Why this value is `@/sbin/ldconfig`, rather than `/sbin/ldconfig`.

[lengrongfu]

• kubectl -n gpu-operator logs -f nvidia-operator-validator-mjzpl -c toolkit-validation

  time="2024-06-20T01:20:55Z" level=info msg="version: 762213f2"
  NVIDIA-SMI couldn't find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system.
  Please also try adding directory that contains libnvidia-ml.so to your system PATH.
  toolkit is not ready
  time="2024-06-20T01:20:55Z" level=info msg="Error: error validating toolkit installation: exit status 12"

Os: ubuntu22.04 Kernel: 5.15.0-72-generic GPU Driver: Host install, 535.54.03

I use gpu-operator install nvidia-container-toolkit, but found nvidia-operator-validator this pod is error, error info is couldn't find libnvidia-ml.so.

I found out through some information that it might be a problem with @/sbin/ldconfig, and found that on the host, /sbin/ldconfig is not a soft link file. and then I change /sbin/ldconfig file to soft link to /sbin/ldconfig.real, nvidia-operator-validator this pod can success running.

my question is why this value is @/sbin/ldconfig, can we optimize this code and first determine whether the file is a soft link?

https://github.com/NVIDIA/nvidia-container-toolkit/blob/6b78c72fec07f4f28b9a5f1072ec86af36fb75bb/internal/config/config.go#L131

[lengrongfu]

This function we should change this to check the file is symlink mode.

func getLdConfigPath() string {
    filePath := "/sbin/ldconfig"
    fileInfo, err := os.Lstat(filePath)
    if err != nil || fileInfo.Mode()&os.ModeSymlink != 0{
        filePath = "@/sbin/ldconfig"
    }
    return NormalizeLDConfigPath(filePath)
}