This reverts commit 1a0e065e9f.
This does not work as expected.
After `losetup --detach`, the kernel lazily removes the loop device.
But, systemd-dissect should gracefully handle that. If it does not, then
it is a bug in systemd-dissect.
Let's not hide the real issue in systemd-dissect.
Set both timeouts to some reasonable values instead of just `infinity`
to provide some form of a safe-net in case the test goes haywire and the
environment didn't set the timeouts itself (like our CIs do).
CIs set QEMU and nspawn timeouts by themselves which reflect their needs
and possibilities, so let's respect that value, instead of using one
pre-set value which might or might not work for all of them.
Both Ubuntu CI and CentOS CI set these values themselves.
to avoid scope clashing in case some previous test crashed and/or didn't
clean up properly. Currently all test machines are called `root`, since
the name is automagically derived from the container path (in this case
`/var/tmp/systemd-test.XXXXXX/root`).
E.g. (from Ubuntu CI):
```
[23:10:12] --x-- Running TEST-71-HOSTNAME --x--
make: Entering directory '/tmp/autopkgtest.5LjnBV/build.0mE/systemd/test/TEST-71-HOSTNAME'
+ make -C TEST-71-HOSTNAME setup run
TEST-71-HOSTNAME SETUP: test hostnamed
Reusing existing cached image /tmp/autopkgtest.5LjnBV/build.0mE/systemd/test/TEST-71-HOSTNAME/../default.img → /tmp/autopkgtest.5LjnBV/build.0mE/systemd/test/default.img
'/var/tmp/systemd-test.1yy2SS/default.img' -> '/tmp/autopkgtest.5LjnBV/build.0mE/systemd/test/default.img'
I: Masking supporting services
'/var/tmp/systemd-test.1yy2SS/root/etc/systemd/system/systemd-hwdb-update.service' -> '/dev/null'
'/var/tmp/systemd-test.1yy2SS/root/etc/systemd/system/systemd-journal-catalog-update.service' -> '/dev/null'
'/var/tmp/systemd-test.1yy2SS/root/etc/systemd/system/systemd-networkd.service' -> '/dev/null'
'/var/tmp/systemd-test.1yy2SS/root/etc/systemd/system/systemd-networkd.socket' -> '/dev/null'
'/var/tmp/systemd-test.1yy2SS/root/etc/systemd/system/systemd-resolved.service' -> '/dev/null'
TEST-71-HOSTNAME RUN: test hostnamed
+ env --unset=UNIFIED_CGROUP_HIERARCHY --unset=SYSTEMD_NSPAWN_UNIFIED_HIERARCHY timeout --foreground 1200 /bin/systemd-nspawn --register=no --kill-signal=SIGKILL --directory=/var/tmp/systemd-test.1yy2SS/root --setenv=SYSTEMD_UNIT_PATH=/usr/lib/systemd/tests/testdata/testsuite-71.units:/usr/lib/systemd/tests/testdata/units: /lib/systemd/systemd systemd.unit=testsuite.target systemd.wants=testsuite-71.service systemd.wants=end.service
Spawning container root on /var/tmp/systemd-test.1yy2SS/root.
Press ^] three times within 1s to kill container.
Failed to allocate scope: Unit root.scope already exists.
E: nspawn failed with exit code 1
```
When we kill a cgroup that is towards the end of the sorted candidate
list (i.e. when we have to resort to killing a candidate with
ManagedOOMPreference=avoid), this cgroup is not logged in the candidate
list. This is due to an off-by-one error when assigning dump_until.
Add a new test function, test_oomd_fetch_cgroup_oom_preference, to test
the ManagedOOMPreference logic. For starters, cut the relevant tests out
of test_oomd_cgroup_context_acquire_and_insert, and add them to the new
function. Then, expand these tests to cover the new behavior.
The ManagedOOMPreference property is only honored on cgroups which are
owned by root. This precludes anyone from setting ManagedOOMPreference
on cgroups managed by user managers.
Loosen this restriction in the following way: when processing a
monitored cgroup for kill candidates, honor the ManagedOOMPreference
setting if the monitored cgroup and cgroup candidate are owned by the
same user. This allows unprivileged users to configure
ManagedOOMPreference on their cgroups without affecting the kill
priority of ancestor cgroups.
N.B. that since swap kill operates globally to kill the largest
candidate, it is not appropriate to apply this logic to the swap kill
scenario. Therefore, the existing restriction on ManagedOOMPreference
will remain when calculating candidates for swap kill.
Add a new function, oomd_fetch_cgroup_oom_preference, to assist with
this new logic. To simplify things, move the `user.oomd_{avoid,omit}`
xattr reads to this function so that the xattr reads and uid checks are
performed all at once.
The extended attributes 'user.oomd_avoid' and 'user.oomd_omit' are used
to store the setting of the ManagedOOMPrefernce property. Since these
are user extended attributes, the access permissions are defined by the
file permissions, and there is no need to guard this from user managers.
Yu Watanabe
adrian5
Luca Boccassi
Uriel Corfa
Swapnil Devesh
Frantisek Sumsal
Devendra Tewari
Nick Rosbrook
Sonali Srivastava