[LTP] [PATCH] memcg_subgroup_charge: Remove memory limitation

[LTP] [PATCH] memcg_subgroup_charge: Remove memory limitation

[Joerg Vehlow]

From: Joerg Vehlow <joerg.vehlow@aox-tech.de>

The test is supposed to verify, that memory used by a task
is not migrated to a subgroup, when the task is migrated.
In order to test this, it is not required, to limit
the memory at all, because accounting is done anyway and
the test checks the memory statistics.

Therefore all memory limiting is removed. This also gets
rid of a faulty assumption when writing the test:
The test allocates 33 pages using mmap. It is expected,
that it does not need more than 33 * pagesize of memory.
But sometimes the kernel has to allocate new page tables.
This is accounted as kmem AND user memory.
(See kernel doc cgroup-v1/memory.txt section 2.7)
If this happend, the test failed, if the system has no swap,
because the oom killer kills the process.

Signed-off-by: Joerg Vehlow <joerg.vehlow@aox-tech.de>
---
 .../kernel/controllers/memcg/functional/memcg_lib.sh |  4 ----
 .../memcg/functional/memcg_subgroup_charge.sh        | 12 +-----------
 2 files changed, 1 insertion(+), 15 deletions(-)

diff --git a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
index aadaae4d2..aec8a00ab 100755
--- a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
+++ b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
@@ -404,13 +404,9 @@ test_hugepage()
 }
 
 # Test the memory charge won't move to subgroup
-# $1 - memory.limit_in_bytes in parent group
-# $2 - memory.limit_in_bytes in sub group
 test_subgroup()
 {
 	mkdir subgroup
-	echo $1 > memory.limit_in_bytes
-	echo $2 > subgroup/memory.limit_in_bytes
 
 	tst_resm TINFO "Running memcg_process --mmap-anon -s $PAGESIZES"
 	memcg_process --mmap-anon -s $PAGESIZES &
diff --git a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
index 9b11f7b58..98e5ad68d 100755
--- a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
+++ b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
@@ -26,7 +26,7 @@
 ################################################################################
 
 TCID="memcg_subgroup_charge"
-TST_TOTAL=3
+TST_TOTAL=1
 
 . memcg_lib.sh
 
@@ -36,15 +36,5 @@ testcase_1()
 	test_subgroup $PAGESIZES $((2*PAGESIZES))
 }
 
-testcase_2()
-{
-	test_subgroup $PAGESIZES $PAGESIZES
-}
-
-testcase_3()
-{
-	test_subgroup $PAGESIZES 0
-}
-
 run_tests
 tst_exit
-- 
2.20.1

[LTP] [PATCH] memcg_subgroup_charge: Remove memory limitation

[Joerg Vehlow]

Hi,

a little ping for this patch?
I recently updated the ltp suite for our tests to 20200930 and stumbled 
over this error again.
Any reason for not applying it?

J?rg

On 11/6/2019 7:18 AM, Joerg Vehlow wrote:
> From: Joerg Vehlow <joerg.vehlow@aox-tech.de>
>
> The test is supposed to verify, that memory used by a task
> is not migrated to a subgroup, when the task is migrated.
> In order to test this, it is not required, to limit
> the memory at all, because accounting is done anyway and
> the test checks the memory statistics.
>
> Therefore all memory limiting is removed. This also gets
> rid of a faulty assumption when writing the test:
> The test allocates 33 pages using mmap. It is expected,
> that it does not need more than 33 * pagesize of memory.
> But sometimes the kernel has to allocate new page tables.
> This is accounted as kmem AND user memory.
> (See kernel doc cgroup-v1/memory.txt section 2.7)
> If this happend, the test failed, if the system has no swap,
> because the oom killer kills the process.
>
> Signed-off-by: Joerg Vehlow <joerg.vehlow@aox-tech.de>
> ---
>   .../kernel/controllers/memcg/functional/memcg_lib.sh |  4 ----
>   .../memcg/functional/memcg_subgroup_charge.sh        | 12 +-----------
>   2 files changed, 1 insertion(+), 15 deletions(-)
>
> diff --git a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
> index aadaae4d2..aec8a00ab 100755
> --- a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
> +++ b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
> @@ -404,13 +404,9 @@ test_hugepage()
>   }
>   
>   # Test the memory charge won't move to subgroup
> -# $1 - memory.limit_in_bytes in parent group
> -# $2 - memory.limit_in_bytes in sub group
>   test_subgroup()
>   {
>   	mkdir subgroup
> -	echo $1 > memory.limit_in_bytes
> -	echo $2 > subgroup/memory.limit_in_bytes
>   
>   	tst_resm TINFO "Running memcg_process --mmap-anon -s $PAGESIZES"
>   	memcg_process --mmap-anon -s $PAGESIZES &
> diff --git a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
> index 9b11f7b58..98e5ad68d 100755
> --- a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
> +++ b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
> @@ -26,7 +26,7 @@
>   ################################################################################
>   
>   TCID="memcg_subgroup_charge"
> -TST_TOTAL=3
> +TST_TOTAL=1
>   
>   . memcg_lib.sh
>   
> @@ -36,15 +36,5 @@ testcase_1()
>   	test_subgroup $PAGESIZES $((2*PAGESIZES))
>   }
>   
> -testcase_2()
> -{
> -	test_subgroup $PAGESIZES $PAGESIZES
> -}
> -
> -testcase_3()
> -{
> -	test_subgroup $PAGESIZES 0
> -}
> -
>   run_tests
>   tst_exit

