When a container is in a PID namespace, it is enought to send
the stop signal to the PID 1 of the namespace, only send signals
to all processes in the container when the container is not in
a pid namespace.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
When trying to reproduce #4704 I noticed that the named volumes
from the Postgres containers in the reproducer weren't being
removed by `podman pod rm -f` saying that the container they were
attached to was still in use. This was rather odd, considering
they were only in use by one container, and that container was in
the process of being removed with the pod.
After a bit of tracing, I realized that the cause is the ordering
of container removal when we remove a pod. Normally, it's done
in removeContainer() before volume removal (which is the last
thing in that function). However, when we are removing a pod, we
remove containers all at once, after removeContainer has already
finished - meaning the container still exists when we try to
remove its volumes, and thus the volume can't be removed.
Solution: collect a list of all named volumes in use by the pod,
and remove them all at once after every container in the pod is
gone. This ensures that there are no dependency issues.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
The earlier attempt to re-add config migration only worked with
user-specified configs (podman run --config). This version works
more in line with that we want - the first rootless config file
will be changed from runc to crun.
Verified on my system after an F31 migration - everything seems
to be working well.
Signed-off-by: Matthew Heon <mheon@redhat.com>
Store the full command plus arguments of the process the container has
been created with. Expose this data as a `Config.CreateCommand` field
in the container-inspect data as well.
This information can be useful for debugging, as we can find out which
command has created the container, and, if being created via the Podman
CLI, we know exactly with which flags the container has been created
with.
The immediate motivation for this change is to use this information for
`podman-generate-systemd` to generate systemd-service files that allow
for creating new containers (in contrast to only starting existing
ones).
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
During writing the tests I found it would be probably useful to have the
tag history part of the inspect data.
Signed-off-by: Sascha Grunert <sgrunert@suse.com>
Updates the podman info command to show registries from v1 config file
in the search table format.
Signed-off-by: José Guilherme Vanz <jvanz@jvanz.com>
We currently rely on exec sessions being removed from the state
by the Exec() API itself, on detecting the session stopping. This
is not a reliable method, though. The Podman frontend for exec
could be killed before the session ended, or another Podman
process could be holding the lock and prevent update (most
notable in `run --rm`, when a container with an active exec
session is stopped).
To resolve this, add a function to reap active exec sessions from
the state, and use it on cleanup (to clear sessions after the
container stops) and remove (to do the same when --rm is passed).
This is a bit more complicated than it ought to be because Kata
and company exist, and we can't guarantee the exec session has a
PID on the host, so we have to plumb this through to the OCI
runtime.
Fixes#4666
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
to make things more effecient for the api work we are doing, we should
process image filters internally (as opposed to in main). this allows
for better api responses and more closely affiliated functions.
Signed-off-by: baude <bbaude@redhat.com>
In the process, make everything in the config omitempty in TOML.
We're seeing issues (notably [1]) where, after rewriting
libpod.conf, fields that were not previously populated are
written - and, because they were not previously written, they are
included as empty. This is unfortunately different from not
included at all - it means that we need to assume the user
explicitly unset the value, and we can't use defaults. Setting
omitempty prevents us from writing things that should not be
written as they were not set originally.
[1] https://github.com/containers/libpod/issues/4210
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
When doing a checkpoint with --export the root file-system diff was not
working as expected. Instead of getting the changes from the running
container to the highest storage layer it got the changes from the
highest layer to that parent's layer. For a one layer container this
could mean that the complete root file-system is part of the checkpoint.
With this commit this changes to use the same functionality as 'podman
diff'. This actually enables to correctly diff the root file-system
including tracking deleted files.
This also removes the non-working helper functions from libpod/diff.go.
Signed-off-by: Adrian Reber <areber@redhat.com>
Return types had to change a bit for this, but since we can wrap
the old v1.ImageConfig, changes are overall not particularly bad.
At present, I believe this only works with commit, not import.
