How to use VESA linear frame buffer

[SayedMohsen64]

Hello.
I'm not a newbie. I've programmed OSes before, but I haven't done any programming in two years. Well, now I'm writing my OS from scratch. I have a custom boot loader that loads my 32-bit kernel. My OS will be graphical like Mac OS X, with no command-line at all. In my boot loader, I set the VESA mode and get the VESA mode info for the linear frame buffer. This is what my code looks like:

Code: Select all

setup_vesa:
	mov ax, 0x4F01
	mov cx, 8113h
	mov di, vesa_info
	int 0x10

	mov ax, 0x4F02
	mov bx, 8113h
	int 0x10

	jmp go_to_pmode

--- some more code for entering 32-bit mode ---

vesa_info:
	dw 0
	db 0
	db 0
	dw 0
	dw 0
	dw 0
	dw 0
	dd 0
	dw 0

	dw 0
	dw 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0

	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0
	db 0

	.linear_frame_buffer_start		dd 0
	dd 0
	dw 0
	times 206 db 0

Supposedly, vesa_info.linear_frame_buffer_start should contains the start of the frame buffer. In my kernel:

Code: Select all

mov edi, ebx             ; boot loader passes linear frame buffer start in EBX
mov ax, 0xFFFF
mov ecx, 0xEA600
rep stosw

This results in crashing Bochs and Qemu. Any help on how to use the VESA frame buffer? Thanks in advance!

- Sayed Mohsen

[Muazzam]

I have tested it in both qemu and bochs. I worked perfectly in both. I think there is bug in loading kernel or switching to protected mode.

[SayedMohsen64]

I have tested it in both qemu and bochs. I worked perfectly in both. I think there is bug in loading kernel or switching to protected mode.

Have you tested this specific code? It crashes my Bochs and Qemu.

[Muazzam]

I have tested it in both qemu and bochs. I worked perfectly in both. I think there is bug in loading kernel or switching to protected mode.

Have you tested this specific code? It crashes my Bochs and Qemu.

Yes I have tested all code you provided.

[SayedMohsen64]

Yes I have tested all code you provided.

This is what I get instead..
And there is no bug in going to protected mode because if I take away the VESA code and have my kernel print a string in 32-bit mode, it prints and works fine. The bug is in my VESA code, it just doesn't like my specific PC.

[BrightLight]

Your exact code works perfectly in Bochs, Qemu and VMware. I don't see why it doesn't work with you, all I did was copy the code into my other project, which is a 32-bit boot loader.
About your bootloader, it's VERY nice to see someone who has written a bootloader for my filesystem!

[SayedMohsen64]

About your bootloader, it's VERY nice to see someone who has written a bootloader for my filesystem!

Yeah, your file system is very nice and comfortable! Really simple. You just need to finish up the missing parts of your specs and release it.

[BrightLight]

Yeah, your file system is very nice and comfortable! Really simple.

Isn't it surprising and weird that you finished a boot loader so quickly for a filesystem you're new to?

[SayedMohsen64]

Yeah, your file system is very nice and comfortable! Really simple.

Isn't it surprising and weird that you finished a boot loader so quickly for a filesystem you're new to?

Heheheh, I used some of Vector OS code. I am allowed to, right?

[BrightLight]

Heheheh, I used some of Vector OS code. I am allowed to, right?

Of course, any one is allowed to. My GitHub repository clearly says that my project is GPL'd.

[alexfru]

Code: Select all

	setup_vesa:
	mov ax, 0x4F01
	mov cx, 8113h
	mov di, vesa_info

What's in ES here?

Code: Select all

	int 0x10

	mov ax, 0x4F02
	mov bx, 8113h

What's in ES:DI here?

Code: Select all

	int 0x10

Supposedly, vesa_info.linear_frame_buffer_start should contains the start of the frame buffer. In my kernel:

Code: Select all

	mov edi, ebx             ; boot loader passes linear frame buffer start in EBX
	mov ax, 0xFFFF
	mov ecx, 0xEA600

What's in EFLAGS.DF here?

0xEA600 = 800*600*2. Are you sure you don't want 800*600 = 0x75300?

Code: Select all

	rep stosw

Exactly how do you extract and pass the value of vesa_info.linear_frame_buffer_start?

[SayedMohsen64]

Code: Select all

	setup_vesa:
	mov ax, 0x4F01
	mov cx, 8113h
	mov di, vesa_info

What's in ES here?

ES is zero because I have the ORG statement at the beginning of my file.

Code: Select all

	int 0x10

	mov ax, 0x4F02
	mov bx, 8113h

What's in ES:DI here?

Well, as the parameters for INT 10h function 0x4F02, ES:DI is undefined. So, ES:DI is whatever garbage was left from before.

Code: Select all

	int 0x10

Supposedly, vesa_info.linear_frame_buffer_start should contains the start of the frame buffer. In my kernel:

Code: Select all

	mov edi, ebx             ; boot loader passes linear frame buffer start in EBX
	mov ax, 0xFFFF
	mov ecx, 0xEA600

What's in EFLAGS.DF here?

My EFLAGS.DF Is always clear. In any of my files after setting up a stack, I always do "cld".

0xEA600 = 800*600*2. Are you sure you don't want 800*600 = 0x75300?

This is just a test. At least if Bochs crashed from this, at least I would see a white screen first, I probably forget to divide it by two.

Code: Select all

	rep stosw

Exactly how do you extract and pass the value of vesa_info.linear_frame_buffer_start?

I do this, before I give control to the kernel:

Code: Select all

	mov esi, vesa_info.linear_frame_buffer_start
	mov ebx, dword[esi]
	mov ax, [mem_blocks]
	mov dl, [bootdisk]

So when my kernel is executed, EBX contains the start of the linear frame buffer.
No worries, it works now in Qemu and VMware, but not in Bochs.

[Icee]

ES is zero because I have the ORG statement at the beginning of my file.

There's no connection between an ORG directive and segment register values.

[BrightLight]

ES is zero because I have the ORG statement at the beginning of my file.

There's no connection between an ORG directive and segment register values.

There definitely is a connection. Say we don't have an ORG statement, and we start at location 0x0. If we set segment registers to 0, instructions like LODSB, STOSB and other segment register-related stuff won't work properly. For example, if our string is at offset 0x100 in our file, and we have our file loaded to 0x5000, with segment registers to zero, with no ORG statement, 0:0x100 won't properly point to our string. So, DS and ES are related to ORG.

[Bender]

ES is zero because I have the ORG statement at the beginning of my file.

There's no connection between an ORG directive and segment register values.

There definitely is a connection. Say we don't have an ORG statement, and we start at location 0x0. If we set segment registers to 0, instructions like LODSB, STOSB and other segment register-related stuff won't work properly. For example, if our string is at offset 0x100 in our file, and we have our file loaded to 0x5000, with segment registers to zero, with no ORG statement, 0:0x100 won't properly point to our string. So, DS and ES are related to ORG.

No, the ORG directive in any sane assembler would add the ORGed value while referencing an internal address.

For example,

Code: Select all

org 0x100
var: dw 1234
...
mov ax, var

Would translate to something like:

Code: Select all

org 0
var: dw 1234
...
mov ax, var
add ax, 0x100

It doesn't mess with segment registers.