Project post originally published in the Flux blog by Daniel Holbach

Next up in our blog series about Flux Security is how we implemented fuzzing in Flux and its controllers and how that makes things safer for you.

Wikipedia explains Fuzzing like so:

Fuzzing or fuzz testing is an automated software testing technique that involves providing invalid, unexpected, or random data as inputs to a computer program. The program is then monitored for exceptions such as crashes, failing built-in code assertions, or potential memory leaks. Typically, fuzzers are used to test programs that take structured inputs. This structure is specified, e.g., in a file format or protocol and distinguishes valid from invalid input. An effective fuzzer generates semi-valid inputs that are “valid enough” in that they are not directly rejected by the parser, but do create unexpected behaviors deeper in the program and are “invalid enough” to expose corner cases that have not been properly dealt with.

We already have quite a good coverage of unit and end-to-end tests across the controllers. Adding fuzzing to the mix will further extend the scope of tests to scenarios and payloads not previously covered. Together with the fuzzing that’s already being done within the Kubernetes repositories, e.g. kubernetesclient-go and apimachinery we feel ever more confident in our code.

We are happy to share that since the 0.27 release of Flux all Flux controllers and libraries are now tested by Google’s continuous fuzzing for open source software.

How we got here

When we announced the results of the security audit back in November, we already shared that the team at ADA Logics had helped put together an initial implementation of Fuzzing for some of the Flux controllers. In this first inception three issues were already found (1x slice out-of-bounds, 2x nil-dereference), and immediately fixed. Naturally we were very interested in merging the fuzzing integration.

In order for us to fully land the fuzzers, we needed to make some architectural changes to the build process, especially for the controllers that rely on C bindings to libgit2, such as source-controller and image-automation-controller, which are now statically built. In addition to that, we extended the scope of the fuzzers considerably. If you take a look at the related pull request for notification-controller you get a good idea of what this all entailed, e.g. fuzzing for all notifiers.

Fuzzers are now run for every commit which lands in the Flux controllers and libraries.

Thanks again ADA Logics for contributing and to everyone else who helped integrate this! We are also very grateful to Google and OpenSSF who provide and maintain the required infrastructure.

What’s next

As Go will see built-in Fuzz support in 1.18, we were very interested in structuring everything closely to the new format, so that the transition from dvyukov/go-fuzz (which is currently being used) goes smoothly. (We can recommend Jay Conrod’s blog post about the Internals of Go’s new fuzzing system, if you are curious!)

The move of Flux to go native fuzzing is being tracked in this issue. We also hope to add new fuzzers soon, so if you want to contribute there: come and find us on Slack! It’s an easy way to get to know and extend the Flux codebase.

This is just one more measure we are taking to keep you more secure.

Talk to us

We love feedback, questions and ideas, so please let us know your personal use-cases today. Ask us if you have any questions and please

See you around!