0

Currently, I'm reading the Linux kernel source code of v0.01. I'm trying to imitate the code and make it work. But got stuck at the very beginning of the bootsector. In bootsect.s, the program first load at 0x7c00 and move itself to 0x90000, then jump to 0x90000, but somehow the 'jmpi' instruction is not work.

SETUPLEN = 4                 ! len of setup-sectors, 4 sectors
BOOTSEG  = 0x07c0            ! original address of boot sector
INITSEG  = 0x9000            ! move bootsect here
SETUPSEG = 0x9020            ! setup starts here
SYSSEG   = 0x1000            ! system loaded at 0x10000 (65536)
ENDSEG   = SYSSEG + SYSSIZE  ! where to stop loading

entry start
start:
    mov ax, #BOOTSEG
    mov ds, ax
    mov ax, #INITSEG
    mov es, ax
    mov cx, #256
    sub si, si
    sub di, di
    rep
    movw                       ! move bootsect itself to address 0x90000  (256 * 2 Bytes)
    jmpi go, INITSEG           ! jump to address 0x90000 and execute
    !j go

go:
    mov ax, cs
    mov ds, ax
    mov es, ax
    mov ss, ax
    mov sp, #0xFF00            ! Top of the stack - ss:sp address is 0x9FF00
    
    ! print something
    mov ah, #0x03             ! read cursor pos
    xor bh, bh
    int 0x10
    
    mov cx, #7
    mov bx, #0x0007           ! page 0, attribute 7 (normal)
    mov bp, #msg
    mov ax, #0x1300           ! write string, move cursor
    int 0x10
 msg:
    .byte 13,10
    .ascii "!"
    .byte 13,10,13,10

one more thing I'd like to mention is if I replace the 'jmpi go, INITSEG' with 'j go', it will definitely work and jump to the label 'go', but something weird is the character printed on the screen is not what I expected as '!', but showing something like "≡S"

I am not sure if 'jmpi' does totally not work or just because there's something wrong with the print(0x10) functions so that I can not see anything in the screen.

I am a beginner in assembly language and I have no idea what's wrong in the code, even the code is almost copied from Linux kernel.

Peter Cordes
  • 328,167
  • 45
  • 605
  • 847
Hu Vincent
  • 11
  • 2
  • This is as86 code, IIRC. I guess you must be using that, otherwise it wouldn't assemble. I don't know as86 syntax, but I can guess that's a FAR jump, which would set CS. (GAS uses `jmpl` for that). You can check with a disassembler how it assembles. Yes you could just to a relative `jmp`, but then you'd still be running the copy at `0x7c00`, not `0x9000`. – Peter Cordes May 27 '22 at 09:17
  • @peter, Yes, I'm using as86. I'm not familiar with that, but by using ojbdump86 -d -l boot.o, the result shows `SEG0 00000200 SYMS 21 ... SYM 8 0000007a ....T msg2 SYM 9 00000092 ....T msg ... SYM 14 00000015 ....T go ... BYTECODE SEG 0 SEG 3 SEG 0 : b8 c0 07 8e d8 b8 00 90 8e c0 b9 00 01 29 f6 29 .............).) : ff f3 a5 eb 00 8c c8 8e d8 8e c0 b4 03 30 ff b8 .............0.. : e4 07 89 c2 b9 03 00 bb 07 00 bd ........... RELSZ 2 DW SEG0+0x0092 ...` does this meaning when I print the 'msg', it haven't jumped to another SEG at 0x9000 ? – Hu Vincent May 27 '22 at 10:05
  • 1
    @PeterSmith, thanks for the idea about disassembler, I found I haven't removed the Minix header which is 32 bytes at the front of the binary. works for me now! – Hu Vincent May 27 '22 at 13:49

1 Answers1

1

Problem resolved!

I was using as86 & ld86 to compile and link. In order to make it bootable, I tried to set the 510 & 511 bytes to 0xAA, and 0x55. I expected there were only 512 bytes for the binary, and rewriting the 510&511 bytes would make it work.

But actually, there are 32 bytes MINIX header in the front of the binary which made the total size of the binary 544 bytes. rewriting the bytes on 510&511 makes the whole binary a mess.

Hu Vincent
  • 11
  • 2