All of lore.kernel.org
 help / color / mirror / Atom feed
* [LTP] [PATCH] memcg_use_hierarchy_test.sh: skip setting use_hierarchy if not available
@ 2020-09-11  9:21 Chen Hanxiao
  2020-09-17  5:21 ` Yang Xu
  0 siblings, 1 reply; 6+ messages in thread
From: Chen Hanxiao @ 2020-09-11  9:21 UTC (permalink / raw)
  To: ltp

The precondition of this case is that we can disable use_hierarchy.
But some distributions such as CentOS 8 had enabled it in root
cgroup and hard to disabled.

As [1] describe:
    NOTE1: Enabling/disabling will fail if either the cgroup already has other
       cgroups created below it, or if the parent cgroup has use_hierarchy
       enabled.

We should check the precondition before testing.

[1] https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt

Signed-off-by: Chen Hanxiao <chenhx.fnst@cn.fujitsu.com>
---
 .../controllers/memcg/functional/memcg_lib.sh | 25 +++++++++++++++----
 .../functional/memcg_use_hierarchy_test.sh    | 12 +++++++--
 2 files changed, 30 insertions(+), 7 deletions(-)

diff --git a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
index 22ef4f5e2..4f99c6f59 100755
--- a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
+++ b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
@@ -491,6 +491,10 @@ cleanup_test()
 		orig_memory_use_hierarchy=""
 	fi
 
+	if [ -n "$root_memory_use_hierarchy" ];then
+                root_memory_use_hierarchy=""
+	fi
+
 	killall -9 memcg_process 2>/dev/null
 	wait
 
@@ -514,16 +518,27 @@ setup_test()
 	# while there are distributions (RHEL7U0Beta for example) that sets
 	# it to 1.
 	orig_memory_use_hierarchy=$(cat /dev/memcg/memory.use_hierarchy)
+        MEMCGROUP_PATH="/sys/fs/cgroup/memory"
+        if [ -e "$MEMCGROUP_PATH" ];then
+                root_memory_use_hierarchy=$(cat "$MEMCGROUP_PATH/memory.use_hierarchy")
+        fi
+
 	if [ -z "$orig_memory_use_hierarchy" ];then
 		tst_resm TINFO "cat /dev/memcg/memory.use_hierarchy failed"
 	elif [ "$orig_memory_use_hierarchy" = "0" ];then
 		orig_memory_use_hierarchy=""
 	else
-		echo 0 > /dev/memcg/memory.use_hierarchy
-		if [ $? -ne 0 ];then
-			tst_resm TINFO "set /dev/memcg/memory.use_hierarchy" \
-				"to 0 failed"
-		fi
+                if [ "$root_memory_use_hierarchy" = "1" ]; then
+                        tst_resm TINFO "root cgroup has use_hierarchy enabled, " \
+                            "can't set /dev/memcg/memory.use_hierarchy to 0"
+                        export root_memory_use_hierarchy
+                else
+                        echo 0 > /dev/memcg/memory.use_hierarchy
+                        if [ $? -ne 0 ];then
+                                tst_resm TINFO "set /dev/memcg/memory.use_hierarchy" \
+                                    "to 0 failed"
+                        fi
+                fi
 	fi
 
 	ROD mkdir "/dev/memcg/$TEST_ID"
diff --git a/testcases/kernel/controllers/memcg/functional/memcg_use_hierarchy_test.sh b/testcases/kernel/controllers/memcg/functional/memcg_use_hierarchy_test.sh
index 4cf6b9fc2..1439b6352 100755
--- a/testcases/kernel/controllers/memcg/functional/memcg_use_hierarchy_test.sh
+++ b/testcases/kernel/controllers/memcg/functional/memcg_use_hierarchy_test.sh
@@ -34,7 +34,9 @@ TST_TOTAL=3
 # test if one of the ancestors goes over its limit, the proces will be killed
 testcase_1()
 {
-	echo 1 > memory.use_hierarchy
+        if [ "$root_memory_use_hierarchy" != "1" ]; then
+                echo 1 > memory.use_hierarchy
+        fi
 	echo $PAGESIZE > memory.limit_in_bytes
 
 	mkdir subgroup
@@ -48,6 +50,10 @@ testcase_1()
 # test Enabling will fail if the cgroup already has other cgroups
 testcase_2()
 {
+        if [ "$root_memory_use_hierarchy" = "1" ]; then
+               tst_resm TCONF "root cgroup has use_hierarchy enabled, skip"
+               return
+        fi
 	mkdir subgroup
 	EXPECT_FAIL echo 1 \> memory.use_hierarchy
 
