Created
April 11, 2018 12:23
-
-
Save dwaq/fb9afcb0ad97068251d2e22f4b26d1e0 to your computer and use it in GitHub Desktop.
Notes for YouTube Video: https://www.youtube.com/watch?v=KguyjVDlH1o
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
https://www.youtube.com/watch?v=KguyjVDlH1o | |
## Master Boot Record | |
copy image of flash card | |
move to ubunutu | |
full disc image so has MBR | |
MBR tells about partitions, code to boot bootloader (512 bytes) | |
> xxd file.img | less | |
> xxd -g1 | |
(seperates bytes) | |
error message at beginning, 0x55AA at end (ways to tell it's a 1st stage bootloader) | |
https://en.wikipedia.org/wiki/Master_boot_record | |
> sudo apt install nasm | |
> ndisasm -b 16 file.img | |
(included with nasm) | |
setting bit width | |
code that bios moves into memory and runs | |
0x7C00 - MBR location | |
this code moves itself (the MBR) to 0x6000, then the 2nd stage bootloader overwrites the original data at | |
0x7c00 | |
https://wiki.osdev.org/Expanded_Main_Page | |
uses xxd to find where data begins (the text of the file: Invalid...) and code ends | |
http://www.ctyme.com/intr/int.htm | |
code: | |
int 0x18 ; just a boot hook: http://www.ctyme.com/intr/rb-2241.htm | |
; WRONG: but uses data in AX (AX=0 from above) xoring itself generates a 0 (xor ax,ax) | |
; so it's keyboard - get stroke (that doesn't seem right...???) | |
; legacy boot to ROM BASIC? | |
(interupt 18, go to page above and see that interrupt) | |
int 0x10 ; ah=0xe, teletype out with options | |
int 0x13 ; ax=0201, 02 = read sectors into memory, 01 = 1 sector | |
int 0x13 ; ax=00, recalibrate drive heads | |
## Extract partition | |
# mount file for VM: | |
> losetup ; set up and control loop devices | |
> mount -o loop file.img /mnt | |
> sudo losetup -f ; find unused loop devices | |
responds with /dev/loop0 | |
> sudo losetup /dev/loop0 -P -r file.img; partitioned, read only | |
> dmesg ; check device driver messages? | |
> ls -l /dev/loop0 ; tab autocomplete, shows partition table was mounted (now a loop0p1) | |
> sudo fdisk /dev/loop0 ; now able to see partition table (same as we were looking at in hex editor) | |
## Mount filesystem: | |
> mkdir mnt | |
> sudo mount /dev/loop0p1 mnt | |
> cd mnt | |
> ls -al ; viewing contents | |
> tar zcvpf archive-fs.tar.gz mnt ; archive the folder | |
> sudo xxd /dev/loop0 ; exactly the same as looking at hex dump from before | |
> sudo xxd /dev/loop0p1 | less; look at first partition | |
ends with 0x55AA at 0x200. this is a PSOS bootloader, not a DOS bootloader | |
PSOS is an old RTOS | |
this data is a partition boot record, not a MBR, but can look at it using the same method | |
> sudo dd if=/dev/loop0p1 of=partitionboot.img bs=512 count=1 ; copy single 512 byte block | |
> xxd partitionboot.img | |
> ndisasm -b 16 partitionboot.img > partitionboot.asm | |
> xxd -g1 partitionboot.img >> partititonboot.asm ; append hexdump to end of asm file | |
disassembler automatically assumed wrong instructions | |
1st instruction in code jumps to 0x2b, which is disassembled as the middle of an instruction | |
need to fix this | |
> ndisasm -b 16 -e 0x2b -o 0x7c2b partitionboot.img ; -e moves to start point (where the first | |
instruction jumped to). and origin to 0x7c2b - 0x7c00 is start point, plus 0x2b offset | |
; looks alright now, try to disassemble again | |
> ndisasm -b 16 partitionboot.img > partitionboot.asm | |
calling a function: example: | |
"call word 0x1234" | |
then go to that address | |
and it will end with | |
"ret" | |
lgdt ; load Global Descriptor Table - tells processor how memory is set up | |
https://en.wikipedia.org/wiki/Global_Descriptor_Table | |
loops in place, error condition (halt loop) | |
sti ; enable interrupts | |
hlt ; sleep until next interrupt | |
jmp ; jumps to the sti intstuction | |
# Disassemble the files from the filesystem | |
> xxd RAM.ABS | less ; expect it to be a mix of data and x86 code | |
find an application (code and data structures that code would point to) | |
> nsdisasm -b 32 RAM.ABS | less | |
# filesystem info | |
BIOS & MBR, partititon boot record -> 512 bytes sitting at the front of the disc | |
PSOSBOOT.SYS & RAM.ABS -> inside FAT filesystem | |
## x86 architexture reference | |
https://en.wikipedia.org/wiki/Control_register#CR0 | |
looking at: | |
mov eax,cr0 | |
or eax, 0x60000000 | |
bits 30, 29 being set -> cache disable, not write thru (globally turning off cache) | |
-> "about to do serious business with the processor" | |
lidt [0x123] | |
lgdt [0x123] | |
Load Global/Interrupt Descriptor Table | |
loading these with absolute address | |
assumption was that these addresses were inside data | |
but the total file size is 0x1c..and idt is at 0x22 which is past the end | |
## assuming that the ABS file starts at 1MB offset | |
(based on the above assumptions about the interrupt table's location) | |
> nsdisasm -b 32 -o 0x100000 RAM.ABS | less | |
in less: /term to search | |
create symbolic link (shortcut) | |
> ln -sf orig.file new.location | |
# Binary Ninja | |
needs python script to correctly format hex file as code (view of partial code at 2:55:38) | |
viewing log seems to be helpul | |
could find strings (in hex view) and then go back and try to find a related print function | |
strings with 0x00 between characters are 16 bit strings on 32 bit system | |
4:14:00 | |
Binary Ninja | |
8:28:35 | |
## booting from hard disk | |
qemu-system-i386 -hda tmp.img | |
or can do -hdb for 2nd hdd | |
qemu-system-i386 -kernal RAM.ABS | |
## custom bootloader | |
git clone github.com/scanlime/metalkit | |
cd metalkit/examples/scanpci | |
make | |
generates scanpci.img | |
qemu-system-i386 -fda scanpci.img | |
fda is floppy | |
hex editor software: https://www.sweetscape.com/010editor/ | |
multiboot is interface between bootloader and linux kernel |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment