Writing Win32 programs in assembly language using TASM

written by Malfunction


 This tutorial is for the beginner who can already code in assembly
 language and who has already coded real mode DOS programs.
 So it's for someone like me a half year ago. :)
 At that time I was searching for some documentation on Win32 assembly.
 As I searched for this I mostly found assembler tutorials for
 real mode programs. And I found lots of links pointing to Iczelion's
 Win32 assembler tutorial, which is written for MASM and uses
 lots of macro shit. The only Win32 ASM tutorial for TASM I have seen
 so far was written by ... let me think ... I believe he called himself
 Masta ... yes, Masta's Win95 ASM tutorial. That wasn't bad, but it
 didn't explain all the stuff I wanted to know. So I decided to
 write my own little tutorial on the subject. I wrote this with the
 aim to write a very complete tutorial. I hope you'll like it! ;)

 Coding in Win32 environment

 As you may know Windows runs in protected mode and so our code will
 do so as well. Windows provides a virtual address space of
 theoretically 4GB of memory for every process. The use of this virtual memory
 allows the system to use the hard disk for swapping when the physical
 memory ain't enough. When you code, you code in a so called "flat"
 memory model. This means you don't need to care for the segment registers
 anymore and that makes the ASM coding a hell easier. You only need
 DWORD offsets when you address memory in Win32. In contrast to 16-bit
 systems like DOS and Win 3.1, 32-bit systems use DWORDs as offsets.
 Do not modify the segment registers or your program will fuck up
 with a chance of 99,99%.
 You will use the 32-bit registers much more than before (if you haven't used
 them already before). Let's take the LOOP instruction for example:
 Now the whole ECX will decrement and not only CX. Remember that!
 In protected mode (as the name suggests) the memory can be protected.
 So you may have read/write access, read only access or no access at all.
 Maybe you have coded COM files in the past and you always had all
 your code and your data in one segment. If you try the same here
 it won't work because:
 1) there MUST be something in the data section or the linker will fail
 2) the code section is write protected, so don't put any variables in here
 Many people tried to use interrupts in Win32 inline ASM code. But this
 doesn't work because you don't call REAL MODE interrupts. You would call
 the protected mode INTs and the good old DOS INTs aren't available anymore.
 Instead of INTs you need to use the Windows API. For a complete documentation
 take a look at Microsoft's MSDN (http://msdn.microsoft.com).
 It is a similar case with the I/O ports. Because your program will
 run in priviledge mode 3 (also called RING-3) you won't be able to access
 some ports. Win95/98/ME don't protect all the I/O ports, but WinNT/2K/XP
 do. In your DOS programs you might still be able to use some ports
 because WinNT/2K/XP allow to use them in the Virtual x86 mode for
 compatiblity reasons.
 And at last I wanna remind you that you will code CASE SENSITIVE from
 now on! It's just like in C++. :)
 This is really important and so write MessageBoxA please and not mESSAGEboXa
 for example! ;)

 Hello World! in Win32 ASM

 Enough theoretical stuff, let's see some code!

 ; ------ CUT here ----------------------------------------------

.model flat

        extrn ExitProcess:proc
        extrn MessageBoxA:proc


        msg_title   DB "MessageBox title",0
        msg_message DB "Hello World!",0


        push 0
        push offset msg_title
        push offset msg_message
        push 0
        call MessageBoxA

        push 0
        call ExitProcess

end start

 ; ------ CUT here ----------------------------------------------

 And now the explanations. :)

 - .386
 - .model flat

 I think this is obvious. The processor directive MUST be before the
 memory model and it must be at least a 386. The model directive
 says we use a flat memory model.

 - extrn ExitProcess:proc
 - extrn MessageBoxA:proc

 Here we import 2 APIs from Kernel32.dll. Do not forget the :proc after
 the API names! The linker will give you no error, but your program
 will definitively fuck up!

 - msg_title DB "MessageBox title",0

 Note that almost every string in Windows is zero terminated.

 - push 0
 - push offset msg_title
 - push offset msg_message
 - push 0
 - call MessageBoxA

 At this time we call an API, the MessageBoxA API to be exactly.
 See below for more info.

 - push 0
 - call ExitProcess

 Yes, no INTs anymore. We use the ExitProcess API to quit. In this
 code example I used 0 as exit code.

