Merge branch 'fix-hook-removal' into 'main'

Improve locating NVIDIA Container Runtime Hook

See merge request nvidia/container-toolkit/container-toolkit!215

cmd/nvidia-container-runtime/main.go

@@ -48,7 +48,12 @@ func run(argv []string) (rerr error) {
 	if err != nil {
 		return fmt.Errorf("failed to set up logger: %v", err)
 	}
-	defer logger.Reset()
+	defer func() {
+		if rerr != nil {
+			logger.Errorf("%v", rerr)
+		}
+		logger.Reset()
+	}()
 
 	logger.Debugf("Command line arguments: %v", argv)
 	runtime, err := newNVIDIAContainerRuntime(logger.Logger, cfg, argv)

internal/modifier/stable.go

@@ -17,11 +17,10 @@
 package modifier
 
 import (
-	"os"
+	"fmt"
-	"os/exec"
-	"path/filepath"
 
 	"github.com/NVIDIA/nvidia-container-toolkit/internal/config"
+	"github.com/NVIDIA/nvidia-container-toolkit/internal/lookup"
 	"github.com/NVIDIA/nvidia-container-toolkit/internal/oci"
 	"github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/sirupsen/logrus"
@@ -44,21 +43,8 @@ type stableRuntimeModifier struct {
 // Modify applies the required modification to the incoming OCI spec, inserting the nvidia-container-runtime-hook
 // as a prestart hook.
 func (m stableRuntimeModifier) Modify(spec *specs.Spec) error {
-	path, err := exec.LookPath(config.NVIDIAContainerRuntimeHookExecutable)
+	// If an NVIDIA Container Runtime Hook already exists, we don't make any modifications to the spec.
-	if err != nil {
+	if spec.Hooks != nil {
-		path = filepath.Join(config.DefaultExecutableDir, config.NVIDIAContainerRuntimeHookExecutable)
-		_, err = os.Stat(path)
-		if err != nil {
-			return err
-		}
-	}
-
-	m.logger.Infof("Using prestart hook path: %s", path)
-
-	args := []string{path}
-	if spec.Hooks == nil {
-		spec.Hooks = &specs.Hooks{}
-	} else if len(spec.Hooks.Prestart) != 0 {
 		for _, hook := range spec.Hooks.Prestart {
 			if isNVIDIAContainerRuntimeHook(&hook) {
 				m.logger.Infof("Existing nvidia prestart hook (%v) found in OCI spec", hook.Path)
@@ -67,6 +53,21 @@ func (m stableRuntimeModifier) Modify(spec *specs.Spec) error {
 		}
 	}
 
+	// We create a locator and look for the NVIDIA Container Runtime Hook in the path.
+	candidates, err := lookup.NewExecutableLocator(m.logger, "").Locate(config.NVIDIAContainerRuntimeHookExecutable)
+	if err != nil {
+		return fmt.Errorf("failed to locate NVIDIA Container Runtime Hook: %v", err)
+	}
+	path := candidates[0]
+	if len(candidates) > 1 {
+		m.logger.Debugf("Using %v from multiple NVIDIA Container Runtime Hook candidates: %v", path, candidates)
+	}
+
+	m.logger.Infof("Using prestart hook path: %v", path)
+	args := []string{path}
+	if spec.Hooks == nil {
+		spec.Hooks = &specs.Hooks{}
+	}
 	spec.Hooks.Prestart = append(spec.Hooks.Prestart, specs.Hook{
 		Path: path,
 		Args: append(args, "prestart"),

packaging/rpm/SPECS/nvidia-container-toolkit.spec

@@ -56,7 +56,7 @@ install -m 644 -t %{buildroot}/usr/share/containers/oci/hooks.d oci-nvidia-hook.
 ln -sf %{_bindir}/nvidia-container-runtime-hook %{_bindir}/nvidia-container-toolkit
 
 %postun
-rm -f %{_bindir}/nvidia-container-toolkit
+if [ -L %{_bindir}/nvidia-container-toolkit ] then; rm -f %{_bindir}/nvidia-container-toolkit; fi
 
 %files
 %license LICENSE