Invoke create-device-nodes directly

Signed-off-by: Evan Lezar <[email protected]>

Author: elezar

cmd/nvidia-mig-manager/main.go

@@ -418,7 +418,7 @@ func migReconfigure(ctx context.Context, migConfigValue string, clientset *kuber
 		return fmt.Errorf("error creating reconfigure instance: %w", err)
 	}
 
-	return rcfg.Run()
+	return rcfg.Run(ctx)
 }
 
 func ContinuouslySyncMigConfigChanges(clientset *kubernetes.Clientset, migConfig *SyncableMigConfig) chan struct{} {

go.mod

@@ -46,6 +46,7 @@ require (
 	github.com/russross/blackfriday/v2 v2.1.0 // indirect
 	github.com/spf13/pflag v1.0.5 // indirect
 	github.com/syndtr/gocapability v0.0.0-20200815063812-42c35b437635 // indirect
+	github.com/urfave/cli/v3 v3.4.1 // indirect
 	github.com/x448/float16 v0.8.4 // indirect
 	github.com/xrash/smetrics v0.0.0-20240521201337-686a1a2994c1 // indirect
 	go.yaml.in/yaml/v2 v2.4.2 // indirect

go.sum

@@ -121,6 +121,8 @@ github.com/syndtr/gocapability v0.0.0-20200815063812-42c35b437635/go.mod h1:hkRG
 github.com/urfave/cli v1.19.1/go.mod h1:70zkFmudgCuE/ngEzBv17Jvp/497gISqfk5gWijbERA=
 github.com/urfave/cli/v2 v2.27.7 h1:bH59vdhbjLv3LAvIu6gd0usJHgoTTPhCFib8qqOwXYU=
 github.com/urfave/cli/v2 v2.27.7/go.mod h1:CyNAG/xg+iAOg0N4MPGZqVmv2rCoP267496AOXUZjA4=
+github.com/urfave/cli/v3 v3.4.1 h1:1M9UOCy5bLmGnuu1yn3t3CB4rG79Rtoxuv1sPhnm6qM=
+github.com/urfave/cli/v3 v3.4.1/go.mod h1:FJSKtM/9AiiTOJL4fJ6TbMUkxBXn7GO9guZqoZtpYpo=
 github.com/x448/float16 v0.8.4 h1:qLwI1I70+NjRFUR3zs1JPUCgaCXSh3SW62uAKT1mSBM=
 github.com/x448/float16 v0.8.4/go.mod h1:14CWIYCyZA/cWjXOioeEpHeN/83MdbZDRQHoFcYsOfg=
 github.com/xeipuuv/gojsonpointer v0.0.0-20180127040702-4e3ac2762d5f/go.mod h1:N2zxlSyiKSe5eX1tZViRH5QA0qijqEDrYZiPEAiq3wU=

pkg/mig/reconfigure/reconfigure.go

@@ -31,6 +31,7 @@ import (
 	"k8s.io/client-go/kubernetes"
 	"tags.cncf.io/container-device-interface/pkg/cdi"
 
+	devicenodes "github.com/NVIDIA/nvidia-container-toolkit/cmd/nvidia-ctk/system/create-device-nodes"
 	"github.com/NVIDIA/nvidia-container-toolkit/pkg/nvcdi"
 	transformroot "github.com/NVIDIA/nvidia-container-toolkit/pkg/nvcdi/transform/root"
 
@@ -122,7 +123,7 @@ func New(ctx context.Context, clientset *kubernetes.Clientset, migPartedBinary [
 }
 
 // Run executes the complete MIG reconfiguration process
-func (r *Reconfigure) Run() error {
+func (r *Reconfigure) Run(ctx context.Context) error {
 
 	// Ensure systemd managers are cleaned up
 	defer r.cleanup()
@@ -212,7 +213,7 @@ func (r *Reconfigure) Run() error {
 			_ = r.setState(migStateFailed)
 			return fmt.Errorf("failed to run nvidia-smi: %w", err)
 		}
-		if err := r.handleCDI(); err != nil {
+		if err := r.handleCDI(ctx); err != nil {
 			_ = r.setState(migStateFailed)
 			return fmt.Errorf("failed to handle CDI: %w", err)
 		}
@@ -469,14 +470,16 @@ func (r *Reconfigure) applyMigConfig() error {
 }
 
 // handleCDI handles CDI operations if enabled
-func (r *Reconfigure) handleCDI() error {
+func (r *Reconfigure) handleCDI(ctx context.Context) error {
 
 	log.Info("Creating NVIDIA control device nodes")
-	// TODO: Instead of shelling out, we need to invoke the method via Go. The Toolkit code needs to be refactored first.
-	cmd := exec.Command("nvidia-ctk", "system", "create-device-nodes", "--control-devices", "--dev-root="+r.opts.DevRootCtrPath)
-	cmd.Stdout = os.Stdout
-	cmd.Stderr = os.Stderr
-	if err := cmd.Run(); err != nil {
+
+	// TODO: Instead of abusing CLI command we generate here, we should expose
+	// this API and use that instead. This would require refactoring in the
+	// toolkit.
+	cmd := devicenodes.NewCommand(log.StandardLogger())
+	err := cmd.Run(ctx, []string{"--control-devices", "--dev-root" + r.opts.DevRootCtrPath})
+	if err != nil {
 		return fmt.Errorf("failed to create control device nodes: %w", err)
 	}