We can walk through the steps taken to build a simple application step by step.
Note that when you type gcc that actually runs a driver program that hides most of the steps from you. Under normal circumstances this is exactly what you want, because the exact commands and options to get a real life working executable on a real system can be quite complicated and architecture specific.
We will show the compliation process with the two following examples. Both are C source files, one defined the main() function for the inital program entry point, and another declares a helper type function. There is one global variable too, just for illustration.
Example 7-4. Hello World
#include <stdio.h>
/* We need a prototype so the compiler knows what types function() takes */
int function(char *input);
5
/* Since this is static, we can define it in both hello.c and function.c */
static int i = 100;
/* This is a global variable */
10 int global = 10;
int main(void)
{
/* function() should return the value of global */
15 int ret = function("Hello, World!");
exit(ret);
};Example 7-5. Function Example
#include <stdio.h>
static int i = 100;
5 /* Declard as extern since defined in hello.c */
extern int global;
int function(char *input)
{
10 printf("%s\n", input);
return global;
};All compilers have an option to only execute the first step of compilation. Usually this is something like -S and the output will generally be put into a file with the same name as the input file but with a .s extension.
Thus we can show the first step with gcc -S as illustrated in the example below.
Example 7-6. Compilation Example
ianw@lime:~/programs/csbu/wk7/code$ gcc -S hello.c ianw@lime:~/programs/csbu/wk7/code$ gcc -S function.c ianw@lime:~/programs/csbu/wk7/code$ cat function.s .file "function.c" 5 .pred.safe_across_calls p1-p5,p16-p63 .section .sdata,"aw",@progbits .align 4 .type i#, @object .size i#, 4 10 i: data4 100 .section .rodata .align 8 .LC0: 15 stringz "%s\n" .text .align 16 .global function# .proc function# 20 function: .prologue 14, 33 .save ar.pfs, r34 alloc r34 = ar.pfs, 1, 4, 2, 0 .vframe r35 25 mov r35 = r12 adds r12 = -16, r12 mov r36 = r1 .save rp, r33 mov r33 = b0 30 .body ;; st8 [r35] = r32 addl r14 = @ltoffx(.LC0), r1 ;; 35 ld8.mov r37 = [r14], .LC0 ld8 r38 = [r35] br.call.sptk.many b0 = printf# mov r1 = r36 ;; 40 addl r15 = @ltoffx(global#), r1 ;; ld8.mov r14 = [r15], global# ;; ld4 r14 = [r14] 45 ;; mov r8 = r14 mov ar.pfs = r34 mov b0 = r33 .restore sp 50 mov r12 = r35 br.ret.sptk.many b0 ;; .endp function# .ident "GCC: (GNU) 3.3.5 (Debian 1:3.3.5-11)";
The assembly is a little to complex to fully describe, but you should be able to see where i is defined as a data4 (i.e. 4 bytes or 32 bits, the size of an int), where function is defined (function:) and a call to printf().
We now have two assembly files ready to be assembled into machine code!
Assembly is a fairly straight forward process. The assembler is usually called as and takes arguments in a similar fasion to gcc
Example 7-7. Assembly Example
ianw@lime:~/programs/csbu/wk7/code$ as -o function.o function.s ianw@lime:~/programs/csbu/wk7/code$ as -o hello.o hello.s ianw@lime:~/programs/csbu/wk7/code$ ls function.c function.o function.s hello.c hello.o hello.s;
After assembling we have object code, which is ready to be linked together into the final executable. You can usually skip having to use the assembler by hand by calling the compiler with -c, which will directly convert the input file to object code, putting it in a file with the same prefix but .o as an extension.
We can't inspect the object code directly, as it is in a binary format (in future weeks we will learn about this binary format). However we can use some tools to inspect the object files, for example readelf --symbols will show us symbols in the object file.