 Something more about APIs

 The MessageBoxA call might look a little strange to you.
 Let's see what the MSDN tells us about this API:

 int MessageBox(HWND  hwndOwner,       // handle of owner window
                LPCTSTR  lpszText,     // address of text in message box
                LPCTSTR  lpszTitle,    // address of title of message box
                UINT  fuStyle          // style of message box

 In Win32, parameters aren't passed in registers anymore. Instead they are
 pushed on the stack. You really can assume that every parameter
 is DWORD size. If you code 'push 0' this instruction will push a
 DWORD 0 on the stack, not a WORD.
 If you take a closer look you will notice that the parameters are
 pushed on the stack in the wrong order. As far as I know is this pascal
 calling convention. So you have to push the last parameter as the first
 one and the first parameter as the last one.
 Then simply call the API. The return value will always be in EAX.

 If you have already coded Win32 in C++, you may have wondered about
 that A behind the MessageBox API: "In my C++ code I never typed this ...".
 Lot's of APIs that use strings are available in two versions:
 ANSI and UNICODE. The ones with the A are ANSI and the ones with
 a W at the end are UNICODE (W = Wide chars).

 Do not forget to save register values which you need before you call an API.
 In good old DOS times you knew exactly which registers will be destroyed
 by an INT call, but in the case of APIs you never know. So this is
 especially important in loops because ECX can be anything after the API call.
 You can only be sure that EBP won't be changed by an API call.
 The reason why EBP won't ever be changed by any API is simple:
 most programs use EBP to build the stack frame.

 One more code example

 Let's have another simple code example. This little program will show
 the system time in a message box. Here we go:

 ; ------ CUT here ----------------------------------------------

.model flat

        extrn ExitProcess:proc
        extrn MessageBoxA:proc
        extrn GetSystemTime:proc


        _SYSTEMTIME struc
                wYear DW ?
                wMonth DW ?
                wDayOfWeek DW ?
                wDay DW ?
                wHour DW ?
                wMinute DW ?
                wSecond DW ?
                wMilliseconds DW ?
        _SYSTEMTIME ends


        myTitle DB "tell me what time it is ...",0
        myMessage DB "The system time is: "
        time_string DB "00:00 h",0


        push offset SYSTEMTIME
        call GetSystemTime

        lea edi,[time_string+4]
        xor eax,eax
        mov ax,[SYSTEMTIME.wMinute]
        call convert_to_string

        lea edi,[time_string+1]
        xor eax,eax
        mov ax,[SYSTEMTIME.wHour]
        call convert_to_string

        push 0
        push offset myTitle
        push offset myMessage
        push 0
        call MessageBoxA

        push 0
        call ExitProcess

        xor edx,edx
        mov ecx,10
        div ecx
        or dl,30h
        mov byte ptr [edi],dl
        xor edx,edx
        div ecx
        or dl,30h
        dec edi
        mov byte ptr [edi],dl

end start

 ; ------ CUT here ----------------------------------------------

 How to compile and link a Win32 program?

 For our 'hello world' program (hello.asm) we would compile it as the following:

 tasm32 /ml hello.asm
 tlink32 /Tpe /aa /c hello.obj,,,import32.lib

 As you can see you need to use tasm32.exe and tlink32.exe and not the
 DOS verions (it's the same for td32.exe). Let's discuss the parameters

 /ml - compile case sensitive
 /Tpe - set's output to PE (Portable EXE), /Tpd would be DLL
 /aa - uses Windows API
 /c - case sensitive linking
 import32.lib - see below ...

 How to use APIs from other DLLs?

 Normally, you specify only the import32.lib file for the linker. This
 is the standard file and it's used by the linker for our API references.
 Import32.lib contains all APIs from kernel32.dll, user32.dll and gdi32.dll
 (maybe more, but at least these ones). Let's imagine we want to use the
 registry in our program. For that purpose we need some APIs like
 RegOpenKeyExA. These registry APIs are in advapi32.dll. In your program
 code you declare them as normal APIs, but how to tell the linker that
 we wanna use it? At first, we need to make our own '.lib' file. For that
 purpose we take the implib.exe from TASM's BIN directory:

 Implib -c advapi32.lib C:\windows\system\advapi32.dll

 Do not forget the -c for case sensitive. Now we need to copy the '.lib'
 file to TASM's LIB directory. And now we can give the linker this
 additional '.lib' file:

 tlink32 /Tpe /aa /c program.obj,,,import32.lib advapi32.lib

 stdcall - does is make the nasty coding easier?

