VESA list on real hardware

[jmkerdal]

Hello, I have a question about VESA and real hardware.

whit this code:

Code: Select all

	cli                         # BIOS enabled interrupts; disable
	cld
	
	# Zero data segment registers DS, ES, and SS.
	xorw    %ax, %ax             # Set %ax to zero
	movw    %ax, %ds             # -> Data Segment
	movw    %ax, %es             # -> Extra Segment
	movw    %ax, %ss             # -> Stack Segment

	/* Don't lose dl */
	mov %dl, boot_disk

	/* Initialize stack to just below us */
	mov $0x7c00, %ax
	mov %ax, %sp

	/* Ask for drive params */
	mov $0x48, %ah
#	mov boot_disk, %dl
	mov $drive_params, %si
	int $0x13

	/* Get video mode info */
	mov $0, %ax
	mov %ax, %es
	mov $vbe_cont_info, %di
	mov $0x4F00, %ax
	int $0x10

On my PC with QEMU emulation when I run my boot, VESA table is stored at 0x7E04:

PC: 7E04
56 45 53 41 00 03 D9 58 00 C0 01 00 00 00 26 7E
00 00 00 01 00 00 ED 58 00 C0 00 59 00 C0 14 59
00 C0 00 01 01 01 02 01 03 01 04 01 05 01 06 01
07 01 0D 01 0E 01 0F 01 10 01 11 01 12 01 13 01

This is correct, I got 7E26 for starting list of available mode.
But with two real machine I got this:

Lenovo:
56 45 53 41 00 03 BC 94 00 C0 01 00 00 00 93 95
00 C0 FF 01 00 00 00 00 00 00 00 00 00 00 00 00

Server:
56 45 53 41 00 03 3A 59 00 C0 01 00 00 00 A7 55
00 C0 00 01 09 03 4F 59 00 C0 56 59 00 C0 5F 59
00 C0 00 00 00 00 00 00 00 00 00 00 00 00 00 00

What's wrong?

[sebihepp]

What should be wrong? You will find the list of video modes on 0x0:0x9593 and 0x0:0x55A7

[Octocontrabass]

What's wrong?

It's a far pointer, not a near pointer. You got 0x0000:0x7E26 on QEMU, 0xC000:0x9593 on the first machine, and 0xC000:0x55A7 on the second machine. Those sound like reasonable addresses to me.

You will find the list of video modes on 0x0:0x9593 and 0x0:0x55A7

...I don't think so...

[jmkerdal]

What I understand is that function fill the 512 bytes table vbe_cont_info.
With emulation like QEMU, VirtualBox I get 26 7E that is the first mode available (here 00 01 is the 0x100 mode 640×400 256 colors).

With two real PC I got a different adresse and it's out of the table. And when I read I got plenty of 0 and strange values and never got the 0xFFFF end of table.

Possibly it's not the good way to retreive and set video mode on real hardware?

PS: I try to force video mode to 0x 0115 and that crash on real PC.

[sebihepp]

sebihepp wrote: ↑20 May 2025, 17:58
You will find the list of video modes on 0x0:0x9593 and 0x0:0x55A7
...I don't think so...

Oh, you are right. I used the wrong offset and switched segment and address. My fault

What I understand is that function fill the 512 bytes table vbe_cont_info.
With emulation like QEMU, VirtualBox I get 26 7E that is the first mode available (here 00 01 is the 0x100 mode 640×400 256 colors).

It fills the table VbeInfoBlock. My documentation says it gives you a far pointer to a list of video modes at offset 0xE. The list of video modes supported doesn't have to be inside that table. It can be everywhere. Do you also use the correct segment?

PS: I try to force video mode to 0x 0115 and that crash on real PC.

Maybe the real PCs don't support mode 0x115?

Possibly it's not the good way to retreive and set video mode on real hardware?

Video mode switching on modern PCs is quite difficult. I have been told that most modern graphics cards don't implement VBE3.0 PMode interface anymore. And switching to RMode and back is no option, as it is error prone and very difficult and expensive to do. I rely on limine (bootloader) to set the graphics mode for me and then just use it. Later I need to implement drivers for each video card... sadly...
Maybe then I try again VBE3.0 PMode, just to see where it is supported.

[Octocontrabass]

With emulation like QEMU, VirtualBox I get 26 7E that is the first mode available

The pointer is four bytes, not two: 26 7E 00 00. You need to use all four bytes. If you're only using two bytes, it won't work on real hardware.

[jmkerdal]

Ho yes you are right! It is not a ushort but a uint, as written in the documentation.

Example I took on GitHub is not correct, as usual I must say...
Real hardware is another level.

I will test tomorrow because today I am working at the client's site. I tell you if it's working or not.

[nullplan]

Example I took on GitHub is not correct, as usual I must say...

And that's the stuff they train Copilot on. Another reason to pass on that.

[jmkerdal]

Thank's values are correct now.