Merge pull request #376 from elezar/cherry-pick-v1.14.6

Cherry pick v1.14.6

Author: elezar

CHANGELOG.md

@@ -1,6 +1,8 @@
 # NVIDIA Container Toolkit Changelog
 
+## v1.14.6
 * Add support for extracting device major number from `/proc/devices` if `nvidia` is used as a device name over `nvidia-frontend`.
+* Add support for selecting IMEX channels using the NVIDIA_IMEX_CHANNELS environement variable.
 
 ## v1.14.5
 * Fix `nvidia-ctk runtime configure --cdi.enabled` for Docker. This was incorrectly setting `experimental = true` instead

cmd/nvidia-container-runtime-hook/container_config.go

@@ -9,9 +9,10 @@ import (
 	"path/filepath"
 	"strings"
 
-	"github.com/NVIDIA/nvidia-container-toolkit/internal/config/image"
 	"github.com/opencontainers/runtime-spec/specs-go"
 	"golang.org/x/mod/semver"
+
+	"github.com/NVIDIA/nvidia-container-toolkit/internal/config/image"
 )
 
 const (
@@ -22,6 +23,7 @@ const (
 	envNVVisibleDevices     = "NVIDIA_VISIBLE_DEVICES"
 	envNVMigConfigDevices   = "NVIDIA_MIG_CONFIG_DEVICES"
 	envNVMigMonitorDevices  = "NVIDIA_MIG_MONITOR_DEVICES"
+	envNVImexChannels       = "NVIDIA_IMEX_CHANNELS"
 	envNVDriverCapabilities = "NVIDIA_DRIVER_CAPABILITIES"
 )
 
@@ -37,6 +39,7 @@ type nvidiaConfig struct {
 	Devices            string
 	MigConfigDevices   string
 	MigMonitorDevices  string
+	ImexChannels       string
 	DriverCapabilities string
 	// Requirements defines the requirements DSL for the container to run.
 	// This is empty if no specific requirements are needed, or if requirements are
@@ -271,6 +274,13 @@ func getMigMonitorDevices(env map[string]string) *string {
 	return nil
 }
 
+func getImexChannels(env map[string]string) *string {
+	if chans, ok := env[envNVImexChannels]; ok {
+		return &chans
+	}
+	return nil
+}
+
 func (c *HookConfig) getDriverCapabilities(env map[string]string, legacyImage bool) image.DriverCapabilities {
 	// We use the default driver capabilities by default. This is filtered to only include the
 	// supported capabilities
@@ -324,6 +334,11 @@ func getNvidiaConfig(hookConfig *HookConfig, image image.CUDA, mounts []Mount, p
 		log.Panicln("cannot set MIG_MONITOR_DEVICES in non privileged container")
 	}
 
+	var imexChannels string
+	if c := getImexChannels(image); c != nil {
+		imexChannels = *c
+	}
+
 	driverCapabilities := hookConfig.getDriverCapabilities(image, legacyImage).String()
 
 	requirements, err := image.GetRequirements()
@@ -335,6 +350,7 @@ func getNvidiaConfig(hookConfig *HookConfig, image image.CUDA, mounts []Mount, p
 		Devices:            devices,
 		MigConfigDevices:   migConfigDevices,
 		MigMonitorDevices:  migMonitorDevices,
+		ImexChannels:       imexChannels,
 		DriverCapabilities: driverCapabilities,
 		Requirements:       requirements,
 	}

cmd/nvidia-container-runtime-hook/main.go

@@ -126,6 +126,9 @@ func doPrestart() {
 	if len(nvidia.MigMonitorDevices) > 0 {
 		args = append(args, fmt.Sprintf("--mig-monitor=%s", nvidia.MigMonitorDevices))
 	}
+	if len(nvidia.ImexChannels) > 0 {
+		args = append(args, fmt.Sprintf("--imex-channel=%s", nvidia.ImexChannels))
+	}
 
 	for _, cap := range strings.Split(nvidia.DriverCapabilities, ",") {
 		if len(cap) == 0 {

third_party/libnvidia-container

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 LIB_NAME := nvidia-container-toolkit
-LIB_VERSION := 1.14.5
+LIB_VERSION := 1.14.6
 LIB_TAG :=
 
 # The package version is the combination of the library version and tag.