 Lot's of Win32 ASM sources use a model directive like the following:

 .model flat, stdcall

 Hmm ... what does stdcall mean? Most coders don't seem to know that.
 They type it because they have seen it somewhere and there's no
 problem using it. I may be wrong here, but it seems to me that this
 is only something that shall make parameter pushing easier.
 All the documentation on the APIs is written for C++ and it is really
 nasty to begin with the last parameter. Let's take the call to the
 MessageBoxA API from the 'hello world' program above. Using the stdcall
 we could write it like this:

 call MessageBoxA, 0, offset msg_message, offset msg_title, 0

 Yes, all in one line. The compiler will produce the push instructions
 for us. The special thing here is that the parameters are given in the
 correct order. In my opinion, this makes the code less readable and
 makes some little optimizations impossible. If you want to call an API
 that needs lots of parameters the line with the call could be very long.
 To continue the call in the next line you can use a '\' at the end of
 a line. Example:

 call CreateProcessA, 0, offset commandline, 0, 0, 0, 0, 0, 0,\
                      offset startupinfo, offset processinformation

 Writing your own DLL

 Let's imagine you want to write your own DLL and you want to export
 some of it's functions. Just write it like a normal program. The
 exported function should be written like this:

 public myFunction

 myFunction PROC
     ; your code goes here ...
 myFunction ENDP

 If you don't declare your function as public the linker will give you
 a warning. The initialization stuff at the entry point of your program
 must quit with a 'ret 0Ch' and NOT with ExitProcess! The reason is simple:
 The loader calls the entry point like this:

 BOOL WINAPI DllEntryPoint(
                HINSTANCE  hinstDLL,        // handle of DLL module
                DWORD  fdwReason,           // reason for calling function
                LPVOID  lpvReserved         // reserved

 In your DllEntryPoint you can do some initialization stuff. This
 entrypoint is called several times. It is called when the DLL is
 being attached to process or thread or when it's being detached
 from a process or thread. Check the MSDN for the different
 values of the fdwReason parameter. Some of the registers must
 be preserved in your DLL entrypoint. This is very important because
 if you don't preserve them the process which loaded the DLL
 will be terminated without any error message after the DLL
 entrypoint was run. I don't know exactly which registers must
 be preserved, but ESI for sure. It's a good idea to preserve simply
 all register by using PUSHAD and POPAD. The return value
 is only of importance when the entrypoint is called with the
 DLL_PROCESS_ATTACH value for the fwdReason. It must be nonzero
 (true) to signalize the LoadLibrary API that the initialization
 was successful. If you return zero the DLL will be removed
 from the process. Construct your entrypoint like this:

        ; ...
        ; code
        ; ...
        mov eax,1
        ret 0Ch

 To export your function you need to write a '.def' file.
 These definition files seem to be very similar to the C++ ones. I don't
 know much about them, but I know that you can write the following
 to export your function:


 That's all. To link the file you need to specify the '.def' file
 and you must use /Tpd instead of /Tpe.

 Using resources

 The standard application icon looks boring ...
 Let's give your program another icon! All we need is an icon (of course *g*)
 and a '.rc' file. Again, these resource scripts are very similar (maybe
 even equal) to the C++ ones. Again, I don't know much about '.rc' files,
 I only used icons so far. :(
 The contents of your resource script should look like:

 100 ICON "C:\path\filename.ico"

 Having this resource files you need to compile it to a *.res file.
 Use the brcc32.exe to do this:

 brcc32.exe myfile.rc

 Then you only need to give the filename of your *.res file as
 a parameter to the linker. Simply start tlink32 /? to see how
 to do this.
 (I'm too lazy to type this and it's now 04:05 o'clock here *g*).

 Last words

 I really hope you liked this tutorial. It really took me some time to write
 all this stuff and two beer, one cigarette, one potion of Snus (swedish tobacco)
 and noisy music were needed to help me writing. :))
 Please mail if you think this tutorial is great, if you think this
 tutorial suckz (but then tell me WHY) or if you have a question
 about Win32 assembly (but do not expect that I can answer it, hehe).
 I'm happy about every mail I receive and I promise to answer.


(c) 2001 Malfunction

Копия на Mega.nz (TXT, ZIP)
Копия на Google.Drive (TXT, ZIP)

Добавить комментарий

Ваш адрес email не будет опубликован. Обязательные поля помечены *