In an earlier article, we have seen C runtime: before starting main & How C program stored in RAM memory. Here we will see “How C program converts into assembly?” and different aspect of its working at the machine level.
A Bit About Functions 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 :
- CPU registers of x86 machine.
- 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:
32-bit SFR | 64-bit SFR | Name |
---|---|---|
eax | rax | Accumulator uses for arithmetic |
ebx | rbx | Base uses for memory address calculations |
ecx | rcx | Counter uses to hold loop count |
edx | rdx | Double-word Accumulator or data register use for I/O port access |
Pointer Register:
32-bit SFR | 64-bit SFR | Name |
---|---|---|
esp | rsp | Stack pointer |
ebp | rbp | Frame/base pointer points current stack frame |
eip | rip | Instruction pointer points to the next instruction to execute |
Segment Register:
SFR | Name |
---|---|
cs | Code segment |
ds | Data segment |
ss | Stack segment |
es | Extra segment |
Index Registers:
32-bit SFR | 64-bit SFR | Name |
---|---|---|
esi | rsi | Source Index uses to point index in sequential memory operations |
edi | rdi | Destination Index uses to point index in sequential memory operations |
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 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 :
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 :
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 inedi
. This is how parameters are passed & stored in the current stack frame.
|
|
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 ofg
from the ultimate location of the current instruction which is pointed byrip
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:
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:
If any return value specified then it will be stored in eax
register which you can see in Line 16.
So, this is it for “How C program converts into assembly?”. Although this kind of information is strictly coupled with compiler & ABI. But most of the compilers, ABI & instruction set architecture follows the same more or less. In case, you have not gone through my previous articles, here are simple FAQs helps you to understand better:
FAQs
Q. How do you determine the stack growth direction
A. Simple…! by comparing the address of two different function’s local variables.
Q. How do you corrupt stack deliberately
A. Corrupt the SFR values stored in the stack frame.
Q. How you can increase stack frame size
A. alloca()
is the answer. Google about it or see this. Although this is not recommended.