[PATCH 0/2] selftests/cgroup: misc cleanups for test_cpu.c

[PATCH 0/2] selftests/cgroup: misc cleanups for test_cpu.c

[Shaojie Sun]

This series contains two minor cleanups to the cgroup cpu
controller selftests:

  - Patch 1 removes a redundant user_usec read in test_cpucg_nice
    that is never used after being read.

  - Patch 2 makes test_cpucg_max and test_cpucg_max_nested write
    both quota and period explicitly to cpu.max instead of relying
    on the kernel default period, making the tests self-documenting
    and the remainder calculation genuinely useful. Misleading
    variable names are also cleaned up.

Shaojie Sun (2):
  selftests/cgroup: remove unused user_usec read in test_cpucg_nice
  selftests/cgroup: write explicit period to cpu.max in max tests

 tools/testing/selftests/cgroup/test_cpu.c | 27 ++++++++++++-----------
 1 file changed, 14 insertions(+), 13 deletions(-)

-- 
2.25.1

[PATCH 1/2] selftests/cgroup: remove unused user_usec read in test_cpucg_nice

[Shaojie Sun]

In test_cpucg_nice, after the child process exits, user_usec is read
from cpu.stat but never used. Only nice_usec is validated against the
expected value. Remove this redundant read.

Signed-off-by: Shaojie Sun <sunshaojie@kylinos.cn>
---
 tools/testing/selftests/cgroup/test_cpu.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tools/testing/selftests/cgroup/test_cpu.c b/tools/testing/selftests/cgroup/test_cpu.c
index 7a40d76b9548..fe637e1d2fcd 100644
--- a/tools/testing/selftests/cgroup/test_cpu.c
+++ b/tools/testing/selftests/cgroup/test_cpu.c
@@ -289,7 +289,6 @@ static int test_cpucg_nice(const char *root)
 		if (!WIFEXITED(status))
 			goto cleanup;
 
-		user_usec = cg_read_key_long(cpucg, "cpu.stat", "user_usec");
 		nice_usec = cg_read_key_long(cpucg, "cpu.stat", "nice_usec");
 		if (!values_close_report(nice_usec, expected_nice_usec, 1))
 			goto cleanup;
-- 
2.25.1

[PATCH 2/2] selftests/cgroup: write explicit period to cpu.max in max tests

[Shaojie Sun]

The cpu.max interface accepts "quota period" as two values, but
test_cpucg_max and test_cpucg_max_nested only wrote the quota,
relying on the kernel default period of 100ms. Since the test
duration is exactly divisible by the default period (1s / 100ms
= 10, no remainder), the remainder calculation was effectively
dead code.

Write both values explicitly so that the period can be easily
adjusted for testing purposes and the remainder logic becomes
genuinely useful.

While at it, rename misleading variable names:
  - default_period_usec -> period_usec (no longer relying on a
    kernel default, as it is now explicitly written to cpu.max)
  - quota_buf -> max_param_buf (since the buffer holds both quota
    and period)

Signed-off-by: Shaojie Sun <sunshaojie@kylinos.cn>
---
 tools/testing/selftests/cgroup/test_cpu.c | 26 ++++++++++++-----------
 1 file changed, 14 insertions(+), 12 deletions(-)

diff --git a/tools/testing/selftests/cgroup/test_cpu.c b/tools/testing/selftests/cgroup/test_cpu.c
index fe637e1d2fcd..10f9254cd08e 100644
--- a/tools/testing/selftests/cgroup/test_cpu.c
+++ b/tools/testing/selftests/cgroup/test_cpu.c
@@ -646,15 +646,16 @@ static int test_cpucg_max(const char *root)
 {
 	int ret = KSFT_FAIL;
 	long quota_usec = 1000;
-	long default_period_usec = 100000; /* cpu.max's default period */
+	long period_usec = 100000;
 	long duration_seconds = 1;
 
 	long duration_usec = duration_seconds * USEC_PER_SEC;
 	long usage_usec, n_periods, remainder_usec, expected_usage_usec;
 	char *cpucg;
-	char quota_buf[32];
+	char max_param_buf[32];
 
-	snprintf(quota_buf, sizeof(quota_buf), "%ld", quota_usec);
+	snprintf(max_param_buf, sizeof(max_param_buf), "%ld %ld", quota_usec,
+		 period_usec);
 
 	cpucg = cg_name(root, "cpucg_test");
 	if (!cpucg)
@@ -663,7 +664,7 @@ static int test_cpucg_max(const char *root)
 	if (cg_create(cpucg))
 		goto cleanup;
 
