The AVH device type in LAVA refers to Arm Virtual Hardware.

AVH for third-party boards provides cloud-based functionally accurate virtual models of popular IoT development kits, simulating CPU, peripherals, sensors and board components. It allows developers to execute the same binaries as on the real hardware and so leverage the board SDKs and software code examples.

AVH integration in LAVA enables efficient CI workflow for modern agile software development.

LAVA AVH Dispatcher Setup

LAVA uses avh-api Python library to communicate with AVH’s REST API for managing virtual devices. If you are a LAVA docker worker user, the library is already pre-installed in lava dispatcher docker image. If your LAVA dispatcher is installed via APT, you will need to install the avh-api library manually using the following commands. This is mainly because the library isn’t available via APT installation yet.

apt-get install --no-install-recommends --yes python3-pip
python3 -m pip install avh-api==1.0.5

API Authentication

The following steps are required to authorize LAVA for API access to AVH.

  1. Generate AVH API token.

    Log in to AVH with your Arm account at https://app.avh.arm.com. Navigate to your profile by clicking on your name at the top right corner. Change to API tab. Then click GENERATE button to generate your AVH API Token.

  2. Add LAVA remote artifact tokens.

    Log in to LAVA server. Click on your name at the top right corner, and then click Profile. On the profile page, click the Remote artifact tokens Create button and input avh_api_token for Token name and the above AVH API token for Token string. Save the token.

  3. Define secrets block in LAVA job definition.

      avh_api_token: avh_api_token

    The secrets block should include the avh_api_token key. LAVA dispatcher needs it for AVH API authentication. The key value should be the name of the above LAVA remote artifact token name. At run time, the token name will be replaced with the token string by LAVA server. This is mainly for hiding the real token in a public LAVA job.


Concurrent uploads of AVH firmware images could be only 5 at a time. The limitation is per AVH project. After 5 images uploaded to the same project, when another one is uploaded it overrides the first. Trying to run more than 5 AVH LAVA jobs using the same AVH project may hit instance creating error as the image uploaded could be overridden by another one at any time.

Multiple projects are possible but only in enterprise/domain accounts so far. For these accounts, lab admin could extend the base AVH device type by setting the avh_project_name variable to a different AVH project name to create another device type. The variable defaults to Default Project. The number of devices for these device types should be always equal to or less than 5.

LAVA users also could overwrite the AVH project name in the device dictionary or in job definition using the deploy.options.project_name key.

Job Example

device_type: avh
job_name: avh-rpi4b-health-check

    minutes: 60

priority: medium
visibility: public

  avh_api_token: avh_api_token

- deploy:
    to: avh
      model: rpi4b
      minutes: 30
        url: https://example.com/rpi4b/nand
        format: ext4
        root_partition: 1
        url: https://example.com/rpi4b/kernel
        url: https://example.com/rpi4b/devicetree

- boot:
    method: avh
      minutes: 20
    - "pi@raspberrypi:"
    - "root@raspberrypi:"
      login_prompt: "login:"
      username: pi
      password_prompt: 'Password:'
      password: raspberry
      - sudo su

- test:
      minutes: 10
    - from: inline
          format: Lava-Test Test Definition 1.0
          name: health checks
          - lava-test-case kernel-info --shell uname -a
          - lava-test-case network-info --shell ip a
      name: health-checks
      path: inline/health-checks.yaml