How C Program Converts Into Assembly

/!: Originally published @ 

www.vishalchovatiya.com.

A bit about function stack frames

  • During function code execution, a new stack frame is created in stack memory to allow access to function parameters and local variables.
  • The direction of stack frame growth totally depends on compiler ABI which is out of our scope for this article.
  • The complete information on stack frame size, memory allocation, returning from stack frame is decided at compile time.
  • Before diving into assembly code you should be aware of two things :
  1. CPU registers of x86 machine.
  2. x86 assembly instructions: As this is a very vast topic & updating quite frequently, we will only see the instructions needed for our examples.

x86 CPU Registers

General Purpose Registers

Pointer Register

Segment Register

Index Registers

Apart from all these, there are many other registers as well which even I don’t know about. But above-mentioned registers are sufficient to understand the subsequent topics.

How C program converts into assembly?

We will consider the following example with its disassembly inlined to understand its different aspect of its working at machine level :

We will focus on a stack frame of the function

func()

. But before analysing stack frame of it, we will see how the calling of function happens

Function calling

Function calling is done by call instruction(see Line 15) which is subroutine instruction equivalent to:

push rip + 1 ; return address is address of next instructions
jmp func
Here,

call

store the

rip+1

(not that +1 is just for simplicity, technically this will be substituted by the size of instruction) in the stack which is return address once call to

func()

ends.

Function stack frame

A function stack frame is divided into three parts
1. Prologue/Entry: As you can see instructions(line 2 to 4) generated against start bracket

{

is prologue which is setting up the stack frame for

func()

, Line 2 is pushing the previous frame pointer into the stack & Line 3 is updating the current frame pointer with stack end which is going to be a new frame start.

push 

is basically equivalent to :

sub esp, 4   ; decrements ESP by 4 which is kind of space allocation
mov [esp], X ; put new stack item value X in

Parameter passing

Argument of 

func()

 is stored in 

edi 

register on Line 14 before calling 

call 

instruction. If there is more argument then it will be stored in a subsequent register or stack & address will be used.

Line 4 in 

func()

 is reserving space by pulling frame pointer(pointed by 

rbp

 register) down by 4 bytes for the parameter 

arg 

as it is of type 

int

. Then 

mov 

instruction will initialize it with value store in

edi

. This is how parameters are passed & stored in the current stack frame.

          ---|-------------------------|--- main()
             |                         |          
             |                         |          
             |                         |          
             |-------------------------|          
             |    main frame pointer   |          
rbp & rsp ---|-------------------------|--- func()
in func()    |           arg           |          
             |-------------------------|          
             |            a            |          
             |-------------------------|    stack 
             |            +            |      |   
             |            +            |      |   
             |            +            |      |   
          ---|-------------------------|---  |/  
             |                         |          
             |                         |          
                                                   

Allocating space for local variables

2. User code: Line 5 is reserving space for a local variable 

a

, again by pulling frame pointer further down by 4 bytes. 

mov

 instruction will initialize that memory with a value 

5

.

Accessing global & local static variables

  • As you can see above, 
    g 

    is addressed directly with its absolute addressing because its address is fixed which lies in the data segment.

  • This is not the case all the time. Here we have compiled our code for  x86 mode, that’s why it is accessing it with an absolute address.
  • In the case of x64 mode, the address is resolved using 
    rip

     register which meant that the assembler and linker should cooperate to compute the offset of 

    g 

    from the ultimate location of the current instruction which is pointed by 

    rip

     register.

  • The same statement stands true for the local static variables also.
3. Epilogue/Exit: After the user code execution, the previous frame pointer is retrieved from the stack by 

pop

 instruction which we have stored in Line 2. 

pop 

is equivalent to:

mov X, [esp] ; put top stack item value into X 
add esp, 4   ; increments ESP by 4 which is kind of deallocation

Return from function

ret

instruction jumps back to the next instruction from where

func() 

called by retrieving the jump address from stack stored by 

call 

instruction. 

ret

 is subroutine instruction which is equivalent to:

pop rip ; 
jmp rip ;
If any return value specified then it will be stored in 

eax

register which you can see in Line 16.

Intuitive FAQs 

Q. How do you determine the stack growth direction ?

A. Simple…! by comparing the address of two different function’s local variables.

int *main_ptr = NULL;
int *func_ptr = NULL;
void func() { int a; func_ptr = &a; }
int main()
{
    int a; main_ptr = &a;
    func();
    (main_ptr > func_ptr) ? printf("DOWNn") : printf("UPn");
    return 0;
}

Q. How do you corrupt stack deliberately ?

A. Corrupt the SFR values stored in the stack frame.

void func()
{
    int a;
    memset(&a, 0, 100); // Corrupt SFR values stored in stack frame
}
int main()
{
    func();
    return 0;
}

Q. How you can increase stack frame size ?

A. 

alloca()

 is the answer. Google about it or see this. Although this is not recommended.

read original article here