What Is ACRN¶
IoT and Edge system developers face mounting demands on the systems they build, as connected devices are increasingly expected to support a range of hardware resources, operating systems, and software tools and applications. Virtualization is key to meeting these broad needs. Most existing hypervisor and Virtual Machine Manager solutions don’t offer the right size, boot speed, real-time support, and flexibility for IoT and Edge systems. Data center hypervisor code is too big, doesn’t offer safety or hard real-time capabilities, and requires too much performance overhead for embedded development. The ACRN hypervisor was built to fill this need.
ACRN is a type 1 reference hypervisor stack that runs on bare-metal hardware, with fast booting, and is configurable for a variety of IoT, Edge, and embedded device solutions. It provides a flexible, lightweight hypervisor, built with real-time and safety-criticality in mind, optimized to streamline embedded development through an open-source, scalable reference platform. It has an architecture that can run multiple OSs and VMs, managed securely, on a consolidated system by means of efficient virtualization. Resource partitioning ensures co-existing heterogeneous workloads on one system hardware platform do not interfere with each other.
ACRN defines a reference framework implementation for virtual device emulation, called the ACRN Device Model or DM, with rich I/O mediators. It also supports non-emulated device passthrough access to satisfy time-sensitive requirements and low-latency access needs of real-time applications. To keep the hypervisor code base as small and efficient as possible, the bulk of the Device Model implementation resides in the Service VM to provide sharing and other capabilities.
ACRN is built to virtualize embedded IoT and Edge development functions (for a camera, audio, graphics, storage, networking, and more), so it’s ideal for a broad range of IoT and Edge uses, including industrial, automotive, and retail applications.
The ACRN hypervisor and ACRN Device Model software are provided under the permissive BSD-3-Clause license, which allows “redistribution and use in source and binary forms, with or without modification” together with the intact copyright notice and disclaimers noted in the license.
ACRN has these key capabilities and benefits:
Small Footprint: The hypervisor is optimized for resource-constrained devices with significantly fewer lines of code (about 40K) than datacenter-centric hypervisors (over 150K).
Built with Real-time in Mind: Low-latency, fast boot times, and responsive hardware device communication supporting near bare-metal performance. Both soft and hard real-time VM needs are supported including no VMExit during runtime operations, LAPIC and PCI passthrough, static CPU assignment, and more.
Built for Embedded IoT and Edge Virtualization: ACRN supports virtualization beyond the basics and includes CPU, I/O, and networking virtualization of embedded IoT and Edge device functions and a rich set of I/O mediators to share devices across multiple VMs. The Service VM communicates directly with the system hardware and devices ensuring low latency access. The hypervisor is booted directly by the bootloader for fast and secure booting.
Built with Safety-Critical Virtualization in Mind: Safety-critical workloads can be isolated from the rest of the VMs and have priority to meet their design needs. Partitioning of resources supports safety-critical and non-safety-critical domains coexisting on one SoC using Intel VT-backed isolation.
Adaptable and Flexible: ACRN has multi-OS support with efficient virtualization for VM OSs including Linux, Zephyr, and Windows, as needed for a variety of application use cases. ACRN scenario configurations support shared, partitioned, and hybrid VM models to support a variety of application use cases.
Truly Open Source: With its permissive BSD licensing and reference implementation, ACRN offers scalable support with a significant up-front R&D cost saving, code transparency, and collaborative software development with industry leaders.
The ACRN architecture has evolved since its initial v0.1 release in July 2018. Beginning with the v1.1 release, the ACRN architecture has flexibility to support VMs with shared HW resources, partitioned HW resources, and a hybrid VM model that simultaneously supported shared and partitioned resources. It enables a workload consolidation solution taking multiple separate systems and running them on a single compute platform to run heterogeneous workloads, with hard and soft real-time support.
Workload management and orchestration are also enabled with ACRN, allowing open-source orchestrators such as OpenStack to manage ACRN VMs. ACRN supports secure container runtimes such as Kata Containers orchestrated via Docker or Kubernetes.
