tests: add VM backend dimension + small test fixes - #6149
Open
JamesC1305 wants to merge 9 commits into
Open
Conversation
Move the body of Microvm.basic_config into a standalone kvm_basic_config helper, alongside the has_kvm probe. Microvm still carries its inline copy until the vm_backend dimension lands and rewires it to delegate. This prepares for the vm_backend dimension, so its commit contains only new logic rather than relocated KVM code. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
Introduce GuestKernel as the logical guest kernel artifact. It parses the canonical vmlinux-<version>[-no-acpi] artifact name into version and ACPI metadata and locates the matching EFI-loadable boot image (bzImage on x86_64; on aarch64 the Image boots both ways). The guest_kernel fixture and pytest properties derive kernel identity from the logical artifact instead of re-parsing file names; the fixture still yields the vmlinux path and the reported identity is unchanged. This prepares the vm_backend dimension, whose backends resolve a GuestKernel to the concrete image they can boot. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
HugePagesConfig describes the supported huge-page modes and is shared by the test framework and its tests. Move it from framework.microvm to a dedicated utils_hugepages module, following the utils_<topic> convention, and update its import sites. Pure move: no functional change. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
The kernel catalogues already carry logical GuestKernel artifacts, but
the fixture still yields the bare vmlinux path, so consumers keep
re-deriving kernel identity from the file name. Yield the GuestKernel
itself; consumers read its parsed fields, and the few that still want
the path read its vmlinux field.
MicroVMFactory.build takes the GuestKernel directly, stores it as
vm.guest_kernel, and boots its vmlinux, so the parsed identity reaches
the VM rather than being re-derived from the file name later. Accepting
a path and coercing it inside build would put that parsing back; instead
the two path callers (tools/sandbox.py, tools/test-popular-containers)
convert at their boundary, where a bad name fails against the file the
user actually named. pin_guest_kernel now takes a catalogue or
pytest.param; every callsite already passed one.
Consumers that treated the fixture value as a path now read what they
mean: the GDB target names the vmlinux, snapshot artifact directories
key on the stable pytest_id, the MSR baseline lookup reads the parsed
version, and the fcmetrics guest_kernel dimension reads the object's
metric_id.
Microvm.kernel_file becomes a property derived from guest_kernel rather
than a second copy of the same answer stored by build(). Snapshot meta
records that concrete image, so restore recovers the logical kernel it
came from: guest_kernel is then the single source of truth for both
built and restored VMs, and the path-derived metric fallbacks go away.
A restored VM now emits the same guest_kernel dimension as a built one
("linux-6.1", previously the patch-level "linux-6.1.168"). A snapshot
naming an artifact the catalogue cannot parse now fails with that value
named rather than degrading the dimension silently; only the out-of-tree
nonci artifact flows can produce one.
Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
The network throughput test recomputed the first free CPU as `vcpus_count + 2`, duplicating pin_threads' internal thread layout. Move pin_threads() out of the network_microvm fixture into the tests so test_network_tcp_throughput can consume pin_threads' documented return value instead, and assert against it in the iperf helper. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
Add a VmBackend field to the Microvm class. Microvm delegates basic_config to the backend-specific helper introduced by the previous commits and asks the backend to perform any required preparation before starting. Microvm owns the shared InstanceStart, state validation, and SSH readiness sequence. Route CPU-template application through backend-owned basic_config so backend-specific machine configuration stays behind that interface. Add the vm_backend fixture as a dimension of the uvm fixtures. Tests using it multiply over the backends the host can run. Backend discovery may return an empty set so backend-independent pytest targets can collect without KVM. After test selection, fail cleanly if any selected test requires a VM backend and none is available. MicroVMFactory now takes its backend explicitly; custom-factory callsites pass the fixture through or stay on KVM. The kernel_file property now resolves through the backend interface; KVM continues to boot the guest kernel's vmlinux. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
Replace the stringly-typed uvm_lifecycle with a VmLifecycle str-enum. Keep pytest IDs and dynamic fixture names unchanged by using enum values explicitly, and update the A/B fixture to compare enum members. This makes the dimension states explicit without changing the generated test matrix. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
Report host backend capabilities in the pytest header: /dev/kvm existence, access and mode. Backend selection failures are then diagnosable from the pytest/CI log alone. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
Modules that need customisation between lifecycle stages - e.g. an extra device or a file pushed into the guest - previously reimplemented the whole spawn/basic_config/add_net_iface/start sequence locally. Document same-name stage shadowing (including decoration via same-name chaining) as a supported extension point in the fixture-system comment block, and adopt it: - test_rng.py: replace the local boot/restore constructors, the uvm_ctor parametrization, the rate_limiter fixture and the private uvm_any with two stage overrides: uvm_configured mirrors the conftest stage but spawns at INFO log level (preserving b755a67, "test: set log level to INFO for test_rng"), and uvm_booted configures the entropy device before boot. - test_sysgenid.py, test_vmclock.py: decorate uvm_booted via same-name chaining to scp the guest test binary, deleting the local boot fixtures. - test_fips.py: override uvm_configured to supply the command line parameters to the guest kernel, rather than maintaining an additional uvm_with_fips fixture. Collected test variants are unchanged across all touched modules. Signed-off-by: James Curtis <jxcurtis@amazon.co.uk>
JamesC1305
added
the
Status: Awaiting review
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #6149 +/- ##
=======================================
Coverage 82.98% 82.98%
=======================================
Files 277 277
Lines 30885 30885
=======================================
Hits 25630 25630
Misses 5255 5255 Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Manciukic
approved these changes
Aug 26, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Changes
Major changes:
uvmfixture,vm_backendGuestKernelwrapper and replaceguest_kernelfixture's raw path parameterisation withGuestKernelobjectSmall changes / fixes:
uvm_configuredlocally, which the booted/restoreduvmfixtures then consume for that modulenum_vcpus + 2)Reason
Introduce the VM backend dimension to abstract away KVM-specific behaviour from the microvm. Anything KVM specific is handled by methods on the backend.
The new
GuestKerneldataclass provides richer access to information about the guest kernel, and allows selection between direct-boot and EFI images.License Acceptance
By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache 2.0 license. For more information on following Developer
Certificate of Origin and signing off your commits, please check
CONTRIBUTING.md.PR Checklist
tools/devtool checkbuild --allto verify that the PR passesbuild checks on all supported architectures.
tools/devtool checkstyleto verify that the PR passes theautomated style checks.
how they are solving the problem in a clear and encompassing way.
in the PR.
CHANGELOG.md.Runbook for Firecracker API changes.
integration tests.
TODO.rust-vmm.