Merge branch 'detect-gpus-flag' into 'master'

Detect use of --gpus flag in experimental mode

See merge request nvidia/container-toolkit/container-toolkit!125

cmd/nvidia-container-runtime/modifier/experimental_test.go

@@ -231,7 +231,7 @@ func TestExperimentalModifier(t *testing.T) {
 			},
 		},
 		{
-			description: "modification fails for existing nvidia-container-runtime-hook",
+			description: "modification removes existing nvidia-container-runtime-hook",
 			spec: &specs.Spec{
 				Hooks: &specs.Hooks{
 					Prestart: []specs.Hook{
@@ -254,20 +254,19 @@ func TestExperimentalModifier(t *testing.T) {
 					return hooks, nil
 				},
 			},
-			expectedError: fmt.Errorf("nvidia-container-runtime-hook already exists"),
 			expectedSpec: &specs.Spec{
 				Hooks: &specs.Hooks{
 					Prestart: []specs.Hook{
 						{
-							Path: "/path/to/nvidia-container-runtime-hook",
+							Path: "/hook/b",
-							Args: []string{"/path/to/nvidia-container-runtime-hook", "prestart"},
+							Args: []string{"/hook/b", "argb"},
 						},
 					},
 				},
 			},
 		},
 		{
-			description: "modification fails for existing nvidia-container-toolkit",
+			description: "modification removes existing nvidia-container-toolkit",
 			spec: &specs.Spec{
 				Hooks: &specs.Hooks{
 					Prestart: []specs.Hook{
@@ -290,13 +289,12 @@ func TestExperimentalModifier(t *testing.T) {
 					return hooks, nil
 				},
 			},
-			expectedError: fmt.Errorf("nvidia-container-toolkit already exists"),
 			expectedSpec: &specs.Spec{
 				Hooks: &specs.Hooks{
 					Prestart: []specs.Hook{
 						{
-							Path: "/path/to/nvidia-container-toolkit",
+							Path: "/hook/b",
-							Args: []string{"/path/to/nvidia-container-toolkit", "prestart"},
+							Args: []string{"/hook/b", "argb"},
 						},
 					},
 				},

cmd/nvidia-container-runtime/modifier/hook_remover.go

@@ -17,7 +17,6 @@
 package modifier
 
 import (
-	"fmt"
 	"path/filepath"
 
 	"github.com/NVIDIA/nvidia-container-toolkit/internal/config"
@@ -47,10 +46,19 @@ func (m nvidiaContainerRuntimeHookRemover) Modify(spec *specs.Spec) error {
 		return nil
 	}
 
+	var newPrestart []specs.Hook
+
 	for _, hook := range spec.Hooks.Prestart {
 		if isNVIDIAContainerRuntimeHook(&hook) {
-			return fmt.Errorf("spec already contains required 'prestart' hook")
+			m.logger.Debugf("Removing hook %v", hook)
+			continue
 		}
+		newPrestart = append(newPrestart, hook)
+	}
+
+	if len(newPrestart) != len(spec.Hooks.Prestart) {
+		m.logger.Debugf("Updating 'prestart' hooks to %v", newPrestart)
+		spec.Hooks.Prestart = newPrestart
 	}
 
 	return nil

cmd/nvidia-container-toolkit/hook_config.go

@@ -7,6 +7,7 @@ import (
 	"reflect"
 
 	"github.com/BurntSushi/toml"
+	"github.com/NVIDIA/nvidia-container-toolkit/internal/config"
 )
 
 const (
@@ -41,10 +42,11 @@ type HookConfig struct {
 	AcceptDeviceListAsVolumeMounts bool               `toml:"accept-nvidia-visible-devices-as-volume-mounts"`
 	SupportedDriverCapabilities    DriverCapabilities `toml:"supported-driver-capabilities"`
 
-	NvidiaContainerCLI CLIConfig `toml:"nvidia-container-cli"`
+	NvidiaContainerCLI     CLIConfig            `toml:"nvidia-container-cli"`
+	NVIDIAContainerRuntime config.RuntimeConfig `toml:"nvidia-container-runtime"`
 }
 
-func getDefaultHookConfig() (config HookConfig) {
+func getDefaultHookConfig() HookConfig {
 	return HookConfig{
 		DisableRequire:                 false,
 		SwarmResource:                  nil,
@@ -63,6 +65,7 @@ func getDefaultHookConfig() (config HookConfig) {
 			User:        nil,
 			Ldconfig:    nil,
 		},
+		NVIDIAContainerRuntime: *config.GetDefaultRuntimeConfig(),
 	}
 }
 

cmd/nvidia-container-toolkit/main.go

@@ -17,6 +17,7 @@ import (
 
 var (
 	debugflag  = flag.Bool("debug", false, "enable debug output")
+	forceflag  = flag.Bool("force", false, "force execution of prestart hook in experimental mode")
 	configflag = flag.String("config", "", "configuration file")
 
 	defaultPATH = []string{"/usr/local/sbin", "/usr/local/bin", "/usr/sbin", "/usr/bin", "/sbin", "/bin"}
@@ -85,6 +86,10 @@ func doPrestart() {
 	hook := getHookConfig()
 	cli := hook.NvidiaContainerCLI
 
+	if hook.NVIDIAContainerRuntime.Experimental && !*forceflag {
+		log.Panicln("invoking the NVIDIA Container Runtime Hook directly (e.g. specifying the docker --gpus flag) is not supported. Please use the NVIDIA Container Runtime instead.")
+	}
+
 	container := getContainerConfig(hook)
 	nvidia := container.Nvidia
 	if nvidia == nil {

internal/config/config.go

@@ -103,7 +103,7 @@ func getDefaultConfig() *Config {
 	c := Config{
 		NVIDIAContainerCLIConfig:     *getDefaultContainerCLIConfig(),
 		NVIDIACTKConfig:              *getDefaultCTKConfig(),
-		NVIDIAContainerRuntimeConfig: *getDefaultRuntimeConfig(),
+		NVIDIAContainerRuntimeConfig: *GetDefaultRuntimeConfig(),
 	}
 
 	return &c

internal/config/runtime.go

@@ -29,7 +29,7 @@ type RuntimeConfig struct {
 
 // getRuntimeConfigFrom reads the nvidia container runtime config from the specified toml Tree.
 func getRuntimeConfigFrom(toml *toml.Tree) *RuntimeConfig {
-	cfg := getDefaultRuntimeConfig()
+	cfg := GetDefaultRuntimeConfig()
 
 	if toml == nil {
 		return cfg
@@ -42,8 +42,8 @@ func getRuntimeConfigFrom(toml *toml.Tree) *RuntimeConfig {
 	return cfg
 }
 
-// getDefaultRuntimeConfig defines the default values for the config
+// GetDefaultRuntimeConfig defines the default values for the config
-func getDefaultRuntimeConfig() *RuntimeConfig {
+func GetDefaultRuntimeConfig() *RuntimeConfig {
 	c := RuntimeConfig{
 		DebugFilePath: "/dev/null",
 		Experimental:  false,

internal/discover/legacy.go

@@ -59,7 +59,7 @@ func (d legacy) Hooks() ([]Hook, error) {
 	}
 	d.logger.Debugf("Using NVIDIA Container Runtime Hook path %v", hookPath)
 
-	args := []string{hookPath, "prestart"}
+	args := []string{hookPath, "--force", "prestart"}
 	h := Hook{
 		Lifecycle: cdi.PrestartHook,
 		Path:      hookPath,