Also change directory when not running pivot root

When passing the no-pivot option to the CLI, the ldcache in the container is not properly updated.

This change ensures that the rootfs of the container is used as the root for running ldconfig.

$ cat /etc/nvidia-container-runtime/config.toml | grep pivot
no-pivot = true

Before the fix:

$ docker run --rm --runtime nvidia nvidia/cuda:11.0-base nvidia-smi -L
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.

After:

$ docker run --rm --runtime nvidia nvidia/cuda:11.0-base nvidia-smi -L
GPU 0: Tesla V100-SXM2-16GB-N (UUID: GPU-edfee158-11c1-52b8-0517-92f30e7fac88)
GPU 1: Tesla V100-SXM2-16GB-N (UUID: GPU-f22fb098-d1b3-3806-2655-ba25f02229c1)
GPU 2: Tesla V100-SXM2-16GB-N (UUID: GPU-f613f823-1032-b3ec-a876-50f2e35e6f9e)
GPU 3: Tesla V100-SXM2-16GB-N (UUID: GPU-3109fa37-4445-73c7-b695-1b5a4d13f58e)
GPU 4: Tesla V100-SXM2-16GB-N (UUID: GPU-e28a6529-288c-7ddf-8fea-68c4833cda70)
GPU 5: Tesla V100-SXM2-16GB-N (UUID: GPU-a27fb382-bad2-c02a-95ba-f6a1da38e76c)
GPU 6: Tesla V100-SXM2-16GB-N (UUID: GPU-f5bb8d07-ee19-1787-4d9a-a84c4ac6b086)
GPU 7: Tesla V100-SXM2-16GB-N (UUID: GPU-1ba0ca0e-6d1d-d9db-07d8-c1c5a8c32814)

Signed-off-by: Evan Lezar elezar@nvidia.com