Machine Code to WASM Compiler

[cretz]

• Use Capstone to disassemble
• Write out WASM
• Build in language that itself can compile to WASM (i.e. Rust) so it can be used (hopefully) w/ statically compiled Capstone in the browser
• Build proof of concept allowing user to drag exe onto webpage and watch it run
• While a fun project, the question is "why"?
  • Can't really think of a good reason
  • The jslinux project exists that is qemu in the browser which can handle apps that live update instructions which this couldn't
  • Are there a bunch of exes sans source that would benefit from this? Otherwise, using the original source language is way better
  • Maybe a nice quick tool that will turn a CLI tool into a usable JS call? Meh
  • TODO: come up with a good benefit or I ain't working on it

[cretz]

Well, now I'm wondering if there is value in ARM compilation for Android apps. I wonder how many phone apps are native yet don't use JITs tho...probably few since most use the Android JVM which uses JIT.

[cretz]

Well, one use case is for an online emulator that does a JIT to WASM on ROM load. Ideally, the device to emulate would:

• Need the speed given by WASM compilation that interpretation cannot give (so a bit more modern)
• Disallow JIT or other runtime machine code generation (some consoles do this)

[cretz]

Another use case - Go, D, and any other non-LLVM compiler

I did some research on disasm libs:

• https://github.com/trailofbits/mcsema
  • Would be nice, but unacceptable due to IDA requirement
  • There is a branch for using binary ninja but that's commercial too
  • But maybe could use https://github.com/trailofbits/remill after disassembling and getting CFG with something else
• https://github.com/repzret/dagger
  • Not completely maintained, would have to test
• https://github.com/Cr4sh/openreil
  • GPL, boo
  • Uses other things like libvex and capstone anyways
• https://github.com/programa-stic/barf-project
  • Not the biggest fan of Python
  • It's pretty clean tho
• https://github.com/angr/angr
  • Also Python
  • It does have bindings to Valgrind's libvex, nice: https://github.com/angr/pyvex
• https://sourceware.org/git/?p=valgrind.git;a=tree;f=VEX;
  • Ug, GPL
• http://www.capstone-engine.org/
  • May not do full dissasembly we might want
• https://github.com/radare/radare2
  • Decent license (LGPL)
  • Nice bindings, including https://github.com/radare/radeco-lib and https://github.com/radare/radare2-bindings/tree/master/go
  • ESIL - https://radare.gitbooks.io/radare2book/content/disassembling/esil.html
  • Rust binding - https://github.com/radare/esil-rs
• https://github.com/BinaryAnalysisPlatform/bap
  • Nice license, language, etc
  • C bindings at https://github.com/BinaryAnalysisPlatform/bap-bindings which are used by https://github.com/maurer/bap-rust
  • BIL is good - https://github.com/BinaryAnalysisPlatform/bil/releases/download/v0.1/bil.pdf
  • I like this one...

[TylerSeppala]

Hey, I know I'm finding this years later but if you actually have a solid approach to making this happen, or maybe even have a start on it, I've got a use-case. I'm building a program in wasm to run in the browser, but there are a lot of pre-existing plugins people in this particular market are dead set on using. If I want to win over a decent portion of said market, I'm going to have to support these plugins (I'm being vague because I don't want to inspire any competition in this space). Problem is, they are all pre-compiled DLLs, originally written in C++. The only way to truly get them to work with what I'm doing would be to convert the machine code directly into wasm and dynamically link it to my project. I'm sure there are others out there with projects in entirely different markets that also rely on commonly used DLLs, and that would benefit greatly from a solution for porting them to the web and linking them to other wasm modules. Let me know if you still have any intention of doing this, or maybe some pointers you could give me beyond this thread. I'm probably gonna end up trying to build something like this one way or another.

[cretz]

Nice! Afraid I don't have any intentions or insight on doing this, but I do agree it would be cool.

[TylerSeppala]

Oh man, well I figured it couldn't hurt to ask. I'll come back when I have a decent start on it, maybe some people watching this will be interested.