Update registry initialization to read on-disk CDI spec

Signed-off-by: Ivan Sim <[email protected]>

go.mod

@@ -6,7 +6,6 @@ toolchain go1.22.2
 
 require (
 	github.com/container-orchestrated-devices/container-device-interface v0.5.4
-	github.com/google/uuid v1.4.0
 	github.com/prometheus/client_golang v1.19.0
 	github.com/rs/zerolog v1.32.0
 	github.com/spf13/cobra v1.8.0
@@ -40,6 +39,7 @@ require (
 	github.com/google/gnostic-models v0.6.8 // indirect
 	github.com/google/go-cmp v0.6.0 // indirect
 	github.com/google/gofuzz v1.2.0 // indirect
+	github.com/google/uuid v1.4.0 // indirect
 	github.com/hashicorp/hcl v1.0.0 // indirect
 	github.com/imdario/mergo v0.3.6 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect

pkg/drivers/gpu/kubelet/cdi/cdi.go

@@ -1,6 +1,7 @@
 package cdi
 
 import (
+	"errors"
 	"fmt"
 	"sync"
 
@@ -64,3 +65,15 @@ func DeleteCDISpec(claimUID string) error {
 	specName := cdiapi.GenerateTransientSpecName(cdiVendor, cdiClass, claimUID)
 	return registry.SpecDB().RemoveSpec(specName)
 }
+
+func Specs() ([]*cdiapi.Spec, error) {
+	specs := registry.SpecDB().GetVendorSpecs(cdiVendor)
+
+	var errs error
+	for _, spec := range specs {
+		specErrs := registry.SpecDB().GetSpecErrors(spec)
+		specErrs = append(specErrs, errs)
+		errs = errors.Join(specErrs...)
+	}
+	return specs, errs
+}

pkg/drivers/gpu/kubelet/nodeserver.go

@@ -5,10 +5,9 @@ import (
 	"errors"
 	"fmt"
 
-	"github.com/google/uuid"
 	draclientset "github.com/ihcsim/k8s-dra/pkg/apis/clientset/versioned"
 	gpuv1alpha1 "github.com/ihcsim/k8s-dra/pkg/apis/gpu/v1alpha1"
-	cdi "github.com/ihcsim/k8s-dra/pkg/drivers/gpu/kubelet/cdi"
+	"github.com/ihcsim/k8s-dra/pkg/drivers/gpu/kubelet/cdi"
 	zlog "github.com/rs/zerolog"
 	resourcev1alpha2 "k8s.io/api/resource/v1alpha2"
 	apierrs "k8s.io/apimachinery/pkg/api/errors"
@@ -21,6 +20,11 @@ const AvailableGPUsCount = 4
 
 var _ kubeletdrav1.NodeServer = &NodeServer{}
 
+type gpuDevice struct {
+	*gpuv1alpha1.GPUDevice
+	cdiAnnotations map[string]string
+}
+
 // NodeServer provides the API implementation of the node server.
 // see https://pkg.go.dev/k8s.io/kubelet/pkg/apis/dra/v1alpha3#NodeServer
 type NodeServer struct {
@@ -42,10 +46,14 @@ func NewNodeServer(
 	nodeName string,
 	availableGPUs int,
 	log zlog.Logger) (*NodeServer, error) {
-	gpus := discoverGPUs(availableGPUs)
+	gpus, err := discoverGPUs(availableGPUs)
+	if err != nil {
+		return nil, err
+	}
+
 	var gpuDevices []*gpuv1alpha1.GPUDevice
 	for _, gpu := range gpus {
-		gpuDevices = append(gpuDevices, gpu.device)
+		gpuDevices = append(gpuDevices, gpu.GPUDevice)
 	}
 
 	if _, err := clientSets.GpuV1alpha1().NodeDevices(namespace).Get(ctx, nodeName, metav1.GetOptions{}); err != nil && apierrs.IsNotFound(err) {
@@ -214,20 +222,23 @@ func (n *NodeServer) NodeListAndWatchResources(req *kubeletdrav1.NodeListAndWatc
 	return s.Send(res)
 }
 
-type gpu struct {
-	device *gpuv1alpha1.GPUDevice
-}
+func discoverGPUs(maxAvailableGPU int) ([]*gpuDevice, error) {
+	specs, err := cdi.Specs()
+	if err != nil {
+		return nil, err
+	}
 
-func discoverGPUs(maxAvailableGPU int) []*gpu {
-	gpus := make([]*gpu, maxAvailableGPU)
-	for i := 0; i < maxAvailableGPU; i++ {
-		gpus = append(gpus, &gpu{
-			device: &gpuv1alpha1.GPUDevice{
-				UUID:        uuid.NewString(),
-				ProductName: "NVIDIA Tesla V100",
-			},
-		})
+	var gpuDevices []*gpuDevice
+	for _, spec := range specs {
+		for _, device := range spec.Devices {
+			gpuDevices = append(gpuDevices, &gpuDevice{
+				GPUDevice: &gpuv1alpha1.GPUDevice{
+					UUID:        device.Name,
+					ProductName: device.ContainerEdits.Env[1],
+				},
+			})
+		}
 	}
 
-	return gpus
+	return gpuDevices, nil
 }