the intended functions of its
related program. To accomplish its task, a process needs certain resources like
CPU, memory, files & I/O devices. In multiprogramming environment, there will a
number of simultaneous processes existing in the system. The OS is responsible
for the following processor/ process management functions:
¾ Provides mechanisms for process synchronization for sharing of resources
amongst concurrent processes.
¾ Keeps track of processor & status of processes. The program that does this
has been called the traffic controller.
¾ Decide which process will have a chance to use the processor; the job
scheduler chooses from all the submitted jobs & decides which one will be
allowed into the system. If multiprogramming, decide which process gets the
processor, when, for how much of time. The module that does this is called a
process scheduler.
¾ Allocate the processor to a process by setting up the necessary hardware
registers. This module is widely known as the dispatcher.
¾ Providing mechanisms for deadlock handling.
¾ Reclaim processor when process ceases to use a processor, or exceeds the
allowed amount of usage.

Lesson No. 1 Intro. to Operating System 9

1.2.1.3 I/O Device Management Functions
An OS will have device drivers to facilitate I/O functions involving I/O devices.
These device drivers are software routines that control respective I/O devices
through their controllers. The OS is responsible for the following I/O Device
Management Functions:
¾ Keep track of the I/O devices, I/O channels, etc. This module is typically
called I/O traffic controller.
¾ Decide what is an efficient way to allocate the I/O resource. If it is to be
shared, then decide who gets it, how much of it is to be allocated, & for how
long. This is called I/O scheduling.
¾ Allocate the I/O device & initiate the I/O operation.
¾ Reclaim device as & when its use is through. In most cases I/O terminates
automatically.
1.2.1.4 Information Management Functions
¾ Keeps track of the information, its location, its usage, status, etc. The module
called a file system provides these facilities.
¾ Decides who gets hold of information, enforce protection mechanism, &
provides for information access mechanism, etc.
¾ Allocate the information to a requesting process, e.g., open a file.
¾ De-allocate the resource, e.g., close a file.
1.2.2 Network Management Functions
An OS is responsible for the computer system networking via a distributed
environment. A distributed system is a collection of processors, which do not
share memory, clock pulse or any peripheral devices. Instead, each processor is
having its own clock pulse, & RAM & they communicate through network. Access
to shared resource permits increased speed, increased functionality & enhanced
reliability. Various networking protocols are TCP/IP (Transmission Control
Protocol/ Internet Protocol), UDP (User Datagram Protocol), FTP (File Transfer
Protocol), HTTP (Hyper Text Transfer protocol), NFS (Network File System) etc.
1.2.3 EXTENDED MACHINE VIEW OF AN OPERATING SYSTEM

Lesson No. 1 Intro. to Operating System 10

As discussed in previous section, there arises a need to identify the system
resources that must be managed by the OS & using the process viewpoint, we
indicate when the corresponding resource manager comes into play. We now
answer the question, “How are these resource managers activated, & where do
they reside?” Does memory manager ever invoke the process scheduler? Does
scheduler ever call upon the services of memory manager? Is the process
concept only for the user or is it used by OS also?
The OS provides many instructions in addition to the Bare machine instructions
(A Bare machine is a machine without its software clothing, & it does not provide
the environment which most programmers are desired for). Instructions that form
a part of Bare machine plus those provided by the OS constitute the instruction
set of the extended machine. The situation is pictorially represented in figure 5.
The OS kernel runs on the bare machine; user programs run on the extended
machine. This means that the kernel of OS is written by using the instructions of
bare machine only; whereas the users can write their programs by making use of
instructions provided by the extended machine.

Process 3 Process 1
Extended Machine

Bare
Machine

Process 4 Process 2

Figure 5. Extended Machine View

The OS kernel runs on the bare machine; user programs run on the extended
machine. This means that the kernel of OS is written by using the instructions of
bare machine only; whereas the users can write their programs by making use of
instructions provided by the extended machine.

Lesson No. 1 Intro. to Operating System 11

1.2.4 EVOLUTION OF PROCESSING TRENDS
Starting from the bare machine approach to its present forms, the OS has
evolved through a number of stages of its development like serial processing,
batch processing multiprocessing etc. as mentioned below:
1.2.4.1 Serial Processing
In theory, every computer system may be programmed in its machine language,
with no systems software support. Programming of the bare machine was
customary for early computer systems. A slightly more advanced version of this
mode of operation is common for the simple evaluation boards that are
sometimes used in introductory microprocessor design & interfacing courses.
Programs for the bare machine can be developed by manually translating
sequences of instructions into binary or some other code whose base is usually
an integer power of 2. Instructions & data are then entered into the computer by
means of console switches, or perhaps through a hexadecimal keyboard.
Loading the program counter with the address of the first instruction starts
programs. Results of execution are obtained by examining the contents of the
relevant registers & memory locations. The executing program, if any, must
control Input/output devices, directly, say, by reading & writing the related I/O
ports. Evidently, programming of the bare machine results in low productivity of
both users & hardware. The long & tedious process of program & data entry
practically precludes execution of all but very short programs in such an
environment.
The next significant evolutionary step in computer-system usage came about
with the advent of input/output devices, such as punched cards & paper tape, &
of language translators. Programs, now coded in a programming language, are
translated into executable form by a computer program, such as a compiler or an
interpreter. Another program, called the loader, automates the process of loading
executable programs into memory. The user places a program & its input data
on an input device, & the loader transfers information from that input device into
memory. After transferring control to the loader program by manual or automatic
means, execution of the program commences. The executing program reads its

Lesson No. 1 Intro. to Operating System 12

input from the designated input device & may produce some output on an output
device. Once in memory, the program may be rerun with a different set of input
data.
The mechanics of development & preparation of programs in such environments
are quite slow & cumbersome due to serial execution of programs & to numerous
manual operations involved in the process. In a typical sequence, the editor
program is loaded to prepare the source code of the user program. The next
step is to load & execute the language translator & to provide it with the source
code of the user program. When serial input devices, such as card reader, are
used, multiple-pass language translators may require the source code to be
repositioned for reading during each pass. If syntax errors are detected, the
whole process must be repeated from the beginning. Eventually, the object code
produced from the syntactically correct source code is loaded & executed. If run-
time errors are detected, the state of the machine can be examined & modified
by means of console switches, or with the assistance of a program called a
debugger.
1.2.4.2 Batch Processing
With the invention of hard disk drive, the things were much better. The batch
processing was relied on punched cards or tape for the input when assembling
the cards into a deck & running the entire deck of cards through a card reader as
a batch. Present batch systems aren’t limited to cards or tapes, but the jobs are
still processed serially, without the interaction of the user. The efficiency of these
systems was measured in the number of jobs completed in a given amount of
time called as throughput. Today’s operating systems are not limited to batch
programs. This was the next logical step in the evolution of operating systems to
automate the sequencing of operations involved in program execution & in the
mechanical aspects of program development. The intent was to increase system
resource utilization & programmer productivity by reducing or eliminating
component idle times caused by comparatively lengthy manual operations.
Furthermore, even when automated, housekeeping operations such as mounting
of tapes & filling out log forms take a long time relative to processors & memory

Lesson No. 1 Intro. to Operating System 13

speeds. Since there is not much that can be done to reduce these operations,
system performance may be increased by dividing this overhead among a
number of programs. More specifically, if several programs are batched together
on a single input tape for which housekeeping operations are performed only
once, the overhead per program is reduced accordingly. A related concept,
sometimes called phasing, is to prearrange submitted jobs so that similar ones
are placed in the same batch. For example, by batching several Fortran
compilation jobs together, the Fortran compiler can be loaded only once to
process all of them in a row. To realize the resource-utilization potential of batch
processing, a mounted batch of jobs must be executed automatically, without
slow human intervention. Generally, OS commands are statements written in
Job Control Language (JCL). These commands are embedded in the job stream,
together with user programs & data. A memory-resident portion of the batch
operating system- sometimes called the batch monitor- reads, interprets, &
executes these commands.
Moreover, the sequencing of program execution mostly automated by batch
operating systems, the speed discrepancy between fast processors &
comparatively slow I/O devices, such as card readers & printers, emerged as a
major performance bottleneck. Further improvements in batch processing were
mostly along the lines of increasing the throughput & resource utilization by
overlapping input & output operations. These developments have coincided with
the introduction of direct memory access (DMA) channels, peripheral controllers,
& later dedicated input/output processors. As a result, satellite computers for
offline processing were often replaced by sophisticated input/output programs
executed on the same computer with the batch monitor.
Many single-user operating systems for personal computers basically provide for
serial processing. User programs are commonly loaded into memory & executed
in response to user commands typed on the console. A file management system
is often provided for program & data storage. A form of batch processing is made
possible by means of files consisting of commands to the OS that are executed

Lesson No. 1 Intro. to Operating System 14

in sequence. Command files are primarily used to automate complicated
customization & operational sequences of frequent operations.
1.2.4.3 Multiprogramming
In multiprogramming, many processes are simultaneously resident in memory, &
execution switches between processes. The advantages of multiprogramming
are the same as the commonsense reasons that in life you don't always wait until
one thing has finished before starting the next thing. Specifically:
¾ More efficient use of computer time. If the computer is running a single
process, & the process does a lot of I/O, then the CPU is idle most of the
time. This is a gain as long as some of the jobs are I/O bound -- spend most
of their time waiting for I/O.
¾ Faster turnaround if there are jobs of different lengths. Consideration (1)
applies only if some jobs are I/O bound. Consideration (2) applies even if all
jobs are CPU bound. For instance, suppose that first job A, which takes an
hour, starts to run, & then immediately afterward job B, which takes 1 minute,
is submitted. If the computer has to wait until it finishes A before it starts B,
then user A must wait an hour; user B must wait 61 minutes; so the average
waiting time is 60-1/2 minutes. If the computer can switch back & forth
between A & B until B is complete, then B will complete after 2 minutes; A will
complete after 61 minutes; so the average waiting time will be 31-1/2 minutes.
If all jobs are CPU bound & the same length, then there is no advantage in
multiprogramming; you do better to run a batch system. The
multiprogramming environment is supposed to be invisible to the user
processes; that is, the actions carried out by each process should proceed in
the same was as if the process had the entire machine to itself.
This raises the following issues:
¾ Process model: The state of an inactive process has to be encoded & saved
in a process table so that the process can be resumed when made active.
¾ Context switching: How does one carry out the change from one process to
another?

Lesson No. 1 Intro. to Operating System 15

¾ Memory translation: Each process treats the computer's memory as its own
private playground. How can we give each process the illusion that it can
reference addresses in memory as it wants, but not have them step on each
other's toes? The trick is by distinguishing between virtual addresses -- the
addresses used in the process code -- & physical addresses -- the actual
addresses in memory. Each process is actually given a fraction of physical
memory. The memory management unit translates the virtual address in the
code to a physical address within the user's space. This translation is invisible
to the process.
¾ Memory management: How does the OS assign sections of physical memory
to each process?
¾ Scheduling: How does the OS choose which process to run when?
Let us briefly review some aspects of program behavior in order to motivate the
basic idea of multiprogramming. This is illustrated in Figure 6, indicated by
dashed boxes. Idealized serial execution of two programs, with no inter-program
idle times, is depicted in Figure 6(a). For comparison purposes, both programs
are assumed to have identical behavior with regard to processor & I/O times &
their relative distributions. As Figure 6(a) suggests, serial execution of programs
causes either the processor or the I/O devices to be idle at some time even if the
input job stream is never empty. One way to attack this problem is to assign
some other work to the processor & I/O devices when they would otherwise be
idling.
Program 1 Program 2

P1 Idle P1 Idle P2 Idle P2 Idle

CPU- activity Time

Figure 6 (a) Sequential execution

Figure 6(b) illustrates a possible scenario of concurrent execution of the two

programs introduced in Figure 6(a). It starts with the processor executing the first

Lesson No. 1 Intro. to Operating System 16

computational sequence of Program 1. Instead of idling during the subsequent
I/O sequence of Program 1, the processor is assigned to the first computational
sequence of the Program 2, which is assumed to be in memory & awaiting
execution. When this work is done, the processor is assigned to Program 1
again, then to Program 2, & so forth.
Program 1

Program 2

P1 P2 P1 P2 P1 Time
CPU- activity

Figure 6(b) Multiprogrammed execution

As Figure 6 suggests, significant performance gains may be achieved by
interleaved executing of programs, or multiprogramming, as this mode of
operation is usually called. With a single processor, parallel execution of
programs is not possible, & at most one program can be in control of the
processor at any time. The example presented in Figure 6(b) achieves 100%
processor utilization with only two active programs. The number of programs
actively competing for resources of a multi-programmed computer system is
called the degree of multiprogramming. In principle, higher degrees of
multiprogramming should result in higher resource utilization. Time-sharing
systems found in many university computer centers provide a typical example of
a multiprogramming system.
1.2.5 TYPES OF OPERATING SYSTEMS
Operating system can be classified into various categories on the basis of
several criteria, viz. number of simultaneously active programs, number of users
working simultaneously, number of processors in the computer system, etc. In
the following discussion several types of operating systems are discussed.

Lesson No. 1 Intro. to Operating System 17

1.2.5.1 Batch Operating System
Batch processing is the most primitive type of operating system. Batch
processing generally requires the program, data, & appropriate system
commands to be submitted together in the form of a job. Batch operating
systems usually allow little or no interaction between users & executing
programs. Batch processing has a greater potential for resource utilization than
simple serial processing in computer systems serving multiple users. Due to
turnaround delays & offline debugging, batch is not very convenient for program
development. Programs that do not require interaction & programs with long
execution times may be served well by a batch operating system. Examples of
such programs include payroll, forecasting, statistical analysis, & large scientific
number-crunching programs. Serial processing combined with batch like
command files is also found on many personal computers. Scheduling in batch is
very simple. Jobs are typically processed in order of their submission, that is,
first-come first-served fashion.
Memory management in batch systems is also very simple. Memory is usually
divided into two areas. The resident portion of the OS permanently occupies one
of them, & the other is used to load transient programs for execution. When a
transient program terminates, a new program is loaded into the same area of
memory. Since at most one program is in execution at any time, batch systems
do not require any time-critical device management. For this reason, many serial
& I/O & ordinary batch operating systems use simple, program controlled method
of I/O. The lack of contention for I/O devices makes their allocation &
deallocation trivial.
Batch systems often provide simple forms of file management. Since access to
files is also serial, little protection & no concurrency control of file access in
required.
1.2.5.2 Multiprogramming Operating System
A multiprogramming system permits multiple programs to be loaded into memory
& execute the programs concurrently. Concurrent execution of programs has a
significant potential for improving system throughput & resource utilization

Lesson No. 1 Intro. to Operating System 18

relative to batch & serial processing. This potential is realized by a class of
operating systems that multiplex resources of a computer system among a
multitude of active programs. Such operating systems usually have the prefix
multi in their names, such as multitasking or multiprogramming.
1.2.5.3 Multitasking Operating System
An instance of a program in execution is called a process or a task. A
multitasking OS is distinguished by its ability to support concurrent execution of
two or more active processes. Multitasking is usually implemented by maintaining
code & data of several processes in memory simultaneously, & by multiplexing
processor & I/O devices among them. Multitasking is often coupled with
hardware & software support for memory protection in order to prevent erroneous
processes from corrupting address spaces & behavior of other resident
processes. Allows more than one program to run concurrently. The ability to
execute more than one task at the same time, a task being a program is called
as multitasking. The terms multitasking & multiprocessing are often used
interchangeably, although multiprocessing sometimes implies that more than one
CPU is involved. In multitasking, only one CPU is involved, but it switches from
one program to another so quickly that it gives the appearance of executing all of
the programs at the same time. There are two basic types of multitasking:
preemptive & cooperative. In preemptive multitasking, the OS parcels out CPU
time slices to each program. In cooperative multitasking, each program can
control the CPU for as long as it needs it. If a program is not using the CPU,
however, it can allow another program to use it temporarily. OS/2, Windows 95,
Windows NT, & UNIX use preemptive multitasking, whereas Microsoft Windows
3.x & the MultiFinder use cooperative multitasking.
1.2.5.4 Multi-user Operating System
Multiprogramming operating systems usually support multiple users, in which
case they are also called multi-user systems. Multi-user operating systems
provide facilities for maintenance of individual user environments & therefore
require user accounting. In general, multiprogramming implies multitasking, but
multitasking does not imply multi-programming. In effect, multitasking operation

Lesson No. 1 Intro. to Operating System 19

is one of the mechanisms that a multiprogramming OS employs in managing the
totality of computer-system resources, including processor, memory, & I/O
devices. Multitasking operation without multi-user support can be found in
operating systems of some advanced personal computers & in real-time
systems. Multi-access operating systems allow simultaneous access to a
computer system through two or more terminals. In general, multi-access
operation does not necessarily imply multiprogramming. An example is provided
by some dedicated transaction-processing systems, such as airline ticket
reservation systems, that support hundreds of active terminals under control of a
single program.
In general, the multiprocessing or multiprocessor operating systems manage the
operation of computer systems that incorporate multiple processors.
Multiprocessor operating systems are multitasking operating systems by
definition because they support simultaneous execution of multiple tasks
(processes) on different processors. Depending on implementation, multitasking
may or may not be allowed on individual processors. Except for management &
scheduling of multiple processors, multiprocessor operating systems provide the
usual complement of other system services that may qualify them as time-
sharing, real-time, or a combination operating system.
1.2.5.5 Multithreading
Allows different parts of a single program to run concurrently. The programmer
must carefully design the program in such a way that all the threads can run at
the same time without interfering with each other.
1.2.5.6 Time-sharing system
Time-sharing is a popular representative of multi-programmed, multi-user
systems. In addition to general program-development environments, many large
computer-aided design & text-processing systems belong to this category. One
of the primary objectives of multi-user systems in general, & time-sharing in
particular, is good terminal response time. Giving the illusion to each user of
having a machine to oneself, time-sharing systems often attempt to provide
equitable sharing of common resources. For example, when the system is

Lesson No. 1 Intro. to Operating System 20

loaded, users with more demanding processing requirements are made to wait
longer.
This philosophy is reflected in the choice of scheduling algorithm. Most time-
sharing systems use time-slicing scheduling. In this approach, programs are
executed with rotating priority that increases during waiting & drops after the
service is granted. In order to prevent programs from monopolizing the
processor, a program executing longer than the system-defined time slice is
interrupted by the OS & placed at the end of the queue of waiting programs. This
mode of operation generally provides quick response time to interactive
programs. Memory management in time-sharing systems provides for isolation &
protection of co-resident programs. Some forms of controlled sharing are
sometimes provided to conserve memory & possibly to exchange data between
programs. Being executed on behalf of different users, programs in time-sharing
systems generally do not have much need to communicate with each other.
As in most multi-user environments, allocation & de-allocation of devices
must be done in a manner that preserves system integrity & provides for good
performance.
1.2.5.7 Real-time systems
Real time systems are used in time critical environments where data must be
processed extremely quickly because the output influences immediate decisions.
Real time systems are used for space flights, airport traffic control, industrial
processes, sophisticated medical equipments, telephone switching etc. A real
time system must be 100 percent responsive in time. Response time is
measured in fractions of seconds. In real time systems the correctness of the
computations not only depends upon the logical correctness of the computation
but also upon the time at which the results is produced. If the timing constraints
of the system are not met, system failure is said to have occurred. Real-time
operating systems are used in environments where a large number of events,
mostly external to the computer system, must be accepted & processed in a
short time or within certain deadlines.

Lesson No. 1 Intro. to Operating System 21

A primary objective of real-time systems is to provide quick event-response
times, & thus meet the scheduling deadlines. User convenience & resource
utilization are of secondary concern to real-time system designers. It is not
uncommon for a real-time system to be expected to process bursts of thousands
of interrupts per second without missing a single event. Such requirements
usually cannot be met by multi-programming alone, & real-time operating
systems usually rely on some specific policies & techniques for doing their job.
The Multitasking operation is accomplished by scheduling processes for
execution independently of each other. Each process is assigned a certain level
of priority that corresponds to the relative importance of the event that it services.
The processor is normally allocated to the highest-priority process among those
that are ready to execute. Higher-priority processes usually preempt execution of
the lower-priority processes. This form of scheduling, called priority-based
preemptive scheduling, is used by a majority of real-time systems. Unlike, say,
time-sharing, the process population in real-time systems is fairly static, & there
is comparatively little moving of programs between primary & secondary storage.
On the other hand, processes in real-time systems tend to cooperate closely,
thus necessitating support for both separation & sharing of memory. Moreover,
as already suggested, time-critical device management is one of the main
characteristics of real-time systems. In addition to providing sophisticated forms
of interrupt management & I/O buffering, real-time operating systems often
provide system calls to allow user processes to connect themselves to interrupt
vectors & to service events directly. File management is usually found only in
larger installations of real-time systems. In fact, some embedded real-time
systems, such as an onboard automotive controller, may not even have any
secondary storage. The primary objective of file management in real-time
systems is usually speed of access, rather then efficient utilization of secondary
storage.
1.2.5.8 Combination of operating systems
Different types of OS are optimized or geared up to serve the needs of specific
environments. In practice, however, a given environment may not exactly fit any

Lesson No. 1 Intro. to Operating System 22

of the described molds. For instance, both interactive program development &
lengthy simulations are often encountered in university computing centers. For
this reason, some commercial operating systems provide a combination of
described services. For example, a time-sharing system may support interactive
users & also incorporate a full-fledged batch monitor. This allows computationally
intensive non-interactive programs to be run concurrently with interactive
programs. The common practice is to assign low priority to batch jobs & thus
execute batched programs only when the processor would otherwise be idle. In
other words, batch may be used as a filler to improve processor utilization while
accomplishing a useful service of its own. Similarly, some time-critical events,
such as receipt & transmission of network data packets, may be handled in real-
time fashion on systems that otherwise provide time-sharing services to their
terminal users.
1.2.5.9 Distributed Operating Systems
A distributed computer system is a collection of autonomous computer systems
capable of communication & cooperation via their hardware & software
interconnections. Historically, distributed computer systems evolved from
computer networks in which a number of largely independent hosts are
connected by communication links & protocols. A distributed OS governs the
operation of a distributed computer system & provides a virtual machine
abstraction to its users. The key objective of a distributed OS is transparency.
Ideally, component & resource distribution should be hidden from users &
application programs unless they explicitly demand otherwise. Distributed
operating systems usually provide the means for system-wide sharing of
resources, such as computational capacity, files, & I/O devices. In addition to
typical operating-system services provided at each node for the benefit of local
clients, a distributed OS may facilitate access to remote resources,
communication with remote processes, & distribution of computations. The
added services necessary for pooling of shared system resources include global
naming, distributed file system, & facilities for distribution.

Lesson No. 1 Intro. to Operating System 23

1.2.5.6 SYSTEM CALLS
System calls are kernel level service routines for implementing basic operations
performed by the operating system. Below are mentioned some of several
generic system calls that most operating systems provide.
CREATE (processID, attributes);
In response to the CREATE call, the OS creates a new process with the
specified or default attributes & identifier. A process cannot create itself-because
it would have to be running in order to invoke the OS, & it cannot run before
being created. So a process must be created by another process. In response to
the CREATE call, the OS obtains a new PCB from the pool of free memory, fills
the fields with provided and/or default parameters, & inserts the PCB into the
ready list-thus making the specified process eligible to run. Some of the
parameters definable at the process-creation time include: (a) Level of privilege,
such as system or user (b) Priority (c) Size & memory requirements (d) Maximum
data area and/or stack size (e) Memory protection information & access rights (f)
Other system-dependent data
Typical error returns, implying that the process was not created as a result of this
call, include: wrongID (illegal, or process already active), no space for PCB
(usually transient; the call may be retries later), & calling process not authorized
to invoke this function.
DELETE (process ID);
DELETE invocation causes the OS to destroy the designated process & remove
it from the system. A process may delete itself or another process. The OS
reacts by reclaiming all resources allocated to the specified process, closing files
opened by or for the process, & performing whatever other housekeeping is
necessary. Following this process, the PCB is removed from its place of
residence in the list & is returned to the free pool. This makes the designated
process dormant. The DELETE service is normally invoked as a part of orderly
program termination.

Lesson No. 1 Intro. to Operating System 24

To relieve users of the burden & to enhance probability of programs across
different environments, many compilers compile the last END statement of a
main program into a DELETE system call.
Almost all multiprogramming operating systems allow processes to terminate
themselves, provided none of their spawned processes is active. OS designers
differ in their attitude toward allowing one process to terminate others. The issue
here is none of convenience & efficiency versus system integrity. Allowing
uncontrolled use of this function provides a malfunctioning or a malevolent
process with the means of wiping out all other processes in the system. On the
other hand, terminating a hierarchy of processes in a strictly guarded system
where each process can only delete itself, & where the parent must wait for
children to terminate first, could be a lengthy operation indeed. The usual
compromise is to permit deletion of other processes but to restrict the range to
the members of the family, to lower-priority processes only, or to some other
subclass of processes.
Possible error returns from the DELETE call include: a child of this process is
active (should terminate first), wrongID (the process does not exist), & calling
process not authorized to invoke this function.
Abort (processID);
ABORT is a forced termination of a process. Although a process could
conceivably abort itself, the most frequent use of this call is for involuntary
terminations, such as removal of a malfunctioning process from the system. The
OS performs much the same actions as in DELETE, except that it usually
furnishes a register & memory dump, together with some information about the
identity of the aborting process & the reason for the action. This information may
be provided in a file, as a message on a terminal, or as an input to the system
crash-dump analyzer utility. Obviously, the issue of restricting the authority to
abort other processes, discussed in relation to the DELETE, is even more
pronounced in relation to the ABORT call.
Error returns for ABORT are practically the same as those listed in the discussion
of the DELETE call.

Lesson No. 1 Intro. to Operating System 25

FORK/JOIN
Another method of process creation & termination is by means of the
FORK/JOIN pair, originally introduced as primitives for multiprocessor systems.
The FORK operation is used to split a sequence of instructions into two
concurrently executable sequences. After reaching the identifier specified in
FORK, a new process (child) is created to execute one branch of the forked code
while the creating (parent) process continues to execute the other. FORK usually
returns the identity of the child to the parent process, & the parent can use that
identifier to designate the identity of the child whose termination it wishes to await
before invoking a JOIN operation. JOIN is used to merge the two sequences of
code divided by the FORK, & it is available to a parent process for
synchronization with a child.
The relationship between processes created by FORK is rather symbiotic in the
sense that they execute from a single segment of code, & that a child usually
initially obtains a copy of the variables of its parent.
SUSPEND (processKD);
The SUSPEND service is called SLEEP or BLOCK in some systems. The
designated process is suspended indefinitely & placed in the suspended state. It
does, however, remain in the system. A process may suspend itself or another
process when authorized to do so by virtue of its level of privilege, priority, or
family membership. When the running process suspends itself, it in effect
voluntarily surrenders control to the operating system. The OS responds by
inserting the target process's PCB into the suspended list & updating the PCB
state field accordingly.
Suspending a suspended process usually has no effect, except in systems that
keep track of the depth of suspension. In such systems, a process must be
resumed at least as many times as if was suspended in order to become ready.
To implement this feature, a suspend-count field has to be maintained in each
PCB. Typical error returns include: process already suspended, wrongID, & caller
not authorized.
RESUME (processID)

Lesson No. 1 Intro. to Operating System 26

The RESUME service is called WAKEUP is some systems. This call resumes the
target process, which is presumably suspended. Obviously, a suspended
process cannot resume itself, because a process must be running to have its OS
call processed. So a suspended process depends on a partner process to issue
the RESUME. The OS responds by inserting the target process's PCB into the
ready list, with the state updated. In systems that keep track of the depth of
suspension, the OS first increments the suspend count, moving the PCB only
when the count reaches zero.
The SUSPEND/RESUME mechanism is convenient for relatively primitive &
unstructured form of inter-process synchronization. It is often used in systems
that do not support exchange of signals. Error returns include: process already
active, wrongID, & caller not authorized.
DELAY (processID, time);
The system call DELAY is also known as SLEEP. The target process is
suspended for the duration of the specified time period. The time may be
expressed in terms of system clock ticks that are system-dependent & not
portable, or in standard time units such as seconds & minutes. A process may
delay itself or, optionally, delay some other process.
The actions of the OS in handling this call depend on processing interrupts from
the programmable interval timer. The timed delay is a very useful system call for
implementing time-outs. In this application a process initiates an action & puts
itself to sleep for the duration of the time-out. When the delay (time-out) expires,
control is given back to the calling process, which tests the outcome of the
initiated action. Two other varieties of timed delay are cyclic rescheduling of a
process at given intervals (e.g,. running it once every 5 minutes) & time-of-day
scheduling, where a process is run at a specific time of the day. Examples of the
latter are printing a shift log in a process-control system when a new crew is
scheduled to take over, & backing up a database at midnight.
The error returns include: illegal time interval or unit, wrongID, & called not
authorized. In Ada, a task may delay itself for a number of system clock ticks

Lesson No. 1 Intro. to Operating System 27

(system-dependent) or for a specified time period using the pre-declared floating-
point type TIME. The DELAY statement is used for this purpose.
GET_ATTRIBUTES (processID, attribute_set);
GET_ATTRIBUTES is an inquiry to which the OS responds by providing the
current values of the process attributes, or their specified subset, from the PCB.
This is normally the only way for a process to find out what its current attributes
are, because it neither knows where its PCB is nor can access the protected OS
space where the PCBs are usually kept.
This call may be used to monitor the status of a process, its resource usage &
accounting information, or other public data stored in a PCB. The error returns
include: no such attribute, wrongID, & caller not authorized. In Ada, a task may
examine the values of certain task attributes by means of reading the pre-
declared task attribute variables, such as T'ACTIVE, T'CALLABLE, T'PRIORITY,
& T'TERMINATED, where T is the identity of the target task.
CHANGE_PRIORITY (processID, new_priority);
CHANGE_PRIORITY is an instance of a more general
SET_PROCESS_ATTRIBUTES system call. Obviously, this call is not
implemented in systems where process priority is static.
Run-time modifications of a process's priority may be used to increase or
decrease a process's ability to compete for system resources. The idea is that
priority of a process should rise & fall according to the relative importance of its
momentary activity, thus making scheduling more responsive to changes of the
global system state. Low-priority processes may abuse this call, & processes
competing with the OS itself may corrupt the whole system. For these reasons,
the authority to increase priority is usually restricted to changes within a certain
range. For example, maximum may be specified, or the process may not exceed
its parent's or group priority. Although changing priorities of other processes
could be useful, most implementations restrict the calling process to manipulate
its own priority only.

Lesson No. 1 Intro. to Operating System 28

The error returns include: caller not authorized for the requested change & wrong
ID. In Ada, a task may change its own priority by calling the SET_PRIORITY
procedure, which is pre-declared in the language.
1.4 SUMMARY
Operating system is also known as resource manager because its prime
responsibility is to manage the resources of the computer system i.e. memory,
processor, devices and files. In addition to these, operating system provides an
interface between the user and the bare machine. Following the course of the
conceptual evolution of operating systems, we have identified the main
characteristics of the program-execution & development environments provided
by the bare machine, serial processing, including batch & multiprogramming.
On the basis of their attributes & design objectives, different types of operating
systems were defined & characterized with respect to scheduling & management
of memory, devices, & files. The primary concerns of a time-sharing system are
equitable sharing of resources & responsiveness to interactive requests. Real-
time operating systems are mostly concerned with responsive handling of
external events generated by the controlled system. Distributed operating
systems provide facilities for global naming & accessing of resources, for
resource migration, & for distribution of computation.
Typical services provided by an OS to its users were presented from the point of
view of command-language users & system-call users. In general, system calls
provide functions similar to those of the command language but allow finer
gradation of control.
1.5. SELF ASSESMENT QUESTIONS (SAQ)
1. What are the objectives of an operating system? Discuss.
2. Discuss modular approach of development of an operating system.
3. Present a hierarchical structure of an operating system.
4. What is an extended machine view of an operating system?
5. Discuss whether there are any advantages of using a multitasking
operating system, as opposed to a serial processing one.
6. What are the major functions performed by an operating system? Explain.

Lesson No. 1 Intro. to Operating System 29

1.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.
3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company
Ltd., New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 30

Lesson number: 2 Writer: Dr. Rakesh Kumar
File System - I Vetter: Prof. Dharminder Kr.

2.0 Objectives
A file is a logical collection of information and file system is a collection of files. The
objective of this lesson is to discuss the various concepts of file system and make the
students familiar with the different techniques of file allocation and access methods. We also
discuss the ways to handle file protection, which is necessary in an environment where
multiple users have access to files and where it is usually desirable to control by whom and
in what ways files may be accessed.
2.1 Introduction
The file system is the most visible aspect of an operating system. While the memory
manager is responsible for the maintenance of primary memory, the file manager is
responsible for the maintenance of secondary storage (e.g., hard disks). It provides the
mechanism for on-line storage of and access to both data and programs of the operating
system and all the users of the computer system. The file system consists to two distinct
parts: a collection of files, each storing related data and a directory structure, which
organizes and provides information about all the files in the system. Some file systems have
a third part, partitions, which are used to separate physically or logically large collections of
directories.
Nutt describes the responsibility of the file manager and defines the file, the
fundamental abstraction of secondary storage:
"Each file is a named collection of data stored in a device. The file manager
implements this abstraction and provides directories for organizing files. It also
provides a spectrum of commands to read and write the contents of a file, to set the
file read/write position, to set and use the protection mechanism, to change the
ownership, to list files in a directory, and to remove a file...The file manager provides
a protection mechanism to allow machine users to administer how processes
executing on behalf of different users can access the information in files. File
protection is a fundamental property of files because it allows different people to

Lesson No. 1 Intro. to Operating System 31

store their information on a shared computer, with the confidence that the
information can be kept confidential."
2.2 Presentation of Contents
2.2.1 File Concepts
2.2.1.1 File Operations
2.2.1.2 File Naming
2.2.1.3 File Types
2.2.1.4 Symbolic Link
2.2.1.5 File Sharing & Locking
2.2.1.6 File-System Structure
2.2.1.7 File-System Mounting
2.2.1.8 File Space Allocations
2.2.1.8.1 Contagious Space Allocation
2.2.1.8.2 Linked Allocation
2.2.1.8.3 Indexed Allocation
2.2.1.8.4 Performance
2.2.1.9 File Attributes
2.2.2 Access Methods
2.2.2.1 Sequential Access
2.2.2.2 Index-sequential
2.2.2.3 Direct Access
2.2 PRESENTATION OF CONTENTS
2.2.1 FILE CONCEPTS
The most important function of an operating system is the effective management of
information. The modules of the operating system dealing with the management of
information are known as file system. The file system provides the mechanism for online
storage and access to both data and programs. The file system resides permanently on
secondary storage, which has the main requirement that it must be able to hold a large
amount of data, permanently. The desirable features of a file system are:
4. Minimal I/O operations.
5. Flexible file naming facilities.

Lesson No. 1 Intro. to Operating System 32

 6. Automatic allocation of file space.
7. Dynamic allocation of file space.
8. Unrestricted flexibility between logical record size and physical block size.
9. Protection of files against illegal forms of access.
10. Static and dynamic sharing of files.
11. Reliable storage of files.
This lesson is primarily concerned with issues concerning file storage and access on the
most common secondary storage medium, the disk.
A file is a collection of related information units (records) treated as a unit. A record is itself
a collection of related data elements (fields) treated as a unit. A field contains a single data
item.
So file processing refers to reading/writing of records in a file and processing of the
information in the fields of a record.
2.2.1.1 File operations
Major file operations are performed are as follows:
¾ Read operation: Read information contained in the file.
¾ Write operation: Write new information into a file at any point or overwriting
existing information in a file.
¾ Deleting file: Delete a file and release its storage space for use in other files.
¾ Appending file: Write new information at the end of a file.
¾ Execute
¾ Coping file
¾ Renaming file
¾ Moving file
¾ Creating file
¾ Merging files
¾ Sorting file
¾ Comparing file
2.2.1.2 File Naming
Each file is a distinct entity and therefore a naming convention is required to distinguish one
from another. The operating systems generally employ a naming system for this purpose. In

Lesson No. 1 Intro. to Operating System 33

fact, there is a naming convention to identify each resource in the computer system and not
files alone.
2.2.1.3 File Types
The files under UNIX can be categorized as follows:
¾ Ordinary files.
¾ Directory files.
¾ Special files.
¾ FIFO files.
Ordinary Files
Ordinary files are the one, with which we all are familiar. They may contain executable
programs, text or databases. You can add, modify or delete them or remove the file entirely.
Directory Files
Directory files, as discussed earlier also represent a group of files. They contain list of file
names and other information related to these files. Some of the commands, which manipulate
these directory files, differ from those for ordinary files.
Special Files
Special files are also referred to as device files. These files represent physical devices such as
terminals, disks, printers and tape-drives etc. These files are read from or written into just
like ordinary files, except that operation on these files activates some physical devices. These
files can be of two types (i) character device files and (ii) block device file. In character
device files data are handled character by character, as in case of terminals and printers. In
block device files, data are handled in large chunks of blocks, as in the case of disks and
tapes.
FIFO Files
FIFO (first-in-first-out) are files that allow unrelated processes to communicate with each
other. They are generally used in applications where the communication path is in only one
direction, and several processes need to communicate with a single process. For an example
of FIFO file, take the pipe in UNIX. This allows transfer of data between processes in a first-
in-first-out manner. A pipe takes the output of the first process as the input to the next
process, and so on.
2.2.1.4 Symbolic Link
A link is effectively a pointer or an alias to another file or subdirectory. For example, a link
may be implemented as an absolute or relative path name (a symbolic link). When a

Lesson No. 1 Intro. to Operating System 34

reference to a file is made, we search the directory. The directory entry is marked as a link
and the name of the real file (or directory) is given. We resolve the link by using the path
name to locate the real file. Links are easily identified by their format in the directory entry
(or by their having a special type on systems that support types), and are effectively named
indirect pointers.
A symbolic link can be deleted without deleting the actual file it links. There can be any
number of symbolic links attached to a single file.
Symbolic links are helpful in sharing a single file called by different names. Each time a link
is created, the reference count in its inode is incremented by one. Whereas deletion of link
decreases file the count by one. The operating system denies deletion of such files whose
reference count is not 0, thereby meaning that the file is in use.
In a system where sharing is implemented by symbolic links, this situation is somewhat
easier to handle. The deletion of a link does not need to affect the original file; only the link
is removed. If the file entry itself is deleted, the space for the file is deallocated, leaving the
links dangling. We can search for these links and remove them also, but unless a list of the
associated link is kept with each file, this search can be expensive. Alternatively, we can
leave the links until an attempt is made to use them. At that time, we can determine that the
file of the name given by the link does not exist, and can fail to resolve the link name; the
access is treated just like any other illegal file name. (In this case, the system designer should
consider carefully what to do when a file is deleted and another file of the same name is
created, before a symbolic link to the original file is used.) In the case of UNIX, symbolic
links are left when a file is deleted, and it is up to the user to realize that the original file is
gone or has been replaced.
Another approach to deletion is to preserve the file until all references to it are deleted. To
implement this approach, we must have some mechanism for determining that the last
reference to the file has been deleted. We could keep a list of all references to a file
(directory entries or symbolic links). When a link or a copy of the directory entry is
established, a new entry is added to the file-reference list. When a link or directory entry is
deleted, we remove its entry on the list. The file is deleted when its file-reference list is
empty.

Lesson No. 1 Intro. to Operating System 35

The trouble with this approach is the variable and potentially large size of the file-reference
list. However, we really do not need to keep the entire list - we need to keep only a count of
the number of references. A new link or directory entry increments the reference counts;
deleting a link or entry decrements the count. When the count is 0, the file can be deleted;
there are no remaining references to it. The UNIX operating system uses this approach for
non-symbolic links, or hard links, keeping a reference count in the file information block or
inode). By effectively prohibiting multiple references to directories, we maintain an acyclic-
graph structure.
To avoid these problems, some systems do not allow shared directories links. For example,
in MS-DOS, the directory structure is a tree structure.
2.2.1.5 File Sharing and Locking
The owner of a file uses the access control list of the file to authorize some other users to
access the file. In a multi-user environment a file is required to be shared among more than
one user. There are several techniques and approaches to affect this operation. File sharing
can occur in two modes (i) sequential sharing and (ii) concurrent sharing. Sequential sharing
occurs when authorized users access a shared file one after another. So any change made by
a user is reflected to other users also. Concurrent sharing occurs when two or more users
access a file over the same period of time. Concurrent sharing may be implemented in one of
the following three forms:
(a) Concurrent sharing using immutable files: In it any program cannot modify the file being
shared.
(b) Concurrent sharing using single image mutable files: An image is a view of a file. All
programs concurrently sharing the file see the same image of the file. So changes made
by one program are also visible to other programs sharing the file.
(c) Concurrent sharing using multiple image mutable files: Each program accessing the file
has its own image of the file. So many versions of the file at a time may exist and updates
made by a user may not be visible to some concurrent user.
There are three different modes to share a file:
¾ Read only: In this mode the user can only read or copy the file.
¾ Linked shared: In this mode all the users sharing the file can make changes in
this file but the changes are reflected in the order determined by the operating

Lesson No. 1 Intro. to Operating System 36

 systems.
¾ Exclusive mode: In this mode a single user who can make the changes (while
others can only read or copy it) acquires the file.
Another approach is to share a file through symbolic links. This approach poses a couple of
problems - concurrent updation problem, deletion problem. If two users try to update the
same file, the updating of one of them will be reflected at a time. Besides, another user must
not delete a file while it is in use.
File locking gives processes the ability to implement mutually exclusive access to a file.
Locking is mechanism through which operating systems ensure that the user making changes
to the file is the one who has the lock on the file. As long as the lock remains with this user,
no other user can alter the file. Locking can be limited to files as a whole or parts of a file.
Locking may apply to any access or different levels of locks may exist such as read/write
locks etc.
2.2.1.6 File-System Structure
Disks provide the bulk of secondary storage on which a file system is maintained. To
improve I/O efficiency, I/O transfers between memory and disks are performed in units of
blocks. Each block is one or more sectors. Depending on the disk drive, sectors vary
from 32 bytes to 4096 bytes; usually, they are 512 bytes. The blocking method
determines how a file’s records are allocated into blocks:
Fixed blocking: An integral number of fixed-size records are stored in each block. No
record may be larger than a block.
Unspanned blocking: Multiple variable size records can be stored in each block but no
record may span multiple blocks.
Spanned blocking: Records may be stored in multiple blocks. There is no limit on the
size of a record.
Disks have two important characteristics that make them a convenient medium for
storing multiple files:
(a) They can be rewritten in place; it is possible to read a block from the disk, to modify the
block, and to write it back into the same place.
(b) One can access directly any given block of information on the disk. Thus, it is simple to
access any file either sequentially or randomly, and switching from one file to another
added requires only moving the read-write heads and waiting for the disk to rotate.

Lesson No. 1 Intro. to Operating System 37

To provide an efficient and convenient access to the disk, the operating system imposes a file
system to allow the data to be stored, located, and retrieved easily. A file system poses two
quite different design problems.
(a) How the file system should look to the user? This task involves the definition of a file
and its attributes, operations allowed on a file and the directory structure for organizing
the files.
(b) Algorithms and data structure must be created to map the logical file system onto the
physical secondary storage devices.
2.2.1.7 File-System Mounting
Just as a file must be opened before it is used, a file system must be mounted before it can be
available to processes on the system. The mount procedure is straightforward. The operating
system is given the name of the device and the location within the file structure at which to
attach the file system (called the mount point). For instance, on the UNIX system, a file
system containing user’s home directory might be mounted as /home; then, to access the
directory structure within that file system, one could precede the directory names with
/home, as in /home/jane. Mounting that file system under /users would result in the path
name /users/jane to reach the same directory.
Next, the operating system verifies that the device contains a valid file system. It does so by
asking the device driver to read the device directory and verifying that the directory has the
expected format. Finally, the operating system notes its directory structure that a file system
is mounted at the specified mount point. This scheme enables the operating system to
traverse its directory structure, switching among file systems as appropriate. Consider the
actions of the Macintosh Operating System.
Whenever the system encounters a disk for the first time (hard disks are found at boot time,
floppy disks ate seen when they are inserted into the drive), the Macintosh Operating System
searches for a file system on the device. If it finds one, it automatically mounts the file
system at the boot-level, adds a folder icon to the screen labeled with the name of the file
system (as stored in the device directory). The user is then able to click on the icon and thus
to display the newly mounted file system.
2.2.1.8 File space allocations

Lesson No. 1 Intro. to Operating System 38

The direct-access nature of disks allows flexibility in the implementation of files. In
almost every case, many files will be stored on the same disk. The main problem is how
to allocate space to these files so that disk space is utilized effectively and files can be
accessed quickly. There are three major methods of allocating disk space:
(a) Contiguous space allocation
(b) Linked allocation
(c) Indexed allocation
Each method has its advantages and disadvantages. Accordingly some systems support all
three. More common system will use one particular method for all files.
2.2.1.8.1 Contiguous space Allocation
The simplest scheme is contiguous allocation. The logical blocks pf a file are stored in a
partition of contiguous physical blocks. Disk addresses define a linear ordering on the disk.
With this ordering, assuming that only one job is accessing the disk, accessing block b+1
after block b normally requites no head movement. When head movement is needed (from
the last sector of one cylinder to the first sector of the next cylinder), it is only one-track
movement. Thus, the number of disk seeks required for accessing contiguously allocated
files is minimal. The disk address and length (in block units) of the first block define
contiguous allocation of the file. If the file is n blocks long, and starts at location b, then it
occupies block b, b+1, b+2, ..., b+n-1. The directory entry for each file indicates the address
of the starting block and the length of the area allocated for this file. Accessing a file that has
been allocated contiguously is easy. For sequential access, the file system remembers the
disk address of the last block referenced and, when necessary, reads the next block. For
direct access to block i of a file that starts at block b, we can immediately access block b+i.
So contiguous space allocation easily supports both sequential and direct access.

User Directory
File Locations Length

Lesson No. 1 Intro. to Operating System 39

 Data
Data
Data
.
.
.
Data
Data

The major problem with contiguous allocation is locating the space for a new file. The
contiguous disk space-allocation problem can be seen to be particular application of the
general dynamic storage-allocation problem, which is how to satisfy a request of size n
from a list of free holes. First-fit (This strategy allocates the first available space that is
big enough to accommodate file. Search may start at beginning of set of holes or where
previous first-fit ended. Searching stops as soon as it finds a free hole that is large
enough) and best-fit (This strategy allocates the smallest hole that is big enough to
accommodate file. Entire list ordered by size is searched & matching smallest left over
hole is chosen) are the most common strategies used to select a free hole from the set of
available holes. Simulations have shown that both first-fit and best-fit are more efficient
than worst-fit (This strategy allocates the largest hole. Entire list is searched. It chooses
largest left over hole) in terms of both time and storage utilization. Neither first-fit nor
best-fit is clearly best in terms of storage utilization, but first-fit is generally faster.
These algorithms suffer from the problem of external fragmentation i.e. the tendency to
develop a large number of small holes. As files are allocated and deleted, the free disk space
is broken into little pieces. External fragmentation exists whenever free space is broken into
chunks. It becomes a problem when the largest contiguous chunk is insufficient for a request;
storage is fragmented into a number of holes, no one of which is large enough to store the
data. Depending on the total amount of disk storage and the average file size, external
fragmentation may be either a minor or a major problem.
Some older microcomputer systems used contiguous allocation on floppy disks. To prevent
loss of significant amounts of disk space to external fragmentation, the user had to run a

Lesson No. 1 Intro. to Operating System 40

repacking routine that copied the entire file system onto another floppy disk or onto a tape.
The original floppy disk was then freed completely, creating one large contiguous free space.
The routine then copied the files back onto the floppy disk by allocating contiguous space
from this one large hole.
The scheme effectively compacts all free space into one contiguous space, solving the
fragmentation problem. The cost of this compaction is time. The time cost is particularly
severe for large hard disks that use contiguous allocation, where compacting all the space
may take hours and may be necessary on a weekly basis. During this down time, normal
system operation generally cannot be permitted, so such compaction is avoided at all costs
on production machines.
This is not all, there are other problems with contiguous allocation. A major problem is
determining how much space is needed for a file. When the file is created, the total amount
of space it will need must be found and allocated. How does the creator (program or person)
know the size of the file to be created? In some cases, this determination may be fairly
simple (copying an existing file, for example); in general, however, the size of an output file
may be difficult to estimate.
If too little space is allocated to a file, it may be found that file cannot be extended.
Especially with only a best-fit allocation strategy, the space on both sides of the file may be
in use. Hence, we cannot make the file larger in space. Two possibilities then exist. First, the
user program can be terminated, with an appropriate error message. The user must then
allocate more space and run the program again. These repeated runs may prove costly. To
prevent them, the user will normally overestimate the amount of space needed, resulting in
considerable wasted space.
The other possibility is to find a larger hole, to copy the contents of the file to the new space
and release the previous space. This series of actions may be repeated as long as space exists,
although it can also be time-consuming. Notice, however, that in this case the user never
needs to be informed explicitly about what is happening; the system continues despite the
problem, although more and more slowly.
Even if the total amount of space needed for a file is known in advance, pre-allocation may
be inefficient. A file that grows slowly over a long period (months or years) must be

Lesson No. 1 Intro. to Operating System 41

allocated enough space for its final size, even though much of that space may be unused for a
long time. The file, therefore, has a large amount of internal fragmentation.
To avoid several of these drawbacks, some operating systems use a modified contiguous
allocation scheme, in which a contiguous chunk of space is allocated initially, and then,
when that amount is not large enough, another chunk of contiguous space, called an extent, is
added to the initial allocation. The location of a file's blocks is then recorded as a location
and a block count, plus a link to the first block of the next extent. On some systems, the
owner of the file can set the extent size, but this setting results in inefficiencies if the owner
is incorrect. Internal fragmentation can still be a problem if the extents are too large, and
external fragmentation can be a problem as extents of varying sizes are allocated and de-
allocated in turn.
2.2.1.8.2 Linked Allocation
In linked allocation, file is not stored on a contiguous set of blocks, rather the physical blocks
in which a file is stored may be scattered throughout the secondary storage devices. Linked
allocation solves all problems of contiguous allocation. With linked allocation, each file is a
linked list of disk blocks; the disk blocks may be scattered anywhere on the disk. The
directory contains a pointer to the first and last blocks of the file. For example, a file of five
blocks might start at block 9, continue at block 16, then block 1, block 10, and finally block
25. Each block contains a pointer to the next block. These pointers are not made available to
the user. Thus, if each block is 512 bytes, and a disk address (the pointer) requires 4 bytes,
then the user sees blocks of 508 bytes. To create anew file, we simply create a new entry in
the directory. With linked allocation, each directory entry has a pointer to the first disk block
of the file. This pointer is initialised to nil (the end-of-list pointer value) to signify an empty
file. The size field is also set to 0. A write to the file causes a free block to be found via the
free-space management system, and this new block is then written to, and is linked to the end
of the file. To read a file, we simply read blocks by following the pointers from block to
block.
There is no external fragmentation with linked allocation, and any free block on the free-
space list can be used to satisfy a request. Notice also that there is no need to declare the size
of a file when a file is created. A file can continue to grow as long as there are free blocks.
Consequently, it is never necessary to compact disk space. Linked allocation suffers from

Lesson No. 1 Intro. to Operating System 42

some disadvantages, however. The major problem is that it can be used effectively for only
sequential-access files. To find the ith block of a file, we must start at the beginning of that
file, and follow the pointers until we get to the ith block. Each access to a pointer requires a
disk read, and sometimes a disk seek also. Consequently, it is inefficient to support a direct-
access capability for linked allocation files.

User Directory
File Location

Data Data

Data Data

Space required for the pointers is another disadvantage to linked allocation. If a pointer
requires 4 bytes out of a 512-byte block, then ((4 / 512) * 100 = 0.78) percent of the disk is
being used for pointers, rather than for information. Each file requires slightly more space
than it otherwise would. The usual solution to this problem is to collect blocks into multiples,
called clusters, and to allocate the clusters rather than blocks. For instance, the file system
may define a cluster as 4 blocks, and operate on the disk in only cluster units. Pointers then
use a much smaller percentage of the file's disk space. This method allows the logical-to-
physical block mapping to remain simple, but improves disk throughput (fewer disk head-
seeks) and decreases the space needed for block allocation and free-list management. The
cost of this approach is an increase in internal fragmentation, because more space is wasted if
a cluster is partially fully than when a block is partially full. Clusters can be used to improve
the disk access time for many other algorithms, so they are used in most operating systems.

Lesson No. 1 Intro. to Operating System 43

Yet another problem is reliability. Since the files are linked together by pointers scattered all
over the disk, consider what would happen if a pointer were lost or damaged. A bug in the
operating- system software or a disk hardware failure might result in picking up the wrong
pointer. This error could result in linking into the free-space list or into another file. Partial
solutions are to use doubly linked lists or, to store the file name and relative block number in
each block; however, these schemes require even more overhead for each file.
An important variation on the linked allocation method is the use of a file-allocation table
(FAT). This simple but efficient method of disk-space allocation is used by the MS-DOS and
OS/2 operating systems. A section of disk at the beginning of each partition is reserved to
contain the table. The table has one entry for each disk block, and is indexed by block
number. The FAT is used much as is a linked list. The directory entry contains the block
number of the first block of the file. The table entry indexed by that block number then
contains the block number of the next block in the file. This chain continues until the last
block, which has a special end-of-file value as the table entry. Unused blocks are indicated
by a 0 table value. Allocating a new block to a file is a simple matter of finding the first 0-
valued table entry, and replacing the previous end-of-file value.
Note that the FAT allocation scheme can result in a significant number of disk head seeks,
unless the FAT is cached. The disk head must move to the start of the partition to read the
FAT and find the location of the block in question, then move to the location of the block
itself. In the worst case, both moves occur for each of the blocks. A benefit is that random
access time is improved, because the disk head can find the location of any block by reading
the information in the FAT.
2.2.1.8.3 Indexed Allocation
Although linked allocation solves the external-fragmentation and size-declaration problems
of contiguous allocation. However, in the absence of a FAT, linked allocation cannot support
efficient direct access, since the pointers to the blocks are scattered with the blocks
themselves all over the disk and need to be retrieved in order. Indexed allocation solves this
problem by bringing all the pointers together into one location, called the index block.
Indexed allocation is a variant of linked allocation.
Each file has its own index block, which is an array of disk-block addresses. The ith entry in
the index block points to the ith block of the file. The directory contains the address of the
index block (See following figure). To read the ith block, we use the pointer in the ith index-
block entry to find and read the desired block.

Lesson No. 1 Intro. to Operating System 44

When the file is created, all pointers in the index block are set to nil. When the ith block is
first written, a block is obtained from the free-space manager, and its address is put in
the ith index-block entry.
Indexed allocation supports direct access, without suffering from external fragmentation,
because any free block on the disk may satisfy a request for more space.
User Directory
File Location

Index block Index block

Data Data

Data Data

Data Data

Indexed allocation of disk space

Indexed allocation does suffer from wasted space. The pointer overhead of the index block is
generally greater than the pointer overhead of linked allocation.

