For non-Podman users of Libpod, we don't want to force the exit
command to use ARGV[0], which probably does not support a cleanup
command.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
even if the system is using cgroups v2, rootless is not able to setup
limits when the cgroup-manager is not systemd.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
"init" is a quite common name for the command executed in a container
image and Podman ends up using the systemd mode also when not
required.
Be stricter on enabling the systemd mode and not enable it
automatically when the basename is "init" but expect the full path
"/usr/sbin/init".
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
CRI-O defaults to 1024 for the maximum pids in a container. Podman
should have a similar limit. Once we have a containers.conf, we can
set the limit in this file, and have it easily customizable.
Currently the documentation says that -1 sets pids-limit=max, but -1 fails.
This patch allows -1, but also indicates that 0 also sets the max pids limit.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
This requires updating all import paths throughout, and a matching
buildah update to interoperate.
I can't figure out the reason for go.mod tracking
github.com/containers/image v3.0.2+incompatible // indirect
((go mod graph) lists it as a direct dependency of libpod, but
(go list -json -m all) lists it as an indirect dependency),
but at least looking at the vendor subdirectory, it doesn't seem
to be actually used in the built binaries.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
if there are no resources specified, make sure the OCI resources block
is empty so that the OCI runtime won't complain.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
When a named volume is mounted on any of the tmpfs filesystems
created by read-only tmpfs, it caused a conflict that was not
resolved prior to this.
Fixes BZ1755119
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
In order to run Podman with VM-based runtimes unprivileged, the
network must be set up prior to the container creation. Therefore
this commit modifies Podman to run rootless containers by:
1. create a network namespace
2. pass the netns persistent mount path to the slirp4netns
to create the tap inferface
3. pass the netns path to the OCI spec, so the runtime can
enter the netns
Closes#2897
Signed-off-by: Gabi Beyer <gabrielle.n.beyer@intel.com>
We identify and resolve conflicts in paths using destination path
matches. We require exact matches, largely for performance
reasons (we use maps to efficiently access, keyed by
destination). This usually works fine, until you get mounts that
are targetted at /output and /output/ - the same path, but not
the same string.
Use filepath.Clean() aggressively to try and solve this.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Currently if a user specifies a --mount option, their is no way to tell SELinux
to relabel the mount point.
This patch addes the relabel=shared and relabel=private options.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
This is mostly used with Systemd, which really wants to manage
CGroups itself when managing containers via unit file.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
provide an implementation for getDevices that skip unreadable
directories for the current user.
Based on the implementation from runc/libcontainer.
Closes: https://github.com/containers/libpod/issues/3919
Signed-off-by: Giuseppe Scrivano <giuseppe@scrivano.org>
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
when running in rootless mode, --device creates a bind mount from the
host instead of specifying the device in the OCI configuration. This
is required as an unprivileged user cannot use mknod, even when root
in a user namespace.
Closes: https://github.com/containers/libpod/issues/3905
Signed-off-by: Giuseppe Scrivano <giuseppe@scrivano.org>
For read-only containers set to create tmpfs filesystems over
/run and other common destinations, we were incorrectly setting
mount options, resulting in duplicate mount options.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
If I mount, say, /usr/bin into my container - I expect to be able
to run the executables in that mount. Unconditionally applying
noexec would be a bad idea.
Before my patches to change mount options and allow exec/dev/suid
being set explicitly, we inferred the mount options from where on
the base system the mount originated, and the options it had
there. Implement the same functionality for the new option
handling.
There's a lot of performance left on the table here, but I don't
know that this is ever going to take enough time to make it worth
optimizing.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
We already process the options on all tmpfs filesystems during
final addition of mounts to the spec. We don't need to do it
before that in parseVolumes.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Previously, we explicitly set noexec/nosuid/nodev on every mount,
with no ability to disable them. The 'mount' command on Linux
will accept their inverses without complaint, though - 'noexec'
is counteracted by 'exec', 'nosuid' by 'suid', etc. Add support
for passing these options at the command line to disable our
explicit forcing of security options.
This also cleans up mount option handling significantly. We are
still parsing options in more than one place, which isn't good,
but option parsing for bind and tmpfs mounts has been unified.
Fixes: #3819Fixes: #3803
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Docker has unlimited tmpfs size where Podman had it set to 64mb. Should be standard between the two.
Remove noexec default
Signed-off-by: Ashley Cui <ashleycui16@gmail.com>
The 'podman run --mount' flag previously allowed the 'ro' option
to be specified, but was missing the ability to set it to a bool
(as is allowed by docker). Add that. While we're at it, allow
setting 'rw' explicitly as well.
Fixes#2980
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Sharing a UTS namespace means sharing the hostname. Fix situations where a container in a pod didn't properly share the hostname of the pod.
Signed-off-by: Peter Hunt <pehunt@redhat.com>
Previously, we use CreateConfig's Command to populate container
Command (which is used as CMD for Inspect and Commit).
Unfortunately, CreateConfig's Command is the container's full
command, including a prepend of Entrypoint - so we duplicate
Entrypoint for images that include it.
Maintain a separate UserCommand in CreateConfig that does not
include the entrypoint, and use that instead.
Fixes#3708
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
In order to run Podman with VM-based runtimes unprivileged, the
network must be set up prior to the container creation. Therefore
this commit modifies Podman to run rootless containers by:
1. create a network namespace
2. pass the netns persistent mount path to the slirp4netns
to create the tap inferface
3. pass the netns path to the OCI spec, so the runtime can
enter the netns
Closes#2897
Signed-off-by: Gabi Beyer <gabrielle.n.beyer@intel.com>
allow a container to run in a new cgroup namespace.
When running in a new cgroup namespace, the current cgroup appears to
be the root, so that there is no way for the container to access
cgroups outside of its own subtree.
By default it uses --cgroup=host to keep the previous behavior.
To create a new namespace, --cgroup=private must be provided.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
We can infer no-new-privileges. For now, manually populate
seccomp (can't infer what file we sourced from) and
SELinux/Apparmor (hard to tell if they're enabled or not).
Signed-off-by: Matthew Heon <mheon@redhat.com>
When we first began writing Podman, we ran into a major issue
when implementing Inspect. Libpod deliberately does not tie its
internal data structures to Docker, and stores most information
about containers encoded within the OCI spec. However, Podman
must present a CLI compatible with Docker, which means it must
expose all the information in 'docker inspect' - most of which is
not contained in the OCI spec or libpod's Config struct.
Our solution at the time was the create artifact. We JSON'd the
complete CreateConfig (a parsed form of the CLI arguments to
'podman run') and stored it with the container, restoring it when
we needed to run commands that required the extra info.
Over the past month, I've been looking more at Inspect, and
refactored large portions of it into Libpod - generating them
from what we know about the OCI config and libpod's (now much
expanded, versus previously) container configuration. This path
comes close to completing the process, moving the last part of
inspect into libpod and removing the need for the create
artifact.
This improves libpod's compatability with non-Podman containers.
We no longer require an arbitrarily-formatted JSON blob to be
present to run inspect.
Fixes: #3500
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
OpenShift Merge Robot
Matthew Heon
Giuseppe Scrivano
Daniel J Walsh
Miloslav Trmač
Gabi Beyer
Danila Kiver
Giuseppe Scrivano
Ashley Cui
Peter Hunt
baude
Matthew Heon