UniqAI/nvidia-container-toolkit
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvidia-container-toolkit synced 2025-06-26 18:18:24 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Create device nodes in JIT-CDI mode
Some checks failed
CI Pipeline / code-scanning (push) Has been cancelled
Details
CI Pipeline / variables (push) Has been cancelled
Details
CI Pipeline / golang (push) Has been cancelled
Details
CI Pipeline / image (push) Has been cancelled
Details
CI Pipeline / e2e-test (push) Has been cancelled
Details

Browse Source 
Signed-off-by: Evan Lezar <elezar@nvidia.com>
This commit is contained in:
Evan Lezar
2025-03-09 13:49:56 +02:00
parent 76b6d4d38f
commit 1cfaef4b01
11 changed files with 82 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
internal/modifier/cdi.go
View File

@@ -22,12 +22,15 @@ import (
"tags.cncf.io/container-device-interface/pkg/parser"
"github.com/NVIDIA/go-nvlib/pkg/nvlib/device"
"github.com/NVIDIA/nvidia-container-toolkit/internal/config"
"github.com/NVIDIA/nvidia-container-toolkit/internal/config/image"
"github.com/NVIDIA/nvidia-container-toolkit/internal/logger"
"github.com/NVIDIA/nvidia-container-toolkit/internal/lookup/root"
"github.com/NVIDIA/nvidia-container-toolkit/internal/modifier/cdi"
"github.com/NVIDIA/nvidia-container-toolkit/internal/oci"
"github.com/NVIDIA/nvidia-container-toolkit/internal/system/nvdevices"
"github.com/NVIDIA/nvidia-container-toolkit/pkg/nvcdi"
"github.com/NVIDIA/nvidia-container-toolkit/pkg/nvcdi/spec"
)
@@ -198,12 +201,14 @@ func generateAutomaticCDISpec(logger logger.Interface, cfg *config.Config, drive
logger.Warningf("Ignoring error(s) loading kernel modules: %v", err)
}
identifiers := []string{}
var identifiers []string
for _, device := range devices {
_, _, id := parser.ParseDevice(device)
identifiers = append(identifiers, id)
}
tryCreateDeviceNodes(logger, driver, identifiers...)
deviceSpecs, err := cdilib.GetDeviceSpecsByID(identifiers...)
if err != nil {
return nil, fmt.Errorf("failed to get CDI device specs: %w", err)
@@ -221,3 +226,27 @@ func generateAutomaticCDISpec(logger logger.Interface, cfg *config.Config, drive
spec.WithClass("gpu"),
)
}
func tryCreateDeviceNodes(logger logger.Interface, driver *root.Driver, identifiers ...string) {
devices, err := nvdevices.New(
nvdevices.WithLogger(logger),
nvdevices.WithDevRoot(driver.Root),
)
if err != nil {
logger.Warningf("Failed to create devices library: %v", err)
return
}
if err := devices.CreateNVIDIAControlDevices(); err != nil {
logger.Warningf("Failed to create control devices: %v", err)
}
if err := devices.CreateNVIDIACapsControlDeviceNodes(); err != nil {
logger.Warningf("Failed to create nvidia-caps control devices: %v", err)
}
for _, id := range identifiers {
identifier := device.Identifier(id)
if err := devices.CreateDeviceNodes(identifier); err != nil {
logger.Warningf("Error creating device nodes for %v: %v", identifier, err)
}
}
}