In essence, a
disassembler is the exact opposite of an assembler. Where an assembler converts code written in an assembly language into binary machine code, a disassembler reverses the process and attempts to recreate the assembly code from the binary machine code.
Since most assembly languages have a one-to-one correspondence with underlying machine instructions, the process of disassembly is relatively straight-forward, and a basic disassembler can often be implemented simply by reading in bytes, and performing a table lookup. Of course, disassembly has its own problems and pitfalls, and they are covered later in this chapter.
Many disassemblers have the option to output assembly language instructions in Intel, AT&T, or (occasionally) HLA syntax. Examples in this book will use Intel and AT&T syntax interchangeably. We will typically not use HLA syntax for code examples, but that may change in the future.
Here we are going to list some commonly available disassembler tools. Notice that there are professional disassemblers (which cost money for a license) and there are freeware/shareware disassemblers. Each disassembler will have different features, so it is up to you as the reader to determine which tools you prefer to use.
ida37fw) A DOS GUI tool that behaves very much like IDA Pro, but is considerably more limited. It can disassemble code for the Z80, 6502, Intel 8051, Intel i860, and PDP-11 processors, as well as x86 instructions up to the 486.
Many of the Unix disassemblers, especially the open source ones, have been ported to other platforms, like Windows (mostly using MinGW or Cygwin). Some Disassemblers like otool ([OS X) are distro-specific.
computer_intelligence_assembler_disassembler. This Forth-based tool allows to incrementally and interactively build knowledge about a code body. It is unique that all disassembled code can be re-assembled to the exact same code. Processors are 8080, 6809, 8086, 80386, Pentium I en DEC Alpha. A scripting facility aids in analyzing Elf and MSDOS headers and makes this tool extendable. The Pentium I ciasdis is available as a binary image, others are in source form, loadable onto lina Forth, available from the same site.
$ echo '1 2' | llvm-mc -disassemble -triple=x86_64-apple-darwin9 addl %eax, (%rdx) $ echo '0x0f 0x1 0x9' | llvm-mc -disassemble -triple=x86_64-apple-darwin9 sidt (%rcx) $ echo '0x0f 0xa2' | llvm-mc -disassemble -triple=x86_64-apple-darwin9 cpuidr $ echo '0xd9 0xff' | llvm-mc -disassemble -triple=i386-apple-darwin9 fcos
As we have alluded to before, there are a number of issues and difficulties associated with the disassembly process. The two most important difficulties are the division between code and data, and the loss of text information.
Since data and instructions are all stored in an executable as binary data, the obvious question arises: how can a disassembler tell code from data? Is any given byte a variable, or part of an instruction?
The problem wouldn't be as difficult if data were limited to the .data section (segment) of an executable (explained in a later chapter) and if executable code were limited to the .code section of an executable, but this is often not the case. Data may be inserted directly into the code section (e.g. jump address tables, constant strings), and executable code may be stored in the data section (although new systems are working to prevent this for security reasons). AI programs, LISP or Forth compilers may not contain .text and .data sections to help decide, and have code and data interspersed in a single section that is readable, writable and executable, Boot code may even require substantial effort to identify sections. A technique that is often used is to identify the entry point of an executable, and find all code reachable from there, recursively. This is known as “code crawling”.
Many interactive disassemblers will give the user the option to render segments of code as either code or data, but non-interactive disassemblers will make the separation automatically. Disassemblers often will provide the instruction AND the corresponding hex data on the same line, shifting the burden for decisions about the nature of the code to the user. Some disassemblers (e.g. ciasdis) will allow you to specify rules about whether to disassemble as data or code and invent label names, based on the content of the object under scrutiny. Scripting your own “crawler” in this way is more efficient; for large programs interactive disassembling may be impractical to the point of being unfeasible.
The general problem of separating code from data in arbitrary executable programs is equivalent to the halting problem. As a consequence, it is not possible to write a disassembler that will correctly separate code and data for all possible input programs. Reverse engineering is full of such theoretical limitations, although by Rice's theorem all interesting questions about program properties are undecidable (so compilers and many other tools that deal with programs in any form run into such limits as well). In practice a combination of interactive and automatic analysis and perseverance can handle all but programs specifically designed to thwart reverse engineering, like using encryption and decrypting code just prior to use, and moving code around in memory.
