Gleb Chesnokov
3d4f88273b
Keep CI selection policy in shared ci-pr and ci-push label sets. Resolve them to ordinary Ginkgo expressions for local discovery and execution, and before booting CI guests so report validation still checks the exact filter. Preserve ordinary expressions and the full manual selection. Mark both ACG teardown scenarios as race and exclude that category from master push runs. Keep their strict overlap, recovery and cleanup checks unchanged. PR runs retain basic/smoke coverage through ci-pr. Document the named selections and check their filtering behavior and propagation through the VM supervisor.
3.5 KiB
Tests in GitHub Actions
KUnit and Ginkgo have independent workflows. Each builds its own SCST modules and runs tests in disposable QEMU guests within a single job:
.github/workflows/kunit.yml: KUnit modules, supervisor checks for guest success, panic and timeout, thenmake kunit-run..github/workflows/ginkgo.yml: ordinary SCST modules,make ginkgo-check, then local and TCP iSCSI tests throughmake ginkgo-test. Pull requests use theci-prlabel set; pushes tomasteruseci-push. Define these sets in labels.json by combining ordinary Ginkgo labels. Manual runs select all tests, including timing-dependent race scenarios.
Both run on pushes and pull requests to master, or via workflow_dispatch.
The Debian package build in .github/workflows/ci.yml remains independent.
Ginkgo uses its locally built image directly; no image artifact is uploaded.
Docker builds a committed Git snapshot with the pinned Fedora kernel and
matching headers. QEMU loads SCST only into the guest kernel. It tries KVM
first and falls back to TCG; use --accel tcg or --accel kvm to select one.
The kernel pins are in scripts/ci/kernel.env and kernel.SHA256SUMS; update
them together. The Fedora base image is pinned in scripts/ci/Dockerfile.
Ginkgo uses separate target and initiator VMs on a private QEMU network, a disposable 1 GiB block disk and guest ext4 backing files. No host block devices or external initiators are used.
Reports and failures
The supervisor keeps serial consoles outside the guests and enforces a hard deadline. Kernel diagnostics, missing or incomplete test reports, a lost initiator and timeouts fail the job, even if QEMU exits with status zero. The KUnit workflow checks panic and timeout handling explicitly under TCG.
After each VM run, the step prints the native test output, including summaries and failures. Failed runs also print kernel diagnostics and the tail of each serial console; full logs remain in the artifacts.
Build logs, test reports and serial consoles are uploaded with if: always()
and retained for seven days. Ginkgo also saves build metadata and package
versions. VM disks and private SSH keys are not uploaded. A guest crash may
prevent its final report, but the external serial log survives. Upload still
requires a working GitHub runner.
Local reproduction
Image construction requires Docker with Buildx, fakeroot, tar and mkfs.ext4. Execution requires Python 3.11+, QEMU with the stream network backend, qemu-img, ssh-keygen and sha256sum. Use new output directories for each run. These commands boot disposable guests; self-test intentionally crashes one.
bash scripts/ci/build-image kunit /tmp/scst-kunit-image
python3 scripts/ci/vm.py self-test /tmp/scst-kunit-image /tmp/scst-vm-selftest --timeout 180 --accel tcg
python3 scripts/ci/vm.py kunit /tmp/scst-kunit-image /tmp/scst-kunit-results --timeout 1200
bash scripts/ci/build-image ginkgo /tmp/scst-ginkgo-image
python3 scripts/ci/vm.py ginkgo /tmp/scst-ginkgo-image /tmp/scst-ginkgo-results --timeout 10000
Add --ginkgo-label ci-pr to reproduce the pull-request selection.
Use --ginkgo-label ci-push for the master push selection; omit the option
for the full manual selection.
Ordinary Ginkgo expressions are also accepted. The resolved filter is printed
in stdout and recorded in result.json; both transport reports must confirm
that filter.
Supervisor checks without VM or SCST access:
python3 -m unittest discover -s scripts/ci -p 'test_*.py'