This matches how things were before we changed to the new parsing
so I think this is fine.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
It turns out we had two independent parsing impkementations for
Dockerfile instructions out of --change. My previous commit fixed
the one used in --change, but as I discovered to my dismay,
commit used a different implementation. Remove that and use the
new parsing implementation instead.
While we're at it, fix some bugs in the current commit code. The
addition of anonymous named volumes to Libpod recently means we
can now include those in the image config when committing. Some
changes (VOLUME, ENV, EXPOSE, LABEL) previously cleared the
config of the former image when used; Docker does not do this, so
I removed that behavior.
Still needs fixing: the new implementation does not support
ONBUILD, while the old one did.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
The code currently assumes that the container we delegate network
namespace to will never further delegate to another container, so
when looking up things like /etc/hosts and /etc/resolv.conf we
won't pull the correct files from the chained dependency. The
changes to resolve this are relatively simple - just need to keep
looking until we find a container without NetNsCtr set.
Fixes#4626
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
After a restart, pods and containers both run a refresh()
function to prepare to run after a reboot. Until now, volumes
have not had a similar function, because they had no per-boot
setup to perform.
Unfortunately, this was not noticed when in-memory locking was
introduced to volumes. The refresh() routine is, among other
things, responsible for ensuring that locks are reserved after a
reboot, ensuring they cannot be taken by a freshly-created
container, pod, or volume. If this reservation is not done, we
can end up with two objects using the same lock, potentially
needing to lock each other for some operations - classic recipe
for deadlocks.
Add a refresh() function to volumes to perform lock reservation
and ensure it is called as part of overall refresh().
Fixes#4605Fixes#4621
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
do not change the runtime error to be lowercase, but use a case
insensitive regex matching. In this way the original error from the
OCI runtime is reported back.
regression introduced by bc485bce47
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
This command will destroy all data created via podman.
It will remove containers, images, volumes, pods.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Trying to checkpoint a container started with --rm works, but it makes
no sense as the container, including the checkpoint, will be deleted
after writing the checkpoint. This commit inhibits checkpointing
containers started with '--rm' unless '--export' is used. If the
checkpoint is exported it can easily be restored from the exported
checkpoint, even if '--rm' is used. To restore a container from a
checkpoint it is even necessary to manually run 'podman rm' if the
container is not started with '--rm'.
Signed-off-by: Adrian Reber <areber@redhat.com>
when parsing the OCI error, be sure to discard any other output that
is not matched. The full output is still printed with
--log-level=debug.
Closes: https://github.com/containers/libpod/issues/4574
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
We leverage the containers/storage image history tracking feature to
show the previously used image names when running:
`podman images --history`
Signed-off-by: Sascha Grunert <sgrunert@suse.com>
These only conflict when joining more than one network. We can
still set a single CNI network and set a static IP and/or static
MAC.
Fixes#4500
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
filter option accepts two filters.
- label
- until
label supports "label=value" or "label=key=value" format
until supports all golang compatible time/duration formats.
Signed-off-by: Kunal Kushwaha <kunal.kushwaha@gmail.com>
If the container is running and we need to get its netns and
can't, that is a serious bug deserving of errors.
If it's not running, that's not really a big deal. Log an error
and continue.
Signed-off-by: Matthew Heon <mheon@redhat.com>
When Libpod removes a container, there is the possibility that
removal will not fully succeed. The most notable problems are
storage issues, where the container cannot be removed from
c/storage.
When this occurs, we were faced with a choice. We can keep the
container in the state, appearing in `podman ps` and available for
other API operations, but likely unable to do any of them as it's
been partially removed. Or we can remove it very early and clean
up after it's already gone. We have, until now, used the second
approach.
The problem that arises is intermittent problems removing
storage. We end up removing a container, failing to remove its
storage, and ending up with a container permanently stuck in
c/storage that we can't remove with the normal Podman CLI, can't
use the name of, and generally can't interact with. A notable
cause is when Podman is hit by a SIGKILL midway through removal,
which can consistently cause `podman rm` to fail to remove
storage.
