3150 Chapter 2 Part 1

Introduction to Operating Systems
Lecturer: Dr. T.Y. Wong

Chapter 2, part 1
Basic Process Management

Outline
• Process vs Program
– What are they? Any difference?

• Process Creation
– How to get it done using system calls?

• Program Execution
– How to execute a program?

Fall semester CEG/CSC 3150 Page 2
2008-2009

Program vs Process

What is a Program?

2008-2009

What is a program?
• What is a program?
– A program is a just a piece of code.

• But, which code do you mean?
– High level language code: C, C++, etc.
– Low level language code: assembly code.
– Not-yet an executable: object code.
– Executable: contains only machine code.

2008-2009

Building a program – initial phase…
hello.c hello.c

High-level Expanded
programming Pre-processor
code
language

#define TXT “hello” Compiler hello.s
int main(void) {
printf(“%sn”, TXT);
Assembly
return 0; Optimizer
} code

Program: hello.c

2008-2009

Pre-processor
• The pre-processor expands:
– #define, #include, #ifdef, #ifndef,
#endif, etc.
– Try: gcc –E define.c

#define TXT “hello”
int main(void) {
int main(void) { printf(“%sn”, “hello” );
printf(“%sn”, TXT ); return 0;
return 0; }
} Pre-processor

Program: hello.c

2008-2009

Pre-processor
• Another example: the macro!
#define SWAP(a,b) { int c; c = a; a = b; b = c; }

int main(void) {
int i = 10, j = 20;
printf("before swap: i = %d, j = %dn", i, j);
SWAP(i, j);
printf("after swap: i = %d, j = %dn", i, j);
} Program: swap.c

int main(void) {
int i = 10, j = 20;
printf("before swap: i = %d, j = %dn", i, j)
Pre-processor { int c; c = i; i = j; j = c; };
printf("after swap: i = %d, j = %dn", i, j);
}

2008-2009

Pre-processor
• How about #include?
#include “header.h” int add_fun(int a, int b) {
return (a + b);
int main(void) { }
add_fun(1,2);
return 0; Program: header.h
}
Program: include.c
int add_fun(int a, int b) {
return (a + b);
}

int main(void) {
Pre-processor add_fun(1,2);
return 0;
}
2008-2009

Compiler and Optimizer
• The compiler performs:
– Syntax checking and analyzing;
– If there is no syntax error, construct
intermediate codes, i.e., assembly
codes;

– For syntax checking topics, we have
the courses: CSC3130 and CSC3120.

– For assembly language: we have the
courses: CSC2510 and CSC3420.
2008-2009

Compiler and Optimizer
• The optimizer optimizes codes.
– In other words: it improves stupid “-O” means optimize.
The number is the
codes! optimization level.
– Try:
-O0: means no optimization.
• gcc –S add.c –O0 –o add_O0.s
• gcc –S add.c –O1 –o add_O1.s Max is level 3, i.e., -O3

int main(void) {
• Important notice: Don’t write int x =1;
x = x + 1;
stupid codes because there is x = x + 1;
return x;
an optimizer! }
Program: add.c
An example of stupid codes.
2008-2009

Building a program – next phase…
hello.s hello.o

Assembly Assembler Object
code “as” code

Static/Dynamic Linker Executable
library “ld”

hello.exe
lib*.a or lib*.so

2008-2009

Assembler and Linker
• The assembler assembles “hello.s” and
generates an object code “hello.o”.
– A step closer to machine code.
– Try: as hello.s –o hello.o

• The linker puts together all object files as
well as the libraries.
– There are two kinds of libraries: statically-
linked ones and dynamically-linked ones
2008-2009

Library File
• A library file is nothing special.
– It is just a bunch of function implementations.
– Of course, the linker knows how to look for the
function that the C program needs.

A library file.
A bunch of “dot-o” files.
2008-2009

Library File
Static library file. Dynamic library file.

A smaller
The final program!
program is the
combination of
the above two
codes. The OS will look for this file
when the program runs.

2008-2009

Library File
• The linker also verifies if the symbols
(functions and variables) that are needed
by the program are completely satisfied.
– A compiled executable does not include any
shared libraries.
– E.g. “gcc math.c –lm” is to link libm.so.6,
the math library.
SO – Shared Object

Jargon
Alert

2008-2009

Extra Notes

How gcc compiles programs?
• gcc by default hides all the intermediate steps.
– “gcc -o hello hello.c” generates the
executable “hello” directly.
– “gcc -c hello.c” generates the object code
“hello.o” directly.

