Bootloader doesn't get to halt instruction

[nekros]

Ok, So I've been working on my bootloader lately. I tested it on qemu for the first time, and it did absolutely nothing.
I have this code at the beginning:

Code: Select all

begin:
mov sp,stackstart
push cs; THANKS TROY!
pop ds
push cs
pop es    ;done setting segments
mov BYTE[drive],dl     ; store the boot drive
;;;;;;;;;;;;;;;Load next 3 Sectors To complete stage1;;;;;;;;;;;;;;;;;;
mov ah,0x08
mov dl,BYTE [drive] 
int 0x13        ; get drive parameters
jc FAIL     ;did this faiL?
;else store info for LBA TO CHS

Really nothing here that should hold anything up. And FAIL is supposed to print a string and then hlt. (I haven't used my readsector code yet).
After that is this section of code:

Code: Select all

mov BYTE [HPC], dh ;//store heads per cylinder 

mov ax,0 ;; start counter for all the bits

testandset: ; tests and sets all the sector bits in cl
bt cx,ax     ; test bit
jc set       ; is the bit set? if yes set it in SPT
inc ax    ; if no go to next bit and test set
cmp ax,6
je continue ;if we are finished, continue with the rest of  the code
jmp testandset


set:
bts WORD [SPT],ax
inc ax
cmp ax,6
je continue ;if done setting, continue
jmp testandset ;we are not done

Continue does the same thing as FAIL, I think their might be something wrong with the loop above and the loop never ends. Should I use test instead of cmp?

EDIT:
I tried using test, it doesn't change anything.

EDIT2:
I bet I'll be facepalming soon.

[Love4Boobies]

Erm, why the heck don't you

Code: Select all

PUSH CS
POP DS
MOV ES,DS

instead of pushing CS twice?

And HEADS PER CYLINDER? ROFL! Dude, I'm not sure OSDev is the right thing for you

[Troy Martin]

Heh, I was thinking quickly at the time (remembered the opcode, wasn't thinking "optimizations" xD)

[nekros]

LBA to CHS formula I got from wikipedia. I'm going to bonafide to check if they are different...

[nekros]

Wikipedia:

CYL = LBA / (HPC * SPT)
TEMP = LBA % (HPC * SPT)
HEAD = TEMP / SPT
SECT = TEMP % SPT + 1

Bonafide:

Sector = (LBA/SectorsPerTrack) Remainder value + 1
Cylinder = (LBA/SectorsPerTrack)/NumHeads (Take Remainder value)
Head = (LBA/SectorsPerTrack)/NumHeads (Take quotient value)

\

That still doesn't explain why I don't get to the print and then halt instructions.

[nekros]

I'm starting to think that there is a problem with qemu. Since I switched the print string call to the start and it only prints one character.... The code for printing a string is from the babystep2 tutorial. I've used that little bit of code in the past and it worked so somethings not right here...

EDIT:
Wrong again, I tested it on real hardware...

[nekros]

I now know that something is causing a halt. I tried executing the print string function twice and I only got one character. It's not the assembler (I tried nasm), what would cause a halt then?

[nekros]

I found the bug, it was setting sp to stackstart. I changed it to (0x7c00-1) and it worked. I don't understand why though, so if anyone has an answer please tell me...

[JohnnyTheDon]

You always need to set the stack pointer to the end of the stack, not the start.

[JohnnyTheDon]

Erm, why the heck don't you

Code: Select all

PUSH CS
POP DS
MOV ES,DS

instead of pushing CS twice?

And HEADS PER CYLINDER? ROFL! Dude, I'm not sure OSDev is the right thing for you

That code results in:

Code: Select all

[john@john-arch ~]$ yasm -f bin test.s
test.s:5: error: invalid combination of opcode and operands

So unless yasm sucks that bad, I don't think you get to tell him to stop doing os dev.

[nekros]

I officially facepalm myself for missing that...
Oh well, back to happily coding until another facepalm...

[JohnnyTheDon]

Don't worry, every time I start a bootloader I make that mistake

[Love4Boobies]

I was referring to the thing with the heads per cylinder

[xenos]

Erm, why the heck don't you

Code: Select all

PUSH CS
POP DS
MOV ES,DS

MOV ES, DS not a valid instruction. You can't move the content of one segment register into another. But this should work:

Code: Select all

MOV AX, CS
MOV DS, AX
MOV ES, AX

[Love4Boobies]

I was referring to the thing with the heads per cylinder

The number of heads is indeed counted per cylinder. Think for a minute what a cylinder is.

No, it's actually the other way around.