Example 7-8. Readelf Example
ianw@lime:~/programs/csbu/wk7/code$ readelf --symbols ./hello.o Symbol table '.symtab' contains 15 entries: Num: Value Size Type Bind Vis Ndx Name 5 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS hello.c 2: 0000000000000000 0 SECTION LOCAL DEFAULT 1 3: 0000000000000000 0 SECTION LOCAL DEFAULT 3 4: 0000000000000000 0 SECTION LOCAL DEFAULT 4 10 5: 0000000000000000 0 SECTION LOCAL DEFAULT 5 6: 0000000000000000 4 OBJECT LOCAL DEFAULT 5 i 7: 0000000000000000 0 SECTION LOCAL DEFAULT 6 8: 0000000000000000 0 SECTION LOCAL DEFAULT 7 9: 0000000000000000 0 SECTION LOCAL DEFAULT 8 15 10: 0000000000000000 0 SECTION LOCAL DEFAULT 10 11: 0000000000000004 4 OBJECT GLOBAL DEFAULT 5 global 12: 0000000000000000 96 FUNC GLOBAL DEFAULT 1 main 13: 0000000000000000 0 NOTYPE GLOBAL DEFAULT UND function 14: 0000000000000000 0 NOTYPE GLOBAL DEFAULT UND exit 20 ianw@lime:~/programs/csbu/wk7/code$ readelf --symbols ./function.o Symbol table '.symtab' contains 14 entries: Num: Value Size Type Bind Vis Ndx Name 25 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 0000000000000000 0 FILE LOCAL DEFAULT ABS function.c 2: 0000000000000000 0 SECTION LOCAL DEFAULT 1 3: 0000000000000000 0 SECTION LOCAL DEFAULT 3 4: 0000000000000000 0 SECTION LOCAL DEFAULT 4 30 5: 0000000000000000 0 SECTION LOCAL DEFAULT 5 6: 0000000000000000 4 OBJECT LOCAL DEFAULT 5 i 7: 0000000000000000 0 SECTION LOCAL DEFAULT 6 8: 0000000000000000 0 SECTION LOCAL DEFAULT 7 9: 0000000000000000 0 SECTION LOCAL DEFAULT 8 35 10: 0000000000000000 0 SECTION LOCAL DEFAULT 10 11: 0000000000000000 128 FUNC GLOBAL DEFAULT 1 function 12: 0000000000000000 0 NOTYPE GLOBAL DEFAULT UND printf 13: 0000000000000000 0 NOTYPE GLOBAL DEFAULT UND global;
Although the output is quite complicated (again!) you should be able to understand much of it. For example
In the output of hello.o have a look at the symbol with name i. Notice how it says it is LOCAL? That is because we declared it static and as such it has been flagged as being local to this object file.
In the same output, notice that the global variable is defined as a GLOBAL, meaning that it is visible outside this file. Similarly the main() function is externally visable.
Notice that the function symbol (for the call to function() is left has UND or undefined. This means that it has been left for the linker to find the address of the function.
Have a look at the symbols in the function.c file and how they fit into the output.
Actually invoking the linker, called ld, is a very complicated process on a real system (are you sick of hearing this yet?). This is why we leave the linking process up to gcc.
But of course we can spy on what gcc is doing under the hood with the -v (verbose) flag.
Example 7-9. Linking Example
/usr/lib/gcc-lib/ia64-linux/3.3.5/collect2 -static /usr/lib/gcc-lib/ia64-linux/3.3.5/../../../crt1.o /usr/lib/gcc-lib/ia64-linux/3.3.5/../../../crti.o /usr/lib/gcc-lib/ia64-linux/3.3.5/crtbegin.o 5 -L/usr/lib/gcc-lib/ia64-linux/3.3.5 -L/usr/lib/gcc-lib/ia64-linux/3.3.5/../../.. hello.o function.o --start-group 10 -lgcc -lgcc_eh -lunwind -lc --end-group 15 /usr/lib/gcc-lib/ia64-linux/3.3.5/crtend.o /usr/lib/gcc-lib/ia64-linux/3.3.5/../../../crtn.o;
The first thing you notice is that a program called collect2 is being called. This is a simple wrapper around ld that is used internally by gcc.