• How about working with multiple files?

Extra Notes are the materials that the lecture will cover them, but
they are not included in examinations. Nevertheless, those
materials may be useful to assignments.

2008-2009

Extra Notes

Compiling Multiple Files
Step 1.
Step 2.
Step 3.
$ gcc –c code.c
......

$ gcc –o prog *.o

*.c
*.o
Prepare all the source
files.
Important: there must be
one and only one file prog
containing the main
function. Compile them into object
codes one by one. Construct the program
together with all the object
codes.
2008-2009

Extra Notes

int add_int(int a, int b) {
return (a + b); gcc –Wall –c add_int.c
}
Program: add_int.c
File: add_int.o

int multi_int(int a, int b) {
if(b == 0)
“add_int” is a new
return 0;
vocabulary, and gcc
else if(b == 1)
doesn’t know where it is.
return a;
else
return add_int(a, multi_int(a, b-1));
}
Program: multi_int.c

$ gcc –Wall –c multi_int.c
multi_int.c: In function ‘multi_int’:
multi_int.c:8: warning: implicit declaration of function ‘add_int’
$ _

2008-2009

Extra Notes

int add_int(int a, int b);

Header file: int_header.h The header file provides the
declaration of a function.
#include “int_header.h”
It is the same as saying: “there is
int multi_int(int a, int b) { a function named add_int()”.
if(b == 0)
return 0;
You don’t have to provide the
else if(b == 1)
return a; implementation in the header file.
else Implementation should be
return add_int(a, multi_int(a, b-1)); provided in program files.
}
Of course, as we know what the
pre-processor will do, the
declaration of the function will be
written into the program file.

2008-2009

Extra Notes

int add_int(int a, int b); Using double quotes
int multi_int(int a, int b); means looking for the file
Header file: int_header.h in the current working
directory.
#include “int_header.h”
#include <stdio.h> Using “<......>”
means looking for the file
int main(void) { in the default include
printf(“3 + 10 = %dn”, add_int(3, 10) ); directories, e.g.
printf(“3 * 10 = %dn”, multi_int(3, 10) ); “/usr/include/”.
return 0;
}
Program: int_main.c

We expand the header file so that the main
function knows that “there are functions
gcc –c add_int.c
gcc –c multi_int.c
gcc –c int_main.c called add_int() and multi_int()”.
gcc –o int_main *.o

Executable: int_main

2008-2009

Conclusion on Programs
• A program is just a file!
– It is static;
– It may be associated with dynamically-linked
files;
• “*.so” in Linux and “*.dll” in Windows.
– It may be compiled from more than one files.

2008-2009

Process

System calls about processes

2008-2009

Process
• Process is irreplaceable in an OS.
– It associates with all the files opened by that process.
– It attaches to all the memory that is allocated for it.
– It contains every accounting information,
• such as the running time, current memory usage, who owns the
process, etc.

• You can’t operate any things in an OS without processes.
– So, this chapter is worth spending more than a month to cover.

• We start with some basic and important system calls.

2008-2009

Process Identification
• How can we identify processes from one to
another?
– Each process is given an unique ID number, and is
called the process ID, or the PID.
– The system call, getpid(), prints the PID of the
calling process.
$ getpid
int main(void) { My PID is 1234
printf(“My PID is %dn”, getpid() ); $ getpid
return 0; My PID is 1235
} $ getpid
My PID is 1240
$ _
Program: getpid.c

2008-2009

Process Creation
• To create a process, we use the system
call fork().
– 分叉 in Chinese. The process splits into two!

Original execution flow
of a process
The process
invokes fork().

Flow of original process
Flow of newly-created process
Not this fork anyway…
2008-2009

Process Creation
• So, how does fork() behave?
Process 1234 is the
$ ./fork_example_1 Process 1234 original process, and we
Ready to fork! call it the parent process.
My PID is 1234
My PID is 1235
$
Process 1235 is created by
Process 1235 the fork() system call, and
we call it the child process.

int main(void) {
printf(“Ready to fork!n”);
fork();
printf(“My PID is %dn”, getpid() );
return 0;
}

Program: fork_example_1.c

2008-2009

Process Creation
• So, how does fork() behave?
Process 1234 is the
$ ./fork_example_1 Process 1234 original process, and we
Ready to fork! call it the parent process.
My PID is 1234
My PID is 1235
$
Process 1235 is created by
Process 1235 the fork() system call, and
we call it the child process.

