UniqAI/nvidia-container-toolkit
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvidia-container-toolkit synced 2024-11-25 21:39:10 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Check requirements before creating CSV discoverer

Browse Source 
Signed-off-by: Evan Lezar <elezar@nvidia.com>
This commit is contained in:
Evan Lezar 2022-05-02 10:28:25 +02:00
parent 583793b7ae
commit 49f4bb3198
2 changed files with 42 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
cmd/nvidia-container-runtime/modifier/experimental.go
View File

@ -22,10 +22,13 @@ import (
"strings"
"github.com/NVIDIA/nvidia-container-toolkit/internal/config"
"github.com/NVIDIA/nvidia-container-toolkit/internal/config/image"
"github.com/NVIDIA/nvidia-container-toolkit/internal/cuda"
"github.com/NVIDIA/nvidia-container-toolkit/internal/discover"
"github.com/NVIDIA/nvidia-container-toolkit/internal/discover/csv"
"github.com/NVIDIA/nvidia-container-toolkit/internal/edits"
"github.com/NVIDIA/nvidia-container-toolkit/internal/oci"
"github.com/NVIDIA/nvidia-container-toolkit/internal/requirements"
"github.com/opencontainers/runtime-spec/specs-go"
"github.com/sirupsen/logrus"
)
@ -46,7 +49,7 @@ const (
// NewExperimentalModifier creates a modifier that applies the experimental
// modications to an OCI spec if required by the runtime wrapper.
func NewExperimentalModifier(logger *logrus.Logger, cfg *config.Config, ociSpec oci.Spec) (oci.SpecModifier, error) {
_, err := ociSpec.Load()
rawSpec, err := ociSpec.Load()
if err != nil {
return nil, fmt.Errorf("failed to load OCI spec: %v", err)
}
@ -75,6 +78,17 @@ func NewExperimentalModifier(logger *logrus.Logger, cfg *config.Config, ociSpec
}
d = legacyDiscoverer
case "csv":
// TODO: Once the devices have been encapsulated in the CUDA image, this can be moved to before the
// visible devices are checked.
image, err := image.NewCUDAImageFromSpec(rawSpec)
if err != nil {
return nil, err
}
if err := checkRequirements(logger, &image); err != nil {
return nil, fmt.Errorf("requirements not met: %v", err)
}
csvFiles, err := csv.GetFileList(csv.DefaultMountSpecPath)
if err != nil {
return nil, fmt.Errorf("failed to get list of CSV files: %v", err)
@ -134,6 +148,31 @@ func (m experimental) Modify(spec *specs.Spec) error {
return specEdits.Modify(spec)
}
func checkRequirements(logger *logrus.Logger, image *image.CUDA) error {
if image.HasDisableRequire() {
// TODO: We could print the real value here instead
logger.Debugf("NVIDIA_DISABLE_REQUIRE=%v; skipping requirement checks", true)
return nil
}
imageRequirements, err := image.GetRequirements()
if err != nil {
// TODO: Should we treat this as a failure, or just issue a warning?
return fmt.Errorf("failed to get image requirements: %v", err)
}
r := requirements.New(logger, imageRequirements)
cudaVersion, err := cuda.Version()
if err != nil {
logger.Warnf("Failed to get CUDA version: %v", err)
} else {
r.AddVersionProperty(requirements.CUDA, cudaVersion)
}
return r.Assert()
}
// resolveAutoDiscoverMode determines the correct discover mode for the specified platform if set to "auto"
func resolveAutoDiscoverMode(logger *logrus.Logger, mode string) (rmode string) {
if mode != "auto" {

4
cmd/nvidia-container-runtime/modifier/experimental_test.go
View File

@ -42,8 +42,8 @@ func TestNewExperimentalModifier(t *testing.T) {
{
description: "spec load error returns error",
spec: &oci.SpecMock{
LoadFunc: func() error {
return fmt.Errorf("load failed")
LoadFunc: func() (*specs.Spec, error) {
return nil, fmt.Errorf("load failed")
},
},
expectedError: fmt.Errorf("load failed"),