Lesson No. 1 Intro. to Operating System 45

Note that indexed allocation schemes suffer from some of the same performance problems,
as does linked allocation. Specifically, the index blocks can be cached in memory, but the
data blocks may be spread all over a partition.
2.2.1.8.4 Performance
To evaluate the performance of allocation methods, two important criteria are storage
efficiency and data-block access times. Both are important criteria in selecting the proper
method or methods for an operating system to implement.
One difficulty in comparing in performance of the various systems is determining how the
systems will be used – in a sequential access manner or random access. A system with
mostly sequential access should use a method different from that for a system with mostly
random access. For any type of access, contiguous allocation requires only one access to get
a disk block. Since we can easily keep the initial address of the file in memory, we can
calculate immediately the disk address of the ith block (or the next block) and read it directly.
For linked allocation, we can also keep the address of the next block in memory and read it
directly. This method is fine for sequential access; for direct access, however, an access to
the ith block might require i disk reads. This problem indicates why linked allocation should
not be used for an application requiring direct access.
As a result, some systems support direct-access files by using contiguous allocation and
sequential access by linked allocation. For these systems, the type of access to be made must
be declared when the file is created. A file created for sequential access will be linked and
cannot be used for direct access. A file created for direct access will be contiguous and can
support both direct access and sequential access, but its maximum length must be declared
with it.
2.2.1.9 File attributes
Attributes are properties of a file. The operating system treats a file according to its
attributes. Following are a few common attributes of a file:
¾ H for hidden
¾ A for archive
¾ D for directory
¾ X for executable
¾ R for read only

Lesson No. 1 Intro. to Operating System 46

These attributes can be used in combination also.
2.2.2 ACCESS METHODS
Files store information, which is when required, may be read into the main memory. There
are several different ways in which the data stored in a file may be accessed for reading and
writing. The operating system is responsible for supporting these file access methods. The
fundamental methods for accessing information in the file are (a) sequential access: in it
information in the file must be accessed in the order it is stored in the file, (b) direct access,
and (c) index sequential access.
2.2.2.1 Sequential access
A sequential file is the most primitive of all file structures. It has no directory and no linking
pointers. The records are generally organized in a specific sequence according to the key
field. In other words, a particular attribute is chosen whose value will determine the order of
the records. Access proceeds sequentially from start to finish. Operations to read or write the
file need not specify the logical location within the file, because operating system maintains
a file pointer that determines the location of the next access. Sometimes when the attribute
value is constant for a large number of records a second key is chosen to give an order when
the first key fails to discriminate. Use of sequential file requires data to be sorted in a desired
sequence according to the key field before storing or processing them.
Its main advantages are:
¾ It is easy to implement
¾ It provides fast access to the next record if the records are to be accessed using
lexicographic order.
Its disadvantages are:
¾ It is difficult to update and insertion of a new record may require moving a large
proportion of the file
¾ Random access is extremely slow.
Sometimes a file is considered to be sequentially organised despite the fact that it is not
ordered according to any key. Perhaps the date of acquisition is considered to be the key
value, the newest entries are added to the end of the file and therefore pose no difficulty to
updating. Sequential files are advisable if the applications are sequential by nature.
2.2.2.2 Index-sequential

Lesson No. 1 Intro. to Operating System 47

An index-sequential file each record is supposed to have a unique key and the set of records
may be ordered sequentially by a key. An index is maintained to determine the location of a
record from its key value. Each key value appears in the index with the associated address of
its record. To access a record with key k, the index entry containing k is found by searching
the index and the disk address mentioned in the entry is used to access the record.
In the following figure an employee file is illustrated where records are arranged in
ascending order according to the employee #.
Track #
1 1 2 5 8 16 20 25 30 32 36
2 38 40 41 43 44 45 50 52
3 53 57 59 60 62 64 67 70
A track index is maintained as shown in the following figure to speed up the search:
Track Low High
1 1 36
2 38 52
3 53 70
For example, to locate the record of employee # 41, index is searched. It is evident from
the index that the record of employee #41 will be on track no.2 because it has the lowest
key value 48 and highest key value 52.
In the literature an index-sequential file is usually
thought of as a sequential file with a hierarchy of indices.
For example, there might be three levels of indexing: track,
cylinder and master. Each entry in the track index will
contain enough information to locate the start of the track,
and the key of the last record in the track, which is also
normally the highest value on that track. There is a track
index for each cylinder. Each entry in the cylinder index
gives the last record on each cylinder and the address of the
track index for that cylinder. If the cylinder index itself is
stored on tracks, then the master index will give the highest
key referenced for each track of the cylinder index and the

Lesson No. 1 Intro. to Operating System 48

starting address of that track. No mention has been made of
the possibility of overflow during an updating process.
Normally provision is made in the directory to administer an
overflow area. This of course increases the number of book-
keeping entries in each entry of the index.
2.2.2.3 Direct access
In direct access file organization, any records can be
accessed irrespective of the current position in the file.
Direct access files are created on direct access storage
devices. Whenever a record is to be inserted, its key value is
mapped into an address using a hashing function. On that
address record is stored. The advantage of direct access file
organization is realized when the records are to be accessed
randomly (not sequentially). Otherwise this organization has a
number of limitations such as (a) poor utilization of the I/O
medium and (b) Time consumption during record address
calculation.
2.3 Key words
Contiguous space Allocation: The logical blocks pf a file are stored in a partition of
contiguous physical blocks.
Linked allocation: In it each file is a linked list of disk blocks; the disk blocks may be
scattered anywhere on the disk. The directory contains a pointer to the first and last blocks of
the file.
Indexed Allocation: In Indexed allocation all the pointers together are stored into one
location, called the index block. Each file has its own index block, which is an array of disk-
block addresses.
Sequential access: In it information in the file must be accessed in the order it is stored in
the file.
Index-sequential: In index-sequential file each record is supposed to have a unique key and
the set of records may be ordered sequentially by a key. An index is maintained to determine
the location of a record from its key value.

Lesson No. 1 Intro. to Operating System 49

 Direct access: In direct access file organization, records
can be accessed randomly. The key value of the record is
mapped into an address using a hashing function. On that
address record is stored.
2.4 SUMMARY
The file system resides permanently on secondary storage, which has the main requirement
that it must be able to hold a large amount of data, permanently.
The various files can be allocated space on the disk in three ways: through contagious, linked
or indexed allocation. Contagious allocation can suffer from external fragmentation. Direct-
access is very inefficient with linked-allocation. Indexed allocation may require substantial
overhead for its index block. There are many ways in which these algorithms can be
optimised.
Free space allocation methods also influence the efficiency of the use of disk space, the
performance of the file system and the reliability of secondary storage.
2.5 SELF-ASSESSMENT QUESTIONS (SAQ)
1. What do you understand by a file? What is a file system?
2. What are the different modes to share a file?
3. What are the different methods to access the information from a file? Discuss their
advantages and disadvantages.
4. What are the advantages of indexed allocation over linked allocation and contiguous
space allocation? Explain.
5. Differentiate between first fit, best fit and worst fit storage allocation strategies.
2.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B., John Wiley &
Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition, Dhamdhere D.M.,
Tata McGraw Hill Publishing Company Ltd., New Delhi.
3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill Publishing
Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company Ltd., New

Lesson No. 1 Intro. to Operating System 50

 Delhi, 2002.

Lesson No. 1 Intro. to Operating System 51

Lesson number: 3 Writer: Dr. Rakesh Kumar
File System - II Vetter: Prof. Dharminder Kr.

3.0 Objectives
The objectives of this lesson are to make the students familiar with directory system and
file protection mechanism. After studying this lesson students will become familiar with:
(a) Different types of directory structures.
(b) Different protection structures such as:
- Access Control Matrix
- Access Control Lists
3.1 Introduction
A file system provides the following facilities to its users: (a) Directory structure and file
naming facilities, (b) Protection of files against illegal form of access, (c) Static and dynamic
sharing of files, and (d) Reliable storage of files. A file system helps the user in organizing
the files through the use of directories. A directory may be defined as an object that contains
the names of the file system objects. Entries in the directory determine the names associated
with a file system object. A directory contains information about a group of files. A typical
structure of a directory entry is as under:
File name – Locations Information – Protection Information – Flags
The presences of directories enable file system to support file sharing and protection.
Sharing is simply a matter of permitting a user to access the files of other user stored in
some other directory. Protection is implemented by permitting the owner of a file to
specify which other users may access his files and in what manner. All these issues are
discussed in detail in this lesson.
3.2 Presentation of Contents
3.2.1 Hierarchical Directory Systems
3.2.1.1 Directory Structure
3.2.1.2 The Logical Structure of a Directory
3.2.1.2.1 Single-level Directory
3.2.1.2.2 Two-level Directory
3.2.1.2.3 Tree-structured Directories

Lesson No. 1 Intro. to Operating System 52

 3.2.1.2.4 Acyclic-Graph Directories
3.2.1.2.5 General Graph Directory
3.2.1.3 Directory Operations
3.2.2 File Protection and Security
3.2.2.1 Type of Access
3.2.2.2 Protection Structure
3.2.2.2.1 Access Control Matrix
3.2.2.2.2 Access Lists and Groups
3.2.2.2.3 Other Protection Approaches
3.2.1 HIERARCHICAL DIRECTORY SYSTEMS
Files are generally stored on secondary storage devices. Numerous files are to be stored on
storage of giga-byte capacity. To handle such a huge size of data, there is a need to properly
organize the files. The organization, usually, done in two parts. In the first part, a file system
may incorporate the notion of a partition, which determines on which device a file will be
stored. The file system is broken into partitions, also known as minidisks or volumes.
Typically, a disk contains at least one partition, which is a low-level structure in which files
and directories reside. Sometimes, there may be more than one partition on a disk, each
partition acting as a virtual disk. The users do not have to concern themselves with the
translating the physical address; the system does the required job.

root

x y z

Z1 Z2 Z3

Z11

Z111 Z112 Z113

Lesson No. 1 Intro. to Operating System 53

Figure 1 Directory Hierarchy
Partitions contain information about itself in a file called partition table. It also contains
information about files and directories on it. Typical file information is name, size, type,
location etc. The entries are kept in a device directory or volume table of contents (VTOC).
Directories may be created within another directory. Directories have parent-child
relationship as shown in the above Figure 1.
3.2.1.1 Directory Structure
The file systems of computers can be extensive. Some systems store thousands of files on
hundreds of gigabytes of disk. To manage all these data, we need to organize them. This
organization is usually done in two parts; first, the file system is broken into in the IBM
world or volumes in the PC and Macintosh arenas. Sometimes, partitions are used to provide
several separate areas within one disk, each treated as a separate storage device, whereas
other systems allow partitions to be larger than a disk to group disks into one logical
structure. In this way, the user needs to be concerned with only the logical directory and file
structure, and can ignore completely the problems of physically allocating space for files.
For this reason partitions can be thought of as virtual disks.
Second, each partition contains information about files within it. This information is kept in a
device directory or volume table of contents. The device directory (more commonly known
simply as a "directory") records information such as name, location, size, and type for all
files on that partition.
3.2.1.2 The Logical Structure of a Directory
3.2.1.2.1 Single-Level Directory
The simplest directory structure is the single-level tree. A single level tree system has only
one directory. All files are contained in the same directory, which is easy to support and
understand. Names in that directory refer to files or other non-directory objects. Such a
system is practical only on systems with very limited numbers of files. A single-level
directory has significant limitations, when the number of files increases or when there is
more than one user. Since all files are stored in the same directory, the name given to each
file should be unique. If there are two users and they give the same name to their file, then
there is a problem.

Lesson No. 1 Intro. to Operating System 54

 ab x test yy ram Z2 pq Directory

Files

Figure 2 Single-level Directory

Even with a single user, as the number of files increase, it becomes difficult to remember the
names of all the files, so as to create only files with unique names. It is not uncommon for a
user to have hundreds of files on one computer system and an equal number of additional
files on another system. In such an environment, keeping track of so many files is a daunting
task.
3.2.1.2.2 Two-Level Directory
The disadvantage of a single-level directory is confusion of file names. The standard solution
is to create a separate directory for each user. In a two level system, only the root level
directory may contain names of directories and all other directories refer only to non-
directory objects. In the two-level directory structure, each user has his/her own user file
directory (UFD). Each UFD has a similar structure, but lists only the files of a single user.
When a user starts or a user logs in, the system's master file directory is searched. The master
file directory is indexed by user name or account.

Master file directory ajay mohan raj

green blue yellow pink xraj yraj User

dir

Lesson No. 1 Intro. to Operating System 55

Figure 3 Two-Level Directory

When in a UFD a user refers to a particular file, only his own UFD is searched. Thus,
different users may have files with the same name, as long as till the filenames within each
UFD are unique.
To create a file for a user, the operating system searches only that user's UFD to ascertain
whether another file of that name exists. To delete a file, the operating system confines its
search to the local UFD; thus, it cannot accidentally delete another user's file that has the
same name.
The user directories themselves must be created and deleted as necessary. A special system
program is run with the appropriate user name and account information. The program creates
a new user file directory and adds an entry for it to the master file directory. The execution of
this program might be restricted to system administrators.
The two-level directory structure solves the name-collision problem, but it still has problems.
This structure effectively isolates one user from another. This isolation is an advantage when
the users are completely independent, but is a disadvantage when the users co-operate on
some task and to access one user's account by other users is not allowed.
If access is to be permitted, one user must have the ability to name a file in another user's
directory.
A two-level directory can be thought of as a tree, or at least an inverted tree. The root of the
tree is the master file directory. Its direct descendants are the UFDs. The descendants of the
user file directories are the files themselves.
Thus, a user name and a file name define a path name. Every file in the system has a path
name. To name a file uniquely, user must know the path name of the file desired.
For example, if user A wishes to access her own test file named test, she can simply refer to
test. To access the test file of user B (with directory-entry name userb), however, she might
have to refer to /userb/test. Every system has its own syntax for naming files in directories
other than the user's own.
There is additional syntax to specify the partition of a file. For instance, in MS-DOS a letter
followed by a colon specifies a partition. Thus, file specification might be "C:\userb\bs.test".

Lesson No. 1 Intro. to Operating System 56

Some systems go even further and separate the partition, directory name, and file name parts
of the specification. For instance, in VMS, the file "login.com" might be specified as:
"u:[sstdeck1]login.com;"
where "u" is the name of the partition, "sst" is the name of the directory, "deck" is the name
of subdirectory, and "1", is the version number. Other systems simply treat the partition
name as part of the directory name. The first name given is that of the partition, and the rest
is the directory and file. For instance, "/u/pbg/test" might specify partition "u”, directory
"pbg", and file "test". A special case of this situation occurs in regard to the system files.
Those programs provided as a part of the system (loaders, assemblers, compilers, utility
routines, libraries, and so on) are generally defined as files. When the appropriate commands
are given to the operating system, these files are read by the loader and are executed. Many
command interpreters act by simply treating the command as the name of a file to load and
execute. As the directory system is defined presently, this file name would be searched for in
the current user file directory .One solution would be to copy the system files into each user
file directory. However, copying all the system files would be enormously wasteful of space.
The standard solution is to complicate the search procedure slightly. A special user directory
is defined to contain the system files.
Whenever a file name is given to be loaded, the operating system first searches the local user
file directory. If the file is found, it is used. If it is not found, the system automatically
searches the special user directory that contains the system files. The sequence of directories
searched when a file is named is called the search path. This idea can be extended, such that
the search path contains an unlimited list of directories to search when a command name is
given. This method is used in UNlX and MS-DOS.
3.2.1.2.3 Tree-Structured Directories
A tree system allows growth of the tree beyond the second level. Any directory may contain
names of additional directories as well as non-directory objects. This generalization allows
users to create their own sub-directories and to organize their files accordingly. The MS-
DOS system, for instance, is structured as a tree. In fact, a tree is the most common directory
structure. The tree has a root directory. Every file in the system has a unique path name. A
path name is the path from the root, through all the subdirectories, to a specified file.

Lesson No. 1 Intro. to Operating System 57

 a b c

A1 A2 B1 B2 C1 C2 C3

A21 A22 A23 C21 C22

Figure 4 Tree-Structured Directories

A directory (or subdirectory) contains a set of files or subdirectories. A directory is simply
another file but it is treated in a special way. All directories have the same internal format,
one bit in each directory entry defines the entry as a file (0) or as a subdirectory (1) Special
system calls are used to create and delete directories.
In normal use, each user has a current directory .The current directory should contain most of
the files that are of current interest to the user. When reference is made to a file, the current
directory is searched. If a file is needed that is not in the current directory, then the user must
either specify a path name or change the current directory to be the directory holding that
file. To change the current directory to a different directory, a system call is provided that
takes a directory name as a parameter and uses it to redefine the current directory.
Thus, the user can change his current directory whenever he desires. From one change
directory system call to the next, all open system calls search the current directory for the
specified file.
The initial current directory of a user is designated when the user job starts or the user logs
in. The operating system searches the accounting file (or ask) other predefined location to
find an entry for this user (for accounting). In the accounting file is a pointer to (or the name
of) the user's initial directory. This pointer is copied to a local variable for this user, which
specifies the user's initial current directory.
Path names can be of two types: absolute path names or relative path names.
(a) Absolute path: An absolute path name begins at the root and follows a path down to the
desired file, giving the directory names on the path. An absolute path name is an

Lesson No. 1 Intro. to Operating System 58

 unambiguous way of referring to a file. Thus identically named files created by different
users differ in their absolute path names.
(b) Relative path: A relative path name defines a path from the current directory.
Allowing the user to define his own subdirectories permits him to impose a structure on
his files. This structure might result in separate directories for files associated with
different topics (for example, a subdirectory was created to hold the text of this book or
different forms of information for example, the directory programs may contain source
programs; the directory bin may store all the binary files. An interesting policy decision in
a tree-structured directory structure is how to handle the deletion of a directory. If a
directory is empty, its entry in its containing directory can simply be deleted. But if the
directory to be deleted is not empty, containing files and subdirectories then one of two
approaches can be taken. As in MS-DOS, if we want to delete a directory then first of all
we have to empty it i.e. delete its contents and If there are any subdirectories, the
procedure must be applied recursively to them, so that they can be deleted also. But this
approach may be time consuming.
An alternative approach, such as that taken by the UNIX rm command, to provide the option
that, when a request is made to delete a directory, and that directory's files and subdirectories
are also to be deleted. Note that either approach is fairly easy to implement; the choice is one
of policy. The latter policy is more convenient, but more dangerous, because an entire
director structure may be removed with one command. If that command was issued in error,
a large number of files and directories would need to be restored from backup tapes.
With a tree-structured directory system, users can access, in addition their files, the files of
other users. For example, user B can access files of user A by specifying their path names.
User B can specify either an absolute or relative path name. Alternatively, user B could
change her current directory be user A's directory, and access the files by their file names.
Some systems also allow users to define their own search paths. In this case, user B could
define her search path to be (1) her local directory, (2) the system file directory, and user A's
directory, in that order. As long as the name of a file of user A did not conflict with the name
of a local file or system file, it could be referred to simply by its name.
A path to a file in a tree-structured directory can be longer than that in a two-level directory.
To allow users to access programs without having to remember these long paths, the
Macintosh operating system automates the search for executable programs. It maintains a file

Lesson No. 1 Intro. to Operating System 59

called the "Desktop File”, containing the name and location of all executable programs it has
seen. Where a new hard disk or floppy disk is added to the system, or the network accessed,
the operating system traverses the directory structure, searching for executable programs on
the device and recording the pertinent information. This mechanism supports the double-
click execution functionality. A double-click on a file causes its creator attribute to be read,
and the "Desktop File" to be searched for a match.
3.2.1.2.4 Acyclic-Graph Directories
Sharing of file is another important issue in deciding the directory structure. If more than one
user are working on some common project. So the files associated with that project should be
placed in a common directory that can be shared among a number of users.

A B

A1 A2 A3 A4 B1 B2 B3

B21 B22 B23

Figure 5 Acyclic-Graph Directories

The important characteristic of sharing is that if a user is making a change in a shared file
that is to be reflected to other user also. In this way a shared file is not the same as two
copies of the file. With two copies, each programmer can view the copy rather than the
original, but if one programmer changes the file, the changes will not appear in the other's
copy. With a shared file, there is only one actual file, so any changes made by the person
would be immediately visible to the other.

Lesson No. 1 Intro. to Operating System 60

This form of sharing is particularly important for shared subdirectories; a new file created by
one person will automatically appear in all the shared subdirectories. File sharing is
facilitated by acyclic graph structure. The tree structure doesn’t permit the sharing of files.
In a situation where several people are working as a team, all the files to be shared may be
put together into one directory. The user file directories of all the team members would each
contain this directory of shared files as a subdirectory. Even when there is a single user, his
file organization may require that some files be put into several different subdirectories. For
example, a program written for a particular project should be both in the directory of all
programs and in the directory for that project.
Shared files and subdirectories can be implemented in several ways. A common used in
UNIX systems, is to create a new directory entry called a link. A link is a pointer to another
file or subdirectory. For example, a link may be implemented as an absolute or relative path
name. When a reference to a file is made, we search the directory. The directory entry is
marked as a link and the name of the real file (or directory) is given. We resolve the link by
using the path name to locate the real file. Links are easily identified by their format in the
directory entry (or by their having a special type on systems that support types), and are
effectively named indirect pointers. The operating system ignores these links when
traversing directory trees to preserve the acyclic structure of the system.
The other approach to implementing shared files is simply to duplicate all information about
them in both sharing directories. Thus, both entries are identical and equal. A link is clearly
different from the original directory entry; thus, the two are not equal. Duplicate directory
entries, however, make the original and the copy indistinguishable. A major problem with
duplicate directory entries is maintaining consistency if the file is modified. An acyclic-
graph directory structure is more flexible than is a simple tree structure, but is also more
complex. Several problems must be considered carefully. Notice that a file may now have
multiple absolute path names. Consequently, distinct file names may refer to the same file.
This situation is similar to the aliasing problem for programming languages. If we are trying
to traverse the entire file system this problem becomes significant, since we do not want to
traverse shared structures more than once.
Another problem involves deletion. When can the space allocated to a shared file be de-
allocated and reused? One possibility is to remove the file whenever anyone deletes it, but

Lesson No. 1 Intro. to Operating System 61

this action may leave dangling pointers to the now non-existent file. Worse, if the remaining
file pointers contain actual disk addresses, and the space is subsequently reused for other
files, these dangling pointers may point into the middle of other files.
In a system where sharing is implemented by symbolic links, this situation is somewhat
easier to handle. The deletion of a link does not need to affect the original file; only the link
is removed. If the file entry itself is deleted, the space for the file is de-allocated, leaving the
links dangling. We can search for these links and remove them also, but unless a list of the
associated link is kept with each file, this search can be expensive. Alternatively, we can
leave the links until an attempt is made to use them. At that time, we can determine that the
file of the name given by the link does not exist, and can fail to resolve the link name; the
access is treated just like any other illegal file name. (In this case, the system designer should
consider carefully what to do when a file is deleted and another file of the same name is
created, before a symbolic link to the original file is used.) In the case of UNIX, symbolic
links are left when a file is deleted, and it is up to the user to realize that the original file is
gone or has been replaced.
Another approach to deletion is to preserve the file until all references to it are deleted. To
implement this approach, we must have some mechanism for determining that the last
reference to the file has been deleted. We could keep a list of all references to a file
(directory entries or symbolic links). When a link or a copy of the directory entry is
established, a new entry is added to the file-reference list. When a link or directory entry is
deleted, we remove its entry on the list. The file is deleted when its file-reference list is
empty. The trouble with this approach is the variable and potentially large size of the file-
reference list.
However, we really do not need to keep the entire list -we need to keep only a count of the
number of references. So a reference count is maintained with shared file, whenever a
reference is made to it, it is incremented by one. On deleting a link, the reference count is
decremented by one, when it becomes zero the file can be deleted. The UNIX operating
system uses this approach for non-symbolic links, or hard links, keeping a reference count in
the file information block or inode. By effectively prohibiting multiple references to
directories, we maintain an acyclic-graph structure.

Lesson No. 1 Intro. to Operating System 62

To avoid these problems, some systems do not allow shared directories links. For example,
in MS-DOS, the directory structure is a tree structure, rather than an acyclic graph, thereby
avoiding the problems associated with file deletion in an acyclic-graph directory structure.
3.2.1.2.5 General Graph Directory
One serious problem with using an acyclic graph structure is ensuring that there are no
cycles. If we start with a two-level directory and allow users to create subdirectories, a tree-
structured directory results. It should be fairly easy to see that simply adding new files and
subdirectories to existing tree structure preserves the tree-structured nature. However, when
we add links to an existing tree-structured directory, the tree structure is destroyed, resulting
in a simple graph structure.
The primary advantage of an acyclic graph is the relative simplicity of the algorithms to
traverse file in the graph and to determine when there are no more references to a file. We
want to avoid file is traversing shared sections of an acyclic graph twice, mainly for
performance reasons. If we have just searched a major shared subdirectory for a particular
file, without finding that file, we want to avoid searching that subdirectory again; the second
search would be a waste of time.

A B C

A1 A2 A3 C1 C2 C3 C4

X Y C41 C42

Figure 6 General Graph Directory

To improve the performance of the system we should avoid searching any component twice
in the systems where cycles are permitted. If cycles are not identified by the algorithm then it

Lesson No. 1 Intro. to Operating System 63

can be trapped in an infinite loop. One solution is to arbitrarily limit the number of
directories, which will be accessed during a search.
A similar problem exists when we are trying to determine when a file can be deleted. As with
acyclic-graph directory structures, a value zero in the reference count means that there are no
more references to the file or directory, and the file can be deleted. However, it is also
possible, when cycles exist, that the reference count may be nonzero, even when it is no
longer possible to refer to a directory or file. This anomaly is due to the self-referencing (a
cycle) in the directory structure. In this case, it is generally necessary to use a garbage
collection scheme to determine when the last reference has been deleted and the disk space
can be reallocated.
Garbage collection involves traversing the entire file system, marking everything that can be
accessed. Then, a second pass collects everything that is not marked onto a list of free space.
Garbage collection for a disk based file system, however, is extremely time-consuming and
is thus seldom attempted. Garbage collection is necessary only because of possible cycles in
the graph. Thus, an acyclic-graph structure is much easier to work with. The difficulty is to
avoid cycles, as new links are added to the structure. There are algorithms to detect cycles in
graphs. However, they are computationally expensive, especially when the graph is on disk
storage. Generally, tree directory structures are more common than are acyclic-graph
structures.
3.2.1.3 Directory Operations
The directory can be viewed as a symbol table that translates file names into their directory
entries. If we take such a view, then it becomes apparent that the directory itself can be
organized in many ways. The different operations that are to be carried out on directories are:
(a) To insert entries.
(b) To delete entries.
(c) To search for a named entry.
(d) To list all the entries in the directory.
When considering a particular directory structure, we need to keep in mind the operations
that are to be performed on a directory:
¾ Search for a directory: We need to be able to search a directory structure to find the
entry for a particular file. Since files have symbolic names and similar names may

Lesson No. 1 Intro. to Operating System 64

 indicate a relationship between files, we may want to be able to find all files whose
names match a particular pattern.
¾ Create a directory: New files need to be created and added to the directory.
¾ Delete a directory: When a file is no longer needed, we want to remove it from the
directory.
¾ List a directory: We need to be able to list the files in a directory and the contents of the
directory entry for each file in the list.
¾ Rename a directory: Because the name of a file represents its contents to its users, the
name must be changeable when the contents or use of the file changes. Renaming a file
may also allow its position within the directory structure to be changed.
¾ Traverse the file system: It is useful to be able to access every directory and every file
within a directory structure. For reliability it is a good idea to save the contents and
structure of the entire file system at regular intervals. This saving often consists of
copying all files to magnetic tape. This technique provides a backup copy in case of
system failure or if the file is simply no longer in use. In this case, the file can be copied
to tape, and the disk space of that file released for reuse by another file.
¾ Copying a directory: A directory may be copied from one location to another.
¾ Moving a directory: A directory may be moved from one location to a new location
with all its contents.
3.2.2 FILE PROTECTION & SECURITY
The security of the information is a major issue in file system. The files are to be protected
from the physical damage as well as improper access. One way of ensuring the security is
through backup. By maintaining the duplicate copy of the files, the reliability is improved. In
many systems this is done automatically without human intervention. The backup of the files
is done at regular interval automatically. So if a copy of the file is accidentally destroyed, we
have its backup copy.
There are a number of factors causing the damage to the file system such as:
(a) Hardware problems.
(b) Power failure
(c) Head crashes
(d) Dirt

Lesson No. 1 Intro. to Operating System 65

 (e) Temperature
(f) Bugs in the software
These things can result into the loss of contents of files. Protection can be provided in many
ways. For a small single-user system, we might provide protection by physically removing
the floppy disks and locking them in a desk drawer or file cabinet. In a multi-user system,
however, other mechanisms are needed.
3.2.2.1 Types of Access
The need for protecting files is a direct result of the ability to access files. On systems that do
not permit access to the files of other users, protection is not needed. Thus, one extreme
would be to provide complete protection by prohibiting access. The other extreme is to
provide free access with no protection. Both of these approaches are too extreme for general
use. What is needed is the controlled access.
Protection mechanisms provide controlled access by limiting
the types of file access that can be made. Access is permitted
or denied depending on several factors, one of which is the
type of access requested. Several different types of
operations may be controlled:
¾ Read - Read information contained in the file.
¾ Write - Write new information into a file at any point or overwrite existing information
in a file.
¾ Execute - Load the contents of a file into main memory and create a process to execute
it.
¾ Append - Write new information at the end of the file.
¾ Delete - Delete the file and release its storage space for use in other files.
¾ List – Read the names contained in a directory.
¾ Change access - Change some user’s access rights for some controlled operation.
Other operations, such as renaming, copying, or editing the file, may also be controlled. For
many systems, however, these higher-level functions (such as copying) may be implemented
by a system program that makes lower-level system calls. Protection is provided at only the
lower level. For instance, copying a file may be implemented simply by a sequence of read
requests. In his case, a user with read access can also cause the file to be copied, printed, and
so on.

Lesson No. 1 Intro. to Operating System 66

Many different protection mechanisms have been proposed. Each scheme has its advantages
and disadvantages and must be selected as appropriate for intended application. A small
computer system that is used by only a few members of a research group may not need the
same types of protection as will a large corporate computer that is used for research, finance,
and personnel iterations.
3.2.2.2 Protection structures
An access privilege is a right to make a specific form of access to a file. An access descriptor
describes access privileges for a file. The common accesses privileges read, write, and
execute are generally represented by r, w, and x descriptors. A user holds access privileges to
one or more files and a file is accessible to one or more users. Access control information for
a file is a collection of access descriptors for access privileges held by various users. Access
control information can be organized in various forms such as Access Control Matrix, access
Control Lists etc. which are discussed in the following section:
3.2.2.2.1 Access Control Matrix
Access control matrix (ACM) consists of rows and columns as shown in the following
figure. Each row describes the access privileges held by a user. Each column describes the
access control information for a file. Thus ACM (ui, fj) =aij implies that user ui can access file
fj in accordance with access privileges aij.
FilesÆ f1 f2 f3
Users
↓
u1 {r} (r, w} {r, w, x}
u2 {r} {r, x}
u3 {w} {r, w, x} {r}
Figure 7 Access Control Matrix
The important advantages of ACM are:
(a) Simplicity and efficiency of access.
(b) All information is stored in one structure.
But its main drawback is its size and sparseness. The size can be reduced by assigning access
privileges to group of users rather than the individual users resulting in the reduction of

Lesson No. 1 Intro. to Operating System 67

number of rows. The solution of sparseness is the use of lists instead of matrix as discussed
following.
3.2.2.2.2. Access control Lists and Groups
The most common approach to the protection problem is to make access dependent on the
stems identity of the user. Various users may need different types of access to a file or
directory. The most general scheme to implement identity-dependent access is to associate
with each file and directory an access control list (ACL), specifying the user name and the
types of access allowed for each user. Each element of the access control list is an access
control pair (<user name>, <list of access privileges>).
When a user requests access to a particular file, the operating system checks the access list
associated with that file. If that user is listed for the requested access, the access is allowed.
Otherwise, a protection violation occurs, and the user job is denied access to the file.
The main problem with access lists is their length. It depends on the number of users and the
number of access privileges defined in the system. Most file systems uses three kinds of
access privileges: (a) Read - file can be read, (b) write – file can be modified and new data
can be added, and (c) execute – permits the execution of the program. If we want to allow
everyone to read a file, we must list all users with read access. This technique has two
undesirable consequences:
(a) Constructing such a list may be a tedious and unrewarding task, especially if we do not
know in advance the list of users in the system.
(b) The directory entry that previously was of fixed size needs now to be of variable size,
resulting in space management being more complicated.
To reduce the size of protection information, users can be classified in some convenient
manner and an access control pair can be specified for each class of user rather than for
individual users. Now an access control list has only as many pairs as the number of user
classes. To condense the length of the access list, many systems recognize three
classifications of users in connection with each file (e.g. in UNIX):
1. Owner - The user who created the file is the owner
2. Group - A set of users who are sharing the file and need similar access is a
group or workgroup.
3. Universe - All other users in the system constitute the universe.
Note that, for this scheme to work properly, group membership must be controlled tightly.
This control can be accomplished in a number of different ways. For example, in the UNIX

Lesson No. 1 Intro. to Operating System 68

system, groups can be created and modified by only the manager of the facility (or by any
super-user). Thus, this control is achieved through human interaction. In the VMS system,
with each file, an access list (also known as an access control list) may be associated, listing
those users who can access the file. The owner of the file can create and modify this access
lists are discussed above.
With this more limited protection classification, only three fields are needed to define
protection. Each field is often a collection of bits, each of which either allows or prevents the
access associated with it. For example, the UNIX system defines three fields of 3 bits each:
rwx, where r controls read access, w controls write access, and x controls execution. A
separate field is kept for the file owner, for the owner's group and for all other users. In this
scheme, 9 bits per file are needed to record protection information.
3.2.2.2.3 Other Protection Approaches
Another approach to the protection problem is to associate
a password with each file. Access to each file can be
controlled by a password. If the passwords are chosen
randomly and changed often, this scheme may be effective in
limiting access to a file to only those users who know the
password.
There are several disadvantages to this scheme.
(a) First, if we associate a separate password with each file, then the number of passwords
that a user needs to remember may become large, making the scheme impractical.
(b) If only one password is used for all the files, then, once it is discovered, all files are
accessible.
Some systems allow a user to associate a password with a subdirectory, rather than with an
individual file, to deal with this problem. The IBM VM/CMS operating system allows three
passwords for a minidisk: one each for read, write, and multi write access. Second,
commonly, only one password is associated with each file. Thus, protection is on an all-or-
nothing basis. To provide protection on a more detailed level, we must use multiple
passwords.
Limited file protection is also currently available on single user systems, such as MS-DOS
and Macintosh operating system. These operating systems, when originally designed,
essentially ignored dealing with the protection problem. However, since these systems are
being placed on networks where file sharing and communication is necessary, protection

Lesson No. 1 Intro. to Operating System 69

mechanisms have to be retrofitted into the operating system. Note that it is almost always
easier to design a feature into an new operating system than it is to add a feature to an
existing one. Such updates are usually less effective and are not seamless.
We note that, in a multilevel directory structure, we need not only to protect individual files,
but also to protect collections of files contained in a subdirectory, that is, we need to provide
a mechanism for directory protection.
The directory operations that must be protected are somewhat different from the file
operations. We want to control the creation and deletion of files in a directory. In
addition, we probably want to control whether a user can determine the existence of a
file in a directory. Sometimes, knowledge of the existence and name of a file may be
significant in itself. Thus, listing the contents of a directory must be a protected
operation. Therefore, if a path name refers to a file in a directory, the user must be
allowed access to both the directory and the file. In systems where files may have
numerous path names (such as acyclic or general graphs), a given user may have
different access rights to a file, depending on the path name used.
3.3 Keywords
Directory: A directory may be defined as an object that contains the names of the file
system objects.
Access Control Matrix: ACM is a matrix in which each row describes the access privileges
held by a user and each column, access control information for a file.
Access Control List: It is a structure to implement identity-dependent access to each file and
directory where each element of the ACL is an access control pair (<user name>, <list of
access privileges>).
3.4 SUMMARY
A file system helps the user in organizing the files through the use of directories that
contains information about a group of files. A number of directory structures are used
such as Single-level Directory Two-level Directory, Tree-structured Directories, Acyclic-
Graph Directories, and General Graph Directory. Each approach has its merits and
demerits. A number of operations are carried out on directories such as insertion, deletion,
search, rename, traversal etc. So file system should facilitate these operations. Another
important issue in file system is the protection of the information from physical damage
and unauthorized access. To provide the access privileges to the files to different users

Lesson No. 1 Intro. to Operating System 70

two common mechanisms Access Control Matrix and Access Control Lists were
discussed. ACM are characterized by their simplicity and efficiency but suffers from large
size and sparseness. The problem of size is tackled by using the groups. If there are a
number of blank entries in the ACM, then ACL can be a preferred solution.
3.5 SUGGESTED READINGS / REFERENCE MATERIAL
6. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B., John Wiley &
Sons.
7. Systems Programming & Operating Systems, 2nd Revised Edition, Dhamdhere D.M.,
Tata McGraw Hill Publishing Company Ltd., New Delhi.
8. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill Publishing
Company Ltd., New Delhi.
9. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education Asia, 2000.
10. Operating Systems, Har ris J.A., Tata McGraw Hill Publishing Company Ltd., New
Delhi, 2002.
3.6 SELF-ASSESSMENT QUESTIONS (SAQ)
1. Define field, record, file, file sharing, and file protection.
2. What are the limitations of acyclic directory structure?
3. Which file operations are applicable to directories? Which are not?
4. How is a directory different from a file?
5. What are the different logical structures of the directory? Discuss their merits and
demerits?
6. Discuss the advantages and disadvantages of Access Control Lists (ACL) and Access
Control Matrix (ACM).
7. Discuss the advantages and disadvantages of two-level directory structure over
single-level directory structure.

Lesson No. 1 Intro. to Operating System 71

Lesson number: 4 Writer: Dr. Rakesh Kumar
CPU Scheduling Vetter: Prof. Dharminder Kr.

3.0 OBJECTIVE
The objective of this lesson is to make the students familiar with the various
issues of CPU scheduling. After studying this lesson, they will be familiar with:
1. Process states & transitions.
2. Different types of scheduler
3. Scheduling criteria
4. Scheduling algorithms
3.1 INTRODUCTION
In nearly every computer, the resource that is most often requested is the CPU or

processor. Many computers have only one processor, so this processor must be

shared via time-multiplexing among all the programs that need to execute on the

computer. Here we need to make an important distinction between a program &

an executing program.

"One of the most fundamental concepts of modern operating systems is the

distinction between a program & the activity of executing a program. The former

is merely a static set of directions; the latter is a dynamic activity whose

properties change as time progresses. This activity is knows as a process. A

process encompasses the current status of the activity, called the process state.

This state includes the current position in the program being executed (the value

of the program counter) as well as the values in the other CPU registers & the

associated memory cells. Roughly speaking, the process state is a snapshot of

the machine at that time. At different times during the execution of a program (at

Lesson No. 1 Intro. to Operating System 72

different times in a process) different snapshots (different process states) will be

observed."

The operating system is responsible for managing all the processes that are
running on a computer & allocated each process a certain amount of time to use
the processor. In addition, the operating system also allocates various other
resources that processes will need such as computer memory or disks. To keep
track of the state of all the processes, the operating system maintains a table
known as the process table. Inside this table, every process is listed along with
the resources the processes are using & the current state of the process.
Processes can be in one of three states: running, ready, or waiting (blocked).
The running state means that the process has all the resources it need for
execution & it has been given permission by the operating system to use the
processor. Only one process can be in the running state at any given time. The
remaining processes are either in a waiting state (i.e., waiting for some external
event to occur such as user input or a disk access) or a ready state (i.e., waiting
for permission to use the processor). In a real operating system, the waiting &
ready states are implemented as queues, which hold the processes in these
states.
The assignment of physical processors to processes allows processors to
accomplish work. The problem of determining when processors should be
assigned & to which processes is called processor scheduling or CPU
scheduling.
When more than one process is runable, the operating system must decide
which one first. The part of the operating system concerned with this decision is
called the scheduler, & algorithm it uses is called the scheduling algorithm. In
operating system literature, the term “scheduling” refers to a set of policies &
mechanisms built into the operating system that govern the order in which the
work to be done by a computer system is completed. A scheduler is an OS
module that selects the next job to be admitted into the system & the next
process to run. The primary objective of scheduling is to optimize system

Lesson No. 1 Intro. to Operating System 73

performance in accordance with the criteria deemed most important by the
system designers.
3.2 PRESENTATION OF CONTENTS

3.2.1 Definition of Process

3.2.2 Process States & Transitions
3.2.3 Types of schedulers

3.2.3.1 The long-term scheduler

3.2.3.2 The medium-term scheduler
3.2.3.3 The short-term scheduler
3.2.4 Scheduling & Performance Criteria
3.2.4.1 User-oriented Scheduling Criteria
3.2.4.2 System-oriented Scheduling Criteria
3.2.5 Scheduler Design
3.2.6 Scheduling Algorithms
3.2.6.1 First-Come, First-Served (FCFS) Scheduling
3.2.6.2 Shortest Job First (SJF)
3.2.6.3 Shortest Remaining Time Next (SRTN) Scheduling
3.2.6.4 Round Robin
3.2.6.5 Priority-Based Preemptive Scheduling (Event-Driven, ED)
3.2.6.6 Multiple-Level Queues (MLQ) Scheduling
3.2.6.7 Multiple-Level Queues with Feedback Scheduling

3.2.1 Definition of Process

The notion of process is central to the understanding of operating systems. There
are quite a few definitions presented in the literature, but no "perfect" definition
has yet appeared.
The term "process" was first used by the designers of the MULTICS in 1960's.
Since then, the term process, used somewhat interchangeably with 'task' or 'job'.
The process has been given many definitions for instance

Lesson No. 1 Intro. to Operating System 74

 ¾ A program in Execution.
¾ An asynchronous activity.
¾ The 'animated sprit' of a procedure in execution.
¾ The entity to which processors are assigned.
¾ The 'dispatchable' unit.
and many more definitions have been given. As we can see from above that
there is no universally agreed upon definition, but the definition "Program in
Execution" seem to be most frequently used. Now that we agreed upon the
definition of process, the question is what is the relation between process &
program. Process is not the same as program. In the following discussion we
point out some of the difference between process & program. As we have
mentioned earlier Process is not the same as program. A process is more than a
program code. A process is an active entity as oppose to program which
consider being a 'passive' entity. As we all know that a program is an algorithm
expressed in some suitable notation, (e.g., programming language). Being a
passive, a program is only a part of process. Process, on the other hand,
includes:
¾ Current value of Program Counter (PC)
¾ Contents of the processors registers
¾ Value of the variables
¾ The process stack (SP) which typically contains temporary data such as
subroutine parameter, return address, & temporary variables.
¾ A data section that contains global variables.
A process is the unit of work in a system.

In Process model, all software on the computer

is organized into a number of sequential
processes. A process includes PC, registers, &
variables. Conceptually, each process has its

Lesson No. 1 Intro. to Operating System 75

own virtual CPU. In reality, the CPU switches
back & forth among processes. The process
state consist of everything necessary to resume
the process execution if it is somehow put aside
temporarily. The process state consists of at
least following:
¾ Code for the program.
¾ Program's static data.
¾ Program's dynamic data.
¾ Program's procedure call stack.
¾ Contents of general purpose register.
¾ Contents of program counter (PC)
¾ Contents of program status word (PSW).
¾ Operating Systems resource in use.
A process goes through a series of discrete process states.
¾ New State: The process being created.
¾ Running State: A process is said to be running if it has the CPU, that is,
process actually using the CPU at that particular instant.
¾ Blocked (or waiting) State: A process is said to be blocked if it is waiting for
some event to happen such that as an I/O completion before it can proceed.
Note that a process is unable to run until some external event happens.
¾ Ready State: A process is said to be ready if it use a CPU if one were
available. A ready state process is runable but temporarily stopped running to
let another process run.
¾ Terminated state: The process has finished execution.
3.2.2 Process States & Transitions
The diagram below contains much information.

Lesson No. 1 Intro. to Operating System 76

Consider a running process P that issues an I/O request
¾ The process blocks
¾ At some later point, a disk interrupt occurs & the driver detects that P's
request is satisfied.
¾ P is unblocked, i.e. is moved from blocked to ready
¾ At some later time the operating system looks for a ready job to run & picks P
& P moved to running state.
¾ A suspended process may be removed from the main memory & placed in
the backup memory. Subsequently they may be released & moved to the
ready state by the medium term scheduler.

3.2.3 TYPES OF SCHEDULERS

Lesson No. 1 Intro. to Operating System 77

The schedulers may be categorized as long term scheduler, medium term
scheduler, & short term scheduler as shown in Figure 1 & Figure 2. Figure 1
shows the possible traversal paths of jobs & programs through the components &
queues, depicted by rectangles, of a computer system. The primary places of
action of the three types of schedulers are marked with down-arrows. As shown
in Figure 2, a submitted batch job joins the batch queue while waiting to be
processed by the long-term scheduler. Once scheduled for execution, processes
spawned by the batch job enter the ready queue to await processor allocation by
the short-term scheduler. After becoming suspended, the running process may
be removed from memory & swapped out to secondary storage. Such processes
are subsequently admitted to main memory by the medium-term scheduler in
order to be considered for execution by the short-term scheduler.

Suspended
Medium Term Scheduler & Swapped
Interactive Out Queue

Programs
Short Term Scheduler
Batch
Jobs
CPU
Batch Queue Ready Queue
Exit
Exit Exit

Suspended
Queue

Long Term Scheduler

Figure 1- Process Schedulers

3.2.3.1 The long-term scheduler

The long-term scheduler decides when to start jobs, i.e., do not necessarily start
them when submitted. CTSS (an early time sharing system at MIT) did this to
insure decent interactive response time. The long-term scheduler, when present,

Lesson No. 1 Intro. to Operating System 78

works with the batch queue & selects the next batch job to be executed. Batch is
usually reserved for resource-intensive (processor time, memory, special I/O
devices), low-priority programs that may be used as fillers to keep the system
resources busy during periods of low activity of interactive jobs. As pointed out
earlier, batch jobs contain all necessary data & commands for their execution.
Batch jobs usually also contain programmer-assigned estimates of their resource
needs, such as memory size, expected execution time, & device requirements.
Knowledge about the anticipated job behavior facilitates the work of the long-
term scheduler.
The primary objective of the long-term scheduler is to provide a balanced mix of
jobs, such as processor-bound & I/O-bound, to the short-term scheduler. In a
way, the long-term scheduler acts as a first-level throttle in keeping resource
utilization at the desired level. For example, when the processor utilization is low,
the scheduler may admit more jobs to increase the number of processes in a
ready queue, & with it the probability of having some useful work awaiting
processor allocation. Conversely, when the utilization factor becomes high as
reflected in the response time, the long-term scheduler may opt to reduce the
rate of batch-job admission accordingly. In addition, the long-term scheduler is
usually invoked whenever a completed job departs the system. The frequency of
invocation of the long-term scheduler is thus both system-and workload-
dependent; but it is generally much lower than for the other two types of
schedulers. As a result of the relatively infrequent execution & the availability of
an estimate of its workload's characteristics, the long-term scheduler may
incorporate rather complex & computationally intensive algorithms for admitting
jobs into the system. In terms of the process state-transition diagram, the long-
term scheduler is basically in charge of the dormant-to-ready transitions. Ready
processes are placed in the ready queue (ready list, in our earlier discussion) for
consideration by the short-term scheduler.

Lesson No. 1 Intro. to Operating System 79

3.2.3.2 The medium-term scheduler
The medium term scheduler suspend (swap out) some process if memory is
over-committed. The criteria for choosing a victim may be (a) How long since
previously suspended? (b) How much CPU time used recently? (c) How much
memory does it use? (d) External priority (pay more, get swapped out less) etc.
A running process may become suspended by making an I/O request or by
issuing a system call. Given that suspended processes cannot make any
progress towards completion until the related suspending condition is removed, it
is sometimes beneficial to remove them from main memory to make room for
other processes. In practice, the main-memory capacity may impose a limit on
the number of active processes in the system. When a number of those
processes become suspended, the remaining supply of ready processes in
systems where all suspended processes remain resident in memory may
become reduced to a level that impairs functioning of the short-term scheduler by
leaving it few or no options for selection. In systems with no support for virtual
memory, moving suspended processes to secondary storage may alleviate this
problem. Saving the image of a suspended process in secondary storage is
called swapping, & the process is said to be swapped out or rolled out.
The medium-term scheduler is in charge of handling the swapped-out processes.
It has little to do while a process remains suspended. However, once the
suspending condition is removed, the medium-term scheduler attempts to
allocate the required amount of main memory, & swap the process in & make it
ready. To work properly, the medium-term scheduler must be provided with
information about the memory requirements of swapped-out processes.
In terms of the state-transition diagram, the medium-term scheduler controls
suspended-to-ready transitions of swapped processes. This scheduler may be
invoked when memory space is vacated by a departing process or when the
supply of ready processes falls below a specified limit.
Medium-term scheduling is really part of the swapping function of an operating
system. The success of the medium-term scheduler is based on the degree of

Lesson No. 1 Intro. to Operating System 80

multiprogramming that it can maintain, by keeping as many processes “runnable”
as possible. More processes can remain executable if we reduce the resident set
size of all processes. The medium-term scheduler makes decisions as to which
pages of which processes need stay resident, & which pages must be swapped
out to make room for other processes. The sharing of some pages of memory,
either explicitly or through the use of shared or dynamic link libraries complicates
the task of the medium-term scheduler, which now must maintain reference
counts on each page. The responsibilities of the medium-term scheduler may be
further complicated in some operating systems, in which some processes may
request (demand?) that their pages remain locked in physical memory:

3.2.3.3 The short-term scheduler

The long-term scheduler runs relatively infrequently, when a decision must be

made as to the admission of new processes: maybe on average every ten

seconds. The medium-term scheduler runs more frequently, deciding which

process’s pages to swap to & from the swapping device: typically once a second.

The short-term scheduler, often termed the dispatcher, executes most frequently

(every few hundredths of a second) making fine-grained decisions as to which

process to move to Running next. The short-term scheduler is invoked whenever

an event occurs which provides the opportunity, or requires, the interruption of

the current process & the new (or continued) execution of another process. Such

opportunities include:

¾ Clock interrupts, provide the opportunity to reschedule every few

milliseconds,

¾ Expected I/O interrupts, when previous I/O requests are finally satisfied,

¾ Operating system calls, when the running process asks the operating system

to perform an activity on its behalf, and

Lesson No. 1 Intro. to Operating System 81

¾ Unexpected, asynchronous, events, such as unexpected input, user-interrupt,

or a fault condition in the running program.

The short-term scheduler allocates the processor among the pool of ready
processes resident in memory. Its main objective is to maximize system
performance in accordance with the chosen set of criteria. Since it is in charge of
ready-to-running state transitions, the short-term scheduler must be invoked for
each process switch to select the next process to be run. In practice, the short-
term scheduler is invoked whenever an event (internal or external) causes the
global state of the system to change. Given that any such change could result in
making the running process suspended or in making one or more suspended
processes ready, the short-term scheduler should be run to determine whether
such significant changes have indeed occurred and, if so, to select the next
process to be run. Some of the events occurred and, if so, to select the next
process to be run.
Most of the process-management OS services discussed in this lesson requires
invocation of the short-term scheduler as part of their processing. For example,
creating a process or resuming a suspended one adds another entry to the ready
list (queue), & the scheduler is invoked to determine whether the new entry
should also become the running process. Suspending a running process,
changing priority of the running process, & exiting or aborting a process are also
events that may necessitate selection of a new running process, changing priority
of the running process, & exiting or aborting a process are also events that may
necessitate selection of a new running process. Some operating systems include
an OS call that allows system programmers to cause invocation of the short-term
scheduler explicitly, such as the DECLARE_SIGNIFICANT_EVENT call in the
RSX-11M operating system. Among other things, this service is useful for
invoking the scheduler from user-written event-processing routines, such as
device (I/O) drivers.
As indicated in Figure 2, interactive programs often enter the ready queue
directly after being submitted to the OS, which then creates the corresponding

Lesson No. 1 Intro. to Operating System 82

process. Unlike-batch jobs, the influx of interactive programs are not throttled, &
they may conceivably saturate the system. The necessary control is usually
provided indirectly by deterioration response time, which tempts the users to give
up & try again later, or at least to reduce the rate of incoming requests.
Figure 2 illustrates the roles & the interplay among the various types of
schedulers in an operating system. It depicts the most general case of all three
types being present. For example, a larger operating system might support both
batch & interactive programs & rely on swapping to maintain a well-behaved mix
of active processes. Smaller or special-purpose operating systems may have
only one or two types of schedulers available. Along-term scheduler is normally
not found in systems without support for batch, & the medium-term scheduler is
needed only when swapping is used by the underlying operating system. When
more than one type of scheduler exists in an operating system, proper support
for communication & interaction is very important for attaining satisfactory &
balanced performance. For example, the long-term & the medium-term
schedulers prepare workload for the short-term scheduler. If they do not provide
a balanced mixed of compute-bound & I/O-bound processes, the short-term
scheduler is not likely to perform well no matter how sophisticated it may be on
its own merit.
3.2.4 SCHEDULING & PERFORMANCE CRITERIA
The objectives of a good scheduling policy include
¾ Fairness.
¾ Efficiency.
¾ Low response time (important for interactive jobs).
¾ Low turnaround time (important for batch jobs).
¾ High throughput
¾ Repeatability.
¾ Fair across projects.
¾ Degrade gracefully under load.
The success of the short-term scheduler can be characterized by its success
against user-oriented criteria under which a single user (selfishly) evaluates their

Lesson No. 1 Intro. to Operating System 83

perceived response, or system-oriented criteria where the focus is on efficient
global use of resources such as the processor & memory. A common measure of
the system-oriented criteria is throughput, the rate at which tasks are completed.
On a single-user, interactive operating system, & the user-oriented criteria take
precedence: it is unlikely that an individual will exhaust resource consumption,
but responsiveness remains all important. On a multi-user, multi-tasking system,
the global system-oriented criteria are more important as they attempt to provide
fair scheduling for all, subject to priorities & available resources.
3.2.4.1 User-oriented Scheduling Criteria
Response time
In an interactive system this measures the time between submissions of a new
process request & the commencement of its execution. Alternatively, it can
measure the time between a user issuing a request to interactive input (such as a
prompt) & the time to echo the user’s input or accept the carriage return.
Turnaround time
This is the time between submission of a new process & its completion.
Depending on the mixture of current tasks, two submissions of identical
processes will likely have different turnaround times. Turnaround time is the sum
of execution & waiting times.
Deadlines
In a genuine real-time operating system, hard deadlines may be requested by
processes. These either demands that the process is completed with a
guaranteed upper-bound on its turnaround time, or provide a guarantee that the
process will receive the processor in a guaranteed maximum time in the event of
an interrupt. A real-time long-term scheduler should only accept a new process if
it can guarantee required deadlines. In combination, the short-term scheduler
must also meet these deadlines.
Predictability
With lower importance, users expect similar tasks to take similar times. Wild
variations in response & turnaround times are distracting.

