debian-mirror-gitlab/doc/development/sec/analyzer_development_guide.md
2022-11-25 23:54:43 +05:30

10 KiB

stage group info
Secure Static Analysis To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/product/ux/technical-writing/#assignments

Sec section analyzer development

Analyzers are shipped as Docker images to execute within a CI pipeline context. This guide describes development and testing practices across analyzers.

Shared modules

There are a number of shared Go modules shared across analyzers for common behavior and interfaces:

  • The command Go package implements a CLI interface.
  • The common project provides miscellaneous shared modules for logging, certificate handling, and directory search capabilities.
  • The report Go package's Report and Finding structs marshal JSON reports.
  • The template project scaffolds new analyzers.

How to use the analyzers

Analyzers are shipped as Docker images. For example, to run the semgrep Docker image to scan the working directory:

  1. cd into the directory of the source code you want to scan.

  2. Run docker login registry.gitlab.com and provide username plus personal or project access token with at least the read_registry scope.

  3. Run the Docker image:

    docker run \
        --interactive --tty --rm \
        --volume "$PWD":/tmp/app \
        --env CI_PROJECT_DIR=/tmp/app \
        -w /tmp/app \
        registry.gitlab.com/gitlab-org/security-products/analyzers/semgrep:latest /analyzer run
    
  4. The Docker container generates a report in the mounted project directory with a report filename corresponding to the analyzer category. For example, SAST generates a file named gl-sast-report.json.

Analyzers development

To update the analyzer:

  1. Modify the Go source code.
  2. Build a new Docker image.
  3. Run the analyzer against its test project.
  4. Compare the generated report with what's expected.

Here's how to create a Docker image named analyzer:

docker build -t analyzer .

For example, to test Secret Detection run the following:

wget https://gitlab.com/gitlab-org/security-products/ci-templates/-/raw/master/scripts/compare_reports.sh
sh ./compare_reports.sh sd test/fixtures/gl-secret-detection-report.json test/expect/gl-secret-detection-report.json \
| patch -Np1 test/expect/gl-secret-detection-report.json && Git commit -m 'Update expectation' test/expect/gl-secret-detection-report.json
rm compare_reports.sh

You can also compile the binary for your own environment and run it locally but analyze and run probably won't work since the runtime dependencies of the analyzer are missing.

Here's an example based on SpotBugs:

go build -o analyzer
./analyzer search test/fixtures
./analyzer convert test/fixtures/app/spotbugsXml.Xml > ./gl-sast-report.json

How to test the analyzers

Video walkthrough of how Dependency Scanning analyzers are using downstream pipeline feature to test analyzers using test projects:

How Sec leverages the downstream pipeline feature of GitLab to test analyzers end to end

Testing local changes

To test local changes in the shared modules (such as command or report) for an analyzer you can use the go mod replace directive to load command with your local changes instead of using the version of command that has been tagged remotely. For example:

go mod edit -replace gitlab.com/gitlab-org/security-products/analyzers/command/v3=/local/path/to/command

Alternatively you can achieve the same result by manually updating the go.mod file:

module gitlab.com/gitlab-org/security-products/analyzers/awesome-analyzer/v2

replace gitlab.com/gitlab-org/security-products/analyzers/command/v3 => /path/to/command

require (
    ...
    gitlab.com/gitlab-org/security-products/analyzers/command/v3 v2.19.0
)

Testing local changes in Docker

To use Docker with replace in the go.mod file:

  1. Copy the contents of command into the directory of the analyzer. cp -r /path/to/command path/to/analyzer/command.
  2. Add a copy statement in the analyzer's Dockerfile: COPY command /command.
  3. Update the replace statement to make sure it matches the destination of the COPY statement in the step above: replace gitlab.com/gitlab-org/security-products/analyzers/command/v3 => /command

Versioning and release process

Analyzers are independent projects that follow their own versioning. Patch version bumps tend to correspond to a Minor version bump of the underlying tools (i.e. bandit), allowing us greater flexibility in reserving Minor bumps for more significant changes to our scanners. In case of breaking changes imposed by the wrapped scanner, creating a new analyzer on a separate repository must be considered.

The analyzers are released as Docker images following this scheme:

  • each push to the master branch will override the edge image tag
  • each push to any awesome-feature branch will generate a matching awesome-feature image tag
  • each Git tag will generate the corresponding Major.Minor.Patch image tag. A manual job allows to override the corresponding Major and the latest image tags to point to this Major.Minor.Patch.

To release a new analyzer Docker image, there are two different options:

  • Manual release process
  • Automatic release process

Manual release process

  1. Ensure that the CHANGELOG.md entry for the new analyzer is correct.
  2. Ensure that the release source (typically the master or main branch) has a passing pipeline.
  3. Create a new release for the analyzer project by selecting the Deployments menu on the left-hand side of the project window, then selecting the Releases sub-menu.
  4. Select New release to open the New Release page.
    1. In the Tag name drop down, enter the same version used in the CHANGELOG.md, for example v2.4.2, and select the option to create the tag (Create tag v2.4.2 here).
    2. In the Release title text box enter the same version used above, for example v2.4.2.
    3. In the Release notes text box, copy and paste the notes from the corresponding version in the CHANGELOG.md.
    4. Leave all other settings as the default values.
    5. Select Create release.

After following the above process and creating a new release, a new Git tag is created with the Tag name provided above. This triggers a new pipeline with the given tag version and a new analyzer Docker image is built.

If the analyzer uses the analyzer.yml template, then the pipeline triggered as part of the New release process above automatically tags and deploys a new version of the analyzer Docker image.

If the analyzer does not use the analyzer.yml template, you'll need to manually tag and deploy a new version of the analyzer Docker image:

  1. Select the CI/CD menu on the left-hand side of the project window, then select the Pipelines sub-menu.
  2. A new pipeline should currently be running with the same tag used previously, for example v2.4.2.
  3. After the pipeline has completed, it will be in a blocked state.
  4. Select the Manual job play button on the right hand side of the window and select tag version to tag and deploy a new version of the analyzer Docker image.

Use your best judgment to decide when to create a Git tag, which will then trigger the release job. If you can't decide, then ask for other's input.

Automatic release process

The following must be performed before the automatic release process can be used:

  1. Configure CREATE_GIT_TAG: true as a CI/CD environment variable.
  2. Check the Variables in the CI/CD project settings. Unless the project already inherits the GITLAB_TOKEN environment variable from the project group, create a project access token with complete read/write access to the API and configure GITLAB_TOKEN as a CI/CD environment variable which refers to this token.

After the above steps have been completed, the automatic release process executes as follows:

  1. A project maintainer merges an MR into the default branch.
  2. The default pipeline is triggered, and the upsert git tag job is executed.
    • If the most recent version in the CHANGELOG.md matches one of the Git tags, the job is a no-op.
    • Else, this job automatically creates a new release and Git tag using the releases API. The version and message is obtained from the most recent entry in the CHANGELOG.md file for the project.
  3. A pipeline is automatically triggered for the new Git tag. This pipeline releases the latest, major, minor and patch Docker images of the analyzer.

Steps to perform after releasing an analyzer

  1. After a new version of the analyzer Docker image has been tagged and deployed, please test it with the corresponding test project.

  2. Announce the release on the relevant group Slack channel. Example message:

    FYI I've just released ANALYZER_NAME ANALYZER_VERSION. LINK_TO_RELEASE

Never delete a Git tag that has been pushed as there is a good chance that the tag will be used and/or cached by the Go package registry.