Table of content
1. Who may be interested
2. What is Boot Loader
3. Be ready to go deeper
3.1. So what language you should know to develop Boot Loader
3.2. What compiler you need
3.3. How system boots
4. Let’s code
4.1 Program architecture
4.2 Development environment
4.3 BIOS interrupts and screen clearing
4.4 «Mixed code»
4.5 CString implementation
4.6 CDisplay implementation
4.7 Types.h implementation
4.8 BootMain.cpp implementation
4.9 StartPoint.asm implementation
5. Let’s assemble everything
5.1 Creation of COM file
5.2 Assembly automation
6. Testing and Demonstration
6.1 How to test boot loader.
6.2 Testing with the virtual machine VmWare
6.2.1 Creation of the virtual machine
6.2.2 Working with Disk Explorer for NTFS
6.3 Testing on the real hardware
7. Information Sources
Who may be interested
Most of all I’ve written this article for those who have been always interested in the way the different things work. It is for those developers who usually create their applications in high-level languages such as C, C++ or Java, but faced with the necessity to develop something at low-level. We will consider low-level programming on the example of working at system loading.
We will describe what is going after you turn on a computer; how the system is loading. As the practical example we will consider how you can develop your own boot loader which is actually the first point of the system booting process.
What is Boot Loader
Boot loader is a program situated at the first sector of the hard drive; and it is the sector where the boot starts from. BIOS automatically reads all content of the first sector to the memory just after the power is turned on, and jump to it. The first sector is also called Master Boot Record. Actually it is not obligatory for the first sector of the hard drive to boot something. This name has been formed historically because developers used to boot their operating systems with such mechanism.
Be ready to go deeper
In this section I will tell about knowledge and tools you need to develop your own boot loader and also remind some useful information about system boot.
So what language you should know to develop Boot Loader
On the first stage on the computer work the control of hardware is performed mainly by means of BIOS functions known as interrupts. The implementation of interrupts is given only in Assembler – so it is great if you know it at least a little bit. But it’s not the necessary condition. Why? We will use the technology of “mixed code” where it is possible to combine high-level constructions with low-level commands. It makes our task a little simpler.
In this article the main development languages is C++. But if you have brilliant knowledge of C then it will be easy to learn required C++ elements. In general even the C knowledge will be enough but then you will have to modify the source code of the examples that I will descried here.
If you know Java or C# well unfortunately it won’t help for our task. The matter is that the code of Java and C# languages that is produced after compilation is intermediate. The special virtual machine is used to process it (Java Machine for Java, and .NET for C#) which transform intermediate code into processor instructions. After that transformation it can be executed. Such architecture makes it impossible to use mixed code technology – and we are going to use it to make our life easier, so Java and C# don’t work here.
So to develop the simple boot loader you need to know C or C++ and also it would be good if you know something about Assembler – language into which all high-level code is transformed it the end.
What compiler you need
To use mixed code technology you need at least two compilers: for Assembler and C/C++, and also the linker to join object files (.obj) into the one executable.
Now let’s talk about some special moments. There are two modes of processor functioning: real mode and protected mode. Real mode is 16-bit and has some limitations. Protected mode is 32-bit and is fully used in OS work. When it starts processor works in 16-bit mode. So to build the program and obtain executable file you will need the compiler and linker of Assembler for 16-bit mode. For C/C++ you will need only the compiler that can create object files for 16-bit mode.
The modern compilers are made for 32-bit applications only so we won’t able to use them.
I tried a number of free and commercial compilers for 16-bit mode and choose Microsoft product. Compiler along with the linker for Assembler, C, C++ are included into the Microsoft Visual Studio 1.52 package and also can be downloaded from the official site of the company. Some details about compilers we need are given below.
ML 6.15 – Assembler compiler by Microsoft for 16-bit mode;
LINK 5.16 – the linker that can create .com files for 16-bit mode;
CL – С, С++ compiler for 16-bit mode.
You can also use some alternative variants:
DMC – free compile for Assembler, C, C++ for 16 and 32-bit mode by Digital Mars;
LINK – free linker for DMC compiler;
There are also some products by Borland:
BCC 3.5 – С, С++ compiler that can create files for 16-bit mode;
TASM - Assembler compiler for 16-bit mode;
TLINK – linker that can create .com files for 16-bit mode.
