Re: [PATCH v2 4/5] selftests: prctl: introduce tests for disabling THPs completely

[David Hildenbrand <david@redhat.com>]

>>> +
>>> +struct test_results {
>>> +    int prctl_get_thp_disable;
>>
>> The result is always one, does that here make sense?
> 
> Its 3 in the next patch for PR_THP_DISABLE_EXCEPT_ADVISED :)
> 
> I will remove this struct, but I think maybe it might have been a good idea to squash this
> with the next patch to show why the struct was useful.

I think it's reasonable to keep them separate.

> 
>>
>>> +    int prctl_applied_collapse_none;
>>
>> "prctl_applied" is a bit confusing. And most of these always have the same value.
>>
>> Can't we special case the remaining two cases on the current policy and avoid this struct compeltely?
>>
> 
> The values are different in the next patch when PR_THP_DISABLE_EXCEPT_ADVISED is used.
> 
> Just to explain how I came about using this struct test_results (though it doesnt matter as
> I will remove it for the next revision):
> I wanted to maximise code reuse and only wanted to have one instance of prctl_thp_disable_test.
> I actually started with special casing, but went the brute force way of adding too many if else
> statements and it was looking quite messy after I added the tests for PR_THP_DISABLE_EXCEPT_ADVISED.
> I saw this struct test_results in another kselftest and thought this should make it much better and
> extendable.
> 
> I have removed struct test_results and changed prctl_thp_disable_test to the following for next revision:

Yeah, or just duplicate that function and call it 
prctl_thp_disable_unless_advised_test() in the next patch.

Makes the code easier to read and the duplication is limited.

-- 
Cheers,

David / dhildenb