Lesson No. 1 Intro. to Operating System 84

3.2.4.2 System-oriented Scheduling Criteria
Throughput
The short-term scheduler attempts to maximize the number of completed jobs
per unit time. While this is constrained by the mixture of jobs, & their execution
profiles, the policy affects utilization & thus completion.
Processor utilization
The percentage of time that the processor may be fed with work from Ready. In a
single-user, interactive system, processor utilization is very unlikely to exceed a
few percent.
Fairness
Subject to priorities, all processes should be treated fairly, & none should suffer
processor starvation. This simply implies, in most cases, that all processes are
moved to the ends of their respective state queues, & may not “jump the queue”.
Priorities
Conversely, when processes are assigned priorities, the scheduling policy should
favor higher priorities.
3.2.5 SCHEDULER DESIGN
Design process of a typical scheduler consists of selecting one or more primary
performance criteria & ranking them in relative order of importance. The next
step is to design a scheduling strategy that maximizes performance for the
specified set of criteria while obeying the design constraints. One should
intentionally avoid the word "optimization" because most scheduling algorithms
actually implemented do not schedule optimally. They are based on heuristic
techniques that yield good or near-optimal performance but rarely achieve
absolutely optimal performance. The primary reason for this situation lies in the
overhead that would be incurred by computing the optimal strategy at run-time, &
by collecting the performance statistics necessary to perform the optimization. Of
course, the optimization algorithms remain important, at least as a yardstick in
evaluating the heuristics. Schedulers typically attempt to maximize the average
performance of a system, relative to a given criterion. However, due
consideration must be given to controlling the variance & limiting the worst-case

Lesson No. 1 Intro. to Operating System 85

behavior. For example, a user experiencing 10-second response time to simple
queries has little consolation in knowing that the system's average response time
is under 2 seconds.
One of the problems in selecting a set of performance criteria is that they often
conflict with each other. For example, increased processor utilization is usually
achieved by increasing the number of active processes, but then response time
deteriorates. As is the case with most engineering problems, the design of a
scheduler usually requires careful balance of all the different requirements &
constraints. With the knowledge of the primary intended use of a given system,
operating-system designers tend to maximize the criteria most important in a
given environment. For example, throughput & component utilization are the
primary design objectives in a batch system. Multi-user systems are dominated
by concerns regarding the terminal response time, & real-time operating systems
are designed for the ability to handle burst of external events responsively.
3.2.7 SCHEDULING ALGORITHMS
The scheduling mechanisms described in this section may, at least in theory, be
used by any of the three types of schedulers. As pointed out earlier, some
algorithms are better suited to the needs of a particular type of scheduler.
Depending on whether a particular scheduling discipline is primarily used by the
long-term or by the short-term scheduler, we illustrate its working by using the
term job or process for a unit of work, respectively.
The scheduling policies may be categorized as preemptive & non-preemptive. So
it is important to distinguish preemptive from non-preemptive scheduling
algorithms. Preemption means the operating system moves a process from
running to ready without the process requesting it. Without preemption, the
system implements “run to completion''. Preemption needs a clock interrupt (or
equivalent). Preemption is needed to guarantee fairness & it is found in all
modern general-purpose operating systems.
Non-pre-emptive: In non-preemptive scheduling, once a process is executing, it
will continue to execute until
¾ It terminates, or

Lesson No. 1 Intro. to Operating System 86

¾ It makes an I/O request which would block the process, or
¾ It makes an operating system call.
Pre-emptive: In the preemptive scheduling, the same three conditions as above
apply, & in addition the process may be pre-empted by the operating system
when
¾ A new process arrives (perhaps at a higher priority), or
¾ An interrupt or signal occurs, or
¾ A (frequent) clock interrupt occurs.
CPU Scheduling deals with the problem of deciding which of the processes in the
ready queue is to be allocated the CPU. Following are some scheduling
algorithms we will study: FCFS Scheduling, Round Robin Scheduling, SJF
Scheduling, SRTN Scheduling, Priority Scheduling, Multilevel Queue Scheduling,
& Multilevel Feedback Queue Scheduling.
3.2.6.1 First-Come, First-Served (FCFS) Scheduling
The simplest selection function is the First-Come-First-Served (FCFS) scheduling
policy. In it
1. The operating system kernel maintains all Ready processes in a single
queue,
2. The process at the head of the queue is always selected to execute next,
3. The Running process runs to completion, unless it requests blocking I/O,
4. If the Running process blocks, it is placed at the end of the Ready queue.
Clearly, once a process commences execution, it will run as fast as possible
(having 100% of the CPU, & being non-pre-emptive), but there are some obvious
problems. By failing to take into consideration the state of the system & the
resource requirements of the individual scheduling entities, FCFS scheduling
may result in poor performance. As a consequence of no preemption, component
utilization & the system throughput rate may be quite low.
Processes of short duration suffer when “stuck” behind very long-running
processes. Since there is no discrimination on the basis of the required service,
short jobs may suffer considerable turnaround delays & waiting times when one
or more long jobs are in the system. For example, consider a system with two

Lesson No. 1 Intro. to Operating System 87

jobs, J1 & J2, with total execution times of 20 & 2 time units, respectively. If they
arrive shortly one after the other in the order J1-J2, the turnaround times are 20
& 22 time units, respectively (J2 must wait for J1 to complete), thus yielding an
average of 21 time units. The corresponding waiting times are 0 & 20 unit,
yielding an average of 10 time units. However, when the same two jobs arrive in
the opposite order, J2-J1, the average turnaround time drops to 11, & the
average waiting time is only 1 time unit.
Compute-bound processes are favored over I/O-bound processes.
We can measure the effect of FCFS by examining:
¾ The average turnaround time of each task (the sum of its waiting & running
times), or
¾ The normalized turnaround time (the ratio of running to waiting times).
3.2.6.2 Shortest Job First (SJF)
In this scheduling policy, the jobs are sorted on the basis of total execution time
needed & then it run the shortest job first. It is a non-preemptive scheduling
policy. Now First consider a static situation where all jobs are available in the
beginning, & we know how long each one takes to run, & we implement “run-to-
completion'' (i.e., we don't even switch to another process on I/O). In this
situation, SJF has the shortest average waiting time. Assume you have a
schedule with a long job right before a short job. Now if we swap the two jobs,
this decreases the wait for the short by the length of the long job & increases the
wait of the long job by the length of the short job. & this in turn decreases the
total waiting time for these two. Hence decreases the total waiting for all jobs &
hence decreases the average waiting time as well. So in this policy whenever a
long job is right before a short job, we swap them & decrease the average
waiting time. Thus the lowest average waiting time occurs when there are no
short jobs rights before long jobs. This is an example of priority scheduling. This
scheduling policy can starve processes that require a long burst.
3.2.6.3 Shortest Remaining Time Next (SRTN) Scheduling
Shortest remaining time next is a scheduling discipline in which the next
scheduling entity, a job or a process, is selected on the basis of the shortest

Lesson No. 1 Intro. to Operating System 88

remaining execution time. SRTN scheduling may be implemented in either the
non-preemptive or the preemptive variety. The non-preemptive version of SRTN
is called shortest job first (SJF). In either case, whenever the SRTN scheduler is
invoked, it searches the corresponding queue (batch or ready) to find the job or
the process with the shortest remaining execution time. The difference between
the two cases lies in the conditions that lead to invocation of the scheduler and,
consequently, the frequency of its execution. Without preemption, the SRTN
scheduler is invoked whenever a job is completed or the running process
surrenders control to the OS. In the preemptive version, whenever an event
occurs that makes a new process ready, the scheduler is invoked to compare the
remaining processor execution time of the running process with the time needed
to complete the next processor burst of the newcomer. Depending on the
outcome, the running process may continue, or it may be preempted & replaced
by the shortest-remaining-time process. If preempted, the running process joins
the ready queue.
SRTN is a provably optimal scheduling discipline in terms of minimizing the
average waiting time of a given workload. SRTN scheduling is done in a
consistent & predictable manner, with a bias towards short jobs. With the addition
of preemption, an SRTN scheduler can accommodate short jobs that arrive after
commencement of a long job. Preferred treatment of short jobs in SRTN tends to
result in increased waiting times of long jobs in comparison with FCFS
scheduling, but this is usually acceptable.
The SRTN discipline schedules optimally assuming that the exact future
execution times of jobs or processes are known at the time of scheduling. In the
case of short-term scheduling & preemption’s, even more detailed knowledge of
the duration of each individual processor burst is required. Dependence on future
knowledge tends to limit the effectiveness of SRTN implementations in practice,
because future process behavior is unknown in general & difficult to estimate
reliably, except for some very specialized deterministic cases.
Predictions of process execution requirements are usually based on observed
past behavior, perhaps coupled with some other knowledge of the nature of the

Lesson No. 1 Intro. to Operating System 89

process & its long-term statistical properties, if available. A relatively simple
predictor, called the exponential smoothing predictor, has the following form:
Pn = α0n-1 + (1 - α)P-1
where 0n is the observed length of the (n-1)th execution interval, Pn-1 is the
predictor for the same interval, & α is a number between 0 & 1. The parameter α
controls the relative weight assigned to the past observations & predictions. For
the extreme case of α = 1, the past predictor is ignored, & the new prediction
equals the last observation. For α = 0, the last observation is ignored. In general,
expansion of the recursive relationship yields
n-1
Pn = α ∑ (1 - α)i0n-i-1
I=0
Thus the predictor includes the entire process history, with its more recent history
weighted more.
Many operating systems measure & record elapsed execution time of a process
in its PCB. This information is used for scheduling & accounting purposes.
Implementation of SRTN scheduling obviously requires rather precise
measurement & imposes the overhead of predictor calculation at run time.
Moreover, some additional feedback mechanism is usually necessary for
corrections when the predictor is grossly incorrect.
SRTN scheduling has important theoretical implications, & it can serve as a
yardstick for assessing performance of other, realizable scheduling disciplines in
terms of their deviation from the optimum. Its practical application depends on
the accuracy of prediction of the job & process behavior, with increased accuracy
calling for more sophisticated methods & thus resulting in greater overhead. The
preemptive variety of SRTN incurs the additional overhead of frequent process
switching & scheduler invocation to examine each & every process transition into
the ready state. This work is wasted when the new ready process has a longer
remaining execution time than the running process.
3.2.6.4 Round Robin

Lesson No. 1 Intro. to Operating System 90

In interactive environments, such as time-sharing systems, the primary
requirement is to provide reasonably good response time and, in general, to
share system resources equitably among all users. Obviously, only preemptive
disciplines may be considered in such environments, & one of the most popular
is time slicing, also known as round robin (RR).
It is a preemptive scheduling policy. This scheduling policy gives each process a
slice of time (i.e., one quantum) before being preempted. As each process
becomes ready, it joins the ready queue. A clock interrupt is generated at
periodic intervals. When the interrupt occurs, the currently running process is
preempted, & the oldest process in the ready queue is selected to run next. The
time interval between each interrupt may vary.
It is one of the most common & most important scheduler. This is not the
simplest scheduler, but it is the simplest preemptive scheduler. It works as
follows:
¾ The processes that are ready to run (i.e. not blocked) are kept in a FIFO
queue, called the "Ready" queue.
¾ There is a fixed time quantum (50 msec is a typical number) which is the
maximum length that any process runs at a time.
¾ The currently active process P runs until one of two things happens:
• P blocks (e.g. waiting for input). In that case, P is taken off the ready
queue; it is in the "blocked" state.
• P exhausts its time quantum. In this case, P is pre-empted, even though it
is still able to run. It is put at the end of the ready queue.
In either case, the process at the head of the ready queue is now made
the active process.
¾ When a process unblocks (e.g. the input it's waiting for is complete) it is put at
the end of the ready queue.
Suppose the time quantum is 50 msec, process P is executing, & it blocks after
20 msec. When it unblocks, & gets through the ready queue, it gets the standard
50 msec again; it doesn't somehow "save" the 30 msec that it missed last time.

Lesson No. 1 Intro. to Operating System 91

It is an important preemptive scheduling policy. It is essentially the preemptive
version of FCFS. The key parameter here is the quantum size q. When a
process is put into the running state a timer is set to q. If the timer goes off & the
process is still running, the OS preempts the process. This process is moved to
the ready state where it is placed at the rear of the ready queue. The process at
the front of the ready list is removed from the ready list & run (i.e., moves to state
running). When a process is created, it is placed at the rear of the ready list. As
q gets large, RR approaches FCFS. As q gets small, RR approaches PS
(Processor Sharing).
What value of q should we choose? Actually it is a tradeoff (1) Small q makes
system more responsive, (2) Large q makes system more efficient since less
process switching.
Round robin scheduling achieves equitable sharing of system resources. Short
processes may be executed within a single time quantum & thus exhibit good
response times. Long processes may require several quanta & thus be forced to
cycle through the ready queue a few times before completion. With RR
scheduling, response time of long processes is directly proportional to their
resource requirements. For long processes that consist of a number of interactive
sequences with the user, primarily the response time between the two
consecutive interactions matters. If the computational requirements between two
such sequences may be completed within a single time slice, the user should
experience good response time. RR tends to subject long processes without
interactive sequences to relatively long turnaround & waiting times. Such
processes, however, may best be run in the batch mode, & it might even be
desirable to discourage users from submitting them to the interactive scheduler.
Implementation of round robin scheduling requires support of an interval timer-
preferably a dedicated one, as opposed to sharing the system time base. The
timer is usually set to interrupt the operating system whenever a time slice
expires & thus force the scheduler to be invoked. The scheduler itself simply
stores the context of the running process, moves it to the end of the ready queue,
& dispatches the process at the head of the ready queue. The scheduler is also

Lesson No. 1 Intro. to Operating System 92

invoked to dispatch a new process whenever the running process surrenders
control to the operating system before expiration of its time quantum, say, by
requesting I/O. The interval timer is usually reset at that point, in order to provide
the full time slot to the new running process. The frequent setting & resetting of a
dedicated interval timer makes hardware support desirable in systems that use
time slicing.
Round robin scheduling is often regarded as a "fair" scheduling discipline. It is
also one of the best-known scheduling disciplines for achieving good & relatively
evenly distributed terminal response time. The performance of round robin
scheduling is very sensitive to the choice of the time slice. For this reason,
duration of the time slice is often made user-tunable by means of the system
generation process.
The relationship between the time slice & performance is markedly nonlinear.
Reduction of the time slice should not be carried too far in anticipation of better
response time. Too short a time slice may result in significant overhead due to
the frequent timer interrupts & process switches. On the other hand, too long a
time slice reduces the preemption overhead but increases response time.
Too short a time slice results in excessive overhead, & too long a time slice
degenerates from round-robin to FCFS scheduling, as processes surrender
control to the OS rather than being preempted by the interval timer. The "optimal"
value of the time slice lies somewhere in between, but it is both system-
dependent & workload-dependent. For example, the best value of time slice for
our example may not turn out to be so good when other processes with different
behavior are introduced in the system, that is, when characteristics of the
workload change. This, unfortunately, is commonly the case with time-sharing
systems where different types of programs may be submitted at different times.
In summary, round robin is primarily used in time-sharing & multi-user systems
where terminal response time is important. Round robin scheduling generally
discriminates against long non-interactive jobs & depends on the judicious choice
of time slice for adequate performance. Duration of a time slice is a tunable
system parameter that may be changed during system generation.

Lesson No. 1 Intro. to Operating System 93

Variants of Round Robin
State dependent RR
It is same as RR but q is varied dynamically depending on the state of the
system. It favors processes holding important resources. For example, non-
swappable memory.
External priorities
In it a user can pay more & get bigger q. That is one process can be given a
higher priority than another. But this is not an absolute priority, i.e., the lower
priority (i.e., less important) process does get to run, but not as much as the high
priority process.
3.2.6.5 Priority-Based Preemptive Scheduling (Event-Driven, ED)
In it each job is assigned a priority (externally, perhaps by charging more for
higher priority) & the highest priority ready job is run. In this policy, If many
processes have the highest priority, it uses RR among them. In principle, each
process in the system is assigned a priority level, & the scheduler always
chooses the highest-priority ready process. Priorities may be static or dynamic. In
either case, the user or the system assigns their initial values at the process-
creating time. The level of priority may be determined as an aggregate figure on
the basis of an initial value, characteristic, resource requirements, & run-time
behavior of the process. In this sense, many scheduling disciplines may be
regarded as being priority-driven, where the priority of a process represents its
likelihood of being scheduled next. Priority-based scheduling may be preemptive
or non-preemptive.
A common problem with priority-based scheduling is the possibility that low-
priority processes may be effectively locked out by the higher priority ones. In
general, completion of a process within finite time of its creation cannot be
guaranteed with this scheduling policy. In systems where such uncertainty
cannot be tolerated, the usually remedy is provided by the aging priority, in which
the priority of each process is gradually increased after the process spends a
certain amount of time in the system. Eventually, the older processes attain high
priority & are ensured of completion in finite time.

Lesson No. 1 Intro. to Operating System 94

By means of assigning priorities to processes, system programmers can
influence the order in which an ED scheduler services coincident external events.
However, the high-priority ones may starve low-priority processes. Since it gives
little consideration to resource requirements of processes, event-driven
scheduling cannot be expected to excel in general-purpose systems, such as
university computing centers, where a large number of user processes are run at
the same (default) level of priority.
Another variant of priority-based scheduling is used in the so-called hard real-
time systems, where each process must be guaranteed execution before
expiration of its deadline. In such systems, time-critical processes are assumed
to be assigned execution deadlines. The system workload consists of a
combination of periodic processes, executed cyclically with a known period, & of
periodic processes, executed cyclically with a known period, & of a periodic
processes whose arrival times are generally not predictable. An optimal
scheduling discipline in such environments is the earliest-deadline scheduler,
which schedules for execution the ready process with the earliest deadline.
Another form of scheduler, called the least laxity scheduler or the least slack
scheduler, has also been shown to be optimal in single-processor systems. This
scheduler selects the ready process with the least difference between its
deadline & computation time. Interestingly, neither of these schedulers is optimal
in multiprocessor environments.
Priority aging
It is a solution to the problem of starvation. As a job is waiting, raise its priority so
eventually it will have the maximum priority. This prevents starvation. It is
preemptive policy. If there are many processes with the maximum priority, it uses
FCFS among those with max priority (risks starvation if a job doesn't terminate)
or can use RR.
3.2.6.6 Multiple-Level Queues (MLQ) Scheduling
The scheduling policies discussed so far are more or less suited to particular
applications, with potentially poor performance when applied inappropriately.
What should one use in a mixed system, with some time-critical events, a

Lesson No. 1 Intro. to Operating System 95

multitude of interactive users, & some very long non-interactive jobs? One
approach is to combine several scheduling disciplines. A mix of scheduling
disciplines may best service a mixed environment, each charged with what it
does best. For example, operating-system processes & device interrupts may be
subjected to event-driven scheduling, interactive programs to round robin
scheduling, & batch jobs to FCFS or STRN.

Event Driven Scheduling

High priority queue
System Processes Round Robin Scheduling
Medium priority queue CPU
Interactive Jobs FCFS Scheduling Between Queue Scheduling
Low priority queue
Batch Processes

Multilevel Queue Scheduling

One way to implement complex scheduling is to classify the workload according

to its characteristics, & to maintain separate process queues serviced by different
schedulers. This approach is often called multiple-level queues (MLQ)
scheduling. A division of the workload might be into system processes,
interactive programs, & batch jobs. This would result in three ready queues, as
depicted in above Figure. A process may be assigned to a specific queue on the
basis of its attributes, which may be user-or system-supplied. Each queue may
then be serviced by the scheduling discipline best suited to the type of workload
that it contains. Given a single server, some discipline must also be devised for
scheduling between queues. Typical approaches are to use absolute priority or
time slicing with some bias reflecting relative priority of the processes within
specific queues. In the absolute priority case, the processes from the highest-
priority queue (e.g. system processes) are serviced until that queue becomes
empty. The scheduling discipline may be event-driven, although FCFS should not
be ruled out given its low overhead & the similar characteristics of processes in
that queue. When the highest-priority queue becomes empty, the next queue
may be serviced using its own scheduling discipline (e.g., RR for interactive

Lesson No. 1 Intro. to Operating System 96

processes). Finally, when both higher-priority queues become empty, a batch-
spawned process may be selected. A lower-priority process may, of course, be
preempted by a higher-priority arrival in one of the upper-level queues. This
discipline maintains responsiveness to external events & interrupts at the
expense of frequent preemption’s. An alternative approach is to assign a certain
percentage of the processor time to each queue, commensurate with its priority.
Multiple queues scheduling is a very general discipline that combines the
advantages of the "pure" mechanisms discussed earlier. MLQ scheduling may
also impose the combined overhead of its constituent scheduling disciplines.
However, assigning classes of processes that a particular discipline handles
poorly by itself to a more appropriate queue may offset the worst-case behavior
of each individual discipline. Potential advantages of MLQ were recognized early
on by the O/S designers who have employed it in the so-called fore-
ground/background (F/B) system. An F/B system, in its usual form, uses a two-
level queue-scheduling discipline. The workload of the system is divided into two
queues-a high-priority queue of interactive & time-critical processes & other
processes that do not service external events. The foreground queue is serviced
in the event-driven manner, & it can preempt processes executing in the
background.
3.2.6.7 Multiple-Level Queues with Feedback Scheduling
Multiple queues in a system may be used to increase the effectiveness &
adaptive ness of scheduling in the form of multiple-level queues with feedback.
Rather than having fixed classes of processes allocated to specific queues, the
idea is to make traversal of a process through the system dependent on its run-
time behavior. For example, each process may start at the top-level queue. If the
process is completed within a given time slice, it departs the system after having
received the royal treatment. Processes that need more than one time slice may
be reassigned by the operating system to a lower-priority queue, which gets a
lower percentage of the processor time. If the process is still now finished after
having run a few times in that queue, it may be moved to yet another, lower-level
queue. The idea is to give preferential treatment to short processes & have the

Lesson No. 1 Intro. to Operating System 97

resource-consuming ones slowly "sink" into lower-level queues, to be used as
fillers to keep the processor utilization high. This philosophy is supported by
program-behavior research findings suggesting that completion rate has a
tendency to decrease with attained service. In other words, the more service a
process receives, the less likely it is to complete if given a little more service.
Thus the feedback in MLQ mechanisms tends to rank the processes dynamically
according to the observed amount of attained service, with a preference for those
that have received less.
On the other hand, if a process surrenders control to the OS before its time slice
expires, being moved up in the hierarchy of queues may reward it. As before,
different queues may be serviced using different scheduling discipline. In contrast
to the ordinary multiple-level queues, the introduction of feedback makes
scheduling adaptive & responsive to the actual, measured run-time behavior of
processes, as opposed to the fixed classification that may be defeated by
incorrect guessing or abuse of authority. A multiple-level queue with feedback is
the most general scheduling discipline that may incorporate any or all of the
simple scheduling strategies discussed earlier. Its overhead may also combine
the elements of each constituent scheduler, in addition to the overhead imposed
by the global queue manipulation & the process-behavior monitoring necessary
to implement this scheduling discipline.
3.3 SUMMARY
An important, although rarely explicit, function of process management is
processor allocation. Three different schedulers may coexist & interact in a
complex operating system: long-term scheduler, medium-term scheduler, &
short-term scheduler. Of the presented scheduling disciplines, FCFS scheduling
is the easiest to implement but is a poor performer. SRTN scheduling is optimal
but unrealizable. RR scheduling is most popular in time-sharing environments, &
event-driven & earliest-deadline scheduling are dominant in real-time & other
systems with time-critical requirements. Multiple-level queue scheduling, & its
adaptive variant with feedback, is the most general scheduling discipline suitable

Lesson No. 1 Intro. to Operating System 98

for complex environments that serve a mixture of processes with different
characteristics.
3.4 Keywords
Long-term scheduling: the decisions to introduce new processes for execution, or

re-execution.

Medium-term scheduling: the decision to add to (grow) the processes that are

fully or partially in memory.

Short-term scheduling: the decisions as to which (Ready) process to execute

next.

Non-preemptive scheduling: In non-preemptive scheduling, process will continue

to execute until it terminates, or makes an I/O request which would block the
process, or makes an operating system call.
In preemptive scheduling, the process may be pre-empted by the operating
system when a new process arrives (perhaps at a higher priority), or an interrupt
or signal occurs, or a (frequent) clock interrupt occurs.
3.5 SELF-ASSESSMENT QUESTIONS (SAQ)
1. Discuss various process scheduling policies with their cons & pros.
2. Define process. What is the difference between a process & a program?
Explain.
3. What are the different states of a process? Explain using a process state
transition diagram.
4. Which type of scheduling is used in real life operating systems? Why?
5. Which action should the short-term scheduler take when it is invoked but no
process is in the ready state? Is this situation possible?
6. How can we compare performance of various scheduling policies before
actually implementing them in an operating system?
7. SJF is a sort of priority scheduling. Comment.
8. What do you understand by starvation? How does SJF cause starvation?
What is the solution of this problem?

Lesson No. 1 Intro. to Operating System 99

9. What qualities are to be there in a scheduling policy? Explain.
10. Differentiate between user-oriented scheduling criteria & system-oriented
scheduling criteria.
3.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin
P.B., John Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New
Delhi.
3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson
Education Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing
Company Ltd., New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 100

Lesson number: 5 Writer: Dr. Rakesh Kumar
Memory Management - I Vetter: Prof. Dharminder Kr.

5.0 OBJECTIVE
The lesson presents the principles of managing the main memory, one of the most precious resources in a
multiprogramming system. In our sample hierarchy of OS layers, memory management belongs to layer 3. Memory
management is primarily concerned with allocation of physical memory of finite capacity to requesting processes. No
process may be activated before a certain amount of memory can be allocated to it. The objective of this lesson is to
make the students acquainted with the concepts of contiguous memory management.
5.1 INTRODUCTION
Memory is large array of words or bytes, each having its unique address. CPU fetches instructions from memory
according to value of program counter. The instructions undergo instruction execution cycle. To increase both CPU
utilization & speed of its response to users, computers must keep several processes in memory. Specifically, the memory
management modules are concerned with following four functions:

1. Keeping track of whether each location is allocated or unallocated, to which

process & how much.
2. Deciding to whom should the memory is allocated, how much, when & where.
If memory is to be shared by more than one process concurrently, it must be
determined which process’ request should be satisfied.
3. Once it is decided to allocate memory, the specific locations must be selected
& allocated. Memory status information is updated.
4. Handling the deallocation/reclamation of memory. After the process holding
memory is finished, memory locations held by it are declared free by
changing the status information.
There are varieties of memory management systems. They are:
1. Contiguous, real memory management system such as:
• Single, contiguous memory management system
• Fixed partitioned memory management system
• Variable Partitioned memory management system
2. Non-Contiguous, real memory management system
• Paged memory management system
• Segmented memory management system
• Combined memory management system
3. Non-Contiguous, virtual memory management system
• Virtual memory management system

Lesson No. 1 Intro. to Operating System 101

These systems can be divided into two major parts (i) Contiguous & (ii) Non-
Contiguous
Contiguous Memory Management: In this approach, each program occupies a
single contiguous block of storage locations.
Non-Contiguous Memory Management: In these, a program is divided into
several blocks or segments that may be placed throughout main storage in
pieces or chunks not necessarily adjacent to one another. It is the function of OS
to manage these different chunks in such a way that they appear to be
contiguous to the user.
Various issues to be considered in various memory management schemes are
relocation, address translation, protection, sharing, & evaluation.
Relocation & address translation: The process of associating program
instructions & data to physical memory addresses is called address binding or
relocation. So binding is mapping from one address to another. It is of two types:
¾ Static Binding: It is taking place before execution; it may be (i) Compile time:
where the compiler or assembler translates symbolic addresses to absolute
addresses & (ii) Load time where the compiler translates symbolic addresses
to relative addresses. The loader translates these to absolute addresses.
¾ Dynamic Binding: In it new locations are determined during execution. The
program retains its relative addresses. The absolute addresses are generated
by hardware.
Memory Protection & Sharing: Protection is used to avoid interference
between programs existing in memory. Sharing is the opposite of protection.
Evaluation: Evaluation of these schemes is done on various parameters such
as:
¾ Wasted memory: It is the amount of physical memory, which remains
unused & thus wasted.
¾ Access time is the time to access the physical memory by the OS.
¾ Time complexity is related to overheads of the allocation or deallocation
methods.
5.2 PRESENTATION OF CONTENTS

Lesson No. 1 Intro. to Operating System 102

5.2.1 SINGLE CONTIGUOUS MEMORY MANAGEMENT
5.2.2 FIXED PARTITIONED MEMORY MANGEMENT SYSTEM
5.2.2.1 Principles of Operation
5.2.2.2 Fragmentation
5.2.2.3 Swapping
5.2.2.4 Relocation
5.2.2.4.1 Static Relocation
5.2.2.4.2 Dynamic Relocation
5.2.2.5 Protection

5.2.2.6 Sharing
5.2.2.7 Evaluation
5.2.3 VARIABLE PARTITIONED MEMORY ALLOCATION
5.2.3.1 Principles of Operation
52.3.2 Compaction
5.2.3.3 Protection
5.2.3.4 Sharing

5.2.3.5 Evaluation
5.2.4 SEGMENTATION
5.2.4.1 Principles of Operation
5.2.4.2 Protection
5.2.4.3 Sharing
5.2.1 SINGLE CONTIGUOUS MEMORY MANAGEMENT
In this scheme, the physical memory is divided into two contiguous areas. One of
them is permanently allocated to the resident portion of the OS. Mostly, the OS
resides in low memory (0 to P as shown in Figure 1). The remaining memory is
allocated to transient or user processes, which are loaded & executed one at a
time, in response to user commands. This process is run to completion & then
the next process is brought in memory.
In this scheme, the starting physical address of the program is known at the time of compilation. The machine contains
absolute addresses. They do not need to be changed or translated at the time of execution. So there is no issue of
relocation or address translation.

00

Lesson No. 1 Intro. to Operating System 103

 OS (monitor)

Transient-Process Area

Max
Figure 1. Single contiguous memory management
In this scheme as there is at most one process is in memory at any given time so there is a rare issue of interference
between programs. However, it is desirable to protect the OS code from being tampered by the executing transient
process.

A common way used in embedded systems to protect the OS code from user
programs is to place the OS in read-only memory. This method is rarely used
because of its inflexibility & inability to patch & update the OS code. In systems
where the OS is in read-write memory, protection from user processes usually
requires some sort of hardware assistance such as the fence registers &
protection bits.
Fence registers are used to draw a boundary between the OS & the transient-process area. Assuming that the resident
portion of the OS is in low memory, the fence register is set to the highest address occupied by OS code. Each memory
address generated by a user process is compared against the fence. Any attempt to read or write the space below the
fence may thus be detected & denied before completion of the related memory reference. Such violations usually trap to
the OS, which in turn may abort the offending program. To serve the purpose of protection, modification of the fence
register must be a privileged operation not executable by user processes. Consequently, this method requires the
hardware ability to distinguish between execution of the OS & of user processes, such as the one provided by user &
supervisor modes of operation.

Another approach to memory protection is to record the access rights in the

memory itself. One possibility is to associate a protection bit with each word in
memory. The memory may then easily be divided into two zones of arbitrary size
by setting all protection bits in one area, & resetting them in the other area. For
example, initially all protection bits may be reset. During system startup,
protection bits may be set in all locations where the OS is loaded. User programs
may then be loaded & executed in the remaining memory locations. Prohibiting
user processes from accessing any memory location whose protection bit is set
may enforce OS protection. At the same time, the OS & system utilities, such as
the loader, may be allowed unrestricted access to memory necessary for their
activities. This approach requires a hardware-supported distinction between at
least two distinct levels of privilege in the execution of machine instructions.

Lesson No. 1 Intro. to Operating System 104

Sharing of code & data in memory does not make much sense in single-process
environments, & single-process OS hardly ever support it. Users’ programs may
of course, pass data to teach other in private arrangements, say, by means of
memory locations known to be safe from being overwritten between executions
of participating processes. Such schemes are obviously unreliable, & their use
should be avoided whenever possible.
Single-process OS are relatively simple to design & to comprehend. They are
often used in systems with little hardware support. But the lack of support for
multiprogramming reduces utilization of both processor & memory. Processor
cycles are wasted because there is no pending work that may be executed while
the running process is waiting for completion of its I/O operations. Memory is
underutilized because its portion not devoted to the OS & the single active user is
wasted. On the average, wasted memory in a specific system is related to the
difference between the size of the transient-process area & the average process
size weighted by the respective process-execution (and residence) times. This
method has fast access time & very little time-complexity. Its usage is limited due
to lack of multi-user facility.
One additional problem is sometimes encountered in systems with simplistic

static forms of memory management. To be useable across a wide range of

configurations with different capacities of installed memory, system programs in

such environments tend to be designed to use the least amount of memory

possible. Besides sacrificing speed & functionality, such programs usually take

little advantage of additional memory when it is available.

5.2.2 FIXED PARTITIONED MEMORY MANGEMENT SYSTEM

In this scheme, memory is divided into number of contiguous regions called
partitions, could be of different sizes. But once decided, they could not be
changed. Partitions are fixed at the time of system generation. System
generation is a process of setting the OS to specific requirements. Various

Lesson No. 1 Intro. to Operating System 105

processes of the OS are allotted different partitions. There are two forms of
memory partitioning (i) Fixed Partitioning & (ii) Variable Partitioning.
In fixed partitioning the main memory is divided into fixed number of partitions
during system startup. The number & sizes of individual partitions are decided by
the factors like capacity of the available physical memory, desired degree of
multiprogramming, & the typical sizes of processes most frequently run on a
given installation. Since, in principle, at most one process may execute out of a
given partition at any time, the number of partitions represents an upper limit on
the number of active processes in a system i.e. degree of multiprogramming.
Given the impact of memory partitioning on overall performance, some systems
allow for manual redefinition of partition sizes.
Programs are queued to run in the smallest available partition. An executable
prepared to run in one partition may not be able to run in another without being
relinked. This technique is called absolute loading.
5.2.2.1 Principles of Operation
An example of partitioned memory is depicted in Figure 2. Out of the six
partitions, one is assumed to be occupied by the resident portion of the OS, &
three others by user processes Pi, Pj, & Pk, as indicated. The remaining two
partitions, shaded in Figure 2, are free & available for allocation.
0K

P0 OS Area
100K

P1
300K
P2 Process Pi
400K
P3
Process Pj
700K
P4
800K
P5 Process Pk
1000K

Lesson No. 1 Intro. to Operating System 106

 Figure 2 – Fixed Partitions
On declaring fixed partitions, the OS creates a Partition Description Table (PDT)
to keep track of status of each partition for allocation purposes. A sample PDT
format is given in Figure 3 according to the partitions given in Figure 2.

Partition Number Partition Base Partition size Partition Status

0 0K 100K Allocated
1 100K 200K Free
2 300K 100K Allocated
3 400K 300K Allocated
4 700K 100K Free
5 800K 200K Allocated
Figure 3 – Partition description table
Each partition is described by its base address, size, & status. When fixed

partitioning is used, only the status field of each entry varies i.e. free or allocated,

in the course of system operation. Initially, all the entries are marked “FREE”. As

& when process is loaded into partitions, the status entry for that partition is

changed to “ALLOCATED”.

Initially, all memory is available for user processes & is called hole. On arrival of

a process, a hole large enough for that process is allocated to it. The OS then

reads the program image from disk to the space reserved. After becoming

resident in memory, the newly loaded process makes a transition to the ready

state & thus becomes eligible for execution.

When a nonresident process is to be activated, the OS searches a free memory

partition of sufficient size in the PDT. If the search is successful, the status field
of the selected entry is marked ALLOCATED, & the process image is loaded into
the corresponding partition. Since the assumed format of the PDT does not
provide any indication as to which process is occupying a given partition, the

Lesson No. 1 Intro. to Operating System 107

identity of the assigned partition may be recorded in the PCB. When the process
departs, using this information the status of related partition is made FREE. To
implement these ideas, two questions are to be answered, (i) how to select a
specific partition for a given process, (ii) What to do when no suitable partition is
available for allocation. The strategies of partition allocation are:
First-fit: This strategy allocates the first available space that is big enough to
accommodate process. Search may start at beginning of set of holes or where
previous first-fit ended. Searching stops as soon as it finds a free hole that is
large enough.
Best-fit: This strategy allocates the smallest hole that is big enough to
accommodate process. Entire list ordered by size is searched & matching
smallest left over hole is chosen.
Worst fit: This strategy allocates the largest hole. Entire list is searched. It
chooses largest left over hole.
These strategies may be compared on the basis of execution speed & memory
utilization must be made. These algorithms have to search the PDT to identify a
free partition of adequate size. However, while the first fit terminates upon finding
the first such partition, the best fit must process all PDT entries to identify the
tightest fit. So first fit tend to execute faster but best fit may achieve higher
utilization of memory by creating the smallest possible gap resulting from the
difference in size between the process & its allocated partition. Both first-fit &
best fit are better than worst-fit in terms of time & storage utilization, but first-fit is
faster.
In case of a relatively small number of fixed partitions in a system, the execution
time differences between the these approaches may not be large enough to
outweigh the lower degree of memory utilization attributable to the first fit. When
the number of partitions is large neither first fit nor best fit is clearly superior.
Request for partitions may be due to (1) creations of new processes or (2)
reactivations of swapped-out processes. The memory manager attempts to
satisfy these requests from the pool of free partitions. Common obstacles faced
by it are:

Lesson No. 1 Intro. to Operating System 108

1. No free partition is large enough to accommodate the incoming process.
2. All partitions are allocated.
3. Some partitions are free, but none of them is large enough to accommodate
the incoming process.
If the process to be created is too large to fit into any of the system partitions, the
OS produces an error message. This is basically a configuration error that may
be remedied by redefining the partitions accordingly. Another option is to reduce
a program's memory requirements by recording & possibly using some sort of
overlays.
The case when all partitions are allocated may be handled by deferring loading of
the incoming process until a suitable partition can be allocated to it. An
alternative is to force a memory-resident process to vacate a sufficiently large
partition. Eviction to free the necessary space incurs the additional overhead of
selecting a suitable victim & rolling it out to disk. This technique is called
swapping. Both deferring & swapping are also applicable to handling the third
case, where free but unsuitable partitions are available. If the deferring option is
chosen, memory utilization may be kept high if the OS continues to allocate free
partitions to other waiting processes with smaller memory requirements.
However, doing so may violate the ordering of process activation’s intended by
the scheduling algorithm and, in turn, affect performance of the system.
The described memory-allocation situations illustrate the close relationship &
interaction between memory management & scheduling functions of the OS.
Although the division of labor in actual systems may vary, the memory manager
is generally charged with implementing memory allocation & replacement
policies. Processor scheduling, on the other hand, determines which process
gets the processor, when, & for how long. The short-term scheduler considers
only the set of ready processes, that is, those that have all the needed resources
except for the processor. Ready processes are, by definition, resident in memory.
By influencing the membership of the set of resident processes, a memory
manager may affect the scheduler's ability to perform. On the other hand, the

Lesson No. 1 Intro. to Operating System 109

effectiveness of the short-term scheduler influences the memory manager by
affecting the average memory-residence times of processes.
In systems with fixed partitioning of memory, the number of partitions effectively
sets an upper limit on the degree of multiprogramming. Within the confines of this
limit, processor utilization may be improved by increasing the ratio of ready to
resident processes. This may be accomplished by removing suspended
processes from memory when otherwise ready ones are available for loading in
the related partitions. A removed process is usually kept in secondary storage
until all resources needed for its execution, except for memory & the processor
may be allocated to it. At that point, the process in question becomes eligible for
loading into the main memory. The medium-term scheduler & the memory
manager cooperate in further processing of such processes.
The OS holds the processes waiting to be loaded in the memory in a queue. The
two methods of maintaining this queue are (i) Multiple Queues & (ii) Single
Queues.
Multiple Queues: In this method there are as many queues as the number of
partitions. Separate queue for each partition is maintained in which processes
are added as they arrive. When a process wants to occupy memory, it is added
to a proper queue depending upon size of processes. Benefit of this method is
that a small process is not loaded in large partition so as to avoid memory
wastage. This leads to longer queue for small partitions.
Single Queue: In this method, there is only one queue for all ready processes.
The order of processes in the queue depends on the scheduling algorithm. In this
case, first fit allocation strategy is more efficient & fast.
5.2.2.2 Fragmentation
Some amount of memory is wasted both in single & multiple partition allocation
techniques. Fragmentation refers to the unused memory that the memory
management system cannot allocate. It is of two types: External & Internal.
External Fragmentation is waste of memory between partitions caused by
scattered non-contiguous free space. It occurs when total available memory
space is enough to satisfy the request for a process to be allocated, but it is not

Lesson No. 1 Intro. to Operating System 110

continuous. Selection of first fit & best fit can affect the amount of fragmentation.
It is severe in variable size partitioning schemes. Compaction is a technique that
is used to overcome this.
Internal fragmentation is waste of memory within a partition caused by
difference between size of partition & the process allocated. It refers to the
amount of memory, which is not being used & is allocated along with a process
request i.e. available memory internal to partition. It is severe in fixed partitioning
schemes.
5.2.2.3 Swapping
Removing suspended or preempted processes from memory & their subsequent
bringing back is called swapping. The basic idea of swapping is to treat main
memory as a ‘pre-emptable’ resource. Lifting the program from the memory &
placing it on the disk is called ‘Swapping out’. To bring the program again from
the disk into the main memory is called ‘Swapping in’. Normally, a blocked
process is swapped out so as to create available space for a ready process. This
results in improving CPU utilization. Swapping has traditionally been used to
implement multiprogramming in systems with restrictive memory capacity.
Swapping may also be helpful for improving processor utilization in partitioned
memory environments by increasing the ratio of ready to resident processes.
Swapping is usually employed in memory-management systems with contiguous
allocation, such as fixed & variable partitioned memory & segmentation.
Somewhat modified forms of swapping may also be present in virtual memory
systems based on segmentation or on paging. Swapping brings flexibility even to
systems with fixed partitions.
When the scheduler decides to admit a new process for which no suitable free partition can be found, the swapper may
be invoked to vacate such a partition. The swapper is an OS process whose major responsibilities include:

¾ Selection of processes to swap out: Its criteria is suspended/blocked state,

low priority, time spent in memory.
¾ Selection of processes to swap in: Its criteria are time spent on swapping
device & priority.
¾ Allocation & management of swap space on a swapping device. Swap space
can be system wide or dedicated.

Lesson No. 1 Intro. to Operating System 111

Thus the swapper performs most of the functions of the medium-term scheduler.
The swapper usually selects a victim among the suspended processes that
occupy partitions large enough to satisfy the needs of the incoming process.
Although the mechanics of swapping out following the choice of a victim process
is fairly simple in principle, implementation of swapping requires some specific
provisions & considerations in OS that support it. These generally include the file
system, specific OS services, & relocation.

Figure 4 showing process of Swapping

A process is typically prepared for execution & submitted to the OS in the form of
a file that contains a program in executable form & the related data. This file may
also contain process attributes, such as priority & memory requirements. Such a
file is sometimes called a process image. Since a process usually modifies its
stack & data when executing, a partially executed process generally has a run-
time image different from its initial static process image recorded on disk.
Therefore, the dynamic run-time state of the process to be swapped out must be
recorded for its proper subsequent resumption. In general, the modifiable portion
of a process's state consists of the contents of its data & stack locations, as well
as of the processor registers. Code is also subject to run-time modifications in
systems that permit the code to modify itself. Therefore, the contents of a sizable
portion or of the entire address space of a victim process must be copied to disk
during the swapping-out operation. Since the static process image is used for
initial activation, the (modified) run-time image should not overwrite the static
process image on disk. Consequently, a separate swap file must be available for

Lesson No. 1 Intro. to Operating System 112

storing the dynamic image of a rolled-out process. There are two basic options
regarding placement of a swap file:
• System-wide swap file
• Dedicated, per-process, swap files
In either case, swapping space for each swappable process is usually reserved &
allocated statically, at process creation time, to avoid the overhead of this
potentially lengthy operation at swap time.
In the system-wide swap file approach, a single large file is created, usually in
the course of system initialization, to handle swapping requirements of all
processes. The swap file is commonly placed on a fast secondary-storage device
so as to reduce the latency of swapping. The location of each swapped out
process image is noted within that file. An important trade-off in implementing a
system-wide swap file is the choice of its size. If a smaller area is reserved for
this file, the OS may not be able to swap out processes beyond a certain limit,
thus affecting the performance.
An alternative is to have a dedicated swap file for each swappable process in the
system. These swap files may be created either dynamically at process creation
time or statically at program preparation time. This method is very flexible, but
can be very inefficient due to the increased number of files & directories. In either
case, the advantages of maintenance of separate swap files include elimination
of the system swap-file dimensioning problem & of that file's overflow errors at
run-time, & non-imposition of restrictions on the number of active processes. The
disadvantages include more disk space expended on swapping, slower access,
& more complicated addressing of swapping files scattered on the secondary
storage.
Regardless of the type of swapping file used, the need to access secondary
storage makes swapping a lengthy operation relative to processor instruction
execution. This overhead must be taken into consideration in the decision of
whether to swap a process in order to make room for another one.
Delays of this magnitude may be unacceptable for interrupt-service routines or
other time-critical processes. For example, swapping out of a momentarily

Lesson No. 1 Intro. to Operating System 113

inactive terminal driver in a time-sharing system is certainly a questionable
"optimization." OS that support swapping usually cope with this problem by
providing some means for system programmers to declare a given process as
being swappable or not. In effect, after the initial loading, an unswappable
process remains fixed in memory even when it is temporarily suspended.
Although this service is useful, a programmer may abuse it by declaring an
excessive number of processes as fixed, thereby reducing the benefits of
swapping. For this reason, the authority to designate a process as being un-
swappable is usually restricted to a given class of privileged processes & users.
All other processes, by default, may be treated as swappable.
An important issue in systems that support swapping is whether process-to-
partition binding is static or dynamic, i.e., whether a swapped-out process can
subsequently be loaded only into the specific partition from which it was removed
or into any partition of adequate size. In general, static binding of processes to
partitions may be done in any system with static partitioning of memory,
irrespective of whether swapping is supported or not. Static process-to-partition
binding eliminates the run-time overhead of partition allocation at the expense of
lower utilization of memory due to potentially unbalanced use of partitions. On
the other hand, systems in which processes are not permanently bound to
specific partitions are much more flexible & have a greater potential for efficient
use of memory. The price paid for dynamic binding of processes to partitions is
the overhead incurred by partition allocation whenever a new process or a
swapped process is to be loaded into main memory. Moreover, dynamic
allocation of partitions usually requires some sort of hardware support for
dynamic relocation.
5.2.2.4 Relocation
The term program relocatability refers to the ability to load & execute a given
program into an arbitrary place in memory. Since different load addresses may
be assigned during different executions of a single relocatable program, a
distinction is often made between virtual addresses (or logical address) & the
physical addresses where the program & its data are stored in memory during a

Lesson No. 1 Intro. to Operating System 114

given execution. In reality, the program may be loaded at different memory
locations, which are called physical addresses. The problem of relocation &
address translation is to find a way to map virtual addresses onto physical
addresses. Depending on when & how the mapping from the virtual address
space to the physical address space takes place in a given relocation scheme,
there are two basic types of relocation: (i) Static relocation & (ii) Dynamic
relocation.
5.2.2.4.1 Static Relocation
Static relocation is performed before or during the loading of the program into
memory, by a relocating linker/ loader. Constants, physical I/O port addresses, &
offsets relative to the program counter are examples of values that are not
location-sensitive & that do not need to be adjusted for relocation. Other forms of
addresses of operands may depend on the location of a program in memory so
must be adjusted accordingly when the program is being loaded or moved to a
different area of memory.
A language translator typically prepares the object module by assuming the
virtual address 0 to be the starting address of the program, thus making virtual
addresses relative to the program loading address. Relocation information,
including virtual addresses that need adjustment following determination of the
physical load address, is provided for subsequent processing by the linker &
loader. Either when the linker combines object modules or when the process
image is being loaded, all program locations that need relocation are adjusted in
accordance with the actual starting physical address allocated to the program.
Once the program is in memory, values that need relocation are indistinguishable
from those that do not.
Since relocation information in memory is usually lost following the loading, a
partially executed statically relocatable program cannot be simply copied from
one area of memory into another & be expected to continue to execute properly.
In systems with static relocation a swapped-out process must either be swapped
back into the same partition from which it was evicted, or software relocation
must be repeated whenever the process is to be loaded into a different partition.

Lesson No. 1 Intro. to Operating System 115

Given the considerable space & time complexity of software relocation, systems
with static relocation are practically restricted to supporting only static binding of
processes to partitions. This method is slow process because it involves software
translation. It is used only once before the initial loading of the program.
5.2.2.4.2 Dynamic Relocation
In it, mapping from the virtual address space to the physical address space is
performed at run-time. Process images in systems with dynamic relocation are
also prepared assuming the starting location to be a virtual address 0, & they are
loaded in memory without any relocation adjustments. When the related process
is being executed, all of its memory references are relocated during instruction
execution before physical memory is actually accesses. This process is often
implemented by means of specialized base registers. After allocating a suitable
partition & loading a process image in memory, the OS sets a base register to
the starting physical load address. This value is normally obtained from the
relevant entry of the PDT. Each memory reference generated by the executing
process is mapped into the corresponding physical address by having the
contents of the base register added to it.
Dynamic relocation is illustrated in Figure 5. A sample process image prepared
with an assumed starting address of virtual address 0 is shown unchanged
before & after being loaded in memory. In this particular example, it is assumed
that address 1000 is allocated as the starting address for loading the process
image. This base address is normally available from the corresponding entry of
the PDT, which is reachable by means of the link to the allocated partition in the
PCB. Whenever the process in question is scheduled to run, the base register is
loaded with this value in the course of process switching.
Base Register
1000 0
IR MAR 500
1000 LDA
500 + 1500
1500 500
2000
Virtual Physical
Address Address 2500

Physical
Lesson No. 1 Intro. to Operating System 116 Memory
 Figure 5 – Dynamic relocation
Relocation of memory references at run-time is illustrated by means of the
instruction LDA 500, which is supposed to load the contents of the virtual
address 500 (relative to program beginning) into the accumulator. As indicated,
the target item actually resides at the physical address 1500 in memory. This
address is produced by hardware by adding the contents of the base register to
the virtual address given by the processor at run-time.
As suggested by Figure 4, relocation is performed by hardware & is invisible to
programmers. In effect, all addresses in the process image are prepared by
counting on the implicit based addressing to complete the relocation process at
run-time. This approach makes a clear distinction between the virtual & the
physical address space.
This is the most commonly used scheme amongst the schemes using fixed
partitions due to its enhanced speed & flexibility. Its advantage is that it supports
swapping easily. Only the base register value needs to be changed before
dispatching.
5.2.2.5 Protection
Not only must the OS be protected from unauthorized tampering by user processes, but each user process must also be
prevented from accessing the areas of memory allocated to other processes. Otherwise, a single erroneous or malevolent
process may easily corrupt any or all other resident processes. There are two approaches for preventing such
interference & achieving protection. These approaches involve the use of Limit Register & Protection Bits.

Implementation of memory protection in a given system tends to be greatly

influenced by the available hardware support. In systems that use base registers

for relocation, a common practice is to use limit registers for protection. The

primary function of a limit register is to detect attempts to access address space

beyond the boundary assigned to the executing program by the OS. The limit

register is usually set to the highest virtual address in a program. As illustrated by

Figure 6, each intended memory reference of an executing program is checked

Lesson No. 1 Intro. to Operating System 117

against the contents of the limit register before being forwarded to memory. In

this way, any attempt to access a memory location outside of the specified area

is detected & aborted by the protection hardware before being allowed to reach

the memory. This violation usually traps to the OS, which may then take a

remedial action, such as to terminate the offending process. The base & limit

values for each process are normally kept in its PBC. Upon each process switch,

the hardware base & limit registers are loaded with the values required for the

new running process. Another approach to protection is to record the access

rights in the memory itself. The bit-per-word approach described earlier, is not

suitable for multiprogramming systems because it can separate only two distinct

address spaces. Adding more bits to designate the identity of each word’s owner

may solve this problem, but this approach is rather costly. A more economical

version of this idea has been implemented by associating a few individual words.

For example, some models of the IBM 360 series use four such bits, called keys,

per each 2 KB block of memory. When a process is loaded in memory, its identity

is recorded in the protection bits of the occupied blocks. The validity of memory

references is established at run-time by comparison of the running process's

identity to the contents of protection bits of the memory block being accessed. If

no match is found, the access is illegal & hardware traps to the OS for

processing of the protection-violation exception. The OS is usually assigned a

unique "master" key, say 0, that gives it unrestricted access to all blocks of

memory. Note that this protection mechanism imposes certain restrictions on

operating-system designers. For example, with 4-bit keys the maximum number

Lesson No. 1 Intro. to Operating System 118

of static partitions & of resident processes is 16. Likewise, associating protection

bits with fixed-sized blocks forces partition sizes to be an integral number of such

blocks.

Base Register 0

<= YES
+
CPU

NO

Protection Violation max

Memory
Figure 6 – Base-limit register
5.2.2.6 Sharing
Sharing of code & data poses a serious problem in fixed partitions because it
might compromise on protection. There are three basic approaches to sharing in
systems with fixed partitioning of memory:
• Entrust shared objects to OS.
• Maintain multiple copies, one per participating partition, of shared objects.
• Use shared memory partitions.
The easiest way to implement sharing without significantly compromising
protection is to entrust shared objects to the OS. It means that any code or data
goes through the OS for any request because the OS has the controlling access
to shared resources. No additional provisions may be needed to support sharing.
This scheme is possible but very tedious. Unfortunately, this simple approach
increases the burden on the OS. Therefore, it is not followed in practice.

Lesson No. 1 Intro. to Operating System 119

Unless objects are entrusted to the OS, sharing is quite difficult in systems with
fixed partitioning of memory. The primary reason is their reliance on rather
straightforward protection mechanisms based mostly on the strict isolation of
distinct address spaces. Since memory partitions are fixed, disjoint, & usually
difficult to access by processes not belonging to the OS, static partitioning of
memory is not very conducive to sharing.
Another approach is to keep copies of the sharable code/ data in all partitions
where required. It is wasteful & leads to inconsistencies. Since there is no
commonly accessible original, each process runs using its copy of the shared
object. Consequently, updates are made only to copies of the shared object. For
consistency, updates made to any must be propagated to all other copies, by
copying the shared data from the address space of the running process to all
participating partitions upon every process switch. Swapping, when supported,
introduces the additional complexity of potentially having one or more
participating address spaces absent from main memory. This approach of
sharing does not make much sense in view of the fact that no saving of memory
may be expected.
Another traditional simple approach to sharing is to place the data in a dedicated
"common" partition. However, any attempt by a participating process to access
memory outside of its own partition is normally regarded as a protection violation.
In systems with protection keys, changing the keys of all shared blocks upon
every process switch in order to grant access rights to the currently running
process may circumvent this obstacle. Keeping track of which blocks are shared
& by whom, as well as the potentially frequent need to modify keys, results in
notable OS overhead necessary to support this form of sharing. With base-limit
registers, the use of shared partitions outside of-and potentially discontiguous to-
the running process's partition requires some special provisions.
5.2.2.7 Evaluation
¾ Wasted memory: In fixed partitions, lot of memory is wasted due to both kinds
of fragmentation.

Lesson No. 1 Intro. to Operating System 120

¾ Access Time: Access time is not very high due to the assistance of special
hardware. The translation from virtual address to physical address is done by
hardware itself, thus enabling rapid access.
¾ Time complexity is very low because allocation/ deallocation routines are
simple as the partitions are fixed.
5.2.3 VARIABLE PARTITIONED MEMORY ALLOCATION
In variable partitions, the number of partitions & their sizes are variable as they

are not defined at the time of system generation. Starting with the initial state of

the system, partitions are created dynamically to fit the needs of each requesting

process. When a process departs, the memory manager returns the vacated

space to the pool of free memory areas from which partition allocations are

made. Process is allocated exactly as much memory as required.

5.2.3.1 Principles of Operation

When instructed to load a process image, the memory-management module
attempts to create a suitable partition for allocation to the process in question.
The first step is to locate a contiguous free area of memory, which is equal to or
larger than the process's size declared with the submitted image. If a suitable
free area is found, the OS carves out a partition from it to provide an exact fit to
the process's needs. The leftover chunk of free memory is returned to the pool of
free memory for later use by the allocation module. The partition is created by
entering its base, size, & status into the system PDT. A copy of, or some link to,
this information is normally recorded in the PCB. After loading the process image
into the created partition, the process may be turned over to the OS module
appropriate for its further processing, such as the short-term scheduler. If no
suitable free area can be allocated, the OS returns an error indication.
When a resident process departs, OS returns the partition's space to the pool of
free memory & invalidates the corresponding PDT entry. For swapped-out
processes, the OS also invalidates the PCB field where the identity of the
allocated partition is normally held.

Lesson No. 1 Intro. to Operating System 121

The OS obviously needs to keep track of both partitions & free memory. Once
created, a partition is defined by its base address & size. Those attributes remain
essentially unchanged for as long as the related partition exists. In addition, for
the purposes of process switching & swapping, it is important to know which
partition belongs to a given process.

Operating Operating Operating Operating

System 128K System System System 320K
Process 1 320K Process 1 320K Process 1

Process 2 224K Process 2 224K

896K
576K
Process 3 288K
352K

Operating Operating Operating Operating 64K

System System System System
Process 1 320K Process 1 320K 320K Process 2 224K
96K
224K Process 4 128K Process 4 198K Process 4 224K
96K 96K 96K
Process 3 288K Process 3 288K Process 3 288K Process 3 288K

64K 64K 64K 64K

Figure 7– Partitions in dynamic memory partitioning

Free areas of memory are produced upon termination of partitions & as leftovers
in the partition creation process. For allocation & for partition creation purpose,
the OS must keep track of the starting address & size of each free area of
memory. This information may need to be updated each time a partition is
created or terminated. The highly dynamic nature of both the number & the
attributes of free areas suggest the use of some sort of a linked list to describe
them. It is common to conserve space by building the free list within the free
memory itself. For example, the first few words of each free area may be used to
indicate the size of the area & to house a link to the successor area.
Figure 6 illustrates the working of variable partitioned memory. In this example, first Process 1,Process 2 & Process 3 are
allocated memory as they arrive. When Process 2 is swapped out, the memory freed by Process 2 is available for any
other process. So it is allocated to Process 4 & the size of partition for process 4 also varies. Again when process 2
arrives, it is allocated memory at different location that was freed by Process 1. Moreover, the size of partition also differs

Lesson No. 1 Intro. to Operating System 122

from the size of partition of process 1.From the given example, it is clear that memory is allocated to processes as they
arrive & on availability of memory. Partitions are created at the time of allocation according to size of process & not at the
time of system generation.

