Help with activating pmode

tadada · Post by **tadada** » Mon May 26, 2008 12:11 pm

Sorry about that. If I use the code shown to print a "P" on a purple background it works in Real Monde only. When I try to use it in Protected Mode it doesn't work.

Code: Select all

mov ax,0xb800 
mov es,ax 
mov word [es:0],0x5050

neon · Post by **neon** » Mon May 26, 2008 12:43 pm

Try this instead:

Code: Select all

mov ebx,0xb8000
mov byte [ebx],'A'
mov byte [ebx+1], 1

This prints a blue 'A' character on the upper left corner of display. This assumes that DS contains the correct data descriptor. Not doing so will result in #GPF.

Remember that protected mode does NOT use segment:offset addressing as in real mode. It uses desc:linear, where desc is the offset in your GDT (Code descriptor, data descriptor, et al..); while linear is a 32 bit linear address.

Looking at the code you posted, you set ES to 0 (Null selector). Referencing any data using ES will generate a #GPF.

DrLink · Post by **DrLink** » Mon May 26, 2008 1:51 pm

I think you can use to copy it to a disk in Linux

dd if=./floppy.img of=/dev/fd0

If that doesn't work, re-build the floppy image and try reversing the if= and of= parameters.

svdmeer · Post by **svdmeer** » Mon May 26, 2008 3:33 pm

tadada wrote:Sorry about that. If I use the code shown to print a "P" on a purple background it works in Real Monde only. When I try to use it in Protected Mode it doesn't work.
Code: Select all
mov ax,0xb800 
mov es,ax 
mov word [es:0],0x5050

True.

In protectedmode you need only mov word [es:0xb8000],0x5050

Combuster · Post by **Combuster** » Mon May 26, 2008 4:15 pm

Would you try running the program in bochs' debugger. Stop it at the mov [0xb8000], 0x5050, close it and post the resulting logfile.

tadada · Post by **tadada** » Mon May 26, 2008 7:29 pm

I have two files:

OSDEV.log (the log. Big. 7.25 MBs. had to use a link it said the file was empty...)
http://www.mediafire.com/?nmyfuuxxbxr

Error.txt (The last thing on the Bochs debug console.)

In the Log I saw this which I thought was interesting:

00002643656d[KBD ] keyboard: 8-bit write to 0064 = ad
00002643656d[KBD ] keyboard disabled
00002643663d[KBD ] keyboard: 8-bit write to 0064 = d0
00002643663d[KBD ] io write to port 64h, command d0h (partial)
00002643663d[KBD ] controller_enQ(03) source=00
00002643669d[IOAP ] set_irq_level(): INTIN1: level=0
00002643669d[KBD ] READ(60) = 03
00002643677d[KBD ] keyboard: 8-bit write to 0064 = d1
00002643677d[KBD ] write output port
00002643685d[KBD ] keyboard: 8-bit write to 0060 = 03
00002643685d[KBD ] write output port with value 03h
00002643685d[KBD ] write output port : enable A20
00002643685d[SYS ] A20: set() = 1
00002643692d[KBD ] keyboard: 8-bit write to 0064 = ae
00002643692d[KBD ] keyboard enabled
00002643700d[CPU0 ] BxError: i with opcode=0xff
00002643700d[CPU0 ] mod was c0, nnn was 7, rm was 7
00002643700d[CPU0 ] WARNING: Encountered an unknown i (signalling illegal i)
00002643700d[CPU0 ] UndefinedOpcode: 0xff causes exception 6
00002643700d[CPU0 ] exception(0x06): error_code=0000
00002643700d[CPU0 ] interrupt(): vector = 6, INT = 0, EXT = 1

If I understood it correctly it activated A20 and then hit an error. This would mean that the loading pmode problem was never fixed and that would exsplain why it never printed.

neon · Post by **neon** » Mon May 26, 2008 8:15 pm

Post your current code, in full if it is not too large.

kmtdk · Post by **kmtdk** » Tue May 27, 2008 2:12 pm

As neon said, we need to see the current code, or else we cant find the error.
for me it looks like a BIG loop- everything is reapeaded

KMT

tadada · Post by **tadada** » Tue May 27, 2008 2:36 pm

I posted everything as one file but normally the GDT is in its own file, enable pmode is in its own file, and enable A20 is in its own file. (.inc files) during assembly I have nasm load them into the kernel as one file using this: %include <file name>

Other then the fact that everything is in one file it is the same as my current kernel.