• Summarize from the above printout:
– Both the parent and the child processes execute the same program
code after fork().
– The child process is not starting its execution from the beginning of the
program code.
– The child process starts its execution after the location that fork() is
called.
2008-2009

Process Creation
• Another example. Program: fork_example_2.c

1 int main(void) {
2 int result; $ ./fork_example_2
3 printf(“before fork …n”); before fork …
4 result = fork();
5 printf(“result = %d.n”, result);
6
7 if(result == 0) {
8 printf(“I’m the child.n”);
9 printf(“My PID is %dn”, getpid());
10 }
11 else
12 printf(“I’m the parent.n”);
14 }
15 Process
16 printf(“program terminated.n”); 1234
17 }

2008-2009

Process Creation

1 int main(void) {
4 result = fork();
6
7 if(result == 0) {
10 }
11 else
14 }
15 Process Process
16 printf(“program terminated.n”); 1234 1235
17 }
Process 1234 produces
Process 1235

2008-2009

Process Creation
• Let there be only one CPU. Then…
1 int main(void) {
2 – Only one process is allowed to be executed./fork_example_2
int result; $ at one time.
3 printf(“before fork …n”);
– However, we can’t predict which process before fork … by
will be chosen the
4 result = fork();
5 OS.
printf(“result = %d.n”, result);
6 – This mechanism is called Process Scheduling. It will be
7
8
discussed== 0) { child.n”);
if(result
in details in the next chapter.
printf(“I’m the
10 }
• In this case, let’s assume that Process 1234 will be
11 else
executed first.
14 }
15 Process Process
17 }
Process 1234 produces
Process 1235

2008-2009

Process Creation

1 int main(void) {
4 result = fork(); result = 1235.
6
7 if(result == 0) {
10 }
11 else
14 }
15 Process Process
17 }

“result” is set to the PID of the child process. Process scheduling

2008-2009

Process Creation

1 int main(void) {
5 printf(“result = %d.n”, result); I’m the parent.
6 My PID is 1234
7 if(result == 0) { program terminated.
10 }
11 else
14 }
15 Process Process
17 }

Process 1234: rest in peace. Process scheduling

2008-2009

Process Creation

1 int main(void) {
6 My PID is 1234
8 printf(“I’m the child.n”); result = 0.
10 }
11 else
14 }
15 Process Process
17 }

“result” is set to zero! Process scheduling

2008-2009

Process Creation

1 int main(void) {
6 My PID is 1234
8 printf(“I’m the child.n”); result = 0.
9 printf(“My PID is %dn”, getpid()); I’m the child.
10 } My PID is 1235
11 else program terminated.
12 printf(“I’m the parent.n”); $ _
14 }
15 Process Process
17 }

Process 1235: rest in peace. Process scheduling

2008-2009

Process Creation
• fork() behaves like “cell division”.
– fork() creates the child process by cloning
from the parent process, including…

Copied entities Description

Program counter That’s why they both execute from the same line of
code after fork() returns.

Program code They are sharing the same piece of code.

Memory Including local variables, global variables, and allocated
memory.

Opened file If the parent process has opened a file “A”, then the
child process has also opened the file “A” automatically.

2008-2009

Process Creation
• However, fork() clones nearly all the
things from the parent process, except…

Distinct entities Parent process Child process
Return value of fork() PID of the child process Always 0.

PID number unchanged Different, not necessarily
“parent PID +1”.

The parent unchanged Doesn’t have the same
parent process as that of
the parent process.

Running time unchanged Just created, so it is 0.

2008-2009

Process Creation
• Simple Exercise

– The command “pstree” is able to show the
process relationship among the processes in
the system.

– Modify “fork_example_*.c” so that you can
see the process relationship of your created
processes.

2008-2009

Program Execution
• fork() is rather boring…
– If a process can only duplicate itself and
always runs the same program, then…
– how can we execute other programs?

• We have another set of system calls to do
this job.
– The exec() system call family.

