Ramdump for Trusted VMs: Efficient Crash Debugging Without Reboot Device
summary
A ramdump system captures only registered critical regions from a secure VM so the hypervisor can encrypt the memory and hand back small, targeted dumps without rebooting the device. Client drivers register regions in a shared mini dump table, the VMM translates GPAs to host offsets and writes ELF-backed files. Measured saves: full kdump ≈ ~200 MB vs mini dump ~1 MB, kdump reserves ~31 MB and takes ~8.5 s while mini dump writes in ~270 μs; upstreaming via mam_inspect and PKVM/virtio support is in progress.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:21Background on trusted VMs
- 1:05Challenges with secure VM dumps
- 2:09Mini dump design overview
- 3:15Registering crash regions
- 5:00Comparison with kdump
- 6:28Performance data analysis
- 8:13Future plans for mini dump
- 9:15Q&A session
related talks
19mKdump memory usage
43mEfficient and Practical Capturing of Crash Data on Embedded Systems
28mMinidump to debug end user device crashes
26mStateless Linux Kernel Crashdumps
49mKernel dump analysis made quick and easy!
27mWhen kdump is way too much
37mPracticing Linux Crash/Panic Issue on Production and Cloud Server: Using Kdump + Crash
37mHardware Error Handling Improvement for Reliable KVM Hypervisor