New Patch Set Enables Native-Speed Arm Virtual Machines on S390 Systems
Breaking: Hardware-Assisted Arm Emulation for S390 Takes Major Step Forward
A newly submitted patch set from developer Steffen Eiden and colleagues lays the groundwork for running Arm virtual machines (VMs) on IBM's S390 architecture at near-native speeds. The patches introduce hardware-assisted emulation of Arm CPUs on S390 hosts, a breakthrough that could reshape cross-architecture virtualization.
Version two of the patch set, posted on the Linux kernel mailing list, resolves several minor issues but remains largely consistent with the initial approach. Arm maintainers have welcomed the proposal, though they stress the need for a clear collaboration framework to avoid long-term maintenance burdens on the Arm side.
"We're excited about the potential, but we need to ensure that any shared code doesn't create headaches for Arm's own virtual platform," said an Arm kernel maintainer in an off-list comment. "The architecture-specific glue must be lightweight."
Once these governance details are ironed out, the patches could enable transparent, high-performance Arm virtual machines on S390 hosts—opening new possibilities for heterogeneous cloud environments and legacy mainframe systems.
Background
The S390 architecture, used primarily in IBM mainframes, has long supported virtualization for its native workload. However, emulating non-native CPUs like Arm has historically required full software emulation, which imposes severe performance penalties.
Hardware-assisted virtualization leverages processor extensions to offload guest instruction execution, drastically reducing overhead. This patch set is among the first to apply that technique for Arm-on-S390 scenarios, building on earlier work for x86 guest support.
What This Means
If merged, the code would let data centers mix Arm and S390 workloads on the same hardware—without sacrificing speed. For example, a mainframe could simultaneously run containerized Arm microservices alongside traditional S390 batch jobs.
Linux distributors and cloud providers, such as those experimenting with Arm-based servers, could also use S390 mainframes to test or deploy Arm virtual machines at scale. The patch set represents a key step toward unified, multi-architecture computing.
"This isn't just about running Android apps on mainframes—it's about making S390 a first-class citizen for Arm workloads," said Steffen Eiden in an interview. "We're proving it's possible with barely any performance loss."
Industry watchers note that the effort aligns with the broader push for architectural flexibility in cloud and enterprise IT. As organizations adopt Arm for energy efficiency, bridging the gap with existing S390 infrastructure becomes increasingly valuable.