We now add a new state for containers that are in the process of
being removed, ContainerStateRemoving. We set this at the
beginning of the removal process. It notifies Podman that the
container cannot be used anymore, but preserves it in the DB
until it is fully removed. This will allow Remove to be run on
these containers again, which should successfully remove storage
if it fails.
Fixes#3906
Signed-off-by: Matthew Heon <mheon@redhat.com>
When restoring a container with user namespace, the user namespace is
created by the OCI runtime, and the network namespace is created after
the user namespace to ensure correct ownership.
In this case PostConfigureNetNS will be set and the value of
c.state.NetNS would be nil. Hence, the following error occurs:
$ sudo podman run --name cr \
--uidmap 0:1000:500 \
-d docker.io/library/alpine \
/bin/sh -c 'i=0; while true; do echo $i; i=$(expr $i + 1); sleep 1; done'
$ sudo podman container checkpoint cr
$ sudo podman container restore cr
...
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x30 pc=0x13a5e3c]
Signed-off-by: Radostin Stoyanov <rstoyanov1@gmail.com>
In conmon 2.0.3, we add another fifo to handle window resizing. This needs to be cleaned up for commands like restore, where the same path is used.
Signed-off-by: Peter Hunt <pehunt@redhat.com>
Rewrite the backend for displaying the history of an image to simplify
the code and be closer to docker's behaviour. Instead of driving
index-based heuristics, create a reverse mapping from top-layers to the
corresponding image IDs and lookup the layers on-demand. Also use the
uncompressed layer size to be closer to Docker's behaviour.
Note that intermediate images from local builds are not considered for
the ID lookups anymore.
Fixes: #3359
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
When running on a node with Cgroups v2, default to using `crun` instead
of `runc`. Note that this only impacts the hard-coded default config.
No user config will be over-written.
Fixes: #4463
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
Currently podman generate kube does not generate the correct RunAsUser and RunAsGroup
options in the yaml file. This patch fixes this.
This patch also make `podman play kube` use the RunAdUser and RunAsGroup options if
they are specified in the yaml file.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
`go get github.com/cri-o/ocicni@deac903fd99b6c52d781c9f42b8db3af7dcfd00a`
I had to fix compilation errors in libpod/networking_linux.go
---
ocicni.Networks has changed from string to the structure NetAttachment
with the member Name (the former string value) and the member Ifname
(optional).
I don't think we can make use of Ifname here, so I just map the array of
structures to array of strings - e.g. dropping Ifname.
---
The function GetPodNetworkStatus no longer returns Result but it returns
the wrapper structure NetResult which contains the former Result plus
NetAttachment (Network name and Interface name).
Again, I don't think we can make use of that information here, so I
just added `.Result` to fix the build.
---
Issue: #1136
Signed-off-by: Jakub Filak <jakub.filak@sap.com>
If user specifies --detach-keys="", this will disable the feature.
Adding define.DefaultDetachKeys to help screen to help identify detach keys.
Updated man pages with additonal information.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
When pulling an unqualified reference (e.g., `fedora`) make sure that
the reference is not using a non-docker transport to avoid iterating
over the search registries and trying to pull from them.
Fixes: #4434
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
If the kube.yaml specifieds the SELinux type or Level, we need the container
to be launched with the correct label.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
the pidWaitTimeout is already a Duration so do not multiply it again
by time.Millisecond.
Closes: https://github.com/containers/libpod/issues/4344
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Pull in changes to pkg/secrets/secrets.go that adds the
logic to disable fips mode if a pod/container has a
label set.
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
change the default to -1, so that we can change the semantic of
"--tail 0" to not print any existing log line.
Closes: https://github.com/containers/libpod/issues/4396
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Refactor the `RuntimeConfig` along with related code from libpod into
libpod/config. Note that this is a first step of consolidating code
into more coherent packages to make the code more maintainable and less
prone to regressions on the long runs.