All code examples in this article were built with the Microsoft tools.
In order to solve our task we should recall how the system is booting.
Let’s consider briefly how the system components are interacting when the system is booting (see Fig.1).
Fig.1 – “How it boots”
After the control has been passed to the address 0000:7C00, Master Boot Record (MBR) starts its work and triggers the Operating System boot. You can learn more about MBR structure for example here.
In the next sections we will be directly occupied with the low-level programming – we will develop our own boot loader.
Boot loader that we are developing is for the training only. Its tasks are just the following:
- Correct loading to the memory by 0000:7C00 address.
- Calling the
BootMain function that is developed in the high-level language.
- Show “”Hello, world…”, from low-level” message on the display.
Program architecture is described on the Fig.2 that is followed by the text description.
Fig.2. – Program architecture description
The first entity is
StartPoint that is developed purely in Assembler as far as high-level languages don’t have the necessary instructions. It tells compiler what memory model should be used, and what address the loading to the RAM should be performed by after the reading from the disk. It also corrects processor registers and passes control to the
BootMain that is written in high-level language.
BootMain – is an analogue of
main that is in its turn the main function where all program functioning is concentrated.
CString classes take care of functional part of the program and show message on the screen. As you can see from the Fig.2
CDisplay class uses
CStringclass in its work.
Here I use the standard development environment Microsoft Visual Studio 2005 or 2008. You can use any other tools but I made sure that these two with some settings made the compiling and work easy and handy.
First we should create the project of Makefile Project type where the main work will be performed (see Fig.3).
Fig.3 – Create the project of Makefile type
BIOS interrupts and screen clearing
To show our message on the screen we should clear it first. We will use special BIOS interrupt for this purpose.
BIOS proposes a number of interrupts for the work with computer hardware such as video adapter, keyboard, disk system. Each interrupt has the following structure:
where number_of_interrupt is the number of interrupt
Each interrupt has the certain number of parameters that should be set before calling it. The
ah processor register is always responsible for the number of function for the current interrupt, and the other registers are usually used for the other parameters of the current operation. Let’s see how the work of
int 10h interrupt is performed in Assembler. We will use the 00 function that changes the video mode and clears screen:
mov al, 02h mov ah, 00h int 10h
We will consider only those interrupts and functions that will be used in our application. We will need:
int 10h, function 00h – performs changing of video mode and clears screenint 10h, function 01h – sets the cursor typeint 10h, function 13h – shows the string on the screen
Compiler for C++ supports the inbuilt Assembler i.e. when writing code in igh-level language you can use also low level language. Assembler Instructions that are used in the high level code are also called asm insertions. They consist of the key word
__asm and the block of the Assembler instructions in braces:
… } ; end of the block
To demonstrate mixed code let’s use the previously mentioned Assembler code that performed the screen clearing and combine it with C++ code.
mov al, 02h ; setting the graphical mode 80x25(text)
mov ah, 00h ; code of function of changing video mode
int 10h ; call interrupt
CString class is designed to work with strings. It includes
Strlen() method that obtains pointer to the string as the parameter and returns the number of symbols in that string.
static byte Strlen(
const char far* inStrSource
#endif // __CSTRING__
const char far* inStrSource
byte lenghtOfString = 0;
while(*inStrSource++ != '\0')
CDisplay class is designed for the work with the screen. It includes several methods:
– it prints the string on the screen.
– it manages the cursor representation on the screen: show, hide.
– it changes the video mode and thus clears screen.
