rust-vmm-ci

The rust-vmm-ci repository contains integration tests and Buildkite pipeline definitions that are used for running the CI for all rust-vmm crates.

CI tests are executed on the container image maintained at rust-vmm/rust-vmm-container repo with builds available on Docker Hub.

Having a centralized place for the tests is one of the enablers for keeping the same quality standard for all crates in rust-vmm.

Getting Started with rust-vmm-ci

To run the integration tests defined in the pipeline as part of the CI:

1. Add rust-vmm-ci as a git submodule to your repository

# Add rust-vmm-ci as a submodule. This will point to the latest rust-vmm-ci
# commit from the main branch. The following command will also add a
# `.gitmodules` file and the `rust-vmm-ci` to the index.
git submodule add https://github.com/rust-vmm/rust-vmm-ci.git
# Commit the changes to your repository so that the CI can run using the
# rust-vmm-ci pipeline and tests.
git commit -s -m "Added rust-vmm-ci as submodule"

2. Create the coverage test configuration file named coverage_config_ARCH.json in the root of the repository, where ARCH is the architecture of the machine. There are two coverage test configuration files, one per each platform. The example of the configuration file for the x86_64 architecture can be found in coverage_config_x86_64.json.sample, and the example of the configuration file for the aarch64 architecture can be found in coverage_config_aarch64.json.sample.

The json must have the following fields:

• coverage_score: The coverage of the repository.
• exclude_path: This field is used for excluding files from the report. File paths that match the given regular expression are skipped. (for example, if multiple files are to be skipped, they must be separated with |). It should be used to exclude autogenerated files. If the repository does not have any autogenerated files, exclude_path should be an empty string.
• crate_features: cargo kcov does not build crate features by default. To get the coverage report including optional features, these need to be specified in crate_features separated by comma. If the crate does not have any features, this field should be empty.

This file is required for the coverage integration so it needs to be added to the repository as well.

3. Copy one of the two provided dependabot configurations to .github/dependabot.yml, e.g. run cp rust-vmm-ci/dependabot-{weekly,monthly}.yml .github/dependabot.yml. Note that just symlinking the file does not work, as dependabot will not follow symlinks into submodules. This means that updates to these files made in rust-vmm-ci will need to be manually consumed for now. We recommend setting up weekly dependabot updates only if the crate receives multiple contributions a week, and if you expect to have the bandwidth to address weekly dependency PRs.

4. Create a new pipeline definition in Buildkite. For this step ask one of the rust-vmm Buildkite admins to create one for you. The process is explained here.

5. There is a script that autogenerates a dynamic Buildkite pipeline. Each step in the pipeline has a default timeout of 5 minutes. To run the CI using this dynamic pipeline, you need to add a step that is uploading the rust-vmm-ci pipeline:

./rust-vmm-ci/.buildkite/autogenerate_pipeline.py | buildkite-agent pipeline upload

This allows overriding some values and extending others through environment variables.

• X86_LINUX_AGENT_TAGS: overrides the tags by which the x86_64 linux agent is selected; the default values are

  {"os": "linux", "platform": "x86.metal"}

• AARCH64_LINUX_AGENT_TAGS: overrides the tags by which the aarch64 linux agent is selected. The default values are

  {"os": "linux", "platform": "arm.metal"}

• DOCKER_PLUGIN_CONFIG: specifies additional configuration for the docker plugin. For available configuration, please check the https://github.com/buildkite-plugins/docker-buildkite-plugin.

• TESTS_TO_SKIP: specifies a list of tests to be skipped.

• TIMEOUTS_MIN: overrides the timeout value for specific tests.

• DEFAULT_AGENT_TAG_HYPERVISOR: sets the hypervisor on which all the tests in the pipeline run. By default, the selected hypervisor is KVM because the hosts running KVM at the time of this change showed better performance and experienced timeouts less often. NOTE: This will not override the hypervisor defined at the test step level. If a test already defines a hypervisor tag that will remain intact.