2008-2009

Program Execution
• A member of exec system call family –
execl()
Program name
Program arguments
int main(void) {
$ ./exec_example_1
printf(“before execl …n”); before execl ...

execl(“/bin/ls”, “/bin/ls”, NULL);

printf(“after execl ...n”);

return 0;
}
Process
Program: exec_example.c 1234

2008-2009

Program Execution
execl()
Result from “/bin/ls”

int main(void) {
$ ./exec_example_1
exec_example_1
execl(“/bin/ls”, “/bin/ls”, NULL); exec_example_1.c


return 0;
}
Process

2008-2009

Program Execution
execl() What? Terminated?!
What’d happened to these two lines of codes?

int main(void) {
$ ./exec_example_1
exec_example_1
execl(“/bin/ls”, “/bin/ls”, NULL); exec_example_1.c
$ _

return 0;
}
Process

2008-2009

Program Execution
• The exec system call family is not simply a
function that “invokes” a command.
– It is doing a very complex thing…

int main(void) {

printf(“before execl …n”);



return 0;
}
Originally, I’m Process
executing this code. 1234

2008-2009

Program Execution
• The exec system call family is not simply a
function that “invokes” a command.
– It is doing a very complex thing… file: /bin/ls

int main(void) {

printf(“before execl …n”);


printf(“after execl ...n”); Now, I change to
execute the program:
return 0; /bin/ls.
}
Process
1234

2008-2009

Program Execution
• The process is changing the code that it is executing and
never returns to the original code.
– The last two lines of codes are therefore not executed.

• The process that calls any one of the member of the
exec system call family will throw away many things,
e.g.,
– Memory: including local variables, global variables, and allocated
memory;
– Register value: e.g., the program counter;

• But, the process will preserve something, such as
– PID;
– Process relationship;
– Running time, etc.
2008-2009

Program Execution
• The exec system call family has 6 members and they are
named as…
Function Path Argument Argument Original Provided
File name
Name name list Array ENV ENV

execl Yes Yes Yes

execlp Yes Yes Yes

execle Yes Yes Yes

execv Yes Yes Yes

execvp Yes Yes Yes

execve Yes Yes Yes

letter
p l v e
in name

2008-2009

Program Execution
• Pathname vs Filename
Environment will be
covered in tutorials
/home/tywong/os/shell.c

• Argument list vs array
array[0] /bin/ls
execl(“/bin/ls”, “/bin/ls”, “-l”, NULL);

array[1] /bin/ls
execle(“/bin/ls”, array);
array[2] -l

array[3] NULL

2008-2009

Program Execution
• Simple Challenge
– Can you write two programs that behave as
follows by using the exec system call family?
Program: Program:
yoyo1.c yoyo2.c

$ ./yoyo1
running yoyo1 for the 1 time ... Code execution is
running yoyo2 for the 1 time ... swinging back and forth.
running yoyo1 for the 2 time ...
......(going indefinitely) Process
1234

2008-2009

The system() call…
• Someone may ask:
– Hey! I know there is a function called “system()”
and it is convenient to use.
– Why would there be some stupid guys inventing the
troublesome exec system call family?

$ ./system_example
int main(void) { before system ...
printf(“before system …n”); exec_example_1
system(“ls”); exec_example_1.c
printf(“after system …n”); after system ...
} $ _

Program: system_example.c
system(): what a great function!
2008-2009

The system() library function call
• Well…you’re reversing the order…
– “system()” is a library function call, and
– it is built based on “fork()”, and exec system call
family.
– (That’s why I’ve told you that system calls are not
programmer-friendly.)

int main(void) {
We are now going to
printf(“before system …n”);
system(“ls”); implement the system()
printf(“after system …n”); function using system calls.
}

Program: system_example.c

2008-2009

system() Implementation
Program: system_implement_1.c
1 int system_ver_3150(const char *cmd_str) {
2 if(cmd_str == NULL)
3 return -1;
4
“/bin/sh” is the shell
5 if(fork() == 0) { program.
6 execl("/bin/sh", "/bin/sh", "-c", cmd_str, NULL);
7 fprintf(stderr, We actually want to
8 "%s: command not foundn", execute the command:
9 cmd_str);
10 exit(-1);
/bin/sh -c [cmd_str]
11 }
12
13 return 0;
14 }
15 $ ./system_implement_1
16 int main(void) { before system_ver_3150 ...
17 printf("before system_ver_3150 ...n"); system_implement_1
18 system_ver_3150("/bin/ls"); system_implement_1.c
19 printf("after system_ver_3150 ...n"); after system_ver_3150 ...
20 return 0; $ _
21 }

