No loop device in container even with loop module loaded

Question

I am trying to make it possible to make and mount loop devices from within a container. This happens to work on my own development system, but is failing to work on our build server where it must be done for an automated build.

I am ensuring that I'm starting the container as privileged. My container start line: docker run --privileged -it --rm <my_build_image>:latest /bin/bash. From within the container I try the following steps from the losetup man page:

# dd if=/dev/zero of=/var/tmp/file.img bs=1024k count=4
...
# losetup --show --find /var/tmp/file.img
...

This should provide me the next unused loop device and have associated it to /dev/loopn. However, instead I'm presented with the following (and showing that the loop module is loaded and /dev/loop-control is present):

[root@64a3a6900e0d /]# losetup --show --find /var/tmp/file.img
losetup: Could not find any loop device. Maybe this kernel does not know
       about the loop device? (If so, recompile or `modprobe loop`.)
[root@64a3a6900e0d /]# ls /dev/loop*
/dev/loop-control
[root@64a3a6900e0d /]# lsmod | grep loop
loop                   28072  0

On my own dev box, this works. I loaded loop and started the container as privileged and was able to make loop devices. What should I check for now?

Answer 1

I have a solution to the problem. Some of the issue is that, although I'm not a docker neophyte, I am not a well seasoned user of docker yet. I think that when the container is spawned the /dev directory that existed at that point is somehow mapped into the container. I thought it was a "real-time" depiction of the /dev directory but this is not the case. If there were no /dev/loop# files none were permitted to exist even if the container was privileged.

So, the solution is that in the Jenkins pipeline, before the container is spawned, there is a stage which arranges for a /dev/loop# file to be made. Then, the container is spawned and a /dev/loop# is available.