'Read-only file system' Error in Docker Containers

Hi all,

There is a file system problem with the new base image provided by Nvidia which is “l4t-base”. When I start a container with the flow as shown below:

xhost +local:docker
XSOCK=/tmp/.X11-unix
XAUTH=/tmp/.docker.xauth
xauth nlist $DISPLAY | sed -e 's/^..../ffff/' | xauth -f $XAUTH nmerge -

and then Docker command:

docker run --runtime nvidia -it --device=/dev/video0 -e DISPLAY=$DISPLAY -v $XSOCK:$XSOCK -v $XAUTH:$XAUTH -e XAUTHORITY=$XAUTH nvcr.io/nvidia/l4t-base:r32.2

The problem is that when I try to edit the “/usr/local/cuda/include/cuda_gl_interop.h” file to build opencv, it throws “cp: cannot create regular file ‘/usr/local/cuda/include/cuda_gl_interop.h.bak’: Read-only file system” or “File ‘/usr/local/cuda/include/cuda_gl_interop.h’ is unwritable”.

I can’t change the permissions of this file within the container either. Is there any workaround for this problem?

Hi,

There is a related issue in the topic 1061539:
[url]nvidia-docker seems unable to use GPU as non-root user - Jetson TX2 - NVIDIA Developer Forums

We have passed this issue to our internal team. Will keep you updated once we have any further conclusion.
Thanks.

Hi, could you get any returns from the team? I can’t move on with my development right now.

Can you mount an external volume on /usr/local into the docker container, and make that mount read/write?

Still, no luck on my side. Did you try your solution and it worked?

Hi, doruk898

Sorry for keeping you waiting.

We have checked this issue with our internal team.
However, due to the limited resource, our team still need some time to have a feedback.

Will keep you updated and thanks for your patience.
Thanks.

OK, thanks for looking into it. I also will be keep digging to come up with a better workaround (or actual solution maybe) too.

Hi,

We found some firmware related issue on our docker image.
However, we haven’t verified that if your issues is caused by the same source or not.

Will keep you updated with the progress.
Thanks for your patience.

Please check the following comment for the suggestion:
[url]https://devtalk.nvidia.com/default/topic/1062955/jetson-nano/unable-to-use-tensorrt-inside-the-l4t-base-container/post/5391772/#5391772[/url]

Thanks.

Hi AastaLLL,

Thanks for the feedback. In case of someone comes to this topic, it is updated here: https://devtalk.nvidia.com/default/topic/1062955/jetson-nano/unable-to-use-tensorrt-inside-the-l4t-base-container/post/5395502/#5395502