Multikernel Linux 7.0-mk5 adds ARM64 support, enables multi-kernel on AArch64 servers
The out-of-tree Multikernel Linux patches now support ARM64 in addition to x86_64, allowing multiple independent kernel instances to coexist on a single physical AArch64 machine.
Summary
The Multikernel Linux project has released mklinux 7.0-mk5, extending its multi-kernel architecture to 64-bit ARM (ARM64/AArch64) systems for the first time. Previously limited to x86_64 hardware, this out-of-tree patch set allows multiple independent Linux kernel instances to run simultaneously on the same physical machine, enabling configurations such as dedicating specific CPU cores to a real-time kernel while running a general-purpose kernel on others.
This release is based on the upstream Linux 7.0 kernel and includes additional improvements to device loss handling, DMA confinement, VT-d integration, and chained takeovers. The architecture has been demonstrated with Kmorph for reboot-less kernel crash recovery, making it relevant for high-availability and safety-critical deployments.
What changes in mklinux 7.0-mk5
- ARM64 architecture support: The Multikernel patches now function on 64-bit ARM (ARM64/AArch64) hardware, expanding beyond the previous x86_64-only limitation.
- DMA confinement improvements: Updated code provides better isolation and handling of direct memory access operations between independent kernel instances.
- VT-d integration work: The release includes continued development on Intel Virtualization Technology for Directed I/O support within the multi-kernel context.
- Chained takeovers: New mechanisms allow sequential or cascading control transfers between kernel instances during runtime operations.
- Device loss handling: Refined logic for managing device availability and failure scenarios when hardware resources are partitioned across multiple kernels.
Why it matters
ARM64 support significantly broadens Multikernel Linux's addressable hardware base to include cloud server instances, embedded industrial controllers, and ARM-based edge infrastructure where mixed-criticality workloads are common. For Linux distribution maintainers and systems integrators, this provides a foundation for deploying consolidated workloads—such as pairing real-time control loops with general-purpose compute—without hardware partitioning or virtualization overhead, though the out-of-tree status means substantial integration work remains before mainstream adoption.
Original reference: www.phoronix.com →