The backstory for this is that runc 1.2 (opencontainers/runc#3967)
fixed a long-standing bug in our mount flag handling (a bug that crun
still has). Before runc 1.2, when dealing with locked mount flags that
user namespaced containers cannot clear, trying to explicitly clearing
locked flags (like rw clearing MS_RDONLY) would silently ignore the rw
flag in most cases and would result in a read-only mount. This is
obviously not what the user expects.
What runc 1.2 did is that it made it so that passing clearing flags
like rw would always result in an attempt to clear the flag (which was
not the case before), and would (in all cases) explicitly return an
error if we try to clear locking flags. (This also let us finally fix a
bunch of other long-standing issues with locked mount flags causing
seemingly spurious errors).
The problem is that podman sets rw on all mounts by default (even if
the user doesn't specify anything). This is actually a no-op in
runc 1.1 and crun because of a bug in how clearing flags were handled
(rw is the absence of MS_RDONLY but until runc 1.2 we didn't correctly
track clearing flags like that, meaning that rw would literally be
handled as if it were not set at all by users) but in runc 1.2 leads to
unfortunate breakages and a subtle change in behaviour (before, a ro
mount being bind-mounted into a container would also be ro -- though
due to the above bug even setting rw explicitly would result in ro in
most cases -- but with runc 1.2 the mount will always be rw even if
the user didn't explicitly request it which most users would find
surprising). By the way, this "always set rw" behaviour is a departure
from Docker and it is not necesssary.
Fixes: https://issues.redhat.com/browse/RHEL-152635,
https://issues.redhat.com/browse/RHEL-152631
Signed-off-by: rcmadhankumar <madhankumar.chellamuthu@suse.com>
(cherry picked from commit bf7dcd5619)
Signed-off-by: Tom Sweeney <tsweeney@redhat.com>
and remote
There is no reason for local and rmeote to use a different error that
just differs in quoting. Make them consitent to simplify the tests.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
There doesn't seem any reason why the system commands should not join
the userns. In particular the main commands use ParentNSRequired and
UnshareNSRequired when they don't want to be joined to the main userns.
Since the system command don't set these the go code does the join and
re-exec anyway so might as well use the shortcut to speed that up.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
When trying to join the conmon pid to recreate the pause process based
on the namespace it can be that the pid is no longer valid, i.e. when
conmon crashed or was killed.
Currently we have a big issue that can be reproduced using:
$ podman run -d quay.io/libpod/testimage:20241011 sleep 100
$ killall -9 conmon
$ killall catatonit
All commands would fail as we keep trying to rejoin the namespace of the
non existing conmon process.
So to address that fall back to creating a new namespace if we fail to
join the conmon pids.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
Just a minor improvement as we know the size needed for the slice we can
allocate it only once instead of the append having to resize it.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
Based on the description in commit 63ef557 this was added so that the
migrate command does not move the pause process into a separate cgroup.
It should however not disable the rejoining of the userns when the pause
process join failed. BEcause of this we end up calling migrate without a
userns and that then can fail if there are actual contianer it tries to
cleanup.
Fixes: 63ef5576ed ("command: migrate doesn't move process to cgroup")
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
There is no good reason to use logrus and os.Exit() here, other parts of
this function already return the error so do the same. The main podman
process will exit then with the normal formatted error message.
And also log an error about the last return which should never happen as
we should have exited above if the re-exec worked or errored out.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
With go 1.23 the maps package was added in the std library. The linter
now wants us to use that. However the API chnaged as it returns an
iterator, thus we also have to use slices.Collect().
Also update the Fedora IMAGE_SUFFIX value in .cirrus.yml
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
The location of some functions have changed in the
updated selinux that we are now using in v5.4-rhel.
Make adjustments to calls to acount for that.
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
Hard coding to none without checking containers.conf is not a good idea
as users who liked the previous behavior and the podman default behavior
of keeping the hosts entries can no longer do that.
With this commit they can set base_hosts_file = "/etc/hosts" to restore
the previous behavior.
Originally Fixes: https://issues.redhat.com/browse/RHEL-92995
This cherry-pick Fixes: https://issues.redhat.com/browse/ACCELFIX-535
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: tomsweeneyredhat <tsweeney@redhat.com>
avoid using the glibc fork() function after using directly the clone()
syscall, as it confuses glibc causing the fork() to hang in some
cases.
The issue has been observed only on s390x, and the fix was confirmed
in the issue discussion.
Closes: https://github.com/containers/podman/issues/25184
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Allowing for multiple manifest per artifact just makes the code and cli
design harder to work with it. It is not clear how mounting, extracting
or edit on a multi manifest artifact should have worked.
A single manifest should make the code much easier to work with.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
podman artifact add now supports two new command line switches.
--type string that describes the type of artifact
--annotation string slice in the form of key=val
These new options allow users to "tag" information in on their artifacts
for any number of purposes down the line
RUN-2446
Signed-off-by: Brent Baude <bbaude@redhat.com>
`argv[0]`, ie: the full buffer allocated by `get_cmd_line_args`,
was going to be freed only if `can_use_shortcut()` was called.
Instead, let `init()` always manage `argv0` lifecycle.
Signed-off-by: Federico Di Pierro <nierro92@gmail.com>
Allow the user to provide an Ansible playbook file on init which will
then be run on boot.
