The source file is in intel 32 bit .hex format, it was built for a nuvoton NUC220LE3AE. I want to decompile it to C++, or at least assembly. I have some experience with assembly, but mostly with MIPS and that was years ago - so high level code is preferred.

Link to .hex file: http://www.3fvape.com/images/3fvape-blog-img/20150806-4384-xcubeII-upgrade/SMOK_X_CUBE_II_firmware_v1.07.hex

The device it resides in is a SMOK Xcube 2 (e-cig box mod.)

My goal is to add an option to control the LED color based on the atomizer temperature.

I searched but could not find an applicable how-to.


Short answer: You can't.

At least, don't expect a readable, and compilable, C source code. There's discussion why elsewhere on this site, so i won't get into details.

Also, note there's not an easy walk-through or how-to. You need to experiment, and you'll need some experience as well.

To get you started, you might:

  • convert the .hex file into a raw binary file, for example using Hex2bin
  • use that binary file with the retargetable Decompiler selecting "raw machine code" and probably "ARM+Thumb" as architecture
  • If the results of of the retargetable decompiler are unsatisfactory (it didnt work well for me, last time i tried), you might want to try the Online Disassembler to get assembly code
  • Of course, the ultimate tool is IDA, but the freeware version can't handle ARM, and the price is probably a bit steep for a hobby project.
| improve this answer | |
  • The online disassebmler only supports up to armv5, but the coretex m0 this is based on uses armv6. I think my company has a license for IDA, but I've gotta ask around on the specifics. – KING_BOZAR Dec 31 '15 at 18:46
  • 1
    You could just use <yourarch>-objdump -b binary -xD file.bin on the raw binary. – domen Jan 22 '16 at 12:49

radare2 supports these ihex files directly

note the command in radare2 in my original answer uses switch -b32 SYS_V commented that it should be -b16 and posted an answer with ample details how to proceed from the end of my answer to a tangible result here is the link to that thread how to find usefull info from a bin file

:\>ls -l
total 172
-rw-rw-rw-  1 Admin 0 172401 2016-01-01 00:44 SMOK_X_CUBE_II_firmware_v1.07.hex

:\>rahash2 -a md5 SMOK_X_CUBE_II_firmware_v1.07.hex
SMOK_X_CUBE_II_firmware_v1.07.hex:0x00000000-0x0002a170 md5: 351660a42b846d19e35f54f75530e2d9

:\>radare2 -A -a arm -b 32 ihex://SMOK_X_CUBE_II_firmware_v1.07.hex
Function too big at 0xa50e54
Function too big at 0xfe25a2ac
Function too big at 0x1648234
Function too big at 0x13ed738
[0x00000000]> s 0xc1
[0x000000c1]> pd 10
|          0x000000c1    4885460c       mcrreq p5, 4, r8, r6, c8
|          0x000000c5    f070fc00       ldrshteq r7, [ip], 0
|          0x000000c9    480047e9       stmdb r7, {r3, r6} ^
|          0x000000cd    1b0000e8       stmda r0, {r0, r1, r3, r4}
|          0x000000d1    0e002004       strteq r0, [r0], -0xe
|          0x000000d5    48804704       strbeq r8, [r7], -0x48
|          0x000000d9    480047fe       cdp2 p0, 4, c0, c7, c8, 2
|          0x000000dd    e7fee7fe       cdp2 p14, 0xe, c15, c7, c7, 7
|          0x000000e1    e7fee7fe       cdp2 p14, 0xe, c15, c7, c7, 7
|          0x000000e5    e7fee75d       stclpl p14, c15, [r7, 0x39c]!

if you were wondering how did radare2 get the 4885460c at address 0xxxxc1 then read further

intel seems to have published the specs of ihex file format i got hold of one pdf from microsym named intelhex.pdf dont know if any latest version are available this is my first brush with ihex or arm for that matter

