Revert "Merge pull request #694 from elezar/add-opt-in-to-sockets"

This reverts commit b061446694, reversing
changes made to c490baab63.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
This commit is contained in:
Evan Lezar
2024-09-20 20:26:16 +02:00
parent a819cfdab4
commit 5145b0a4b6
12 changed files with 52 additions and 103 deletions

View File

@@ -19,11 +19,10 @@ package config
type featureName string
const (
FeatureGDS = featureName("gds")
FeatureMOFED = featureName("mofed")
FeatureNVSWITCH = featureName("nvswitch")
FeatureGDRCopy = featureName("gdrcopy")
FeatureIncludePersistencedSocket = featureName("include-persistenced-socket")
FeatureGDS = featureName("gds")
FeatureMOFED = featureName("mofed")
FeatureNVSWITCH = featureName("nvswitch")
FeatureGDRCopy = featureName("gdrcopy")
)
// features specifies a set of named features.
@@ -32,57 +31,53 @@ type features struct {
MOFED *feature `toml:"mofed,omitempty"`
NVSWITCH *feature `toml:"nvswitch,omitempty"`
GDRCopy *feature `toml:"gdrcopy,omitempty"`
// IncludePersistencedSocket enables the injection of the nvidia-persistenced
// socket into containers.
IncludePersistencedSocket *feature `toml:"include-persistenced-socket,omitempty"`
}
type feature bool
// IsEnabledInEnvironment checks whether a specified named feature is enabled.
// IsEnabled checks whether a specified named feature is enabled.
// An optional list of environments to check for feature-specific environment
// variables can also be supplied.
func (fs features) IsEnabledInEnvironment(n featureName, in ...getenver) bool {
func (fs features) IsEnabled(n featureName, in ...getenver) bool {
featureEnvvars := map[featureName]string{
FeatureGDS: "NVIDIA_GDS",
FeatureMOFED: "NVIDIA_MOFED",
FeatureNVSWITCH: "NVIDIA_NVSWITCH",
FeatureGDRCopy: "NVIDIA_GDRCOPY",
}
envvar := featureEnvvars[n]
switch n {
// Features with envvar overrides
case FeatureGDS:
return fs.GDS.isEnabledWithEnvvarOverride("NVIDIA_GDS", in...)
return fs.GDS.isEnabled(envvar, in...)
case FeatureMOFED:
return fs.MOFED.isEnabledWithEnvvarOverride("NVIDIA_MOFED", in...)
return fs.MOFED.isEnabled(envvar, in...)
case FeatureNVSWITCH:
return fs.NVSWITCH.isEnabledWithEnvvarOverride("NVIDIA_NVSWITCH", in...)
return fs.NVSWITCH.isEnabled(envvar, in...)
case FeatureGDRCopy:
return fs.GDRCopy.isEnabledWithEnvvarOverride("NVIDIA_GDRCOPY", in...)
// Features without envvar overrides
case FeatureIncludePersistencedSocket:
return fs.IncludePersistencedSocket.IsEnabled()
return fs.GDRCopy.isEnabled(envvar, in...)
default:
return false
}
}
// IsEnabled checks whether a feature is enabled.
func (f *feature) IsEnabled() bool {
if f != nil {
return bool(*f)
}
return false
}
// isEnabledWithEnvvarOverride checks whether a feature is enabled and allows an envvar to overide the feature.
// isEnabled checks whether a feature is enabled.
// If the enabled value is explicitly set, this is returned, otherwise the
// associated envvar is checked in the specified getenver for the string "enabled"
// A CUDA container / image can be passed here.
func (f *feature) isEnabledWithEnvvarOverride(envvar string, ins ...getenver) bool {
if envvar != "" {
for _, in := range ins {
if in.Getenv(envvar) == "enabled" {
return true
}
func (f *feature) isEnabled(envvar string, ins ...getenver) bool {
if f != nil {
return bool(*f)
}
if envvar == "" {
return false
}
for _, in := range ins {
if in.Getenv(envvar) == "enabled" {
return true
}
}
return f.IsEnabled()
return false
}
type getenver interface {

