Merge branch 'revert-nvlink' into 'master'

Revert support for NVIDIA_FABRIC_DEVICES

See merge request nvidia/container-toolkit/container-toolkit!41

Makefile

@@ -17,8 +17,8 @@ MKDIR    ?= mkdir
 DIST_DIR ?= $(CURDIR)/dist
 
 LIB_NAME := nvidia-container-toolkit
-LIB_VERSION := 1.6.0
-LIB_TAG ?= rc.1
+LIB_VERSION := 1.5.1
+LIB_TAG ?=
 
 GOLANG_VERSION := 1.16.3
 MODULE := github.com/NVIDIA/nvidia-container-toolkit

cmd/nvidia-container-toolkit/container_config.go

@@ -23,7 +23,6 @@ const (
 	envNVVisibleDevices     = "NVIDIA_VISIBLE_DEVICES"
 	envNVMigConfigDevices   = "NVIDIA_MIG_CONFIG_DEVICES"
 	envNVMigMonitorDevices  = "NVIDIA_MIG_MONITOR_DEVICES"
-	envNVFabricDevices      = "NVIDIA_FABRIC_DEVICES"
 	envNVDriverCapabilities = "NVIDIA_DRIVER_CAPABILITIES"
 )
 
@@ -44,7 +43,6 @@ type nvidiaConfig struct {
 	Devices            string
 	MigConfigDevices   string
 	MigMonitorDevices  string
-	FabricDevices      string
 	DriverCapabilities string
 	Requirements       []string
 	DisableRequire     bool
@@ -318,13 +316,6 @@ func getMigMonitorDevices(env map[string]string) *string {
 	return nil
 }
 
-func getFabricDevices(env map[string]string) *string {
-	if devices, ok := env[envNVFabricDevices]; ok {
-		return &devices
-	}
-	return nil
-}
-
 func getDriverCapabilities(env map[string]string, legacyImage bool) *string {
 	// Grab a reference to the capabilities from the envvar
 	// if it actually exists in the environment.
@@ -403,11 +394,6 @@ func getNvidiaConfig(hookConfig *HookConfig, env map[string]string, mounts []Mou
 		driverCapabilities = *c
 	}
 
-	var nvFabricDevices string
-	if d := getFabricDevices(env); d != nil {
-		nvFabricDevices = *d
-	}
-
 	requirements := getRequirements(env, legacyImage)
 
 	// Don't fail on invalid values.
@@ -417,7 +403,6 @@ func getNvidiaConfig(hookConfig *HookConfig, env map[string]string, mounts []Mou
 		Devices:            devices,
 		MigConfigDevices:   migConfigDevices,
 		MigMonitorDevices:  migMonitorDevices,
-		FabricDevices:      nvFabricDevices,
 		DriverCapabilities: driverCapabilities,
 		Requirements:       requirements,
 		DisableRequire:     disableRequire,

cmd/nvidia-container-toolkit/container_test.go

@@ -403,30 +403,6 @@ func TestGetNvidiaConfig(t *testing.T) {
 			privileged:    false,
 			expectedPanic: true,
 		},
-		{
-			description: "fabric devices selected",
-			env: map[string]string{
-				envNVVisibleDevices: "all",
-				envNVFabricDevices:  "all",
-			},
-			expectedConfig: &nvidiaConfig{
-				Devices:            "all",
-				FabricDevices:      "all",
-				DriverCapabilities: defaultDriverCapabilities,
-			},
-		},
-		{
-			description: "fabric devices selected empty",
-			env: map[string]string{
-				envNVVisibleDevices: "all",
-				envNVFabricDevices:  "",
-			},
-			expectedConfig: &nvidiaConfig{
-				Devices:            "all",
-				FabricDevices:      "",
-				DriverCapabilities: defaultDriverCapabilities,
-			},
-		},
 	}
 	for _, tc := range tests {
 		t.Run(tc.description, func(t *testing.T) {

cmd/nvidia-container-toolkit/main.go

@@ -132,9 +132,6 @@ func doPrestart() {
 	if len(nvidia.MigMonitorDevices) > 0 {
 		args = append(args, fmt.Sprintf("--mig-monitor=%s", nvidia.MigMonitorDevices))
 	}
-	if len(nvidia.FabricDevices) > 0 {
-		args = append(args, fmt.Sprintf("--fabric-device=%s", nvidia.FabricDevices))
-	}
 
 	for _, cap := range strings.Split(nvidia.DriverCapabilities, ",") {
 		if len(cap) == 0 {

packaging/debian/changelog

@@ -1,9 +1,3 @@
-nvidia-container-toolkit (1.6.0~rc.1-1) UNRELEASED; urgency=medium
-
-  * Add support for NVIDIA_FABRIC_DEVICES to select nvlink and nvswitch devices
-
- -- NVIDIA CORPORATION <cudatools@nvidia.com>  Fri, 13 Aug 2021 09:00:00 -0700
-
 nvidia-container-toolkit (1.5.1-1) UNRELEASED; urgency=medium
 
   * Fix bug where Docker Swarm device selection is ignored if

packaging/debian/control

@@ -10,7 +10,7 @@ Build-Depends: debhelper (>= 9)
 
 Package: nvidia-container-toolkit
 Architecture: any
-Depends: ${misc:Depends}, libnvidia-container-tools (>= 1.5.0~rc.1), libnvidia-container-tools (<< 2.0.0)
+Depends: ${misc:Depends}, libnvidia-container-tools (>= 1.4.0), libnvidia-container-tools (<< 2.0.0)
 Breaks: nvidia-container-runtime (<< 2.0.0), nvidia-container-runtime-hook
 Replaces: nvidia-container-runtime (<< 2.0.0), nvidia-container-runtime-hook
 Description: NVIDIA container runtime hook

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

@@ -18,7 +18,7 @@ Source4: LICENSE
 
 Obsoletes: nvidia-container-runtime < 2.0.0, nvidia-container-runtime-hook
 Provides: nvidia-container-runtime-hook
-Requires: libnvidia-container-tools >= 1.5.0-0.1.rc.1, libnvidia-container-tools < 2.0.0
+Requires: libnvidia-container-tools >= 1.4.0, libnvidia-container-tools < 2.0.0
 
 %description
 Provides a OCI hook to enable GPU support in containers.
@@ -53,10 +53,8 @@ rm -f %{_bindir}/nvidia-container-runtime-hook
 /usr/share/containers/oci/hooks.d/oci-nvidia-hook.json
 
 %changelog
-* Fri Aug 13 2021 NVIDIA CORPORATION <cudatools@nvidia.com> 1.6.0-0.1.rc.1
-- Add support for NVIDIA_FABRIC_DEVICES to select nvlink and nvswitch devices
-
 * Mon Jun 14 2021 NVIDIA CORPORATION <cudatools@nvidia.com> 1.5.1-1
+
 - Fix bug where Docker Swarm device selection is ignored if NVIDIA_VISIBLE_DEVICES is also set
 - Improve unit testing by using require package and adding coverage reports
 - Remove unneeded go dependencies by running go mod tidy