ACRN is a Type 1 hypervisor, meaning it runs directly on bare-metal hardware. It implements a hybrid Virtual Machine Manager (VMM) architecture, using a privileged Service VM that manages the I/O devices and provides I/O mediation. Multiple User VMs are supported with each of them potentially running different OSs. By running systems in separate VMs, you can isolate VMs and their applications, reducing potential attack surfaces and minimizing interference, but potentially introducing additional latency for applications.
ACRN relies on Intel Virtualization Technology (Intel VT) and runs in Virtual Machine Extension (VMX) root operation, host mode, or VMM mode. All the User VMs and the Service VM run in VMX non-root operation, or guest mode.
The Service VM runs with the system’s highest virtual machine priority to ensure required device time-sensitive requirements and system quality of service (QoS). Service VM tasks run with mixed priority. Upon a callback servicing a particular User VM request, the corresponding software (or mediator) in the Service VM inherits the User VM priority.
As mentioned earlier, hardware resources used by VMs can be configured into two parts, as shown in this hybrid VM sample configuration:
Shown on the left of Figure 1, we’ve partitioned resources dedicated to a User VM launched by the hypervisor and before the Service VM is started. This pre-launched VM runs independently of other virtual machines and owns dedicated hardware resources, such as a CPU core, memory, and I/O devices. Other VMs may not even be aware of the pre-launched VM’s existence. Because of this, it can be used as a Safety VM that runs hardware failure detection code and can take emergency actions when system critical failures occur. Failures in other VMs or rebooting the Service VM will not directly impact execution of this pre-launched Safety VM.
Shown on the right of Figure 1, the remaining hardware resources are shared among the Service VM and User VMs. The Service VM is launched by the hypervisor after any pre-launched VMs are launched. The Service VM can access remaining hardware resources directly by running native drivers and provides device sharing services to the User VMs, through the Device Model. These post-launched User VMs can run one of many OSs including Ubuntu or Windows, or a real-time OS such as Zephyr, VxWorks, or Xenomai. Because of its real-time capability, a real-time VM (RTVM) can be used for software programmable logic controller (PLC), inter-process communication (IPC), or Robotics applications. These shared User VMs could be impacted by a failure in the Service VM since they may rely on its mediation services for device access.
The Service VM owns most of the devices including the platform devices, and provides I/O mediation. The notable exceptions are the devices assigned to the pre-launched User VM. Some PCIe devices may be passed through to the post-launched User VMs via the VM configuration.
The ACRN hypervisor also runs the ACRN VM manager to collect running information of the User VMs, and controls the User VMs such as starting, stopping, and pausing a VM, and pausing or resuming a virtual CPU.
See the ACRN High-Level Design Overview developer reference material for more in-depth information.
ACRN Device Model Architecture¶
Because devices may need to be shared between VMs, device emulation is used to give VM applications (and their OSs) access to these shared devices. Traditionally there are three architectural approaches to device emulation:
Device emulation within the hypervisor: a common method implemented within the VMware workstation product (an operating system-based hypervisor). In this method, the hypervisor includes emulations of common devices that the various guest operating systems can share, including virtual disks, virtual network adapters, and other necessary platform elements.
User space device emulation: rather than the device emulation embedded within the hypervisor, it is implemented in a separate user space application. QEMU, for example, provides this kind of device emulation also used by other hypervisors. This model is advantageous, because the device emulation is independent of the hypervisor and can therefore be shared for other hypervisors. It also permits arbitrary device emulation without having to burden the hypervisor (which operates in a privileged state) with this functionality.
Paravirtualized (PV) drivers: a hypervisor-based device emulation model introduced by the XEN Project. In this model, the hypervisor includes the physical device drivers, and each guest operating system includes a hypervisor-aware driver that works in concert with the hypervisor drivers.
There’s a price to pay for sharing devices. Whether device emulation is performed in the hypervisor, or in user space within an independent VM, overhead exists. This overhead is worthwhile as long as the devices need to be shared by multiple guest operating systems. If sharing is not necessary, then there are more efficient methods for accessing devices, for example, “passthrough.”