Common algorithms for selection of a free area of memory for creation of a

partition (Step 1) are (i) First fit & its variant, next fit, (ii) Best fit, (iii) Worst fit.
Next fit is a modification of first fit whereby the pointer to the free list is saved
following an allocation & used to begin the search for the subsequent allocation.
as opposed to always starting from the beginning of the free list, as is the case
with first fit. The idea is to reduce the search by avoiding examination of smaller
blocks that tend to be created at the beginning of the free list as a result of
previous allocations. In general, next fit was not found to be superior to the first fit
in reducing the amount of wasted memory.
First fit is generally faster because it terminates as soon as a free block large
enough to house a new partition is found. Thus, on the average, first fit searches
half of the free list per allocation. Best fit, on the other hand, searches the entire
free list to find the smallest free block large enough to hold a partition being
created. First fit is faster, but it does not minimize wasted memory for a given
allocation. Best fit is slower, & it tends to produce small leftover free blocks that
may be too small for subsequent allocations. However, when processing a series
of request starting with an initially free memory, neither algorithm has been
shown to be superior to the other in terms of wasted memory.
Worst fit is an antipode of best fit, as it allocates the largest free block, provided
the block size exceeds the requested partition size. The idea behind the worst fit
is to reduce the rate of production of small holes, which are quite common in best
fit. However, some simulation studies indicate that worst fit allocation is not very
effective in reducing wasted memory in the processing of a series of requests.
Termination of partitions in a system with dynamic allocation of memory may be performed by means of the procedure
that recombines free areas, if possible, to reduce fragmentation of memory. When first fit or best fit is used, the free list
may be sorted by address to facilitate recombination of free areas when partitions are deallocated.

Buddy System: This is another method of allocation-deallocation which speeds

up merging of adjacent holes. This method facilitates merging of free space by
allocating free areas with an affinity to recombine. It treats entire space available
as a single block of 2k, Requests for free areas are rounded up to the next integer
power of base 2. When a free block of size 2k is requested, the memory allocator

Lesson No. 1 Intro. to Operating System 123

attempts to satisfy it by allocating a free block from the list of free blocks of size
2k. If none is available, the block of the next larger size, 2k+1, is split in two halves
(buddies) to satisfy the request. An important property of this allocation scheme
is that the base address of the other buddy can be determined given the base
address & size of one buddy (for a block of size 2k, the two addresses differ only
in the binary digit whose weight is 2k). Thus, when a block is freed, a simple test
of the status bit can reveal whether its buddy is also free. If so, the two blocks
can be recombined to form the twice-larger original block. In addition to the free-
list links, a status field is associated with each area of memory to indicate
whether it is in use or not. Free blocks of equal size are often kept in separate
free lists. Advantage of Buddy System is that it coalesces adjacent buffers or
holes. Its major disadvantage is that this method is very inefficient in terms of
memory utilisation.
As an example, consider a system with 1MB of memory (100000H) managed
using the buddy allocation scheme. An initial request for a 192 KB block in such
a system would require allocation of a 256 KB block (rounded up to the size that
is a power of 2). Since no such block is initially available, the memory manager
would form it by splitting the 1 MB block into two 512 KB buddies, & then splitting
one of them to form two 256 KB blocks. The first split produces two 512 KB
blocks (buddies) with starting addresses of 00000H (H stands for hexadecimal) &
80000H, respectively. The second split of the block at 80000H yields two 256 KB
blocks (buddies) that start at 80000H & A0000H, respectively. Assume that the
block starting at A0000H is allocated to the user. When the partition starting as
A0000H is eventually terminated, the memory manager can identify the base
address of its 256 KB buddy (buddies are of the same size) by toggling the
address bit in the position that corresponds to the size of the block being
released. In the presented example, 256 KB = 218, & toggling of that bit yields a 0
(in this example) in bit position 18, which, with all other bits unchanged, produces
the address of the original buddy, 80000H. A quick inspection of the associated
status word indicates whether the buddy at that address is free or not. If it is, the
two buddies are coalesced to reform the 512 KB block starting at address

Lesson No. 1 Intro. to Operating System 124

80000H, which was originally used to produce the smaller blocks to satisfy the
pending request.
5.2.3.2 Compaction
It is one solution to problem of external fragmentation. The goal here is to shuffle
memory contents & place all free memory together in one block. Compaction is
possible only if relocation is dynamic. This technique shifts the necessary
process images to bring the free chunks of memory to adjacent positions to
coalesce. Coalescing of adjacent free areas is a method often used to reduce
fragmentation and the amount of wasted memory. However, such remedies tend
to defer the impact of, rather than to prevent, the problem. The primary reason
for fragmentation is that, due to different lifetimes of resident objects, the pattern
of returns of free areas is generally different from the order of allocations.
When memory becomes seriously fragmented, the only way out may be to
relocate some or all partitions into one end of memory & thus combine the holes
into one large free area. Since affected processes must be suspended & actually
copied from one area of memory into another. It is important to decide when &
how memory compaction is to be performed.
Memory compaction may be performed whenever possible or only when needed. Some systems compact memory
whenever a process departs, thus collecting most of the free memory into a single large area. An alternative is to compact
only upon a failure to allocate a suitable partition, provided that the combined size of free areas exceeds the needs of the
request at hand.

Compaction involves a high overhead, but it increases the degree of

multiprogramming. That is why; OS can accommodate a process with a larger
size, which would have been impossible before compaction.
5.2.3.3 Protection
Protection & sharing in systems with dynamic partitioning of memory are not
significantly different from their counterparts in static partitioning environments,
since they both rely on virtually identical hardware support. One difference is that
dynamic partitioning potentially allows adjacent partitions in physical memory to
overlap. Consequently, a single physical copy of a shared object may be
accessible from two distinct address spaces. This possibility is illustrated in
Figure 8, where partitions A & B overlap to include the shared object placed in
the doubly shaded area. The relevant portion of the partition of the partition

Lesson No. 1 Intro. to Operating System 125

definition table is also shown in Figure 8. As indicated, 500 locations starting
from the physical address 5500 are shared & included in both partitions.
Although perhaps conceptually appealing, this form of sharing is quite restrictive
in practice. Sharing of objects is limited to two processes; when several
processes are in play, one of the more involved schemes described in above
must be used.
5.2.3.4 Sharing
Sharing of code is generally more restrictive than sharing of data. One of the
reasons for this is that shared code must be either reentrant or executed in a
strictly mutually exclusive fashion with no preemption’s. Otherwise, serious
problems may result if a process in the middle of execution of the shared code is
switched off, & another process begins to execute the same section of the
shared code. Reentrancy generally requires that variables be kept on stack or in
registers, so that new activation’s do not affect the state of the preempted,
incomplete executions. Additional complexities in sharing of code are imposed by
the need for shared code to ensure that references to itself-such as local jumps &
access to local data-are mapped properly during executions on behalf of any of
the participating processes. When dynamic relocation with base registers is
used, this means that all references to addresses within a shared-code object
from instructions within that code must reach the same set of physical addresses
where the shared code is stored at run-time, no matter which particular base is
used for a given relocation. This may be accomplished in different ways, such as
by making the shared code position independent or by having shared code
occupy identical virtual offsets in address spaces of all processes that reference
it.

Base Size 4000

A 4000 2000 Area A
A
B 5500 2500 5500 Shared
Area

Lesson No. 1 Intro. to Operating System 126

 6000
B Area B
8000

Figure 8 – Overlapping partitions

Some aspects of the issues involved in self-referencing of shared code are
illustrated in Figure 9, where a subroutine SUB is assumed to be shared by two
processes, PA & PB, whose respective partitions overlap in physical memory as
indicated in Figure 8. Let us assume that the system in question used dynamic
relocation & dynamic memory allocation, thus allowing the two partitions to
overlap. The sizes of the address spaces of the two processes are 2000 & 2500
locations, respectively. The shared subroutine, SUB, occupies 500 locations, & it
is placed in locations 5500 to 5999 in physical memory. The subroutine starts at
virtual addresses 1500 & 0 in the address spaces of processes PA & PB,
respectively. Being shared by the two processes, SUB may be linked with &
loaded with either process image.
Figure 9 also shows the references to SUB from within the two processes. As indicated in Figure 9(a), the CALL SUB at
virtual address 100 of process PA is mapped to the proper physical address of 5500 at run-time by adding the contents of
PA's base register. Likewise the CASS SUB at virtual address 1800 in process PB is mapped to 5500 at run-time by
adding PB's value of the base register. This is illustrated in Figure 9(b). Thus proper referencing of SUB from the two
processes is accomplished even when the two partitions are relocated due to swapping or compaction, provided that they
overlap in the same way in the new set of physical addresses. However, making references from within SUB to itself
poses a problem unless some special provisions are made. For example, a jump using absolute addressing from location
50 to location 100 within SUB should read JUMP 1600 for proper transfer of control when invoked by PA, but JMP 100 if
PB’s invocation is to work properly. Since the JMP instruction may have only one of these two addresses in its
displacement field, there is a problem in executing SUB correctly in both possible contexts.

One way to solve this problem is to use relative references instead of absolute
references within shared code. For example, the jump in question may read JMP
$+50, where $ denotes the address of the JMP instruction. Since it is relative to
the program counter, the JMP is mapped properly when invoked by either
process, that is, to virtual address 1600 or 100, respectively. At run-time,
however, both references map to the same physical address, 5600, as they
should. This is illustrated in Figure 9.
Code that executes correctly regardless of its load address is often referred to as position-independent code. One of its
properties is that references to portions of the position-independent code itself are always relative, say, to the program
counter or to a base when based addressing is used. Position-independent coding is often used for shared code, such as
memory-resident subroutine libraries. In our example, use of position-independent code solves the problem of self-
referencing.

Lesson No. 1 Intro. to Operating System 127

 Base
4000
Register
0 1500
Call Sub (call 1500) 5500 +
100
Base
4000
Register (a) Process
A SUB
1500 JMP $+50 1550+50
+
1550 5600
2000

0 Base
5500
50 SUB Register
JMP $+50
50+50
+
500 5600
Base
5500
Register (b) Process B
0
Call Sub (call 0) +
1800 5500
2000
Figure 9 – Accessing shared code (a) Process A (b) Process B
Position-independent coding is one way to handle the problem of self-referencing
of shared code. The main point of our example, however, is that sharing of code
is more restrictive than sharing of data. In particular, both forms of sharing of
code is more restrictive than sharing of data. In particular, both forms of sharing
require the shared object to be accessible from all address spaces of which it is a
part. ;in addition, shared code must also be reentrant or executed on a mutually
exclusive basis, & some special provisions-such as position-independent coding-
must be made in order to ensure proper code references to itself. Since ordinary
(non-shared) code does not automatically meet these requirements, some
special language provisions must be in place, or assembly language coding may

Lesson No. 1 Intro. to Operating System 128

be necessary to prepare shared code for execution in systems with partitioned
allocation of memory.

5.2.3.5 Evaluation
™ Wasted memory: This memory management scheme wastes less memory
than fixed partitions because there is no internal fragmentation as the partition
size can be of any length. By using compaction, external fragmentation can
also be eliminated.
™ Access Time: Access time is same as of fixed partitions as the same scheme
of address translation using base register is used.
™ Time Complexity: Time complexity is higher in variable partitions due to
various data structures & algorithms used, for eg: Partition Description Table
(PDT) is no more of fixed length.
5.2.4 SEGMENTATION
The external fragmentation & its negative impact should be reduced in systems where the average size of a request for
allocation is smaller. OS cannot reduce the average process size, but a way to reduce the average size of a request for
memory is to divide the address space of a single process into blocks that may be placed into noncontiguous areas of
memory. This can be accomplished by segmentation. Segmentation provides breaking of an address space into several
logical segments, dynamic relocation & sophisticated forms of protection & sharing.

According to user’s view, programs are collections of subroutines, stacks,

functions etc. Each of this component is of variable length & are logically related
entities. Elements within segment are identified by their offset from beginning of
the segment. Segments are formed at program translation time by grouping
together logically related items. For example, a typical process may have
separate code, data, & stack segments. Data or code shared with other
processes may be placed in their own dedicated segments.
All segments of all programs do not have to be of the same length
since the segments are formed as a result of logical division. There is
a maximum segment length. Although different segments may be
placed in separate, noncontiguous areas of physical memory, items
belonging to a single segment must be placed in contiguous areas of
physical memory. Since segments are not equal, segmentation is
similar to dynamic partitioning. Thus segmentation possesses some

Lesson No. 1 Intro. to Operating System 129

properties of both contiguous (with regard to individual segments) &
noncontiguous (with regard to address space of a process) schemes
for memory management.
DATA SEGMENT (DATA)
. .
. .
datum x dw xx
datum y dw yy (STACK)
.
.
(CODE)
DATA ENDS
STACK SEGMENT
Ds 500 (SHARED)
STACK ENDS
CODE SEGMENT
Psub ….
.
.

main …
.

. Segment Map
CODE ENDS
SHARED SEGMENT Segment# Size Type
ssub1 … 0 d data

. 1 500 stack

. 2 c code
ssub2 3 s code
SHARED ENDS

Lesson No. 1 Intro. to Operating System 130

 (a) Segment Definition (b) Load Module

Figure 10 - Segments
Segmentation is quite natural for programmers who tend to think of their programs in terms of logically related entities,
such as subroutines & global or local data areas. A segment is essential a collection of such entities. The segmented
address space of a single process is illustrated in Figure 10(a). In that particular example, four different segments are
defined: DATA, STACK, CODE, & SHARED. Except for SHARED, the name of each segment is chosen to indicate the
type of information that it contains. The STACK segment is assumed to consist of 500 locations reserved for stack. The
SHARED segment consists of two subroutines, SSUB1 & SSUB2, shared with other processes. The definition of the
segments follows the typical assembly-language notation, in which programmers usually have the freedom to define
segments directly in whatever way they feel best suits the needs of the program at hand. As a result, a specific process
may have several different segments of the same generic type, such as code or data. For example, both CODE &
SHARED segments contain executable instructions & thus belong to the generic type "code".

4.2.4.1 Principles of Operation

Segmentation is mapping of user’s view onto physical memory. A logical address
space is a collection of segments. Each segment has a name & length. User
specifies each address by segment name (no.) & offset within segment.
Segments are numbered & are referenced by segment number. For relocation
purposes, each segment is compiled to begin at its own virtual address 0. An
individual item within a segment is then identifiable by its offset relative to the
beginning of the enclosing segment. Thus, logical address consists of <segment
no., offset>. In segmented systems, components belonging to a single segment
reside in one contiguous area. But different segments belonging to same process
occupy non-contiguous area of physical memory because each segment is
individually relocated.
For example, the subroutine SSUB2 in segment SHARED is assumed to begin at
offset 100. However, the unique designation of an item in a segmented address
space requires the specification of both its segment & the relative offset therein.
Offset 100 may fetch the first instruction of the subroutine SSUB2 within the
segment SHARED, but the same relative offset may designate an entirely
unrelated datum in the DATA segment.
To simplify processing, segment names are usually mapped to (virtual) segment
numbers. This mapping is static, & systems programs in the course of
preparation of process images may perform it.
A sample linker-produced load module for the segments defined in Figure 10(a)
is depicted in Figure 10(b). Virtual segment numbers are shown as part of the

Lesson No. 1 Intro. to Operating System 131

segment map that systems programs prepare to facilitate loading of segments in
memory by the OS. When segment numbers & relative offsets within the
segments are defined, two-component virtual addresses uniquely identify all
items within a process's address space. For example, if the SHARED segment is
assigned number 3, the subroutine SSUB2 may be uniquely identified by its
virtual address (3100), where 100 is the offset within the segment number 3-
SHARED.
Address Translation
Since physical memory in segmented systems generally retains its linear-array
organization, some address translation mechanism is needed to convert a two-
dimensional virtual-segment address into its one-dimensional physical
equivalent. In segmented systems, items belonging to a single segment reside in
one contiguous area of physical memory. With each segment compiled as if
starting from the virtual address zero, segments are generally individually
relocatable. As a result, different segments of the same process need not occupy
contiguous areas of physical memory.

limit base

Segment
table

CPU s d

yes
< + Physical
memory

no

Trap: addressing error

Hardware support for Segmentation

Lesson No. 1 Intro. to Operating System 132

 Figure 11 – Address translation in segmented systems
When requested to load a segmented process, the OS attempts to allocate
memory for the supplied segments. Using logic similar to that used for dynamic
partitioning, it may create a separate partition to suit the needs of each particular
segment. The base (obtained during partition creation) & size (specified in the
load module) of a loaded segment are recorded as a tuple called the segment
descriptor. All segment descriptors of a given process are collected in a table
called the segment descriptor table (SDT). An important component of address
mapping in segmented system is segment descriptor table. Two dimensional
user defined address is mapped to one dimensional physical address by
segment descriptor table. Each entry of this table has segment base & segment
limit. Segment base contains the starting physical address of the segment &
segment limit specifies the length of the segment.
Figure 11 illustrates a sample placement of the segments defined in Figure 10

into physical memory, & the resulting SDT formed by the OS. With the physical

base address of each segment defined, the process of translation of a virtual,

two-component address into its physical equivalent basically follows the

mechanics of based addressing. The segment number provided in the virtual

address is used to index the segment descriptor table & to obtain the physical

base address of the related segment. Adding the offset of the desired item to the

base of its enclosing segment then produces the physical address. This process

is illustrated in Figure 11 for the example of the virtual address (3100). To access

Segment 3, the number 3 is used to index the SDT & to obtain the physical base

address, 20000, of the segment SHARED. The size field of the same segment

Lesson No. 1 Intro. to Operating System 133

descriptor is used to check whether the supplied offset is within the legal bounds

of its enclosing segment. If so, the base & offset are added to produce the target

physical address. In our example that value is 20100, the first instruction word of

the shared subroutine SSUB2.

In general, the size of a segment descriptor table is related to the size of the
virtual address space of a process. For example, Intel's iAPX 286 processor is
capable of supporting up to 16K segments of 64 KB each per process, thus
requiring 16K entries per SDT. Given their potential size, segment descriptor
tables are not kept in registers. Being a collection of logically related items, the
SDTs themselves are often treated as special types of segments. Their
accessing is usually facilitated by means of a dedicated hardware register called
the segment descriptor table base register (SDTBR), which is set to point to the
base of the running process's SDT. Since the size of an SDT may vary from a
few entries to several thousand, another dedicated hardware register, called the
segment descriptor table limit register (SDTLR), is provided to mark the end of
the SDT pointed to by the SDTBR. In this way, an SDT need contain only as
many entries as there are segments actually defined in a given process.
Attempts to access nonexistent segments may be detected & dealt with as
nonexistent-segment exceptions.
From the OS's point of view, segmentation is essentially a multiple-base-limit
version of dynamically partitioned memory. Memory is allocated in the form of
variable partitions; the main difference is that one such partition is allocated to
each individual segment. Bases & limits of segments belonging to a given
process are collected into an SDT are normally kept in the PCB of the owner
process. Upon each process switch, the SDTBR & SDTLR are loaded with the
base & size, respectively, of the SDT of the new running process. In addition to
the process-loading time, SDT entries may also need to be updated whenever a
process is swapped out or relocated for compaction purposes. Swapping out
requires invalidation of all SDT entries that describe the affected segments.
When the process is swapped back in, the base fields of its segment descriptors

Lesson No. 1 Intro. to Operating System 134

must be updated to reflect new load addresses. For this reason, swapping out of
the SDT itself is rarely useful. Instead, the SDT of the swapped-out process may
be discarded, & the static segment map-such as the one depicted in Figure
10(b)-may be used for creation of an up-to-date SDT whenever the related
process is loaded in memory. Compaction, when supported, requires updating of
the related SDT entry for each segment moved. In such systems, some
additional or revised data structures may be needed to facilitate identification of
the SDT entry that describes the segment scheduled to be moved.
The price paid for segmenting the address space of a process is the overhead of
storing & accessing segment descriptor tables. Mapping each virtual address
requires two physical memory references for a single virtual (program) reference,
as follows:
• Memory reference to access the segment descriptor in the SDT
• Memory reference to access the target item in physical memory
In other words, segmentation may cut the effective memory bandwidth in half by
making the effective virtual-access time twice as long as the physical memory
access time.
Segment Descriptor Caching
With performance of segmented systems so critically dependent on the duration
of the address translation process, computer system designers often provide
some hardware accelerators to speed the translation. Memory references
expended on mapping may be avoided by keeping segment descriptors in
registers. However, the potential size of an SDT & the overhead of process
switching make it too costly to keep an entire SDT of the running process in
registers. A reasonable compromise is to keep a few of the most frequently used
segment descriptors in registers. In this way, most of the memory references
may be mapped with the aid of registers. The rest may be mapped using the
SDT in memory, as usual. This scheme is dependent on the OS's ability to select
the proper segment descriptors for storing into registers. In order to provide the
intuitive motivation for one possible implementation of systematic descriptor

Lesson No. 1 Intro. to Operating System 135

selection., let us investigate the types of segments referenced by the executing
process.
Memory references may be functionally categorized as accesses to (i)
Instructions, (ii) Data, & (iii) Stack
A typical instruction execution sequence consists of a mixture of the outline types
of memory references. In fact, completion of a single stack manipulation
instruction, such as a push of a datum from memory onto stack, may require all
three types of references. Thus the working space of a process normally
encompasses one each of code, data, & stack segments. Therefore, keeping the
current code, data, & stack segment descriptors in registers may accelerate
address translation. Depending on its type, a particular memory reference may
then be mapped using the appropriate register. But can we know the exact type
of each memory reference as the processor is making it? The answer is yes, with
the proper hardware support. Namely, in most segmented machines the CPU
emits a few status bits to indicate the type of each memory reference. The
memory management hardware uses this information to select the appropriate
mapping register.

Lesson No. 1 Intro. to Operating System 136

 Figure 12 – Segment-descriptor cache registers

Register-assisted translation of virtual to physical addresses is illustrated in

Figure 12. As indicated, the CPU status lines are used to select the appropriate
segment descriptor register (SDR). The size field of the selected segment
descriptor is used to check whether the intended reference is within the bounds
of the target segment. If so, the base field is added with the offset to produce the
physical address. By making the choice of the appropriate segment register
implicit in the type of memory reference being made, segment typing may
eliminate the need to keep track of segment numbers during address
translations. Though segment typing is certainly useful, it may become restrictive
at times. For example, copying an instruction sequence from one segment into
another may confuse the selector logic into believing that source & target
segments should be of type data rather than code. Using the so-called segment

Lesson No. 1 Intro. to Operating System 137

override of type prefixes, which allows the programmer to explicitly indicate the
particular segment descriptor register to be used for mapping the memory
reference in question, may alleviate this problem.
Segment descriptor registers are initially loaded from the SDT. Whenever the
running process makes an intersegment reference, the corresponding segment
descriptor is loaded into the appropriate register from the SDT. For example, an
intersegment JUMP or CALL causes the segment descriptor of the target (code)
segment to be copied from the SDT to the code segment descriptor register.
When segment typing is used as described, segment descriptor register. When
segment typing is used as described, segment descriptor caching becomes
deterministic as opposed to probabilistic. Segment descriptors stored in the three
segment descriptor registers; define the current working set of the executing
process. Since membership in the working set of segments of a process changes
with time, segment descriptor registers are normally included in the process
state. Upon each process switch, the contents of the SDRs of the departing
process are stored with the rest of its context. Before dispatching the new
running process, the OS loads segment descriptor registers with their images
recorded in the related PCB.
4.2.4.2 Protection
The base-limit form of protection is obviously the most natural choice for
segmented systems. The legal address space of a process is the collection of
segments defined by its SDT. Except for shared segments. Placing different
segments in disjoint areas of memory enforces separation of distinct address
space. Thus most of the discussion of protection in systems with dynamic
allocation of memory is applicable to segmented environments as well.
An interesting possibility in segmented systems is to provide protection within the
address space of a single process, in addition to the more usually protection
between different processes. Given that the type of each segment is defined
commensurate with the nature of information stored in its constituent elements,
access rights to each segment can be defined accordingly. For instance, though
both reading & writing of stack segments may be necessary, accessing of code

Lesson No. 1 Intro. to Operating System 138

segments can be permitted in execute-only or perhaps in the read-only mode.
Data segments can be read-only, write-only, or read-write. Thus, segmented
systems may be able to prohibit some meaningless operations, such as
execution of data or modifications of code. Additional examples include
prevention of stack growth into the adjacent code or data areas, & other errors
resulting from mismatching of segment types & intended references to them. An
important observation is that access rights to different portions of a single
address space may vary in accordance with the type of information stored
therein. Due to the grouping of logically related items, segmentation is one of the
rare memory-management schemes that allow such finely grained delineation of
access rights. The mechanism for enforcement of declared access rights in
segmented systems is usually coupled with the address translation hardware.
Typically, access-rights bits are included in segment descriptors. In the course of
address mapping, the intended type of reference is checked against the access
rights for the segment in question. Any mismatch results in abortion of the
memory reference in progress, & a trap to the OS.
4.2.4.3 Sharing
Shared objects are usually placed in separate, dedicated segments. A shared
segment may be mapped, via the appropriate segment descriptor tables, to the
virtual-address spaces of all processes that are authorized to reference it. The
deliberate use of offsets & of bases addressing facilitate sharing since the virtual
offset of a given item is identical in all processes that share it. The virtual number
of a shared segment, on the other hand, need not be identical in all address
spaces of which it is a member. These points are illustrated in Figure 13, where a
code segment EMACS is assumed to be shared by three processes. The
relevant portions of the segment descriptor tables of the participating processes,
SDT1, SDT2, & SDT3, are shown. As indicated, the segment EMACS is
assumed to have different virtual numbers in the three address spaces of which it
is part. The placement of access-rights bits in segment descriptor tables is also
shown. Figure 13 illustrates the fact that different processes can have different
access rights to the same shared segment. For example, whereas processes P1

Lesson No. 1 Intro. to Operating System 139

 & P2 can execute only the shared segment EMACS, process P3 is allowed both
reading & writing.
Figure 13 also illustrates the ability of segmented systems to conserve memory
by sharing the code of programs executed by many users. In particular, each
participating process can execute the shared code from EMACS using its own
private data segment. Assuming that EXACS is an editor, this means that a

(DATA1)
RW

EO
EMCS

SD1
BASE SIZE ACCESS RIGHT

DATA2
RW
EO

RW
RW DATA3

SD3
SD2
MEMORY

Figure 13 – Sharing in segmented systems

single copy of it may serve the entire user population of a time-sharing system.
Naturally, execution of EMACS on behalf of each user is stored in a private data
segment of its corresponding process. For example, users 1, 2, & 3 can have
their respective texts buffers stored in data segments DATA1, DATA2, & DATA3.
Depending on which of the three processes is active at a given time, the
hardware data segment descriptor register points to data segment DATA1,
DATA2, or DATA3, & the code segment descriptor register points to EMACS in
all cases. Of course, the current instruction to be executed by the particular
process is indicated by the program counter, which is saved & restored as a part

Lesson No. 1 Intro. to Operating System 140

of each process's state. In segmented systems, the program counter register
usually contains offsets of instructions within the current code segment. This
facilitates sharing by making all code self-references relative to the beginning of
the current code segment. When coupled with segment typing, this feature
makes it possible to assign different virtual segment numbers to the same
(physical) shared segment in virtual-address spaces of different processes of
which it is a part. Alternatively, the problem of making direct self-references in
shared routines described in Section 3.1.3.4 of this lesson (sharing) restricts the
type of code that may safely be shared.
As described, sharing is encouraged in segmented systems. This presents some

problems in systems that also support swapping, which is normally done to

increase processor utilization. For example, a shared segment may need to

maintain its memory residence while being actively used by any of the processes

authorized to reference it. Swapping in this case opens up the possibility that a

participating process may be swapped out while its shared segment remains

resident. When such a process is swapped back in, the construction of its SDT

must take into consideration the fact that the shared segment may already be

resident. In other words, the OS must keep track of shared segments & of

processes that access them. When a participating process is loaded in memory,

the OS is expected to identify the location of the shared segment in memory, if

any, & to ensure its proper mapping from all virtual address spaces of which it is

a part.

5.3 Keywords
Contiguous Memory Management: In this approach, each program occupies a
single contiguous block of storage locations.
First-fit: This allocates the first available space that is big enough to
accommodate process.

Lesson No. 1 Intro. to Operating System 141

Best-fit: This allocates the smallest hole that is big enough to accommodate
process.
Worst fit: This strategy allocates the largest hole.
External Fragmentation is waste of memory between partitions caused by
scattered non-contiguous free space.
Internal fragmentation is waste of memory within a partition caused by
difference between size of partition & the process allocated.
Compaction is to shuffle memory contents & place all free memory together in
one block.
Program relocatability refers to the ability to load & execute a given program
into an arbitrary place in memory.
4.4 SUMMARY
In this lesson, we have presented several schemes for management of main
memory that are characterized by contiguous allocation of memory. Except for
the single partition, which is inefficient in terms of both CPU & memory utilization,
all schemes support multiprogramming by allowing address spaces of several
processes to reside in main memory simultaneously. One approach is to
statically divide the available physical memory into a number of fixed partitions &
to satisfy requests for memory by granting suitable free partitions, if any. Fixed
partition sizes limit the maximum allowable virtual-address space of any given
process to the size of the largest partition (unless overlays are used). The total
number of partitions in a given system limits the number of resident processes.
Within the confines of this limit, the effectiveness of the short-term scheduler may
be improved by employing swapping to increase the ratio of resident to ready
processes. Systems with static partitioning suffer from internal fragmentation of
memory.
Variable (dynamic) partitioning allows allocation of the entire physical memory,
except for the resident part of the OS, to a single process. Thus, in systems with
dynamic partitioning, the virtual-address space of any given process or an
overlay is limited only by the capacity of the physical memory in a given system.
Dynamic creation of partitions according to the specific needs of requesting

Lesson No. 1 Intro. to Operating System 142

processes also eliminates the problem of internal fragmentation. Dynamic
allocation of partitions requires the use of more complex algorithms for de-
allocation of partitions & coalescing of free memory in order to combat external
fragmentation. The need for occasional compaction of memory is also a major
contributor to the increased time & space complexity of dynamic partitioning.
Both fixed & variable partitioning of memory rely on virtually identical hardware
support for relocation & protection. Sharing is quite restrictive in both systems.
Segmentation allows breaking of the virtual address space of a single process
into separate entities (segments) that may be placed in noncontiguous areas of
physical memory. As a result, the virtual-to-physical address translation at
instruction execution time in such systems is more complex, & some dedicated
hardware support is necessary to avoid a drastic reduction in effective memory
bandwidth. Since average segment sizes are usually smaller then average
process sizes, segmentation can reduce the impact of external fragmentation on
the performance of systems with dynamically partitioned memory. Other
advantages of segmentation include dynamic relocation, finely grained protection
both within & between address spaces, ease of sharing, & facilitation of dynamic
linking & loading. Unless virtual segmentation is supported, segmentation does
not remove the problem of limiting the size of a process's virtual space by the
size of the available physical memory.
4.5 SELF ASSESSMENT QUESTIONS (SAQ)
1. What functions does a memory manager perform?
2. How is user address space loaded in one partition of memory protected
from others?
3. What is the problem of fragmentation? How is it dealt with?
4. What do you understand by program relocatability? Differentiate between
static and dynamic relocation.
5. Differentiate between first fit, best fit, and worst fit memory allocation
strategies. Discuss their merits and demerits.
6. How is the tracks of status of memory is kept in partitioned memory
management?

Lesson No. 1 Intro. to Operating System 143

7. What do you mean by relocation of address space? What problems does
it cause?
8. Differentiate between internal fragmentation and external fragmentation.
9. What is external fragmentation? What is compaction? What are the merits
and demerits of compaction?
10. What do you understand by segmentation? Discuss the address
translation in segmented systems.
11. What are three basic approaches to sharing in systems with fixed
partitioning of memory?
4.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin
P.B., John Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New
Delhi.
3. Operating Systems, Godbole A.S., Tata McGraw Hill Publishing
Company Ltd., New Delhi.
4. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
5. Operating Systems-A Modern Perspective, Gary Nutt, Pearson
Education Asia, 2000.
6. Operating Systems, Harris J.A., Tata McGraw Hill Publishing
Company Ltd., New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 144

Lesson number: 6 Writer: Dr. Rakesh Kumar
Memory Management - II Vetter: Prof. Dharminder Kr.

6.0 OBJECTIVE
The objective of this lesson is to make the students familiar with the following
concepts of (1) Non-contiguous memory management, (2) Paging, and (3) Virtual
memory.
6.1 INTRODUCTION
In noncontiguous memory management memory is allocated in such a way that
parts of single logical object may be placed in noncontiguous areas of physical
memory, whereas the address space of a process always remains contiguous. At
run time contiguous virtual address space is mapped to noncontiguous physical
address space. This type of memory management is done in various ways:
4. Non-Contiguous, real memory management system
¾ Paged memory management system
¾ Segmented memory management system
¾ Combined memory management system
5. Non-Contiguous, virtual memory management system
¾ Virtual memory management system
6.2 Presentation of contents
6.2.1 Paging
6.2.1.1 Principles of Operation
6.2.1.2 Page Allocation
6.2.1.3 Hardware Support for Paging
6.2.1.4 Protection & Sharing
6.2.2 Virtual Memory
6.2.2.1 Principles of Operation
6.2.2.2 Management of Virtual Memory
6.2.2.4 Program Behavior
6.2.2.5 Replacement Policies

Lesson No. 1 Intro. to Operating System 145

 6.2.2.5.1 Replacement Algorithms
6.2.2.5.2 Global & Local Replacement Policies
6.2.2.6 Allocation Policies
6.2.2.7 Hardware Support & Considerations
6.2.2.8 Protection & Sharing
6.2.3 Segmentation & Paging
6.2.1 PAGING
Paging is another solution to the problem of memory fragmentation. It removes
the requirement of contiguous allocation of physical memory. In it, the physical
memory is conceptually divided into a number of fixed-size slots, called page
frames. The virtual-address space of a process is also split into fixed-size blocks
of the same size, called pages. Memory management module identifies sufficient
number of unused page frames for loading of the requesting process's pages. An
address translation mechanism is used to map virtual pages to their physical
counterparts. Since each page is mapped separately, different page frames
allocated to a single process need not occupy contiguous areas of physical
memory.
6.2.1.1 Principles of Operation
Figure 1 demonstrates the basic principle of paging. It illustrates a sample 16 MB
system where virtual & physical addresses are assumed to be 24 bits long each.
The page size is assumed to be 4096 bytes. Thus, the physical memory can
accommodate 4096 page frames of 4096 bytes each. After reserving 1 MB of
physical memory for the resident portion of the OS, the remaining 3840 page
frames are available for allocation to user processes. The addresses are given in
hexadecimal notation. Each page is 1000H bytes long, & the first user-allocatable
page frame starts at the physical address 100000H.

Lesson No. 1 Intro. to Operating System 146

 Virtual Virtual Physical Physical
Address Page Page Page Frame Address Frame
0000 0 No. 000000
01000 1 0 FFD .
02000 LDA 003 200 2 1 100 .
03000 3 2 103 .
3 FFF .
.
PMT
100000 (P1) 100
101000 101
102000 102
103000 (P2) 103
.
:
200 :
:
FFF
:
:
:
FFF200
:
:
FFC000 FFC
FFD000 (P0) FFD
C FFE000 FFE
FFF000 (P3) FFF
Memory

Figure 1 - Paging

The virtual-address space of a sample user process that is 14,848 bytes (3A00H)
long is divided into four virtual pages numbered from 0 to 3. A possible
placement of those pages into physical memory is depicted in Figure 1. The
mapping of virtual addresses to physical addresses in paging systems is
performed at the page level. Each virtual address is divided into two parts: the
page number & the offset within that page. Since pages & page frames have
identical sizes, offsets within each are identical & need not be mapped. So each
24-bit virtual address consist of a 12-bit page number (high-order bits) & a 12-bit
offset within the page.
Address translation is performed with the help of the page-map table (PMT),
constructed at process-loading time. As indicated in figure 1, there is one PMT
entry for each virtual page of a process. The value of each entry is the number of
the page frame in the physical memory where the corresponding virtual page is
placed. Since offsets are not mapped, only the page frame number need be
stored in a PMT entry. E.g., virtual page 0 is assumed to be placed in the

Lesson No. 1 Intro. to Operating System 147

physical page frame whose starting address is FFD000H (16,764,928 decimal).
With each frame being 1000H bytes long, the corresponding page frame number
is FFDH, as indicated on the right-hand side of the physical memory layout in
Figure 1. This value is stored in the first entry of the PMT. All other PMT entries
are filled with page frame numbers of the region where the corresponding pages
are actually loaded.
The logic of the address translation process in paged systems is illustrated in
Figure 1 on the example of the virtual address 03200H. The virtual address is
split by hardware into the page number 003H, & the offset within that page
(200H). The page number is used to index the PMT & to obtain the
corresponding physical frame number, i.e. FFF. This value is then concatenated
with the offset to produce the physical address, FFF200H, which is used to
reference the target item in memory.
000 ALLOCATED
:
:
:
0FF ALLOCATED
100 ALLOCATED
101 FREE
102 FREE
103 ALLOCATED
:
:
:
FFC FREE
FFD ALLOCATED
FFE FREE
FFF ALLOCATED
Figure 2 – Memory-map table (MMT)
The OS keeps track of the status of each page frame by using a memory-map table (MMT). Format of an MMT is
illustrated in Figure 2, assuming that only the process depicted in Figure 1 & the OS are resident in memory. Each entry of
the MMT described the status of page frame as FREE or ALLOCATED. The number of MMT entries i.e. f is computed as f
= m/p where m is the size of the physical memory, & p is page size. Both m & p are usually an integer power of base 2,
thus resulting in f being an integer. When requested to load a process of size s, the OS must allocate n free page frames,
so that n = Round(s/p) where p is the page size. The OS allocates memory in terms of an integral number of page
frames. If the size of a given process is not a multiple of the page size, the last page frame may be partly unused resulting
into page fragmentation.

After selecting n free page frames, the OS loads process pages into them &
constructs the page-map table of the process. Thus, there is one MMT per
system, & as many PMTs as there are active processes. When a process

Lesson No. 1 Intro. to Operating System 148

terminates or becomes swapped out, memory is deallocated by releasing the
frame holdings of the departing process to the pool of free page frames.
6.2.1.2 Page Allocation
The efficiency of the memory allocation algorithm depends on the speed with
which it can locate free page frames. To facilitate this, a list of free pages is
maintained instead of the static-table format of the memory map assumed earlier.
In that case, n free frames may be identified & allocated by unlinking the first n
nodes of the free list. Deallocation of memory in systems without the free list
consists of marking in the MMT as FREE all frames found in the PMT of the
departing process a time consuming operation. Frames identified in the PMT of
the departing process can be linked to the beginning of the freed list. Linking at
the beginning is the fastest way of adding entries to an unordered singly linked
list. Since the time complexity of deallocation is not significantly affected by the
choice of data structure of free pages, the free-list approach has a performance
advantage as its time complexity of deallocation is not significantly affected by
the choice of data structure of free pages, & is not affected by the variation of
memory utilization.
6.2.1.3 Hardware Support for Paging
Hardware support for paging, concentrates on saving the memory necessary for
storing of the mapping tables, & on speeding up the mapping of virtual to
physical addresses. In principle, each PMT must be large enough to
accommodate the maximum size allowed for the address space of a process in a
given system. In theory, this may be the entire physical memory. So in a 16 MB
system with 256-byte pages, the size of a PMT should be 64k entries. Individual
PMT entries are page numbers that are 16 bits long in the sample system, thus
requiring 128 KB of physical memory to store a PMT. With one PMT needed for
each active process, the total PMT storage can consume a significant portion of
physical memory.
Since the actual address space of a process may be well below its allowable
maximum, it is reasonable to construct each PMT with only as many entries as
its related process has pages. This may be accomplished by means of a

Lesson No. 1 Intro. to Operating System 149

dedicated hardware page-map table limit register (PMTLR). A PMTLR is set to
the highest virtual page number defined in the PMT of the running process.
Accessing of the PMT of the running process may be facilitated by means of the
page-map table base register (PMTBR), which points to the base address of the
PMT of the running process. The respective values of these two registers for
each process are defined at process-loading time & stored in the related PCB.
Upon each process switch, the PCB of the new running process provides the
values to be loaded in the PMTBR & PMTLR registers.
Even with the assistance of these registers, address translations in paging
systems still require two memory references; one to access the PMT for
mapping, & the other to reference the target item in physical memory. To speed it
up, a high-speed associative memory for storing a subset of often-used page-
map table entries is used. This memory is called the translation look aside buffer
(TLB), or mapping cache.
Associative memories can be searched by contents rather than by address. So,
the main-memory reference for mapping can be substituted by a TLB reference.
Given that the TLB cycle time is very small, the memory-access overhead
incurred by mapping can be significantly reduced. The role of the cache in the
mapping process is depicted in Figure 3.

Lesson No. 1 Intro. to Operating System 150

Virtual address
.
Page Offset
.
.
C 40000 (i)
Yes 400 I
Page .
600 M
In .
cache 900 K
TLB .
60000 (m)
No
.
.
.
90000 (k)
.
.
I 400 C .
PMTBR
.
.
Yes .
< M 900
PMT +

K 600
No
PMT

Non-existent page exception

Figure 3 – Translation-lookaside buffer (TLB)

As indicated, the TLB entries contain pairs of virtual page numbers & the
corresponding page frame numbers where the related pages are stored in
physical memory. The page number is necessary to define each particular entry,
because a TLB contains only a subset of page-map table entries. Address
translation begins by presenting the page-number portion of the virtual address
to the TLB. If the desired entry is found in the TLB, the corresponding page frame
number is combined with the offset to produce the physical address.
Alternatively, if the target entry is not in TLB, the PMT in memory must be
accessed to complete the mapping. This process begins by consulting the
PMTLR to verify that the page number provided in the virtual address is within
the bounds of the related process's address space. If so, the page number is
added to the contents of the PMTBR to obtain the address of the corresponding

Lesson No. 1 Intro. to Operating System 151

PMT entry where the physical page frame number is stored. This value is then
concatenated with the offset portion of the virtual address to produce the physical
memory address of the desired item.
Figure 3 demonstrates that the overhead of TLB search is added to all mappings,
regardless of whether they are eventually completed using the TLB or the PMT in
main memory. In order for the TLB to be effective, it must satisfy a large portion
of all address mappings. Given the generally small size of a TLB because of the
high price of associative memories, only the PMT entries most likely to be
needed ought to reside in the TLB.
The effective memory-access time, teff in systems with run-time address
translation is the sum of the address translation time, tTR & the subsequent
access time needed to fetch the target item from memory, tM.
teff = tTR + tM
With TLB used to assist in address translation, tTR becomes
TTR = h tTLB + (1 - h) (tTLB + tM) = tTLB + (1 - h)tM
where h is the TLB hit ratio, that is, the ratio of address translations that are
contained the TLB over all translations, & thus 0≤h≤1;tTLB is the TLB access time;
& tM is the main-memory access time. Therefore, effective memory-access time
in systems with a TLB is
teff = tTLB + (2 - h)tM
It is observed that the hardware used to accelerate address translations in
paging systems (TLB) is managed by means of probabilistic algorithms, as
opposed to the deterministic mapping-register typing described in relation to
segmentation. The reason is that the mechanical splitting of a process’s address
space into fixed-size chunks produces pages. As a result, a page, unlike a
segment, in general does not bear any relationship to the logical entities of the
underlying program. For example, a single page may contain a mixture of data,
stack, & code. This makes typing & other forms of deterministic loading of TLB
entries extremely difficult, in view of the stringent timing restrictions imposed on
TLB manipulation.
6.2.1.4 Protection & Sharing

Lesson No. 1 Intro. to Operating System 152

Unless specifically declared as shared, distinct address spaces are placed in
disjoint areas of physical memory. Memory references of the running process are
restricted to its own address space by means of the address translation
mechanism, which uses the dedicated PMT. The PMTLR is used to detect & to
abort attempts to access any memory beyond the legal boundaries of a process.
Modifications of the PMTBR & PMTLR registers are usually possible only by
means of privileged instructions, which trap to the OS if attempted in user mode.
By adding the access bits to the PMT entries & appropriate hardware for testing
these bits, access to a given page may be allowed only in certain programmer-
defined modes such as read-only, execute-only, or other restricted forms of
access. This feature is much less flexible in paging systems than segmentation.
The primary difference is that paging is supposed to be entirely transparent to
programmers. Mechanical splitting of an address space into pages is performed
without any regard for the possible logical relationships between the items under
consideration. Since there is no notion of typing, code & data may be mixed
within one page. As we shall see, specification of the access rights in paging
systems is useful for pages shared by several processes, but it is of much less
value inside the boundaries of a given address space.
Protection in paging systems may also be accomplished by means of the
protection keys. In principle, the page size should correspond to the size of the
memory block protected by the single key. This allows pages belonging to a
single process to be scattered throughout memory-a perfect match for paged
allocation. By associating access-rights bits with protection keys, access to a
given page may be restricted when necessary.
Sharing of pages is quite straightforward with paged memory management. A
single physical copy of a shared page can be easily mapped into as many
distinct address spaces as desired. Since each such mapping is performed via a
dedicated entry in the PMT of the related process, different processes may have
different access rights to the shared page. Given that paging is transparent to
users, sharing at the page level must be recognized & supported by systems

Lesson No. 1 Intro. to Operating System 153

programs. Systems programs must ensure that virtual offsets of each item within
a shared page are identical in all participating address spaces.
Like data, shared code must have the same within-page offsets in all address
spaces of which it is a part. As usual, shared code that is not executed in
mutually exclusive fashion must be reentrant. In addition, unless the shared code
is position-independent, it must have the same virtual page numbers in all
processes that invoke it. This property must be preserved even in cases when
the shared code spans several pages.
6.2.2 VIRTUAL MEMORY
Virtual memory allows execution of partially loaded processes. As a
consequence, virtual address spaces of active processes in a virtual-memory
system can exceed the capacity of the physical memory. This is accomplished by
maintaining an image of the entire virtual-address space of a process on
secondary storage, & by bringing its sections into main memory when needed.
The OS decides which sections to bring in, when to bring them in, & where to
place them. Thus, virtual-memory systems provide for automatic migration of
portions of address spaces between secondary & primary storage. Virtual
memory provides the illusion of a much larger memory than may actually be
available, so programmers are relieved of the burden of trying to fit a program
into limited memory.
Due to the ability to execute a partially loaded process, a process may be loaded
into a space of arbitrary size resulting into the reduction of external
fragmentation. Moreover, the amount of space in use by a given process may be
varied during its memory residence. As a result, the OS may speed up the
execution of important processes by allocating them more real memory.
Alternatively, by reducing the real-memory holdings of resident processes, the
degree of multi-programming can be increased by using the vacated space to
activate more processes.
The speed of program execution in virtual-memory systems is bounded from
above by the execution speed of the same program run in a non-virtual memory

Lesson No. 1 Intro. to Operating System 154

management system. That is due to delays caused by fetching of missing
portions of program's address space at run-time.
Virtual memory provides execution of partially loaded programs. But an
instruction can be completed only if all code, data, & stack locations that it
references reside in physical memory. When we reference an out-of-memory
item, the running process must be suspended to fetch the target item from disk.
So what is the performance penalty?
An analysis of program behavior provides an answer to the question. Most
programs consist of alternate execution paths, some of which do not span the
entire address space. On any given run, external & internal program conditions
cause only one specific execution path to be followed. Dynamic linking & loading
exploits this aspect of program behavior by loading into memory only those
procedures that are actually referenced on a particular run. Moreover, many
programs tend to favor specific portions of their address spaces during
execution. So it is reasonable to keep in memory only those routines that make
up the code of the current pass. When another pass over the source code
commences, the memory manager can bring new routines into the main memory
& return those of the previous pass back to disk.
6.2.2.1 Principles of Operation
Virtual memory can be implemented as an extension of paged or segmented
memory management or as a combination of both. Accordingly, address
translation is performed by means of PMT, segment descriptor tables, or both.
The important characteristic is that in virtual-memory systems some portions of
the address space of the running process can be absent from main memory.
The process of address mapping in virtual-memory systems is more formally
defined as follows. Let the virtual-address space be V = {0, 1, … , v-1}, & the
physical memory space by M = {0, 1, ….m-1}. The OS dynamically allocates real
memory to portions of the virtual-address space. The address translation
mechanism must be able to associate virtual names with physical locations. In
other words, at any time the mapping hardware must realize the function f: V →
M such that

Lesson No. 1 Intro. to Operating System 155

 r if item x is in real memory at location r
f(x) =
missing-item exception if item x is not in real memory
Thus, the additional task of address translation hardware in virtual systems is to
detect whether the target item is in real memory or not. If the referenced item is
in memory, the process of address translation is completed.
We present the operation of virtual memory assuming that paging is the basic
underlying memory-management scheme. The detection of missing items is
rather straightforward. It is usually handled by adding the presence indicator, a
bit, to each entry of page-map tables. The presence bit, when set, indicates that
the corresponding page is in memory; otherwise the corresponding virtual page
is not in real memory. Before loading the process, the OS clears all the presence
bits in the related page-map table. As & when specific pages are brought into the
main memory, its presence bit is reset.
A possible implementation is illustrated in Figure 4. The presented process's
virtual space is assumed to consisting of only six pages. As indicated, the
complete process image is present in secondary memory. The PMT contains an
entry for each virtual page of the related process. For each page actually present
in real memory, the presence bit is set (IN), & the PMT points to the physical
frame that contains the corresponding page. Alternatively, the presence bit is
cleared (OUT), & the PMT entry is invalid.
The address translation hardware checks the presence bit during the mapping of
each memory reference if the bit is set, the mapping is completed as usual.
However, if the corresponding presence bit in the PMT is reset, the hardware
generates a missing-item exception. In paged virtual-memory systems, this
exception is called a page fault. When the running process experiences a page
fault, it must be suspended until the missing page is brought into main memory.
The disk address of the faulted page is usually provided in the file-map table
(FMT). This table is parallel to the PMT. Thus, when processing a page fault, the
OS uses the virtual page number provided by the mapping hardware to index the

Lesson No. 1 Intro. to Operating System 156

FMT & to obtain the related disk address. A possible format & use of the FMT is
depicted in Figure 4.

Figure 4 – Virtual memory

6.2.2.2 Management of Virtual Memory
The implementation of virtual memory requires maintenance of one PMT per
active process. Given that the virtual-address space of a process may exceed
the capacity of real memory, the size of an individual PMT can be much larger in
a virtual than in a real paging system with identical page sizes. The OS maintains
one MMT or a free-frame list to keep track of Free/allocated page frames.
A new component of the memory manager's data structures is the FMT. FMT
contains secondary-storage addresses of all pages. The memory manager used
the FMT to load the missing items into the main memory. One FMT is maintained
for each active process. Its base may be kept in the control block of the related

Lesson No. 1 Intro. to Operating System 157

process. An FMT has a number of entries identical to that of the related PMT. A
pair of page-map table base & page-map length registers may be provided in
hardware to expedite the address translation process & to reduce the size of
PMT for smaller processes. As with paging, the existence of a TLB is highly
desirable to reduce the negative effects of mapping on the effective memory
bandwidth.
The allocation of only a subset of real page frames to the virtual-address space
of a process requires the incorporation of certain policies into the virtual-memory
manager. We may classify these policies as follows:
1. Allocation policy: How much real memory to allocate to each active process
2. Fetch policy: Which items to bring & when to bring them from secondary
storage into the main memory
3. Replacement policy: When a new item is to be brought in & there is no free
real memory, which item to evict in order to make room.
4. Placement policy: Where to place an incoming item
6.2.2.4 Program Behavior
Program behavior is of extreme importance for the performance of virtual-
memory systems. Execution of partially loaded programs generally leads to
longer turnaround times due to the processing of page faults. By minimizing the
number of page faults, the effective processor utilization, effective disk I/O
bandwidth, & program turnaround times may be improved.
It is observed that there is a strong tendency of programs to favor subsets of their
address spaces during execution. This phenomenon is known as locality of
reference. Both temporal & spatial locality of reference has been observed.
Spatial locality is the tendency for a program to reference clustered locations in
preference to randomly distributed locations. Temporal locality is the tendency for
a program to reference the same location or a cluster several times during brief
intervals of time. Temporal locality of reference is exhibited by program loops.
Spatial locality suggests that once an item is referenced, there is a high
probability that it or its neighboring items are going to be referenced in the near
future. A locality is a small cluster of not necessarily adjacent pages to which

Lesson No. 1 Intro. to Operating System 158

most memory references are made during a period of time. Both temporal &
spatial locality of reference is dynamic properties in the sense that the identity of
the particular pages that compose the actively used set varies with time. As
observed, the executing program moves from one locality to another in the
course of its execution. Statistically speaking, the probability that a particular
memory reference is going to be made to a specific page is a time-varying
function. It increases when pages in its current locality are being referenced, & it
decreases otherwise. The evidence also suggests that the executing program
moves slowly from one locality to another.
Locality of reference basically suggests that a significant portion of memory
references of the running process may be made to a subset of its pages. These
findings may be utilized for implementation of replacement & allocation policies.
6.2.2.5 Replacement Policies
If a page fault is there, then it is to be brought into the main memory
necessitating creation of a room for it. There are two options for this situation:
• The faulted process may be suspended until availability of memory.
• A page may be removed to make room for the incoming one.
Suspending a process is not an acceptable solution. Thus, removal is commonly
used to free the memory needed to load the missing items. A replacement policy
decides the victim page for eviction. In virtual memory systems all pages are kept
on the secondary storage. As & when needed, some of those pages are copied
into the main memory. While executing, the running process may modify its data
or stack areas, thus making some resident pages different from their disk images
(dirty page). So it must be written back to disk in place of its obsolete copy. When
a page that has not been modified (clean page) during its residence in memory is
to be evicted, if can simply be discarded. Tracking of page modifications is
usually performed in hardware by adding a written-into bit called as dirty bit, to
each entry of the PMT. It indicates whether the page is dirty or clean.
6.2.2.5.1 Replacement Algorithms
First-In-First-Out (FIFO). The FIFO algorithm replaces oldest pages i.e. the
resident page that has spent the longest time in memory. To implement the FIFO

Lesson No. 1 Intro. to Operating System 159

page-replacement algorithm, the memory manager must keep track of the
relative order of the loading of pages into the main memory. One way to
accomplish this is to maintain a FIFO queue of pages.
FIFO fails to take into account the pattern of usage of a given page; FIFO tends
to throw away frequently used pages because they naturally tend to stay longer
in memory. Another problem with FIFO is that it may defy intuition by increasing
the number of page faults when more real pages are allocated to the program.
This behavior is known as Belady’s anomaly.
Least Recently Used (LRU). The least recently used algorithm replaces the
least recently used resident page. LRU algorithm performs better than FIFO
because it takes into account the patterns of program behavior by assuming that
the page used in the most distant past is least likely to be referenced in the near
future. The LRU algorithm belongs to a larger class of stack replacement
algorithms. A stack algorithm is distinguished by the property of performing
better, or at least not worse, when more real memory is made available to the
executing program. Stack algorithms therefore do not suffer from Belady's
anomaly.
The implementation of the LRU algorithm imposes too much overhead to be
handled by software alone. One possible implementation is to record the usage
of pages by means of a structure similar to the stack. Whenever a resident page
is referenced, it is removed from its current stack position & placed at the top of
the stack. When a page eviction is in order, the page at the bottom of the stack is
removed from memory.
Maintenance of the page-referencing stack requires it’s updating for each page
reference, regardless of whether it results in a page fault or not. So the overhead
of searching the stack, moving the reference page to the top, & updating the rest
of the stack accordingly must be added to all memory references. But the FIFO
queue needs to be updated only when page faults occur-overhead almost
negligible in comparison to the time required for processing of a page fault.
Optimal (OPT): The algorithm by Belady, removes the page to be reference in
the most distant future. Since it requires future knowledge, the OPT algorithm is

Lesson No. 1 Intro. to Operating System 160

not realizable. Its significance is theoretical, as it can serve as a yardstick for
comparison with other algorithms.
Approximations-Clock. One popular algorithm combines the relatively low
overhead of FIFO with tracking of the resident-page usage, which accounts for
the better performance of LRU. This algorithm is sometimes referred to as Clock,
& it is also known as not recently used (NRU).
The algorithm makes use of the referenced bit, which is associated with each
resident page. The referenced bit is set whenever the related page is reference &
cleared occasionally by software. Its setting indicates whether a given page has
been referenced in the recent past. How recent this past is depends on the
frequency of the referenced-bit resetting. The page-replacement routine makes
use of this information when selecting a victim for removal.
The algorithm is usually implemented by maintaining a circular list of the resident
pages & a pointer to the page where it left off. The algorithm works by sweeping
the page list & resetting the presence bit of the pages that it encounters. This
sweeping motion of the circular list resembles the movement of the clock hand,
hence the name clock. The clock algorithm seeks & evicts pages not recently
used in order to free page frames for allocation to demanding processes. When it
encounters a page whose reference bit is cleared, which means that the related
page has not been referenced since the last sweep, the algorithm acts as
follows: (1) If the page is modified, it is marked for clearing & scheduled for
writing to disk. (2) If the page is not modified, it is declared non-resident, & the
page frames that it occupies are feed. The algorithm continues its operation until
the required numbers of page frames are freed. The algorithm may be invoked at
regular time intervals or when the number of free page frames drops below a
specified threshold.
Other approximations & variations on this theme are possible. Some of them
track page usage more accurately by means of a reference counter that counts
the number of sweeps during which a given page is found to be un-referenced.
Another possibility is to record the states of referenced bits by shifting them
occasionally into related bit arrays. When a page is to be evicted, the victim is

Lesson No. 1 Intro. to Operating System 161

chosen by comparing counters or bit arrays in order to find the least frequently
reference page. The general idea is to devise an implementable algorithm that
bases its decisions on measured page usage & thus takes into account the
program behavior patterns.

