Compilation, Linking, Loading
Abhijit A M
Review of last few lectures
Boot sequence: BIOS, boot-loader, kernel
● Boot sequence: Process world
–kernel->init -> many forks+execs() -> ....
● Hardware interrupts, system calls, exceptions
● Event driven kernel
● System calls
2
What are compiler, assembler, linker
and loader, and C library
System Programs/Utilities
Most essential to make a kernel really usable
3
Standard C Library
A collection of some of the most frequently
●
needed functions for C programs
–scanf, printf, getchar, system-call wrappers (open,
read, fork, exec, etc.), ...
An machine/object code file containing the
●
machine code of all these functions
–Not a source code! Neither a header file. More 4
Compiler
●application program, which converts one
(programming) language to another
–Most typically compilers convert
main.c a high gcc
levelmain
language like C, C++, etc. toSource
Machine
code file code Machine code fi
language
● E.g. GCC /usr/bin/gcc
–Usage: e.g. 5
Assembler
application program, converts assembly code into
●
machine code
● What is assembly language? main.s as main
–Human readable machine code language.
assembly code file Machine code fi
● E.g. x86 assembly code
–mov 50, r1
6
Compilation Process
preprocessor
compiler +
assembler
m/c linking
Exectable
C Code m/c Code
try.c try.o try
Handles all #
Code Checks syntax
Combines machine
errors, grammatical
directives in the code of input file(S)
errors, missing
input C code and library files
declarations,e tc.
And converts the
input code to
machine code
gcc
●From the
textbook
Example
try.c f.c g.c
#include <stdio.h> int g(int); int g(int x) {
#define MAX 30 #define ADD(a, b) (a + b) return x + 10;
int f(int, int); int f(int m, int n) { }
int main() { return ADD(m,n) + g(10);
int i, j, k; }
scanf("%d%d", &i, &j);
k = f(i, j) + MAX; Try these commands, observe the output/errors/warnings, and try to unde
printf("%d\n", k); $ gcc try.c
return 0; $ gcc -c try.c
} $ gcc -c f.c
$ gcc -c g.c
$ gcc try.o f.o g.o -o try
$ gcc -E try.c
$ gcc -E f.c
More about the steps
● Pre-processor
–#define ABC XYZ
cut ABC and paste XYZ
●
–# include <stdio.h>
copy-paste the file stdio.h
●
There is no CODE in stdio.h, only typedefs, #includes, #define, #ifdef, etc.
●
● Linking
–Normally links with the standard C-library by default
–To link with other libraries, use the -l option of gcc
Using gcc itself to understand the
process
Run only the preprocessor
●
–cc -E test.c
–Shows the output on the screen
Run only till compilation (no linking)
●
–cc -c test.c
–Generates the “test.o” file , runs compilation + assembler
–gcc -S main.c
–One step before machine code generation, stops at assembly code
Combine multiple .o files (only linking part)
●
Linking steps
Linker is an application program
●
–On linux, it's the "ld" program
–E.g. you can run commands like $ ld a.o b.o -o c.o
–Normally you have to specify some options to ld to get a proper
executable file.
When you run gcc
●
–$ cc main.o f.o g.o -o try
–the CC will internally invoke "ld" . ld does the job of linking
Linking steps
●The resultatnt file "try" here, will contain the codes of all
the functions and linkages also.
● What is linking?
–"connecting" the call of a function with the code of the
function.
● What happens with the code of printf()?
–The linker or CC will automatically pick up code from the
libc.so.6 file for the functions.
Executable file format
●An executable file ● ELF : The format on Linux.
needs to execute in an ● Try this
environment created by
–$ file /bin/ls
OS and on a particular
processor –$ file /usr/lib/x86_64-
linux-gnu/libc-2.31.so
–Contains machine code +
other information for OS
–Need for a structured-way of
storing machine code in it
Exec() and ELF
When you run a program
● ● ELF is used not only for
executable (complete machine
–$ ./try
code) programs, but also for
–Essentially there willl be a fork() partially compiled files e.g.
and exec("./try", ...") main.o and library files like
libc.so.6
–So the kernel has to read the
file "./try" and understand it. ●What is a.out?
–Soeach kernel will demand it's –"a.out"was the name of a
own object code file format. format used on earleir Unixes.
–Hence ELF, EXE, etc. Formats –It so happened that the early
Utilities to play with object code files
objdump
●
ar
●
–$ objdump -D -x /bin/ls –To create a “statically linked”
–Shows all disassembled library file
machine instructions and –$ ar -crs libmine.a one.o two.o
“headers” ●Gcc to create shared library
hexdump
–$ gcc hello.o -shared -o
●
–$ hexdump /bin/ls libhello.so
–Justshows the file in ●To see how gcc invokes as, ld,
hexadecimal etc; do this
–$ gcc -v hello.c -o hello
Linker, Loader, Link-Loader
● Linker or linkage-editor or link-editor
–The “ld” program. Does linking.
● Loader
–The exec(). It loads an executable in the memory.
● Link-Loader
–Oftenthe linker is called link-loader in literature.
Because where were days when the linker and
Static, dynamic / linking, loading
● Both linking and loading can be
–Static or dynamic
–Moreabout this when we learn memory
management
● An important fundamental:
–memorymanagement features of processor,
memory management architecture of kernel,
Cross-compiler
●Compiler on system-A, but generate object-code
file for system-B (target system)
–E.g.compile on Ubuntu, but create an EXE for
windows
Normally used when there is no compiler available
●
on target system
–see gcc -m option
