FASMARM generates object file with invalid architecture

Index > Non-x86 architectures > FASMARM generates object file with invalid architecture

Author

Thread

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

I'm not sure what's wrong here. For the following source:

Code:

format ELF
section '.text' executable
nop

FASM generates the right architecture i386 (objdump output):

Quote:

elf32_x86.o: file format elf32-i386
architecture: i386, flags 0x00000010:
HAS_SYMS
start address 0x00000000

FASMARM seems to set the architecture to an invalid value:

Quote:

elf32.o: file format elf32-little
architecture: UNKNOWN!, flags 0x00000010:
HAS_SYMS
start address 0x00000000

I tried to add PROCESSOR and COPROCESSOR directives but no luck.

05 May 2020, 20:03

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

fasmarm is designed as set of patches and with replacing instruction handling to ARM arch.
I found bug in fasmarm for 2 years ago in if-blocks handling (absent in fasm & specific to fasmarm - it is still unresolved) (in opposition to thour case that not so trivial as thours).
If thou could find what value should be thou could do additional patch thourself.

maybe even I could provide patch for thou: what value? by which value to replace?

05 May 2020, 21:47

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

for x86 these:

Code:

format_elf:
        test    [format_flags],8
        jnz     format_elf64
        mov     edx,edi
        mov     ecx,34h shr 2
        lea     eax,[edi+ecx*4]
        cmp     eax,[tagged_blocks]
        jae     out_of_memory
        xor     eax,eax
        rep     stos dword [edi]
        mov     dword [edx],7Fh + 'ELF' shl 8
        mov     al,1
        mov     [edx+4],al
        mov     [edx+5],al
        mov     [edx+6],al
        mov     [edx+14h],al
        mov     byte [edx+12h],3  ; patch here
        mov     byte [edx+28h],34h
        mov     byte [edx+2Eh],28h
        mov     [code_type],32   ; patch here
        mov     byte [edx+10h],2
        cmp     word [esi],1D19h
        je      format_elf_exe
        mov     byte [edx+10h],3
        cmp     word [esi],021Eh
        je      format_elf_exe
      elf_header_ok:

and provided patches for ARM:

Code:

        patch format_elf,       <<mov byte[edx+12h],3>>,        <<mov byte[edx+12h],40>>,       62      ;ARM
        patch format_elf,       <<mov [code_type],32>>,         <<mov [code_type],ARM_code>>,   74      ;CODE32
        patch elf_exe_brand_ok, <<mov [image_base],8048000h>>,  <<mov [image_base],8000h>>,     0

try
replace:

Code:

        patch format_elf,       <<mov byte[edx+12h],3>>,        <<mov byte[edx+12h],40>>,       62      ;ARM

with this:

Code:

        patch format_elf,       <<mov     byte[edx+12h],3>>,        <<call ARM_elf_fix_machine_flags>>,       62      ;ARM

and somewhere in the bottom of ARMv8.INC add this

Code:

ARM_elf_fix_machine_flags:
        mov byte[edx+12h],40 ; EM_ARM
        mov byte[edx+27h],5  ; ABI version 5
        ret

no guaranties (I am a last man on this forum for whom thou could address linux related questions)

05 May 2020, 22:02

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 18112
Location: In your JS exploiting you and your system

revolution

Okay, another bug I need to fix. Thanks for the report.

For various legal and virus related reasons I can't access any of the code right now. Sorry, It might be another two months before I can get back to my dev machine.

06 May 2020, 04:14

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

All right, thanks for the info. I'd like to fix it myself.

Can you guys help me recompile FASMARM? It seems that FASM sources are not a part of FASMARM sources and I have no idea where should I place the FASM sources.

06 May 2020, 07:37

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

use sources of fasmarm in https://yadi.sk/d/HTs1wxZvp6BmYw (only they comptibable with last fasm 1.72-1.73.24)
for compilation use fasmw.exe in BIN (it same as official one but in fasmw.ini there are paths to proper assembly)
And source has no patch suggested in this topic. By the way there are fasmarm sources for all OSes, except 64 bit variants.

Good Luck in arm elf research process. Hope thou share results - they could be usefull for community.

06 May 2020, 08:19

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

ProMiNick, thank you for your answers.

