Modularization for Lockdep
summary
Proposes splitting lockdep into a front-end that keeps a per-task lock array even when lockdep is off, a pipeline that can cache dependencies, and a back end that can run in-kernel, in userspace, or on another host; supplies a working Rust backend prototype. Claims front-end-only bookkeeping is cheap, printing and stack-transfer are the bottlenecks, and a userspace back end risks missing hard deadlocks that happen before it reports. Seeks upstreamable front-end refactor and protocol work for serialized stack dumps.
AI-written summary · about
chapters (8)
- 0:00Introduction to Lockdep
- 4:01Problems with current Lockdep
- 10:28Modular design proposal
- 14:29User space vs kernel space debate
- 20:52Middle layer and pipeline
- 24:35Use cases and CI benefits
- 28:32Q&A and future directions
- 32:51Conclusion and thanks
related talks
48mMore Effective Approach to Detecting Potential Deadlocks, DEPT(DEPendency Track...
26mMoving to modern languages in long-living projects
01h 08mFireside chat
25mResponsive Parallelism with Synchronization
44mModular Design in Elixir
19mLeaf: Modularity for Temporary Sharing in Separation Logic
50mSmashing the Stack: Let’s Make It Less Fun And Unprofitable!
37mModularization - flatten your graph and get the real benefits
