Devicetree: The Disaster so Far
summary
Device tree adoption breaks when bindings land without review and then force ABI changes that make firmware‑shipped DTBs stop booting, with several emails a week reporting regressions. Review bottlenecks and vague binding docs cause vendors to commit incompatible bindings; product teams already ship devices and even a million devices per model. Concrete fixes proposed and partly implemented are schemas and validators, deprecation warnings in the device tree compiler, and a dummy Linux platform (Macvert) plus KVM guest DT generation, which work today to reduce future breakage.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:20Agenda Disaster is a strong word. Let's talk about.
- 3:17Device Tree - Overview
- 6:30Device Tree - Bindings
- 7:50Device Tree - Source
- 8:50Unfamiliarity
- 10:02Inconsistency How do we refer to interrupts? Interrupt connection The single IRQ line
- 10:33Get acquainted with device tree PAPR stil online
- 12:10We are used to board files Compiled into kernel Atomic updates
- 14:02Cleanup is breakage
- 18:26We make mistakes
- 19:49Design for extension and correction
- 22:29The conversion process
- 24:21A fresh start: mach-virt
- 31:39Better binding review
- 33:53Missed opportunity - We're not sharing FreeBSD
- 35:07ACPI is on the horizon But
- 36:41How to help
related talks
01h 10mBoF: Devicetree
34mDTS 101: From Roots To Trees, Aka Devicetree for Beginners
50mRethinking U-Boot Devicetree Story
44mBoF: Devicetree
44mSolving Devicetree Issues, part 3.0
33mStructured Documentation and Validation for Device Tree
51mShipping Multiple Devicetrees: How to Identify Which DTB Is for My Board?
50mDevicetree Hardware Autoconfiguration