ret, retn, retf - how to use them

Question

I have the following asm code:

; int __stdcall wWinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPWSTR lpCmdLine, int nShowCmd)
_wWinMain@16 proc near
         


        

Answer 1

There are actually only two different returns, retn (near return) and retf (far return). When you just use ret, the assembler or compiler is smart enough to pick which one is necessary. A near return is a jump to within the existing code segment, a far return is a jump to a different code segment. On Windows you only have a single code segment, and thus ret should just be a mnemonic for retn. Separate retn and retf instructions are a throwback to older days when segmented memory models were common. Pretty much all 32-bit x86 systems running today uses a flat, not segmented, memory model.

Ret with no argument pops the return address off of the stack and jumps to it. Some calling conventions (like __stdcall) specify that the callee function cleans up the stack. In this case, they call ret with number of bytes to pop those parameters off of the stack. The 16 bytes are the parameters to the winmain function.

Answer 2

In the mnemonic ret N, N is the size of parameters on the stack. In this case it is 4 * 4 = 16 (10h) for 4 DWORDs.
But this only applies to calling conventions when the callee is responsible for stack cleanup. In case of cdecl convention the ret should be without any numbers, as the caller is responsible for stack cleanup.

Answer 3

It's actually two types: retn and retf. The third one ret is coded by the assembler into one of the first two.

The difference is that retn (return near) will pop the instruction pointer (IP) only. While the retf (return far) will pop both the instruction pointer (IP) and the code segment (CS).