UniqAI/nvidia-container-toolkit
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvidia-container-toolkit synced 2024-11-25 13:35:00 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Use devRoot to resolve MIG device nodes

Browse Source 
Signed-off-by: Evan Lezar <elezar@nvidia.com>
This commit is contained in:
Evan Lezar 2023-12-15 13:36:38 +01:00
parent 9396858834
commit 21fc1f24e4
2 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
CHANGELOG.md
View File

@ -6,6 +6,7 @@
* Log explicitly requested runtime mode. * Log explicitly requested runtime mode.
* Remove package dependency on libseccomp. * Remove package dependency on libseccomp.
* Added detection of libnvdxgdmal.so.1 on WSL2 * Added detection of libnvdxgdmal.so.1 on WSL2
* Use devRoot to resolve MIG device nodes.
## v1.15.0-rc.1 ## v1.15.0-rc.1
* Skip update of ldcache in containers without ldconfig. The .so.SONAME symlinks are still created. * Skip update of ldcache in containers without ldconfig. The .so.SONAME symlinks are still created.

10
pkg/nvcdi/mig-device-nvml.go
View File

@ -67,7 +67,7 @@ func (l *nvmllib) GetMIGDeviceEdits(parent device.Device, mig device.MigDevice)
return nil, fmt.Errorf("error getting Compute Instance ID: %v", ret) return nil, fmt.Errorf("error getting Compute Instance ID: %v", ret)
} }
editsForDevice, err := GetEditsForComputeInstance(l.logger, l.driverRoot, gpu, gi, ci) editsForDevice, err := l.GetEditsForComputeInstance(gpu, gi, ci)
if err != nil { if err != nil {
return nil, fmt.Errorf("failed to create container edits for MIG device: %v", err) return nil, fmt.Errorf("failed to create container edits for MIG device: %v", err)
} }
@ -76,8 +76,8 @@ func (l *nvmllib) GetMIGDeviceEdits(parent device.Device, mig device.MigDevice)
} }
// GetEditsForComputeInstance returns the CDI edits for a particular compute instance defined by the (gpu, gi, ci) tuple // GetEditsForComputeInstance returns the CDI edits for a particular compute instance defined by the (gpu, gi, ci) tuple
func GetEditsForComputeInstance(logger logger.Interface, driverRoot string, gpu int, gi int, ci int) (*cdi.ContainerEdits, error) { func (l *nvmllib) GetEditsForComputeInstance(gpu int, gi int, ci int) (*cdi.ContainerEdits, error) {
computeInstance, err := newComputeInstanceDiscoverer(logger, driverRoot, gpu, gi, ci) computeInstance, err := newComputeInstanceDiscoverer(l.logger, l.devRoot, gpu, gi, ci)
if err != nil { if err != nil {
return nil, fmt.Errorf("failed to create discoverer for Compute Instance: %v", err) return nil, fmt.Errorf("failed to create discoverer for Compute Instance: %v", err)
} }
@ -91,7 +91,7 @@ func GetEditsForComputeInstance(logger logger.Interface, driverRoot string, gpu
} }
// newComputeInstanceDiscoverer returns a discoverer for the specified compute instance // newComputeInstanceDiscoverer returns a discoverer for the specified compute instance
func newComputeInstanceDiscoverer(logger logger.Interface, driverRoot string, gpu int, gi int, ci int) (discover.Discover, error) { func newComputeInstanceDiscoverer(logger logger.Interface, devRoot string, gpu int, gi int, ci int) (discover.Discover, error) {
parentPath := fmt.Sprintf("/dev/nvidia%d", gpu) parentPath := fmt.Sprintf("/dev/nvidia%d", gpu)
migCaps, err := nvcaps.NewMigCaps() migCaps, err := nvcaps.NewMigCaps()
@ -113,7 +113,7 @@ func newComputeInstanceDiscoverer(logger logger.Interface, driverRoot string, gp
deviceNodes := discover.NewCharDeviceDiscoverer( deviceNodes := discover.NewCharDeviceDiscoverer(
logger, logger,
driverRoot, devRoot,
[]string{ []string{
parentPath, parentPath,
giCapDevicePath, giCapDevicePath,