Detecting failed device probes
summary
Proposal: a generic regression detector that verifies devices actually instantiate and bind to drivers by parsing kernel driver match tables and checking the runtime CFS/device tree state. Practiced results: a device-tree kselftest merged upstream that extracts compatibles from driver match structs and flags missing probes; an ACPI test prototype covers ACPI-enumerated devices; discoverable buses (USB/PCI) require a per-platform discoverable-bus descriptor (lab-maintained or generated) to catch probe regressions, work-in-progress but already prevents false positives and catches config/probe-binding regressions.
AI-written summary · about
chapters (10)
- 0:00Introduction to device probe regressions
- 1:04Problem with unprobed devices
- 3:57Current testing approaches
- 6:48Device tree and ACPI tests
- 10:44Handling discoverable buses
- 16:50Describing hardware topology
- 21:04Discussion on PCI and firmware
- 26:03Testing as baseline in CI
- 30:07Maintaining device catalogs
- 35:09Q&A and closing remarks
related talks
34mLinux Kernel Testing: Providing Better Results to the Upstream Community- 34mLinux Kernel Testing: Providing Better Results to the Upstream Community
39mRegression Testing Boot-time Performance in the Linux Kernel
39mDay-to-day Testing of Linux 'Next' Kernel Branch - My Story
26mUnit Testing Memory Mapped Device Drivers
26mUnit Testing Memory Mapped Device Drivers
18mPreview: Verification (Correctness & Debugging)
31mHow to make kernel and user space CI for input devices?