Kernel CI – How Far Can It Go?
summary
KernelCI's KCIDB aggregates JSON results from many systems (0day, Syzbot, LKFT, CKI) into a Grafana prototype dashboard and subscription/notification service, ingesting about 300,000 test reports/day, ~20,000 builds and ~100 revisions/day. Key limits are hardware-driven coverage, high latency and low reliability from flaky infra and test drift. Concrete proposals: enable pre-merge testing via Patchwork or GitHub/GitLab integration, prioritize a small set of canonical tests in-tree, automate result grouping (like Syzbot) and shrink human review to shorten feedback loops.
AI-written summary · about
chapters (7)
- 0:00Kernel CI landscape
- 3:33Continuous integration in kernel
- 6:46Hardware limits and CI reliability
- 12:11Improving Kernel CI realistically
- 17:51Test reliability and sync solutions
- 20:00Impact on developer trust
- 23:09Q&A
related talks
37mKernel CI – How Far Can It Go?
23mUnifying Kernel Test Reporting with KernelCI
40mKernelCI for Embedded: Enhancing Reliability through Scalable Test Automation
27mPlugging into the Red Hat kernel CI ecosystem
40mOne year into the new KernelCI
38mQuality Beyond Kernelci.Org in Upstream Linux for TI SoCs
40mGetting Started With New KernelCI CLI Tools for Automating Linux Kernel Testing...
20m"GKernelCI v.2 new features and KernelCI compatibility"