diff options
| -rw-r--r-- | docs/podman-system-migrate.1.md | 15 | ||||
| -rw-r--r-- | docs/tutorials/rootless_tutorial.md | 85 | ||||
| -rw-r--r-- | libpod/boltdb_state.go | 70 | ||||
| -rw-r--r-- | pkg/rootless/rootless_linux.c | 3 |
4 files changed, 139 insertions, 34 deletions
diff --git a/docs/podman-system-migrate.1.md b/docs/podman-system-migrate.1.md index 48f3bca8f..1efa779ce 100644 --- a/docs/podman-system-migrate.1.md +++ b/docs/podman-system-migrate.1.md @@ -11,11 +11,24 @@ podman\-system\-migrate - Migrate container to the latest version of podman **podman system migrate** takes care of migrating existing containers to the latest version of podman if any change is necessary. +"Rootless Podman uses a pause process to keep the unprivileged +namespaces alive. This prevents any change to the `/etc/subuid` and +`/etc/subgid` files from being propagated to the rootless containers +while the pause process is running. + +For these changes to be propagated, it is necessary to first stop all +running containers associated with the user and to also stop the pause +process and delete its pid file. Instead of doing it manually, `podman +system migrate` can be used to stop both the running containers and the +pause process. The `/etc/subuid` and `/etc/subgid` files can then be +edited or changed with usermod to recreate the user namespace with the +newly configured mappings. + ## SYNOPSIS **podman system migrate** ## SEE ALSO -`podman(1)`, `libpod.conf(5)` +`podman(1)`, `libpod.conf(5)`, `usermod(8)` ## HISTORY April 2019, Originally compiled by Giuseppe Scrivano (gscrivan at redhat dot com) diff --git a/docs/tutorials/rootless_tutorial.md b/docs/tutorials/rootless_tutorial.md new file mode 100644 index 000000000..a32c1f15b --- /dev/null +++ b/docs/tutorials/rootless_tutorial.md @@ -0,0 +1,85 @@ + + +# Basic Setup and Use of Podman in a Rootless environment. + +Prior to allowing users without root privileges to run Podman, the administrator must install or build Podman and complete the following configurations. + +## Administrator Actions + +### Installing Podman + +For installing Podman, please see the [installation instructions](https://github.com/containers/libpod/blob/master/install.md). + +### Building Podman + +For building Podman, please see the [installation instructions](https://github.com/containers/libpod/blob/master/install.md#building-from-scratch). + +### Install slirp4netns + +The [slirp4netns](https://github.com/rootless-containers/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](https://github.com/rootless-containers/slirp4netns). + +### Ensure fuse-overlayfs is installed + +When using Podman in a rootless environment, it is recommended to use fuse-overlayfs rather than the VFS file system. Installing the fuse3-devel package gives Podman the dependencies it needs to install, build and use fuse-overlayfs in a rootless environment for you. The fuse-overlayfs project is also available from [GitHub](https://github.com/containers/fuse-overlayfs). This especially needs to be checked on Ubuntu distributions as fuse-overlayfs is not generally installed by default. + +### Enable user namespaces (on RHEL7 machines) + +The number of user namespaces that are allowed on the system is specified in the file `/proc/sys/user/max_user_namespaces`. On most Linux platforms this is preset by default and no adjustment is necessary. However on RHEL7 machines a user with root privileges may need to set that to a reasonable value by using this command: `sysctl user.max_user_namespaces=15000`. + +### /etc/subuid and /etc/subgid configuration + +Rootless podman requires the user running it to have a range of UIDs listed in /etc/subuid and /etc/subgid files. The `shadows-utils` or `newuid` package provides these files on different distributions and they must be installed on the system. These files will need someone with root privileges on the system to add or update the entries within them. The following is a summarization from the [How does rootless Podman work?](https://opensource.com/article/19/2/how-does-rootless-podman-work) article by Dan Walsh on [opensource.com](https://opensource.com) + +Update the /etc/subuid and /etc/subgid with fields for each user that will be allowed to create containers that look like the following. Note that the values for each user must be unique and without any overlap. If there is an overlap, there is a potential for a user to use another’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/passwd or getpwent. +* 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. + +If you update either the /etc/subuid or the /etc/subgid file, 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 using the `[podman system migrate](https://github.com/containers/libpod/blob/master/docs/podman-system-migrate.1.md)` command which will stop all the containers for the user and will kill the pause