Rudra is a static analyzer to detect common undefined behaviors in Rust programs. It is capable of analyzing single Rust packages as well as all the packages on crates.io.
Rudra and its associated paper received the Distinguished Artifact Award at the 28th ACM Symposium on Operating Systems Principles 2021 (SOSP '21). (PDF, short talk, long talk)
You can find the list of bugs found by Rudra at Rudra-PoC repository.
The easiest way to use Rudra is to use Docker.
- First, make sure your system has Docker and Python 3 installed.
- Add
rudra:latest
image on your system. There are two ways of doing this:docker pull ghcr.io/sslab-gatech/rudra:master && docker tag ghcr.io/sslab-gatech/rudra:master rudra:latest
- Alternatively, you can build your own image with
docker build . -t rudra:latest
- Run
./setup_rudra_runner_home.py <directory>
and setRUDRA_RUNNER_HOME
to that directory. Example:./setup_rudra_runner_home.py ~/rudra-home && export RUDRA_RUNNER_HOME=$HOME/rudra-home
.- There are two scripts,
./setup_rudra_runner_home.py
and./setup_rudra_runner_home_fixed.py
. In general,./setup_rudra_runner_home.py
should be used unless you want to reproduce the result of the paper with a fixed cargo index.
- There are two scripts,
- Add
docker-helper
in Rudra repository to$PATH
. Now you are ready to test Rudra!
For development, you might want to install Rudra on your host system. See DEV.md for advanced usage and development guide.
docker-cargo-rudra <directory>
The log and report are printed to stderr by default.
Rudra can be run as a GitHub Action allowing the static analyze to be used in an Action workflow.
# Run Rudra
- name: Rudra
uses: sslab-gatech/Rudra@master
Rudra is tied to a specific Rust compiler version,
and it can only analyze projects that compiles with this version of the compiler.
master
branch uses nightly-2021-10-21
version of Rust right now.
Check the version page for all supported versions.
- Rudra does not support workspaces (#11). You can install Rudra on your host system (see DEV.md) and run analysis in the subdirectories to sidestep the problem for now.
- Rudra does not support suppressing warnings in specific locations. This could cause a usability issue when used in CI/CD due to false positives.
Rudra currently detects the following bug types. For the full detail, please check our SOSP 2021 paper.
Detects when unsafe code may lead to memory safety issues if a user provided
closure or trait panics. For example, consider a function that dereferences a
pointer with ptr::read
, duplicating its ownership and then calls a user
provided function f
. This can lead to a double-free if the function f
panics.
See this section of the Rustonomicon for more details.
while idx < len {
let ch = unsafe { self.get_unchecked(idx..len).chars().next().unwrap() };
let ch_len = ch.len_utf8();
// Call to user provided predicate function f that can panic.
if !f(ch) {
del_bytes += ch_len;
} else if del_bytes > 0 {
unsafe {
ptr::copy(
self.vec.as_ptr().add(idx),
self.vec.as_mut_ptr().add(idx - del_bytes),
ch_len,
);
}
}
// Point idx to the next char
idx += ch_len;
}
Example: rust#78498
When code assumes certain properties about trait methods that aren't enforced,
such as expecting the Borrow
trait to return the same reference on multiple
calls to borrow
.
let mut g = Guard { len: buf.len(), buf };
// ...
Ok(n) => g.len += n,
Example: rust#80894
This occurs when a type generic over T
implements Send or Sync without having
correct bounds on T
.
unsafe impl<T: ?Sized + Send, U: ?Sized> Send for MappedMutexGuard<'_, T, U> {}
unsafe impl<T: ?Sized + Sync, U: ?Sized> Sync for MappedMutexGuard<'_, T, U> {}
Example: futures#2239
Rudra was ran on the entirety of crates.io state as of July 4th, 2020 as well
as the Rust standard library from nightly-2020-08-26
. It managed to find 264
new memory safety issues across the Rust ecosystem which resulted in 76 CVEs.
The details of these bugs can be found in the Rudra-PoC repo.
Licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)
at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.