Tegra libraries are missing on Nvidia L4Tbase docker Image

[florianC31]

Hello,

I am currently working on the Jetson Nano Developer Kit 2GB with docker and a ZED2i camera from Stereolabs.
I’m using the Nvidia Docker base image “nvcr.io/nvidia/l4t-base:r32.7.1”.
When I try to compile my C++ project using ZED SDK, I need several libraries which should be present in/usr/lib/aarch64-linux-gnu/tegra. However, the entire tegra folder is missing in this docker image.

When I check on the Jetson system (which has been flashed with the Jetpack 4.6.1), the /usr/lib/aarch64-linux-gnu/tegra folder exists with all expected libraries.

As a temporary solution, I copied all this folder from the Jetson to my Docker container and it works properly.

However, I would want to know if it’s normal that this folder is missing, and, in this case, how install all these Tegra libraries properly.

Tank you for your help.

[dusty_nv]

Hi @florianC31, you can check which libraries from /usr/lib/aarch64-linux-gnu/tegra/ that should be mounted into l4t-base container (and containers that derive from l4t-base) when --runtime nvidia is used by checking the file /etc/nvidia-container-runtime/host-files-for-container.d.csv

Are you starting l4t-base with --runtime nvidia ? Or if you are doing a docker build operation, have you set your default docker runtime to nvidia like shown here: https://github.com/dusty-nv/jetson-containers#docker-default-runtime

[florianC31]

You’re right, the issue came from missing “–runtime nvidia”.

I already added this parameter in my Docker Compose a few weeks ago, but yesterday I create an intermediary compilation docker and I forgot to add it…
And when I made some tests directly on Nvidia images I forgot too…

So, I added in default parameters to not be stuck any more in the future.
Thanks for the tips !

[dusty_nv]

OK no worries, glad that you got it figured out!