1
0
mirror of https://github.com/systemd/systemd.git synced 2025-01-11 09:18:07 +03:00
systemd/test/units/testsuite-55.sh
Luca Boccassi bb5232b6a3 core: add systemd-executor binary
Currently we spawn services by forking a child process, doing a bunch
of work, and then exec'ing the service executable.

There are some advantages to this approach:

- quick: we immediately have access to all the enourmous amount of
  state simply by virtue of sharing the memory with the parent
- easy to refactor and add features
- part of the same binary, will never be out of sync

There are however significant drawbacks:

- doing work after fork and before exec is against glibc's supported
  case for several APIs we call
- copy-on-write trap: anytime any memory is touched in either parent
  or child, a copy of that page will be triggered
- memory footprint of the child process will be memory footprint of
  PID1, but using the cgroup memory limits of the unit

The last issue is especially problematic on resource constrained
systems where hard memory caps are enforced and swap is not allowed.
As soon as PID1 is under load, with no page out due to no swap, and a
service with a low MemoryMax= tries to start, hilarity ensues.

Add a new systemd-executor binary, that is able to receive all the
required state via memfd, deserialize it, prepare the appropriate
data structures and call exec_child.

Use posix_spawn which uses CLONE_VM + CLONE_VFORK, to ensure there is
no copy-on-write (same address space will be used, and parent process
will be frozen, until exec).
The sd-executor binary is pinned by FD on startup, so that we can
guarantee there will be no incompatibilities during upgrades.
2023-10-12 15:01:51 +01:00

