Using Yocto to Debug Embedded Device Crashes
summary
Shows how Yocto uses minidebuginfo to give symbolic user-space stacks with only ~4% image overhead (example: 70 MB → 73 MB vs ~390 MB with full debug packages), and how systemd-coredump plus debuginfod/debug-info-daemon fetches symbols by build ID. Describes kernel crash capture via kexec, creating VM core then processing with makedumpfile and crash, noting a 1.6 GB VM core reduced to 16 MB with filters. Calls out gotchas: user namespace and sandboxing (protect_home/private_tmp), service stop timeouts that truncate cores, and KASLR and config requirements for kernel symbols.
AI-written summary · about
chapters (5)
- 0:00Introduction to Yocto debugging
- 1:16Debug symbol management
- 7:49Handling user space coredumps
- 22:37Debugging kernel crashes
- 38:56Q&A and troubleshooting
related talks
42mUsing Yocto to Debug Embedded Device Crashes
46mTools and Techniques to Debug an Embedded Linux System
43mTools and Techniques to Debug an Embedded Linux System
38mEfficient On-Device Core Dump Processing for IoT: A Rusty Implementation
28mMinidump to debug end user device crashes
40mI heart the Yocto Project
31mIf (oops) { Do_not_panic(); }
39mHow to Investigate Issues on Linux Embedded System?