View File

@@ -24,13 +24,7 @@ import (
type ipcMounts mounts
// NewIPCDiscoverer creats a discoverer for NVIDIA IPC sockets.
func NewIPCDiscoverer(logger logger.Interface, driverRoot string, includePersistencedSocket bool) (Discover, error) {
var requiredSockets []string
if includePersistencedSocket {
requiredSockets = append(requiredSockets, "/nvidia-persistenced/socket")
}
requiredSockets = append(requiredSockets, "/nvidia-fabricmanager/socket")
func NewIPCDiscoverer(logger logger.Interface, driverRoot string) (Discover, error) {
sockets := newMounts(
logger,
lookup.NewFileLocator(
@@ -40,7 +34,10 @@ func NewIPCDiscoverer(logger logger.Interface, driverRoot string, includePersist
lookup.WithCount(1),
),
driverRoot,
requiredSockets,
[]string{
"/nvidia-persistenced/socket",
"/nvidia-fabricmanager/socket",
},
)
mps := newMounts(

View File

@@ -189,7 +189,6 @@ func generateAutomaticCDISpec(logger logger.Interface, cfg *config.Config, devic
nvcdi.WithDriverRoot(cfg.NVIDIAContainerCLIConfig.Root),
nvcdi.WithVendor("runtime.nvidia.com"),
nvcdi.WithClass("gpu"),
nvcdi.WithOptInFeature("include-persistenced-socket", cfg.Features.IncludePersistencedSocket.IsEnabled()),
)
if err != nil {
return nil, fmt.Errorf("failed to construct CDI library: %w", err)

View File

@@ -46,7 +46,7 @@ func NewFeatureGatedModifier(logger logger.Interface, cfg *config.Config, image
driverRoot := cfg.NVIDIAContainerCLIConfig.Root
devRoot := cfg.NVIDIAContainerCLIConfig.Root
if cfg.Features.IsEnabledInEnvironment(config.FeatureGDS, image) {
if cfg.Features.IsEnabled(config.FeatureGDS, image) {
d, err := discover.NewGDSDiscoverer(logger, driverRoot, devRoot)
if err != nil {
return nil, fmt.Errorf("failed to construct discoverer for GDS devices: %w", err)
@@ -54,7 +54,7 @@ func NewFeatureGatedModifier(logger logger.Interface, cfg *config.Config, image
discoverers = append(discoverers, d)
}
if cfg.Features.IsEnabledInEnvironment(config.FeatureMOFED, image) {
if cfg.Features.IsEnabled(config.FeatureMOFED, image) {
d, err := discover.NewMOFEDDiscoverer(logger, devRoot)
if err != nil {
return nil, fmt.Errorf("failed to construct discoverer for MOFED devices: %w", err)
@@ -62,7 +62,7 @@ func NewFeatureGatedModifier(logger logger.Interface, cfg *config.Config, image
discoverers = append(discoverers, d)
}
if cfg.Features.IsEnabledInEnvironment(config.FeatureNVSWITCH, image) {
if cfg.Features.IsEnabled(config.FeatureNVSWITCH, image) {
d, err := discover.NewNvSwitchDiscoverer(logger, devRoot)
if err != nil {
return nil, fmt.Errorf("failed to construct discoverer for NVSWITCH devices: %w", err)
@@ -70,7 +70,7 @@ func NewFeatureGatedModifier(logger logger.Interface, cfg *config.Config, image
discoverers = append(discoverers, d)
}
if cfg.Features.IsEnabledInEnvironment(config.FeatureGDRCopy, image) {
if cfg.Features.IsEnabled(config.FeatureGDRCopy, image) {
d, err := discover.NewGDRCopyDiscoverer(logger, devRoot)
if err != nil {
return nil, fmt.Errorf("failed to construct discoverer for GDRCopy devices: %w", err)