cdi: Add CDI spec generation

[zvonkok]

After enumerating the needed devices via go-nvlib, create the CDI spec before starting the device-plugin.

Depends On: #25

[Copilot]

Pull request overview

This PR adds CDI (Container Device Interface) spec generation for NVIDIA GPUs and NVSwitches after device enumeration. The changes refactor device discovery to use go-nvlib's nvpci interface instead of manual filesystem operations, and generate CDI specifications before starting device plugins.

Changes:

• Replaced manual PCI device discovery with go-nvlib's nvpci library for device enumeration
• Added CDI spec generation for GPUs and NVSwitches with support for both IOMMUFD and legacy VFIO modes
• Refactored device structures to include additional metadata (device name, IommuFD, IsNVSwitch flag)

Reviewed changes

Copilot reviewed 7 out of 3089 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
pkg/device_plugin/generic_device_plugin.go	Improved context cancellation handling and removed manual VFIO device discovery logic
pkg/device_plugin/generic_device_plugin_test.go	Updated tests to use structured device data with complete PCI addresses and device metadata
pkg/device_plugin/device_plugin.go	Replaced filesystem-based device discovery with nvpci library and added device name formatting
pkg/device_plugin/device_plugin_test.go	Rewrote tests to use nvpci mocks instead of filesystem operations
pkg/device_plugin/constants.go	Added CDI-related constants and removed unused basePath variable
pkg/device_plugin/cdi.go	New file implementing CDI spec generation for discovered devices
go.mod	Updated Go version and added dependencies for nvpci and CDI libraries

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[zvonkok]

NvSwitches and GPUs are sorted

❯ sudo cat /var/run/cdi/nvidia.com-nvswitch.yaml
---
cdiVersion: 1.1.0
kind: nvidia.com/nvswitch
devices:
    - name: "0"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio8
    - name: "1"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio9
    - name: "2"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio10
    - name: "3"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio11

❯ sudo cat /var/run/cdi/nvidia.com-pgpu.yaml 
---
cdiVersion: 1.1.0
kind: nvidia.com/pgpu
devices:
    - name: "0"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio0
    - name: "1"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio1
    - name: "2"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio2
    - name: "3"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio3
    - name: "4"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio4
    - name: "5"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio5
    - name: "6"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio6
    - name: "7"
      containerEdits:
        deviceNodes:
            - path: /dev/vfio/devices/vfio7

[zvonkok]

❯ k describe node | grep nvidia.com                                        
                    nvidia.com/cc.capable=true
                    nvidia.com/cc.mode=ppcie
                    nvidia.com/cc.mode.state=ppcie
                    nvidia.com/gpu.deploy.cc-manager=true
                    nvidia.com/gpu.deploy.kata-manager=true
                    nvidia.com/gpu.deploy.nvsm=
                    nvidia.com/gpu.deploy.sandbox-device-plugin=true
                    nvidia.com/gpu.deploy.sandbox-validator=true
                    nvidia.com/gpu.deploy.vfio-manager=true
                    nvidia.com/gpu.family=hopper
                    nvidia.com/gpu.present=true
                    nvidia.com/gpu.workload.config=vm-passthrough
  nvidia.com/GH100_H100_NVSWITCH:             4
  nvidia.com/pgpu:                            8
  nvidia.com/GH100_H100_NVSWITCH:             4
  nvidia.com/pgpu:                            8
  nvidia.com/GH100_H100_NVSWITCH             0            0
  nvidia.com/pgpu                            0            0

Next step NVSWITCH_ALIAS

[rajatchopra]

LGTM
We can optimize the structures in a follow up if needed.