Refactor cdi api

[elezar]

The original CDI spec generation API was focussed on NVML device specifically. Since then we have replaced the more specific functions (for GPU and MIG devices) in the API with more generally applicable functions based on mode and device IDs.

This organic growth of APIs also means that for the NVML case specifically we had multiple different implementations of CDI spec generation making keeping things consistent more difficult.

Thes changes remove the redundant functions in the nvcdi.Interface allowing devices to be requested by ID across all use cases. It also refactors the CDI spec generation for NVML devices to ensure that the same generation logic is used for all cases.

[coveralls]

Pull Request Test Coverage Report for Build 16001110444

Details

• 112 of 248 (45.16%) changed or added relevant lines in 12 files are covered.
• 9 unchanged lines in 5 files lost coverage.
• Overall coverage increased (+0.9%) to 34.13%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
pkg/nvcdi/lib-wsl.go	0	3	0.0%
pkg/nvcdi/gds.go	0	4	0.0%
pkg/nvcdi/lib-imex.go	41	45	91.11%
pkg/nvcdi/management.go	0	4	0.0%
pkg/nvcdi/mofed.go	0	4	0.0%
pkg/nvcdi/wrapper.go	12	17	70.59%
pkg/nvcdi/lib.go	3	9	33.33%
pkg/nvcdi/lib-csv.go	0	10	0.0%
pkg/nvcdi/full-gpu-nvml.go	8	26	30.77%
pkg/nvcdi/mig-device-nvml.go	14	33	42.42%

Files with Coverage Reduction	New Missed Lines	%
pkg/nvcdi/lib-wsl.go	1	0.0%
pkg/nvcdi/mig-device-nvml.go	1	25.0%
pkg/nvcdi/lib-imex.go	2	82.76%
pkg/nvcdi/wrapper.go	2	70.21%
pkg/nvcdi/lib-nvml.go	3	42.48%

Totals
Change from base Build 15971468852:	0.9%
Covered Lines:	4485
Relevant Lines:	13141

---

💛 - Coveralls

[Copilot]

Pull Request Overview

Refactors the CDI API to use a unified factory-based SpecGenerator pattern and consolidates NVML spec generation logic.

• Introduce deviceSpecGeneratorFactory, SpecGenerator, and DeviceSpecGenerator types and update all implementations accordingly
• Replace redundant per-device methods with fullGPUDeviceSpecGenerator / migDeviceSpecGenerator and a combined DeviceSpecGenerators type
• Update wrapper and CLI to use GetDeviceSpecsByID("all") and deprecate GetAllDeviceSpecs

Reviewed Changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
pkg/nvcdi/wrapper.go	Added factory field and new SpecGenerator types
pkg/nvcdi/mofed.go	Implement factory interface for MOFED
pkg/nvcdi/mig-device-nvml.go	Extracted MIG spec generator type
pkg/nvcdi/management.go	Implement factory interface for management mode
pkg/nvcdi/lib.go	Wire deviceSpecGeneratorFactory into New
pkg/nvcdi/lib-wsl.go	Implement factory interface for WSL mode
pkg/nvcdi/lib-nvml_test.go	Add tests for NVML DeviceSpecGenerators
pkg/nvcdi/lib-nvml.go	Major refactor: split Init/Shutdown and generators
pkg/nvcdi/lib-imex.go	Implement factory interface for IMEX
pkg/nvcdi/lib-csv.go	Implement factory interface for CSV mode
pkg/nvcdi/gds.go	Implement factory interface for GDS mode
pkg/nvcdi/full-gpu-nvml.go	Extracted full GPU spec generator type
pkg/nvcdi/api.go	Updated Interface to use SpecGenerator
cmd/nvidia-ctk/cdi/generate/generate.go	Replace deprecated GetAllDeviceSpecs call

Comments suppressed due to low confidence (4)

pkg/nvcdi/lib-nvml.go:108

• [nitpick] The local variable DeviceSpecGenerators shadows the type DeviceSpecGenerators, which may confuse readers. Rename the variable (e.g., generators) to avoid shadowing.

	var DeviceSpecGenerators DeviceSpecGenerators

pkg/nvcdi/wrapper.go:39

• [nitpick] Add a Go doc comment for deviceSpecGeneratorFactory to explain its role in producing DeviceSpecGenerator instances.

// TODO: Rename this type

pkg/nvcdi/lib-nvml.go:47

• [nitpick] Expand this comment to specify the exact ID formats supported (e.g., gpuIndex, uuid, or gpuIndex:migIndex for MIG devices) so consumers know how to request each device.

// DeviceSpecGenerators returns the CDI device spec generators for NVML devices

pkg/nvcdi/lib-nvml_test.go:37

• Consider adding a test case for invalid device IDs (e.g., an unsupported string) to verify that getDeviceSpecGeneratorsForIDs returns an appropriate error.

		expectedLength     int

[Copilot]

The fields index and migIndex on migDeviceSpecGenerator are never initialized, causing getNames() to always use zero indices. Pass and set the correct index and migIndex when constructing the generator.

Suggested change

	func (l *nvmllib) newMIGDeviceSpecGeneratorFromNVMLDevice(id string, nvmlDevice nvml.Device) (DeviceSpecGenerator, error) {
	func (l *nvmllib) newMIGDeviceSpecGeneratorFromNVMLDevice(id string, nvmlDevice nvml.Device, index int, migIndex int) (DeviceSpecGenerator, error) {