process. + +Rather than updating the files directly, the usermod program can be used to assign UIDs and GIDs to a user. + +``` +usermod --add-subuids 200000-201000 --add-subgids 200000-201000 johndoe +grep johndoe /etc/subuid /etc/subgid +/etc/subuid:johndoe:200000:1001 +/etc/subgid:johndoe:200000:1001 +``` + +### Enable unprivileged `ping` + +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 persistent, the administrator will need to add a file in `/etc/sysctl.d` that contains `net.ipv4.ping_group_range=0 $MAX_UID`. + + +## User Actions + +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. + +### User Configuration Files. + +The Podman configuration files for root reside in /etc/containers. In the rootless environment they reside in ${XDG_RUNTIME_DIR}/containers and are owned by each individual user. The user can modify these files as they wish. + +## More information + +If you are still experiencing problems running Podman in a rootless environment, please refer to the [Shortcomings of Rootless Podman](https://github.com/containers/libpod/blob/master/rootless.md) page which lists known issues and solutions to known issues in this environment. + +For more information on Podman and its subcommands, checkout the asciiart demos on the [README.md](../../README.md#commands) page or the [podman.io](https://podman.io) web site. diff --git a/libpod/boltdb_state.go b/libpod/boltdb_state.go index 63e40a98f..12c364993 100644 --- a/libpod/boltdb_state.go +++ b/libpod/boltdb_state.go @@ -73,42 +73,50 @@ func NewBoltState(path string, runtime *Runtime) (State, error) { // As such, just a db.Close() is fine here. defer db.Close() - // Perform initial database setup - err = db.Update(func(tx *bolt.Tx) error { - if _, err := tx.CreateBucketIfNotExists(idRegistryBkt); err != nil { - return errors.Wrapf(err, "error creating id-registry bucket") - } - if _, err := tx.CreateBucketIfNotExists(nameRegistryBkt); err != nil { - return errors.Wrapf(err, "error creating name-registry bucket") - } - if _, err := tx.CreateBucketIfNotExists(nsRegistryBkt); err != nil { - return errors.Wrapf(err, "error creating ns-registry bucket") - } - if _, err := tx.CreateBucketIfNotExists(ctrBkt); err != nil { - return errors.Wrapf(err, "error creating containers bucket") - } - if _, err := tx.CreateBucketIfNotExists(allCtrsBkt); err != nil { - return errors.Wrapf(err, "error creating all containers bucket") - } - if _, err := tx.CreateBucketIfNotExists(podBkt); err != nil { - return errors.Wrapf(err, "error creating pods bucket") - } - if _, err := tx.CreateBucketIfNotExists(allPodsBkt); err != nil { - return errors.Wrapf(err, "error creating all pods bucket") - } - if _, err := tx.CreateBucketIfNotExists(volBkt); err != nil { - return errors.Wrapf(err, "error creating volume bucket") - } - if _, err := tx.CreateBucketIfNotExists(allVolsBkt); err != nil { - return errors.Wrapf(err, "error creating all volumes bucket") + createBuckets := [][]byte{ + idRegistryBkt, + nameRegistryBkt, + nsRegistryBkt, + ctrBkt, + allCtrsBkt, + podBkt, + allPodsBkt, + volBkt, + allVolsBkt, + runtimeConfigBkt, + } + + // Does the DB need an update? + needsUpdate := false + err = db.View(func(tx *bolt.Tx) error { + for _, bkt := range createBuckets { + if test := tx.Bucket(bkt); test == nil { + needsUpdate = true + break + } } - if _, err := tx.CreateBucketIfNotExists(runtimeConfigBkt); err != nil { - return errors.Wrapf(err, "error creating runtime-config bucket") + return nil + }) + if err != nil { + return nil, errors.Wrapf(err, "error checking DB schema") + } + + if !needsUpdate { + state.valid = true + return state, nil + } + + // Ensure schema is properly created in DB + err = db.Update(func(tx *bolt.Tx) error { + for _, bkt := range createBuckets { + if _, err := tx.CreateBucketIfNotExists(bkt); err != nil { + return errors.Wrapf(err, "error creating bucket %s", string(bkt)) + } } return nil }) if err != nil { - return nil, errors.Wrapf(err, "error creating initial database layout") + return nil, errors.Wrapf(err, "error creating buckets for DB") } state.valid = true diff --git a/pkg/rootless/rootless_linux.c b/pkg/rootless/rootless_linux.c index 5ab368691..26dfc7b31 100644 --- a/pkg/rootless/rootless_linux.c +++ b/pkg/rootless/rootless_linux.c @@ -480,7 +480,7 @@ create_pause_process (const char *pause_pid_file_path, char **argv) close (fd); setenv ("_PODMAN_PAUSE", "1", 1); - execlp (argv[0], NULL); + execlp (argv[0], argv[0], NULL); /* If the execve fails, then do the pause here. */ do_pause (); @@ -702,7 +702,6 @@ reexec_in_user_namespace (int ready, char *pause_pid_file_path, char *file_to_re pid = syscall_clone (CLONE_NEWUSER|CLONE_NEWNS|SIGCHLD, NULL); if (pid < 0) { - FILE *fp; fprintf (stderr, "cannot clone: %s\n", strerror (errno)); check_proc_sys_userns_file (_max_user_namespaces); check_proc_sys_userns_file (_unprivileged_user_namespaces); |