Merge branch 'backport-install-nvidia-ctk' into 'release-1.12'

Backport changes for 1.12.1 release

See merge request nvidia/container-toolkit/container-toolkit!332

CHANGELOG.md

@@ -9,6 +9,9 @@
 * Discover all `gsb*.bin` GSP firmware files when generating CDI specification.
 * Remove `fedora35` packaging targets
 * [libnvidia-container] Include all `gsp*.bin` firmware files if present
+* [toolkit-container] Install `nvidia-ctk` from toolkit container
+* [toolkit-container] Use installed `nvidia-ctk` path in NVIDIA Container Toolkit config
+* [toolkit-container] Bump CUDA base images to 12.1.0
 
 ## v1.12.0
 

test/container/toolkit_test.sh

@@ -61,6 +61,7 @@ testing::toolkit::install() {
 	grep -q -E "^\s*ldconfig = \"@${nvidia_run_dir}/driver/sbin/ldconfig(.real)?\"" "${shared_dir}/usr/local/nvidia/toolkit/.config/nvidia-container-runtime/config.toml"
 	grep -q -E "^\s*root = \"${nvidia_run_dir}/driver\"" "${shared_dir}/usr/local/nvidia/toolkit/.config/nvidia-container-runtime/config.toml"
 	grep -q -E "^\s*path = \"/usr/local/nvidia/toolkit/nvidia-container-cli\"" "${shared_dir}/usr/local/nvidia/toolkit/.config/nvidia-container-runtime/config.toml"
+	grep -q -E "^\s*path = \"/usr/local/nvidia/toolkit/nvidia-ctk\"" "${shared_dir}/usr/local/nvidia/toolkit/.config/nvidia-container-runtime/config.toml"
 }
 
 testing::toolkit::delete() {

tools/container/toolkit/toolkit.go

@@ -204,7 +204,12 @@ func Install(cli *cli.Context, opts *options) error {
 		return fmt.Errorf("error installing NVIDIA container runtime hook: %v", err)
 	}
 
-	err = installToolkitConfig(toolkitConfigPath, nvidiaContainerCliExecutable, opts)
+	nvidiaCTKPath, err := installContainerToolkitCLI(opts.toolkitRoot)
+	if err != nil {
+		return fmt.Errorf("error installing NVIDIA Container Toolkit CLI: %v", err)
+	}
+
+	err = installToolkitConfig(toolkitConfigPath, nvidiaContainerCliExecutable, nvidiaCTKPath, opts)
 	if err != nil {
 		return fmt.Errorf("error installing NVIDIA container toolkit config: %v", err)
 	}
@@ -262,7 +267,7 @@ func installLibrary(libName string, toolkitRoot string) error {
 
 // installToolkitConfig installs the config file for the NVIDIA container toolkit ensuring
 // that the settings are updated to match the desired install and nvidia driver directories.
-func installToolkitConfig(toolkitConfigPath string, nvidiaContainerCliExecutablePath string, opts *options) error {
+func installToolkitConfig(toolkitConfigPath string, nvidiaContainerCliExecutablePath string, nvidiaCTKPath string, opts *options) error {
 	log.Infof("Installing NVIDIA container toolkit config '%v'", toolkitConfigPath)
 
 	config, err := toml.LoadFile(nvidiaContainerToolkitConfigSource)
@@ -311,6 +316,9 @@ func installToolkitConfig(toolkitConfigPath string, nvidiaContainerCliExecutable
 		config.Set(key, value)
 	}
 
+	// Set nvidia-ctk options
+	config.Set("nvidia-ctk.path", nvidiaCTKPath)
+
 	_, err = config.WriteTo(targetConfig)
 	if err != nil {
 		return fmt.Errorf("error writing config: %v", err)
@@ -322,6 +330,19 @@ func installToolkitConfig(toolkitConfigPath string, nvidiaContainerCliExecutable
 	return nil
 }
 
+// installContainerToolkitCLI installs the nvidia-ctk CLI executable and wrapper.
+func installContainerToolkitCLI(toolkitDir string) (string, error) {
+	e := executable{
+		source: "/usr/bin/nvidia-ctk",
+		target: executableTarget{
+			dotfileName: "nvidia-ctk.real",
+			wrapperName: "nvidia-ctk",
+		},
+	}
+
+	return e.install(toolkitDir)
+}
+
 // installContainerCLI sets up the NVIDIA container CLI executable, copying the executable
 // and implementing the required wrapper
 func installContainerCLI(toolkitRoot string) (string, error) {

versions.mk

@@ -24,7 +24,7 @@ NVIDIA_CONTAINER_RUNTIME_VERSION := 3.12.0
 # Specify the expected libnvidia-container0 version for arm64-based ubuntu builds.
 LIBNVIDIA_CONTAINER0_VERSION := 0.10.0+jetpack
 
-CUDA_VERSION := 12.0.1
+CUDA_VERSION := 12.1.0
 GOLANG_VERSION := 1.18.8
 
 GIT_COMMIT ?= $(shell git describe --match="" --dirty --long --always --abbrev=40 2> /dev/null || echo "")