Automating Linux Kernel Documentation for Safety-Criti...
summary
Claims an automated pipeline that uses locally hosted LLMs to generate safety-oriented documentation for Linux kernel functions, applied to ftrace and evaluated on an Ubuntu 22.04 Intel i7 with an RTX 4000 GPU. Pipeline uses Olama for remote inference, tree-sitter for smart chunking and AST-based no-behavior-change checks, emits patch files, and runs Jenkins + LTP to verify builds. Mistral Devstral gives the most standards-aligned outputs; system reduces authoring time but requires human review and further traceability work.
AI-written summary · about
chapters (6)
- 0:00Introduction and background
- 1:50Motivation and agenda
- 6:05Technical architecture overview
- 21:16Results and model comparison
- 25:13Next steps and limitations
- 27:19Q&A session
related talks
49mNext-Gen Documentation with AI
31mLiterate Tracing- 49mNext-Gen Documentation with AI
41mEfficient DevSecOps workflows with a little help from AI
53mDocumenting the Design of the Linux Kernel
35mFighting logs with Log Detective
47mUsing AI To Bridge the Gap Between Safety Standards and Open Source Development
16mAutomating Technical Privacy Reviews Using LLMs