since we now enter the user namespace prior to read the conmon.pid, we
can write the conmon.pid file again to the runtime dir.
This reverts commit 6c6a865436.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
when running in rootless mode, be sure psgo is honoring the user
namespace settings for huser and hgroup.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
replace two usage of kwait.ExponentialBackoff in favor of WaitForFile
that uses inotify when possible.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
enable polling also when using inotify. It is generally useful to
have it as under high load inotify can lose notifications. It also
solves a race condition where the file is created while the watcher
is configured and it'd wait until the timeout and fail.
Closes: https://github.com/containers/libpod/issues/2942
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Mainly add support for podman build using --overlay mounts.
Updates containers/image also adds better support for new registries.conf
file.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
`string.Split()` splits into slice of size greater than 2
which may result in loss of environment variables
fixes#3132
Signed-off-by: Divyansh Kamboj <kambojdivyansh2000@gmail.com>
use a pause process to keep the user and mount namespace alive.
The pause process is created immediately on reload, and all successive
Podman processes will refer to it for joining the user&mount
namespace.
This solves all the race conditions we had on joining the correct
namespaces using the conmon processes.
As a fallback if the join fails for any reason (e.g. the pause process
was killed), then we try to join the running containers as we were
doing before.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
once the default event logger was removed from libpod.conf, we need to
set the default based on whether the systemd build tag is used or not.
Signed-off-by: baude <bbaude@redhat.com>
StartAndAttach() runs start() in a goroutine, which can allow it
to fire after the caller returns - and thus, after the defer to
unlock the container lock has fired.
The start() call _must_ occur while the container is locked, or
else state inconsistencies may occur.
Fixes#3114
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
If the systemd development files are not present on the system which
builds podman, then `podman events` will error on runtime creation.
Beside this, a warning will be printed when compiling podman.
This commit mainly exists because projects which depend on libpod
would not need the podman event support and therefore do not need to
rely on the systemd headers.
Signed-off-by: Sascha Grunert <sgrunert@suse.com>
When using CGroupfs, we see races during pod removal between
removing the CGroup and the cleanup process starting (in the
CGroup, thus preventing removal).
The simplest way to avoid this is to prevent the forking of the
cleanup process. Conveniently, we can do this via the CGroup that
we already created for Conmon - we just need to update the PID
limit to 0, which completely inhibits new forks.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Instead of rewriting the logic, reuse the standard logic we use
for removing containers, which is much better tested.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Ensure that, if an error occurs somewhere along the way when we
remove a pod, it's preserved until the end and returned, even as
we continue to remove the pod.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Removing a pod must first removal all containers in the pod.
Libpod requires the state to remain consistent at all times, so
references to a deleted pod must all be cleansed first.
Pods can have many containers in them. We presently iterate
through all of them, and if an error occurs trying to clean up
and remove any single container, we abort the entire operation
(but cannot recover anything already removed - pod removal is not
an atomic operation).
Because of this, if a removal error occurs partway through, we
can end up with a pod in an inconsistent state that is no longer
usable. What's worse, if the error is in the infra container, and
it's persistent, we get zombie pods - completely unable to be
removed.
When we saw some of these same issues with containers not in
pods, we modified the removal code there to aggressively purge
containers from the database, then try to clean up afterwards.
Take the same approach here, and make cleanup errors nonfatal.
Once we've gone ahead and removed containers, we need to see
pod deletion through to the end - we'll log errors but keep
going.
Also, fix some other small things (most notably, we didn't make
events for the containers removed).
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
After a reboot, when we refresh Podman's state, we retrieved the
lock from the fresh SHM instance, but we did not mark it as
allocated to prevent it being handed out to other containers and
pods.
Provide a method for marking locks as in-use, and use it when we
refresh Podman state after a reboot.
Fixes#2900
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
We want to start supporting the registries.conf format.
Also start showing blocked registries in podman info
Fix sorting so all registries are listed together in podman info.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
The on-failure restart option supports restarting only a given
number of times. To do this, we need one additional field in the
DB to track restart count (which conveniently fills a field in
Inspect we weren't populating), plus some plumbing logic.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
This field indicates that a container was explciitly stopped by
an API call, and did not exit naturally. It's used when
implementing restart policy for containers.
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
Fallback to executing ps(1) in case we hit an unknown psgo descriptor.
This ensures backwards compatibility with docker-top, which was purely
ps(1) driven.
Also support comma-separated descriptors as input.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