6.2.2.5.2 Global & Local Replacement

Policies
As discussed, all replacement policies choose a victim among the resident pages
owned by the process that experiences the page fault. This is known as local
replacement. However, each of the presented algorithms may be made to
operate globally. A global replacement algorithm processes all resident pages
when selecting a victim. Local replacement tends to localize effects of the
allocation policy to each particular process. Global replacement, on the other
hand, increases the degree of coupling between replacement & allocation
strategies. A global replacement algorithm may take pages allocated to one
process by the allocation algorithm, away. Global replacement is concerned
mostly with the overall state of the system, & much less with the behavior of each
individual process. Global replacement is known to be sub-optimal.
6.2.2.6 Allocation Policies
The allocation policy must compromise among conflicting requirements such as
reduced page-fault frequency, improved turn-around time and processor
utilization, etc. Giving more real pages to a process will result in reduced page-
fault frequency & improved turnaround time. But it reduces the number of active
processes that may coexist in memory at a time resulting into the lower
processor utilization factor. On the other hand, if too few pages are allocated to a
process, its page-fault frequency & turnaround times may deteriorate.
Another problem caused by under-allocation of real pages may be encountered
in systems that opt for restarting of faulted instructions. If fewer pages are
allocated to a process than are necessary for execution of the restartable
instruction that causes the largest number of page faults in a given architecture,

Lesson No. 1 Intro. to Operating System 162

the system might fault continuously on a single instruction & fail to make any real
progress.
Consider a two-address instruction, such as Add @X, @Y, where X & Y are
virtual addresses & @ denotes indirect addressing. Assuming that the operation
code & operand addresses are encoded in one word each, this instruction needs
three words for storage. With the use of indirect addressing, eight memory
references are needed to complete execution of this instruction: three to fetch the
instruction words, two to fetch operand addresses, two to access the operands
themselves (indirect addressing), & one to store the result. In the worst case, six
different pages may have to reside in memory concurrently in order to complete
execution of this instruction: two if the instruction crosses a page boundary, two
holding indirect addresses, & two holding the target operands. A likely
implementation of this instruction calls for the instruction to be restarted after a
page fault. If so, with fewer than six pages allocated to the process that executes
it, the instruction may keep faulting forever. In general, the lower limit on the
number of pages imposed by the described problem is architecture-dependent.
In any particular implementation, the appropriate bound must be evaluated &
built into the logic of the allocation routine.
While we seem to have some guidance as to the minimal number of pages, the
reasonable maximum remains elusive. It is also unclear whether a page
maximum should be fixed for a given system or determined on an individual
basis according to some specific process attributes. Should the maximum be
defined statically or dynamically, in response to system resource utilization &
availability, & perhaps in accordance with the observable behavior of the specific
process?
From the allocation module's point of view, the important conclusion is that each
program has a certain threshold regarding the proportion of real to virtual pages,
below which the number of page faults increases very quickly. At the high end,
there seems to be a certain limit on the number of real pages, above which an
allocation of additional real memory results in little or in moderate performance

Lesson No. 1 Intro. to Operating System 163

improvement. Thus, we want to allocate memory in such a way that each active
program is between these two extremes.
Being program-specific, the upper & lower limits should probably not be fixed but
derived dynamically on the basis of the program faulting behavior measured
during its execution. When resource utilization is low, activating more processes
may increase the degree of multiprogramming. However, the memory manager
must keep track of the program behavior when doing so. A process that
experiences a large number of page faults should be either allocated more
memory or suspended otherwise. Likewise, a few pages may be taken away
from a process with a low page-fault rate without great concern. In addition, the
number of pages allocated to a process may be influenced by its priority (higher
priority may indicate that shorter turnaround time is desirable), the amount of free
memory, fairness, & the like.
Although the complexity & overhead of memory allocation should be within a
reasonable bound, the use of oversimplified allocation algorithms has the
potential of crippling the system throughput. If real memory is over-allocated to
the extent that most of the active programs are above their upper page-fault-rate
thresholds, the system may exhibit a behavior known as thrashing. With very
frequent page faults, the system spends most of its time shuttling pages between
main memory & secondary memory. Although the disk I/O channel may be
overloaded by this activity, but processor utilization is reduced.
One way of introducing thrashing behavior is dangerously logical & simple. After
observing a low processor utilization factor, the OS may attempt to improve it by
activating more processes. if no free pages are available, the holdings of the
already-active processes may be reduced. This may drive some of the processes
into the high page-fault zone. As a result, the processor utilization may drop while
the processes are awaiting their pages to be brought in. In order to improve the
still-decreasing processor utilization, the OS may decide to increase the degree
of multi-programming even further. Still more pages will be taken away from the
already-depleted holdings of the active processes, & the system is hopelessly on

Lesson No. 1 Intro. to Operating System 164

its way to thrashing. It is obvious that global replacement strategies are
susceptible to thrashing.
Thus a good design must make sure that the allocation algorithm is not unstable
& inclined toward thrashing. Knowing the typical patterns of program behavior,
we want to ensure that no process is allocated too few pages for its current
needs. Too few pages may lead to thrashing, & too many pages may unduly
restrict the degree of multi-programming & processor utilization.
Page-Fault Frequency (PFF)
This policy uses an upper & lower page-fault frequency threshold to decide for
allocation of new page frames. The PFF parameter P may be defined as: P = 1/T
Where T is the critical inter-page fault time. P is usually measured in number of
page faults per millisecond. The PFF algorithm may be implemented as follows:
1. The OS defines a system-wide (or per-process) critical page-fault frequency,
P.
2. The OS measures the virtual (process) time & stores the time of the most
recent page fault in the related process control block.
When a page fault occurs, the OS acts as follows:
¾ If the last page fault occurred less than T = 1/P ms ago, the process is
operating above the PFF threshold, & a new page frame is added from the
pool to house the needed page.
¾ Otherwise, the process is operating below the PFF threshold P, & a page
frame occupied by a page whose reference bit & written-into bit are not set is
freed to accommodate the new page.
¾ The OS sweeps & resets referenced bits of all resident pages. Pages that are
found to be unused, unmodified, & not shared since the last sweep are
released, & the freed page frames are returned to the pool for future
allocations.
For completeness, some policies need to be employed for process activation &
deactivation to maintain the size of the pool of free page frames within desired
limits.
6.2.2.7 Hardware Support & Considerations

Lesson No. 1 Intro. to Operating System 165

Virtual memory requires: (1) instruction interruptibility & restartability, (2) a
collection of page status bits associated with each page descriptor, and-if based
on paging-(3) a TLB to accelerate address translations. Choice of the page size
is an important design consideration in that it can have a significant impact on
performance of a virtual-memory system. In most implementations, one each of
the following bits is provided in every page descriptor:
¾ Presence bit, used to aid detection of missing items by the mapping hardware
¾ Written-into (modified) bit, used to reduce the overhead incurred by the
writing of unmodified replaced pages to disk
¾ Referenced bit, used to aid implementation of the replacement policy
An important hardware accelerator in virtual-memory systems is the TLB.
Although system architects & hardware designers primarily determine the details
of the TLB operation, the management of TLB is of interest because it deals with
problems quite similar to those discussed in the more general framework of
virtual memory. TLB hardware must incorporate allocation & replacement policies
so as to make the best use of the limited number of mapping entries that the TLB
can hold. An issue in TLB allocation is whether to devote all TLB entries to the
running process or to distribute them somehow among the set of active
processes. The TLB replacement policy governs the choice of the entry to be
evicted when a miss occurs & another entry needs to be brought in.
Allocation of all TLB entries to the running process can lead to relatively lengthy initial
periods of “loading” the TLB whenever a process is scheduled. This can lead to the
undesirable behavior observed in some systems when an interrupt service routine (ISR)
preempts the running process. Since a typical ISR is only a few hundred instructions
long, it may not have enough time to load the TLB. This can result in slower execution of
the interrupt service routine due to the need to reference PMT in memory while
performing address translations. Moreover, when the interrupted process is resumed, its
performance also suffers from having to load the TLB all over again. One way to combat
this problem is to use multi-context TLBs that can contain & independently manage the
PMT entries of several processes. With a multi-context TLB, when a process is
scheduled for execution, it may find some of its PMT entries left over in the TLB from the
preceding period of activity. Management of such TLBs requires the identity of the

Lesson No. 1 Intro. to Operating System 166

corresponding process to be associated with each entry, in order to make sure that
matches are made only with the TLB entries belonging to the process that produced the
addresses to be mapped.
Removal of TLB entries is usually done after each miss. If PMT entries of several
processes are in the buffer, the victim may be chosen either locally or globally.
Understandably, some preferential treatment is usually given to holdings of the
running process. In either case, least recently used is a popular strategy for
replacement of entries.
The problem of maintaining consistency between the PMT entries & their TLB
copies in the presence of frequent page moves must also be tackled by hardware
designers. Its solution usually relies on some specialized control instructions for
TLB flushing or for it selective invalidation.
Another hardware-related design consideration in virtual-memory systems is
whether I/O devices should operate with real or virtual addresses.
A hardware/software consideration involved in the design of paged systems Is
the choice of the page size. Primary factors that influence this decision are 1)
memory utilization & cost & 2) page-transport efficiency. Page-transport
efficiency refers to the performance cost & overhead of fetching page from the
disk or, in a diskless workstation environment, across the network. Loading of a
page from disk consists of two basic components: the disk-access time
necessary to position the heads over the target track & sector, & the page-
transfer time necessary to transfer the page to main memory thereafter. Head
positioning delays generally exceed disk-memory transfer times by order of
magnitude. Thus, total page-transfer time tends to be dominated by the disk
positioning delay, which is independent of the page size.
Small page size reduces page breakage, & it may make better use of memory by
containing only a specific locality of reference. Research results suggest that
procedures in many applications tend to be smaller than 100 words. On the other
hand, small pages may result in excessive size of mapping tables in virtual
systems with large virtual-address spaces. Page-transport efficiency is also

Lesson No. 1 Intro. to Operating System 167

adversely affected by small page sizes, since the disk-accessing overhead is
imposed for transferring a relatively small group of bytes.
Large pages tend to reduce table fragmentation & to increase page-transport
efficiency. This is because the overhead of disk accessing is amortized over a
larger number of bytes whenever a page is transferred between disk & memory.
On the negative side, larger pages may impact memory utilization by increasing
page breakage & by spanning more than one locality of reference. If multiple
localities contained in a single page have largely dissimilar patterns of reference,
the system may experience reduced effective memory utilization & wasted I/O
bandwidth. In general, the page-size trade-off is technology-dependent, & its
outcome tends to vary as the price & performance of individual components
change.
6.2.2.8 Protection & Sharing
The frequent moves of items between main & secondary memory may
complicate the management of mapping tables in virtual systems. When several
parties share an item in real memory, the mapping tables of all involved
processes must point to it. If the shared item is selected for removal, all
concerned mapping tables must be updated accordingly. The overhead involved
tends to outweigh the potential benefit or removing shared items. Many systems
simplify the management of mapping tables by fixing the shared objects in
memory.
An interesting possibility provided by large virtual-address spaces is to treat the
OS itself as a shared object. As such, the OS is mapped as a part of each user’s
virtual space. To reduce table fragmentation, dedicated mapping registers are
often provided to access a single physical copy of the page-map table reserved
for mapping references to the OS. One or more status bits direct the mapping
hardware to use the public or private mapping table, as appropriate for each
particular memory reference. In this scheme, different users have different
access rights to portions of the OS. Moreover, the OS-calling mechanism may be
simplified by avoiding expensive mode switches between users & the OS code.

Lesson No. 1 Intro. to Operating System 168

With the protection mechanism provided by mapping, a much faster CALL
instruction, or its variant, may be used to invoke the OS.
6.2.3 Segmentation & Paging
It is also possible to implement virtual memory in the form of demand
segmentation inheriting the benefits of sharing & protection provided by
segmentation. Moreover, their placement policies are aided by explicit
awareness of the types of information contained in particular segments. For
example, a “working set” of segments should include at least one each of code,
data, & stack segments. As with segmentation, inter-segment references alert
the OS to changes of locality. However, the variability of segment sizes & the
within-segment memory contiguity requirement complicate the management of
both main & secondary memories. Placement strategies are quite complex in
segmented systems. Moreover, allocation & deallocation of variable-size storage
areas to hold individual segments on disk imposes considerably more overhead
than handling of pages that are usually designed to fit in a single disk block.
On the other hand, paging is very easy for the management of main & secondary
memories, but it is inferior with regard to protection & sharing. The transparency
of paging necessitates the use of probabilistic replacement algorithms which
virtually no guidance from users, they are forced to operate mainly on the basis
of their observations of program behavior.
Both segmented & paged implementations of virtual memory have their
advantages/disadvantages. Some systems combine the two approaches in order
to enjoy the benefits of both. One approach is to use segmentation from the
user’s point of view but to divide each segment into pages of fixed size for
purposes of allocation. In this way, the combined system retains most of the
advantages of segmentation. At the same time, the problems of complex
segment placement & management of secondary memory are eliminated by
using paging.

Lesson No. 1 Intro. to Operating System 169

 Figure 5 – Segmentation & paging

The principle of address translation in combined segmentation & paging systems

is shown in Figure 5. Both segment descriptor tables & PMT are required for
mapping. Instead of containing the base & limit of the corresponding segment,
each entry of the SDT contains the base address & size of the PMT to be used
for mapping of the related segment’s pages. The presence bit in each PMT entry
indicates availability of the corresponding page in the real memory. Access rights
are recorded as a part of segment descriptors, although they may be placed or
refined in the entries of the PMT. Each virtual address consists of three fields:
segment number, page number, & offset within the page. When a virtual address
is presented to the mapping hardware, the segment number is used to locate the
corresponding PMT. Provided that the issuing process is authorized to make the
intended type of reference to the target segment, the page number is used to
index the PMT. If the presence bit is set, obtaining the page-frame address from
the PMT & combining this with the offset part of the virtual address complete the
mapping. If the target page is absent from real memory, the mapping hardware
generates a page-fault exception, which is processed. At both mapping stages,

Lesson No. 1 Intro. to Operating System 170

the length fields are used to verify that the memory references of the running
process lie within the confines of it address space.
Many variations of this powerful scheme are possible. For example, the presence
bit may be included with entries of the SDT. It may be cleared when no pages of
the related segment are in real memory. When such a segment is referenced,
bringing several of lits pages into main memory may process the segment fault.
In general, page re-fetching has been more difficult to implement in a way that
performs better than demand paging. One of the main reasons for this is the
inability to predict the use of previously un-referenced pages. However,
referencing of a particular segment increases the probability of its constituent
pages being referenced.
While the combination of segmentation & paging is certainly appealing, it
requires two memory accesses to complete the mapping of each virtual address
resulting into the reduction of the effective memory bandwidth by two-thirds. It
may be too much to bear even in the face of all the added benefits. Obviously,
hardware designers of such systems must assist the work of the OS by providing
ample support in terms of mapping registers & look aside buffers.
6.3 Keywords

6.4 SUMMARY
The memory-management layer of an OS allocates & reclaims portions of main
memory in response to requests from other users & from other OS modules, & in
accordance with the resource-management objectives of a particular system.
Processes are created & loaded into memory in response to scheduling
decisions that are affected by, among other things, the amount of memory
available for allocation at a given instant. Memory is normally freed when
resident objects terminate. When it is necessary & cost-effective, the memory
manager may increase the amount of available memory by moving inactive or
low-priority objects to lower levels of the memory hierarchy (swapping).
Thus, the memory manager interacts with the scheduler in selecting the objects
to be placed into or evicted from the main memory. The mechanics of memory

Lesson No. 1 Intro. to Operating System 171

management consists of allocating & reclaiming space, & of keeping track of the
state of memory areas. The objective of memory management is to provide
efficient use of memory by minimizing the amount of wasted memory while
imposing little storage, computational, & memory-access overhead. In addition,
the memory manager should provide protection by isolating distinct address
spaces, & facilitate inter-process cooperation by allowing access to shared data
& code. Partitioned allocation of memory imposes relatively little overhead, but it
restricts sharing & suffers from internal or external fragmentation. Segmentation
reduces the impact of fragmentation & offers superior protection & sharing by
dividing each process's address space into logically related entities that may be
placed into dis-contiguous areas of physical memory. Contiguity of the virtual-
address space is maintained by performing address translation at instruction
execution time. Hardware assistance is usually provided for this operation in
order to avoid a drastic reduction of the effective memory bandwidth. Paging
simplifies allocation & de-allocation of memory by dividing address spaces into
fixed-sized chunks. Execution-time translation of virtual to physical addresses,
usually assisted by hardware, is used to bridge the gap between contiguous
virtual addresses & dis-contiguous physical addresses where different pages
may reside.
Virtual memory removes the restriction on the size of address spaces of
individual processes that is imposed by the capacity of the physical memory
installed in a given system. In addition, virtual memory provides for dynamic
migration of portions of address spaces between primary & secondary memory
fin accordance with the relative frequency of usage.
6.5 SUGGESTED READINGS / REFERENCE MATERIAL
6. OS Concepts, 5th Edition, Silberschatz A., Galvin P.B., John Wiley & Sons.
7. Systems Programming & OSs, 2nd Revised Edition, Dhamdhere D.M.,
Tata McGraw Hill Publishing Company Ltd., New Delhi.
8. OSs, Madnick S.E., Donovan J.T., Tata McGraw Hill Publishing Company
Ltd., New Delhi.
9. OSs-A Modern Perspective, Gary Nutt, Pearson Education Asia, 2000.

Lesson No. 1 Intro. to Operating System 172

10. OSs, Harris J.A., Tata McGraw Hill Publishing Company Ltd., New Delhi,
2002.
6.6 SELF ASSESMENT QUESTIONS (SAQ)
1. What is the common drawback of all the real memory management
techniques? How is it overcome in virtual memory management schemes?
2. What is the basic difference between paging & segmentation? Which one
is better & why?
3. What extra hardware do we require for implementing demand paging &
demand segmentation?
4. Differentiate between internal & external fragmentation.
5. What do you understand by thrashing? What are the factors causing it?

Lesson No. 1 Intro. to Operating System 173

Lesson number: 7 Writer: Dr. Rakesh Kumar
Disk Scheduling Vetter: Prof. Dharminder Kr.

7.0 Objectives

The objective of this lesson is to make the students familiar with:

(a) The characteristics of the disk that affect the performance.

(b) A number of disk scheduling algorithms to improve the access time.
(c) Disk scheduling algorithm in fixed-head storage devices.
7.1 Introduction

We know that the processor and memory of the computer should be scheduled such that

the system operates more efficiently. Another very important scheduler is the disk

scheduler. The disk can be considered the one I/O device that is common to every

computer. Most of the processing of computer system centers on the disk system. Disk

provides the primary on-line storage of information, both programs and data. All the

important programs of the system such as compiler, assemblers, loaders, editors, etc. are

stored on the disk until loaded into memory. Hence it becomes all-important to properly

manage the disk storage and it scheduling.

7.2 Presentation of contents

7.2.1 Storage device characteristics

7.2.1.1 Seek time
7.2.1.2 Latency time
7.2.1.3 Transfer time
7.2.2 Disk Scheduling

7.2.2.1 First come first serve (FCFS) scheduling

7.2.2.2 Shortest seek time first (SSTF) scheduling

Lesson No. 1 Intro. to Operating System 174

 7.2.2.3 Scan
7.2.2.4 C-scan (Circular scan)
7.2.2.5 Look

7.2.2.6 N-step scan

7.2.2.7 F-Scan

7.2.3 Scheduling algorithm selection

7.2.4 Sector queuing

7.2.1 Storage device characteristics

Disk comes in many sizes and speeds, and information may be stored optically or

magnetically. However, all disks share a number of important features. A disk is a flat

circular object called a platter. Information may be stored on both sides of a platter

(although some multiplatter disk packs do not use the top most or bottom most surface).

Platter rotates around its own axis. The circular surface of the platter is coated with a

magnetic material on which the information is stored. A read/write head is used to

Lesson No. 1 Intro. to Operating System 175

perform read/write operations on the disk.

Spindle

Platters

Read/write
heads

Moving head disk mechanism

The read write head can move radially over the magnetic surface. For each position of the

head, the recorded information forms a circular track on the disk surface. Within a track

information is written in blocks. The blocks may be of fixed size or variable size

separated by block gaps. The variable length size block scheme is flexible but difficult to

implement. Blocks can be separately read or written. The disk can access any information

randomly using an address of the record of the form (track no, record no.). When the disk

Lesson No. 1 Intro. to Operating System 176

is in use a drive motor spins it at high speed. The read/write head positioned just above

the recording surface stores the information magnetically on the surface.

Tracks and sectors on a disk

On floppy disks and hard disks, the media spins at a constant rate. Sectors are organized

into a number of concentric circles or tracks. As one moves out from the center of the

disk, -the tracks get larger. Some disks store the same number of sectors on each track,

with outer tracks being recorded using lower bit densities. Other disks place more sectors

on outer tracks. On such a disk, more information can be accessed from an outer track

than an inner one during a single rotation of the disk.

The disk speed is composed of three parts:

(a) Seek Time
(b) Latency Time
(c) Transfer time
7.2.1.1 Seek time
To access a block from the disk, first of all the system has to move the read/write
head to the required position. The time consumed in this operation is known as

Lesson No. 1 Intro. to Operating System 177

seek time and the head movement is called seek. Seek time (S) is determined in
terms of:
I: startup delays in initiating head movement.
H: the rate at which read/write head can be moved.
C: How far the head must travel.
S=H*C+I
7.2.1.2 Latency time
Once the head is positioned at the right track, the disk is to be rotated to move
the desired block under the read/write head. This delay is known as latency time.
On average this will be one-half of one revolution. Thus latency can be computed
by dividing the number of revolution per minute (RPM), R, into 30.
L = 30 / R
7.2.1.3 Transfer time
Finally the actual data is transferred from the disk to main memory. The
consumed in this operation is known as transfer time. Transfer time T, is
determined by the amount of information to be read, B; the number of bytes per
track, N; and the rotational speed.
T = 60B/RN
So the total time (A) to service a disk request is the sum of these tree i.e. seek
time, latency time, and transfer time.
A=S+L+T
Since most of the systems depend heavily on the disk, so it become very
important to make the disk service as fast as possible.
So a number of variations have been observed in disk organization motivated by
the desire to reduce the access time, increase the capacity of the disk and to
make optimum use of disk surface. For example there may be one head for
every track on the disk surface. Such arrangement is known as fixed-head disk.
In this arrangement it is very easy for computer to switch from one track to
another quickly but it makes the disk very expensive due to the requirement of a
number of heads. Generally there is one head that moves in and out to access
different tracks because it is cheaper option.

Lesson No. 1 Intro. to Operating System 178

Higher disk capacities are obtained by mounting many platters on the same
spindle to form a disk pack. There is one read/write head per circular surface of a
platter. All heads of the disk are mounted on a single disk arm, which moves
radially to access different tracks. Since the heads are located on the identically
positioned tracks of different surfaces, so such tracks can be accessed without
any further seeks. So placing that data in one cylinder can speed up sequential
access. Cylinder is a collection of identically positioned tracks of different
surfaces.
The hardware for a disk system can be divided into two parts. The disk drive is
the mechanical part, including the device motor, the read/write heads and
associated logic. The other part called the disk controller determines the logical
interaction with the computer. The controller takes instructions from the CPU and
orders the disk drive to carry out the instructions.
Every disk drive has a queue of pending requests to be serviced. Whenever a
process needs I/O to or from the disk, it issues a request to the operating system,
which is placed in the disk queue. The request specifies the disk address,
memory address, amount of information to be transferred, and the type of
operation (input or output).

7.2.2 Disk Scheduling

For a multiprogramming system with many processes, the disk queue may often
be non-empty. Thus, when a request is complete, the disk scheduler has to pick
a new request from the queue and service it. As apparent, the amount of head
movement needed to satisfy a series of I/O requests could affect the
performance. For this reason, a number of scheduling algorithms have been
proposed.

7.2.2.1 First cum first served (FCFS) scheduling

This form of scheduling is the simplest one but may not provide the best service.
The algorithm is very easy to implement. In it the system picks every time the first
request from the disk queue. In this scheduling the total seek time may be
substantially high as evident from the following example:

Lesson No. 1 Intro. to Operating System 179

Considered an ordered disk queue with requests involving tracks:
86,140, 23, 50, 12, 89, 14, 120, 64
The following figure shows the movement of read/write head in First Come First
Serve scheduling.
Queue: 86, 140, 23, 50, 12, 89, 14, 120, 64
Head starts at 58

12 14 23 50 58 64 86 89 120 140

First Come First Serve Disk scheduling

7.2.2.2 Shortest seek time first (SSTF) scheduling

This scheduling algorithm services the request whose track position is closest to
the current track position. Shortest-Seek-Time-First selects the request that is
asking for minimum seek time from the current head position. Since seek time is
generally proportional to the track difference between the requests, this approach
is implemented by moving the head to the closest track in the request queue.
The following figure shows the read/write head movement in Shortest-Seek-
Time-First scheduling for the above example discussed in First Come First Serve
scheduling. It shows a substantial improvement in disk services i.e. reduction in
the total movement of the head resulting into the reduced seek time.

Lesson No. 1 Intro. to Operating System 180

Shortest Seek Time First is just like the Shortest Job First process scheduling.
So it is having the limitations of Shortest Job First also. It may cause starvation of
some requests.
Queue: 86, 140, 23, 50, 12, 89, 14, 120, 64
Head starts at 58

12 14 23 50 58 64 86 89 120 140

Shortest Seek Time First Scheduling

7.2.2.3 Scan
In this algorithm the read/write head moves back and forth between the
innermost and outermost tracks. As the head gets to each track, satisfies all
outstanding requests for that track. In this algorithm also, starvation is possible
only if there are repeated requests for the current track.
The scan algorithm is sometimes called the elevator algorithm. As it is familiar to
the behavior of elevators as they service requests to move from floor to floor in a
building.

Lesson No. 1 Intro. to Operating System 181

 Queue: 86, 140, 23, 50, 12, 89, 14, 120, 64
Head starts at 58

12 14 23 50 58 64 86 89 120 140

7.2.2.4 C-scan (Circular scan)

C-scan is a variant of scan. It is designed to provide a more uniform wait time. C-
scan moves the head from one end of the disk to another, servicing requests as
it goes. When it reaches the other end, however, it immediately return to the
beginning of the disk, without servicing any requests on the return trip. C-scan
treats the disk, as it was circular, with the last track adjacent to the first one.

Lesson No. 1 Intro. to Operating System 182

 Queue: 86, 140, 23, 50, 12, 89, 14, 120, 64
Head starts at 58

12 14 23 50 58 64 86 89 120 140

7.2.2.5 Look
This algorithm is also similar to scan but unlike scan, the head does not
unnecessarily travel to the innermost track and outermost track on each circuit. In
this algorithm, head moves in one direction, satisfying the request for the closest
track like scan in that direction. When there are no more requests in that direction
the head is traveling, head reverse the direction and repeat.

7.2.2.6 N-step scan

In it the request queue is divided into sub queues with each sub queue having a
maximum length of N. Sub queues are processed in FIFO order. Within a sub
queue, requests are processed using Scan. While a sub queue is being serviced,
incoming requests are placed in the next non-filled sub queue. N-step scan
eliminates any possibility of starvation.

7.2.2.7 F-Scan
The "F" stands for "freezing" the request queue at a certain time. It is just like N-
step scan but there are two sub queues only and each is of unlimited length.

Lesson No. 1 Intro. to Operating System 183

While requests in one sub queue are serviced, new requests are placed in other
sub queue.

7.2.3 Scheduling algorithm selection

As there are so many disk-scheduling algorithms, an important question is how to
choose a scheduling algorithm that will optimize the performance. The commonly
used algorithm is Shortest-Seek-Time-First and it has a natural appeal also. San
and its variants are more appropriate for system with a heavy load on the disk. It
is possible to define an optimal scheduling algorithm, but computational
overheads required for that may not justify the savings over Shortest-Seek-Time-
First and scan.
No doubt in any scheduling algorithm the performance depends on the number
and types of the requests. If every time there is only one outstanding request,
then the performance of all the scheduling algorithms will be more or less
equivalent. Studies suggest that even First-Come-First-Serve performance will
also be reasonably well.
It is also observed that performance of scheduling algorithms is also greatly
influenced by the file allocation method. The requests generated by the
contiguously allocated files will result in minimum movement of the head. But in
case of indexed access and direct access where the blocks of a file are scattered
on the disk surface resulting into a better utilization of the storage space, there
may a lot of movement of the head.
In all these algorithms, to improve the performance, the decision is taken on the
basis of head movement i.e. seek time. Latency time is not considered as a
factor. Because it is not possible to predict latency time because rotational
location cannot be determined. However, multiple requests for the same track
may be serviced based on latency.

7.2.4 Sector queuing

In case of disk with one read/write head, the objective of the entire scheduling
algorithm is to minimize the seek time by minimizing the movement o read/write

Lesson No. 1 Intro. to Operating System 184

head. The dick scheduling algorithms like First Come First Serve, Shortest-Seek-
Time-First, Scan, and C-scan centers on this objective. But in case of the storage
devices such as drum, which has the fixed head, this is not an issue at all. So
different scheduling algorithms is used for this type of devices known as sector
queuing.
In fixed-head storage devices the tracks are divided into a fixed number of
blocks, known as sectors. Any I/O requests specify the address composed of
track number and sector number. As seek time is zero for fixed-head storage
devices, the important issue is the latency time to improve the performance.
In sector queuing a separate queue is maintained for each sector of the drum.
When a request arrives for sector i, it is placed in the queue for sector i. As
sector i rotates beneath the read/write head, the first request in its queue is
serviced.
Formatting
Before data can be written to a disk, all the administrative data must be written to
the disks, organizing it into sectors. This low-level formatting or physical
formatting is often done by the manufacturer. In the formatting process, some
sectors may. be found to be defective. Most disks have spare sectors, and a
remapping mechanism substitutes spare sectors for defective ones.
For sectors that fail after formatting, the operating
system may implement a bad block mechanism. Such mechanisms
are usually in terms of blocks and are implemented at a
level above the device driver. The manner in which sectors
are positioned on a track can affect disk performance. If
disk I/O operations are limited to transferring a single
sector at a time, to read multiple sectors in sequence,
separate I/O operations must be performed. After the first
I/O operation completes, its interrupt must be processed
and the second I/O operation must be issued. During this
time, the disk continues to spin. If the start of the next
sector to be read has already spun past the read/write

Lesson No. 1 Intro. to Operating System 185

head, the sector cannot be read until the next revolution
of the disk brings the cylinder by the read/write head. In
a worst-case scenario, the disk must wait almost a full
revolution. To avoid this problem, sectors may be
interleaved. The degree of interleaving is determined by
how far the disk revolves in the time from the end of one
I/O operation until the controller can issue a subsequent
I/O operation. The sector layout, given different degrees
of interleaving, is illustrated in Fig. 5.
For most modem hard-disk controllers, interleaving is not used. The controller
contains sufficient memory to store an entire track, so a single I/O operation may
be used to read all the sectors on a track. Interleaving is more commonly used
on less sophisticated disk systems, like floppy disks.

7 0 7 0 5 0
6 1 3 4 2 3
5 2 6 1 7 6
4 3 2 5 4 1

Non-interleaved Single Inteleaved Double Interleaved

Interleaving
Most operating systems also provide the ability for disks to be divided into one or
more virtual disks called partitions. On personal computers, Unix, Windows, and
DOS all adhere to a common partitioning scheme so that they all may co-reside
on a single disk.
Before files can be written to a disk, an empty file system must be created on a
disk partition. This requires the various data structures required by the file system
to be written to the partition, and is known as a high- level format or logical
format. A file system is not required to make use of a disk. Some operating
systems allow applications to write directly to a disk device. To an application, a
directly accessible disk device is just a large sequential collection of storage
blocks. In such a case, it is the responsibility of the application to impose an
order on the information written there.

Lesson No. 1 Intro. to Operating System 186

7.3 Summary
As processor and main memory speeds increase more rapidly than those of
secondary storage devices, optimizing disk performance has become important
to realizing optimal performance. Magnetic storage records data by changing the
direction of magnetization of regions, each representing a 1 or a 0. To access
data, a current-carrying device called a read/write head hovers above the
medium as it moves.
Most modern computers use hard disks as secondary storage. As the platters
spin, each read-write head sketches out a circular track of data on a disk surface.
All read-write heads are attached to a single disk arm. When the disk arm moves
the read/write heads to a new position, a different set of tracks becomes
accessible. The time it takes for the head to move from its current cylinder to the
one containing the data record being accessed is called the seek time. The time
it takes for data to rotate from its current position to a position adjacent to the
read/write head is called latency time. Then the record must be made to spin by
the read/write head so that the data can be read from or written to the disk. Many
processes can generate requests for reading and writing data on a disk
simultaneously. Because these processes sometimes make requests faster than
they can be serviced by the disk, queues build up to hold disk requests. Some
early computing systems simply serviced these requests on a first-come-first-
served (FCFS) basis. FCFS is a fair method of allocating service, but when the
request rate becomes heavy, FCFS results in long waiting times.
FCFS exhibits a random seek pattern in which successive requests can cause
time-consuming seeks from the innermost to the outermost cylinders. To reduce
the time spent seeking records, it seems reasonable to order the request queue
in some other manner than FCFS. This reordering is called disk scheduling. The
two most common types of scheduling are seek optimization and rotational
optimization. Disk scheduling strategies often are evaluated by comparing their
throughput, mean response time and variance of response times.

Lesson No. 1 Intro. to Operating System 187

Shortest-seek-time-first (SSTF) scheduling next services the request that is
closest to the read-write head's current cylinder, even if that is not the first one in
the queue. By reducing average seek times, SSTF achieves higher throughput
rates than FCFS, and mean response times tend to be lower for moderate loads.
One significant drawback is that higher variances occur on response times
because of the discrimination against the outermost and innermost tracks; in the
extreme, indefinite postponement of requests far from the read-write heads could
occur.
The SCAN scheduling strategy reduces unfairness and variance of response
times by choosing the request that requires the shortest seek distance in a
preferred direction. Thus, if the preferred direction is currently outward, the SCAN
strategy chooses the shortest seek distance in the outward direction. However,
because SCAN ensures that all requests in a given direction will be serviced
before the requests in the opposite direction, it offers a lower variance of
response times than SSTF.
The circular SCAN (C-SCAN) modification to the SCAN strategy moves the arm
from the outer cylinder to the inner cylinder, servicing requests on a shortest-
seek basis. When the arm has completed its inward sweep, it jumps to the
outermost cylinder, then resumes its inward sweep processing requests. C-
SCAN maintains high levels of throughput while further limiting variance of
response times by avoiding the discrimination against the innermost and
outermost cylinders.
The FSCAN and N-Step SCAN modifications to the SCAN strategy eliminate the
possibility of indefinitely postponing requests. FSCAN uses the SCAN strategy to
service only those requests waiting when a particular sweep begins. Requests
arriving during a sweep are grouped together and ordered for optimum service
during the return sweep. N-Step SCAN services the first n requests in the queue
using the SCAN strategy. When the sweep is complete, the next n requests are
serviced. Arriving requests are placed at the end of the request queue, which
prevents requests in the current sweep from being indefinitely postponed.
FSCAN and N-Step SCAN offer good performance due to high throughput, low

Lesson No. 1 Intro. to Operating System 188

mean response times and a lower variance of response times than SSTF and
SCAN.
The LOOK variation of the SCAN strategy "looks" ahead to the end of the current
sweep to determine the next request to service. If there are no more requests in
the current direction, LOOK changes the preferred direction and begins the next
sweep, stopping when passing a cylinder that corresponds to a request in the
queue. This strategy eliminates unnecessary seek operations experienced by
other variations of the SCAN strategy by preventing the read/write head from
moving to the innermost or outermost cylinders unless it is servicing a request to
those locations.
The circular LOOK (C-LOOK) variation of the LOOK strategy uses the same
technique as C-SCAN to reduce the bias against requests located at extreme
ends of the platters. When there are no more requests on a current sweep, the
read/write head moves to the request closest to the outer cylinder and begins the
next sweep. The C-LOOK strategy is characterized by potentially lower variance
of response times compared to LOOK, and by high throughput, although lower
than that of LOOK. Sector queuing is a scheduling algorithm for fixed head
devices such as drums.

7.4 Keywords
¾ Seek time: To access a block from the disk, first of all the system has to
move the read/write head to the required position. The time consumed in
moving the head to access a block from the disk is known as seek time.
¾ Latency time: the time consumed in rotating the disk to move the desired
block under the read/write head.
¾ Transfer time: The consumed in transferring the data from the disk to the
main memory is known as transfer time.
7.5 SELF ASSESMENT QUESTIONS (SAQ)
1. Compare the throughput of scan and C-scan assuming a uniform
distribution of requests.

Lesson No. 1 Intro. to Operating System 189

 2. What do you understand by seek time, latency time, and transfer time?
Explain.
3. Shortest Seek Time First favors tracks in the center of the disk. On an
operating system using Shortest Seek Time First, how might this affect the
design of the file system?
4. When there is no outstanding request in the queue, all the disk-scheduling
algorithms reduce to First Come First Serve scheduling? Explain why?
5. The entire disk scheduling algorithms except First Come First Serve may
cause starvation and hence not truly fair.
- Explain why.
- Come up with a scheme to ensure fairness.
- Why is fairness an important goal in a time-sharing system?
7.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B., John
Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.
3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company Ltd.,
New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 190

 Lesson number: 8 Writer: Dr. Rakesh Kumar
Process Management Vetter: Prof. Dharminder Kr.

8.0 Objectives
The objective of this lesson is get the students acquainted with the concepts of
process synchronization. This lesson will make them familiar with:
(a) Critical section
(b) Mutual exclusion
(c) Classical coordination problems
8.1 Introduction
Since processes frequently need to communicate with other processes therefore,
there is a need for a well-structured communication, without using interrupts,
among processes. Processes use two kinds of synchronization to control their
activities; (a) Control synchronization: it is needed if a process waits to perform
some action only after some other processes have executed some action, (b)
Data access synchronization: It is used to access shared data in a mutually
exclusive manner. The basic technique used to implement this synchronization is
to block a process until an appropriate condition is fulfilled. In this lesson
synchronization in concurrent processes is discussed. Some classical
coordination problems such as dining philosopher problem, producer-consumer
problem etc are also discussed. These classical problems are the abstractions of
the synchronization problems observed in operating systems.
8.2 Presentation of contents
8.2.1 Race Conditions
8.2.2 Critical Section
8.2.3 Mutual Exclusion
8.2.3.1 Mutual Exclusion Conditions
8.2.3.2 Proposals for Achieving Mutual Exclusion
8.2.4 Classical Process Co-Ordination Problems

Lesson No. 1 Intro. to Operating System 191

 8.2.4.1 The readers and writers problem
8.2.4.2 The Bounded Buffer Producers and Consumers
8.2.5 Semaphores
8.2.5.1 Producer-Consumer Problem Using Semaphores
8.2.6 The dining philosophers problem
8.2.1 Race Conditions
Consider the following extremely simple procedure
void deposit(int amount)
{
balance = balance + amount;
}
(Where we assume that balance is a shared variable). If two processes try to call
deposit concurrently, something very bad can happen. The single statement
balance = balance + amount is really implemented, on most computers, by a
sequence of instructions such as
Load Reg, balance
Add Reg, amount
Store Reg, balance
Suppose process P1 calls deposit(10) and process P2 calls deposit(20). If one
completes before the other starts, the combined effect is to add 30 to the
balance, as desired. However, suppose the calls happen at exactly the same
time, and the executions are interleaved. Suppose the initial balance is 100, and
the two processes run on different CPUs. One possible result is
P1 loads 100 into its register
P2 loads 100 into its register
P1 adds 10 to its register, giving 110
P2 adds 20 to its register, giving 120
P1 stores 110 in balance
P2 stores 120 in balance
and the net effect is to add only 20 to the balance!

Lesson No. 1 Intro. to Operating System 192

This kind of bug, which only occurs under certain timing conditions, is called a
race condition. It is an extremely difficult kind of bug to track down (since it may
disappear when you try to debug it) and may be nearly impossible to detect from
testing (since it may occur only extremely rarely). The only way to deal with race
conditions is through very careful coding. To avoid these kinds of problems,
systems that support processes always contain constructs called synchronization
primitives.
In operating systems, processes that are working together share some common
storage (main memory, file etc.) that each process can read and write. When two
or more processes are reading or writing some shared data and the final result
depends on who runs precisely when, are called race conditions. Concurrently
executing threads that share data need to synchronize their operations and
processing in order to avoid race condition on shared data. Only one ‘customer’
thread at a time should be allowed to examine and update the shared variable.
Race conditions are also possible in Operating Systems. If the ready queue is
implemented as a linked list and if the ready queue is being manipulated during
the handling of an interrupt, then interrupts must be disabled to prevent another
interrupt before the first one completes. If interrupts are not disabled than the
linked list could become corrupt.
So a race condition on a data item arises when many processes concurrently
update its value. To maintain consistency, any time only one process should
update the value. How to avoid race conditions? One solution is critical section.
8.2.2 Critical Section

Lesson No. 1 Intro. to Operating System 193

The key to preventing trouble involving shared storage is find some way to
prohibit more than one process from reading and writing the shared data
simultaneously. That part of the program where the shared memory is accessed
is called the Critical Section. To avoid race conditions and flawed results, one
must identify codes in Critical Sections in each thread. The characteristic
properties of the code that form a Critical Section are
¾ Codes that reference one or more variables in a “read-update-write”
fashion while any of those variables is possibly being altered by another
thread.
¾ Codes that alter one or more variables that are possibly being referenced
in “read-update-write” fashion by another thread.
¾ Codes use a data structure while any part of it is possibly being altered by
another thread.
¾ Codes alter any part of a data structure while it is possibly in use by
another thread.
Here, the important point is that when one process is executing shared modifiable data in

its critical section, no other process is to be allowed to execute in its critical section.

Thus, the execution of critical sections by the processes is mutually exclusive in time. So

a critical section for a data item d is defined as a section of code, which cannot be

executed concurrently with itself or with other critical sections of d.

Consider a system consisting of n processes {P0, P1, ..., Pn-i}. Each process has
a segment of code, called a critical section, in which the process may be
changing common variables, updating a table, writing a file, and so on. The
important feature of the system is that, when one process is executing its critical
section, no other process is to be allowed to execute in its critical section. Thus,
the execution of critical sections by the processes is mutually exclusive in time.
The critical-section problem is to design a protocol that the processes can use to
co-operate. Each process must request permission to enter its critical section.
The section of code implementing this request is the entry section. An exit

Lesson No. 1 Intro. to Operating System 194

section may follow the critical section. The remaining code is the remainder
section.
Repeat

Critical section

remainder section
until FALSE
Properties of critical section

The following properties are to be possessed by an implementation of critical section:

¾ Correctness: At most one process may execute a critical section at any given moment.
¾ Progress: When a critical section is not in use, one of the processes visiting to enter it
will be granted entry to the critical section. If no process is executing in its critical
section and there exist some processes that wish to enter their critical sections, then
only those processes that are not executing in their remainder section can participate
in the decision of which will enter its critical section next. Moreover, this decision
cannot be postponed indefinitely. So if no process is in critical section, one can decide
quickly who enters and only one process can enter the critical section so in practice,
others are put on the queue.
¾ Bounded wait: After a process p has indicated its desire to enter a critical section, the
number of times other processes gain entry to the critical section ahead of p is
bounded by a finite integer. So there must exist a bound on the number of times that
other processes are allowed to enter their critical sections after a process has made a
request to enter its critical section and before that request is granted. The wait is the
time from when a process makes a request to enter its critical section until that
request is granted. In practice, once a process enters its critical section, it does not get
another turn until a waiting process gets a turn (managed as a queue)
¾ Deadlock freedom: The implementation is free of deadlock.

Lesson No. 1 Intro. to Operating System 195

8.2.3 Mutual Exclusion
It is a way of making sure that if one process is using a shared modifiable data,
the other processes will be excluded from doing the same thing.
Formally, while one process executes the shared variable, all other processes
desiring to do so at the same time moment should be kept waiting; when that
process has finished executing the shared variable, one of the processes waiting;
one of the processes waiting to do so should be allowed to proceed. In this
fashion, each process executing the shared data (variables) excludes all others
from doing so simultaneously. This is called Mutual Exclusion.
Note that mutual exclusion needs to be enforced only when processes access
shared modifiable data - when processes are performing operations that do not
conflict with one another they should be allowed to proceed concurrently.
8.2.3.1 Mutual Exclusion Conditions
If we could arrange matters such that no two processes were ever in their critical
sections simultaneously, we could avoid race conditions. We need four
conditions to hold to have a good solution for the critical section problem (mutual
exclusion).
¾ No two processes may at the same moment inside their critical sections.
¾ No assumptions are made about relative speeds of processes or number
of CPUs.
¾ No process outside its critical section should block other processes.
¾ No process should wait arbitrary long to enter its critical section.
8.2.3.2 Proposals for Achieving Mutual Exclusion
The mutual exclusion problem is to devise a pre-protocol (or entry protocol) and
a post-protocol (or exist protocol) to keep two or more threads from being in their
critical sections at the same time.
Problem
When one process is updating shared modifiable data in its critical section, no
other process should allow to enter in its critical section.

Lesson No. 1 Intro. to Operating System 196

Proposal 1 -Disabling Interrupts (Hardware Solution)
Each process disables all interrupts just after entering in its critical section and
re-enable all interrupts just before leaving critical section. With interrupts turned
off the CPU could not be switched to other process. Hence, no other process will
enter its critical section and mutual exclusion achieved.
Disabling interrupts is sometimes a useful technique within the kernel of an
operating system, but it is not appropriate as a general mutual exclusion
mechanism for users process. The reason is that it is unwise to give user
process the power to turn off interrupts.
Proposal 2 - Lock Variable (Software Solution)
In this solution, we consider a single, shared, (lock) variable, initially 0. When a
process wants to enter in its critical section, it first tests the lock. If lock is 0, the
process first sets it to 1 and then enters the critical section. If the lock is already
1, the process just waits until (lock) variable becomes 0. Thus, a 0 means that no
process in its critical section, and 1 means hold your horses - some process is in
its critical section.
Repeat
While Lock = 1
Do {nothing};
{Critical section}
Lock = 0
{Remainder of the cycle}
Forever
The flaw in this proposal can be best explained by example. Suppose process A
sees that the lock is 0. Before it can set the lock to 1 another process B is
scheduled, runs, and sets the lock to 1. When the process A runs again, it will
also set the lock to 1, and two processes will be in their critical section
simultaneously.
Proposal 3 - Strict Alteration
In this proposed solution, the integer variable 'turn' keeps track of whose turn is
to enter the critical section. Initially, process A inspects turn, finds it to be 1, and

Lesson No. 1 Intro. to Operating System 197

enters in its critical section. Process B also finds it to be 1 and sits in a loop
continually testing 'turn' to see when it becomes 2.Continuously testing a variable
waiting for some value to appear is called the Busy-Waiting.
Algorithm
Var num: integer
Begin
Turn=1
Concurrent begin
Repeat Repeat
While turn=2 While turn=1
Do {nothing} Do {nothing}
{critical section} {critical section}
turn=2 turn=1
{remainder of the cycle} {remainder of the cycle}
forever forever
concurrent end
end.

Process 1 Process 2
The shared variable turn is used to indicate which process can enter the critical
section next. Let process p1 wish to enter the Critical Section. If turn=1, p1 can
enter straightway. After completing the Critical Section, it sets turn to 2 so as to
enable process p2 to enter the Critical Section. If p1 finds turn=2 when it wishes
to enter Critical Section, it waits in the while loop until p2 exits from the critical
section and executes the assignment turn=1. Thus processes may encounter a
busy wait before gaining entry to the Critical Section.
Taking turns is not a good idea when one of the processes is much slower than
the other. Suppose process 1 finishes its critical section quickly, so both
processes are now in their non-critical section. This situation violates above-
mentioned condition 3(i.e. No process outside its critical section should block
other processes.). E.g. let process p1 be in critical section and p2 in the

Lesson No. 1 Intro. to Operating System 198

remainder section. If p1 exits from the Critical Section, finishes the remainder
section and wishes to enter the critical section again, it will face busy wait until
after p2 uses the Critical Section. So the progress condition is violated since p1
is not able to enter although critical section is available. The solution of this
problem can be:
Var c1, c2: integer
C1=1;c2=1;
{
repeat repeat
while c2=0 do {nothing} while c1=0 do {nothing}
c1=0; c2=0;
{critical section} {critical section}
c1=1; c2=1;
{remainder section} {remainder section}
forever forever
}
process p1 process p2
In this algorithm, c1 is a status flag for p1. p1 sets this flag to 0 while entering the
critical section and change it to 1 upon exit. P2 checks the status of c1, if it is 1
then enter otherwise wait. This check eliminates the progress violation by
enabling a process to enter critical section again any number of times if other
process is not interested in entering the Critical Section. However the busy wait
condition still exists.
Using Systems calls 'sleep' and 'wakeup'
Basically, what above-mentioned solution does is this: when a process wants to
enter in its critical section, it checks to see if the entry is allowed. If it is not, then
it waits until it is allowed to enter. This approach waste CPU-time.
Now look at some interprocess communication primitives i.e. the pair of steep-
wakeup.
™ Sleep

Lesson No. 1 Intro. to Operating System 199

 It is a system call that causes the caller to block, that is, be suspended until
some other process wakes it up.
™ Wakeup
It is a system call that wakes up the process.
Both 'sleep' and 'wakeup' system calls have one parameter that represents a
memory address used to match up 'sleeps' and 'wakeups'.
8.2.4 CLASSICAL PROCESS CO-ORDINATION PROBLEMS
There are a number of different process co-ordination problems arising in
practical situations that exemplify important associated issues. These problems
also provide base for solution testing for process co-ordination problems. In this
section, we will see some of such classical process co-ordination problems.
8.2.4.1 The readers and writers problem
Courtois, Heymans, and Pumas posed an interesting synchronization problem
called the readers-writers problem. Suppose a resource is to be shared among a
community of processes of two distinct types: readers and writers. A reader
process can share the resource with any other reader process but not with any
writer process. A writer process acquires exclusive access to the resource
whenever it requires any access to the resource.
This scenario is similar to one in which a file is to be shared among a set of
processes, If a process wants only to read the file, then it may share the file with
any other process that also wants to read the file. If a writer wants to modify the
file, then no other process should have access to the file writer has access to it.
The correctness conditions are as follows:
1. There are two classes of processes:
(a) Readers, which can work concurrently.
(b) Writers, which need exclusive access.
2. Only one writer can write any time. So we must prevent 2 writers from
being concurrent.
3. We must prevent a reader and a writer from being concurrent. So reading
is prohibited while a writer is writing.

Lesson No. 1 Intro. to Operating System 200

4. We must permit readers to be concurrent when no writer is active. So
many readers can read simultaneously.
5. Perhaps we want fairness (i.e., freedom from starvation).
6. Optional variants can be:
a. Writer should have priority over readers.
b. Readers should have priority over writers.
The “easy way out'' is to treat all processes as writers in which case the problem reduces

to mutual exclusion. The disadvantage of the easy way out is that you give up reader

concurrency. A possible solution is as under:

{ {

repeat repeat

if a writer is writing if readers reading writer is writing

then {wait} then {wait}

{read} {write}

If writer waiting If reader or writer waiting

Then Then

Wake a writer if no Wake all readers

Readers are reading. Or one writer

forever forever

} }

Readers Writers

Lesson No. 1 Intro. to Operating System 201

 8.2.4.2 The Bounded Buffer Producers and Consumers
As an example how sleep-wakeup system calls are used, consider the producer-
consumer problem also known as bounded buffer problem. In the producers and
consumers problem, there are two classes of processes
¾ Producers, which produce items and insert them into a buffer.
¾ Consumers, which remove items and consume them.
These two processes share a common, fixed-size (bounded) buffer. The
producer puts information into the buffer and the consumer takes information out.
Trouble arises when
1. The producer wants to put a new data in the buffer, but buffer is already
full.
Solution: Producer goes to sleep and to be awakened when the consumer
has removed data.
2. The consumer wants to remove data from the buffer but buffer is already
empty.
Solution: Consumer goes to sleep until the producer puts some data in
buffer and wakes consumer up.
The bounded buffer producers and consumers assume that there is a fixed buffer
size i.e., a finite numbers of slots are available.
This approach also leads to same race conditions we have seen in earlier
approaches. Race condition can occur due to the fact that access to 'count' is
unconstrained. The essence of the problem is that a wakeup call, sent to a
process that is not sleeping, is lost.
Another solution is the use of monitors. Monitors is a high-level data abstraction tool that

automatically generates atomic operations on a given data structure. A monitor has:

¾ Shared data.
¾ A set of atomic operations on that data.
¾ A set of condition variables.
Each monitor has one lock. It acquires lock when begin a monitor operation, and
releases lock when operation finishes. It statically identifies operations that only

Lesson No. 1 Intro. to Operating System 202

read data, and then allow these read-only operations to go concurrently. Writers
get mutual exclusion with respect to other writers and to readers. The
advantages of using monitors are (i) it reduces probability of error (never forget to
Acquire or Release the lock), (ii) biases programmer to think about the system in
a certain way (is not ideologically neutral).
Bounded buffer using monitors and signals
¾ Shared State data [num] - a buffer holding produced data. num - tells how
many produced data items there are in the buffer.
¾ Atomic Operations Produce (v) called when producer produces data item
v. Consume (v) called when consumer is ready to consume a data item.

Consumed item put into v.

¾ Condition Variables There are two condition variables (1) bufferAvail -
signalled when a buffer becomes available. (2) dataAvail - signalled when
data becomes available.

