Elfloader – An Architecture-Agnostic ELF File Flattener For Shellcode
elfloader
is a super simple loader for ELF files that generates a flat in-memory representation of the ELF.
Pair this with Rust and now you can write your shellcode in a proper, safe, high-level language. Any target that LLVM can target can be used, including custom target specifications for really exotic platforms and ABIs. Enjoy using things like u64
s on 32-bit systems, bounds checked arrays, drop handling of allocations, etc :)
It simply concatenates all LOAD
sections together, using zero-padding if there are gaps, into one big flat file.
This file includes zero-initialization of .bss
sections, and thus can be used directly as a shellcode payload.
If you don't want to waste time with fail-open linker scripts, this is probably a great way to go.
This doesn't handle any relocations, it's on you to make sure the original ELF is based at the address you want it to be at.
Usage
To use this tool, simply:
Usage: elfloader [--binary] [--base=<addr>] <input ELF> <output>
--binary - Don't output a FELF, output the raw loaded image with no
metadata
--base=<addr> - Force the output to start at `<addr>`, zero padding from
the base to the start of the first LOAD segment if needed.
`<addr>` is default hex, can be overrided with `0d`, `0b`,
`0x`, or `0o` prefixes.
Warning: This does not _relocate_ to base, it simply starts
the output at `<addr>` (adding zero bytes such that the
output image can be loaded at `<addr>` instead of the
original ELF base)
<input ELF> - Path to input ELF
<output> - Path to output file
To install this tool run:
cargo install --path .
Now you can use elfloader
from anywhere in your shell!
Dev
This project was developed live here:
https://www.youtube.com/watch?v=x0V-CEmXQCQ
Example
There's an example in example_small_program
, simply run make
or nmake
and this should generate an example.bin
which is 8 bytes.
pleb@gamey ~/elfloader/example_small_program $ make
cargo build --release
Finished release [optimized] target(s) in 0.03s
elfloader --binary target/aarch64-unknown-none/release/example_small_program example.bin
pleb@gamey ~/elfloader/example_small_program $ ls -l ./example.bin
-rw-r--r-- 1 pleb pleb 8 Nov 8 12:27 ./example.bin
pleb@gamey ~/elfloader/example_small_program $ objdump -d target/aarch64-unknown-none/release/example_small_program
target/aarch64-unknown-none/release/example_small_program: file format elf64-littleaarch64
Disassembly of section .text:
00000000133700b0 <_start>:
133700b0: 8b000020 add x0, x1, x0
133700b4: d65f03c0 ret
Now you can write your shellcode in Rust, and you don't have to worry about whether you emit .data
, .rodata
, .bss
, etc. This will handle it all for you!
There's also an example with .bss
and .rodata
pleb@gamey ~/elfloader/example_program_with_data $ make
cargo build --release
Finished release [optimized] target(s) in 0.04s
elfloader --binary target/aarch64-unknown-none/release/example_program_with_data example.bin
pleb@gamey ~/elfloader/example_program_with_data $ ls -l ./example.bin
-rw-r--r-- 1 pleb pleb 29 Nov 8 12:39 ./example.bin
pleb@gamey ~/elfloader/example_program_with_data $ objdump -d target/aarch64-unknown-none/release/example_program_with_data
target/aarch64-unknown-none/release/example_program_with_data: file format elf64-littleaarch64
Disassembly of section .text:
0000000013370124 <_start>:
13370124: 90000000 adrp x0, 13370000 <_start-0x124>
13370128: 90000008 adrp x8, 13370000 <_start-0x124>
1337012c: 52800029 mov w9, #0x1 // #1
13370130: 91048000 add x0, x0, #0x120
13370134: 3904f109 strb w9, [x8, #316]
13370138: d65f03c0 ret
pleb@gamey ~/elfloader/example_program_with_data $ readelf -l target/aarch64-unknown-none/release/example_program_with_data
Elf file type is EXEC (Executable file)
Entry point 0x13370124
There are 4 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000120 0x0000000013370120 0x0000000013370120
0x0000000000000004 0x0000000000000004 R 0x1
LOAD 0x0000000000000124 0x0000000013370124 0x0000000013370124
0x0000000000000018 0x0000000000000018 R E 0x4
LOAD 0x000000000000013c 0x000000001337013c 0x000000001337013c
0x0000000000000000 0x0000000000000001 RW 0x4
GNU_STAC K 0x0000000000000000 0x0000000000000000 0x0000000000000000
0x0000000000000000 0x0000000000000000 RW 0x0
Section to Segment mapping:
Segment Sections...
00 .rodata
01 .text
02 .bss
03
Internals
This tool doesn't care about anything except for LOAD
sections. It determines the endianness (little vs big) and bitness (32 vs 64) from the ELF header, and from there it creates a flat image based on program header virtual addresses (where it's loaded), file size (number of initialized bytes) and mem size (size of actual memory region). The bytes are initialized from the file based on the offset and file size, and this is then extended with zeros until mem size (or truncated if mem size is smaller than file size).
These LOAD
sections are then concatenated together with zero-byte padding for gaps.
This is designed to be incredibly simple, and agnostic to the ELF input. It could be an executable, object file, shared object, core dump, etc, doesn't really care. It'll simply give you the flat representation of the memory, nothing more.
This allows you to turn any ELF into shellcode, or a simpler file format that is easier to load in hard-to-reach areas, like embedded devices. Personally, I developed this for my MIPS NT 4.0 loader which allows me to run Rust code.
FELF0001 format
This tool by default generates a FELF file format. This is a Falk ELF. This is a simple file format:
FELF0001 - Magic header
entry - 64-bit little endian integer of the entry point address
base - 64-bit little endian integer of the base address to load the image
<image> - Rest of the file is the raw image, to be loaded at `base` and jumped
into at `entry`