Hexadecimal Object File Format Specification Revision A January 6, 1988

based on the description in the file

it seems each line in the ihex file starts with a colon :
followed by ONE BYTE = record length
followed by TWO BYTES = offset to load
followed by ONE BYTE = Record Type
Last BYTE in the line = Checksum

each of the above are hexpairs ie BYTE E8 will be 0x45 0x38 in the file

3A 31 30 30 30 30 30 30 30 45 38 30 45 30 30 :10000000E80E00

the file consists of 3833 lines out of which 3830 lines have a record length of 0x10

wc SMOK_X_CUBE_II_firmware_v1.07.hex
  3833   3833 172401 SMOK_X_CUBE_II_firmware_v1.07.hex

grep -ivn :10 SMOK_X_CUBE_II_firmware_v1.07.hex

dissecting the first line

line1 data size   = 0x02 
load offset       = 0x0000
record type is    = 0x04  (extended linear address 32 bit format)

and it stays as is until another record 04 is encountered lets check if the file contains another record 04

:\>grep -in :......04 SMOK_X_CUBE_II_firmware_v1.07.hex

only one line the last line 3833 denotes end of file record type 0x01
the last but one line denotes start linear address record type 0x05 and EIP = 0xc1 checksum matches (100 - (0xc1 + 0x05 + 0x04 ) ) == 0x36

so that leaves 3830 lines as code data xx

based on the above details we can carve the bytes

delete the first , last . last but one lines from the input file
sed substitute the first nine characters and last two character and rip off the line endings \r\n .

sed s/:........//g < in | sed s/..$//g | tr -d  "\r\n" > out

lets check out if sed magic worked we should have 32 characters per line if it worked concatenated into one big string

:\>wc out
     0      1 122560 out

:\>set /a 32*3830

lets convert the hex pairs to binary

rax2 -s < out > bin 

this doesn't work as it should deliver us half the input size but it is higher than that also the inversion rax2 -S < bin > is_original doesn't get us the original input back

to the developers of radare if you read this can you check if rax2 -s works properly in windows if input is file it seems it suffers from unix / windows line ending quirks (windows seems to convert the binary 0x0A to 0x0d 0x0a when using the redirection operator

:\>rax2 -s < out > bin

:\>wc bin
  616  4126 61896 bin

:\>set /a 61896 * 2

lets cook a python unhexlify

:\>cat makebin.py
import binascii
fp = open("out","rb")
fo = open("bin","wb")
buff = fp.read()
:\>python makebin.py

:\>wc *
   616   4126  61280 bin
  3830   3830 172350 in
     6     14    139 makebin.py
     0      1 122560 out
   616   4126  61896 rax2bin
  5068  12097 418225 total

:\>set /a 61280*2

python seems to deliver us the correct bytes

lets ask radare2 if the size 122560 is right

[0x000000c1]> if
file     ihex://SMOK_X_CUBE_II_firmware_v1.07.hex
fd       2357404
size     0xef60
blksz    0x0
mode     r--
block    0x100
format   any

[0x000000c1]> !rax2 0xef60*2

it seems to agree lets xxd and look if we fished the right bytes at right offset

:\>xxd -g 4 -l 32 -s 0xc1 bin
00000c1: 4885460c f070fc00 480047e9 1b0000e8  H.F..p..H.G.....
00000d1: 0e002004 48804704 480047fe e7fee7fe  .. .H.G.H.G.....

it looks right now we can try and start understanding the mnemonics (google doesn't seem to know much about mccreq i don't know arm so good-luck from here on

| improve this answer | |
  • 1
    The target processor, ARM Cortex-M0, uses a 16-bit Thumb instruction set even though its architecture is 32-bit – julian Mar 26 '17 at 23:06
  • @SYS_V thanks for the info i see you posted an answer to the other thread that refers to this question with lots of details ill add a foot note to the answer referring to that thread and this comment – blabb Mar 27 '17 at 5:07
  • You are welcome. The explanation in this post of how to use r2 to disassemble .hex files was very useful – julian Mar 27 '17 at 13:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.