rewrite tor in rust-lang

Rust provides many features that would benefit Tor like memory safety, etc. Writing C code that is safe is super hard (even for exeperienced Tor devs).

changed milestone to %Tor: very long term

added component::core tor/tor milestone::Tor: very long term priority::medium resolution::invalid rust severity::normal status::closed type::project labels

Not likely to happen except as an effort to refactor Tor into different modules and rewrite piece-by-piece. OTOH, there's nothing stopping somebody from doing a compatible implementation in whatever language they like. Rust enthusiasts: Prove the superiority of your language by writing a simpler, better, bug-free Tor today! ;)

Trac:
Milestone: N/A to Tor: very long term

I'm really interested in getting the ball moving for this. I'm going to look how the Mozilla devs integrated Rust into Firefox. Also, like nickm said in the previous comment, this should be done module by module. nickm gave me suggestions on IRC:

10:32 AM <nickm> frewsxcv: the workqueue.c code is a bit independent, is inherently threaded, and might be a good test of one's rust skills.
10:32 AM <nickm> frewsxcv: di_ops.c might be an even deeper test of whether-rust-can-even-do-that
10:32 AM <nickm> frewsxcv: rewriting as much as possible of routerparse.c could have concrete benefits
10:33 AM <nickm> frewsxcv: oh, in that case, try tor-resolv.c
10:33 AM <nickm> it depends on nothing
10:33 AM <nickm> Yawning: i think a wholesale rewrite is a bad choice, but that a piecemeal rewrite might be clever.
10:33 AM <nickm> frewsxcv: also, dircollate.c is mostly algorithmic
10:34 AM <nickm> Yawning: yeah but you already rewrote it in go
10:34 AM <nickm> as for things where rust is a good choice, though, our inner loops are most important, and not very isolated :/

Also it looks like dhuseby has already written a config parser https://github.com/dhuseby/torrc-rs

Trac:
Username: frewsxcv

Trac:
Cc: N/A to gk

Replacing torrc parsing in rust could be fun but we don't expect hostile input there.

buffers.c would be an interesting candidate, since it's low-level, performance-critical, and security-critical.

Nearly everything in src/tools is interesting but unlikely to be security-critical (in the sense that it doesn't need to handle tricky hostile inputs.)

Here's a relevant metabug from Mozilla's Bugzilla regarding integrating Rust compilation support into the build system for Gecko:

https://bugzilla.mozilla.org/show_bug.cgi?id=1135640

I took a look at the 'configure.ac' file with the hopes of integrating rustc into the build system. Needless to say the contents of it intimidated me a bit, but I haven't worked with many large C projects before so this sort of config file might be pretty common practice (unfortunately?). If anyone has any suggestions on where to start with respect to incorporating rustc into the build process, let me know.

Trac:
Username: frewsxcv

So in theory, the part that would need integration would be the various automake files (the various include.am files, plus Makefile.am).

Searching for "rust automake" gets a few hits, but I don't know if any are relevant.

One possibility is to better split Tor up into sublibraries, so that the C parts and the rust parts get built separately, but that would require a bit of rethinking, maybe.

#18621 (moved) is a duplicate.

Trac:
Reviewer: N/A to N/A
Sponsor: N/A to N/A
Severity: N/A to Normal

Trac:
Cc: gk to gk, erinn

Trac:
Username: jryans
Cc: gk, erinn to gk, erinn, jryans@gmail.com

Update from Sebastian Hahn:

https://lists.torproject.org/pipermail/tor-dev/2017-March/012088.html

Trac:
Username: frewsxcv

Trac:
Cc: gk, erinn, jryans@gmail.com to gk, erinn, jryans@gmail.com, Sebastian

Trac:
Username: Nekit
Cc: gk, erinn, jryans@gmail.com, Sebastian to gk, erinn, jryans@gmail.com, Sebastian, Nekit

Trac:
Username: mgeisler
Cc: gk, erinn, jryans@gmail.com, Sebastian, Nekit to gk, erinn, jryans@gmail.com, Sebastian, Nekit, mgeisler

I'm closing this ticket, since a rewrite in any language is so large as to extend across multiple releases and milestones. The granularity here is just too high for a single ticket.

That doesn't mean no progress has been made, though. Right now, Tor has exploratory support for building with some parts of it in rust. If this works out well, and we manage to get things working nicely on every platform we care about, we can start figuring out which areas are highest-priority for doing in a safe language, and start opening tickets for them.

Trac:
Resolution: N/A to invalid
Status: new to closed
Keywords: N/A deleted, rust added

closed

mentioned in issue #18621 (moved)

moved to tpo/core/tor#11331 (closed)