#define BLACK 0x0
#define BLUE 0x1
#define GREEN 0x2
#define CYAN 0x3
#define RED 0x4
#define MAGENTA 0x5
#define BROWN 0x6
#define GREY 0x7
#define DARK_GREY 0x8
#define LIGHT_BLUE 0x9
#define LIGHT_GREEN 0xA
#define LIGHT_CYAN 0xB
#define LIGHT_RED 0xC
#define LIGHT_MAGENTA 0xD
#define LIGHT_BROWN 0xE
#define WHITE 0xF
static void ClearScreen();
static void TextOut(
const char far* inStrSource,
byte inX = 0,
byte inY = 0,
byte inBackgroundColor = BLACK,
byte inTextColor = WHITE,
bool inUpdateCursor = false
static void ShowCursor(
#endif // __CDISPLAY__
const char far* inStrSource,
byte textAttribute = ((inTextColor) | (inBackgroundColor << 4));
byte lengthOfString = CString::Strlen(inStrSource);
mov al, inUpdateCursor
xor bh, bh
mov bl, textAttribute
xor cx, cx
mov cl, lengthOfString
mov dh, inY
mov dl, inX
mov es, word ptr[inStrSource + 2]
mov bp, word ptr[inStrSource]
mov ah, 13h
mov al, 02h
mov ah, 00h
byte flag = inMode ? 0 : 0x32;
mov ch, flag
mov cl, 0Ah
mov ah, 01h
Types.h is the header file that includes definitions of the data types and macros.
typedef unsigned char byte;
typedef unsigned short word;
typedef unsigned long dword;
typedef char bool;
#define true 0x1
#define false 0x0
#endif // __TYPES__
BootMain() is the main function of the program that is the first entry point (analogue of
main()). Main work is performed here.
#define HELLO_STR "\"Hello, world…\", from low-level..."
extern "C" void BootMain()
.286 .model TINY extrn _BootMain:near .code
org 07c00h ; for BootSector
jmp short start nop
mov ax,cs mov ds,ax mov es,ax mov ss,ax mov bp,7c00h
mov sp,7c00h sti
Creation of COM file
Now when the code is developed we need to transform it to the file for the 16-bit OS. Such files are .com files. We can start each of compilers (for Assembler and C, C++) from the command line, transmit necessary parameters to them and obtain several object files as the result. Next we start linker to transform all .obj files to the one executable file with .com extension. It is working way but it’s not very easy.
Let’s automate the process. In order to do it we create .bat file and put commands with necessary parameters there. Fig.4 represents the full process of application assembling.
Fig.4 – Process of program compilation
Let’s put compilers and linker to the project directory. In the same directory we create .bat file and fill it accordingly to the example (you can use any directory name instead of VC152 where compilers and linker are situated):
.\VC152\CL.EXE /AT /G2 /Gs /Gx /c /Zl *.cpp
.\VC152\ML.EXE /AT /c *.asm
.\VC152\LINK.EXE /T /NOD StartPoint.obj bootmain.obj cdisplay.obj cstring.obj
As the final stage in this section we will describe the way how to turn Microsoft Visual Studio 2005, 2008 into the development environment with any compiler support. Go to the Project Properties: Project->Properties->Configuration Properties\General->Configuration Type.
Configuration Properties tab includes three items: General, Debugging, NMake. Go to NMake and set the path to the build.bat in the Build Command Line and Rebuild Command Line fields – Fig.5.
Fig.5 –NMake project settings
If everything is correct then you can compile in the common way pressing F7 or Ctrl + F7. At that all attendant information will be shown in the Output window. The main advantage here is not only the assembly automation but also navigation thru the code errors if they happen.
Testing and Demonstration
This section will tell how to see the created boot loader in action, perform testing and debug.
How to test boot loader
You can test boot loader on the real hardware or using specially designed for such purposes virtual machine – VmWare. Testing on the real hardware gives you more confidence that it works while testing on the virtual machine makes you confident that it just can work. Surely we can say that VmWare is great method for testing and debug. We will consider both methods.
First of all we need a tool to write our boot loader to the virtual or physical disk. As far as I know there a number of free and commercial, console and GUI applications. I used Disk Explorer for NTFS 3.66 (version for FAT that is named Disk Explorer for FAT) for work in Windows and Norton Disk Editor 2002 for work in MS-DOS.
I will describe only Disk Explorer for NTFS 3.66 because it is the simplest method and suits our purposes the most.
Testing with the virtual machine VmWare
Creation of the virtual machine
We will need VmWare program version 5.0, 6.0 or higher. To test boot loader we will create the new virtual machine with minimal disk size for example 1 Gb. We format it for NTFS file system. Now we need to map the formatted hard drive to VmWare as the virtual drive. To do it:
File->Map or Disconnect Virtual Disks...
After that the window appears. There you should click Map button. In the next appeared window you should set the path to the disk. Now you can also chose the letter for the disk- see Fig.6.