[LTP] [PATCH] memcg_subgroup_charge: Remove memory limitation

[Richard Palethorpe]

Hello,

Joerg Vehlow <lkml@jv-coder.de> writes:

> Hi,
>
> a little ping for this patch?
> I recently updated the ltp suite for our tests to 20200930 and stumbled 
> over this error again.
> Any reason for not applying it?

If there was no response it's safe to assume it was forgotten
about. Thanks for bumping it.

>
> J?rg
>
> On 11/6/2019 7:18 AM, Joerg Vehlow wrote:
>> From: Joerg Vehlow <joerg.vehlow@aox-tech.de>
>>
>> The test is supposed to verify, that memory used by a task
>> is not migrated to a subgroup, when the task is migrated.
>> In order to test this, it is not required, to limit
>> the memory at all, because accounting is done anyway and
>> the test checks the memory statistics.
>>
>> Therefore all memory limiting is removed. This also gets
>> rid of a faulty assumption when writing the test:
>> The test allocates 33 pages using mmap. It is expected,
>> that it does not need more than 33 * pagesize of memory.
>> But sometimes the kernel has to allocate new page tables.
>> This is accounted as kmem AND user memory.
>> (See kernel doc cgroup-v1/memory.txt section 2.7)
>> If this happend, the test failed, if the system has no swap,
>> because the oom killer kills the process.

Perhaps they were checking for side effects from setting the limit... I
think though it is correct to remove these limits.

Both check_stat lines are suspect to me as some unrelated tasklet could
run in the processes context and allocate some memory. Possibly this
should be converted to only an approximate check?

>>
>> Signed-off-by: Joerg Vehlow <joerg.vehlow@aox-tech.de>
>> ---
>>   .../kernel/controllers/memcg/functional/memcg_lib.sh |  4 ----
>>   .../memcg/functional/memcg_subgroup_charge.sh        | 12 +-----------
>>   2 files changed, 1 insertion(+), 15 deletions(-)
>>
>> diff --git a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
>> index aadaae4d2..aec8a00ab 100755
>> --- a/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
>> +++ b/testcases/kernel/controllers/memcg/functional/memcg_lib.sh
>> @@ -404,13 +404,9 @@ test_hugepage()
>>   }
>>   
>>   # Test the memory charge won't move to subgroup
>> -# $1 - memory.limit_in_bytes in parent group
>> -# $2 - memory.limit_in_bytes in sub group
>>   test_subgroup()
>>   {
>>   	mkdir subgroup
>> -	echo $1 > memory.limit_in_bytes
>> -	echo $2 > subgroup/memory.limit_in_bytes
>>   
>>   	tst_resm TINFO "Running memcg_process --mmap-anon -s $PAGESIZES"
>>   	memcg_process --mmap-anon -s $PAGESIZES &
>> diff --git a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
>> index 9b11f7b58..98e5ad68d 100755
>> --- a/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
>> +++ b/testcases/kernel/controllers/memcg/functional/memcg_subgroup_charge.sh
>> @@ -26,7 +26,7 @@
>>   ################################################################################
>>   
>>   TCID="memcg_subgroup_charge"
>> -TST_TOTAL=3
>> +TST_TOTAL=1
>>   
>>   . memcg_lib.sh
>>   
>> @@ -36,15 +36,5 @@ testcase_1()
>>   	test_subgroup $PAGESIZES $((2*PAGESIZES))
>>   }
>>   
>> -testcase_2()
>> -{
>> -	test_subgroup $PAGESIZES $PAGESIZES
>> -}
>> -
>> -testcase_3()
>> -{
>> -	test_subgroup $PAGESIZES 0
>> -}
>> -
>>   run_tests
>>   tst_exit


-- 
Thank you,
Richard.