Skip to contentSkip to conversation

cfs_bandwidth01: Fix cleanup on failure in set_cpu_quota

PATCH New Joerg Vehlow 3 participants 6 messages Aug 31, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 cfs_bandwidth01: Fix cleanup on failure in set_cpu_quota

From: Joerg Vehlow <[email protected]>

If set_cpu_quota failed, mk_cpu_cgroup did not return and cg_workers[n] was not set. This lead to a failure during cleanup, because the worker cgroups were not deleted.

---
 .../sched/cfs-scheduler/cfs_bandwidth01.c       | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +7 -10
index 9301ee458..e8032d65a 100644
@@ -57,17 +57,14 @@ static void set_cpu_quota(const struct tst_cgroup_group *const cg,
 		tst_cgroup_group_name(cg), quota_us, period_us);
 }
 
-static struct tst_cgroup_group *
-mk_cpu_cgroup(const struct tst_cgroup_group *const cg_parent,
+static void mk_cpu_cgroup(struct tst_cgroup_group **cg,
+          const struct tst_cgroup_group *const cg_parent,
 	      const char *const cg_child_name,
 	      const float quota_percent)
 {
-	struct tst_cgroup_group *const cg =
-		tst_cgroup_group_mk(cg_parent, cg_child_name);
+	*cg = tst_cgroup_group_mk(cg_parent, cg_child_name);
 
-	set_cpu_quota(cg, quota_percent);
-
-	return cg;
+	set_cpu_quota(*cg, quota_percent);
 }
 
 static void busy_loop(const unsigned int sleep_ms)
@@ -142,11 +139,11 @@ static void setup(void)
 	cg_level2 = tst_cgroup_group_mk(cg_test, "level2");
 
 	cg_level3a = tst_cgroup_group_mk(cg_level2, "level3a");
-	cg_workers[0] = mk_cpu_cgroup(cg_level3a, "worker1", 30);
-	cg_workers[1] = mk_cpu_cgroup(cg_level3a, "worker2", 20);
+	mk_cpu_cgroup(&cg_workers[0], cg_level3a, "worker1", 30);
+	mk_cpu_cgroup(&cg_workers[1], cg_level3a, "worker2", 20);
 
 	cg_level3b = tst_cgroup_group_mk(cg_level2, "level3b");
-	cg_workers[2] = mk_cpu_cgroup(cg_level3b, "worker3", 30);
+	mk_cpu_cgroup(&cg_workers[2], cg_level3b, "worker3", 30);
 }
 
 static void cleanup(void)
Signature · 1 line
2.25.1
#1 lore raw

PATCH 2/2 cfs_bandwidth01: Add misssing needs_kconfigs

From: Joerg Vehlow <[email protected]>

The test requires CONFIG_CFS_BANDWIDTH, otherwise it fails with
cfs_bandwidth01.c:51: TBROK: openat(13</sys/fs/cgroup/cpu,cpuacct/ltp/test-5666/level2/level3a/worker1>, 'cpu.cfs_period_us', O_WRONLY): ENOENT (2)

---
 testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c | 4 ++++
 1 file changed, 4 insertions(+)
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +4 -0
index e8032d65a..001fb2685 100644
@@ -178,6 +178,10 @@ static struct tst_test test = {
 	.forks_child = 1,
 	.needs_checkpoints = 1,
 	.taint_check = TST_TAINT_W | TST_TAINT_D,
+	.needs_kconfigs = (const char *[]) {
+		"CONFIG_CFS_BANDWIDTH",
+		NULL
+	},
 	.tags = (const struct tst_tag[]) {
 		{"linux-git", "39f23ce07b93"},
 		{"linux-git", "b34cb07dde7c"},
Signature · 1 line
2.25.1
#3 lore raw

PATCH 1/2 cfs_bandwidth01: Fix cleanup on failure in set_cpu_quota

In reply to Joerg Vehlow

Hello Joerg,

Quoted 16 lines · Joerg Vehlow

From: Joerg Vehlow <[email protected]>

If set_cpu_quota failed, mk_cpu_cgroup did not return and cg_workers[n] was not set. This lead to a failure during cleanup, because the worker cgroups were not deleted.

---
 .../sched/cfs-scheduler/cfs_bandwidth01.c       | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +0 -0
index 9301ee458..e8032d65a 100644

testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c static void set_cpu_quota(const struct tst_cgroup_group *const cg, · line 61

Quoted 7 lines
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +2 -2
@@ -57,17 +57,14 @@ static void set_cpu_quota(const struct tst_cgroup_group *const cg,
 		tst_cgroup_group_name(cg), quota_us, period_us);
 }
 
-static struct tst_cgroup_group *
-mk_cpu_cgroup(const struct tst_cgroup_group *const cg_parent,
+static void mk_cpu_cgroup(struct tst_cgroup_group **cg,
+          const struct tst_cgroup_group *const cg_parent,

Seems like there are spaces instead of a tab at the start.

Otherwise LGTM, thanks.

Signature · 2 lines
Thank you,
Richard.
#4 lore raw

PATCH 1/2 cfs_bandwidth01: Fix cleanup on failure in set_cpu_quota

Hi Richard,

Quoted 23 lines · Richard Palethorpe

Hello Joerg,

Quoted 19 lines · Joerg Vehlow

From: Joerg Vehlow <[email protected]>

If set_cpu_quota failed, mk_cpu_cgroup did not return and cg_workers[n] was not set. This lead to a failure during cleanup, because the worker cgroups were not deleted.

---
  .../sched/cfs-scheduler/cfs_bandwidth01.c       | 17 +++++++----------
  1 file changed, 7 insertions(+), 10 deletions(-)
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +0 -0
index 9301ee458..e8032d65a 100644
@@ -57,17 +57,14 @@ static void set_cpu_quota(const struct tst_cgroup_group *const cg,
  		tst_cgroup_group_name(cg), quota_us, period_us);
  }
 
Quoted 5 lines

testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c static void set_cpu_quota(const struct tst_cgroup_group *const cg, · line 61

Quoted 4 lines
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +2 -2
-static struct tst_cgroup_group *
-mk_cpu_cgroup(const struct tst_cgroup_group *const cg_parent,
+static void mk_cpu_cgroup(struct tst_cgroup_group **cg,
+          const struct tst_cgroup_group *const cg_parent,

Seems like there are spaces instead of a tab at the start.

Right, I wanted to keep the same format you had, but did not realize it was one tab and 5 spaces instead of 9 spaces.
I guess this can be fixed during merging.

Quoted 4 lines

Otherwise LGTM, thanks.

Thanks

Joerg

#5 lore raw

PATCH 1/2 cfs_bandwidth01: Fix cleanup on failure in set_cpu_quota

Hi Joerg, Richard,

Quoted 8 lines
Quoted 5 lines

testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c static void set_cpu_quota(const struct tst_cgroup_group *const cg, · line 61

Quoted 4 lines
testcases/kernel/sched/cfs-scheduler/cfs_bandwidth01.c +2 -2
-static struct tst_cgroup_group *
-mk_cpu_cgroup(const struct tst_cgroup_group *const cg_parent,
+static void mk_cpu_cgroup(struct tst_cgroup_group **cg,
+          const struct tst_cgroup_group *const cg_parent,

Seems like there are spaces instead of a tab at the start.

Quoted 4 lines

Right, I wanted to keep the same format you had, but did not realize it was one tab and 5 spaces instead of 9 spaces.
I guess this can be fixed during merging.

I fixed that and pushed those patches. Thanks!

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20210831/33d6120f/attachment-0001.htm>