Re: [PATCH v7 1/2] selftests: memcg: Allow low event with no memory.low and memory_recursiveprot on

[Waiman Long <llong@redhat.com>]

On 4/16/25 5:25 AM, Michal Koutný wrote:
> On Tue, Apr 15, 2025 at 05:04:14PM -0400, Waiman Long <longman@redhat.com> wrote:
>> +	/*
>> +	 * Child 2 has memory.low=0, but some low protection is still being
>> +	 * distributed down from its parent with memory.low=50M if cgroup2
>> +	 * memory_recursiveprot mount option is enabled. So the low event
>> +	 * count will be non-zero in this case.
> I say: Child 2 should have zero effective low value in this test case.
> Johannes says (IIUC): One cannot argue whether there is or isn't
> effective low for Child 2, it depends on siblings.
> (I also say that low events should only be counted for nominal low
> breaches but that's not so important here.)
>
> But together this means no value of memory.events:low is valid or
> invalid in this testcase. Hence I suggested ignoring Child 2's value in
> checks.
I understand your point of view. What I want to do is to document the 
expected behavior and I don't see any example of ignoring a metric for a 
particular child in the test. In this particular test, I did see an elow 
of 17 for child 2.
>
>> +	 */
>>   	for (i = 0; i < ARRAY_SIZE(children); i++) {
>> -		int no_low_events_index = 1;
>> +		int no_low_events_index = has_recursiveprot ? 2 : 1;
>>   		long low, oom;
>>   
>>   		oom = cg_read_key_long(children[i], "memory.events", "oom ");
> But this is not what I Suggested-by: [1]

I was referring to the suggestion that the setting of 
memory_recursiveprot mount option has a material impact of the child 2 
test result. Roman probably didn't have memory_recursiveprot set when 
developing this selftest.

I can take out the Suggested-by tag.

Cheers,
Longman


>
> Michal
>
> [1] https://lore.kernel.org/r/awgbdn6gwnj4kfaezsorvopgsdyoty3yahdeanqvoxstz2w2ke@xc3sv43elkz5