184 lines
6.1 KiB
Bash
Executable File

#!/usr/bin/env bash
# SPDX-License-Identifier: LGPL-2.1-or-later
set -eux
set -o pipefail
systemd-analyze log-level debug
# Ensure that the init.scope.d drop-in is applied on boot
test "$(cat /sys/fs/cgroup/init.scope/memory.high)" != "max"
# Loose checks to ensure the environment has the necessary features for systemd-oomd
[[ -e /proc/pressure ]] || echo "no PSI" >>/skipped
cgroup_type="$(stat -fc %T /sys/fs/cgroup/)"
if [[ "$cgroup_type" != *"cgroup2"* ]] && [[ "$cgroup_type" != *"0x63677270"* ]]; then
echo "no cgroup2" >>/skipped
fi
if [ ! -f /usr/lib/systemd/systemd-oomd ] && [ ! -f /lib/systemd/systemd-oomd ]; then
echo "no oomd" >>/skipped
fi
if [[ -e /skipped ]]; then
exit 0
fi
rm -rf /run/systemd/system/testsuite-55-testbloat.service.d
# Activate swap file if we are in a VM
if systemd-detect-virt --vm --quiet; then
mkswap /swapfile
swapon /swapfile
swapon --show
fi
# Configure oomd explicitly to avoid conflicts with distro dropins
mkdir -p /run/systemd/oomd.conf.d/
cat >/run/systemd/oomd.conf.d/99-oomd-test.conf <<EOF
[OOM]
DefaultMemoryPressureDurationSec=2s
EOF
mkdir -p /run/systemd/system/-.slice.d/
cat >/run/systemd/system/-.slice.d/99-oomd-test.conf <<EOF
[Slice]
ManagedOOMSwap=auto
EOF
mkdir -p /run/systemd/system/user@.service.d/
cat >/run/systemd/system/user@.service.d/99-oomd-test.conf <<EOF
[Service]
ManagedOOMMemoryPressure=auto
ManagedOOMMemoryPressureLimit=0%
EOF
mkdir -p /run/systemd/system/systemd-oomd.service.d/
cat >/run/systemd/system/systemd-oomd.service.d/debug.conf <<EOF
[Service]
Environment=SYSTEMD_LOG_LEVEL=debug
EOF
systemctl daemon-reload
# enable the service to ensure dbus-org.freedesktop.oom1.service exists
# and D-Bus activation works
systemctl enable systemd-oomd.service
# if oomd is already running for some reasons, then restart it to make sure the above settings to be applied
if systemctl is-active systemd-oomd.service; then
systemctl restart systemd-oomd.service
fi
# Ensure that we can start services even with a very low hard memory cap without oom-kills, but skip under
# sanitizers as they balloon memory usage.
if ! [[ -v ASAN_OPTIONS || -v UBSAN_OPTIONS ]]; then
systemd-run -t -p MemoryMax=10M -p MemorySwapMax=0 -p MemoryZSwapMax=0 /bin/true
fi
systemctl start testsuite-55-testchill.service
systemctl start testsuite-55-testbloat.service
# Verify systemd-oomd is monitoring the expected units
# Try to avoid racing the oomctl output check by checking in a loop with a timeout
oomctl_output=$(oomctl)
timeout="$(date -ud "1 minutes" +%s)"
while [[ $(date -u +%s) -le $timeout ]]; do
if grep "/testsuite-55-workload.slice" <<< "$oomctl_output"; then
break
fi
oomctl_output=$(oomctl)
sleep 1
done
grep "/testsuite-55-workload.slice" <<< "$oomctl_output"
grep "20.00%" <<< "$oomctl_output"
grep "Default Memory Pressure Duration: 2s" <<< "$oomctl_output"
systemctl status testsuite-55-testchill.service
# systemd-oomd watches for elevated pressure for 2 seconds before acting.
# It can take time to build up pressure so either wait 2 minutes or for the service to fail.
timeout="$(date -ud "2 minutes" +%s)"
while [[ $(date -u +%s) -le $timeout ]]; do
if ! systemctl status testsuite-55-testbloat.service; then
break
fi
oomctl
sleep 2
done
# testbloat should be killed and testchill should be fine
if systemctl status testsuite-55-testbloat.service; then exit 42; fi
if ! systemctl status testsuite-55-testchill.service; then exit 24; fi
# Make sure we also work correctly on user units.
systemctl start --machine "testuser@.host" --user testsuite-55-testchill.service
systemctl start --machine "testuser@.host" --user testsuite-55-testbloat.service
# Verify systemd-oomd is monitoring the expected units
# Try to avoid racing the oomctl output check by checking in a loop with a timeout
oomctl_output=$(oomctl)
timeout="$(date -ud "1 minutes" +%s)"
while [[ $(date -u +%s) -le $timeout ]]; do
if grep -E "/user.slice.*/testsuite-55-workload.slice" <<< "$oomctl_output"; then
break
fi
oomctl_output=$(oomctl)
sleep 1
done
grep -E "/user.slice.*/testsuite-55-workload.slice" <<< "$oomctl_output"
grep "20.00%" <<< "$oomctl_output"
grep "Default Memory Pressure Duration: 2s" <<< "$oomctl_output"
systemctl --machine "testuser@.host" --user status testsuite-55-testchill.service
# systemd-oomd watches for elevated pressure for 2 seconds before acting.
# It can take time to build up pressure so either wait 2 minutes or for the service to fail.
timeout="$(date -ud "2 minutes" +%s)"
while [[ $(date -u +%s) -le $timeout ]]; do
if ! systemctl --machine "testuser@.host" --user status testsuite-55-testbloat.service; then
break
fi
oomctl
sleep 2
done
# testbloat should be killed and testchill should be fine
if systemctl --machine "testuser@.host" --user status testsuite-55-testbloat.service; then exit 42; fi
if ! systemctl --machine "testuser@.host" --user status testsuite-55-testchill.service; then exit 24; fi
# only run this portion of the test if we can set xattrs
if setfattr -n user.xattr_test -v 1 /sys/fs/cgroup/; then
sleep 120 # wait for systemd-oomd kill cool down and elevated memory pressure to come down
mkdir -p /run/systemd/system/testsuite-55-testbloat.service.d/
cat >/run/systemd/system/testsuite-55-testbloat.service.d/override.conf <<EOF
[Service]
ManagedOOMPreference=avoid
EOF
systemctl daemon-reload
systemctl start testsuite-55-testchill.service
systemctl start testsuite-55-testmunch.service
systemctl start testsuite-55-testbloat.service
timeout="$(date -ud "2 minutes" +%s)"
while [[ "$(date -u +%s)" -le "$timeout" ]]; do
if ! systemctl status testsuite-55-testmunch.service; then
break
fi
oomctl
sleep 2
done
# testmunch should be killed since testbloat had the avoid xattr on it
if ! systemctl status testsuite-55-testbloat.service; then exit 25; fi
if systemctl status testsuite-55-testmunch.service; then exit 43; fi
if ! systemctl status testsuite-55-testchill.service; then exit 24; fi
fi
systemd-analyze log-level info
touch /testok