Compound Interest - Dealing with Two Decades of Technical Debt in Embedded...
summary
GPIO subsystem accumulates two decades of legacy: a dual in-kernel API, two uAPIs and 35% still use the old number-based ABI. Serialization and sleeping vs atomic context races force SRCU adoption and SRCU in 6.9 lands to enforce safe locking. Non‑exclusive GPIOs and ad‑hoc quirks proliferate; a userland persistence gap motivates a new dbus daemon + client to bridge CFS character device expectations. Work is incremental, many drivers still require conversion before old interfaces can be removed.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:05Technical debt overview
- 3:32Project types and debt origins
- 5:41Why debt accumulates
- 8:32Kernel API/ABI stability
- 11:21Case study: 20 years GPIO debt
- 14:33GPIO’s three historical epochs
- 17:45Migrating to modern GPIO
- 20:15GPIO toggling and locking issues
- 24:11Fixing GPIO with SRCU in 6.9
- 24:53Non-exclusive GPIO misuse
- 27:26Worst debt: Userspace exposed
- 29:13New GPIO userspace API
- 30:59Summary
- 32:46Q&A
related talks
42mOpportunities and Challenges on Merging HW Security Features into Linux Mainline Kernel
52mKernel Hardening: Ten Years Deep
28mModernizing Kernel Cryptography: From Complex APIs To Streamlined Libraries
44mBackporting is so 1993
38mHow I started chasing speculative type confusion bugs in the kernel and ended up wi...
42mDefending the Branch: PAC, BTI & GCS on Linux
31mSponsored Session: Can We Do Better than Shipping 4 Year Old Software in Products?
01h 06mIt's Time for Operating Systems to Rediscover Hardware