-	if (cg_write(cpucg, "cpu.max", quota_buf))
+	if (cg_write(cpucg, "cpu.max", max_param_buf))
 		goto cleanup;
 
 	struct cpu_hog_func_param param = {
@@ -685,8 +686,8 @@ static int test_cpucg_max(const char *root)
 	 * The following calculation applies only since
 	 * the cpu hog is set to run as per wall-clock time
 	 */
-	n_periods = duration_usec / default_period_usec;
-	remainder_usec = duration_usec - n_periods * default_period_usec;
+	n_periods = duration_usec / period_usec;
+	remainder_usec = duration_usec - n_periods * period_usec;
 	expected_usage_usec
 		= n_periods * quota_usec + MIN(remainder_usec, quota_usec);
 
@@ -710,15 +711,16 @@ static int test_cpucg_max_nested(const char *root)
 {
 	int ret = KSFT_FAIL;
 	long quota_usec = 1000;
-	long default_period_usec = 100000; /* cpu.max's default period */
+	long period_usec = 100000;
 	long duration_seconds = 1;
 
 	long duration_usec = duration_seconds * USEC_PER_SEC;
 	long usage_usec, n_periods, remainder_usec, expected_usage_usec;
 	char *parent, *child;
-	char quota_buf[32];
+	char max_param_buf[32];
 
-	snprintf(quota_buf, sizeof(quota_buf), "%ld", quota_usec);
+	snprintf(max_param_buf, sizeof(max_param_buf), "%ld %ld", quota_usec,
+		 period_usec);
 
 	parent = cg_name(root, "cpucg_parent");
 	child = cg_name(parent, "cpucg_child");
@@ -734,7 +736,7 @@ static int test_cpucg_max_nested(const char *root)
 	if (cg_create(child))
 		goto cleanup;
 
-	if (cg_write(parent, "cpu.max", quota_buf))
+	if (cg_write(parent, "cpu.max", max_param_buf))
 		goto cleanup;
 
 	struct cpu_hog_func_param param = {
@@ -756,8 +758,8 @@ static int test_cpucg_max_nested(const char *root)
 	 * The following calculation applies only since
 	 * the cpu hog is set to run as per wall-clock time
 	 */
-	n_periods = duration_usec / default_period_usec;
-	remainder_usec = duration_usec - n_periods * default_period_usec;
+	n_periods = duration_usec / period_usec;
+	remainder_usec = duration_usec - n_periods * period_usec;
 	expected_usage_usec
 		= n_periods * quota_usec + MIN(remainder_usec, quota_usec);
 
-- 
2.25.1

Re: [PATCH 2/2] selftests/cgroup: write explicit period to cpu.max in max tests

[Michal Koutný]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

Hi.

Here I cannot appreciate the changes.

On Wed, Jul 29, 2026 at 06:04:51PM +0800, Shaojie Sun <sunshaojie@kylinos.cn> wrote:
> The cpu.max interface accepts "quota period" as two values, but
> test_cpucg_max and test_cpucg_max_nested only wrote the quota,
> relying on the kernel default period of 100ms.

Yes. Is that a problem?

> Write both values explicitly so that the period can be easily
> adjusted for testing purposes and the remainder logic becomes
> genuinely useful.

I have yet to see a system where the default period is modified.
Therefore, I wouldn't open that box with non-default periods
unnecessarily.

Thanks,
Michal

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 265 bytes --]

Re: [PATCH 1/2] selftests/cgroup: remove unused user_usec read in test_cpucg_nice

[Michal Koutný]

[-- Attachment #1: Type: text/plain, Size: 928 bytes --]

On Wed, Jul 29, 2026 at 06:04:50PM +0800, Shaojie Sun <sunshaojie@kylinos.cn> wrote:
> diff --git a/tools/testing/selftests/cgroup/test_cpu.c b/tools/testing/selftests/cgroup/test_cpu.c
> index 7a40d76b9548..fe637e1d2fcd 100644
> --- a/tools/testing/selftests/cgroup/test_cpu.c
> +++ b/tools/testing/selftests/cgroup/test_cpu.c
> @@ -289,7 +289,6 @@ static int test_cpucg_nice(const char *root)
>  		if (!WIFEXITED(status))
>  			goto cleanup;
>  
> -		user_usec = cg_read_key_long(cpucg, "cpu.stat", "user_usec");
>  		nice_usec = cg_read_key_long(cpucg, "cpu.stat", "nice_usec");

Good catch.

This test is anologous to test_cpucg_stats(), so the checks should be
analogous too, like:

 		user_usec = cg_read_key_long(cpucg, "cpu.stat", "user_usec");
 		nice_usec = cg_read_key_long(cpucg, "cpu.stat", "nice_usec");
 		if (user_usec <= 0)
 			goto cleanup;  

instead of the removal.

Thanks,
Michal

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 265 bytes --]