Condition *bufferAvail, *dataAvail;

int num = 0;
int data[10];
Lock *monitorLock;
Produce(v) {
monitorLock->Acquire(); /* Acquire monitor lock - makes operation atomic */
while (num == 10) {
bufferAvail->Wait(monitorLock);
}
put v into data array
num++;
dataAvail->Signal(monitorLock);
monitorLock->Release(); /* Release monitor lock after perform operation */
}
Consume(v) {
monitorLock->Acquire(); /* Acquire monitor lock - makes operation atomic */

Lesson No. 1 Intro. to Operating System 203

 while (num == 0)
dataAvail->Wait(monitorLock);
}
put next data array value into v
num--;
bufferAvail->Signal(monitorLock);
monitorLock->Release(); /* Release monitor lock after perform operation */
}
}
8.2.5 Semaphores
E.W. Dijkstra (1965) abstracted the key notion of mutual exclusion in his
concepts of semaphores.
Definition
A semaphore is a protected variable whose value can be accessed and altered
only by the operations P and V and initialization operation called
'Semaphoiinitislize'.
Binary Semaphores can assume only the value 0 or the value 1, counting
semaphores also called general semaphores can assume only nonnegative
values.
The P (or wait or sleep or down) operation on semaphores S, written as P(S) or
wait (S), operates as follows:
P(S): IF S >0
THEN S = S – 1
ELSE (wait on S)
The V (or signal or wakeup or up) operation on semaphore S, written as V(S) or
signal (S), operates as follows:
V(S): IF (one or more process are waiting on S)
THEN (let one of these processes proceed)
ELSE S = S - 1
Operations P and V are done as single, indivisible, atomic action. It is guaranteed
that once a semaphore operation has stared, no other process can access the

Lesson No. 1 Intro. to Operating System 204

semaphore until operation has completed. Mutual exclusion on the semaphore,
S, is enforced within P(S) and V(S).
If several processes attempt a P(S) simultaneously, only one process will be
allowed to proceed. The other processes will be kept waiting, but the
implementation of P and V guarantees that processes will not suffer indefinite
postponement.
Semaphores solve the lost-wakeup problem.
8.2.5.1 Producer-Consumer Problem Using Semaphores
The Solution to producer-consumer problem uses three semaphores, namely,
full, empty and mutex.
The semaphore 'full' is used for counting the number of slots in the buffer that are
full. The 'empty' for counting the number of slots that are empty and semaphore
'mutex' to make sure that the producer and consumer do not access modifiable
shared section of the buffer simultaneously.
Initialization
• Set full buffer slots to 0 i.e., semaphore Full = 0.
• Set empty buffer slots to N i.e., semaphore empty = N.
• For control access to critical section set mutex to 1 i.e., semaphore mutex
= 1.
Producer ( )
WHILE (true)
produce-Item ( );
P (empty);
P (mutex);
enter-Item ( )
V (mutex)
V (full);
Consumer ( )
WHILE (true)
P (full)
P (mutex);

Lesson No. 1 Intro. to Operating System 205

 remove-Item ( );
V (mutex);
V (empty);
consume-Item (Item)
8.2.6 The dining philosophers problem
The dining philosophers problem is a “classical” synchronization problem. Taken
at face value, it is a pretty meaningless problem, but it is typical of many
synchronization problems that you will see when allocating resources in
operating systems.
The problem is defined as follows: There are 5 philosophers sitting at a round
table. Between each adjacent pair of philosophers is a chopstick. In other words,
there are five chopsticks. Each philosopher does two things: think and eat. The
philosopher thinks for a while, and then stops thinking and becomes hungry.
When the philosopher becomes hungry, he/she cannot eat until he/she owns the
chopsticks to his/her left and right. When the philosopher is done eating he/she
puts down the chopsticks and begins thinking again.
The challenge in the dining philosophers problem is to design a protocol so that
the philosophers do not deadlock (i.e. every philosopher has a chopstick), and so
that no philosopher starves (i.e. when a philosopher is hungry, he/she eventually
gets the chopsticks). Additionally, our protocol should try to be as efficient as
possible -- in other words, we should try to minimize the time that philosophers
spent waiting to eat.
A simple solution to this problem can be of the form
Repeat
Lift the left fork
Lift the right fork
{Eat}
{Think}
Forever
But this solution is not acceptable since it is prone to deadlock (If all the
philosophers lift their left fork).

Lesson No. 1 Intro. to Operating System 206

One solution is to order the forks and require the philosophers to pick up the
forks in increasing order, which mathematically eliminates the possibility of a
deadlock. To illustrate this solution, label the philosophers P1, P2, P3, P4, and P5,
and label the forks F1, F2, F3, F4, and F5. Each philosopher must pick up forks in
a prescribed order and cannot pick up a fork another philosopher already has.
Upon acquiring two forks, a philosopher may eat. Philosophers P1 through P4
follow the rule that Px must pick up fork Fx first and then may pick up fork Fx+1.
For example, P1 must pick up F1 first and F2 second. Philosopher P5 must,
conversely, pick up fork F1 before picking up fork F5, to respect the deadlock-
preventing fork ordering rule.
Although avoiding a deadlock, this solution is inefficient, because one can arrive
to a situation where only one philosopher is eating and everybody else is waiting
for him. For example philosophers P1 to P3 could hold forks F1 to F3, waiting to
get forks F2 to F4 respectively, philosopher P5 could wait on fork F1 having no fork
yet, while philosopher P4 would be eating holding forks F4 and F5. Optimally,
either philosopher P1 or philosopher P2 should be able to eat in such
circumstances.
Preventing starvation depends on the method of mutual exclusion enforcement
used. Implementations using spinlocks or busy waiting can cause starvation
through timing problems inherent in these methods. Other methods of mutual
exclusion that utilize queues can prevent starvation by enforcing equal access to
a fork by the adjacent philosophers.
8.3 Summary
In operating systems, concurrent processes share some common storage that

Lesson No. 1 Intro. to Operating System 207

each process can read and write. Since processes frequently need to
communicate with other processes therefore, there is a need for a well-structured
communication. Processes use two kinds of synchronization to control their
activities; Control synchronization, & Data access synchronization. A race
condition on a data item arises when many processes concurrently update its
value. One solution to race condition is critical section. If we could arrange
matters such that no two processes were ever in their critical sections
simultaneously, we could avoid race conditions. There are four conditions to hold
to have a good solution for the critical section problem (mutual exclusion). In the
approaches to mutual exclusion, if a process wants to enter in its critical section,
it checks to see if the entry is allowed. If it is not allowed to enter, it waits. This
approach waste CPU-time. To avoid this sleep and wakeup calls are used.
A semaphore is a protected variable whose value can be accessed and altered
only by the indivisible operations P and V and initialization operation called
'Semaphoiinitislize' Semaphores helps in avoiding the race condition
8.4 Keywords
Critical section: that part of the program where the shared memory is accessed.
Mutual Exclusion: each process executing the shared data excludes all others
from doing so simultaneously.
Semaphore: an object that hides an integer value and only allows three
operations: initialization to a specified value, increment, or decrement.
8.5 Self assessment questions
1. What do you understand by critical section? What are the charcteristic
properties of it? Explain.
2. What is mutual exclusion? Discuss the different approaches to solve the
problem of mutual exclusion.
3. What do you understand by semphores? Does it satisfy the bounded wait
condition? Explain.
4. What is semaphore? How does its help in avoiding the rce condition? Explain.
5. What do you understand by:
(a) Busy waiting

Lesson No. 1 Intro. to Operating System 208

 (b) Bounded wait
8.6 Suggested readings/rfernce material
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New
Delhi.
3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson
Education Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company
Ltd., New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 209

Lesson number: 9 Writer: Dr. Rakesh Kumar
Deadlocks Vetter: Prof. Dharminder Kr.

Crises and deadlocks when they occur have at least this advantage that
they force us to think.”- Jawaharlal Nehru (1889 - 1964)
9.0 Objectives
The objectives of this lesson are to make the students acquainted with the
problem of deadlocks. In this lesson, we characterize the problem of deadlocks
and discuss policies, which an OS can use to ensure their absence. Deadlock
detection, resolution, prevention and avoidance have been discussed in detail in
the present lesson.
After studying this lesson the students will be familiar with following:
(a) Condition for deadlock.
(b) Deadlock prevention
(c) Deadlock avoidance
(d) Deadlock detection and recovery
9.1 Introduction
If a process is in the need of some resource, physical or logical, it requests the
kernel of operating system. The kernel, being the resource manager, allocates
the resources to the processes. If there is a delay in the allocation of the
resource to the process, it results in the idling of process. The deadlock is a
situation in which some processes in the system faces indefinite delays in
resource allocation. In this lesson, we identify the problems causing deadlocks,
and discuss a number of policies used by the operating system to deal with the
problem of deadlocks.
9.2 Presentation of contents
9.2.1 Definition
9.2.2 Preemptable and Nonpreemptable Resources
9.2.3 Necessary and Sufficient Deadlock Conditions
9.2.4 Resource-Allocation Graph

Lesson No. 1 Intro. to Operating System 210

 9.2.4.1 Interpreting a Resource Allocation Graph with Single Resource
Instances
9.2.5 Dealing with Deadlock
9.2.6 Deadlock Prevention
9.2.6.1 Elimination of “Mutual Exclusion” Condition
9.2.6.2 Elimination of “Hold and Wait” Condition
9.2.6.3 Elimination of “No-preemption” Condition
9.2.6.4 Elimination of “Circular Wait” Condition
9.2.7 Deadlock Avoidance
9.2.7.1 Banker’s Algorithm
9.2.7.2 Evaluation of Deadlock Avoidance Using the Banker’s Algorithm
9.2.8 Deadlock Detection
9.2.9 Deadlock Recovery
9.2.10 Mixed approaches to deadlock handling
9.2.11 Evaluating the Approaches to Dealing with Deadlock
9.2.1 Definition
A deadlock involving a set of processes D is a situation in which:
(a) Every process Pi in D is blocked on some event Ei.
(b) Event Ei can be caused only by action of some process (es) in D.
A set of process is in a deadlock state if each process in the set is waiting for an
event that can be caused by only another process in the set. In other words,
each member of the set of deadlock processes is waiting for a resource that can
be released only by a deadlock process. None of the processes can run, none of
them can release any resources, and none of them can be awakened. It is
important to note that the number of processes and the number and kind of
resources possessed and requested are unimportant.
The resources may be either physical or logical. Examples of physical resources
are Printers, Tape Drivers, Memory Space, and CPU Cycles. Examples of logical
resources are Files, Semaphores, and Monitors.
The simplest example of deadlock is where process 1 has been allocated non-
shareable resources A, say, a tap drive, and process 2 has be allocated non-

Lesson No. 1 Intro. to Operating System 211

sharable resource B, say, a printer. Now, if it turns out that process 1 needs
resource B (printer) to proceed and process 2 needs resource A (the tape drive)
to proceed and these are the only two processes in the system, each is blocked
the other and all useful work in the system stops. This situation ifs termed
deadlock. The system is in deadlock state because each process holds a
resource being requested by the other process neither process is willing to
release the resource it holds.
9.2.2 Preemptable and Nonpreemptable Resources
Resources come in two flavors: preemptable and nonpreemptable. A
preemptable resource is one that can be taken away from the process with no ill
effects. Memory is an example of a preemptable resource. On the other hand, a
nonpreemptable resource is one that cannot be taken away from process
(without causing ill effect). For example, CD resources are not preemptable at an
arbitrary moment.
Reallocating resources can resolve deadlocks that involve preemptable
resources. Deadlocks that involve nonpreemptable resources are difficult to deal
with.
9.2.3 Necessary and Sufficient Deadlock Conditions
Coffman (1971) identified four (4) conditions that must hold simultaneously for
there to be a deadlock.
1. Mutual Exclusion Condition
The resources involved are non-shareable.
Explanation: At least one resource (thread) must be held in a non-shareable
mode, that is, only one process at a time claims exclusive control of the resource.
If another process requests that resource, the requesting process must be
delayed until the resource has been released.
2. Hold and Wait Condition
Requesting process hold already, resources while waiting for requested
resources.

Lesson No. 1 Intro. to Operating System 212

Explanation: There must exist a process that is holding a resource already
allocated to it while waiting for additional resource that are currently being held
by other processes.
3. No-Preemptive Condition
Resources already allocated to a process cannot be preempted.
Explanation: Resources cannot be removed from the processes are used to
completion or released voluntarily by the process holding it.
4. Circular Wait Condition
The processes in the system form a circular list or chain where each process in
the list is waiting for a resource held by the next process in the list.
Conditions 1 and 3 pertain to resource utilization policies, while condition 2
pertains to resource requirements of individual processes. Only condition 4
pertains to relationships between resource requirements of a group of processes.
As an example, consider the traffic deadlock in the following figure

Consider each section of the street as a resource.

Lesson No. 1 Intro. to Operating System 213

 1. Mutual exclusion condition applies, since only one vehicle can be on a
section of the street at a time.
2. Hold-and-wait condition applies, since each vehicle is occupying a section
of the street, and waiting to move on to the next section of the street.
3. No-preemptive condition applies, since a section of the street that is a
section of the street that is occupied by a vehicle cannot be taken away
from it.
4. Circular wait condition applies, since each vehicle is waiting on the next
vehicle to move. That is, each vehicle in the traffic is waiting for a section
of street held by the next vehicle in the traffic.
The simple rule to avoid traffic deadlock is that a vehicle should only enter an
intersection if it is assured that it will not have to stop inside the intersection.
It is not possible to have a deadlock involving only one single process. The
deadlock involves a circular “hold-and-wait” condition between two or more
processes, so “one” process cannot hold a resource, yet be waiting for another
resource that it is holding. In addition, deadlock is not possible between two
threads in a process, because it is the process that holds resources, not the
thread that is, each thread has access to the resources held by the process.
9.2.4 Resource-Allocation Graph
The deadlock conditions can be modeled using a directed graph called a
resource allocation graph (RAG). A resource allocation graph is a directed graph.
It consists of 2 kinds of nodes:
Boxes — Boxes represent resources, and Instances of the resource are
represented as dots within the box i.e. how many units of that resource exist in
the system.
Circles — Circles represent threads / processes. They may be a user process or
a system process.
An edge can exist only between a process node and a resource node. There are
2 kinds of (directed) edges:

Lesson No. 1 Intro. to Operating System 214

Request edge: It represents resource request. It starts from process and
terminates to a resource. It indicates the process has requested the resource,
and is waiting to acquire it.
Assignment edge: It represents resource allocation. It starts from resource
instance and terminates to process. It indicates the process is holding the
resource instance.
When a request is made, a request edge is added.
When request is fulfilled, the request edge is transformed into an assignment
edge.
When process releases the resource, the assignment edge is deleted.
9.2.4.1 Interpreting a Resource Allocation Graph with Single Resource
Instances
Following figure shows a resource allocation graph. If the graph does not contain
a cycle, then no deadlock exists. Following figure is an example of a no deadlock
situation.

R1 R2
O O

P1 P2 P3

O O
R3 R4

If the graph does contain a cycle, then a deadlock does exist. As following
resource allocation graph depicts a deadlock situation.

Lesson No. 1 Intro. to Operating System 215

 R1 R2
O O

P1 P2 P3

O O
R3 R4

With single resource instances, a cycle is a necessary and sufficient condition

for deadlock
9.2.5 Dealing with Deadlock
There are following approaches to deal with the problem of deadlock.
The Ostrich Approach — sticks your head in the sand and ignores the problem.
This approach can be quite useful if you believe that they are rarest chances of
deadlock occurrence. In that situation it is not a justifiable proposition to invest a
lot in identifying deadlocks and tackling with it. Rather a better option is ignore it.
Deadlock prevention: This approach prevents deadlock from occurring by
eliminating one of the four (4) deadlock conditions.
Deadlock detection algorithms: This approach detects when deadlock has
occurred.
Deadlock recovery algorithms: After detecting the deadlock, it breaks the
deadlock.
Deadlock avoidance algorithms: This approach considers resources currently
available, resources allocated to each thread, and possible future requests, and
only fulfill requests that will not lead to deadlock
9.2.6 Deadlock Prevention
Deadlock prevention is based on designing resource allocation policies, which
make deadlocks impossible. Use of the deadlock prevention approach avoids the
over- head of deadlock detection and resolution. However, it incurs two kinds of
costs - overhead of using the resource allocation policy, and cost of resource
idling due to the policy.
As described in earlier section, four conditions must hold for a resource deadlock

Lesson No. 1 Intro. to Operating System 216

to arise in a system:
¾ Non-shareable resources
¾ Hold-and-wait by processes
¾ No preemption of resources
¾ Circular waits.
Ensuring that one of these conditions cannot be satisfied prevents deadlocks.
We first discuss how each of these conditions can be prevented and then discuss
a couple of resource allocation policies based on the prevention approach.
Havender in his pioneering work showed that since all four of the conditions are
necessary for deadlock to occur, it follows that deadlock might be prevented by
denying any one of the conditions.
9.2.6.1 Elimination of “Mutual Exclusion” Condition
The mutual exclusion condition must hold for non-sharable resources. That is,
several processes cannot simultaneously share a single resource. This condition
is difficult to eliminate because some resources, such as the tap drive and
printer, are inherently non-shareable. Note that shareable resources like read-
only-file do not require mutually exclusive access and thus cannot be involved in
deadlock.
9.2.6.2 Elimination of “Hold and Wait” Condition
There are two possibilities for elimination of the second condition. The first
alternative is that a process request be granted all of the resources it needs at
once, prior to execution. The second alternative is to disallow a process from
requesting resources whenever it has previously allocated resources. This
strategy requires that all of the resources a process will need must be requested
at once. The system must grant resources on “all or none” basis. If the complete
set of resources needed by a process is not currently available, then the process
must wait until the complete set is available. While the process waits, however, it
may not hold any resources. Thus the “wait for” condition is denied and
deadlocks simply cannot occur. This strategy can lead to serious waste of
resources. For example, a program requiring ten tap drives must request and
receive all ten derives before it begins executing. If the program needs only one

Lesson No. 1 Intro. to Operating System 217

tap drive to begin execution and then does not need the remaining tap drives for
several hours. Then substantial computer resources (9 tape drives) will sit idle for
several hours. This strategy can cause indefinite postponement (starvation).
Since not all the required resources may become available at once.
9.2.6.3 Elimination of “No-preemption” Condition
The nonpreemption condition can be alleviated by forcing a process waiting for a
resource that cannot immediately be allocated to relinquish all of its currently
held resources, so that other processes may use them to finish. Suppose a
system does allow processes to hold resources while requesting additional
resources. Consider what happens when a request cannot be satisfied. A
process holds resources a second process may need in order to proceed while
second process may hold the resources needed by the first process. This is a
deadlock. This strategy requires that when a process that is holding some
resources is denied a request for additional resources. The process must release
its held resources and, if necessary, request them again together with additional
resources. Implementation of this strategy denies the “no-preemptive” condition
effectively.
The main drawback of this approach is high cost. When a process releases
resources the process may lose all its work to that point. One serious
consequence of this strategy is the possibility of indefinite postponement
(starvation). A process might be held off indefinitely as it repeatedly requests and
releases the same resources.
9.2.6.4 Elimination of “Circular Wait” Condition
Presence of a cycle in resource allocation graph indicates the “circular wait”
condition. The last condition, the circular wait, can be denied by imposing a total
ordering on all of the resource types and than forcing, all processes to request
the resources in order (increasing or decreasing). This strategy impose a total
ordering of all resources types, and to require that each process requests
resources in a numerical order (increasing or decreasing) of enumeration. With
this rule, the resource allocation graph can never have a cycle.
For example, provide a global numbering of all the resources, as shown

Lesson No. 1 Intro. to Operating System 218

 Card
1 ≡
reader
2 ≡ Printer
3 ≡ Plotter
4 ≡ Tape drive
Card
5 ≡
punch
Now the rule is this: processes can request resources whenever they want to, but
all requests must be made in numerical order. A process may request first printer
and then a tape drive (order: 2, 4), but it may not request first a plotter and then a
printer (order: 3, 2). The problem with this strategy is that it may be impossible to
find an ordering that satisfies everyone. The resource ranking policy works best
when all processes require their resources in the order of increasing ranks.
However, difficulty arises when a process requires resources in some other
order. Now processes may tend to circumvent such difficulties by acquiring lower
ranking resources much before they are actually needed. In the worst case this
policy may degenerate into the ‘all requests together’ policy of resource
allocation. Anyway this policy is attractive due to its simplicity once resource
ranks have been assigned.
“All requests together” is the simplest of all deadlock prevention policies. A
process must make its resource requests together-typically, at the start of its
execution. This restriction permits a process to make only one multiple request in
its lifetime. Since resources requested in a multiple request are allocated
together, a blocked process does not hold any resources. The hold-and-wait
condition is satisfied. Hence paths of length larger than 1 cannot exist in the
Resource Allocation Graph, a mutual wait-for relationships cannot develop in the
system. Thus, deadlocks cannot arise.
9.2.7 Deadlock Avoidance
This approach to the deadlock problem anticipates deadlock before it actually
occurs. This approach employs an algorithm to access the possibility that
deadlock could occur and acting accordingly. This method differs from deadlock

Lesson No. 1 Intro. to Operating System 219

prevention, which guarantees that deadlock cannot occur by denying one of the
necessary conditions of deadlock.
If the necessary conditions for a deadlock are in place, it is still possible to avoid
deadlock by being careful when resources are allocated. Perhaps the most
famous deadlock avoidance algorithm, due to Dijkstra [1965], is the Banker’s
algorithm. So named because the process is analogous to that used by a banker
in deciding if a loan can be safely made.
9.2.7.1 Banker’s Algorithm
In this analogy
Customers≡ processes
resources,
Units ≡
say, tape drive
Operating
Banker ≡
System

Customers Used Max

A 0 6
B 0 5 Available
C 0 4 Units = 10
D 0 7

In the above figure, we see four customers each of whom has been granted a
number of credit units. The banker reserved only 10 units rather than 22 units to
service them. At certain moment, the situation becomes
Customers Used Max
A 1 6
B 1 5 Available
C 2 4 Units = 2
D 4 7
Safe State The key to a state being safe is that there is at least one way for all
users to finish. In other analogy, the state of figure 2 is safe because with 2 units

Lesson No. 1 Intro. to Operating System 220

left, the banker can delay any request except C's, thus letting C finish and
release all four resources. With four units in hand, the banker can let either D or
B have the necessary units and so on.
Unsafe State Consider what would happen if a request from B for one more
unit were granted in above figure 2.
We would have following situation
Customers Used Max
A 1 6
B 2 5 Available
C 2 4 Units = 1
D 4 7
This is an unsafe state.
If all the customers namely A, B, C, and D asked for their maximum loans, then
banker could not satisfy any of them and we would have a deadlock.
Important Note: It is important to note that an unsafe state does not imply the
existence or even the eventual existence a deadlock. What an unsafe state does
imply is simply that some unfortunate sequence of events might lead to a
deadlock.
The Banker's algorithm is thus to consider each request as it occurs, and see if
granting it leads to a safe state. If it does, the request is granted, otherwise, it
postponed until later. Haberman [1969] has shown that executing of the
algorithm has complexity proportional to N2 where N is the number of processes
and since the algorithm is executed each time a resource request occurs, the
overhead is significant.
9.2.7.2 Evaluation of Deadlock Avoidance Using the Banker’s Algorithm
There are following advantages and disadvantages of deadlock avoidance using
Banker’s algorithm.
Advantages:
¾ There is no need to preempt resources and rollback state (as in deadlock
detection & recovery)
¾ It is less restrictive than deadlock prevention

Lesson No. 1 Intro. to Operating System 221

Disadvantages:
¾ In this case maximum resource requirement for each process must be
stated in advance.
¾ Processes being considered must be independent (i.e., unconstrained by
synchronization requirements)
¾ There must be a fixed number of resources (i.e., can’t add resources,
resources can’t break) and processes (i.e., can’t add or delete processes)
¾ Huge overhead — Operating system must use the algorithm every time a
resource is requested. So a huge overhead is involved.
9.2.8 Deadlock Detection
Deadlock detection is the process of actually determining that a deadlock exists
and identifying the processes and resources involved in the deadlock.
The basic idea is to check allocation against resource availability for all possible
allocation sequences to determine if the system is in deadlocked state a. Of
course, the deadlock detection algorithm is only half of this strategy. Once a
deadlock is detected, there needs to be a way to recover several alternatives
exists:
• Temporarily prevent resources from deadlocked processes.
• Back off a process to some check point allowing preemption of a needed
resource and restarting the process at the checkpoint later.
• Successively kill processes until the system is deadlock free.
These methods are expensive in the sense that each iteration calls the detection
algorithm until the system proves to be deadlock free. The complexity of
algorithm is O(N2) where N is the number of proceeds. Another potential problem
is starvation; same process killed repeatedly.
9.2.9 Deadlock Recovery
Once you've discovered that there is a deadlock, what do you do about it? One
thing to do is simply re-boot. A less drastic approach is to yank back a resource
from a process to break a cycle. As we saw, if there are no cycles, there is no
deadlock. If the resource is not preemptable, snatching it back from a process

Lesson No. 1 Intro. to Operating System 222

may do irreparable harm to the process. It may be necessary to kill the process,
under the principle that at least that's better than crashing the whole system.
Sometimes, we can do better. For example, if we checkpoint a process from time
to time, we can roll it back to the latest checkpoint, hopefully to a time before it
grabbed the resource in question. Database systems use checkpoints, as well as
a technique called logging, allowing them to run processes “backwards,” undoing
everything they have done. It works like this: Each time the process performs an
action, it writes a log record containing enough information to undo the action.
For example, if the action is to assign a value to a variable, the log record
contains the previous value of the record. When a database discovers a
deadlock, it picks a victim and rolls it back.
Rolling back processes involved in deadlocks can lead to a form of starvation, if
we always choose the same victim. We can avoid this problem by always
choosing the youngest process in a cycle. After being rolled back enough times,
a process will grow old enough that it never gets chosen as the victim--at worst
by the time it is the oldest process in the system. If deadlock recovery involves
killing a process altogether and restarting it, it is important to mark the “starting
time” of the reincarnated process as being that of its original version, so that it
will look older that new processes started since then.
When should you check for deadlock? There is no one best answer to this
question; it depends on the situation. The most “eager” approach is to check
whenever we do something that might create a deadlock. Since a process cannot
create a deadlock when releasing resources, we only have to check on allocation
requests. If the OS always grants requests as soon as possible, a successful
request also cannot create a deadlock. Thus we only have to check for a
deadlock when a process becomes blocked because it made a request that
cannot be immediately granted. However, even that may be too frequent. As we
saw, the deadlock-detection algorithm can be quite expensive if there are a lot of
processes and resources, and if deadlock is rare, we can waste a lot of time
checking for deadlock every time a request has to be blocked.

Lesson No. 1 Intro. to Operating System 223

What's the cost of delaying detection of deadlock? One possible cost is poor
CPU utilization. In an extreme case, if all processes are involved in a deadlock,
the CPU will be completely idle. Even if there are some processes that are not
deadlocked, they may all be blocked for other reasons (e.g. waiting for I/O). Thus
if CPU utilization drops, that might be a sign that it's time to check for deadlock.
Besides, if the CPU isn't being used for other things, you might as well use it to
check for deadlock!
On the other hand, there might be a deadlock, but enough non-deadlocked
processes to keep the system busy. Things look fine from the point of view of the
OS, but from the selfish point of view of the deadlocked processes, things are
definitely not fine. If the processes may represent interactive users, who can't
understand why they are getting no response. Worse still, they may represent
time-critical processes (missile defense, factory control, hospital intensive care
monitoring, etc.) where something disastrous can happen if the deadlock is not
detected and corrected quickly. Thus another reason to check for deadlock is
that a process has been blocked on a resource request “too long.” The definition
of “too long” can vary widely from process to process. It depends both on how
long the process can reasonably expect to wait for the request, and how urgent
the response is. If an overnight run deadlocks at 11pm and nobody is going to
look at its output until 9am the next day, it doesn't matter whether the deadlock is
detected at 11:01pm or 8:59am. If all the processes in a system are sufficiently
similar, it may be adequate simply to check for deadlock at periodic intervals
(e.g., one every 5 minutes in a batch system; once every millisecond in a real-
time control system).
9.2.10 Mixed approaches to deadlock handling
The deadlock handling approaches differ in terms of theirv usage implications.
Hence it is not possible to use a single deadlock handling approach to govern the
allocation of all resources. The following mixed approach is found useful:
1. System control block: Control blocks like JCB, PCB etc. can be acquired
in a specific order. Hence resource ranking can be used here. If a simpler

Lesson No. 1 Intro. to Operating System 224

 strategy is desired, all control blocks for a job or process can be allocated
together at its initiation.
2. I/O devices files: Avoidance is the only practical strategy for these
resources. However, in order to eliminate the overheads of avoidance,
new devices are added as and when needed. This is done using the
concept of spooling. If a system has only one printer, many printers are
created by using some disk area to store a file to be printed. Actual
printing takes place when a printer becomes available.
3. Main memory: No deadlock handling is explicitly necessary. The memory
allocated to a program is simply preempted by swapping out the program
whenever the memory is needed for another program.
9.2.11 Evaluating the Approaches to Dealing with Deadlock
¾ The Ostrich Approach — ignoring the problem
It is a good solution if deadlock is not frequent.
Deadlock prevention — eliminating one of the four (4) deadlock conditions
This approach may be overly restrictive and results into the under utilization of
the resources.
¾ Deadlock detection and recovery — detect when deadlock has occurred, then
break the deadlock
In it there is a tradeoff between frequency of detection and performance /
overhead added.
¾ Deadlock avoidance — only fulfilling requests that will not lead to deadlock
It needs too much a priori information and not very dynamic (can’t add processes
or resources), and involves huge overhead
9.3 Summary
¾ A set of process is in a deadlock state if each process in the set is waiting for
an event that can be caused by only another process in the set. Processes
compete for physical and logical resources in the system. Deadlock affects
the progress of processes by causing indefinite delays in resource allocation.
¾ There are four Necessary and Sufficient Deadlock Conditions (1) Mutual Exclusion
Condition: The resources involved are non-shareable, (2) Hold and Wait Condition:

Lesson No. 1 Intro. to Operating System 225

 Requesting process hold already, resources while waiting for requested resources,(3)
No-Preemptive Condition: Resources already allocated to a process cannot be
preempted,(4) Circular Wait Condition: The processes in the system form a circular
list or chain where each process in the list is waiting for a resource held by the next
process in the list.
¾ The deadlock conditions can be modeled using a directed graph called a
resource allocation graph (RAG) consisting of boxes (resource), circles
(process) and edges (request edge and assignment edge). The resource
allocation graph helps in identifying the deadlocks.
¾ There are following approaches to deal with the problem of deadlock: (1) The
Ostrich Approach — stick your head in the sand and ignore the problem, (2)
Deadlock prevention — prevent deadlock from occurring by eliminating one of
the 4 deadlock conditions, (3) Deadlock detection algorithms — detect when
deadlock has occurred, (4) Deadlock recovery algorithms — break the
deadlock, (5) Deadlock avoidance algorithms — consider resources currently
available, resources allocated to each thread, and possible future requests,
and only fulfill requests that will not lead to deadlock
¾ There are merits/demerits of each approach. The Ostrich Approach is a good
solution if deadlock is not frequent. Deadlock prevention may be overly
restrictive. In Deadlock detection and recovery there is a tradeoff between
frequency of detection and performance / overhead added, Deadlock
avoidance needs too much a priori information and not very dynamic (can’t
add processes or resources), and involves huge overhead
9.4 Keywords
Deadlock: A deadlock is a situation in which some processes in the system face
indefinite delays in resource allocation.
Preemptable resource: A preemptable resource is one that can be taken away
from the process with no ill effects.
Nonpreemptable resource: It is one that cannot be taken away from process
(without causing ill effect).

Lesson No. 1 Intro. to Operating System 226

Mutual exclusion: several processes cannot simultaneously share a single
resource
9.5 SELF-ASSESMENT QUESTIONS (SAQ)
1. What do you understand by deadlock? What are the necessary conditions
for deadlock?
2. What do you understand by resource allocation graph (RAG)? Explain
using suitable examples, how can you use it to detect the deadlock?
3. Compare and contrast the following policies of resource allocation:
(a) All resources requests together.
(b) Allocation using resource ranking.
(c) Allocation using Banker’s algorithm
On the basis of (a) resource idling and (b) overhead of the resource allocation
algorithm.
4. How can pre-emption be used to resolve deadlock?
5. Why Banker’s algorithm is called so?
6. Under what condition(s) a wait state becomes a deadlock?
7. Explain how mutual exclusion prevents deadlock.
8. Discuss the merits and demerits of each approach dealing with the
problem of deadlock.
9. Differentiate between deadlock avoidance and deadlock prevention.
10. A system contains 6 units of a resource, and 3 processes that need to use
this resource. If the maximum resource requirement of each process is 3
units, will the system be free of deadlocks for all time? Explain clearly.
If the system had 7 units of the resource, would the system be deadlock-
free?
9.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
2. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.

Lesson No. 1 Intro. to Operating System 227

3. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
4. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.
5. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company
Ltd., New Delhi, 2002.
6. Operating Systems, A Concept-based Approach, Dhamdhere D.M., Tata
McGraw Hill Publishing Company Ltd., New Delhi.

Lesson No. 1 Intro. to Operating System 228

Lesson Number: 10 Writer: Dr. Rakesh Kumar
Case Study of UNIX Vetter: Prof. Dharminder Kr.

10.0 Objectives
The objective of this lesson is
(A) To give an overview of the important features of UNIX operating system to
the students.
(B) To make the students familiar with some important UNIX commands.
10.1 Introduction
UNIX is written in a high-level language giving it the benefit
of machine independence, portability, understandability, and
modifiability. Multi-tasking (more than one program can be
made to run at the same time) and multi-user (more than one
user can work at the same computer system at the same time)
are the two most important characteristics of UNIX helping it
in gaining widespread acceptance among a large variety of
users. It was the first operating system to bring in the
concept of hierarchical file structure. It uses a uniform
format for files called the byte stream making the application
programs to be written easily. UNIX treats every file as a
stream of bytes so the user can manipulate his file in the
manner he wants. It provides primitives that allow more
complex and complicated programs to be built from the simpler
ones. It provides very simple user interface both character-
based and graphical based. It hides the machine architecture
from the user. This helps the programmer to write different
programs that can be made to run on different hardware
configurations. It provides a simple, uniform interface to
peripheral devices.
10.2 Presentation of contents

Lesson No. 1 Intro. to Operating System 229

10.2.1 Versions
10.2.2 UNIX Architecture
10.2.3 Features of UNIX
10.2.3.1 Portability
10.2.3.2 Machine Independent
10.2.3.3 Multi-user Capability
10.2.3.4 Multitasking Capability
10.2.3.5 Software Development Tools
10.2.3.6 Built-in Networking
10.2.3.7 Security
10.2.4 Implementation of Operating System Functions
10.2.4.1 Process management functions
10.2.4.2 Memory Management
10.2.4.3 Device and File functions
10.2.5 UNIX Kernel
10.2.5.1 Assumptions about Hardware
10.2.5.2 Interrupts and Exceptions
10.2.5.3 Processor Execution Levels
10.2.6 File System and Internal Structure of Files
10.2.6.1 Representation of Data in a File
10.2.6.2 Directories
10.2.6.3 Blocks and Fragments
10.2.7 UNIX Shell
10.2.8 User Interaction with UNIX Operating System
10.2.8.1 Steps to Login
10.2.8.2 Changing your Password
10.2.8.3 UNIX Command Structure

10.2.9 Common UNIX Commands

10.2.10 File System, Permissions Changing Order and Group
10.2.11 UNIX Editors

Lesson No. 1 Intro. to Operating System 230

10.2.1 Versions
Some popular versions of UNIX are AIX IBM, XENIX, ULTRIX,
Sun OS, and BSD. The original version of UNIX came in the
late 1960's designed by Ken Thompson at AT&T Bell
Laboratories. At that time, Bell Labs were busy in designing
an operating system called Multics with an objective to
provide a very sophisticated and complex multi-user operating
system that had support for many advanced features. However,
Multics failed because the state of art provided by it at that
time was too complex. So, Bell Labs had to withdraw themselves
from the Multics project. Ken Thompson then started working on
a simpler project and he named it UNIX. Dennis Ritchie rewrote
the source code of UNIX operating system in C language.
By the year 1977, UNIX I system found its major contribution in the telephone
companies, providing a good environment for program development, network
transaction services and real time services. A large number of institutions and
universities were provided licenses of UNIX system. In the year 1977, the UNIX
system was first ported from a PDP to a non-PDP machine.
As the popularity of UNIX grew, many other companies came out with their own
versions of UNIX and ported it onto other new machines. From the year 1977 to
1982, Bell Laboratories combined many AT & T variants into a single system and
gave it a name UNIX System III. Bell Laboratories in this version brought out many
new features and advancements. It was given the name UNIX System V. The
people at University of California at Berkeley developed a variant to the UNIX
System. Its recent version is called 4.3 BSD for VAX machines.
By the beginning of 1984, UNIX system was installed at about 1,00,000 different
computer sites. It ran on a wide range of computers ranging from a mini computer to
a mainframe.
10.2.2 UNIX Architecture
The high level architecture of the UNIX system is shown in Figure 10.1.

Lesson No. 1 Intro. to Operating System 231

 Figure 10.1 System Architecture of UNIX
The UNIX system seems to be organized as a set of layers. The Kernel surrounds
the hardware. The user programs are independent of the hardware on which they
are running. The programs such as the shell and editors interact with the Kernel by
invoking a well-defined set of system calls. The system calls get various actions
done from the Kernel for the calling program. They interchange data between the
Kernel and the program. There are many other programs in this layer which from a
part of the standard system configurations. These programs are known as
commands. There are several other user created programs present in the same
layer. The outer most layer contains other application programs which can be build
on top of lower level programs. For instance, the C compiler appears in the
outermost layer of the figure. It invokes a C preprocessor, compiler, assembler and
link loader. These are all separate lower level programs. The programming style
offered by the UNIX system helps us to fulfill a task by combining the existing
programs.
10.2.3 Features of UNIX
The popularity of UNIX is due to the following reasons:
10.2.3.1 Portability
UNIX is its portable i.e. it can run successfully on all types of computers. The reason
of this is that it is written in a high-level language. PCs, Macintoshes, Workstations,
Minicomputers, Super Computers and Mainframes run the UNIX operating system
with equal ease.

Lesson No. 1 Intro. to Operating System 232

10.2.3.2 Machine Independent
The UNIX system does not expose the machine architecture to the user. Thus, it
becomes very easy to write applications that can run on micros, minis or
mainframes.
10.2.3.3 Multi-user Capability
UNIX is a multi-user system in which the same computer resources like hard disk,
memory etc can be used or accessed by many users simultaneously. Each user is
given a terminal. Each terminal is an input and an output device for the user. All the
terminals are connected to the main computer. So, a user sitting at any terminal can
not only uses the data or the software of the main computer but also the peripherals
like printers attached to it. The main computer is called the server or the console.
The number of terminals that can be connected to the server depends upon the
number of parts present in the controller card.
10.2.3.4 Multitasking Capability
UNIX has the facility to carry out more than one job at the same time. Multitasking is
achieved by dividing the CPU time in the order of milliseconds/microseconds for
execution between all the jobs that are being carried out. Each job is carried out
according to its priority number. It gives the impression that the tasks are being
carried out simultaneously.
10.2.3.5 Software Development Tools
UNIX offers an excellent environment for developing new software. It provides a
variety of tools ranging from editing a program to maintenance of software. It
exploits the power of hardware to the maximum extent of effectiveness and
efficiency.
10.2.3.6 Built-in Networking
UNIX has got built in networking support with a large number of programs and
utilities. It also offers an excellent media for communication with other users. The
users have the liberty of exchanging mail, data, programs, etc. You can send your
data at any place irrespective of the distance over a computer network.
10.2.3.7 Security
UNIX enforces security at three levels.

Lesson No. 1 Intro. to Operating System 233

(a) Each user is assigned a login name and a password. So, only the valid users
can have access to the files and directories.
(b) Each file is bound around permissions (read, write, execute). The file
permissions decide who can read/modify/execute a particular file. The
permissions once decided for a file can also be changed from time to time.
(c) Then file encryption comes into picture. It encodes file in a format that cannot be
very easily read. So, if anybody happens to open file, even then he will not be
able to read the text of the file. However, you can decode the file for reading its
contents.
10.2.4 Implementation of Operating System Functions
UNIX operating system performs following designated
functions:
(a) Process management functions: creating, destroying and manipulating
processes.
(b) Memory management functions: allocating, de-allocating and manipulating
memory.
(c) Input/Output functions: communicating and controlling I/O device and file
system.
(d) Miscellaneous functions: Network functions etc.
The UNIX System V offers somewhere around 64 system calls, which carry very
simple options with them. So, it becomes easy to make use of these system calls.
The body of the Kernel is formed by the set of system calls and the internal
algorithms that implement them. Kernel provides all the services to the application
programs in the UNIX system. In UNIX, the programs don't have any knowledge of
the internal format in which the Kernel stores file data.
10.2.4.1 Process management functions
The behavior of a UNIX process is defined by its text segment, data segment and
stack segment as shown in Figure 10.2.

Lesson No. 1 Intro. to Operating System 234

 TEXT DATA STACK

PSW
FILES RESOURCE

UNIX KERNEL
Figure 10.2: A Process in UNIX
The text segment contains the compiled object instructions, the data segment
contains static variables, and the stack segment holds the runtime stack used to
store temporary variables. A set of source file that is compiled and linked-into an
executable form is stored in a file with the default name of a. out. If the program
references statically define data, such as C static variables, a template for the data
segment is maintained in the executable file. The data segment will be created and
initialized to contain values and space for variables when the executable file is
loaded and executed. The stack segment is used to allocate storage for dynamic
elements of the program, such as automatic C variables that are created when they
come into scope and are destroyed when pass out of scope.
The compiler and linker create the executable file. These utilities do not define a
process; they define only the program text and a template for the data component
that the process will use when it executes the program. When the loader loads a
program into the computer's memory, the system creates appropriate data and stack
segments, called a process.
A process has a unique process identifier (PID), a pointer to a table of process
descriptors used by the UNIX OS kernel to reference the process's descriptor.
Whenever one process references another process in a system call, it provides the
pointer of the target process. The UNIX pa command lists each pm associated with
the user executing the command. The pm of each process appears as a field in the
descriptor of each process.
UNIX command for creating a new process is the fork system call. Whenever a
process calls fork, a child process is created with its descriptor, including its own

Lesson No. 1 Intro. to Operating System 235

copies of the parent's program text, data, and segments, and access to all open file
descriptors (in the kernel). The child and parent processes execute in their own
separate address spaces. This means that even though they have access to the
same information, both the child and its parent each reference their own copy of the
data. No part of the address space of either process is shared.
Hence, the parent and child cannot communicate by referencing variables stored at
the same address in their respective address space. UNIX systems also provide
several forms of the execve system call to enable a process to reload its address
space with a different program:
execve (char *path, char *avgv[], char *envp[] );
This system call causes the load module stored in the file at path to replace the
program currently being executed by the process. After execve has completed
executing, the program that called it, is no longer loaded. Hence, there is no notion
of returning from an execve call, since the calling program is no longer loaded in
memory. When the new program is started, it is passed the argument list, angv, and
the process uses a new set of environment variables, envp.
UNIX also provides a system call, wait (and a variant, waitpid), to enable a parent
process to detect when one of its child processes terminates. Details of the
terminating child's status may be either returned to the parent via a value parameter
passed to wait or ignored by the parent. The waitpid allows the parent to wait for a
particular child process (based on its PID) to terminate, while the wait command
does not discriminate among child processes. When a process exits, its resources,
including the kernel process descriptor, are released. The operating system signals
the parent that the child has died, but it will not release the process descriptor until
the parent has received the signal. The parent executes the wait call to
acknowledge the termination of a child process.
Following are the rest of UNIX system calls related to
process management:
¾ acct enable/disable process accounting alarm set a process alarm clock exit
terminate a process fork create a new process
¾ getpid get process, process group and parent process ID

Lesson No. 1 Intro. to Operating System 236

¾ getuid get real user, effective user real group and effective group ID
¾ kill send a signal to a process or group of processes
¾ msgctl message control operation msgop message operation
¾ nice change priority of a process pause suspend until
¾ pipe create an inter-process channel
¾ profil execution time profile ptrace process trace
¾ semctl semaphor control operations
¾ semget get set of semaphor
¾ semop semaphor operations
¾ setpgrp: set process group ID
¾ setuid set group and user ID
¾ signal specify what to do when a signal is received
¾ stime set time
¾ sync update super block time get time
¾ times get process and child process times
¾ ulimit get user upper limits
¾ uname get the name of the current operating system
¾ ulink remove directory entry
¾ Wait wait for the child process to stop or terminate
10.2.4.2 Memory Management
Kernel resides in the main memory so long as computer is operational. When a
program is compiled, a set of addresses is generated in the program by the
compiler. These represent addresses of variables or addresses of instructions such
as functions. The addresses generated by the compiler are for a virtual machine.
The addresses are not absolute in terms of memory addresses where they will be
loaded eventually. It assumes that no other program will be executing concurrently.
However, when you run the program the Kernel allocates some space to it in the
main memory. But the virtual addresses generated by the compiler might not
resemble the physical addresses occupied in the machine. Then the Kernel maps
the compiler-generated address with the physical machine addresses.

Lesson No. 1 Intro. to Operating System 237

UNIX divides the available memory into system memory and application memory. It
loads itself into system memory and creates data structures it will use in its
operation into this area of memory. The application memory area contains the user's
programs. The application memory area "is divided into global stack, local / stack
and heap. The global and static type of variables, functions etc. are assigned space
in these memory areas. UNIX provides system calls to affect loading and unloading
of programs into and out of the processes.
¾ Internally UNIX uses paging with segmentation methods to manage memory. In
addition to these primitive operations, UNIX provides library functions like malloc,
for allocating memory to a process and objects dynamically. Other memory
related system calls are:
¾ brk change data segment space allocation
¾ shmop shared memory operations
¾ shmctl shared memory control operation
¾ shmget get shared memory segment
¾ plock lock process, text or data memory addresses,
¾ msgget get message queue
10.2.4.3 Device and File functions
For each device it has device drivers for low-level communication. UNIX treats every
device the same way as it treats the files. Device drivers are intended to be
accessed by user space code. If an application accesses a driver, it uses one of two
standardized interfaces: the block device interface or the character device interface.
Both interfaces provide a fixed set of functions to the user programs.
When a user program calls the driver, it performs a system call. The kernel searches
the entry point for the device in the block or character in direct reference table (the
jump table) and then calls the entry point. The exact semantics of each function
depends on the nature of the device and the intent of the driver design. Hence, the
function names suggest only a purpose for each. The logical contents of the jump
table are kept in the file system in the dev directory.
A Unix driver has three parts:
¾ Code to initiate derive operations

Lesson No. 1 Intro. to Operating System 238

¾ Device interrupt handlers
The initialisation code is run when the system is booted or
started first time. It tests for the physical presence of
respective devices and then initializes them. The API
implements functions for a subset of the entry points. This
part of the code also provides information to the kernel as to
which functions are implemented. The system interrupt handler
that corresponds to the physical device causing the interrupt
calls the device interrupt handler.
System administrators are responsible for installing devices and drivers. The
information necessary to install a driver can be incorporated into a configuration file
by the administrator and then processed by the configuration builder tool /etc/conf.
UNIX has system calls to effect I/O manipulation. Some of them are:
¾ write - write on a file
¾ utime - set file access and modification times
¾ fstat - get file system statistics
¾ ulink - remove directory entry
¾ umount - unmount a file system
¾ umask - set and get file creation mask
¾ stat - get file status
¾ read - read from a file
¾ open - open for reading or writing h
¾ mount - mount a file system
¾ mknod - make a directory or special or ordinary file
¾ lseek - move read/write pointer link link to a file
¾ bfcntl - file control
¾ nexec - execute a file
¾ lytodup - duplicate an open file descriptor
¾ dbycreat - create a new file or rewrite an existing one
¾ the close close a file descriptor let of chroot change to root
¾ chown - change owner or group of file entry

Lesson No. 1 Intro. to Operating System 239

¾ chmod - change mode of file
¾ chdir - change directory device access determine accessibility of a file
10.2.5 UNIX Kernel
The services provided by the Kernel are given below:
1. It controls the fate and state of various processes such as their creation,
termination and I/O suspension.
2. The Kernel allocates main memory for an executing process. The Kernel allows
the processes to share portions of their address space. It keeps the private
space of processes secure and doesn't allow tampering from other processes.
However, if the free memory is low with the system, then the Kernel frees out
some memory by writing a process temporarily to secondary memory. In case
the Kernel writes all the processes to the secondary memory, it is called a
swapping system. However, if only the pages of memory are written onto the
secondary memory, then it is called the paging system.
3. The Kernel schedules processes for execution on the CPU. The time-sharing
concept allows the processes to share the CPU. When the time of a process
has finished, the Kernel suspends it and puts some other ready process for
execution in the CPU. It is again the work of the Kernel to reschedule the
suspended process.
4. The Kernel permits different processes to make use of the peripheral devices
such as terminals, tape drives, disk drives and network devices as and when
requested.
5. The Kernel allocates the secondary memory for efficient storage and retrieval of
user data. The Kernel allocates secondary storage for user files, organizes the
file system in a well-planned manner and provides security to user files from
illegal access.
6. The services provided by the Kernel are absolutely transparent to the user. For
instance, the Kernel formats the data present in a file for internal storage.
However, it hides the internal format from user processes. Similarly, it makes a
distinction between the regular file or a device but hides the distinction from
user processes. Finally, the Kernel provides the services so that the user level

Lesson No. 1 Intro. to Operating System 240

 processes can support the services they must provide. For instance, the Kernel
provides the services that the shell requires to act as a command interpreter.
Therefore, the Kernel allows the shell to read terminal input, to create pipes and
to redirect I/O. The computer users can also create private versions of the shell
so that they can create an environment according to their own requirements
without disturbing the other users.
10.2.5.1 Assumptions about Hardware
Whenever the user on the UNIX system executes a process, it
is divided into two levels: User level and Kernel level. So,
as and when a process executes a system call, the execution
mode of the process changes from the user mode to Kernel mode.
The Kernel tries to process the requests made by the user. It
returns an error message if the process fails. However, if no
requests are given to the operating system to service, even
then the operating system keeps itself busy with other
operations such as handling interrupts, scheduling processes,
managing memory and so on. The main differences between the
user mode and the Kernel mode are given below:
1. Process in a user mode can access their own instructions and data but they
cannot access the instructions and data of the Kernel. But all the processes
present in the Kernel can have the access to both the Kernel and the user
addresses.
2. Some machine instructions give an error message when executed in user
mode. For instance, a machine may contain an instruction that manipulates the
processor status register. This instruction is not allowed to be executed in user
mode. Processes executing in user mode should not have this capability
otherwise they may corrupt the kernel loaded.
It is very true that the system runs in either the user mode or the Kernel mode.
However, the Kernel runs on behalf of the user process. The Kernel is not a
separate process running parallel to user processes. The Kernel forms a part of
each user process.

Lesson No. 1 Intro. to Operating System 241

10.2.5.2 Interrupts and Exceptions
The UNIX system allows devices such as I/O peripherals or the system clock to
interrupt the CPU abruptly. Whenever the Kernel receives the interrupt, it saves the
current work it is doing and services the interrupt. After the interrupt is processed,
the Kernel resumes the interrupted work and proceeds as if nothing had happened.
The hardware gives a priority weightage according to the order in which the
interrupts should be handled. Thus, when the Kernel looks into an interrupt, it keeps
the lower priority interrupts waiting and services the higher priority interrupts.
The term exception is different from the term interrupt. An exception is a condition in
which a process causes an unexpected event. For instance, dividing a number by
zero, illegal address, out of memory, etc. Exceptions occur in the middle of the
execution of an instruction and are the similar to interrupts. The system tries to start
the instruction again after handling the exception. However, interrupts are
considered to happen between the executions of two instructions. The system
continues working on the next instruction after servicing the interrupt.
10.2.5.3 Processor Execution Levels
Sometimes, the Kernel must stop the interrupt from occurring during critical activity
preventing data corruption. For instance, the Kernel might not want to handle an
interrupt when it is working with linked lists because handling the interrupt at this
point of time might lead to corruption of pointers. Therefore, a better technique has
been worked out. The processor execution levels can be set with the help of certain
instructions. If you set the processor execution level to certain value, then it can
keep away the interrupt from that level and lower levels. It will only allow the high
level interrupts to disturb the process.
10.2.6 File System and Internal Structure of Files
Kernel does not impose any structure on files, and no meaning is attached to its
contents - the meaning of bytes depends solely on the program that interprets the
file. This is not true of just disc files but of peripherals devices as well. Magnetic
tapes, mail messages, character typed on the keyboard, line printer output, data
flowing in pipes - each of these is just a sequence of bytes as far as the system and
the programs in it are concerned.

Lesson No. 1 Intro. to Operating System 242

Files are organized in tree-structured directories. Directories are themselves files
that contain information on how to find other files. A path name to a file is a text
string that identifies a file by specifying a path through the directory structure to the
file. Syntactically it contains of individual file name elements separated by the slash
character.
The UNIX file system supports two main objects: files and directories. Directories
are nothing but files, which have a special format.
10.2.6.1 Representation of Data in a File
All the data entered by the user is kept in files. Internally the data blocks take up most of the
data that has been put in files. Each block on the disk is addressable by a number. Associated
with each file in UNIX is a little table called inode, which contains the table of contents to
locate a file's data on disk. The table of contents consists of a set of disk block numbers. An
inode maintains the attributes of a file, including the layout of its data on disk. Disk inodes
consists of the following fields:
¾ Last modification date
¾ Last access date
¾ Time the file last read
¾ Last inode modification
¾ Time the file was last modified
¾ Reference count
¾ Block reference pointer and indirect pointer to blocks in the file
The data on a file is not stored in a contiguous section of the disk. The reason
behind is that the Kernel will have to allocate and reserve continuous space in the
file system before allowing operations that would increase the file size. For instance,
let us suppose that there are three files A, B and C. Each file consists of 10 blocks
of storage and supposes the system allocated storage for the three files
contiguously as shown in Figure 10.3.

Lesson No. 1 Intro. to Operating System 243

 ……………….. File File File File …………
A B C D
40 50 60 70 80
Allocation of contiguous files
Figure 10.3
However, if the user now wishes to add 5 blocks of data in the file B, then the Kernel
will have to copy the file to such a place where a file of 15 blocks can be
accommodated. Moreover, the previously occupied disk block by file B's data can
only be used in a case where the files have data less than 10 blocks.
The Kernel allocates the file space of one block at a time. This allows the data to be
spread with throughout the file system. In this case, locating the data of a file
becomes a complicated process. If a block contains 10K bytes, then such a file
would need an index of 100 block numbers and as the block of 100K bytes would
need an index of 1000 block numbers. Thus, the size of the inode would keep
varying according to the size of the file.
10.2.6.2 Directories
The directories are files that give the file system a hierarchical structure. In a
directory the data is put in a sequence of entries. Each such entry contains an inode
number and the name of a file present in the directory .The pathname is a null
terminated character string. The pathname is divided into separate parts by the /
(slash) character. Each component of the pathname should hold the name of a
directory. However, the very last component can be a non-directory file. The
component names can have a maximum of 14 characters, with a 2 byte entry for the
inode number, the size of a directory entry is 16 bytes. Each directory contains the
file names dot and dot-dot. The inode numbers of these directories are those of the
directory and its parent directory respectively. The inode number of "." in "\etc"
directory is present at offset 0 in the file and its value is 83. The inode number of "..”
is present at the offset 16 and its value is 2. Any directory entry can also be kept
empty. Its inode number is indicated by 0.

Lesson No. 1 Intro. to Operating System 244

The data stored by the Kernel for a directory is similar to the data stored for an
ordinary file. For the directory also the Kernel makes use of the inode structure and
direct and indirect blocks. The access permission of a directory has the following
meaning: the read permission allows a process to read a directory. Write permission
allows a process to create new directory entries and remove the old directory
entries. It accounts for altering the contents of a directory. The execute permission
allows a process to search the directory for a filename.

root

dev bin user etc tmp lib

Figure 10.4
Figure 10.4 shows a typical UNIX File System. The file system is organized as a
tree with a single root node called the root (written "/ "); every non-leaf node of the
file system structure is a directory, and leaf nodes of the tree are either directories or
regular files or special devices.
¾ The /bin directory contains the executable files for most UNIX commands.
¾ The /etc directory contains other additional commands that related to system
maintenance and administration. It also contains several files, which store the
relevant information about the users of the system, the terminals and devices
connected to the system.
¾ The /lib directory contains all the library functions provided by UNIX for the
programmers.
¾ The /dev directory stores files that are related to the devices. UNIX has a file
associated with each of the I/O devices.
¾ The /user directory is created for each user to have a private work area where
the user can store his files. This directory can be given any name. Here it is
named as "user".
¾ The /tmp directory is the directory into which temporary files are kept. The files
stored in this directory are deleted as soon as the system is shutdown and
restarted.