Fig.6 – Parameters of virtual disk mapping
Don’t forget to uncheck the “Open file in read-only mode (recommended)” checkbox. When checked it indicates that the disk should be opened in read-only mode and prevent all recording attempts to avoid data corruption.
After that we can work with the disk of virtual machine as with the usual Windows logical disk. Now we should use Disk Explorer for NTFS 3.66 and record boot loader by the physical offset 0.
Working with Disk Explorer for NTFS
After program starts we go to our disk (File->Drive). In the window appeared we go to the Logical Drives section and chose disk with the specified letter (in my case it is Z) – see Fig.7.
Fig.7 – choosing disk in Disk Explorer for NTFS
Now we use menu item View and As Hex command. It the appeared window we can see the information on the disk represented in the 16-bit view, divided by sectors and offsets. There are only 0s as soon as the disk is empty at the moment. You can see the first sector on the Fig.8.
Fig.8 – Sector 1 of the disk
Now we should write our boot loader program to this first sector. We set the marker to position 00 as it is shown on the Fig.8. To copy boot loader we use Edit menu item, Paste from file command. In the opened window we specify the path to the file and click Open. After that the content of the first sector should change and look like it’s shown on the Fig.9 – if you haven’t changed anything in the example code, of course.
You should also write signature 55AAh by the 1FE offset from the sector beginning. If you don’t do it BIOS will check the last two bytes, won’t find the mentioned signature and will consider this sector as not the boot one and won’t read it to the memory.
To switch to the edit mode press F2 and write the necessary numbers –55AAh signature. To leave edit mode press Esc.
Now we need to confirm data writing.
Fig.9 – Boot Sector appearance
To apply writing we go to Tools->Options. Window will appear; we go to the Mode item and chose the method of writing - Virtual Write and click Write button – Fig.10.
Fig.10 – Choosing writing method in Disk Explorer for NTFS
A great number of routine actions are finished at last and now you can see what we have been developing from the very beginning of this article. Let’s return to the VwWare to disconnect the virtual disk (File->Map or Disconnect Virtual Disks… and click Disconnect).
Let’s execute the virtual machine. We can see now how from the some depth, from the kingdom of machine codes and electrics the familiar string appears ““Hello, world…”, from low-level…” – see Fig.11.
Fig.11 – “Hello world…”
Testing on the real hardware
Testing on the real hardware is almost the same as on the virtual machine except the fact that if something doesn’t work you will need much more time to repair it than to create the new virtual machine. To test boot loader without the threat of existent data corruption (everything can happen), I propose to use flash drive, but first you should reboot your PC, enter BIOS and check if it supports boot from the flash drive. If it does than everything is ok. If it does not than you have to limit your testing to virtual machine test only.
The writing of boot loader to the flash disk in Disk Explorer for NTFS 3.66 is the same to the process for virtual machine. You just should choose the hard drive itself instead of its logical section to perform writing by the correct offset – see Fig.12.
Fig.12 – Choosing physical disk as the device
If something went wrong – and it usually happens – you need some tools to debug your boot loader. I should say at once that it is very complicated, tiring and time-eating process. You will have to grasp in the Assembler machine codes – so good knowledge of this language is required. Any way I give a list of tools for this purpose:
TD (Turbo Debugger) – great debugger for 16-bit real mode by Borland.
CodeView – good debugger for 16-bit mode by Microsoft.
D86 – good debugger for 16-bit real mode developed by Eric Isaacson – honored veteran of development for Intel processor in Assembler.
Bocsh – program-emulator of virtual machine that includes debugger of machine commands.
“Assembly Language for Intel-Based Computers” by Kip R. Irvine is the great book that gives good knowledge of inner structure of the computer and development in Assembler. You ca also find information about installation, configuration and work with the MASM 6.15 compiler.
This link will guide you to the BIOS interrupt list: http://en.wikipedia.org/wiki/BIOS_interrupt_call
In this article we have considered what is boot loader, how BIOS works, and how system components interact when system boots. Practical part gave the information about how to develop your own simple boot loader. We demonstrated the mixed code technology and process of automation of assembly with Microsoft Visual Studio 2005, 2008.
Of course it is just a small piece comparing with the huge theme of low-level programming, but if you get interested of this article – it’s great.
See more case studies and research results at Apriorit site.