Some libpod definitions were moved to `libpod/define` to resolve
circular dependencies.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
There were many situations that made exec act funky with input. pipes didn't work as expected, as well as sending input before the shell opened.
Thinking about it, it seemed as though the issues were because of how os.Stdin buffers (it doesn't). Dropping this input had some weird consequences.
Instead, read from os.Stdin as bufio.Reader, allowing the input to buffer before passing it to the container.
Signed-off-by: Peter Hunt <pehunt@redhat.com>
command.Start() just starts the command. That catches some
errors, but the nasty ones - bad options and similar - happen
when the command runs. Use CombinedOutput() instead - it waits
for the command to exit, and thus catches non-0 exit of the
`mount` command (invalid options, for example).
STDERR from the `mount` command is directly used, which isn't
necessarily the best, but we can't really get much more info on
what went wrong.
Fixes#4303
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
always create a new cgroup for conmon also when running as rootless.
We were previously creating one only when necessary, but that behaves
differently than root containers.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Currently podman play kube is not using the system default seccomp.json file.
This PR will use the default or override location for podman play.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Generate an image's RepoDigests list using all applicable digests, and
refrain from outputting a digest in the tag column of the "images"
output.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
Be prepared to report multiple image digests for images which contain
multiple manifests but, because they continue to have the same set of
layers and the same configuration, are considered to be the same image.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
Add --override-arch and --override-os as hidden flags, in line with the
global flag names that skopeo uses, so that we can test behavior around
manifest lists without having to conditionalize more of it by arch.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
When an image can be opened as an ImageSource but not an Image, handle
the case where it's an image list all by itself, the case where it's an
image for a different architecture/OS combination, or the case where
it's both.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
Move to containers/image v5 and containers/buildah to v1.11.4.
Replace an equality check with a type assertion when checking for a
docker.ErrUnauthorizedForCredentials in `podman login`.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
read the slirp4netns stderr and propagate it in the error when the
process fails.
Replace: https://github.com/containers/libpod/pull/4338
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
We have a lot of checks for container state scattered throughout
libpod. Many of these need to ensure the container is in one of a
given set of states so an operation may safely proceed.
Previously there was no set way of doing this, so we'd use unique
boolean logic for each one. Introduce a helper to standardize
state checks.
Note that this is only intended to replace checks for multiple
states. A simple check for one state (ContainerStateRunning, for
example) should remain a straight equality, and not use this new
helper.
Signed-off-by: Matthew Heon <mheon@redhat.com>
when running in systemd mode on cgroups v1, make sure the
/sys/fs/cgroup/systemd/release_agent is masked otherwise the container
is able to modify it and execute scripts on the host.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
When you try and create a new volume with the name of a volume
that already exists, you presently get a thoroughly unhelpful
error from `mkdir` as the volume attempts to create the
directory it will be mounted at. An EEXIST out of mkdir is not
particularly helpful to Podman users - it doesn't explain that
the name is already taken by another volume.
The solution here is potentially racy as the runtime is not
locked, so someone else could take the name while we're still
getting things set up, but that's a narrow timing window, and we
will still return an error - just an error that's not as good as
this one.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
if the cgroup manager is set to systemd, detect if dbus is available,
otherwise fallback to --cgroup-manager=cgroupfs.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Everything else is a flag to mount, but "uid" and "gid" are not.
We need to parse them out of "o" and handle them separately.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
OpenShift Merge Robot
Daniel J Walsh
Matthew Heon
Matthew Heon
José Guilherme Vanz
Valentin Rothberg
Sascha Grunert
baude
Adrian Reber
Giuseppe Scrivano
Kunal Kushwaha
Radostin Stoyanov
Peter Hunt
Dmitry Smirnov
Jakub Filak
Giuseppe Scrivano
Urvashi Mohnani
Nalin Dahyabhai