The next thing you notice is object files starting with crt being specified to the linker. These functions are provided by gcc and the system libraries and contain code required to start the program. In actuality, the main() function is not the first one called when a program runs, but a function called _start which is in the crt object files. This function does some generic setup which application programmers do not need to worry about.
The path heirarchy is quite complicated, but in essence we can see that the final step is to link in some extra object files, namely
crt1.o : provided by the system libraries (libc) this object file contains the _start function which is actually the first thing called within the program.
crti.o : provided by the system libraries
crtbegin.o
crtsaveres.o
crtend.o
crtn.o
We discuss how these are used to start the program a little later.
Next you can see that we link in our two object files, hello.o and function.o. After that we specify some extra libraries with -l flags. These libraries are system specific and required for every program. The major one is -lc which brings in the C library, which has all common functions like printf().
After that we again link in some more system object files which do some cleanup after programs exit.
Although the details are complicated, the concept is straight forward. All the object files will be linked together into a single executable file, ready to run!
We will go into more details about the executable in the short future, but we can do some inspection in a similar fashion to the object files to see what has happened.
Example 7-10. Executable Example
ianw@lime:~/programs/csbu/wk7/code$ gcc -o program hello.c function.c ianw@lime:~/programs/csbu/wk7/code$ readelf --symbols ./program Symbol table '.dynsym' contains 11 entries: 5 Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 1: 6000000000000de0 0 OBJECT GLOBAL DEFAULT ABS _DYNAMIC 2: 0000000000000000 176 FUNC GLOBAL DEFAULT UND printf@GLIBC_2.2 (2) 3: 600000000000109c 0 NOTYPE GLOBAL DEFAULT ABS __bss_start 10 4: 0000000000000000 704 FUNC GLOBAL DEFAULT UND exit@GLIBC_2.2 (2) 5: 600000000000109c 0 NOTYPE GLOBAL DEFAULT ABS _edata 6: 6000000000000fe8 0 OBJECT GLOBAL DEFAULT ABS _GLOBAL_OFFSET_TABLE_ 7: 60000000000010b0 0 NOTYPE GLOBAL DEFAULT ABS _end 8: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _Jv_RegisterClasses 9: 0000000000000000 544 FUNC GLOBAL DEFAULT UND __libc_start_main@GLIBC_2.2 (2) 15 10: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__ Symbol table '.symtab' contains 127 entries: Num: Value Size Type Bind Vis Ndx Name 0: 0000000000000000 0 NOTYPE LOCAL DEFAULT UND 20 1: 40000000000001c8 0 SECTION LOCAL DEFAULT 1 2: 40000000000001e0 0 SECTION LOCAL DEFAULT 2 3: 4000000000000200 0 SECTION LOCAL DEFAULT 3 4: 4000000000000240 0 SECTION LOCAL DEFAULT 4 5: 4000000000000348 0 SECTION LOCAL DEFAULT 5 25 6: 40000000000003d8 0 SECTION LOCAL DEFAULT 6 7: 40000000000003f0 0 SECTION LOCAL DEFAULT 7 8: 4000000000000410 0 SECTION LOCAL DEFAULT 8 9: 4000000000000440 0 SECTION LOCAL DEFAULT 9 10: 40000000000004a0 0 SECTION LOCAL DEFAULT 10 30 11: 40000000000004e0 0 SECTION LOCAL DEFAULT 11 12: 40000000000005e0 0 SECTION LOCAL DEFAULT 12 13: 4000000000000b00 0 SECTION LOCAL DEFAULT 13 14: 4000000000000b40 0 SECTION LOCAL DEFAULT 14 15: 4000000000000b60 0 SECTION LOCAL DEFAULT 15 35 16: 4000000000000bd0 0 SECTION LOCAL DEFAULT 16 17: 