All emulation, para-virtualization, and passthrough are used in ACRN project. ACRN defines a device emulation model where the Service VM owns all devices not previously partitioned to pre-launched User VMs, and emulates these devices for the User VM via the ACRN Device Model. The ACRN Device Model is thereby a placeholder of the User VM. It allocates memory for the User VM OS, configures and initializes the devices used by the User VM, loads the virtual firmware, initializes the virtual CPU state, and invokes the ACRN hypervisor service to execute the guest instructions. ACRN Device Model is an application running in the Service VM that emulates devices based on command line configuration.
See the Device Model High-Level Design developer reference for more information.
At the highest level, device passthrough is about providing isolation of a device to a given guest operating system so that the device can be used exclusively by that User VM.
Near-native performance can be achieved by using device passthrough. This is ideal for networking applications (or those with high disk I/O needs) that have not adopted virtualization because of contention and performance degradation through the hypervisor (using a driver in the hypervisor or through the hypervisor to a user space emulation). Assigning devices to specific User VMs is also useful when those devices inherently wouldn’t be shared. For example, if a system includes multiple video adapters, those adapters could be passed through to unique User VM domains.
Finally, there may be specialized PCI devices that only one User VM uses,
so they should be passed through to the User VM. Individual USB ports could be
isolated to a given domain too, or a serial port (which is itself not shareable)
could be isolated to a particular User VM. In the ACRN hypervisor, we support USB
controller passthrough only, and we don’t support passthrough for a legacy
serial port (for example,
Hardware Support for Device Passthrough¶
Intel’s processor architectures provide support for device passthrough with Virtual Technology for Directed I/O (VT-d). VT-d maps User VM physical addresses to machine physical addresses, so devices can use User VM physical addresses directly. When this mapping occurs, the hardware takes care of access (and protection), and the User VM OS can use the device as if it were a non-virtualized system. In addition to mapping User VM to physical memory, isolation prevents this device from accessing memory belonging to other VMs or the hypervisor.
Another innovation that helps interrupts scale to large numbers of VMs is called Message Signaled Interrupts (MSI). Rather than relying on physical interrupt pins to be associated with a User VM, MSI transforms interrupts into messages that are more easily virtualized, scaling to thousands of individual interrupts. MSI has been available since PCI version 2.2 and is also available in PCI Express (PCIe). MSI is ideal for I/O virtualization, as it allows isolation of interrupt sources (as opposed to physical pins that must be multiplexed or routed through software).
Hypervisor Support for Device Passthrough¶
By using the latest virtualization-enhanced processor architectures, hypervisors and virtualization solutions can support device passthrough (using VT-d), including Xen, KVM, and ACRN hypervisor. In most cases, the User VM OS must be compiled to support passthrough by using kernel build-time options.
Static Configuration Based on Scenarios¶
Scenarios are a way to describe the system configuration settings of the ACRN hypervisor, VMs, and resources they have access to that meet your specific application’s needs such as compute, memory, storage, graphics, networking, and other devices. Scenario configurations are stored in an XML format file and edited using the ACRN configurator.
Following a general embedded-system programming model, the ACRN hypervisor is designed to be statically customized at build time per hardware and scenario, rather than providing one binary for all scenarios. Dynamic configuration parsing is not used in the ACRN hypervisor for these reasons:
Reduce complexity. ACRN is a lightweight reference hypervisor, built for embedded IoT and Edge. As new platforms for embedded systems are rapidly introduced, support for one binary could require more and more complexity in the hypervisor, which is something we strive to avoid.
Maintain small footprint. Implementing dynamic parsing introduces hundreds or thousands of lines of code. Avoiding dynamic parsing helps keep the hypervisor’s Lines of Code (LOC) in a desirable range (less than 40K).
Improve boot time. Dynamic parsing at runtime increases the boot time. Using a static build-time configuration and not dynamic parsing helps improve the boot time of the hypervisor.
The scenario XML file together with a target board XML file are used to build the ACRN hypervisor image tailored to your hardware and application needs. The ACRN project provides a board inspector tool to automatically create the board XML file by inspecting the target hardware. ACRN also provides a configurator tool to create and edit a tailored scenario XML file based on predefined sample scenario configurations.
Predefined Sample Scenarios¶
Project ACRN provides some predefined sample scenarios to illustrate how you can define your own configuration scenarios.
