Reliable Recovery in Partitioned Embedded Systems Using IVSHMEM and...
Lightningat Open Source Summit Europe 2025 · on Tuesday, 2025-08-26· at 16:20· Emerald Room (Level 1)·
summary
Demonstrates a working pattern that uses IVSHMEM heartbeat between a Linux root cell and inmate under the Jailhouse hypervisor on TI Sitara EVMs to detect unresponsive partitions and trigger selective inmate restart without a full system reboot. Implementation uses Jailhouse static partitioning, IVSHMEM exposed as a PCI/UIO region, a crash-test app that sends interrupts and acknowledgements, and watchdogs for root recovery. Works on multi‑core boards (minimum two CPUs); root-cell failure still requires a root restart via watchdog. The approach materially reduces downtime and isolates faults.
AI-written summary · about
chapters (11)
- 0:00Mixed-critical systems intro
- 1:03Critical system design challenges
- 3:12Existing recovery mechanisms
- 3:35Jailhouse hypervisor
- 4:39IVSHMEM shared memory in Jailhouse
- 6:01Crash detection and recovery
- 7:08Jailhouse use cases and advantages
- 8:35Hypervisor recovery summary
- 9:20Q&A: CPU partitioning
- 10:25Q&A: Root cell crash handling
- 11:30Q&A: Contributing to Jailhouse
related talks
46mBuilding Mixed Criticality Linux Systems with the Jailhouse Hypervisor
40mZero-Overhead Virtualisation: It's a Trap!
01h 45mTutorial: Bootstrapping the Partitioning Hypervisor Jailhouse
29mSecure and Dynamic Hardware Partitioning Management on Heterogeneous SoC
49mGeneric System for Safe Rootfs/Kernel Upgrades without Single-Point of Failure
19mSupporting Trusted Virtual Machines with Hardware-based Secure Remote Memory
39mCPU isolation vs jailbreaking IPIs
09mSmartVisor: User-Friendly Hypervisor for Mobile Robots