Lesson No. 1 Intro. to Operating System 245

Create, open, read, write are system calls, which are used for basic file
manipulation. The create system call; given a path name creates an empty file. An
existing file opened by the open system call, which takes a path name and a node
and returns a small descriptor which may then be passed to a read or write system
call to perform data transfer to or from the file.
A file descriptor is an index into a small table of open files for this process.
Descriptors start at 0 and seldom get higher than 6 or 7 for typical programs,
depending on the maximum number of simultaneously open files. Each read or write
updates the current offset into the file, which is associated with file table entry and is
used to determine the position in the field for the next read or write.
10.2.6.3 Blocks and Fragments
Most of the file system is taken up by data blocks, which contain whatever the users
have put in their files. The hardware disk sector is usually 512 bytes. A block size
larger than 512 bytes is desirable for a speed. However, because UNIX file system
usually contain a very large number of small files, much larger blocks would cause
excessive internal fragmentation. That is why the earlier 4.1 BSD file system was
limited to 1024-byte block. The 4.2 BSD solution is to use two block sizes for files
which have no indirect blocks: all the blocks of the file are large block size except
the last. The last block is an appropriate multiple of a smaller fragment size to fill out
the file. Thus, a file of size 18000 bytes would have two 8K blocks and one 2K block
fragment.
The block and fragment sizes are set during the file creation according to the intended use of
the file system: if many small files are expected, the fragment size should be small; if
repeated transfer of large files are expected, the basic block size should be large.

10.2.7 UNIX Shell

A shell is the user-interface to the UNIX. A shell could use many different strategies to
execute a user's computation. The approach used in modern shells is to create a new process
(or thread) to execute new computation. For example, if a user decides to compile a program,
the process interacting with the user creates a new child process to carry out the compilation
took to execute the compiler program. The initial process (the OS) can use this same
technique when it decides to service a new interactive user in a timesharing environment.

Lesson No. 1 Intro. to Operating System 246

That is, when the user attempts to establish an interactive session, the OS treats this as a new
computation. It awakens a previously created process for the login port or creates a new
process to handle the interaction with the user.

This idea of creating a new process to execute a computation may seem like overkill, but it
has a very important characteristic. When the original process decides to execute a new
computation, it protects itself from any fatal errors that might arise during that execution. If
it did not use a child process to execute the command, a chain of fatal errors could cause the
initial process to fail, thus crashing the entire system.

The Bourne shell and others accept a command line from the user, parse the command line,
and then invoke the OS to run the specified command with the specified arguments. When a
user passes a command line to the shell, it is interpreted as a request to execute a program in
the specified file - even if the file contains a program that the user wrote. That is, a
programmer can write an ordinary C program, compile it, then have the shell execute it just
like it was a normal UNIX command.

For example, you could write a C program in a file named main.c, then compile and execute
it with shell commands like

$ cc main.c
$ a.out
The shell finds the cc command (the C compiler) in the /bin directory, then passes it
the string "main.c" when it creates a child process to execute the cc program. The C
compiler, by default, translates the C program that is stored in main.c, then writes
the resulting executable program into a file named a.out in the current directory. In
the second command, the command line is just the name of the file to be executed,
a.out (without any parameters). The shell finds the a.out file in the current directory,
then executes it.
Consider the detailed steps that a shell must take to accomplish its job:
¾ Printing a prompt. There is a default prompt string, sometimes hard coded into
the shell, e.g., the single character string "%", "#", ">" or other. When the shell is
started, it can look up the name of the machine on which it is running, and
prepare this string name to the standard prompt character, for example giving a

Lesson No. 1 Intro. to Operating System 247

 prompt string such as "kio-wa$". The shell can also be designed to print the
current directory as part of the prompt, meaning that each time the user employs
cd to change to a different directory, the prompt string is redefined.
¾ Once the prompt string is determined, the shell prints it to screen whenever it is
ready to accept a command line.
¾ Getting the command line. To get a command line, the shell performs a blocking
read operation so that the process that executes the shell will be blocked until
the user types a command line in response to the prompt. When the command
has been provided by the user and terminated with a NEWLINE character, the
command line string is returned to the shell.
¾ Parsing the command. The syntax for the command line is trivial. The parser
begins at the left side of the command line and scans until it sees a white space
character. The first such word is treated as the command name, and subsequent
words are treated as parameter string.
¾ Finding the file. The shell provides a set of environment variables for each user-
this variable is first defined in the user's login file, though it can be modified at
any time with the set command. The PATH environment variable is an ordered
list of absolute pathnames that specifies where the shell should search for
command files. If the login file has a line such as
set path=(.:/bin:/usr/bin)
The shell will first look in the current directory (since the first pathname is "." for the
current directory), then in /bin, and finally in /usr/bin. If there is no file with the same
name as the command in any of the specified directories, the sheII responds to the
user that it is unable to find the command.
¾ Preparing the parameters. The shell simply passes the string parameters to the
command as the argv array of pointers to strings.
¾ Executing the command. Finally the shell must execute the binary object
program in the specified file. UNIX shells have always been designed to protect
the original process from crashing when it executes a program. That is, since a
command can be any executable file, the process that is executing the shell
must protect itself in case the executable file has a fatal error in it. Somehow, the

Lesson No. 1 Intro. to Operating System 248

 shell wants to "launch" the executable so that even if the executable contains a
fatal error (which destroys the process executing it), the shell will remain
unharmed.
The Bourne shell uses multiple processes to accomplish what the UNIX-style system calls
fork, execve, and wait. This system call creates a new process, which is a copy of the calling
process except that it has its own process identification (with the correct relationships to the
other processes) and its own pointers to shared kernel entities such as file descriptors. After
fork has been called, two processes will execute the next statement after the fork in their own
address spaces - the parent and the child. If the call succeeds in the parent process, fork
returns the process identification of the newly created child process, and in the child process,
fork() returns a zero value.
execve. This system call is used to change the program that the process is currently
executing. It has the form execve (char *path, char *argv[], char *envp[]. The path
argument is the pathname of a file that contains the new program to be executed.
The argv array is a list of parameter strings, and the envp array is a list of
environment variable strings and values that should be used when the process
begins executing the new program. When a process encounters the execve system
call, the next instruction it executes will be the one at the entry point of the new
executable file. This means that the kernel performs a considerable amount of work
in this system call. It must find the new executable file, load it into the address space
currently being used by the calling process (overwriting area and discarding the
previous program), set the argv array and environment variables for the new
program execution, then start the process executing at the new program’s entry
point.
There are various versions of execve available at the system call interface; they
differ in the way parameters are specified.
wait. A process uses this system call to block itself until the kernel signals the
process to execute again. For example, because one of its children processes has
terminated. When the wait call returns as a result of a child process terminating, the
status of the terminated child is returned as a parameter to the calling process.
10.2.8 User Interaction with UNIX Operating System

Lesson No. 1 Intro. to Operating System 249

To interact with UNIX, the first step is login process in which we use a name and
password initially assigned by the system administrator.
10.2.8.1 Steps to Login
Logging in is a procedure that tells the UNIX System who you
are; the system responds by asking you the password. So, in
order to login, first, connect your PC to the UNIX system.
After a successful connection is established, you would find
the following prompt coming in the on the screen.
Login:
Each user on the UNIX system is assigned an account name, which identifies him
as a unique user. The account name is any combination of eight or less characters.
Now, at the login prompt, enter your account name. Press Enter Key. Type your
account name in lowercase letters. UNIX treats uppercase letters differently from
lowercase letters.
Login: pankaj
Password: ******
Once the login name is entered, UNIX prompts you to enter a
password. While you are entering your password, it will not be
shown on the screen. If you give either the login name or the
password wrong, then UNIX denies you the permission to access
its resources. The system then shows an error message on the
screen, which is given below:
Login: pankaj
Password: ******
Login incorrect:
Login:
Many UNIX systems give you three or four chances to enter your login and password
correct. Once you have successfully logged on by giving a correct login and password, you
are given some information about the system, some news for users and a message whether
you have an electronic mail or not and followed y $ prompt. The dollar sign is the UNIX's
method of telling that it's ready to accept commands from the user. You can have a different

Lesson No. 1 Intro. to Operating System 250

prompt also in a case where your system is configured for showing a different prompt. By
default a $ is shown for the Korn or Bourne Shells.

At this point, you are ready to enter your first UNIX command. Now, when you are
done working on your UNIX system and decide to leave your terminal - then it is
always a good idea to log off the system. In order to log off the system, type the
following command:
$ exit
login:
The above command will work if you are using a Bourne or a Korn shell. However, if
you are working on C shell, you can give another command to log off.
$ logout
login:
UNIX system is very particular not to allow the unauthorized users to access the
system. So, when a message like 'Login denied' comes on the screen, it does not
tell you what was wrong with your login.
10.2.8.2 Changing your Password
You can change your password with the 'passwd' command. The procedure of
changing passwords is very simple. In order to change your password, you first
have to log on to the UNIX system. Then issue the 'passwd' command at the UNIX
prompt.
Syntax: passwd [user-name] Options
-d Deletes your password
-x days. This sets the maximum number of days that the password will be date
active. After the specified number of days you will be required to give a new
password.
-n days. This sets the minimum number of days the password has to be active,
before it can be changed.
-s: This gives you the status of the user's password.
Only the superuser can use the above options.
Example:
$passwd

Lesson No. 1 Intro. to Operating System 251

Changing password for shefali
Enter old password: *******
Enter new password: *******
Re-type new password: *******
Thus UNIX wants you to type in your old password. Then it asks for the new password.
Finally, UNIX confirms your new password by asking you to type in the new password
again. If by any means, any mismatch happens, then the UNIX system warns you that the
information provided by you is inconsistent as shown below:

$ passwd:
Changing password for shefali
Enter old password:
Enter new password:
Re-type new password:
Mismatch-password unchanged
UNIX also offers a variety of tools to maintain security .One such tool is the usage of
the 'lock' command. The lock command locks your keyboard till the time you enter a
valid password, as shown below:
$lock
Password: Sorry
Password:
10.2.8.3 UNIX Command Structure

There are a few of UNIX commands, that you can type them standalone. For
example, ls, date, pwd, logout and so on. But UNIX commands generally require
some additional options and/ or arguments to be supplied in order to extract more
information. Let us find out the basic UNIX command structure. The UNIX
commands follow the following format: Command [options] [arguments]
The options/arguments are specified within square brackets if they are optional. The
options are normally specified by a “-“ (hyphen) followed by letter, one letter per
option.
10.2.9 Common UNIX Commands
Some commonly used UNIX commands are discussed below:

Lesson No. 1 Intro. to Operating System 252

cal Command
The cal command creates a calendar of the specified month for the specified year, if
you do not specify the month, it creates a calendar for the entire year. By default this
command shows the calendar for the current month based on the system date. The
cal writes its output to the standard output.
Syntax: cal [ [mm] yy ]
where mm is the month, an integer between 1 and 12 and yy is the year; an integer
between 1 and 9999, For current years, a 4-digit number must be used, '98' will not
produce a calendar of 1998.
Options: None
date Command
It shows or sets the system date and time. If no argument is specified, it displays the
current date and the current time.
Syntax: date [+options] Options:
%d displays date as mm/dd/yy
%a displays abbreviated weekday (Sun to Sat)
%t displays time as HH:MM:SS
%r displays time as HH:MM:SS(A.M/P.M.)
%d displays only dd
%m displays only mm
If you are working in the superuser mode, you can set the date as shown below:
$ date MMddhhmm[yy]
where MM = Month (1-12)
dd = day (1-31)
hh = hour (1-23)
mm = minutes (1-59)
yy = Year
who Command
The who command lists the users that are currently logged into the system.
Syntax: who [options]
Options:

Lesson No. 1 Intro. to Operating System 253

¾ u - lists the currently logged-in users.
¾ t - gives the status of all logged-in users, am
¾ i - this lists login-id and terminal of the user invoking this command.
finger Command
In larger system, you may get a big list of users shown on the screen. The finger
command with an argument gives you more information about the user. The finger
command followed by an argument can give a complete information for a user who
is not logged onto the system.
Syntax: finger [user-name] Options: none
Examples
(i) $ finger xyz
If you want to know about everyone currently logged onto the system, give the
following command:
$finger
The ls Command
The ls command is used for listing information about files and directories.
Syntax: ls [-options] [filename]
Options:
-a - List all directory entries including dot (.) entries.
-d - Give the name of directories only.
-9 - Print group id only in the long listing.
-i - It Print inode number of each file in the first column.
-l - Lists in the long or detailed format owner's id only in the long listing.
-s - This lists the disk blocks (of 512 bytes each), occupied by a file.
Sort file names by time since last access.
-t - Sort file names by time since last modified.
-r - Recursively lists all subdirectories.
-f - Marks type of each file.
You can make use of more than one option at a given time, just group them together and
precede them with "-".

Lesson No. 1 Intro. to Operating System 254

The cp Command
The cp command creates a duplicate copy of a file.
Syntax: cp file1 file2
Options: None
Here, the file1 is copied as file2. If file2 already exists, the new file overwrites it. The
file names specified may be full path names or just the name (current working
directory, will then be assumed).
The mv Command
This command moves or renames files.
Syntax: mv file1 file2
Here file1 refers to the source filename and 'file2' refers to the destination filename.
Moving a file to another within the same directory is equivalent to renaming the file.
Otherwise also, mv doesn't really move the file, it just renames it and changes
directory entries.
The ln Command
The 'ln' command adds one or more links to a file. Syntax: ln filel file2
The ln command here establishes a link to an existing file. File name 'file1' specifies
the file that has to be linked and file name 'file2' specifies the directory into which the
link has to be established. If the 'file2' is in the same directory as file1 then the file
seems to carry names, but physically there is only one copy. If you use the ls -li
command, you will find that the link count has been incremented by one and that
both the files have the same inode number, as they refer to the same data blocks in
the disk. Any changes that are made to one file will be reflected in the other. And if
'file2' specifies a different directory, the file will be physically present at one place
but will appear as if it is present in other directory, thereby allowing different users to
access the file. It saves a lot of disk space because the file is not duplicated.
But you should note that you should have write permission to the directory under
which the link is to be created.
(i) $ln /usr/mkt/mkt.c /usr/mktl/new-mkt.c
This will create a link for file mkt.c in 'mkt' directory to 'mktl' directory by the name
'new-mkt.c'.

Lesson No. 1 Intro. to Operating System 255

(ii) $ In myfile.prg newfile.prg
The above command links the file 'myfile.prg' as 'new-file.prg' in the same directory.
You can see these files by giving the ls command.
The rm Command
The 'rm' command removes files or directories. Syntax: rm [options] file(s)
When you remove a file, you are actually removing a link. The space occupied by
the file on the disk is freed, only when you remove the last link to a file.
The Options:
c -confirms on each file before deleting it.
f - removes only those files which do not have write permission.
r - deletes the directory and its contents along with all the sub-directories and their
contents.
The cat Command
The cat writes the contents of one of more files, onto the screen in the sequence
specified. If you do not specify an input file, cat reads its data from the standard
input file, generally the keyboard.
Syntax: cat file.
Examples
$ cat /usr/mkt/new-mkt.c
This command will display the contents of ‘c’ program file ‘new-mkt.c’ onto the
screen.
chmod command
Only a superuser can change permissions for any file on the system.
g - the group to which the file owner belongs
o - other users, not part of the group
a - all users
'operation' denotes the options to be done, and can be:
+ add permission
- remove permission
= assign permission
‘permission' can be:

Lesson No. 1 Intro. to Operating System 256

r - read permission
w - write permission
x - execute permission
‘filename(s)’ can be the files on which you want to carry out this command.
Examples - First see the file permissions using the ls -1 command for mkt.c as
shown below:
$ls -l mkt.c
Output: -rwx- -x- -2 root other 1428 May 1507:34 mkt.c
i.e. user has rwx, group has x and all others also have x permission.
(i) Now use the chmod command as illustrated below:
$chmod u-x g+w o+r mkt.c
The above command remove execute (x) permission for user, give write (w) permission to
group and give read (r) permission to all others.
(ii) Then again use the ls -l command to verify, if the permissions have been or
not set.
$1s -1 mkt.c
Output: -rw- -wxr-x 2 root other 1428 May 1507:34 mkt.c
Alternatively, you could have also used the following commands to do the same
work:
$chmod u=rw g=wx o=rx mkt.c
If we use
$ chmod a=rwx mkt.c "
$ ls -l mkt.c
Output:-rwxrwxrwx 2 root other 1428 May 15 07:34 mkt.c
In the above command, a=rwx assigns read, write and execute permission to all
users.
The chown Command
The chown command changes the owner of the specified file(s).
Syntax: chown new-owner filename.
This command requires you to be in the superuser mode. The new owner can be
the user ID of the new owner or the new owner's user number. You can also specify

Lesson No. 1 Intro. to Operating System 257

the owner by his name. But the new owner should have an entry in the /etc/passwd
file. The filename is the name(s) of the file(s), whose owner is to be changed.
Options: None
Examples
$chown bobby sales.c
The above command now makes bobby the owner of sales.c file.
The chgrp Command
The chgrp is used to change the group of a file.
Syntax: chgrp group filename.
Only the superuser can use this command. This command changes the group
ownership of a file. Here group denotes the new group-ID and filename denotes the
file whose group-ID is desired to be changed.
10.2.10 File System, Permissions Changing Order and Group
File is a unit of storing information. All utilities, applications and data are represented
as files. The file may contain executable programs, texts or databases. They are
stored on secondary memory storage such as a disk or magnetic tape.
Naming Files
You can give filenames up to 14 characters long. The name may contain alphabets,
digits and a few special characters. Files in UNIX do not have the concept of primary
or secondary name as in DOS, and therefore file names may contain more than one
period(.). Therefore, the following file names are all valid filenames: mkt.c, name2.c,
.star, a.out
However, UNIX file names are case sensitive. Therefore, the following names
represent two different files in UNIX. mkt.c and Mkt.c
Types
The files under UNIX can be categorized as follows:
1. Ordinary files.
2. Directory files.
3. Special files.
4. FIFO files.
We are discussing about these files below.

Lesson No. 1 Intro. to Operating System 258

Ordinary Files
Ordinary files are the one, with which we all are familiar. They may contain
executable programs, text or databases. You can add, modify or delete them or
remove the file entirely.
Directory Files
Directory files as discussed earlier also represent a group of files. They contain list
of file names and other information related to these files. Some of the commands,
which manipulate these directory files, differ from those for ordinary files.
Special Files
Special files are also referred to as device files. These files represent physical
devices such as terminals, disks, printers and tape-drives etc. These files are read
from or written into just like ordinary files, except that operation on these files
activates some physical devices. These files can be of two types Character device
files and block device files. In character device files data is handled character by
character, as in case of terminals and printers. In block device files, data is handled
in large chunks of blocks, as in the case of disks and tapes.
FIFO Files
FIFO (first-in-first-out) are files that allow unrelated processes to communicate with
each other. They are generally used in applications where the communication path
is in only one direction, and several processes need to communicate with a single
process. For an example of FIFO file, take the pipe in UNIX. This allows transfer of
data between processes in a FIFO manner. A pipe takes the output of the first
process as the input to the next process, and so on.
File Names and Meta characters
In UNIX, we can refer to a group of files with the help of METACHARACTERS. The
valid meta characters are ?, [, and ]. It replaces any number of characters including
a null character.
? - used in place of one and only one character.
[] - brackets are used to specify a set of a range of characters.
Examples
(i) $ls []c

Lesson No. 1 Intro. to Operating System 259

It will list all files starting with any character or characters and ending with the
character c.
(ii) $ ls robin[]
It will list all the files starting with robin and ending with any character.
(iii) $ ls x?yz[]
It will list all those files, in which the first character is x, the second character can be
anything; the third and fourth characters should be respectively y and z and that the
rest of the name can be anything.
(iv) $ls I[abc]mn
It will list all those files, in which the first character is I, the second character can be
either a, b or c the last two characters i.e., 3rd and 4th should be m and n
respectively.
Alternatively the above command can also be given in the following manner.
$ ls I[a-c]mn
File Security and Ownership
The data is centralized on a system working with UNIX. The first step towards data
security is the usage of passwords. The next step should be to guard the data
among users. If the number of users is small, it is not much of a problem. But it can
be problematic on a large system supporting many users.
UNIX can thus differentiate files belonging to an individual, the owner of a file or
group of users or the others with different limited accesses, as the case may be. The
different ways to access a file are:
Read(r) - You can just look through the file.
Write(w) - You can also modify it.
Execute(x) - You can just execute it
Therefore if you have a file called vendor.c and that you are the owner of it, you may
provide; yourself with all the rights rwx [read, write and execute]. You can provide rx (read,
and execute) rights to the members of your group and only the x (execute) right to all others.

Normally, when you create a file, you are the owner of the file and your group
becomes the group id for the file. The system assigns a default set of permissions
for file, as set by the system administrator. The user can also change these

Lesson No. 1 Intro. to Operating System 260

permissions at his will. But only a super user can change these permissions (rwx),
ownership and group id's of any file in the system.
10.2.11 UNIX Editors
UNIX operating system provides several text editors can be classified into two types
-Line Editors and Screen Editors.
(a) Line Editors: The early UNIX editors that edit processes one line at a time are
called Line Editors. So with a line editor, you are required to give many commands
to display or edit a set of lines. The common examples of UNIX line editors are ed
and ex.
(i) ed - ed was the first line editor of UNIX which is still used sometimes, though it is
not popular. ed was popular in those days when most UNIX commands consisted of
only two or three letters. It has become outdated now due to the use of screen
editors that provides much more features.
(ii) ex - ex is more powerful and comprehensive than ed line editor. Some ex line-
oriented commands are also used in few screen editors (such as vi).
(b) Screen Editors: Editors that make use of the whole screen for editing or
processing more than one line at a time, are called screen editors. With screen
editors, you can display and edit many lines by giving a single command. The
common examples of screen editors are vi and emacs.
- vi -vi stands for 'Visual editor'. vi is the standard full-screen UNIX tool and is the
only editor available on SCO UNIX.
- emacs - emacs is another popular screen editor of UNIX. Although most
vendors distribute emacs with UNIX system, emacs is not a part of UNIX.
10.3 Keywords
Multi-tasking: More than one program can be made to run at the same time.
Multi-user: More than one user can work at the same computer system at the same
time.
Kernel: This is the actual operating system, a single large program that always
resides in memory. Sections of the code in this program are executed on behalf of
users to do needed tasks. Strictly speaking, the kernel is UNIX.
UNIX shell: A shell is the user-interface to the UNIX.

Lesson No. 1 Intro. to Operating System 261

10.4 Summary
Unix is a multi-user, multi tasking operating system written in high-level language. It
is portable, modifiable and understandable. Some popular versions of UNIX are AIX
IBM, XENIX, ULTRIX, Sun OS, and BSD. The UNIX is organized as a set of layers.
In the center, Kernel surrounds the hardware, which is surrounded by shell and
editors that interact with the Kernel by invoking a well-defined set of system calls.
Another important feature of UNIX is its security implemented by password, file
permissions, and encryption. Files are organized in tree-structured directories.
Directories are also files that contain information on how to find other files. Kernel
does not impose any structure on files; the meaning of bytes depends solely on the
program that interprets the file. This is not true of just disc files but of peripherals
devices as well, each of these is just a sequence of bytes.
UNIX provides a number of line and screen editors such as ed, ex, and vi.
10.5 SELF-ASSESSMENT QUESTIONS (SAQ)
1. How many types of files there can be in UNIX?
2. How can security of files be maintained?
3. What are different types of users in UNIX for files?
4. What is function of ls command? Give various options.
5. Differentiate between cp & mv Command.
6. What is rm command used for?
7. Differentiate between chmod & chown Command.
8. What is meant by a multi-user and multi-tasking operating system? Is UNIX
multi-user, how?
9. What are various Features of UNIX? What makes UNIX portable and secure?
10. Give different layers of UNIX Architecture. Explain the intended purposes of
each.
11. What do you understand by Kernel? What are the functions provided by it?
12. How can the exceptions be resolved in UNIX?
13. How are file organized in UNIX? What is the difference between a directory &
a file in Unix?

Lesson No. 1 Intro. to Operating System 262

14. What do you understand by Editor? Why it is needed? How many types of
Editors are present?
10.6 SUGGESTED READINGS / REFERENCE MATERIAL
1. The Design of the UNIX Operating System, Bach M.J., PHI, New Delhi, 2000.
2. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
3. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.
4. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
5. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.
6. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company Ltd.,
New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 263

Lesson Number: 11 Writer: Dr. Rakesh Kumar
Case Study of MS-DOS Vetter: Prof. Dharminder Kumar

11.0 Objectives
The objective of this lesson is
(C) To give an overview of the important features of MS-DOS operating system.
(D) To make familiar with some important MS-DOS commands.

11.1 Introduction
MS-DOS is a single-user operating system. It is designed to operate on
machines using the Intel line of 8086 microprocessors. These processors include
the 8088, 8086, 80286, 80386, 80486, and the new Pentium (80586). PCs that
are called 386s or 486s; are based on their processor names, 80386 and 80486.
The 80586 was so different from the previous versions that it was given the name
Pentium as a distinction. MS-DOS cannot support a large network of users.
Because PCs and MS-DOS became so popular, many network operating
systems were designed in the mid-1980s to network MS-DOS machines
together; working around the operating system limitations through software. MS-
DOS is the preferred operating system for most of the Intel processor (currently
Pentium) based PCs of the world. MS-DOS does not break out into neat
compartments as easily as some operating systems do. This is partially due to its
simplicity -- no multi-user or multitasking ability. It is also partly due to the way
MS-DOS has evolved over the years.

So in many respects DOS was a primitive OS. It was based on previous systems,
of course, and echoes of UNIX and CP/M can be seen in it. But if you read
Stephenson's article above, you will realize that it had hidden power, as well,
because you could interact more directly with the components of the computer
than you can with more modern operating systems. It is this power that makes it
valuable to know DOS today. The majority of computer users today use graphical

Lesson No. 1 Intro. to Operating System 264

systems, but users of Windows, for instance, have the ability to interact with the
computer through DOS (or in the case of NT, with a DOS-analogue) to do things
which are difficult or impossible to accomplish through the graphical interface.
For this reason, familiarity with DOS is still considered essential for anyone
supporting Intel-based machines running Windows.

11.2 Presentation of contents

11.2.1 Kernel
11.2.2 COMMAND.COM
11.2.3 MS-DOS Services
11.2.4 Startup Files
11.2.5 Limitations
11.2.6 Security Shortcomings
11.2.7 Ease of Use
11.2.8 DOS Variations
11.2.9 The user’s view of MS-DOS
11.2.9.1 MS-DOS Commands
11.2.9.2 Managing Input and Output
11.2.9.3 Setting up the Environment
11.2.9.4 Batch Files
11.2.9.5 Working with Disks
11.2.9.6 Working with Directories
11.2.9.7 Managing Files
11.2.9.8 Device drivers
11.2.9.9 Memory Management
11.2.9.10 Controlling Program Flow
11.2.10 System calls - The programmer’s view
11.2.10.1 Working with Disks
11.2.10.2 Working with Directories

Lesson No. 1 Intro. to Operating System 265

 11.2.10.3 Managing Files
11.2.10.4 Managing Input and Output
11.2.10.5 Managing Memory
11.2.10.6 Controlling Program Flow

11.2.1 Kernel
MS-DOS uses two hidden files at boot time. These are io.sys and msdos.sys. For
all practical purposes, these files, in conjunction with the firmware BIOS (Basic
Input Output Services) built into every PC, make up the MS-DOS kernel (basic
operating system). They load at the time of startup and allow the command
processor to run. These files are not rebuildable or alterable. Software to run
printers or CD-ROMs (device drivers) can be installed in MS-DOS but the kernel
cannot be changed.

11.2.2 COMMAND.COM
This file starts the command processor. The command processor uses the
commands you enter at the C:> prompt. When you run application programs and
then return to MS-DOS, the system must be able to find COMMAND.COM and
reload it back into memory (RAM). The command processor also supports a
command language knows as the DOS Batch language. The Batch language
files have a .BAT extension and are considered executable by the command
processor. The Batch language is not as powerful as some command languages
but does support conditional statements and variables (if time = next_day
then...). In the early years of PCs, the Batch language was used for many tasks.
Now, low cost utilities often provide many times the functions in addition to direct
support. However, batch files are still very common and very useful.

11.2.3 MS-DOS Services

Because MS-DOS is not able to execute more than one task at a time, it cannot
have multiple jobs working in the background. However, MS-DOS does use

Lesson No. 1 Intro. to Operating System 266

services that allow programs to interact with the computer. These services
include basic input/output services (BIOS), print services, and file services. MS-
DOS services use what is known as an interrupt to the microprocessor. This
interrupt concept allows terminate-and-stay-resident (TSR) programs to function.
TSRs can sit quietly in the background and appear when the user requests them.
TSRs provide MS-DOS with the appearance of running more than one task at a
time.

11.2.4 Startup Files

MS-DOS uses two changeable startup files. They are CONFIG.SYS and
AUTOEXEC.BAT. The CONFIG.SYS has the task of loading installable device
drivers and other system parameters which must run at boot time. The device
drivers might be for a CD-ROM drive or a sound board. The AUTOEXEC.BAT
starts applications automatically for the user, handles logins for network software,
and places information in the PC’s environment. In some cases, there is not a
clear-cut line that exists between which commands go into the CONFIG.SYS and
which go into the AUTOEXEC.BAT. In general, the CONFIG.SYS holds only
device drivers, special commands provided with MS-DOS, and memory
managers.

11.2.5 Limitations
MS-DOS is limited in its memory usage. From the beginning, MS-DOS was
designed to allow only the first 640 kilobytes of memory (RAM) to be used for
application programs, even though only 1 megabyte of memory was
addressable. To current MS-DOS users, 640 KB has proven to be inadequate as
well. The latest version (as of this writing), MS-DOS 6.0, has special memory
management tools and techniques built into allow programs additional memory.
Additionally, and entire counter-culture of memory management tools have been
developed which all programs to use more memory under previous versions of
MS-DOS. These include Quarterdeck’s QEMM386, Qualitas Inc. 386 M, and
Pharlap for the 486.

Lesson No. 1 Intro. to Operating System 267

Closely tied to the limitation on memory is MS-DOS’s inability to run more than
one program at a time. The nature of MS-DOS requires one program to end
before another one begins. Again, the ingenuity of programmers has worked
numerous tricks to get around this limitation. For one, TSRs can load into
memory and stay throughout a session. These are background programs
available for use by other programs. However, many programmers have created
poorly-behaved TSRs, which are unstable, or can cause multiple problems with
other programs. Lack of standards has given TSRs a bad reputation. Sometimes
a TSR itself is reliable, but does not work well with other TSRs. A second solution
to the single application shortcoming has been task-switching programs like
Software Carousel and the original Microsoft Windows. These allow you to
quickly move between different applications that are in RAM or located on a hard
disk swap space. Though certainly not multitasking, they are much better than
nothing. Nevertheless, the only solution that attempts multitasking is Microsoft
Windows. You should understand that no matter what Microsoft says, Microsoft
Windows is not an operating system. MS-DOS remains the operating system and
Windows is simply a user interface with lots of flexibility.

11.2.6 Security Shortcomings

MS-DOS has no concern for security or networking. There are no passwords,
login files, encryption, networking hooks, mail programs, communication
programs, or any of the other tidbits that users have come to take for granted.
This security shortage presents a particularly difficult problem in a networked
environment. Most network operating systems have put security measures on top
of MS-DOS. These typically involve logins and access permissions to the
programs on the server. MS-DOS now comes with a program called SHARE that
controls access to the same file by multiple users over a network. Still, each
individual MS-DOS machine is inherently susceptible to viruses, rogues, and
security leaks.

11.2.7 Ease of Use

Lesson No. 1 Intro. to Operating System 268

How did MS-DOS become so popular? The answer is a combination of
marketplace luck, good advertising, low cost, and ease of use. Precisely because
MS-DOS had no restraints and because IBM put MS-DOS on its PCs, anyone
could easily buy a machine and put MS-DOS on their desk. The DOS batch
language allowed nonprogrammers to write simple programs and interfaces. MS-
DOS itself is not that easy to use, but the programs created to run under MS-
DOS were themselves understandable to the computer illiterate individuals of the
world.

11.2.8 DOS Variations

Though MS-DOS is the most popular, Microsoft does not make the only versions
of DOS. IBM has a license to create DOS, which runs on IBM PS/2 computers
and is called PC-DOS. Up until DOS 4.0, there were only minor differences
between MS-DOS and PC-DOS. Since that time there has been more
divergence between the two companies. In the late 1980s, a number of DOS
multi-user versions appeared on the market including VM-386 and CMOS. The
only major competitor to MS-DOS in the single machine market is DR-DOS from
Digital Research (DRI), that is the same company, now owned by Novell Inc.,
which was involved in the original discussions with IBM that crashed and burned.
Essentially, DR-DOS has jumped a release ahead of MS-DOS when it comes to
new capabilities and technologies. DRI guarantees the compatibility of DR-DOS
with programs designed for MS-DOS.

11.2.9 THE USER’S VIEW OF MS-DOS

MS-DOS provides an interface between the user and the computer system in
form of a set of commands, which are available through a command interpreter.
COMMAND.COM file is the MS-DOS command interpreter. The totality of DOS
commands may be divided in three classes, viz., internal commands, external
commands and batch command. In MS-DOS there are two types of commands.
An Internal command, which is a command embedded into the command.com

Lesson No. 1 Intro. to Operating System 269

file, and an external command, which is not embedded into command.com and
therefore requires a separate file to be used.

12.2.9.1 MS-DOS commands

COMMAND DESCRIPTION
ANSI.SYS Defines functions that change display graphics, control cursor
movement, and reassign keys.
APPEND Causes MS-DOS to look in other directories when editing a file or
running a command.
ARP Displays, adds, and removes arp information from network devices.
ASSIGN Assign a drive letter to an alternate letter.
ASSOC View the file associations.
AT Schedule a time to execute commands or programs.
ATMADM Lists connections and addresses seen by Windows ATM call
manager.
ATTRIB Display and change file attributes.
BATCH Recovery console command that executes a series of commands in
a file.
BOOTCFG Recovery console command that allows a user to view, modify, and
rebuild the boot.ini
BREAK Enable / disable CTRL + C feature.
CACLS View and modify file ACL's.
CALL Calls a batch file from another batch file.
CD Changes directories.
CHCP Supplement the International keyboard and character set information.
CHDIR Changes directories.
CHKDSK Check the hard disk drive running FAT for errors.
CHKNTFS Check the hard disk drive running NTFS for errors.
CHOICE Specify a listing of multiple options within a batch file.

Lesson No. 1 Intro. to Operating System 270

CLS Clears the screen.
CMD Opens the command interpreter.
COLOR Easily change the foreground and background color of the MS-DOS
window.
COMMAND Opens the command interpreter.
COMP Compares files.
COMPACT Compresses and uncompress files.
CONTROL Open Control Panel icons from the MS-DOS prompt.
CONVERT Convert FAT to NTFS.
COPY Copy one or more files to an alternate location.
CTTY Change the computers input/output devices.
DATE View or change the systems date.
DEBUG Debug utility to create assembly programs to modify hardware
settings.
DEFRAG Re-arrange the hard disk drive to help with loading programs.
DEL Deletes one or more files.
DELETE Recovery console command that deletes a file.
DELTREE Deletes one or more files and/or directories.
DIR List the contents of one or more directory.
DISABLE Recovery console command that disables Windows system services
or drivers.
DISKCOMP Compare a disk with another disk.
DISKCOPY Copy the contents of one disk and place them on another disk.
DOSKEY Command to view and execute commands that have been run in the
past.
DOSSHELL A GUI to help with early MS-DOS users.
DRIVPARM Enables overwrite of original device drivers.
ECHO Displays messages and enables and disables echo.
EDIT View and edit files.

Lesson No. 1 Intro. to Operating System 271

EDLIN View and edit files.
EMM386 Load extended Memory Manager.
ENABLE Recovery console command to enable a disable service or driver.
ENDLOCAL Stops the localization of the environment changes enabled by the
setlocal command.
ERASE Erase files from computer.
EXIT Exit from the command interpreter.
EXPAND Expand a Microsoft Windows file back to it's original format.
EXTRACT Extract files from the Microsoft Windows cabinets.
FASTHELP Displays a listing of MS-DOS commands and information about
them.
FC Compare files.
FDISK Utility used to create partitions on the hard disk drive.
FIND Search for text within a file.
FINDSTR Searches for a string of text within a file.
FIXBOOT Writes a new boot sector.
FIXMBR Writes a new boot record to a disk drive.
FOR Boolean used in batch files.
FORMAT Command to erase and prepare a disk drive.
FTP Command to connect and operate on a FTP server.
FTYPE Displays or modifies file types used in file extension associations.
GOTO Moves a batch file to a specific label or location.
GRAFTABL Show extended characters in graphics mode.
HELP Display a listing of commands and brief explanation.
IF Allows for batch files to perform conditional processing.
IFSHLP.SYS 32-bit file manager.
IPCONFIG Network command to view network adapter settings and assigned
values.
KEYB Change layout of keyboard.

Lesson No. 1 Intro. to Operating System 272

LABEL Change the label of a disk drive.
LH Load a device driver in to high memory.
LISTSVC Recovery console command that displays the services and drivers.
LOADFIX Load a program above the first 64k.
LOADHIGH Load a device driver in to high memory.
LOCK Lock the hard disk drive.
LOGON Recovery console command to list installations and enable
administrator login.
MAP Displays the device name of a drive.
MD Command to create a new directory.
MEM Display memory on system.
MKDIR Command to create a new directory.
MODE Modify the port or display settings.
MORE Display one page at a time.
MOVE Move one or more files from one directory to another directory.
MSAV Early Microsoft Virus scanner.
MSD Diagnostics utility.
MSCDEX Utility used to load and provide access to the CD-ROM.
NBTSTAT Displays protocol statistics and current TCP/IP connections using
NBT
NET Update, fix, or view the network or network settings
NETSH Configure dynamic and static network information from MS-DOS.
NETSTAT Display the TCP/IP network protocol statistics and information.
NLSFUNC Load country specific information.
NSLOOKUP Look up an IP address of a domain or host on a network.
PATH View and modify the computers path location.
PATHPING View and locate locations of network latency.
PAUSE Command used in batch files to stop the processing of a command.
PING Test / send information to another network computer or network

Lesson No. 1 Intro. to Operating System 273

 device.
POPD Changes to the directory or network path stored by the pushd
command.
POWER Conserve power with computer portables.
PRINT Prints data to a printer port.
PROMPT View and change the MS-DOS prompt.
PUSHD Stores a directory or network path in memory so it can be returned to
at any time.
QBASIC Open the Qbasic.
RD Removes an empty directory.
REN Renames a file or directory.
RENAME Renames a file or directory.
RMDIR Removes an empty directory.
ROUTE View and configure windows network route tables.
RUNAS Enables a user to execute a program on another computer.
SCANDISK Run the scandisk utility.
SCANREG Scan registry and recover registry from errors.
SET Change one variable or string to another.
SETLOCAL Enables local environments to be changed without affecting anything
else.
SETVER Change MS-DOS version to trick older MS-DOS programs.
SHARE Installs support for file sharing and locking capabilities.
SHIFT Changes the position of replaceable parameters in a batch program.
SHUTDOWN Shutdown the computer from the MS-DOS prompt.
SMARTDRV Create a disk cache in conventional memory or extended memory.
SORT Sorts the input and displays the output to the screen.
START Start a separate window in Windows from the MS-DOS prompt.
SUBST Substitute a folder on your computer for another drive letter.
SWITCHES Remove add functions from MS-DOS.

Lesson No. 1 Intro. to Operating System 274

SYS Transfer system files to disk drive.
TELNET Telnet to another computer / device from the prompt.
TIME View or modify the system time.
TITLE Change the title of their MS-DOS window.
TRACERT Visually view a network packets route across a network.
TREE View a visual tree of the hard disk drive.
TYPE Display the contents of a file.
UNDELETE Undelete a file that has been deleted.
UNFORMAT Unformat a hard disk drive.
UNLOCK Unlock a disk drive.
VER Display the version information.
VERIFY Enables or disables the feature to determine if files have been written
properly.
VOL Displays the volume information about the designated drive.
XCOPY Copy multiple files, directories, and/or drives from one location to
another.

11.2.9.2 Managing Input and Output

MS-DOS normally reads input from the standard input, and normally writes output to
the standard output. The standard input or output may be redirected to a file or a
device. MS-DOS treats devices as files. The output of a command or program can
be piped to another command or program. Following DOS commands are relevant
for I/O manipulation: CLS, CTTY, FIND, GRAFTABL, GRAPHICS, MODE, MORE,
PRINT, SORT, and TYPE.

11.2.9.3 Setting up the Environment

The user can tailor the working environment via the
CONFIG.SYS file that defines operating characteristics of an
MS-DOS system. When MS-DOS starts, it automatically executes
the commands in the CONFIG.SYS file if one is available.

Lesson No. 1 Intro. to Operating System 275

(a) ANSI.SYS - This is a device driver that implements the American National
Standards Institute (ANSI) standard escape codes for screen and keyboard
control.
(b) BREAK - Control how often MS-DOS checks for a Control-C interrupt. (When
MS- DOS encounters a Control-C, it interrupts the current process and returns
to the system prompt.)
(c) BUFFERS - Specify the number of in-memory disk buffers that MS-DOS
allocates each time it starts.
(d) COUNTRY - Specify the country, thus indicating the appropriate date, decimal
sign and currency symbol to be used.
(e) DEVICE - Install a device driver.
(f) DRIVPARM - Redefine the default characteristics defined by a device driver for
a block device.
(g) FCBS - Specify the maximum number of FCB-controlled files that can be open
while file sharing is in effect.
(h) FILES - Specify the maximum number of open files controlled by file handles.
(i) LASTDRIVE - Determine the maximum number of drives by setting the highest
drive letter that MS-DOS recognizes.
(j) SHELL - Specify a command processor to be used in place of
COMMAND.COM.
(k) VDISK.SYS-Create a memory-resident virtual disk. This is a file containing the
device driver that actually creates the virtual disk in RAM.
(l) VDISK.SYS is placed in the CONFIG.SYS file with the DEVICE command. A
virtual disk is much faster than a real disk, but the data is lost when the power
is turned off.
(m) CHKDSK reports the storage allocated to the virtual disk.

11.2.9.4 Batch Files

A batch file is a file that contains a number of DOS commands, each of which
could be run individually from the command prompt. By putting them into a batch
file, they can be run as a group by simply running the batch file. Note that the

Lesson No. 1 Intro. to Operating System 276

commands execute in the order they appear in the batch file, and that anything
that causes a command to halt will also halt the batch file.

You create a batch file by using an ASCII text editor, such as DOS EDIT, or
Windows Notepad. When you have created the batch file, you save it with a file
name, and give it the extension *.bat. Note that you must not use a name that is
the same as any DOS commands or any other program or utility you are likely to
run. If you use a DOS command name, trying to run your batch file will not work
because the DOS command will execute first. If your name matches some other
program or utility, you may never be able to run that program again because your
batch file will run before the program runs. So pick something that is not likely to
match any command name or program file name.

Virtually all internal and external commands can be used in a batch file. The few
exceptions are the commands that are intended only for configuration which are
used in the CONFIG.SYS file. Examples of these include BUFFERS, COUNTRY,
DEVICE, etc.

When you create a batch file, you are beginning to write a program, essentially.
DOS batch files may not have the power of a structured programming language,
but they can be very handy for handling quick tasks. The one good habit for any
programmer is to put comments in the program that explain what the program is
doing. To do so place REM at the beginning of a comment line. The OS will then
ignore that line entirely when it executes the program, but anyone who looks at
the "source code" in the batch file can read your comments and understand what
it is doing.
To have a batch file executed automatically every time MS-
DOS starts, create an AUTOEXEC.BAT file in the root directory.
Common commands used in batch file are:
(a) CALL - Call another batch file with parameters.
(b) ECHO - Display command names or messages as commands are executed

Lesson No. 1 Intro. to Operating System 277

 from a batch file.
(c) FOR - Execute an MS-DOS command iteratively for each file in a set of files.
(d) GOTO - Execute a command in a batch file other than the next in sequence.
(e) IF - Test a condition and execute a command in a batch file, depending on the
result.

11.2.9.5 Working with Disks

Formatting disks with the FORMAT command prepares them to be read or written
by MS-DOS routines. The FORMAT command simply calls the device driver, which
contains the code to perform the formatting of its particular device. Newly installed
device drivers must contain the code to format the devices they control.
A volume is a floppy disk or a partition of a fixed disk. Volumes are divided into
logical sectors. Logical sectors reside on parts of physical tracks. The tracks are
determined by the various fixed positions of the read-write heads. Sectors are
grouped into clusters; the file allocation table chains the clusters to define the
contents of the file. Each volume has an identifying volume label. Volume labels
may be created, modified, or deleted with the LABEL command. Each logical
volume includes five control areas and a files area as follows:
(a) Boot sector
(b) Reserved area
(c) File allocation table #1
(d) File allocation table #2
(e) Root directory
(f) Files area

The boot sector contains all the information MS- DOS needs to interpret the structure of the
disk. The FAT allocates disk clusters to files, each cluster may contain many sectors, but this
number is fixed when the disk is formatted and it must be a power of 2. A file's clusters are
chained together in the FAT. There is an extra copy of the FAT is for integrity and
reliability. The root directory organizes the files on the disk into directories and

Lesson No. 1 Intro. to Operating System 278

subdirectories, The files area is where the files' information is stored. File space is allocated
one cluster at a time as needed.

The FORMAT command causes the following information to be placed in the BIOS
parameter block (BPB) in the boot sector:
i. sector size in bytes
ii. sectors on the disk
iii. sectors per track
iv. cluster size in bytes
v. number of FATs
vi. sectors per FAT
vii. number of directory entries
viii. number of heads
ix. hidden sectors
x. reserved sectors
xi. media identification code.

The boot sector is logical sector 0 of every logical volume and is created when the
disk is formatted. It contains OEM identification, the BIOS parameter block, and
bootstrap loader. When the system is booted, a ROM bootstrap program reads in
the first sector of the disk (this contains the disk bootstrap) and transfers control to it.
The boot strap loader reads MS-DOS BIOS to memory and transfers control to it.

CHKDSK compares the two FATs to make sure they are identical. MS-DOS
maintains copy of the FAT in memory to be able to search it quickly. Each cluster
entry in the FAT contains codes to indicate the following:

(a) The cluster is available.

(b) The cluster is in use.
(c) The cluster is bad (i.e., it has a bad sector); CHKDSK reports these.
(d) The cluster is a file's last one.

Lesson No. 1 Intro. to Operating System 279

(e) A pointer (cluster number) to the file's next cluster. (The directory points to the
file's first cluster.)

Clusters are allocated to files one at a time on demand. Because of the continuing addition,
modification, and deletion of files, the disk tends to become fragmented with free clusters
dispersed throughout the disk. Clusters are allocated sequentially, but in-use clusters are
skipped. No attempt is made to reorganize the disk so that files would consist of sequential
clusters, but this can be accomplished via the MS-DOS commands.

MS-DOS uses memory buffers with disk input/output. With file handle calls, the
location of the buffer may be specified in the call. With FCB calls, MS-DOS uses a
preset buffer called the disk transfer area, which is normally in a program's program
segment and is 118 bytes long.

11.2.9.6 Working with Directories

Initial versions of MS-DOS had a simple linear directory listing all files. Version 2.0
incorporated a hierarchical file structure with directories and subdirectories. A root
directory and subdirectories form the hierarchical structure, and directory entries
describe individual files.

The 32-byte directory entry contains the following:

i. File name (8 bytes) Extension (3 bytes)

ii. File attribute byte (1 byte)
iii. Reserved (10 bytes)
iv. Time created or last updated (2 bytes)
v. Date created or last updated (2 bytes)
vi. Starting cluster (2 bytes)
vii. File size (4 bytes)
The first byte of the file name may contain special codes
indicating the following:

Lesson No. 1 Intro. to Operating System 280

 i. A directory entry that has never been used
ii. The entry is a subdirectory
iii. File has been used, but is now erased.

The file attribute byte may indicate the following:

i. Read-only file (an attempt to open the file for writing or deletion will fail)
ii. Hidden file (excluded from normal searches)
iii. System file (excluded from normal searches)
iv. Volume label (can exist only in the root directory)
v. Subdirectory (excluded from normal searches)
vi. Archive bit (set to "on" whenever a file is modified)

Each disk is formatted to have at least one directory, but

there can be many directories. A directory may include other
directories. Formatting creates an initial root directory.
Each fixed disk partition has its own root directory. With
versions 2.0 and higher, the root directory may contain
directories in addition to files. Directories may contain one
or more file entries or more directories, thus generating a
hierarchical file system directory structure. The volume
label, if present, appears in the root directory.

MS-DOS function calls are available to search directories for files in a hardware
independent manner. FCB functions require that files be specified by a pointer to an
unopened FCB; these functions do not support the hierarchical file system. With the
file handle function request, it is possible to search for a file by specifying an ASCII
string; it is possible to search within any subdirectory on any drive, regardless of the
current subdirectory, by specifying a pathname.

Lesson No. 1 Intro. to Operating System 281

11.2.9.7 Managing Files
Data is organized into files whether that data is in memory or on disk. Files may be
data files or executable files. A file is simply a string of bytes; no record structure is
assumed. Applications impose their own record structure on the string of bytes MS-
DOS requires only a pointer to the data buffer, and a count of the number of bytes to
be read or written in order to do I/O. Executable files must be in either the .COM
format or the .EXE format. Object modules are maintained in the Intel Corporation
object-record format.

Files may be accessed with FCB calls or file handle calls. File handle calls are
designed to work with a hierarchical file system. File handle calls are preferable, but
FCB calls are provided for compatibility with previous versions of DOS. File handle
calls support record locking and sharing. Users interested in writing programs that
will be compatible with future versions of MS-DOS should use handle calls rather
than the FCB calls.

The FCB function requests are as follows:

i. Open file with FCB

ii. Close file with FCB
iii. Delete file with FCB
iv. Read block sequentially
v. Write block sequentially
vi. Create file with FCB
vii. Rename file with FCB
viii. Read block directly
ix. Write block directly
x. Get file size with FCB
xi. Set relative record field
xii. Read multiple blocks directly
xiii. Write multiple blocks directly

Lesson No. 1 Intro. to Operating System 282

The FCB data structure occupies a portion of the application's memory and contains bytes of
data as follows:

i. Drive identifier (1 byte) (0=default drive, 1=drive A, 2=drive B, and so on)

ii. File name (8 bytes)
iii. Extension (3 bytes)
iv. Current block number (2 bytes) (for sequential reads and writes)
v. Record size (2 bytes) (MS-DOS sets to 128, but this may be changed.)
vi. File size (4 bytes)
vii. Date created/updated (2 bytes)
viii. Time created/updated (2 bytes)
ix. Reserved (8 bytes)
x. Current record number (1 byte) (for sequential reads and writes)
xi. Relative record number (4 bytes)

A file handle is a 16-bit integer (usually 0 to 20) created

by MS-DOS and returned to a program that creates or opens a
file or device. A file may be opened with a handle by using a
pathname and attribute. The program saves the handle and uses
it to specify the file in future operations on the file. An
FCB-like data structure is built by MS-DOS for the file, but
this is strictly controlled by the operating system. The file
handle function calls are as follows:

(a) Create file

(b) Open file
(c) Close file
(d) Read from file or device
(e) Write to file or device
(f) Move file pointer

Lesson No. 1 Intro. to Operating System 283

(g) Duplicate file handle (creates a new handle that refers to the same file as an
existing handle)
(h) Match file handle (cause one handle to refer to the same file as another)
(i) Create temporary file
(j) Create file if name is unique (fails if a file with the same name already exists)

MS-DOS maintains a table that relates handles to files or

devices. Eight handles are normally available to programs, but
this can be increased to 20 via the CONFIG.SYS file. Five of
the handles are pre-assigned to standard devices as follows:

Handle Standard Device Device Name Description

0 Standard input device CON reads char from the keyboard
1 Standard output device CON writes characters to the VDU
2 Standard error device CON writes characters to the VDU
3 Standard auxiliary device AUX controls serial port I/O
4 Standard printer device PRN controls parallel port output
------------------------------------------------------------------------------------------------------------
A COM file consists of absolute machine code and data. It
is not relocatable. It loads faster than an .EXE file. An .EXE
file is a relocatable and executable file; it contains a
header with relocation information, and the actual load
module. The .EXE files separate the code and data portions.
This feature is designed to facilitate the sharing of “pure
procedures” among several tasks in evolving MS-DOS systems
that support concurrent tasks. The Microsoft Macro Assembler
translates source code files into object files with the .OBJ
extension.

11.2.9.8 Device drivers

Device drivers are programs that control input and output. They manage

Lesson No. 1 Intro. to Operating System 284

communication between the hardware devices and the operating system.
Regardless of the specific details of performing input/output on particular devices,
device drivers communicate in standard manner with the rest of the operating
system. Five resident device drivers control the standard devices. Other device
drivers may be installed at the command level as needed. The two types of devices
are character devices that handle one byte at a time, and block devices that can
access blocks of information on random access devices such as disks. Devices are
treated as files. The IO.SYS file consists of the five resident device drivers that form
the MS-DOS BIOS.

New or replacement device drivers may be installed by using the DEVICE command
followed by the driver's file name in a CONFIG.SYS file on the boot disk. Thus, the
input/output system may be reconfigured at the command level. MS-DOS has
character device drivers and block device drivers. CON, AUX, and PRN are
character device drivers. The CONFIG.SYS file may be used to notify the operating
system of hardware device changes such as additional hard disks, a clock chip, a
RAM disk, or additional memory.

A device driver is composed of a device header, a strategy routine, and an interrupt

routine. The device header contains a pointer to the next device driver in the chain
of device drivers in IO.SYS, and it points to the strategy routine and the interrupt
routine. The request header is the data structure that MS-DOS uses to give the
driver the information necessary to perform a requested input/output operation. The
strategy routine saves a pointer to the request header data structure. The interrupt
routine performs the I/O, passes status and completion information back to MS-DOS
in the request header. Each driver has a strategy routine entry point and an interrupt
routine entry point. For asynchronous I/O, the strategy routine is called upon to
enqueue a request and return to the caller quickly. The interrupt routine then
performs the I/O when it can.

A filter is a program that processes input data in some particular way to produce

Lesson No. 1 Intro. to Operating System 285

output directed to some file or device. Three filters are built in to MS-DOS. SORT
sorts text data. FIND searches for a character string. MORE displays one screen of
data at a time. Other filters may be created. Filters may be piped together (with the |
symbol) so that the output of one forms the input to another. Inputs or outputs may
be redirected.

Background programs are dormant until they are activated by a signal from the
keyboard. A hot key, or combination of keys, signals a program to take control of
keyboard input.

11.2.9.9 Memory Management

Memory is organized as follows in MS-DOS (from low memory to high memory
locations):

i. Interrupt vector table

ii. Optional extra space (used by IBM for ROM data area
iii. IO.SYS
iv. MSDOS.SYS
v. Buffers, control areas, and installed device drivers
vi. Resident part of COMMAND.COM
vii. External commands or utilities (.COM and .EXE files are loaded here)
viii. User stack for .COM Files (256 bytes)
ix. Transient part of COMMAND.COM

The interrupt vector table contains the addresses of the

interrupt handler routines. IO.SYS is the basic input/output
system; it is the MS-DOS/hardware interface. MSDOS.SYS
contains most of the interrupt handlers and function requests.
The resident part of COMMAND.COM contains certain interrupt
handlers and the code that reloads the transient part of
COMMAND.COM as needed; the transient part of COMMAND.COM

Lesson No. 1 Intro. to Operating System 286

includes the batch processor, the internal commands, and the
command processor.

MS-DOS begins the user's program segment in the lowest address free memory.
The program segment prefix (PSP) occupies the first 256 bytes of the program
segment area. The PSP points to various memory locations the program requires as
it executes.

MS-DOS creates a memory control block at the start of each memory area it
allocates. This data structure specifies the following:

i. the size of the area

ii. the program name (if a program owns the area)
iii. a pointer to the next allocated area of memory

MS-DOS may allocate a new memory block to a program, free a memory block, or
change the size of an allocated memory block. If a program tries to allocate a
memory block of a certain size, MS-DOS searches for an appropriate block. If such
a block is found, it is modified to belong to the requesting process. If the block is too
large, MS-DOS parcels it into an allocated block and a new free block. When a block
of memory is released by program, MS-DOS changes the block to indicate that it is
available. When a program reduces the amount of memory it needs, MS-DOS
creates a new memory control block for the memory being freed. The first memory
block of a program always begins with program segment prefix. Normally when a
program terminates, its memory is released. The program can retain its memory by
issuing function 31, TERMINATE BUT STAY RESIDENT.