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
Signed-off-by: Brent Baude <bbaude@redhat.com>
podman exec support detaching early via the detach key sequence. In that
case the podman process should exit successfully but the container exec
process keeps running.
Now I wrote automated test for both podman run and exec detach but this
uncovered several larger issues:
- detach sequence parsing is broken[1]
- podman-remote exec detach is broken[2]
- detach in general seems to be buggy/racy, seeing lot of flakes that
fail to restore the terminal and get an EIO instead, i.e.
"Unable to restore terminal: input/output error"
Thus I cannot add tests for now but this commit should at least fix the
obvoius case as reported by the user so I like to get this in regardless
and I will work through the other issues once I have more time.
Fixes#24895
[1] https://github.com/containers/common/pull/2302
[2] https://github.com/containers/podman/issues/25089
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
As issue #25112 points out, it was possible to start a machine on one of the darwin providers and then switch providers and start another one with a different name. This PR firstly prevents that use which is a forbidden use case.
Secondarily, performed some minor cleanup on the error messages being used so that the error would be specific to this condition.
This bug fix is for darwin only. In the case of Windows, we probably need to answer the question I raised in #24067 first, which is whether we want to stop allowing WSL to run multiple machines.
Fixes#25112
Signed-off-by: Brent Baude <bbaude@redhat.com>
The Kind() exported function is unused in our code; moreover, the function cannot be accurate because in the case of darwin, applehv and libkrun use the same config in the struct and therefore, we cannot identify the provider via that method.
Signed-off-by: Brent Baude <bbaude@redhat.com>
The behavior of function `path/filepath.EvalSymlinks()` has
changed in Go v1.23:
- https://go-review.googlesource.com/c/go/+/565136
- https://go.dev/doc/go1.23#minor_library_changes
- https://tip.golang.org/doc/godebug
As a consequences, starting with Podman 5.3.0, when installing
on Windows (WSL) using scoop, Podman fails to start because it
fails to find helper binaries. Scoop copies Podman binaries in
a folder of type Junction and `EvalSymlinks` returns an error.
The problem is described in #24557.
To address this problem we are checking if a path is a `Symlink`
before calling `EvalSymlinks` and, if it's not (hardlinks, mount
points or canonical files), we are calling `path/filepath.Clean`
for consistency. In fact `path/filepath.EvalSymlinks`, after
evaluating a symlink target, calls `Clean` too.
Signed-off-by: Mario Loriedo <mario.loriedo@gmail.com>
The `podman system prune` command is able to remove build containers that were created during the build, but were not removed because the build terminated unexpectedly.
By default, build containers are not removed to prevent interference with builds in progress. Use the **--build** flag when running the command to remove build containers as well.
Fixes: https://issues.redhat.com/browse/RHEL-62009
Signed-off-by: Jan Rodák <hony.com@seznam.cz>
This should be set only by podman as it is used for the podman generate
systemd --new command. For the api it was set to the system service
command which is simply pointless. It must be empty in these cases.
Fixes#25026
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
the podman artifact verb is used to manage OCI artifacts. the following
verbs were added to `podman artifact`:
* add
* inspect
* ls
* pull
* push
* rm
Notable items with this PR:
* all artifact commands and their output are subject to change. i.e.
consider all of this tech preview
* there is no way to add a file to an artifact that already exists in
the store. you would need to delete and recreate the artifact.
* all references to artifacts names should be fully qualified names in
the form of repo/name:tag (i.e. quay.io/artifact/foobar:latest)
* i understand that we will likely want to be able to attribute things
like arch, etc to artifact files. this function is not available yet.
Many thanks to Paul Holzinger for autocompletion PRs and review PRs that
fixed issues early on.
Also fix up some Args function to specify the correct number of args.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Signed-off-by: Brent Baude <bbaude@redhat.com>
The regex match would return a horrible error message and is way more
complicated then it should be. Simply check that .exe is not part of the
output.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The test pulls a big disk image every time which is slow. I see no good
way around that. Let's try to use /dev/null as image as we do not have
to run the VM at all and just can pass a NOP file to make the init
command happy.
That pull of that image seems to take over 2m so we safe quite a lot.
Also update the matcher for the slice. BeTrue() produces horrible
errors.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Fixes: https://github.com/containers/podman/issues/25002
Also add the ability to inspect containers for
UseImageHosts and UseImageHostname.
Finally fixed some bugs in handling of --no-hosts for Pods,
which I descovered.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Signed-off-by: Leo Liu <silver886@users.noreply.github.com>
Add comment
In shell completion, there is `.exe` suffix on Windows and this does not provide same experience across platforms, #16499
Signed-off-by: Leo Liu <silver886@users.noreply.github.com>
Create unit test for `.exe` suffix removal
Signed-off-by: Leo Liu <11664880+silver886@users.noreply.github.com>
Update comments
Signed-off-by: Leo Liu <11664880+silver886@users.noreply.github.com>
We now handle CDI qualified names being passed to resources.limits. The
support for that was already in libpod as of ab7f6095a1
when passed via the devices list. this just hooks the kube yaml parser
up to it.
Additionally we introduce `podman.io/device` that accepts device paths
as names and is transparently translated to mimick --device. This allows
bringing arbitrary devices into the container via similar, although
incompatible with, k8s mechanics:
```yaml
resources:
requests:
podman.io/device=/dev/kmsg: 1
```
Fixes: https://github.com/containers/podman/issues/17833
Signed-off-by: Robert Günzler <r@gnzler.io>