4000000000000ce0 0 SECTION LOCAL DEFAULT 17 18: 6000000000000db8 0 SECTION LOCAL DEFAULT 18 19: 6000000000000dd0 0 SECTION LOCAL DEFAULT 19 20: 6000000000000dd8 0 SECTION LOCAL DEFAULT 20 40 21: 6000000000000de0 0 SECTION LOCAL DEFAULT 21 22: 6000000000000fc0 0 SECTION LOCAL DEFAULT 22 23: 6000000000000fd0 0 SECTION LOCAL DEFAULT 23 24: 6000000000000fe0 0 SECTION LOCAL DEFAULT 24 25: 6000000000000fe8 0 SECTION LOCAL DEFAULT 25 45 26: 6000000000001040 0 SECTION LOCAL DEFAULT 26 27: 6000000000001080 0 SECTION LOCAL DEFAULT 27 28: 60000000000010a0 0 SECTION LOCAL DEFAULT 28 29: 60000000000010a8 0 SECTION LOCAL DEFAULT 29 30: 0000000000000000 0 SECTION LOCAL DEFAULT 30 50 31: 0000000000000000 0 SECTION LOCAL DEFAULT 31 32: 0000000000000000 0 SECTION LOCAL DEFAULT 32 33: 0000000000000000 0 SECTION LOCAL DEFAULT 33 34: 0000000000000000 0 SECTION LOCAL DEFAULT 34 35: 0000000000000000 0 SECTION LOCAL DEFAULT 35 55 36: 0000000000000000 0 SECTION LOCAL DEFAULT 36 37: 0000000000000000 0 SECTION LOCAL DEFAULT 37 38: 0000000000000000 0 SECTION LOCAL DEFAULT 38 39: 0000000000000000 0 SECTION LOCAL DEFAULT 39 40: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 60 41: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 42: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 43: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 44: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 45: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 65 46: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 47: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 48: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 49: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 50: 0000000000000000 0 FILE LOCAL DEFAULT ABS abi-note.S 70 51: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 52: 0000000000000000 0 FILE LOCAL DEFAULT ABS abi-note.S 53: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 54: 0000000000000000 0 FILE LOCAL DEFAULT ABS abi-note.S 55: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 75 56: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 57: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 58: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 59: 0000000000000000 0 FILE LOCAL DEFAULT ABS abi-note.S 60: 0000000000000000 0 FILE LOCAL DEFAULT ABS init.c 80 61: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 62: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 63: 0000000000000000 0 FILE LOCAL DEFAULT ABS initfini.c 64: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 65: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 85 66: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 67: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 68: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 69: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 70: 4000000000000670 128 FUNC LOCAL DEFAULT 12 gmon_initializer 90 71: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 72: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 73: 0000000000000000 0 FILE LOCAL DEFAULT ABS initfini.c 74: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 75: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 95 76: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 77: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 78: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 79: 0000000000000000 0 FILE LOCAL DEFAULT ABS /build/buildd/glibc-2.3.2 80: 