The variable TESTS_TO_SKIP is specified as a JSON list with the names of the tests to be skipped. The variable TIMEOUTS_MIN is a dictionary where each key is the name of a test and each value is the number of minutes for the timeout. The other variables are specified as dictionaries, where the first key is tests and its value is a list of test names where the configuration should be applied; the second key is cfg and its value is a dictionary with the actual configuration.

For example, we can skip the test commit-format, have a timeout of 30 minutes for the test style and extend the docker plugin specification as follows:

TESTS_TO_SKIP='["commit-format"]' TIMEOUTS_MIN='{"style": 30}' DOCKER_PLUGIN_CONFIG='{
    "tests": ["coverage"],
    "cfg": {
        "devices": [ "/dev/vhost-vdpa-0" ],
        "privileged": true
    }
}' ./rust-vmm-ci/.buildkite/autogenerate_pipeline.py | buildkite-agent pipeline upload

For most use cases, overriding or extending the configuration is not necessary. We may want to do so if, for example, the platform needs a custom device that is not available on the existing test instances or if we need a specialized hypervisor.

6. Tests will be running on x86_64 and aarch64 platforms by default. To change this, e.g. to enable other experimental platforms like riscv64, a .platform file can be included in the repository root. This file documents what platforms are to be enabled for the repository.

If .platform file is provided, it will be strictly observed. In .platform file, each platform are separated by newline character. Currently, we support x86_64, aarch64 and riscv64 platforms.

For example, we can enable tests to be run on riscv64 platform in addition to x86_64 and aarch64 by:

x86_64
aarch64
riscv64

7. The code owners of the repository will have to setup a WebHook for triggering the CI on pull request and push events.

Buildkite Pipeline

The Buildkite pipeline is the definition of tests to be run as part of the CI. It includes steps for running unit tests and linters (including coding style checks), and computing the coverage.

Currently the tests can run on Linux x86_64 and aarch64 hosts.

Example of step that checks the build:

steps:
- label: build-gnu-x86_64
  command: cargo build --release
  retry:
    automatic: false
  agents:
    os: linux
    platform: x86_64.metal
  plugins:
  - docker#v3.8.0:
      image: rustvmm/dev:v16
      always-pull: true
  timeout_in_minutes: 5

To see all steps in the pipeline check the output of the .buildkite/autogenerate_pipeline.py script.

Custom Pipeline

Some crates might need to test functionality that is specific to that particular component and thus cannot be added to the common pipeline.

In this situation, the repositories need to create a JSON file with a custom test configuration. The preferred path is .buildkite/custom-tests.json.

For example to test the build with one non-default feature enabled, the following configuration can be added:

{
  "tests": [
    {
      "test_name": "build-bzimage",
      "command": "cargo build --release --features bzimage",
      "platform": [
        "x86_64"
      ]
    }
  ]
}

To run this custom pipeline, you need to add a step that is uploading it in Buildkite. The same script that autogenerates the main pipeline can be used with the option -t PATH_TO_CUSTOM_CONFIGURATION:

./rust-vmm-ci/.buildkite/autogenerate_pipeline.py -t .buildkite/custom-tests.json | buildkite-agent pipeline upload

Integration Tests

In addition to the one-liner tests defined in the Buildkite Pipeline, the rust-vmm-ci also has more complex tests defined in integration_tests.

Test Profiles

The integration tests support two test profiles:

• devel: this is the recommended profile for running the integration tests on a local development machine.
• ci (default option): this is the profile used when running the integration tests as part of the Continuous Integration (CI).

The test profiles are applicable to pytest, the integration test framework used with rust-vmm-ci. Currently only the coverage test follows this model as all the other integration tests are run using the Buildkite pipeline.

The difference between is declaring tests as passed or failed:

• with the devel profile the coverage test passes if the current coverage is equal or higher than the upstream coverage value. In case the current coverage is higher, the coverage file is updated to the new coverage value.
• with the ci profile the coverage test passes only if the current coverage is equal to the upstream coverage value.

Further details about the coverage test can be found in the Adaptive Coverage section.

