Blob Blame History Raw
From 294efa52d47be083704da51b148c685d347be4ac Mon Sep 17 00:00:00 2001
From: Anita Zhang <the.anitazha@gmail.com>
Date: Tue, 8 Jun 2021 00:04:35 -0700
Subject: [PATCH] test: add extended test for triggering mount rate limit

It's hard to trigger the failure to exit the rate limit state in
isolation as it needs multiple event sources in order to show that it
gets stuck in the queue. Hence why this is an extended test.

(cherry picked from commit 0c81900965a72b29eb76e0737ed899b925ee75b6)

Related: #2094712
---
 test/TEST-60-MOUNT-RATELIMIT/Makefile     |  1 +
 test/TEST-60-MOUNT-RATELIMIT/test.sh      | 48 +++++++++++++++
 test/TEST-60-MOUNT-RATELIMIT/testsuite.sh | 73 +++++++++++++++++++++++
 3 files changed, 122 insertions(+)
 create mode 120000 test/TEST-60-MOUNT-RATELIMIT/Makefile
 create mode 100755 test/TEST-60-MOUNT-RATELIMIT/test.sh
 create mode 100755 test/TEST-60-MOUNT-RATELIMIT/testsuite.sh

diff --git a/test/TEST-60-MOUNT-RATELIMIT/Makefile b/test/TEST-60-MOUNT-RATELIMIT/Makefile
new file mode 120000
index 0000000000..e9f93b1104
--- /dev/null
+++ b/test/TEST-60-MOUNT-RATELIMIT/Makefile
@@ -0,0 +1 @@
+../TEST-01-BASIC/Makefile
\ No newline at end of file
diff --git a/test/TEST-60-MOUNT-RATELIMIT/test.sh b/test/TEST-60-MOUNT-RATELIMIT/test.sh
new file mode 100755
index 0000000000..e3c9288546
--- /dev/null
+++ b/test/TEST-60-MOUNT-RATELIMIT/test.sh
@@ -0,0 +1,48 @@
+#!/usr/bin/env bash
+set -e
+TEST_DESCRIPTION="Test that mount/unmount storms can enter/exit rate limit state and will not leak units"
+
+. $TEST_BASE_DIR/test-functions
+
+test_setup() {
+    create_empty_image
+    mkdir -p $TESTDIR/root
+    mount ${LOOPDEV}p1 $TESTDIR/root
+
+    (
+        LOG_LEVEL=5
+        eval $(udevadm info --export --query=env --name=${LOOPDEV}p2)
+
+        setup_basic_environment
+
+        # mask some services that we do not want to run in these tests
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-hwdb-update.service
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-journal-catalog-update.service
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-networkd.service
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-networkd.socket
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-resolved.service
+        ln -fs /dev/null $initdir/etc/systemd/system/systemd-machined.service
+
+        # setup the testsuite service
+        cat >$initdir/etc/systemd/system/testsuite.service <<EOF
+[Unit]
+Description=Testsuite service
+
+[Service]
+ExecStart=/bin/bash -x /testsuite.sh
+Type=oneshot
+StandardOutput=tty
+StandardError=tty
+NotifyAccess=all
+EOF
+        cp testsuite.sh $initdir/
+
+        setup_testsuite
+    ) || return 1
+    setup_nspawn_root
+
+    ddebug "umount $TESTDIR/root"
+    umount $TESTDIR/root
+}
+
+do_test "$@"
diff --git a/test/TEST-60-MOUNT-RATELIMIT/testsuite.sh b/test/TEST-60-MOUNT-RATELIMIT/testsuite.sh
new file mode 100755
index 0000000000..8158754667
--- /dev/null
+++ b/test/TEST-60-MOUNT-RATELIMIT/testsuite.sh
@@ -0,0 +1,73 @@
+#!/usr/bin/env bash
+set -eux
+set -o pipefail
+
+systemd-analyze log-level debug
+systemd-analyze log-target journal
+
+NUM_DIRS=20
+
+# mount/unmount enough times to trigger the /proc/self/mountinfo parsing rate limiting
+
+for ((i = 0; i < NUM_DIRS; i++)); do
+    mkdir "/tmp/meow${i}"
+done
+
+for ((i = 0; i < NUM_DIRS; i++)); do
+    mount -t tmpfs tmpfs "/tmp/meow${i}"
+done
+
+systemctl daemon-reload
+systemctl list-units -t mount tmp-meow* | grep -q tmp-meow
+
+for ((i = 0; i < NUM_DIRS; i++)); do
+    umount "/tmp/meow${i}"
+done
+
+# figure out if we have entered the rate limit state
+
+exited_rl=0
+timeout="$(date -ud "2 minutes" +%s)"
+while [[ $(date -u +%s) -le ${timeout} ]]; do
+    if journalctl -u init.scope | grep -q "(mount-monitor-dispatch) entered rate limit"; then
+        entered_rl=1
+        break
+    fi
+    sleep 5
+done
+
+# if the infra is slow we might not enter the rate limit state; in that case skip the exit check
+
+if [ "${entered_rl}" = "1" ]; then
+    exited_rl=0
+    timeout="$(date -ud "2 minutes" +%s)"
+    while [[ $(date -u +%s) -le ${timeout} ]]; do
+        if journalctl -u init.scope | grep -q "(mount-monitor-dispatch) left rate limit"; then
+            exited_rl=1
+            break
+        fi
+        sleep 5
+    done
+
+    if [ "${exited_rl}" = "0" ]; then
+        exit 24
+    fi
+fi
+
+# give some time for units to settle so we don't race between exiting the rate limit state and cleaning up the units
+
+sleep 60
+systemctl daemon-reload
+sleep 60
+
+# verify that the mount units are always cleaned up at the end
+
+if systemctl list-units -t mount tmp-meow* | grep -q tmp-meow; then
+    exit 42
+fi
+
+systemd-analyze log-level info
+
+echo OK >/testok
+
+exit 0