Shared (called Industry in previous releases) is a traditional computing, memory, and device resource sharing model among VMs. The ACRN hypervisor launches the Service VM. The Service VM then launches any post-launched User VMs and provides device and resource sharing mediation through the Device Model. The Service VM runs the native device drivers to access the hardware and provides I/O mediation to the User VMs.
Virtualization is especially important in industrial environments because of device and application longevity. Virtualization enables factories to modernize their control system hardware by using VMs to run older control systems and operating systems far beyond their intended retirement dates.
The ACRN hypervisor needs to run different workloads with little-to-no interference, increase security functions that safeguard the system, run hard real-time sensitive workloads together with general computing workloads, and conduct data analytics for timely actions and predictive maintenance.
In this example, one post-launched User VM provides Human Machine Interface (HMI) capability, another provides Artificial Intelligence (AI) capability, some compute function is run the Kata Container, and the RTVM runs the soft Programmable Logic Controller (PLC) that requires hard real-time characteristics.
The Service VM, provides device sharing functionalities, such as disk and network mediation, to other virtual machines. It can also run an orchestration agent allowing User VM orchestration with tools such as Kubernetes.
The HMI Application OS can be Windows* or Linux*. Windows is dominant in Industrial HMI environments.
ACRN can support a soft real-time OS such as preempt-rt Linux for soft-PLC control, or a hard real-time OS that offers less jitter.
Partitioned is a VM resource partitioning model when a User VM requires independence and isolation from other VMs. A partitioned VM’s resources are statically configured and are not shared with other VMs. Partitioned User VMs can be Real-Time VMs, Safety VMs, or standard VMs and are launched at boot time by the hypervisor. There is no need for the Service VM or Device Model since all partitioned VMs run native device drivers and directly access their configured resources.
This scenario is a simplified configuration showing VM partitioning: both User VMs are independent and isolated, they do not share resources, and both are automatically launched at boot time by the hypervisor. The User VMs can be Real-Time VMs (RTVMs), Safety VMs, or standard User VMs.
Hybrid scenario simultaneously supports both sharing and partitioning on the consolidated system. The pre-launched (partitioned) User VMs, with their statically configured and unshared resources, are started by the hypervisor. The hypervisor then launches the Service VM. The post-launched (shared) User VMs are started by the Device Model in the Service VM and share the remaining resources.
In this Hybrid real-time (RT) scenario, a pre-launched RTVM is started by the hypervisor. The Service VM runs a post-launched User VM that runs non-safety or non-real-time tasks.
You can find the predefined scenario XML files in the misc/config_tools/data folder in the hypervisor source code. The Introduction to ACRN Configuration tutorial explains how to use the ACRN configurator to create your own scenario, or to view and modify an existing one.
The ACRN hypervisor can be booted from a third-party bootloader directly. A popular bootloader is grub and is also widely used by Linux distributions.
Using GRUB to Boot ACRN has an introduction on how to boot ACRN hypervisor with GRUB.
In Figure 6, we show the boot sequence:
The Boot process proceeds as follows:
UEFI boots GRUB.
GRUB boots the ACRN hypervisor and loads the VM kernels as Multi-boot modules.
The ACRN hypervisor verifies and boots kernels of the Pre-launched VM and Service VM.
In the Service VM launch path, the Service VM kernel verifies and loads the ACRN Device Model and Virtual bootloader through
The virtual bootloader starts the User-side verified boot process.
In this boot mode, the boot options of a pre-launched VM and the Service VM are defined
in the variable of
bootargs of struct
in the source code
resides under the hypervisor build directory) by default.
These boot options can be overridden by the GRUB menu. See Using GRUB to Boot ACRN for
details. The boot options of a post-launched VM are not covered by hypervisor
source code or a GRUB menu; they are defined in the User VM’s OS image file or specified by
The ACRN documentation offers more details of topics found in this introduction about the ACRN hypervisor architecture, Device Model, Service VM, and more.
These documents provide introductory information about development with ACRN:
These documents provide more details and in-depth discussions of the ACRN hypervisor architecture and high-level design, and a collection of advanced guides and tutorials: