[LTP] [PATCH v1 1/1] cpuhotplug03.sh: Poll for CPU migration instead of sleeping

[LTP] [PATCH v1 1/1] cpuhotplug03.sh: Poll for CPU migration instead of sleeping

[Benjamin Wheeler via ltp]

The test onlined the target CPU, slept a fixed one second and then
checked once whether the scheduler had migrated a cpuhotplug_do_spin_loop
process onto it. A blind sleep is both racy (the migration may not have
happened yet on a busy or slow system) and wasteful (it always waits the
full second even when the migration is immediate).

Replace the sleep and single check with tst_retry(), which retries the
check a few times until a process is seen on the onlined CPU. The
migration check is factored out into a check_cpu_migrated() helper.

tst_retry() is the retry helper available in the legacy shell test API
(test.sh) that this test uses; TST_RETRY_FUNC() only exists in the new
API (tst_test.sh), so it is not usable here without converting the whole
test, which is out of scope for this change.

The TBROK vs TFAIL distinction is preserved: if no spin loop processes
are running at all it is reported as a broken test (TBROK), whereas
processes running but none migrating to the onlined CPU remains a
genuine failure (TFAIL).

Signed-off-by: Benjamin Wheeler <benjamin.wheeler@canonical.com>
---
 .../cpu_hotplug/functional/cpuhotplug03.sh    | 43 ++++++++++++-------
 1 file changed, 27 insertions(+), 16 deletions(-)

diff --git a/testcases/kernel/hotplug/cpu_hotplug/functional/cpuhotplug03.sh b/testcases/kernel/hotplug/cpu_hotplug/functional/cpuhotplug03.sh
index 3a461e30c..1577f3e6e 100755
--- a/testcases/kernel/hotplug/cpu_hotplug/functional/cpuhotplug03.sh
+++ b/testcases/kernel/hotplug/cpu_hotplug/functional/cpuhotplug03.sh
@@ -50,6 +50,20 @@ do_clean()
 	set_all_cpu_states "$cpu_states"
 }

+# check_cpu_migrated()
+#
+#  Returns 0 if at least one cpuhotplug_do_spin_loop process is running
+#  on CPU_TO_TEST, non-zero otherwise.
+#  Since procps v3.3.15, we need to accurately select command name by -C
+#  option, because procps cannot truncate normal command name to 15
+#  characters by default.
+check_cpu_migrated()
+{
+	ps -o psr -o command --no-headers -C cpuhotplug_do_s \
+		| sed -e "s/^ *//" | cut -d' ' -f 1 \
+		| grep -q "^${CPU_TO_TEST}$"
+}
+
 while getopts c:l: OPTION; do
 case $OPTION in
 	c)
@@ -120,21 +134,18 @@ until [ $LOOP_COUNT -gt $HOTPLUG03_LOOPS ]; do
 		tst_brkm TBROK "CPU${CPU_TO_TEST} cannot be onlined"
 	fi

-	sleep 1
-
-	# Verify at least one process has migrated to the new CPU
-	# Since procps v3.3.15, we need to accurately select command name
-	# by -C option, because procps cannot trucate normal command name
-	# to 15 characters by default).
-	ps -o psr -o command --no-headers -C cpuhotplug_do_s
-	if [ $? -ne 0 ]; then
-		tst_brkm TBROK "No cpuhotplug_do_spin_loop processes \
-			found on any processor"
-	fi
-	NUM=`ps -o psr -o command --no-headers -C cpuhotplug_do_s \
-		| sed -e "s/^ *//" | cut -d' ' -f 1 | grep "^${CPU_TO_TEST}$" \
-		| wc -l`
-	if [ $NUM -lt 1 ]; then
+	# Verify at least one process has migrated to the new CPU.
+	# Retry the check in a loop instead of blindly sleeping, since the
+	# scheduler may take a while to migrate a task to the freshly
+	# onlined CPU.
+	if ! tst_retry "check_cpu_migrated"; then
+		# Distinguish a broken workload (no spin loop processes running
+		# at all) from a genuine failure (processes running, but none
+		# migrated to the onlined CPU).
+		if ! ps --no-headers -C cpuhotplug_do_s >/dev/null; then
+			tst_brkm TBROK "No cpuhotplug_do_spin_loop processes \
+				found on any processor"
+		fi
 		tst_resm TFAIL "No cpuhotplug_do_spin_loop processes found on \
 			CPU${CPU_TO_TEST}"
 		tst_exit
@@ -145,7 +156,7 @@ until [ $LOOP_COUNT -gt $HOTPLUG03_LOOPS ]; do
 	LOOP_COUNT=$((LOOP_COUNT+1))
 done

-tst_resm TPASS "$NUM cpuhotplug_do_spin_loop processes found on \
+tst_resm TPASS "cpuhotplug_do_spin_loop process(es) migrated to \
 	CPU${CPU_TO_TEST}"

 tst_exit
--
2.43.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] cpuhotplug03.sh: Poll for CPU migration instead of sleeping

[linuxtestproject.agent]

Hi Benjamin,

On Tue, 14 Jul 2026 12:28:34 -0400, Benjamin Wheeler wrote:
> cpuhotplug03.sh: Poll for CPU migration instead of sleeping

Verdict - Reviewed

---
Note:

The agent can sometimes produce false positives although often its
findings are genuine. If you find issues with the review, please
comment this email or ignore the suggestions.

Regards,
LTP AI Reviewer

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v1 1/1] cpuhotplug03.sh: Poll for CPU migration instead of sleeping

[Cyril Hrubis]

Hi!
Reviewed-by: Cyril Hrubis <chrubis@suse.cz>

-- 
Cyril Hrubis
chrubis@suse.cz

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp