[rahulait]

Description

Fixes: #2817

Fixes RHEL driver installation when a custom repository is used alongside the host’s Red Hat subscription repositories.
Users can now configure:

driver:
  repoConfig:
    configMapName: custom-repos
    useHostSubscription: true

This mounts both the custom repository and the host RHEL subscription configuration, allowing DNF to access BaseOS/AppStream packages as well as additional custom repositories.
The default behavior is unchanged: when useHostSubscription is omitted, repoConfig remains suitable for air-gapped installations and host subscription mounts are skipped.

Checklist

• No secrets, sensitive information, or unrelated changes
• Lint checks passing (make lint)
• Generated assets in-sync (make validate-generated-assets)
• Go mod artifacts in-sync (make validate-modules)
• Test cases are added for new code paths

Testing

---

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds UseHostSubscription to both driver repository API versions and Helm configuration. RHEL subscription mounts are retained when repository configuration is enabled and host subscription use is explicitly enabled. Controller and driver state tests cover custom repositories, host subscription mounts, repeated transformation, configuration propagation, and unique mount paths.

Priority: ➖ Normal

Severity of issue fixed: Medium

Merge Risk: 🔵 Low · up to dd333

The feature remains usable through CRDs, but OLM metadata and regression coverage are incomplete. Address the generated CSV and test assertions before relying on the combined repository behavior.

---

_{Comment @coderabbitai help to get the list of available commands.}

[rahulait]

@coderabbitai full review

[coderabbitai]

✅ Action performed

Full review finished.

[coderabbitai]

Note

Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.

🟡 Other comments (3)

api/nvidia/v1/clusterpolicy_types.go-1284-1288 (1)

1284-1288: 📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win

Regenerate the CSV metadata for useHostSubscription.

The CRD schemas and deepcopy files already include this field. However, the generated CSV contains no descriptor for either API type, although both Go fields use +operator-sdk:gen-csv markers. Regenerate the CSV so OLM metadata exposes useHostSubscription.

controllers/transforms_test.go-4800-4800 (1)

4800-4800: 📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win

Assert the custom repository volume and mount in the combined case.

TransformDriver adds test-repo-config and mounts custom.repo at /etc/yum.repos.d/custom.repo. The current helpers inspect only subscription-prefixed volumes, so the test still passes if either custom repository contribution is removed. Add assertions for both resources.

internal/state/driver_test.go-671-682 (1)

671-682: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Assert the custom repository volume and mount in the combined case.

When repoConfigEnabled and useHostSubscription are both true, getDriverAdditionalConfigs adds the test-repo-config ConfigMap volume and mount. The test only checks subscription entries and unique mount paths, so removing the custom repository volume and mount would leave every assertion passing. Add assertions for both to cover the required coexistence behavior.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: e10c688b-e835-415e-884b-391d65097ece

📥 Commits

Reviewing files that changed from the base of the PR and between 3fc63e2 and dd333ec.

⛔ Files ignored due to path filters (6)

• bundle/manifests/nvidia.com_clusterpolicies.yaml is excluded by !bundle/manifests/nvidia.com_*.yaml
• bundle/manifests/nvidia.com_nvidiadrivers.yaml is excluded by !bundle/manifests/nvidia.com_*.yaml
• config/crd/bases/nvidia.com_clusterpolicies.yaml is excluded by !config/crd/bases/**
• config/crd/bases/nvidia.com_nvidiadrivers.yaml is excluded by !config/crd/bases/**
• deployments/gpu-operator/crds/nvidia.com_clusterpolicies.yaml is excluded by !deployments/gpu-operator/crds/**
• deployments/gpu-operator/crds/nvidia.com_nvidiadrivers.yaml is excluded by !deployments/gpu-operator/crds/**

📒 Files selected for processing (9)

• api/nvidia/v1/clusterpolicy_types.go
• api/nvidia/v1alpha1/nvidiadriver_types.go
• controllers/object_controls.go
• controllers/transforms_test.go
• deployments/gpu-operator/templates/nvidiadriver.yaml
• deployments/gpu-operator/values.yaml
• internal/consts/consts.go
• internal/state/driver_test.go
• internal/state/driver_volumes.go

Included review availability: Your plan provides up to 12 included reviews per hour; 10 remain after this review.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[kvalliyurnatt]

IIUC this is to ensure if TransformDriver is called twice we don't add the same volumes twice. But I am not sure if this is practically possible with our code currently, We seem to call with an empty manifest always. Also if we want to address the idempotency issue with TransformDriver, shouldn't we do this for other volumes as well ?

[rahulait]

Yes, this is defensive handling just in case a future change accidentally transforms the same object twice. This is not needed by the current reconcile code. We can keep the scope change minimal and do the defensive handling for all volumes if needed in a future change, so just keep this PR focused on the subscription fix.

[kvalliyurnatt]

Yeah, I think I would prefer that since right now it is not possible to hit this issue and the fix is partial right now

[rahulait]

Closing in favor of #3109