2008-2009

3 return -1;
4
5 if(fork() == 0) {
7 fprintf(stderr,
8 "%s: command not foundn",
9 cmd_str);
10 exit(-1); However, when I run it for the
11 } 2nd time…
12 What’d happened?!
13 return 0;
14 }
17 printf("before system_ver_3150 ...n"); after system_ver_3150 ...
18 system_ver_3150("/bin/ls"); system_implement_1
19 printf("after system_ver_3150 ...n"); system_implement_1.c
20 return 0; $ _
21 }

2008-2009

3 return -1; Let’s re-color the program!
4
5 if(fork() == 0) { Parent process
7 fprintf(stderr,
Child process
9 cmd_str); Both processes
10 exit(-1);
11 }
12
13 return 0;
14 }
17 printf("before system_ver_3150 ...n"); after system_ver_3150 ...
18 system_ver_3150("/bin/ls"); system_implement_1
19 printf("after system_ver_3150 ...n"); system_implement_1.c
20 return 0; $ _
21 }

2008-2009

It is because of the OS’
process scheduling…

First Execution Sequence Second Execution Sequence

fork() fork()
Parent Child Parent Parent

Parent Child

$ ./system_implement_1 $ ./system_implement_1
before system_ver_3150 ... before system_ver_3150 ...
system_implement_1 after system_ver_3150 ...
system_implement_1.c system_implement_1
after system_ver_3150 ... system_implement_1.c
$ _ $ _

2008-2009

• Don’t forget that we’re trying to implement a
system()-compatible function…
– it is very weird to allow different execution orders.

• Our current problem is …how to let the child to
execute first?
– Sorry…we can’t control the process scheduling of the
OS to this extent.

• Then, our problem becomes…
– How to suspend the execution of the parent process?
– How to wake the parent up after the child is
terminated?
2008-2009

3 return -1;
4
5 if(fork() == 0) {
7 fprintf(stderr,
9 cmd_str);
10 exit(-1);
11 }
12 Put something useful here! We introduce you
13 return 0; the wait()
14 } system call!
15
16 int main(void) {
17 printf("before system_ver_3150 ...n");
18 system_ver_3150("/bin/ls");
19 printf("after system_ver_3150 ...n");
20 return 0;
21 }

2008-2009

system() Implementation – with wait()
• The wait() system call allows a wake up
parent process to be suspended. wait()

• wait() returns and wakes up the
Parent is
parent process when the status of suspended.
the child process changes from
– running to suspended; or
– running to terminated. fork() Terminate
Case 1.

• Another similar system call is Case 2.
waitpid(). wait()
– wait() returns when one of the
children change status.
– waitpid() returns when the no suspension
child specified by a PID changes is needed.
status.

fork() Terminate

2008-2009

3 return -1; Program: system_implement_2.c
4
5 if(fork() == 0) {
7 fprintf(stderr,
9 cmd_str);
10 exit(-1); When wait() returns, the parameter can
11 }
tell the parent about the current status of
12 wait(NULL);
13 return 0; the child process, i.e., terminated or
14 } suspended.
15
16 int main(void) { The parameter NULL means that “I don’t
17 printf("before system_ver_3150 care about the child’s status”.
...n");
18 system_ver_3150("/bin/ls");
20 return 0;
Look up the man page of wait by yourself.
21 } In Linux, the command is: “man 2 wait”

2008-2009

3 return -1; Program: system_implement_2.c
4
5 if(fork() == 0) {
7 fprintf(stderr,
9 cmd_str);
10 exit(-1);
11 }
12 wait(NULL);
13 return 0; Pop Quiz!
14 }
15
16 int main(void) { Can we remove the “exit()” statement?
17 printf("before system_ver_3150 ...n");
18 system_ver_3150("/bin/ls");Hint. when the path “/bin/sh” cannot be
found, the execl() call will continue
20 return 0; executing this program code.
21 }

2008-2009

Summary
• A process is created by cloning.
– fork() is the system call that clones processes.
– Cloning is copying: the new process inherits many
things from the parent process.
– (But, where is the first process?)

• Program execution is not trivial.
– A process is the place that hosts a program and run it.
– exec system call family changes the program that a
process is running.
– A process can run more than one program.
• as long as there is a set of programs that keeps on calling
the exec system call family.

2008-2009

3150 Chapter 2 Part 1

More Related Content

What's hot (20)

Similar to 3150 Chapter 2 Part 1 (20)

3150 Chapter 2 Part 1