SFrame for Arm64 Reliable Stacktrace
summary
Proposal implements a compiler-generated SFrame section that stores FDE/FRE tables mapping PC ranges to CFA, return-address and frame-pointer offsets so the kernel can unwind across exception boundaries and satisfy livepatch consistency, avoiding objtool reverse-engineering. Current kernel patches mark exception-boundary unwinds unreliable; SFrame aims to restore reliability and can fall back to frame-pointer unwinding for modules lacking SFrame. GCC already emits SFrame today, SFrame V3 and some edge cases (relinked stack holes, register liveness) remain to resolve.
AI-written summary · about
chapters (9)
- 0:00Introduction to reliable stack traces
- 3:14Why reliable stack traces matter
- 6:06Live patch and stack trace reliability
- 10:00Limitations of x86 ORC unwinder
- 13:16ARM64 reliable stack trace progress
- 18:37Introducing SFrame solution
- 22:49Current SFrame implementation
- 27:09SFrame future improvements
- 30:11Q&A on SFrame and frame pointers
related talks
40mThreads in Embedded Linux- 6 Easy Pieces
16mKeynote: The Rise of Frameworks in Zephyr: From Devices to Subsystems to Frameworks
47mLinkage in the Third Millenium. Ideas, Innovations, Experiments and Prototypes for Fl...
27mInsights on the Re-Architecting to a Single Codebase, Running on Different Hardware...
11mBuilding a Test for Diverse Fleet of Platforms, How Hard That Could Be?
01h 03mWhat I Talk About When I Talk About Cross Platform Development
55mC++ Concurrency TS 2 Use Cases and Future Direction
13mDemo: RISC-V as an Enabler of Heterogeneous Compute