Rewriting the GNU Coreutils in Rust

Posted Jun 9, 2021 18:52 UTC (Wed) by jezuch (subscriber, #52988)
In reply to: Rewriting the GNU Coreutils in Rust by helge.bahmann
Parent article: Rewriting the GNU Coreutils in Rust

Well, duh... They had several decades to stamp out the bugs. Which seems like a good thing, until you realize that they reached a plateau of "don't touch it or you'll break it!!". You may still think that the ultimate stability resulting from that is a good thing, though :)

Rewriting the GNU Coreutils in Rust

Posted Jun 10, 2021 2:00 UTC (Thu) by karkhaz (subscriber, #99844) [Link] (1 responses)

Admittedly this paper is from 2008, but:

> KLEE found ten unique bugs in COREUTILS (usually memory error crashes). Figure 7 gives the commandlines used to trigger them. The first three errors existed since at least 1992, so should theoretically crash any COREUTILS distribution up to 6.10

From https://llvm.org/pubs/2008-12-OSDI-KLEE.html.

Rewriting the GNU Coreutils in Rust

Posted Jun 10, 2021 9:22 UTC (Thu) by helge.bahmann (subscriber, #56804) [Link]

Sure, I bet there is also at the very least one more memory-safety bug in coreutils it's just... I'm just not sure if it is the most pressing problem -- e.g. the things klee found 12 years(!) ago: Okay mknod crashes when passing an invalid security context -- that's embarrassing and shouldn't happen and everything, but is it a serious issue problem? After all, if someone manages to call mknod with an invalid security context causing it to crash and misbehave, things have gone so wrong that they might also have called it with a _valid_ but wrong security context and compromised the system anyways. FWIW, nasty coreutil CVEs include things like TOCTOU races regarding symlink handling & such which are not solvable by any language means whatsoever.