I added this code at the bottom of ARMv8.INC and FASMARM.ASM compiles without errors:

Code:

ARM_elf_fix_machine_flags:
        mov byte[edx+12h],40 ; EM_ARM
        mov byte[edx+27h],5  ; ABI version 5
        ret

However, when I replace the line with patch as suggested, I get error in the patch macro:

Quote:

Length original: 4, Length replacement: 5

I have no idea what's wrong. I'm not familiar with fasm macro language. I also dn't understand how the "offset" parameter actually works.

And I'm willing to share everything but I don't do anything special, I just need to create an ELF object file with fasmarm. Feel free to ask if you have questions.

09 May 2020, 15:51

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

Code:

        patch format_elf,       <<mov byte[edx+12h],3>,<mov byte [edx+28h],34h>>,        <<call ARM_elf_fix_machine_flags>,nop,nop,nop>,       62      ;ARM

Code:

ARM_elf_fix_machine_flags:
        mov byte[edx+12h],40 ; EM_ARM
        mov word[edx+27h],3405h  ; ABI version 5 & smthing that was 34h
        ret

patched & patch bytes should be equal size, call 1 byte bigger so increase patch size to fit it.

09 May 2020, 19:34

guignol

Joined: 06 Dec 2008
Posts: 756

guignol

revolution wrote:

For various legal and virus related reasons I can't access any of the code right now. Sorry, It might be another two months before I can get back to my dev machine.

I wonder...

It might as well be two years

11 May 2020, 07:07

guignol

Joined: 06 Dec 2008
Posts: 756

guignol

I do notice you reside on the forum for quite a time, most of your day maybe

11 May 2020, 07:11

guignol

Joined: 06 Dec 2008
Posts: 756

guignol

ProMiNick wrote:

By the way there are fasmarm sources for all OSes, except 64 bit variants.

yeah, i mean, it's the near future of "personal computing"
fasm for Android, too, once again

ProMiNick wrote:

Good Luck in arm elf research process. Hope thou share results - they could be usefull for community.

aww, elves...

11 May 2020, 07:16

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

guignol, modern androids have nothing common with linux(
Thou cant run elf on them.
Jnly dalvik scripts.
fasmarm will not help. dalvik different architecture.

11 May 2020, 14:53

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

ProMiNick, you can run ELF executables on Android easily: https://board.flatassembler.net/topic.php?t=21327

11 May 2020, 15:02

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

Some time ago that was still possible to run elfes.
Modern androids fix that so nothing except dalvic could not be run in android environment. Or thou targeted to android versions that soon became out of date?

11 May 2020, 22:33

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

Let's move the discussion about ELF executables and Android here: https://board.flatassembler.net/topic.php?p=213974#213974

12 May 2020, 14:25

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

ProMiNick, your last patch seems to work, thanks. The byte at offset 0x27 should be 0x05, as documented in ARM manual:

Quote:

EF_ARM_ABIMASK (0xFF000000) (current version is 0x05000000): This masks an 8-bit version number, the version of the ABI to which this ELF file conforms. This ABI is version 5. A value of 0 denotes unknown conformance.

I've just found that the objdump output is confusing, it says "architecture: UNKNOWN!" even when I patch the byte to 0x05.

The readelf utility has better output in this case. For the object file generated with fasmarm, using -h option, it says:

Code:

ELF Header:
  Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF32
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
  Type:                              REL (Relocatable file)
  Machine:                           ARM
  Version:                           0x1
  Entry point address:               0x0
  Start of program headers:          0 (bytes into file)
  Start of section headers:          3146176 (bytes into file)
  Flags:                             0x0
  Size of this header:               52 (bytes)

But the correct Flags is:

Code:

  Flags:                             0x5000000, Version5 EABI

12 May 2020, 15:01

ProMiNick

Joined: 24 Mar 2012
Posts: 640
Location: Russian Federation, Sochi

ProMiNick

try this?

Code:

ARM_elf_fix_machine_flags:
        mov byte[edx+7],97 ; ELFOSABI_ARM
        mov byte[edx+12h],40 ; EM_ARM
        mov word[edx+27h],3405h  ; ABI version 5 & smthing that was 34h
        ret

13 May 2020, 05:38

MazeGen

Joined: 06 Oct 2003
Posts: 977
Location: Czechoslovakia

MazeGen

Thanks, the "Version5 EABI" flag is enough for me because I can link the file now.

25 May 2020, 10:16

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum