If I run test container without --volume, then /mnt directory ownership is root:root. So volume ownership is not set to the one on the directory under which it is mounted. Rrgrds, On 22.09.2021 15:41, Daniel Walsh wrote:

On 9/22/21 03:04, Michael Ivanov wrote:

...

Greetings,

I observe the following strange behavior as regarding volume ownership in rootless mode. I have user oracle with uid 502 and group oinstall with gid 501 both on host system and in my container.

I create a volume and change it's ownership as follows:

    podman volume create data     podman unshare chown 502:501 ~/.local/share/containers/storage/volumes/rdata/_data     ls -ld ~/.local/share/containers/storage/volumes/rdata/_data

    drwxr-xr-x 1 200501 200500 0 мар 14  2021 /home/ivans/.local/share/containers/storage/volumes/rdata/_data/

So far so good. I run my test container:

    podman run --name test --detach --volume rdata:/mnt test

And check the volume ownership inside the container:

    podman exec -it test ls -ld /mnt

The owner reported for /mnt is the one configured with --user, when container was built, not oracle:oinstall!

I stop the container and check volume owner. It has changed:

    ls -ld ~/.local/share/containers/storage/volumes/rdata/_data

    drwxr-xr-x 1 201000 200999 0 мар 14  2021 /home/ivans/.local/share/containers/storage/volumes/phsdata/_data/

I change volume owner again:

    podman unshare chown 502:501 ~/.local/share/containers/storage/volumes/rdata/_data

and run the container using same command:

    podman run --name test --detach --volume rdata:/mnt test

The second time everything is correct:

    podman exec -it test ls -ld /mnt

    drwxr-xr-x 1 oracle oinstall 0 Mar 14  2021 /mnt

And same outside of container:

    drwxr-xr-x 1 200501 200500 0 мар 14  2021 /home/ivans/.local/share/containers/storage/volumes/rdata/_data/

If I remove the volume and create it again, then the ownership is again changed to default container user. So expected ownership is set only after second mount of the volume.

What might be wrong here?

Best regards,

The first time a volume is used with chown the file to the destination directorys permissions if the destination exists, otherwise we chown it to the UID of the default user of the image.

We do this with unnamed volumes, but maybe we are doing it also with named volumes which I think would be a bug.

Especially if that volume was used in a different container with a different UID. _______________________________________________ Podman mailing list -- podman@lists.podman.io To unsubscribe send an email to podman-leave@lists.podman.io

-- \ / | | (OvO) | Михаил Иванов | (^^^) | | \^/ | E-mail: ivans@isle.spb.ru | ^ ^ | |