User defined textual identifiers, such as variable names, label names, and macros are removed by the assembly process. They may still be present in generated object files, for use by tools like debuggers and relocating linkers, but the direct connection is lost and re-establishing that connection requires more than a mere disassembler. Especially small constants may have more than one possible name. Operating system calls (like DLLs in MS-Windows, or syscalls in Unices) may be reconstructed, as their names appear in a separate segment or are known beforehand. Many disassemblers allow the user to attach a name to a label or constant based on his understanding of the code. These identifiers, in addition to comments in the source file, help to make the code more readable to a human, and can also shed some clues on the purpose of the code. Without these comments and identifiers, it is harder to understand the purpose of the source code, and it can be difficult to determine the algorithm being used by that code. When you combine this problem with the possibility that the code you are trying to read may, in reality, be data (as outlined above), then it can be even harder to determine what is going on. Another challenge is posed by modern optimising compilers; they inline small subroutines, then combine instructions over call and return boundaries. This loses valuable information about the way the program is structured.
Akin to Disassembly,
'Decompilers' take the process a step further and actually try to reproduce the code in a high level language. Frequently, this high level language is C, because C is simple and primitive enough to facilitate the decompilation process. Decompilation does have its drawbacks, because lots of data and readability constructs are lost during the original compilation process, and they cannot be reproduced. Since the science of decompilation is still young, and results are “good” but not “great”, this page will limit itself to a listing of decompilers, and a general (but brief) discussion of the possibilities of decompilation. Compared to disassemblers a decompiler generates code that doesnot require that one is familiar at the processor at hand. It may even be that the decompiled code can be compiled on a different processor, or give a reasonable starting point to reproduce the program on a different processor.
In the face of optimizing compilers, it is not uncommon to be asked “Is decompilation even possible?” To some degree, it usually is. Make no mistake, however: an optimizing compiler results in the irretrievable loss of information. An example is in-lining, as explained above, where code called is combined with its surroundings, such that the places where the original subroutine is called cannot even be identified. An optimizer that reverses that process is comparable to an artificial intelligence program that recreates a poem in a different language. So perfectly operational decompilers are a long way off. At most, current Decompilers can be used as simply an aid for the reverse engineering process leaving lots of arduous work.
C-likecode representation. The code is half-way between assembly and C, but it is much more readable than the pure assembly is. Unfortunately the program appears to be rather unstable.
Most embedded CPUs are 8-bit CPUs. Jim Turley. The Two Percent Solution. 2002.
Normally when a subroutine is finished, it returns to executing the next address immediately following the
However, assembly-language programmers occasionally use several different techniques that adjust the return address, making disassembly more difficult:
On 8-bit CPUs, calculated jumps are often implemented by pushing a calculated “return” address to the stack, then jumping to that address using the “return” instruction.
For example, the RTS Trick uses this technique to implement jump tables (branch table).
Instead of picking up their parameters off the stack or out of some fixed global address, some subroutines provide parameters in the addresses of memory that follow the instruction that called that subroutine. Subroutines that use this technique adjust the return address to skip over all the constant parameter data, then return to an address many bytes after the “call” instruction. One of the more famous programs that used this technique is the “Sweet 16” virtual machine.
The technique may make disassembly more difficult.
A simple example of this is the
write() procedure implemented as follows:
; assume ds = cs, e.g like in boot sector code start: call write ; push message's address on top of stack db "Hello, world",0dh,0ah,00h ; return point ret ; back to DOS write proc near pop si ; get string address mov ah,0eh ; BIOS: write teletype w_loop: lodsb ; read char at [ds:si] and increment si or al,al ; is it 00h? jz short w_exit int 10h ; write the character jmp w_loop ; continue writing w_exit: jmp si write endp end start
A macro-assembler like TASM will then use a macro like this one:
_write macro message call write db message db 0 _write endm
From a human disassembler's point of view, this is a nightmare, although this is straightforward to read in the original Assembly source code, as there is no way to decide if the db should be interpreted or not from the binary form, and this may contain various jumps to real executable code area, triggering analysis of code that should never be analysed, and interfering with the analysis of the real code (e.g. disassembling the above code from 0000h or 0001h won't give the same results at all).
However a half-decent tool with possibilities to specifiy rules, and heuristic means to identify texts will have little trouble.
Most 32-bit CPUs use the ARM instruction set.
“Although Intel and AMD receive the bulk of attention in the computing world, ARM’s embedded 32-bit architecture, … has outsold all others.”
“ARM licensed 1.6 billion cores [in 2005]” 2006.
Typical ARM assembly code is a series of subroutines, with literal constants scattered between subroutines.