Prior to allowing users without root privileges to run Podman, the administrator must install or build Podman and complete the following configurations.
For installing Podman, see the installation instructions.
For building Podman, see the build instructions.
The slirp4netns package provides user-mode networking for unprivileged network namespaces and must be installed on the machine in order for Podman to run in a rootless environment. The package is available on most Linux distributions via their package distribution software such as yum, dnf, apt, zypper, etc. If the package is not available, you can build and install slirp4netns from GitHub.
Rootless Podman requires the user running it to have a range of UIDs listed in the files /etc/subuid and /etc/subgid. The shadow-utils or newuid package provides these files on different distributions and they must be installed on the system. Root privileges are required to add or update entries within these files. The following is a summary from the How does rootless Podman work? article by Dan Walsh on opensource.com
For each user that will be allowed to create containers, update /etc/subuid and /etc/subgid for the user with fields that look like the following. Note that the values for each user must be unique. If there is overlap, there is a potential for a user to use another user's namespace and they could corrupt it.
# cat /etc/subuid
johndoe:100000:65536
test:165536:65536
The format of this file is USERNAME:UID:RANGE
- username as listed in
/etc/passwdor in the output ofgetpwent. - The initial UID allocated for the user.
- The size of the range of UIDs allocated for the user.
This means the user johndoe is allocated UIDs 100000-165535 as well as their standard UID in the /etc/passwd file. NOTE: this is not currently supported with network installs; these files must be available locally to the host machine. It is not possible to configure this with LDAP or Active Directory.
Rather than updating the files directly, the usermod program can be used to assign UIDs and GIDs to a user.
# usermod --add-subuids 100000-165535 --add-subgids 100000-165535 johndoe
grep johndoe /etc/subuid /etc/subgid
/etc/subuid:johndoe:100000:65536
/etc/subgid:johndoe:100000:65536
If you update either /etc/subuid or /etc/subgid, you need to stop all the running containers owned by the user and kill the pause process that is running on the system for that user. This can be done automatically by running podman system migrate as that user.
Users can fully map additional groups to a container namespace if those groups subordinated to the user:
# usermod --add-subgids 2000-2000 johndoe
grep johndoe /etc/subgid
This means the user johndoe can "impersonate" the group 2000 inside the
container. Note that it is usually not a good idea to subordinate active
user ids to other users, because it would allow user impersonation.
johndoe can use --group-add keep-groups to preserve the additional
groups, and --gidmap="+g102000:@2000" to map the group 2000 in the host
to the group 102000 in the container:
$ podman run \
--rm \
--group-add keep-groups \
--gidmap="+g102000:@2000" \
--volume "$PWD:/data:ro" \
--workdir /data \
alpine ls -lisa
(It is very unlikely that you will need to do this on a modern distro).
Users running in a non-privileged container may not be able to use the ping utility from that container.
If this is required, the administrator must verify that the UID of the user is part of the range in the /proc/sys/net/ipv4/ping_group_range file.
To change its value the administrator can use a call similar to: sysctl -w "net.ipv4.ping_group_range=0 2000000".
To make the change persist, the administrator will need to add a file with the .conf file extension in /etc/sysctl.d that contains net.ipv4.ping_group_range=0 $MAX_GID, where $MAX_GID is the highest assignable GID of the user running the container.
The majority of the work necessary to run Podman in a rootless environment is on the shoulders of the machine’s administrator.
Once the Administrator has completed the setup on the machine and then the configurations for the user in /etc/subuid and /etc/subgid, the user can just start using any Podman command that they wish.
The Podman configuration files for root reside in /usr/share/containers with overrides in /etc/containers. In the rootless environment they reside in ${XDG_CONFIG_HOME}/containers and are owned by each individual user.
Note: in environments without XDG environment variables, Podman internally sets the following defaults:
$XDG_CONFIG_HOME=$HOME/.config$XDG_DATA_HOME=$HOME/.local/share$XDG_RUNTIME_DIR=/run/user/$UIDonsystemdenvironments$TMPDIR/podman-run-$UIDotherwise
The three main configuration files are containers.conf, storage.conf and registries.conf. The user can modify these files as they wish.
Podman reads
/usr/share/containers/containers.conf/etc/containers/containers.conf${XDG_CONFIG_HOME}/containers/containers.conf
if they exist, in that order. Each file can override the previous for particular fields.
For storage.conf the order is
/etc/containers/storage.conf${XDG_CONFIG_HOME}/containers/storage.conf
In rootless Podman, certain fields in /etc/containers/storage.conf are ignored. These fields are:
graphroot=""
container storage graph dir (default: "/var/lib/containers/storage")
Default directory to store all writable content created by container storage programs.
runroot=""
container storage run dir (default: "/run/containers/storage")
Default directory to store all temporary writable content created by container storage programs.
In rootless Podman these fields default to
graphroot="${XDG_DATA_HOME}/containers/storage"
runroot="${XDG_RUNTIME_DIR}/containers"
$XDG_RUNTIME_DIR defaults on most systems to /run/user/$UID.
Registry configuration is read in this order
/etc/containers/registries.conf/etc/containers/registries.d/*${XDG_CONFIG_HOME}/containers/registries.conf
The files in the home directory should be used to configure rootless Podman for personal needs. These files are not created by default. Users can copy the files from /usr/share/containers or /etc/containers and modify them.
The default authorization file used by the podman login and podman logout commands is ${XDG_RUNTIME_DIR}/containers/auth.json.
Rootless Podman is not, and will never be, root; it's not a setuid binary, and gains no privileges when it runs. Instead, Podman makes use of a user namespace to shift the UIDs and GIDs of a block of users it is given access to on the host (via the newuidmap and newgidmap executables) and your own user within the containers that Podman creates.
If your container runs with the root user, then root in the container is actually your user on the host. UID/GID 1 is the first UID/GID specified in your user's mapping in /etc/subuid and /etc/subgid, etc. If you mount a directory from the host into a container as a rootless user, and create a file in that directory as root in the container, you'll see it's actually owned by your user on the host.
So, for example,
host$ whoami
john
# a folder which is empty
host$ ls /home/john/folder
host$ podman run -it -v /home/john/folder:/container/volume mycontainer /bin/bash
# Now I'm in the container
root@container# whoami
root
root@container# touch /container/volume/test
root@container# ls -l /container/volume
total 0
-rw-r--r-- 1 root root 0 May 20 21:47 test
root@container# exit
# I check again
host$ ls -l /home/john/folder
total 0
-rw-r--r-- 1 john john 0 May 20 21:47 test
We do recognize that this doesn't really match how many people intend to use rootless Podman - they want their UID inside and outside the container to match. Thus, we provide the --userns=keep-id flag, which ensures that your user is mapped to its own UID and GID inside the container.
It is also helpful to distinguish between running Podman as a rootless user, and a container which is built to run rootless. If the container you're trying to run has a USER which is not root, then when mounting volumes you must use --userns=keep-id. This is because the container user would not be able to become root and access the mounted volumes.
Another consideration in regards to volumes:
- When providing the path of a directory you'd like to bind-mount, the path needs to be provided as an absolute path
or a relative path that starts with
.(a dot), otherwise the string will be interpreted as the name of a named volume.
If you are still experiencing problems running Podman in a rootless environment, please refer to the Shortcomings of Rootless Podman page which lists known issues and solutions to known issues in this environment.
For more information on Podman and its subcommands, follow the links on the main README.md page or the podman.io web site.