@@ -57,7 +63,9 @@ testcase_2()
 # test disabling will fail if the parent cgroup has enabled hierarchy.
 testcase_3()
 {
-	echo 1 > memory.use_hierarchy
+        if [ "$root_memory_use_hierarchy" != "1" ]; then
+               echo 1 > memory.use_hierarchy
+        fi
 	mkdir subgroup
 	EXPECT_FAIL echo 0 \> subgroup/memory.use_hierarchy
 
-- 
2.26.0.windows.1




^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [LTP] 答复:  [PATCH] memcg_use_hierarchy_test.sh: skip setting use_hierarchy if not available
@ 2020-10-14  7:31 Chen, Hanxiao
  0 siblings, 0 replies; 6+ messages in thread
From: Chen, Hanxiao @ 2020-10-14  7:31 UTC (permalink / raw)
  To: ltp



Hi, Li

>???: Li Wang <liwang@redhat.com>
>????: 2020?10?14? 14:28
>???: Xu, Yang <xuyang2018.jy@cn.fujitsu.com>; Chen, Hanxiao <chenhx.fnst@cn.fujitsu.com>
>??: LTP List <ltp@lists.linux.it>
>??: Re: ??: [LTP] [PATCH] memcg_use_hierarchy_test.sh: skip setting use_hierarchy if not available
>
>
>
>On Tue, Oct 13, 2020 at 11:19 AM Yang Xu <xuyang2018.jy@cn.fujitsu.com> wrote:
>Hi Hanxiao
>
>> Hi, Yang
>>
>>> -----??????-----
>>> ???: Re: [LTP] [PATCH] memcg_use_hierarchy_test.sh: skip setting
>>> use_hierarchy if not available
[snip]
>> The behavior above looks conflicting with https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt.
>Yes. Centos7 behavior is different from the documentation.
>>> So why not directly use /dev/memcg/memory.use_hierarchy value to judge in
>>> testcase after setting 0.
>> In setup_test from memcg_lib.sh, we actually did:
>>     mount -t cgroup -omemory memcg /dev/memcg
>> Then kernel will find a suitable cgroup root for us in cgroup1_mount.
>>
>> In this case, /sys/fs/cgroup/memory/ would be the good choice.
>> So it's equivalent to read memory.use_hierarchy from either side.
>I understand. Only a minor suggestion, please use tabs instead of spaces
>at the beginning of the line.
>
>
>This patch looks good to me,
>Acked-by: Yang Xu <xuyang2018.jy@cn.fujitsu.com>
>
>@Li, I think this patch is ok, Do you have some comment about it?
>
>I'm ok to go with memory.use_hierarchy checking in the preconditioning phase.
>
>But how can you assert the default memory cgroup is mount at path: "/sys/fs/cgroup/memory", is there any possibility the default path mount at other places(for different distribution)?

https://github.com/torvalds/linux/blob/b5fc7a89e58bcc059a3d5e4db79c481fb437de59/kernel/cgroup/cgroup.c#L5768

WARN_ON(sysfs_create_mount_point(fs_kobj, "cgroup"));

So the default mount place should be /sys/fs/cgroup if distributions
did not modify these lines.

As the discussion above,
mount -t cgroup -omemory memcg /dev/memcg
kernel will find a suitable cgroup root for us.

How about reading /dev/memcg/memory.use_hierarchy as Yang recommended instead?

Regards,

-   Hanxiao

>--
>Regards,
>Li Wang


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20201014/2c23546c/attachment-0001.htm>

^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2020-10-14  7:31 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-09-11  9:21 [LTP] [PATCH] memcg_use_hierarchy_test.sh: skip setting use_hierarchy if not available Chen Hanxiao
2020-09-17  5:21 ` Yang Xu
2020-09-21 10:53   ` [LTP] 答复: " Chen, Hanxiao
2020-10-13  3:19     ` Yang Xu
2020-10-14  6:27       ` Li Wang
2020-10-14  7:31 Chen, Hanxiao

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.