Guillaume Tucker

KernelCI has now been a Linux Foundation project for just over a year. During that time, it has set the basis needed to fulfil its mission of being the de facto upstream kernel test system. We can now build many more kernels, run many more tests and collate results from many more test labs. We also have a growing team of core contributors, an on-going commitment from our member companies as well as more presence in the kernel community. Together, we are gathering the momentum needed to start a trend.

Now we need to make KernelCI a natural part of upstream kernel development and realise its true potential. The Community Survey in June 2020 showed there is great value and interest in having a more test-driven workflow. While this is going to be a long-term goal, we already have a process to let the community shape the KernelCI tools according to their own needs. This talk gives an overview of how it would work, essentially by allowing decisions to be based on feedback from the whole ecosystem (developers, maintainers, OEMs...). We want to hear from you!

KernelCI is a project dedicated to testing the upstream Linux kernel. Originally created by Linaro in 2014, it started a new chapter by becoming a Linux Foundation project in October 2019. Its future looks bright, with plenty of opportunities for new contributors to join.

The chosen one

The upstream kernel testing landscape is pretty wide, rich and diverse, in the same ways that the Linux kernel is. But as there is only one upstream kernel, it became clear that there should also be one main test system associated with it. KernelCI was chosen to fulfil this role, being rather neutral, versatile and based on a distributed architecture.

A welcoming place

While the project now has a governing board via the Linux Foundation membership, its involvement with the kernel community is only getting stronger. It is of utmost importance to keep the roadmap aligned with expectations from maintainers and developers to preserve the integrity and overall purpose of project. In fact, it now needs to become an easy tool to use by anyone who wants to add tests for their subsystem or their hardware and for anyone to reproduce those tests locally.

An exciting year ahead

This is a new beginning for KernelCI, with many of its prior limitations now being removed thanks to the framework provided by the Linux Foundation. Contributors to the code, tests, hardware labs and new project members will all have a great influence by joining the project at this very special point in time. Now is the time to come and help shape it as a successful project for the years to come.

Slides: gtucker-kernelci-fosdem-2020.pdf

KernelCI is a project dedicated to testing the upstream Linux kernel. Originally created by Linaro in 2014, it is now given a second breath by joining the Linux Foundation and has a fast growing community. Results are already starting to show.

On testing the upstream kernel

The upstream Linux kernel is known for having a high rate of changes while supporting a vast range of hardware. How is it being tested? The answer is manifold, with various systems focusing on very different use-cases. From Intel's 0-Day project which has a large test coverage but only runs on x86 platforms to individual hackers who only have their locally available development hardware, there is no comprehensive solution to test the kernel as a whole.

A second breath for KernelCI

KernelCI was originally created by Linaro in 2014 to address a segmented ARM ecosystem. It then went through a long period of slow progress, with a good range of hardware in its test labs but limited build infrastructure, a reduced community and a lack of advanced features. It is now given a second breath: labs are growing with new devices, functional testing is expanding to cover more subsystems, automated bisection is becoming effective. Last but not least, KernelCI is now joining the Linux Foundation with a membership scheme to provide it with a sustainable structure and funding.

In the long run

Overall, the project aims at testing the upstream kernel in its entirety. The hope is to gradually shift the kernel development model to become more test-driven with continuous automated testing. It is a very big task, so to start with it's going off the beaten tracks by looking at areas that are not already very well covered. And of course, there are many things there that need fixing before they become real problems for users.