0000000000000000 0 FILE LOCAL DEFAULT ABS auto-host.h 100 81: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 82: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 83: 6000000000000fc0 0 NOTYPE LOCAL DEFAULT 22 __CTOR_LIST__ 84: 6000000000000fd0 0 NOTYPE LOCAL DEFAULT 23 __DTOR_LIST__ 85: 6000000000000fe0 0 NOTYPE LOCAL DEFAULT 24 __JCR_LIST__ 105 86: 6000000000001088 8 OBJECT LOCAL DEFAULT 27 dtor_ptr 87: 40000000000006f0 128 FUNC LOCAL DEFAULT 12 __do_global_dtors_aux 88: 4000000000000770 128 FUNC LOCAL DEFAULT 12 __do_jv_register_classes 89: 0000000000000000 0 FILE LOCAL DEFAULT ABS hello.c 90: 6000000000001090 4 OBJECT LOCAL DEFAULT 27 i 110 91: 0000000000000000 0 FILE LOCAL DEFAULT ABS function.c 92: 6000000000001098 4 OBJECT LOCAL DEFAULT 27 i 93: 0000000000000000 0 FILE LOCAL DEFAULT ABS auto-host.h 94: 0000000000000000 0 FILE LOCAL DEFAULT ABS <command line> 95: 0000000000000000 0 FILE LOCAL DEFAULT ABS <built-in> 115 96: 6000000000000fc8 0 NOTYPE LOCAL DEFAULT 22 __CTOR_END__ 97: 6000000000000fd8 0 NOTYPE LOCAL DEFAULT 23 __DTOR_END__ 98: 6000000000000fe0 0 NOTYPE LOCAL DEFAULT 24 __JCR_END__ 99: 6000000000000de0 0 OBJECT GLOBAL DEFAULT ABS _DYNAMIC 100: 4000000000000a70 144 FUNC GLOBAL HIDDEN 12 __do_global_ctors_aux 120 101: 6000000000000dd8 0 NOTYPE GLOBAL DEFAULT ABS __fini_array_end 102: 60000000000010a8 8 OBJECT GLOBAL HIDDEN 29 __dso_handle 103: 40000000000009a0 208 FUNC GLOBAL DEFAULT 12 __libc_csu_fini 104: 0000000000000000 176 FUNC GLOBAL DEFAULT UND printf@@GLIBC_2.2 105: 40000000000004a0 32 FUNC GLOBAL DEFAULT 10 _init 125 106: 4000000000000850 128 FUNC GLOBAL DEFAULT 12 function 107: 40000000000005e0 144 FUNC GLOBAL DEFAULT 12 _start 108: 6000000000001094 4 OBJECT GLOBAL DEFAULT 27 global 109: 6000000000000dd0 0 NOTYPE GLOBAL DEFAULT ABS __fini_array_start 110: 40000000000008d0 208 FUNC GLOBAL DEFAULT 12 __libc_csu_init 130 111: 600000000000109c 0 NOTYPE GLOBAL DEFAULT ABS __bss_start 112: 40000000000007f0 96 FUNC GLOBAL DEFAULT 12 main 113: 6000000000000dd0 0 NOTYPE GLOBAL DEFAULT ABS __init_array_end 114: 6000000000000dd8 0 NOTYPE WEAK DEFAULT 20 data_start 115: 4000000000000b00 32 FUNC GLOBAL DEFAULT 13 _fini 135 116: 0000000000000000 704 FUNC GLOBAL DEFAULT UND exit@@GLIBC_2.2 117: 600000000000109c 0 NOTYPE GLOBAL DEFAULT ABS _edata 118: 6000000000000fe8 0 OBJECT GLOBAL DEFAULT ABS _GLOBAL_OFFSET_TABLE_ 119: 60000000000010b0 0 NOTYPE GLOBAL DEFAULT ABS _end 120: 6000000000000db8 0 NOTYPE GLOBAL DEFAULT ABS __init_array_start 140 121: 6000000000001080 4 OBJECT GLOBAL DEFAULT 27 _IO_stdin_used 122: 60000000000010a0 8 OBJECT GLOBAL DEFAULT 28 __libc_ia64_register_back 123: 6000000000000dd8 0 NOTYPE GLOBAL DEFAULT 20 __data_start 124: 0000000000000000 0 NOTYPE WEAK DEFAULT UND _Jv_RegisterClasses 125: 0000000000000000 544 FUNC GLOBAL DEFAULT UND __libc_start_main@@GLIBC_ 145 126: 0000000000000000 0 NOTYPE WEAK DEFAULT UND __gmon_start__;
Some things to note
Note I built the executable the "easy" way!
See there are two symbol tables; the dynsym and symtab ones. We explain how the dynsym symbols work soon, but notice that some of them are versioned with an @ symbol.
Note the many symbols that have been included from the extra object files. Many of them start with __ to avoid clashing with any names the programmer might choose. Read through and pick out the symbols we mentioned before from the object files and see if they have changed in any way.