regression after update to podman v5

[LE0N7]

Hi all,

RHEL9.5 is GA and with this release also podman and friends got updates.

podman-4.9.4 -> 5.2.2

Since then, I have a couple of rootless containers
that have problems with their volumes.

The minimal example is as follows:

$ cat Containerfile
FROM fedora:41
USER 1000:1000
ENV HOME=/data
WORKDIR  /data
VOLUME ["/data"]

So, when started, the volume (as specified in the file) will
be created automatically.

With podmam 4.9

podman run --rm -ti  --userns=keep-id:uid=1000,gid=1000 \
  localhost/test:latest ls -la /data
drwxr-xr-x. 2 1000 1000 4096 Nov 20 12:44 .
dr-xr-xr-x. 1 root root 4096 Nov 20 12:44 ..

shows that the volume is mounted in the container with
the USER id. This works so far for my workload.

With podmam 5.2

podman run --rm -ti --userns=keep-id:uid=1000,gid=1000 \
  localhost/test:latest ls -la /data
drwxr-xr-x. 2  999  999 4096 Nov 20 12:11 .
dr-xr-xr-x. 1 root root 4096 Nov 20 12:46 ..

shows that the volume is mounted in the container with
(USER id - 1). The user can not write to the folder and
the container fails to operate.

Forcing to use a named volume for overlay podman v5 shows
the same output but when the destination is changed to
a different directory (/side), it gets the right owner set:

podman run --rm -ti --userns=keep-id:uid=1000,gid=1000 \
 -v ${RANDOM}:/side localhost:8085/shee/test:latest ls -la /side
drwxr-xr-x. 2 1000 1000 4096 Nov 20 12:51 .
dr-xr-xr-x. 1 root root 4096 Nov 20 12:51 ..

I read the changelogs but didn't find any hints for a solution.

I would appreciate any feedback, thanks!

[LE0N7]

As usual, you will only find a hint after your own question.

#24577

Seems to be resolved in 5.2.3

https://github.com/containers/podman/releases/tag/v5.2.3

Fixed a bug where volume copy-up did not properly chown empty volumes