I thought of something. No matter what on shutdown bochs prints out the register values. So in that thinking I can test how far it gets by putting a value in a register at certain points. The further it gets the higher the value goes. The question here is this: Can I use 64 bit registers even though I'm in 32 bit? If so I would use register R8, It is unused by my OS so is a perfect canidate. If I can't I will use register ebp because it is unused so far.

Martijn · Post by **Martijn** » Tue May 27, 2008 3:51 pm

A few bugs i found so far:
1. 'bits 32' must appear after your far jump to 'Kernel'.
2. Your far jump to 'Kernel' must have the linear address of 'Kernel' as its offset.
3. There is no linear address to the gdt in your gdt_ptr.

JAAman · Post by **JAAman** » Tue May 27, 2008 4:02 pm

The question here is this: Can I use 64 bit registers even though I'm in 32 bit?

no, you cant -- to use the 64bit registers, requires removing existing instructions to turn them into prefixes -- which makes it incompatible with 16/32 bit software (the same is not true for 32bit registers, therefore those work in 16bit modes, but 64bit registers cannot work in 16/32bit modes)

while most code can run unaltered in 64bit mode, not all can, which is why there is a 32bit LMode

tadada · Post by **tadada** » Tue May 27, 2008 7:56 pm

ok I think I got somewhere. Now it says that it indeed loads the GDT and activates pmode. The problem is now with the jump to my kernel.

The exact log:

00001796618i[BIOS ] Booting from 0000:7c00
00002626616e[CPU0 ] fetch_raw_descriptor: GDT: index (f)1 > limit (0)
00002626616i[CPU0 ] CPU is in protected mode (active)
00002626616i[CPU0 ] CS.d_b = 16 bit
00002626616i[CPU0 ] SS.d_b = 16 bit

; few lines later

00002626616i[CPU0 ] >> jmp far 0008:0587 : EA87050800
00002626616e[CPU0 ] exception(): 3rd (13) exception with no resolution, shutdown status is 00h, resetting

I don't understand the error but I'm sure that I can find some Boch docs that exsplain what it means. (hafta look tomorrow though)

neon · Post by **neon** » Tue May 27, 2008 8:31 pm

You are getting a General Protection Fault (#GPF) exception. I can't help much without looking at your current code though..

tadada · Post by **tadada** » Wed May 28, 2008 1:54 pm

I was running through everything in the debugger again and I come to the far jump. When it far jumps it goes to 0008:0587 (EA87050800) but right after it far jumps again. (wrong) If it jumps correctly then it should jump once and then set ds, ss, and es to 0x10. (shouldn't that be to GDT entries that correspond to something more... correct? like for SS to where I want the stack to be and for ds the data entry in the GDT? ES being "extra segment" I believe can be to where I chose though...) I think the second jump might be from the fact that it is loaded into 0x500 BUT is ORG [0x000]. Can someone verify whether that could cause the problem or not?

neon · Post by **neon** » Wed May 28, 2008 4:15 pm

ss,es,ds,gs,fs all should contain the data segment selector offset in your GDT. Normally its usually at offset 0x10. It depends on your GDT though.

As for the second far jump, where does it jump to? Does it set CS to anything besides your 0x8? If it does, then that very well can be the cause of your #GPF.

ES being "extra segment" I believe can be to where I chose though...

As long as its a valid descriptor offset in your GDT, then yes. If not, it will #GPF as soon as you use it.

I think the second jump might be from the fact that it is loaded into 0x500 BUT is ORG [0x000].

Wait... 0x500? What is loaded at 0x500? Your kernel?