When copying files from or to a container we so far opened the host side
fd first, then entered the container (specifically, joined it's mount
namespace) in a forked off child process, and opened the other side
there, followed by the (potentially slow) copying from inside the
container mount namespace.
This commit changes this so that we rejoin the host mount namespace
before doing the copying routine. This is relevant, so that we can rely
on /proc/self/fd/… to work, which is not the case otherwise, as we'll
see /proc/ from a pidns that is not our own, in wich case
/proc/self/fd/… is refused. By moving back to the host mount namespace
our own pidns and the pidns the /proc/ mount belongs to will be in sync
again, and all is good.
This is in particular preparation for the next commit, that makes the
copy routine strictly depending on /proc/ being accessible and working.
This PR introduces io.systemd.Machine.CopyFrom and CopyTo method which
are DBus alternatives of:
- CopyFromMachine
- CopyToMachine
- CopyFromMachineWithFlags
- CopyToMachineWithFlags
The PR also contain tests
unit_file_add_dependency() asserts, so check before calling it that the
type is expected, or return EINVAL to the caller.
root@localhost:~# busctl call org.freedesktop.systemd1 /org/freedesktop/systemd1 org.freedesktop.systemd1.Manager AddDependencyUnitFiles "asssbb" 0 uwhatm8 After 1 1
Broadcast message from systemd-journald@localhost (Mon 2025-01-06 18:12:14 UTC):
systemd[1]: Caught <ABRT>, from our own process.
Fixes https://github.com/systemd/systemd/issues/35882
We can use our upstream fmf definitions to run downstream tests in
the Fedora systemd dist-git repository
(https://src.fedoraproject.org/rpms/systemd). To have access to the
dist-git sources when running the tests, we enable dist-git-source: true
downstream which makes the sources available in $TMT_SOURCE_DIR so
let's make sure we use those sources if they're available.
Let's assume the workflow file containing the mkosi always being
available.
Also fix the detection while we're at it by not leaving the systemd
directory.
…ith fd passing enabled
Let's add a simple flag that enables fd passing for all connections of a
server. It's much easier to use this than to install a connect handler
which manually enables this for each connection.
Devices under /sys/firmware/ do not have subsystems. Hence, the
validation in sd_device_new_from_subsystem_sysname() ->
device_new_from_path_join() always failed.
Fixes a bug introduced by cd7c71154c
(v257).
Fixes#35861.
Supersedes #35308 (cherry-picked one commit and replaced the rest)
(I left a few comments that's folded by GitHub. Please make sure to
check them too.)
Let's add a simple flag that enables fd passing for all connections of a
server. It's much easier to use this than to install a connect handler
which manually enables this for each connection.
This moves https://src.fedoraproject.org/tests/systemd into upstream
so it's one less moving piece that we need to remember. While we're at
it, do various cleanups as well:
- Remove non-packit logic
- Set NO_BUILD in mkosi.local.conf as it's not picked up from the
process environment
- Don't set unused environment variables
- Remove workdir logic as we run in an ephemeral VM already
- Drop --verbose from meson test to reduce noise
from passed tests
- Use mkosi sandbox and drop duplicated deps from metadata file
- Stop running the unit tests as they already run as part of the rpm
build and as part of the integration tests.
- Various style fixes
The systemd rpms we try to install in packit have /usr/bin/bash and
/usr/bin/python3 as dependencies which breaks dnf5 because mkosi
doesn't download filelists metadata and
https://bugzilla.redhat.com/show_bug.cgi?id=2263771 is still not fixed
almost a year after being reported.
Devices under /sys/firmware/ do not have subsystems. Hence, the
validation in sd_device_new_from_subsystem_sysname() ->
device_new_from_path_join() always failed.
Fixes a bug introduced by cd7c71154c (v257).
Fixes#35861.
