Defining a variable inside c++ inline assembly
To do that, you'll need to create a "naked" method with _declspec(naked) and to write yourself the prolog and the epilog that are normally created by the compiler.
The aim of a prolog is to:
- set up EBP and ESP
- reserve space on stack for local variables
- save registers that should be modified in the body of the function
An epilog has to:
- restore the saved register values
- clean up the reserved space for local variables
Here is a standard prolog
push ebp ; Save ebp
mov ebp, esp ; Set stack frame pointer
sub esp, localbytes ; Allocate space for locals
push <registers> ; Save registers
and a standard epilog:
pop <registers> ; Restore registers
mov esp, ebp ; Restore stack pointer
pop ebp ; Restore ebp
ret ; Return from function
Your local variables will then begin at (ebp - 4)
and go downward to (ebp - 4 - localbytes)
. The function parameters will start at (ebp + 8)
and go upward.
It' impossible to create a C variable in assembler: the C compiler has to know about the variable (ie its type and address), which means it has to be declared in the C code.
What can be done is accessing symbols defined in assembler via extern
declarations in C. That won't work for variables with automatic storage duration, though, as these don't have a fixed address but are referenced relative to the base pointer.
If you don't want to access the variables outside of the asm
block, you can use the stack for storing assembler-local data. Just keep in mind that you have to restore the stack pointer to its previous value when leaving the asm
block, eg
sub esp, 12 ; space for 3 asm-local 32bit vars
mov [esp-8], 42 ; set value of local var
[...]
push 0xdeadbeaf ; use stack
[...] ; !!! 42 resides now in [esp-12] !!!
add esp, 16 ; restore esp
If you don't want the relative addresses of the local variables to change whenever you manipulate the stack (ie use push
or pop
), you have to establish a stack frame (ie save the base of the stack in ebp
and address locals relative to this value) as described in cedrou's answer.
Local variables are allocated and freed by manipulating the available space on the call stack via the ESP register, ie:
__asm
{
add esp, 4
mov [esp], 2;
...
sub esp, 4
}
Generally, this is better handled by establishing a "stack frame" for the calling function instead, and then access local variables (and function parameters) using offsets within the frame, instead of using the ESP register directly, ie:
__asm
{
push ebp
mov ebp, esp
add esp, 4
...
mov [ebp-4], 2;
...
mov esp, ebp
pop ebp
}