Adaptive Coverage

The line coverage is saved in tests/coverage. To update the coverage before submitting a PR, run the coverage test:

CRATE="kvm-ioctls"
# NOTE: This might not be the latest container version, you can check which one we're using
# by looking into the .buildkite/autogenerate_pipeline.py file.
LATEST=16
docker run --device=/dev/kvm \
           -it \
           --security-opt seccomp=unconfined \
           --volume $(pwd)/${CRATE}:/${CRATE} \
           rustvmm/dev:v${LATEST}
cd ${crate}
pytest --profile=devel rust-vmm-ci/integration_tests/test_coverage.py

If the PR coverage is higher than the upstream coverage, the coverage file needs to be manually added to the commit before submitting the PR:

git add tests/coverage

Failing to do so will generate a fail on the CI pipeline when publishing the PR.

NOTE: The coverage file is only updated in the devel test profile. In the ci profile the coverage test will fail if the current coverage is higher than the coverage reported in tests/coverage.

Performance tests

rust-vmm-ci includes an integration test that can run a battery of benchmarks at every pull request, comparing the results with the tip of the upstream main branch. The test is not included in the default Buildkite pipeline. Each crate that requires the test to be run as part of the CI must add a custom pipeline.

An example of a pipeline that runs the test for ARM platforms and prints the results:

steps:
- label: bench-aarch64
  command: pytest rust-vmm-ci/integration_tests/test_benchmark.py -s
  retry:
    automatic: false
  agents:
    os: linux
    platform: arm.metal
  plugins:
  - docker#v3.8.0:
      image: rustvmm/dev:v16
      always-pull: true

The test requires criterion benchmarks to be exported by the crate. The test expects the entry point into the performance benchmarks to be named main. In other words, the following configuration is expected in Cargo.toml:

[[bench]]
name = "main"

All benchmarks need to be collected in a main.rs file placed in benches/.

criterion collects performance results by running a function for a user-configured number of iterations, timing the runs, and applying statistics. The individual benchmark tests must be added in the crate. They can be run outside the CI with:

cargo bench [--all-features] OR [--features <features>]

rust-vmm-ci uses critcmp to compare the results yielded by cargo bench --all-features on the PR being tested with those from the tip of the upstream main branch. The test runs cargo bench twice, once on the current HEAD, then again after git checkout origin/main. critcmp takes care of the comparison, making use of criterion's stable format for output files. The results are printed to stdout and can be visually inspected in the pipeline output. In its present form, the test cannot fail.

To run the test locally:

docker run --device=/dev/kvm \
           -it \
           --security-opt seccomp=unconfined \
           --volume $(pwd)/${CRATE}:/${CRATE} \
           rustvmm/dev:v${LATEST}
cd ${CRATE}
pytest rust-vmm-ci/integration_tests/test_benchmark.py -s

Note that performance is highly dependent on the underlying platform that the tests are running on. The raw numbers obtained are likely to differ from their counterparts on a CI instance.

Running the tests locally

To run the integration tests locally, you can run the following from the crate you need to test. You can find the latest container version in the script that autogenerates the pipeline. For example:

cd ~/vm-superio
CRATE="vm-superio"
# NOTE: This might not be the latest container version, you can check which one we're using
# by looking into the .buildkite/autogenerate_pipeline.py file.
LATEST=16
docker run -it \
           --security-opt seccomp=unconfined \
           --volume $(pwd):/${CRATE} \
           --volume ~/.ssh:/root/.ssh \
           rustvmm/dev:v${LATEST}
cd vm-superio
./rust-vmm-ci/test_run.py

Known issues:

• When running the cargo-audit test, the following error may occur:

test_cargo-audit (__main__.TestsContainer) ... error: couldn’t fetch advisory database: git operation failed: reference ‘refs/heads/main’ not found; class=Reference (4); code=NotFound (-3)

A fix for this is to remove ~/.cargo/advisory-db in the container, and then rerun test_run.py:

rm -rf ~/.cargo/advisory-db
./rust-vmm-ci/test_run.py