11.2.9.10 Controlling Program Flow

COMMAND.COM uses the EXEC function to load and execute program files. A
program issuing EXEC causes MS-DOS to allocate memory, write a program
segment prefix, loads the new program, and transfer control to it. The calling

Lesson No. 1 Intro. to Operating System 287

program is the parent program; the called program is the child program. A child
program may use EXEC to load and execute its own child programs. Each child
automatically inherits its parent's active handles, so it can access its parent's active
files. MS-DOS allows programs to load and execute overlays.

MS-DOS handles hardware interrupts from devices and software interrupts caused
by executing instructions. An interrupt vector table causes control to transfer to the
appropriate interrupt handlers. Users may write their own interrupt handlers. The
Control-C interrupt normally terminates the active process and returns control to the
command interpreter.

Certain types of critical errors may occur that prevent a program from continuing.
For example, if a program tries to open a file on a disk drive without a disk, or a disk
drive whose door is open, a critical error is signaled. MS-DOS contains a critical
error handler, but user users may wish to provide their own routines for better
control over errors in specific situations.

11.2.10 SYSTEM CALLS - THE PROGRAMMER'S VIEW

MS-DOS provides the programmer with a rich collection of system calls. These calls
accomplish various manipulations that would be extremely time-consuming and
difficult to program and debug if the operating system were not available. The calls
manipulate files, disks and directories; return information about the hardware and
software environment; and obtain additional memory and relinquish memory no
longer needed. By using MS-DOS calls, programmers realize several advantages.
Applications programming becomes faster and less error prone, and applications
that use system calls are easier to upgrade to new versions of the operating system.
Microsoft is committed to providing upward compatibility for programs that use
system calls properly. The next several sections enumerate most of MS-DOS's
system calls.

Lesson No. 1 Intro. to Operating System 288

11.2.10.1 Working with Disks
¾ Check Status of Verify Flag - Returns the value of the Verify Flag, which
determines whether or not DOS verifies write operations to the disk.
¾ Get Default Drive Data - Returns information about the current default drive,
namely the number of sectors per cluster, information about the type of disk in
the current default drive, bytes per sector, and clusters per drive.
¾ Get Disk Transfer Address - Returns the pointer to the current Disk Transfer
Area (the DT A is the buffer that DOS uses to transfer data to-and-from the disk).
¾ Get Free Disk Space - Determines how many bytes are free on the specified
drive. Get Specified Drive Data-Returns information about the specified drive,
namely the number of sectors per cluster, the offset to the FAT ID byte, the
number of bytes per sector, and the number of clusters per drive.
¾ Identify Current Drive - Returns the default drive ID.
¾ Read Absolute Disk Sectors - Reads the specified number of disk sectors from
the disk drive, starting at a given location.
¾ Reset Disk - Flushes all buffers to the media and marks them "free."
¾ Select Drive - Sets the default drive to the specified drive.
¾ Set Disk Transfer Address - Sets the Disk Transfer Address to the specified
value.
¾ Set or Reset Verify Flag - Sets the value of the Verify Flag to a specified value.
The Verify Flag determines whether or not DOS verifies write operations to the
disk.
¾ Write Absolute Disk Sectors - This interrupt handler writes the specified number
of sectors to the disk drive, starting at a given location.

11.2.10.2 Working with Directories

¾ Change Current Directory - Changes the current directory to the directory in the
user's pathname.
¾ Create Directory - Creates a directory using the name in the user's pathname.
¾ Get Current Directory - Returns the pathname of the current directory on a

Lesson No. 1 Intro. to Operating System 289

 specified drive.
¾ Remove Directory - Removes the directory specified in the user's pathname.

11.2.10.3 Managing Files

¾ Change File Attributes - Gets or sets the attributes of the file specified in the
user's pathname.
¾ Close File - Closes the specified handle.
¾ Close File with FCB - Closes the file pointed to by the user's FCB.
¾ Create File-Creates and assigns a handle to the file in the user's pathname.
¾ Create File if Name is Unique - Creates a new file if a file by that name does not
exist in the specified directory.
¾ Create File with FCB - Creates a file in the current directory using the file name
in the user's FCB.
¾ Create Temporary File - Uses the clock device to provide a unique file name and
appends it to the pathname provided by the user.
¾ Delete File - Deletes the directory specified in the user's pathname.
¾ Delete File with FCB - Deletes a file named in the user's FCB, and removes the
specified file from the directory buffer for the drive specified in the user's FCB.
¾ Duplicate File Handle - Copies the handle of an open file.
¾ Find First Matching File - Searches the specified' or current directory for the first
entry that matches the user's pathname.
¾ Find First Matching File with FCB - Searches the current directory for the first file
name matching the file name contained in the user's FCB.
¾ Find Next Matching File - Searches for the next entry that matches the name and
attributes specified in a previous Find First Matching File.
¾ Find Next Matching File with FCB - Searches for the file named in the user's
FCB. The file name was used originally to conduct the "Find First Matching File
with FCB" function.
¾ Get File Size with FCB - Returns the size of the file specified in the FCB.
¾ Lock or Unlock Region of File - Lock (deny access to) or unlock a specified
region of a file.

Lesson No. 1 Intro. to Operating System 290

¾ Match File Handle - Forces a specified handle to refer to the same file as another
handle already associated with an open file.
¾ Move File Pointer - Moves the read/write pointer of the open file associated with
the specified handle.
¾ Open File - Opens a file and assigns it a handle using the file name specified in
the user's pathname.
¾ Open File with FCB - Opens the file named in the user's FCB.
¾ Parse File Name - Searches the specified string for a valid name. If the file name
is valid, the function returns a pointer to an unopened FCB.
¾ Read Block Directly - Reads the record pointed to by the user's FCB Relative-
Record field.
¾ Read Block Sequentially - Reads the next block from the file named in the user's
FCB.
¾ Read from File or Device - Reads from the specified open file handle.
¾ Read Multiple Blocks Directly - Reads a specified number of records from the file
starting at the block named in the user's FCB relative-record field.
¾ Rename File - Renames the file specified in the user's pathname by changing
the directory entry.
¾ Rename File with FCB - Renames a file in the current directory using the two file
names in the user's modified FCB; the first name is the file in the current
directory, and the second is the new name for the file.
¾ Set Relative Record Field - Sets the relative record field in the user's FCB to the
value' and of the current record field in the FCB. The relative record field is used
by the Read (or Write) Block Directly functions.
¾ Write Block Directly - Writes the record pointed to by the relative record field (in
the user's FCB), to the file named in the user's FCB.
¾ Write Block Sequentially - Writes to the file named in the user's FCB. The FCB
also contains the current block and the record to be written.
¾ Write Multiple Blocks Directly - Writes a specified number of records to the file
named in the user's FCB. The FCB also contains the current relative record
number where writing will start.

Lesson No. 1 Intro. to Operating System 291

11.2.10.4 Managing Input and Output
¾ Check Keyboard Status - Returns a code to indicate if characters are available from
standard input.
¾ Display Character - Sends the specified character to standard output. Display
String-Sends the specified string to standard output.
¾ I/O Through Standard Device - Either reads or sends a character to standard
output, depending on the specified value.
¾ Input from Auxiliary Device - Waits for a character to be read from standard al
and returns the character.
¾ IOCTL - The IOCTL function is a group of 16 related sub functions that
manipulates character and block-device control data. These sub functions can
be used to set or reset the attributes of a certain device, send or receive control
data with either character or block devices, check the status of a device, or
check if a device or file handle has been redirected with Microsoft Networks.
¾ Output to Auxiliary Device - Sends the specified character to the standard
auxiliary device.
¾ Print Character - Sends the specified character to the standard printer device.
¾ Read Keyboard - Waits for a character to be read from the standard input device
and returns the character.
¾ Read Keyboard after Clearing Buffer - First clears the input buffer and then
executes the specified function.
¾ Read Keyboard and Echo - Waits for a character to be read from the standard
device, then echoes it to the standard output device and returns the character.
¾ Read Keyboard to Buffer - Gets a string (terminated with a carriage return) from
standard input device and puts it in the user-specified buffer.
¾ Read Keyboard without Echo - Waits for a character to be read from the standard put
device and returns the character.

11.2.10.5 Managing Memory

¾ Allocate Memory - Attempts to allocate the specified amount of memory to

Lesson No. 1 Intro. to Operating System 292

 current process. DOS coalesces disowned blocks, if necessary, to provide
enough for the user's requested block size.
¾ Change Memory Allocation Block Size - Puts the user's specified block back into
memory pool, coalesces it with other unowned blocks, and then tries to allocate
requested block size; the function returns the size of the resultant block to the
user.
¾ Free Allocated Memory - Releases the specified block of memory.
¾ Get or Set Memory Allocation Strategy - Gets or sets the strategy used by MS-
DOS allocate memory when a process requests it; the three different strategies
are first best fit, and last fit.

11.2.10.6 Controlling Program Flow

¾ Abort Program - Aborts the current process.
¾ Check Control-C - Either returns or sets DOS's internal Control-C checking flag;
the Control-C flag is used by DOS to determine before which function requests it
should check for a Control-C.
¾ Create New Program Segment Prefix - Creates a new program segment prefix
(PSP) at the specified address. (This is an obsolete function except for
compatibility with pre-2.0 versions of DOS.)
¾ Execute Program - Loads and executes a program.
¾ Get Current Program Segment Prefix - Returns the contents of the DOS "Current
PSP" variable. (A PSP is a block of memory that immediately precedes a .COM
or .EXE program.)
¾ Get Extended Error Codes - Returns an extended error code for the function that
immediately preceded it. The extended error code contains details about the type
and location of the error, and the recommended course of action to take when
the error occurs.
¾ Get Interrupt Vector - Returns the address of the specified interrupt handler.
¾ Get Return Code of Child Process - Returns the exit code specifying the reason
for termination, such as normal, Control-C, or critical device error.
¾ Set Interrupt Vector - Sets the address of the specified interrupt in the Interrupt

Lesson No. 1 Intro. to Operating System 293

 Vector Table.
¾ Terminate But Stay Resident - Returns control to the parent process (usually
COMMAND.COM), but keeps the current process resident after it terminates.
¾ Terminate But Stay Resident - Terminates a program of up to 64K in size, but
keeps it in memory. (This interrupt is obsolete.)
¾ Terminate on Control-C - Used by DOS to deal with a Control-C being typed on
the keyboard.
¾ Terminate on Fatal Error - Used by DOS if a critical error occurs during I/O.
¾ Terminate Process - This is the proper way for a program to terminate when
finished. Terminate Program - Terminates the current process and returns
control to the parent process. (This is an obsolete interrupt.)
¾ Terminate Program at Address - Used by DOS to transfer control after a process
terminates.

11.3 Keywords
Batch file: It is an executable file that contains a group of commands that the user
wants to execute in sequence.
Device drivers: These are programs that control input and output.
Internal commands: These are commands, such as COPY and DIR that can be
handled by the COMMAND.COM program.

11.4 Summary
MS-DOS is a single-user, single-process operating system. Due to confinement of
device-independent code into lone layer, porting of MS-DOS is theoretically reduced
to writing of the BIOS code for the new hardware. Although early versions of MS-
DOS show resemblance to the CP/M operating system, later releases of MS-DOS
have Unix-like features. At the command level, MS-DOS provides a hierarchical file
system, 1/0 redirection, pipes and filters. User-written commands can be invoked in
the same way as standard system commands, thus giving the appearance of
extending the basic system functionality.

Lesson No. 1 Intro. to Operating System 294

MS-DOS provides both device-dependent and device-independent versions of
system calls for input/output and file manipulation. Being a single-user system, MS-
DOS provides only rudimentary file protection and access control. Disk space is
allocated in terms of clusters of consecutive sectors. A variant of chaining that
allows for relatively fast random access to files is used for keeping track of both file
blocks and free space.

11.5 SELF-ASSESSMENT QUESTIONS (SAQ)

1. Discuss history and different versions of MS-DOS.
2. Discuss various internal and external MS-DOS commands.
3. Discuss memory management in MS-DOS.
4. What do you understand by batch files? Discuss some common
commands used in it.
5. Differentiate between CONFIG.SYS and AUTOEXEC.BAT files.
6. What are the limitations of DOS? Explain.
7. What do you understand by Terminate and Stay Resident programs
(TSR)? Explain.

11.6 SUGGESTED READINGS / REFERENCE MATERIAL

1. The Design of the UNIX Operating System, Bach M.J., PHI, New Delhi, 2000.
2. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
3. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.
4. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
5. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.

Lesson No. 1 Intro. to Operating System 295

 6. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company Ltd.,
New Delhi, 2002.

Lesson No. 1 Intro. to Operating System 296

Lesson Number: 12 Writer: Dr. Rakesh Kumar
Case Study of MS-WINDOWS NT Vetter: Prof. Dharminder Kr.

12.0 Objectives
The objectives of this lesson are:
(a) To provide a brief overview of the history of Windows operating system.
(b) To discuss the key features of Windows NT operating system.

12.1 Introduction
Microsoft Windows is the name of several families of software operating systems
by Microsoft. Microsoft first introduced an operating environment named
Windows in November 1985 as an add-on to MS-DOS in response to the
growing interest in graphical user interfaces (GUI). The term Windows
collectively describes several generations of Microsoft (MS) operating system
(OS) products categorized as follows:
12.1.1 16-bit operating environments
The early versions of Windows were often thought of as just graphical user
interfaces, because they ran on top of MS-DOS and used it for file system
services. However even the earliest 16-bit Windows versions have many typical
operating system functions, such as having their own executable file format and
providing their own device drivers for applications. Unlike MS-DOS, Windows
allowed users to execute multiple graphical applications at the same time,
through cooperative multitasking. Finally, Windows implemented an elaborate,
segment-based, software virtual memory scheme which allowed it to run
applications larger than available memory: code segments and resources were
swapped in and thrown away when memory became scarce, and data segments
moved in memory when a given application had left processor control, typically
waiting for user input. 16-bit Windows versions include Windows 1.0, Windows
2.0 and Windows/286.

Lesson No. 1 Intro. to Operating System 297

 12.1.2 Hybrid 16/32-bit operating environments
Windows/386 introduced a 32-bit protected mode kernel and virtual machine
monitor. For the duration of a Windows session, it created one or more virtual
8086 environments and provided device virtualization for the video card,
keyboard, mouse, timer and interrupt controller inside each of them. The user-
visible consequence was that it became possible to preemptively multitask
multiple MS-DOS environments in separate Windows. Windows applications
were still multi-tasked cooperatively inside one of such real-mode environments.
Windows 3.0 and Windows 3.1 improved the design, because of virtual memory
and loadable virtual device drivers which allowed them to share arbitrary devices
between multitasked DOS windows. Because of this, Windows applications could
now run in 16-bit protected mode, which gave them access to several megabytes
of memory and removed the obligation to participate in the software virtual
memory scheme. They still ran inside the same address space, where the
segmented memory provided a degree of protection, and multi-tasked
cooperatively
12.1.3 Hybrid 16/32-bit operating systems
With the introduction of 32-bit Windows, Windows finally stop relying on DOS for
file management. Windows 95 introduced Long File Names. The most important
novelty was the possibility of running 32-bit multi-threaded preemptively
multitasked graphical programs. However, the necessity of keeping compatibility
with 16-bit programs meant the GUI components were still 16-bit only and not
fully reentrant, which resulted in reduced performance and stability.
Microsoft's next OS was Windows 98; there were two versions of this. In 2000,
Microsoft released Windows ME, which used the same core as Windows 98 but
adopted the visual appearance of Windows 2000, as well as a new feature called
System Restore, allowing the user to set the computer's settings back to an
earlier date.
12.1.4 32-bit operating systems
This family of Windows systems was designed for higher-reliability business use.
The first release was Windows NT 3.1, followed by NT 3.5, NT 3.51, and NT 4.0.

Lesson No. 1 Intro. to Operating System 298

Microsoft then moved to combine their consumer and business operating
systems. Their first attempt, Windows 2000, failed to meet their goals, and was
released as a business system. The home consumer edition of Windows 2000,
codenamed "Windows Neptune," ceased development and Microsoft released
Windows Me in its place. Eventually "Neptune" was merged into their new
project, Whistler, which later became Windows XP.
12.1.5 64-bit operating systems
Windows NT included support for several different platforms before the x86-
based personal computer became dominant in the professional world. Versions
of NT from 3.1 to 4.0 supported DEC Alpha and MIPS R4000, which were 64-bit
processors, although the operating system treated them as 32-bit processors.
With the introduction of the Intel Itanium architecture, Microsoft released new
versions of Windows 2000 to support it. Itanium versions of Windows XP and
Windows Server 2003 were released at the same time as their mainstream x86 (32-
bit) counterparts. On April 25, 2005, Microsoft released Windows XP Professional
x64 Edition and x64 versions of Windows Server 2003 to support the AMD64/Intel64
architecture. Microsoft dropped support for the Itanium version of Windows XP in
2005. Windows Vista is the first end-user version of Windows that Microsoft has
released simultaneously in 32-bit and x64 editions.

12.2 Presentation of contents

12.2.1 General Architecture
12.2.1.1 Extensibility
12.2.1.2 Portability
12.2.1.3 Reliability and Security
12.2.2 The Hardware Abstraction Layer (HAL)
12.2.2.1 Objects
12.2.2.2 Threads
12.2.2.3 Traps, Interrupts, and Exceptions
12.2.2.4 Thread Scheduling
12.2.3 Multiprocess Synchronization

Lesson No. 1 Intro. to Operating System 299

 12.2.3.1 The NT Executive
12.2.3.1.1 Object Manager

12.2.3.1.2 Process and Thread Manager

12.2.3.1.3 Virtual Memory Manager
12.2.3.1.4 Security Reference Manager
12.2.3.1.5 I/O Manager
12.2.3.1.6 Cache Manager
12.2.3.1.7 The Native API
12.2.4 NT Subsystems
12.2.4.1 Win 32 API: The Programmer’s View of NT
12.2.1 General Architecture
The product goals for Windows NT were that it should be an extensible, portable,
reliable, and secure OS for contemporary computers.
12.2.1.1 Extensibility
There are two dimensions to the extensibility aspect. The first relates to OS
configurations. A Windows NT machine can be configured for a workstation or a
server. In either configuration, the OS uses the same fundamental source code, but
different components are incorporated into each at compile time. This allows
Windrows NT to be optimised to perform best according to the way the machine will
be used-as a workstation or as a server-without building two different OSs.
The second, aspect of extensibility is in the way the OS software is structured.
Windows NT is designed using an extensible nucleus software model. In this
approach, only the most essential OS functions are implemented in a small nucleus
of code (microkernel). Additional mechanisms are then implemented on top of the
nucleus to define policy as needed. This approach has the advantage that key
mechanisms can be carefully designed and tested as one trusted subassembly that
can then be used to implement many different policies. This is a basic approach to
support the goals of good security and reliable operation. The NT Kernel provides
these essential low-level mechanisms as a layer of abstraction from the hardware
(see Figure 12.1). The NT Executive is designed as a layer of abstraction of the NT
Kernel. It provides specific mechanisms for general object and memory

Lesson No. 1 Intro. to Operating System 300

management, process management, file management, and device management.
Together, the NT Kernel and the NT Executive provide the essential elements of the
OS.
The NT Kernel and the NT Executive are combined into a single executable,

NTOSKRNL.EXE, before they are actually executed. NTOSKRNL.EXE also invokes

additional dynamically linked libraries (DLLs) whenever they are needed. Thus the

logical view of Windows NT is quite different from the way the executable actually

appears in memory. It is best to use the logical view for considering different aspects of

the OS, since that is the model under which it is designed.

APPLICATIONS APPLICATIONS APPLICATIONS

OTHER WIN 32 OTHER

SUBSYSTEMS SUBSYSTEMS SUBSYSTEMS

WIN-32

Figure 12.1 NT Organization

The next layer of abstraction of Windows NT is the subsystem layer. Subsystems
provide application portability for windows software. An NT Subsystem is a software
module that uses the services implemented in the Kernel and the Executive to
implement more abstract services, especially the services offered by some target
OS. For example, Version 4.0 has a POSIX subsystem that executes on top of the
Kernel and the Executive that makes Windows NT look like POSIX; such
subsystems are called environment subsystems. Other subsystems implement
specialized services such as the security subsystem. All subsystems (and all
application programs that use the subsystems) execute when the processor is in
user mode. Subsystems are the key component in allowing Microsoft to support
various computational models, such as the MS-DOS and Win 16 program models.
Application programs written to run on MS-DOS use the MS-DOS subsystem
interface. This subsystem provides the same API to the application, as does MS-
DOS, thereby allowing old MS-DOS programs to run on a Windows NT system.

Lesson No. 1 Intro. to Operating System 301

12.2.1.2 Portability
The portability aspect of Windows NT overlaps its extensibility. Subsystems allow
Windows NT to be extended to meet various application support requirement.
Microsoft has built various subsystems to implement OS personalities of interest to
their customers. Besides the MS-DOS subsystem, there are subsystems to support
Win 16 applications and POSIX programs, as well as a new Win32 subsystem. In
general, it is possible for software developers to implement any subsystem to satisfy
their general requirements for OS service; such a subsystem uses the
Executive/Kernel interface. Even so, the Win32 subsystem takes a special role in
Windows NT because it implements various extensions of the NT Executive that are
needed by all other subsystems; every subsystem relies on the presence of the
Win32 subsystem. While it is possible to add new environment subsystems to
Windows NT, and to omit most of them, the Win32 subsystem must always be
present.
Another aspect of portability is the ability to port Windows NT across different
hardware platforms. Microsoft's goal was to be able to reuse the Kernel, Executive,
and subsystems on new microprocessors as they became available without having
to rewrite the Kernel. Windows NT's designers carefully identified the things that
were common across a wide set of microprocessors and the things that were
different. This allowed them to create a hardware abstraction layer (HAL) software
module to isolate the Kernel from hardware differences. The HAL is responsible for
mapping various low-level, processor-specific operations into a fixed interface that is
used by the Windows NT Kernel and Executive. The HAL also executes with the
processor in supervisor mode.
The HAL, Kernel, and Executive are supervisor-mode software that collectively export

an API that is used by subsystem designers (but not by application programmers).

Environment subsystem designers choose a target API (such as the Win 16) and then

build a subsystem to implement the API using the supervisor portion of Windows NT.

Microsoft has even chosen its own preferred API-the Win32 API-which is also the API

Lesson No. 1 Intro. to Operating System 302

for the Win32 subsystem. Windows NT application programs are written to work on the

Win32 API rather than on the interface to NTOSKRNL.EXE.

12.2.1.3 Reliability and Security

Separating the HAL, Kernel, Executive, and subsystem functionality from one
another, thus eliminating unnecessary interactions, supports reliability. Windows NT
is designed to meet standard requirement for trusted OSs. According to Solomon, in
1995 United Kingdom Information Technology Security Evaluation and Certification
Board certified Windows NT at the C2 level by the United States National Computer
Security Centre and in 1996 at F-C2/E3 level. Much of the security mechanism is
implemented in a Security Subsystem that depends upon the Security Reference
Manager in the Executive.
12.2.2 The Hardware Abstraction Layer (HAL)
The HAL is a low-level software module that translates critical hardware behaviours
into a standardized set of behaviours. The HAL functions are exported through a
kernel-mode DLL, HAL.DLL. The OS calls functions in HAL.DLL when it needs to
determine the way the host hardware behaves. This allows the Windows NT code to
call a HAL function everywhere a hardware specific address is needed. For
example, device interrupts usually have addresses determined by the
microprocessor architecture, and they differ from one microprocessor to another.
The HAL interface allows Windows NT to reference the interrupt addresses via
functions rather than by using the hardware addresses directly.
The HAL implementation for any specific microprocessor provides the appropriate
hardware-specific information via the corresponding function on the HAL API. This
means that it is possible to use the same source code on a Digital Equipment Alpha
processor as is used on an Intel Pentium processor. It also means that it is possible
to create device drivers for Windows NT that will also work without change in
Windows 9x.
The use of the HAL is transparent above the Executive/Kernel interface. Subsystem
and application programmers are generally unconcerned with the type of processor

Lesson No. 1 Intro. to Operating System 303

chip in the computer. Windows NT provides a fixed set of services independent of
the hardware platform type.
The NT Kernel creates the basic unit of computation and provides the foundation for
multitasking support. It does so without committing to any particular policy/strategy
for process management, memory management, file management, or device
management. To appreciate the level of support the Kernel provides, think of the
Kernel as offering a collection of building components such as wheels, pistons,
lights, and so on, that could be used to build a sports car, a sedan, a sports utility
vehicle, or a truck. Similarly, the Kernel's clients can combine the components to
build a compound component that defines a policy for how the low-level components
are used.
The Kernel provides objects and threads on top of the HAL and the hardware.
Software that uses the Kernel can be defined using objects and threads as
primitives, that is, these abstractions appear to Kernel client software as natural
parts of the hardware. To implement objects and threads, the Kernel must manage
the hardware interrupts and exceptions, perform processor scheduling, and handle
multiprocessor synchronization.

12.2.2.1 Objects
The NT Kernel defines a set of built-in object types. Some kernel object types are
instantiated by the Kernel itself to form other parts of the overall OS execution
image. These objects collectively save and manipulate the Kernel's state. Other
objects are instantiated and used by the Executive, subsystems, and application
code as the foundation of their computational model. That is, Windows NT and all of
its applications are managed at the Kernel level as objects.
Kernel objects are intended to be fast. They run in supervisor mode in a trusted
context, so there is no security and only limited error checking for Kernel objects, in
contrast to normal objects, which incorporate these features. However, Kernel
objects cannot be manipulated directly by user-mode programs, only through
function calls. Kernel objects are characterized as being either control objects or
dispatcher objects.

Lesson No. 1 Intro. to Operating System 304

A) Control Objects
Control objects implement mechanisms to control the hardware and other Kernel
resources. When an application program creates a new process, it requests that the
Kernel create a type of control object, a new process object; the OS returns a
handle to the object after it has been created. The application refers to the object by
using the handle. When the application program manipulates the process, it is
manipulating the underlying kernel process object. There are other control objects.
The asynchronous procedure call (APC) object is an object that can be used to
interrupt another thread and cause it to execute a specific procedure. An interrupt
object is an object created to match up to each interrupt source so that when a
designated interrupt occurs, the corresponding object will receive a message. A
profile object is an object that can be used to monitor the amount of time a thread
spends executing different parts of the code. Other control objects exist to handle
power failure when it occurs, to check to see if power has failed, to do power
management, and so on.
B) Dispatcher Objects
Dispatcher objects are used to implement threads along with their scheduling and
synchronization operations. Each dispatcher object has built-in characteristics that
are used to support user-level synchronization. A process object creates a
computational abstraction that can have an address space and a set of resources.
However, in Windows NT the process object cannot execute. A dispatcher object
called a thread object is the active element of the computation abstraction; it has its
own stack and can execute within a process. Whenever any application program is
to be executed, it must have an associated process object and a thread object.
Other dispatcher objects are used primarily to implement one form or another of
synchronization. While a single synchronization primitive as this would be sufficient
(ultimately, all synchronization is based on the Kernel spin-lock described shortly),
the Executive will implement a set of variants to simplify the use of dispatcher
objects for synchronization.
12.2.2.2 Threads

Lesson No. 1 Intro. to Operating System 305

A thread is an abstraction of computation. A Windows NT process object defines an
address space in which one or more threads can execute, and each thread object
represents one execution within the process. In a UNIX environment, there can be
only one thread executing in each address space. A UNIX process does not
differentiate between the address space concepts of a process object and the
execution aspects of a thread object. In the Windows NT environment it is common
to have more than one thread - a logical path traversal through the code in an
address space - executing in a process. The separation of the thread concept from
the rest of the process concept has been done so that it is natural to think of several
different "threads of execution" within a single address space, all sharing the same
resources.
12.2.2.3 Traps, Interrupts, and Exceptions
In Windows NT terminology, the Kernel trap handler is responsible for reacting to
hardware interrupts and processor exceptions (such as system service call,
execution errors, and virtual memory faults). Whenever an interrupt or processor
exception is recognized by the hardware, the trap handler (see Figure 12.2)
moves into action. It is responsible for doing the following:
¾ Disabling interrupts
¾ Determining the cause of the interrupt or exception.
¾ Saving processor state in a trap frame
¾ Re-enabling interrupts
¾ Changing the processor to supervisor mode if required
¾ Dispatching specialized code, for example, an Interrupt Service Routine (ISR), a
DLL, an exception

Interrupt Exception dispatcher, or the

virtual memory
handler, to handle
Authentication ISR
the trap.
DLL
Determine

Exception
Dispatch Dispatcher

Lesson No. 1 Intro. to Operating System 306 VM Pager

 Figure 12.2 Trap Handler
In the case of an interrupt, the trap handler will normally run an ISR for the specific
interrupt. For exceptions, the trap handler might address the cause itself or invoke
the appropriate OS code to react to the exception.
As in all "system call interface" OS designs, supervisor-mode functions are invoked
by an application program when it executes an instruction that causes exception; in
many hardware sets, this is the trap instruction. The trap handler must be used to
call system functions, since the processor mode needs to be switched from user to
supervisor. Before the mode can be switched, the OS must be assured that the
code to be executed (while the hardware is in supervisor mode) is trusted code.
Therefore user programs are not allowed to link and call these functions directly.
Instead, they can be invoked only through the trap handle in Windows NT, the trap
handler uses a DLL, NTDLL.DLL, to authenticate the call and start the OS code; the
application links NTDLL.DLL into its address space and then calls entry points in the
DLL. These points are translated into traps (using the host hardware mechanism for
raising an exception) that cause the processor mode to be switched to supervisor
mode and a secure call to be made on the OS code.
Interrupts are used to allow a device to notify the OS when the device completes an
operation. Windows NT's Interrupt management generally follows the same design
that other OSs have used for a number of years. The device’s driver initiates each
device operation. The thread initiating the operation might wait for the I/O call to
complete (said to be a synchronous I/O call) or continue running concurrently with
the I/O operation (said to be an asynchronous I/O call). Traditionally, the API does
allow the application thread to use asynchronous I/O, though asynchronous I/O is

Lesson No. 1 Intro. to Operating System 307

fully supported in Windows NT. The API used with Windows NT extends the normal
C routines so that application program can use asynchronous I/O operations.
In either the synchronous or asynchronous case, the processor continues to execute
software concurrent with the device operation - the calling thread's code, in the
asynchronous case, or another thread's code, in the synchronous case. The device
will eventually signal the processor that it has completed the I/O operation by raising
an interrupt. This causes the trap handler to run and to determine which device has
completed and then to run an ISR that will finish the house-keeping related to
completing the I/O operation. Each time the user moves the mouse or types a key or
information arrives on a connected network, an interrupt is raised, the trap handler
runs, and an ISR is called to manage the incoming information.
12.2.2.4 Thread Scheduling
The Windows NT thread scheduler is a time-sliced, priority-based, pre-emptive
scheduler. The basic unit of processor allocation is a time quantum computed as a
multiple of the number of clock interrupts. On most Windows NT machines, the time
quantum ranges from about 20 to 200 milliseconds. Servers are configured to have
time quanta that are six times longer than for a workstation with the same processor
type. The scheduler supports 32 different scheduling queues. As in all multiple-level
queue schedulers, as long as there are threads in the highest-priority queue, then
only those threads will be allocated the processor. If there are no threads in that
queue, then the scheduler will service the threads in the second highest-priority
queue. If there are no threads ready to run in the second highest-priority queue, the
scheduler will service the third highest-priority queue, and so on.
There are three levels of queues:
¾ Real-time level, consisting of the 16 highest-priority queues
¾ Variable-level, consisting of the next 15 higher-priority queues
¾ System-level, consisting of the lowest-priority queue
The scheduler attempts to limit the number of threads that are entered into the real-
time queues, thereby increasing the probability that there will be little competition
among threads that execute at these high-priority levels. However, Windows NT is
not a real-time system and cannot guarantee that threads running at high priority will

Lesson No. 1 Intro. to Operating System 308

receive the processor before any fixed deadline. The highest-level queue processing
continues through the variable-level queues, down to the system-level queue. The
system-level queue contains a single "zero page thread" to represent an idle
system. That is, when there are no runnable threads in the entire system, it
executes the zero page thread until an interrupt occurs and another thread becomes
runnable. The zero page thread is the single lowest-priority thread in the system, so
it runs whenever there are no other runnable threads.
A thread's base priority is normally inherited from its process. The priority can also
be set with various function calls, provided the caller has the authority to set the
priority. The Win32 API model defines four priority classes:
¾ REAL TIME
¾ HIGH
¾ NORMAL
¾ IDLE
Each thread also has a relative thread priority within the class, any of the following:
¾ TIME CRITICAL
¾ HIGHEST
¾ ABOVE NORMAL
¾ NORMAL
¾ BELOW NORMAL
¾ LOWEST
¾ IDLE
Thus a thread could be in the HIGH class and operating at the ABOVE NORMAL

relative priority at one moment, but then be in the HIGH class and operating at the

BELOW NORMAL relative priority a little later. The thread’s class and the class’s

NORMAL relative priority define base priority. If the priority class is not REAL TIME,

then the thread's priority will be for one of the variable-level queues. In this case,

Windows NT might adjust priorities of threads in the variable level according to system

Lesson No. 1 Intro. to Operating System 309

conditions. Windows NT does not change the priority of a thread that has been placed in

the real-time levels.

The thread scheduler is also preemptive. This means that whenever a thread
becomes ready to run, it is placed in a run queue at a level corresponding to its
current priority. If there is another thread in execution at that time and that thread
has a lower priority, then the lower-priority thread is interrupted and the new, higher-
priority thread is assigned the processor. In a single-processor system, this would
mean that a thread could cause itself to be removed from the processor by enabling
a higher-priority thread. In a multiprocessor system, the situation can be subtler.
Suppose that in a two-processor system, one processor is running a thread at level
10 and the other is running a thread at level 4. If the level 10 thread performs some
action that causes a previously blocked thread to suddenly become runnable at level
6, then the level 4 thread will be halted and the new level 6 thread will begin to use
the processor that the level 4 thread was using.
12.2.3 Multiprocess Synchronization
Single-processor systems can support synchronization by disabling interrupts.
However, Windows NT is designed to also support multiprocessors, so the Kernel
must provide an alternative mechanism to ensure that a thread executing on one
processor does not violate a critical section of a thread on another processor. The
Kernel employs spinlocks by which a thread on one process can wait for a critical
section by actively testing a Kernel lock variable to determine when it can enter the
critical section, if the hardware supports the test-and-set instruction. Spinlocks are
implemented using the hardware. Spinlock synchronization is used only within the
Kernel and Executive. User-mode programs use abstractions that are implemented
by the Executive.
12.2.3.1 The NT Executive
The NT Executive builds on the Kernel to implement the full set of Windows NT
policies and services, including process management, memory management, file
management, and device management. Windows NT uses object-oriented

Lesson No. 1 Intro. to Operating System 310

technology but the NT Executive is designed and implemented at the source code
level as a modularised set of elements.
¾ Object Manager
¾ Process and Thread Manager
¾ Virtual Memory Manager
¾ Security Reference Manager
¾ I/O Manager
¾ Cache Manager
¾ LPC facility
¾ Runtime functions
¾ Executive support functions
12.2.3.1.1 Object Manager

The Executive Object Manager implements another object model on top of the
Kernel Object Manager, Whereas Kernel objects operate in a trusted environment,
Executive objects are used by other parts of the Executive and user-mode software
and must take extra measures to assure secure and reliable operation.
An Executive object exists in supervisor space, though user threads can reference
it. This is accomplished by having the Object Manager provide a handle for each
Executive object. Whenever a thread needs a new Executive object, it calls an
Object Manager function to create the object (in supervisor space), to create a
handle to the object (in the process's address space), and then to return the handle
to the calling thread.
Sometimes a second thread will want to use an Executive object that has already
been created. When the second thread attempts to create the existing object, the
Object Manager notes that the object already exists, so it creates a second handle
for the second thread to use to reference the existing Executive object. The two
threads share the single object. The Object Manager keeps a reference count of all
handles to an Executive object. When all outstanding handles have been closed, the
Executive object is deal located. Thus it is important for each thread to close each
handle it opens, preferably as soon as it no longer needs the handle.

Lesson No. 1 Intro. to Operating System 311

There is a predefined set of about is object types for the Object Manager. When an
object is created, it includes an object header (used by the Object Manager to
manage the object) and a body to contain type-specific information. The header
includes the following:
¾ Object name: Allows the object to be referenced by different processes.
¾ Security descriptor: Contains access permissions.
¾ Open handle information: Contains details of which processes are using the
object
¾ Object type: Contains details of the object's class definition.
¾ Reference count: Holds the count of the number of outstanding handles that
reference the object.
The Object Manager manages the information in the header. For example, when a new

handle is created to an object, the Object Manager updates the open handle information

and reference count. The object type information defines a standard set of methods that

the object implements such as open, close, and delete. Some of these methods are

supplied by the Object Manager, and some must be tailored to the object type; however,

the interface is determined as part of the object header.

The object body format is determined by the Executive component that uses the
object. For example, if the Executive object is a file object, the body format and
contents are managed by the File Manager part of the I/O Manager in the Executive.
12.2.3.1.2 Process and Thread Manager
Manager provides an Executive Process and Thread Manager serves the same purpose in
Windows NT that a process manager serves in any OS. It is the part of the OS responsible
for the following:
¾ Creating and destroying processes and threads
¾ Overseeing resource allocation
¾ Providing synchronization primitives
¾ Controlling process and thread state changes
¾ Keeping track of most of the information that the OS knows about each thread

Lesson No. 1 Intro. to Operating System 312

The Process Manager implements the process abstraction that will be used at the
subsystem and application levels. Implementing the abstraction means that the
Process Manager defines a number of data structures for keeping track of state of
each process and thread. The base process descriptor is called an executive
process control (EPROCESS block). The EPROCESS block contains information
such as identifications, resource lists, and address space descriptions. The
EPROCESS block also references a Kernel-level process control block (PCB, or the
KPROCESS block), which contains the Kernel's view of the process. The NT Kernel
manipulates its portion of the EPROCESS block, and the NT Executive is
responsible for maintaining the remaining fields.
There is also a close relationship between an Executive process and a thread. Just
as there is an EPROCESS block, there is also an executive thread control
(ETHREAD) block for each thread in a process. Since the thread exists within a
process, the EPROCESS block references the list of ETHREAD blocks. Information
about the thread that is managed by the Process Manager is stored in the
ETHREAD block. And because the thread is built on a Kernel-level thread object,
there is also a Kernel thread control (KTHREAD) block containing the information
about the Kernel thread object that is managed by the Kernel-level management.
The EPROCESS block references a KPROCESS block, which references a set of
KTHREAD blocks. The EPROCESS block also references a set of ETHREAD
blocks, each of which references the same KTHREAD blocks.
USER SPACE

SUPERVISOR SPACE ETHREAD

NT-EXECUTIVE
EPROCESS

KTHREAD NT-KERNEL

KPROCESS

Figure 12.3 Process and Thread Descriptors

The NTOSKRNL function NtCreateProcess is called to create a process; that is, the
Win32 API CreateProcess function calls NtCreateProcess. When NtCreateProcess

Lesson No. 1 Intro. to Operating System 313

is called (ordinarily by CreateProcess), it performs the following work in setting up
the process.
¾ Calls the Kernel to have it create a Kernel process object.
¾ Creates and initializes an EPROCESS block.
¾ Creates an address space for the process.
A process has no ability to execute code in its address space; it must have at least one

thread, called the base thread, to execute the code.

The NtCreateThread Executive function creates a thread that can execute within the
process. (The Win32 API CreateProcess function calls both NtCreateProcess and
NtCreateThread; the CreateThread function calls NtCreateThread to create
additional threads within a process.) NtCreateThread performs the following work.
¾ Calls the Kernel to have it create a Kernel thread object.
¾ Creates and initializes an ETHREAD block.
¾ Initializes the thread for execution (sets up its stack, provides it with an
executable start address, and so on).
¾ Places the thread in a scheduling queue.
12.2.3.1.3 Virtual Memory Manager
Windows NT is a paging virtual memory system, which saves a process's address
space contents in secondary storage, loading portions of the image from the
secondary storage into the primary storage on a page-by-page basis whenever it is
needed.
When a process is created, it has 4 GB of virtual addresses available to it, though
none of the addresses are actually allocated at that time. When the process needs
space, it first reserves as much of the address space as it needs at that moment;
reserved addresses do not cause any actual space to be allocated; rather, virtual
addresses are reserved for later use. When the process needs to use the virtual
addresses to store information, it commits the address space, meaning that some
system storage space is then allocated to the process to hold information. A commit
operation causes space on the disk (in the process's page file) to be allocated to the

Lesson No. 1 Intro. to Operating System 314

process; the information is stored on the disk until it is actually referenced by a
thread
When an executing thread references a virtual address, the Virtual Memory
Manager ensures that the page containing that virtual address is read from the page
file and placed at some system-defined location in the physical executable memory.
The Virtual Memory Manager maps the virtual address referenced by the thread into
the physical executable memory.
The Virtual Memory Manager has been designed so that a large portion of each
process's address space (usually half of it, though different configurations of
Windows NT use different fractions) is mapped to the information used by the
system when it is in supervisor mode (Figure 12.4). There are a few important
implications of this decision, as follows.
A process can directly reference every location in the system.
USER SPACE

SUPERVISOR SPACE ETHREAD

NT-EXECUTIVE
EPROCESS

KTHREAD NT-KERNEL

KPROCESS

Figure 12.4. Virtual Memory

Every process shares the same view of the system's space. Such a large, shared
virtual address space makes memory-mapped files feasible. In Figure 12.4, when a
thread references an address in the user space, the virtual memory system loads
the target location into the physical memory prior to its use so that the thread can
read or write the virtual memory address by referencing a physical memory address.
The same mapping takes place for OS memory references, though these references
are protected, and every process's OS addresses map to the OS memory rather
than to the application-specific part of the address space.
12.2.3.1.4 Security Reference Manager
The NT Kernel supports secure operation by including low-level mechanisms for
authentication. The Security Reference Manager is the Executive-level mechanism

Lesson No. 1 Intro. to Operating System 315

to implement the critical parts of certifiable security policies. It is constructed to
check object access according to any give protection policy (specified within
subsystem components that manage the specific access that a process is trying to
perform).
The Security Reference Monitor is a protection mechanism used by the Security
Reference Manager in conjunction with a security policy module executing in user
space. Windows NT includes a user space subsystem component, the Local
Security Authority (LSA) server, to represent the desired security policy. The LSA
uses its own policy database, stored in the machine’s Registry, to hold the details of
the particular machine's policy. The authentication mechanism the LSA server uses
to compare access requests with the database contents can also be provided on an
installation-by-installation basis, though a default mechanism is provided with NT.
The Security Reference Manager authenticates access to Executive objects.
Whenever any thread makes a system call to access an Executive object, the part of
the Executive that handles the access passes a description of the attempted access
to the Security Reference Monitor. The object contains a security descriptor
identifying the object's owner and an access control list (ACL) of processes that are
permitted access to the object. The Security Reference Monitor determines the
thread's identity and access type and then verifies that the thread is allowed to
access the object (according to the information in the (ACL).
12.2.3.1.5 I/O Manager
I/O Manager is responsible for handling all the input/output operations to every device in

the system. The I/O Manager creates an abstraction of all device I/O operations on the

system so that the system's clients can perform operations on a common place data

structure.

The client can perform synchronous and asynchronous I/O.

The client can invoke the Security Reference Monitor whenever security is an issue.
The I/O Manager must accommodate device drivers written in high-level language
by third parties. Those drivers must be able to execute in supervisor mode.
Installation and removal of a device driver must be dynamic.

Lesson No. 1 Intro. to Operating System 316

The I/O Manager can accommodate alternative files systems on the system's disks.
This means that some files systems might use the MS-DOS format others might use
an industry standard CD-ROM format, and yet others might use NT's own file
system (NTFS).
I/O Manager extensions-device drivers and/or file systems-must be consistent with
the memory-mapped file mechanism implemented in the Virtual Memory Manager,
so extension designs are constrained by the facilities provided by the manager.
The I/O Manager is made up of the following components, as shown in Figure 12.5.
Device drivers are at the lowest level. They manipulate the physical I/O devices.
These drivers are described generically in most textbooks.
Intermediate drivers are software modules that work with a low-level device driver to
provide enhanced service. For example, a low-level device driver might simply pass
an error condition "upward" when it detects it, while an intermediate driver might
receive the error and decide to issue a retry operation to the lower-level driver.
File system drivers extend the functionality of the lower-level drivers to implement
the target file system.
A filter driver can be inserted between a device driver and an intermediate, driver,
between an intermediate driver and a file system driver, or between the file system
driver and the I/O Manager API to perform any kind of function that might be
desired. For example, a network redirector filter can intercept file commands
intended for remote files and redirect them to remote file servers.

DATA FLOW

I/O PORTION OF NATIVE API

FILTER DRIVER

INTERMEDIATE DRIVER
N
I/ T
O FILTER DRIVER E
M X
A E
N C
A
DEVICE DRIVER
U
G TI
E V
HAL
Lesson No. 1 Intro.R to Operating System 317 E

DEVICE
 Figure 12.5 The I/O Manager
Drivers are the single component that can be added to the NT Executive to run a
low-level device in supervisor mode. The OS has not been designed to support third
party software, other than drivers, that want to add supervisor mode functionality. In
today's commercial computer marketplace, a consumer can buy a computer from
one vendor and then buy disk drives, graphic adapters, sound boards, and so on,
from other vendors. The OS must be able to accommodate this spectrum of
equipment built by different vendors. Therefore it is mandatory that the OS allow
third parties to add software drivers for each of these hardware components that can
be added to the computer.
The NT I/O Manager defines the framework in which device drivers, intermediate
drivers, file system drivers, and filter drivers are dynamically added to and removed
from the system and are made to work together. The dynamic Stream design allows
one to easily configure complex I/O systems.
The I/O Manager directs modules by issuing I/O request packets (IRPs) into a
stream. If the IRP is intended for a particular module, that module responds to the
IRP; otherwise, it passes the IRP to the next module in the stream. Each driver in
the stream has the responsibility of accepting IRPs, either reacting to the IRP if it is
directed at the driver or passing it on to the next module if it is not.
All information read from or written to the device is managed as a stream of bytes,
called a virtual file. Every driver is written to read and/or write a virtual file. Low-level
device drivers transform information read from the device into a stream & transform
stream information into a device-dependent format before writing it.
As a result of the design of the I/O system architecture, the API to the I/O subsystem
is not complex. For example, subsystems can use NtCreateFile or NtOpen to create
a handle to an Executive file object, NtReadFile and NtWriteFile to read and write an
open file, and Ntlock and NtUnlock to lock a portion of a file.

Lesson No. 1 Intro. to Operating System 318

12.2.3.1.6 Cache Manager
A bottleneck to an application's performance is the time the application must wait for
a physical device to process an I/O command. As processors become faster, that
fraction of the total runtime spent waiting for devices to complete their I/O operations
increasingly dominates the total runtime. The solution to the problem is to devise
ways for the thread to execute concurrently with its own device I/O operations. This
means that the thread is able to predict information that it will need before it actually
needs it and issue an I/O request in anticipation of using data, while concurrently
processing data it has already read.
The Cache Manager is designed to work with the Virtual Memory Manager and the
I/O Manager to perform read-ahead and write-behind on virtual files. The idea is a
classic OS idea. That is, since files are usually accessed sequentially, whenever a
thread reads byte i, it is likely to read byte i+1 soon thereafter. Therefore, on a read-
ahead strategy, when the thread requests that byte i be read from the device, the
Cache Manager asks the Virtual Memory Manager to prepare a buffer to hold K+1
bytes of information from the virtual file and instructs the I/O Manager to read byte i
and the next K bytes into the buffer. Then when the thread requests byte i+1, i+2, ...,
i+K, those bytes will have already been read, so the thread need not wait for a
device operation to complete. The write-behind strategy works similarly.
Most of the Cache Manager's operation is transparent above the NTOSKRNL API.
The Win32 API has only four attributes that it can set when CreateFile is called to
influence the Cache Manager's operation. These attributes are essentially
information to assure the Cache Manager that the thread will access the information
in the file sequentially. The main clients for the Cache Manager are drivers that are
added to the I/O Manager. It is these modules that customize the file system and
use the file-caching facilities provided by this manager.
12.2.3.1.7 The Native API
Executive and Kernel are combined into the NTOSKRNL.EXE executable when NT
is built. The combined Executive and Kernel module (with the underlying HAL)
implements the full NT OS. In Version 5.0, NTOSKRNL exports about 240 functions,
most of which are undocumented, meaning those only subsystem developers

Lesson No. 1 Intro. to Operating System 319

should base their software on the functions. Developers call this interface the NT
Native API or the Executive API. In this lesson, it is called the Native API. Microsoft
provides Windows NT with a set of complementary subsystems, some of which
provide more abstract APIs that application programmers are expected to use.
12.2.4 NT Subsystems
Software systems are often constructed as a layered architecture. Layer i is
constructed using the services provided by layer i -1, creating its own services and
exporting them through its own (layer i) interface. There are several reasons for the
popularity of layered architectures.
¾ It is a simple strategy for dividing and conquering a large problem.
¾ Each layer implements a well-defined subset of the total system functionality.
¾ The functionality at layer i can be designed and tested as a manageable unit.
¾ Layer i+1 services can simplify the way work is done using layer i or lower.
Layer i+1 services can be ported across different
implementations of layer i. In the Windows NT architecture,
subsystems provide a layer of service above the Native API.
There can be many different subsystems, some related, but
others independent of one another, as functionality is
added to the computer system. For example, a typical
Windows NT system includes the Win32 subsystem, the
WinLogon service, a remote procedure call service, and
perhaps a Win 16 subsystem.
If the Windows NT machine were required to support POSIX application programs, a
POSIX subsystem could be added as a component in the subsystem layer. Each
subsystem uses the Native API to provide the services it implements. The
environment subsystems behave as a traditional interior layer. In the layered
architecture approach, they use the Native API, add functionality and services, and
then export their own API. In the Microsoft strategy, subsystem APIs are
documented APIs, meaning that a programmer can write new software at the next
higher-level layer and be assured that the API will be unchanged when
implementations at a lower-level layer in the architecture are changed.

Lesson No. 1 Intro. to Operating System 320

Figure 12.6 shows how this layering works in NT. The Win32 Subsystem exports a
documented interface, the Win32 API, as a set of about 1,000 functions {f0, f1, ..,
fn}. The Win32 API is a documented interface. An application programmer can write
software above the Win32 subsystem that calls the functions, {f0, f1, ..., fn} to
accomplish an application-specific task.
The Win32 Subsystem also provides a user interface management system, since
the Executive/Kernel does not have one of these. This is primarily a matter of
practicality-when the system begins to run, some part of the system software needs
to read the keyboard and mouse and manage the display. Rather than have each
environment subsystem provide its own user interface, the Win32 Subsystem
implements the common window manager for all subsystems. This means that there
is a single human computer interaction model implemented in a single subsystem,
but used by all other subsystems.

WIN 32 API

WIN 32 SUBSYSTEM

NATIVE API

NTOSKRNL

Figure 12.6 The Win32 API

A subsystem's design can be simple or complex. In the simplest case, each function
or service that the subsystem exports is implemented wholly within the subsystem
itself. For example, the subsystem might keep a data structure filled with information
it extracts from information obtained through the Native API. When a program
queries the subsystem, it simply reads the data structure and returns a result without
ever interacting with the OS.
A slightly more complex case occurs when a subsystem function requires that the
subsystem implementation interact with the OS via the Native API. For example, the
Wln32 API function CreateProcess causes the Wln32 subsystem to call the Native
API functions NtCreateProcess and NtCreateThread.

Lesson No. 1 Intro. to Operating System 321

The most complex design requires that the interaction between the subsystem and
the OS be more complicated than a function call or two. The Executive provides a
special interprocess communication facility called the Local Procedure Call (LPC)
facility. The LPC facility allows one process to call a function that is implemented in
another process. This requires special OS activity, since the target procedure is not
known to the compiler and Iink editor and is not determined until the processes are
running. When the calling process starts an LPC, the OS takes the call request,
finds the procedure in the target process’s address space, and calls the target
procedure. Two processes can communicate with one another by making LPCs
back and forth. The most complex subsystem designs use LPCs to invoke Executive
functions.
12.2.4.1 Win 32 API: The Programmer’s View of NT
The Win32 API is the "official OS interface" to all Microsoft OSs. The rationale for
having a single OS API relates to portability. That is, if all Microsoft OSs can export
the same API, then an application writer can produce application software that will
work on all OS versions. Further, enhancements to any of the OS products still
provide the same services via the same, fixed interface. The cost of adopting this
strategy is the need for a subsystem between the OS's native API and the API used
by the application programmers.
MS-DOS created a set of fundamental OS services on which application
programmers came to depend. Unfortunately, the original MS-DOS API is very old.
As a result, it had many built-in dependencies on 16-bit address spaces, single
thread of execution, and so on. The MS-DOS API was upgraded to a Windows
interface, now generally regarded as the Win 16 API. Yet that still was not adequate
to allow programmers to use the full power of Windows NT, Windows 9x, and CE.
All of Microsoft's current OS family implement some variant of the Win32 API.
Whenever an application programmer writes code for a Microsoft OS, the only
documented interfaces available are the Win32 API versions for each OS. There are
few differences between the Windows 9x and Windows NT implementations of
Win32 API. Since CE is aimed at such hardware as palmtop computers and
television set-top boxes, its variant of the Win32 API is distinctly different from the

Lesson No. 1 Intro. to Operating System 322

mainstream API. The Win32 API has about 1,000 function calls. The Win32 API
uses most of the same abstractions that appear at the Native API, including
processes, threads, objects, handles, and files. One reason the Win32 API is so
much larger than the Native API is that the Win32 API also includes the interface to
all of the graphics and user interfaces components, code that is not part of NT.

12.3 Keywords
Microkernel: It is a small nucleus of code comprising of the most essential OS
functions.
NT Kernel: It provides specific mechanisms for general object and memory

management, process management, file management, and device management.

HAL: Hardware Abstraction Layer (HAL) is responsible for mapping various low-
level, processor-specific operations into a fixed interface that is used by the
Windows NT Kernel and Executive.
Virtual Memory Manager: It maps the virtual address referenced by the thread into
the physical executable memory.
I/O Manager: It is responsible for handling all the input/output operations to
every device in the system.
Cache Manager: It is designed to work with the Virtual Memory Manager and the
I/O Manager to perform read-ahead and write-behind on virtual files.

12.4 SUMMARY
Microsoft designed NT to be an extensible, portable operating system, able to take
advantage of new techniques and hardware. NT supports multiple operating
environments and symmetric multiprocessing. The use of kernel objects to provide
basic services, and the support for client-server computing, enable NT to support a
wide variety of application environments. For instance, NT can run programs
compiled for MS-DOS, Win16, Windows 95, NT, and POSIX. It provides virtual
memory, integrated caching, and preemptive scheduling. NT supports a security
model stronger than those of previous Microsoft operating systems, and includes

Lesson No. 1 Intro. to Operating System 323

internationalization features. NT runs on a wide variety of computers, so users can
choose and upgrade hardware to match their budgets and performance
requirements, without needing to alter the applications that they run.

12.5 SUGGESTED READINGS / REFERENCE MATERIAL

7. The Design of the UNIX Operating System, Bach M.J., PHI, New Delhi, 2000.
8. Operating System Concepts, 5th Edition, Silberschatz A., Galvin P.B.,
John Wiley & Sons.
9. Systems Programming & Operating Systems, 2nd Revised Edition,
Dhamdhere D.M., Tata McGraw Hill Publishing Company Ltd., New Delhi.
10. Operating Systems, Madnick S.E., Donovan J.T., Tata McGraw Hill
Publishing Company Ltd., New Delhi.
11. Operating Systems-A Modern Perspective, Gary Nutt, Pearson Education
Asia, 2000.
12. Operating Systems, Harris J.A., Tata McGraw Hill Publishing Company Ltd.,
New Delhi, 2002.

12.6 SELF-ASSESSMENT QUESTIONS (SAQ)

1. What is meant by a multi-user and multi-tasking operating system?
2. How is Unix different from other Operating systems like DOS, Windows,
etc.?
3. What are the various services provided by an Operating System?
4. What is function of a Kernel?
5. What is the general structure of Windows-NT operating system?
6. Windows-NT is a paging virtual memory system. Comment.
7. Discuss Win 32 API.

Lesson No. 1 Intro. to Operating System 324