Re: [PATCH] mm/huge_memory: prevent THP_ZERO_PAGE_ALLOC increased twice

[Liu Shixin]

On 2022/9/8 7:35, Andrew Morton wrote:
> On Tue, 6 Sep 2022 09:52:23 +0800 Liu Shixin <liushixin2@xxxxxxxxxx> wrote:
>
>> On 2022/9/6 4:07, Andrew Morton wrote:
>>> On Mon, 5 Sep 2022 21:38:13 +0800 Liu Shixin <liushixin2@xxxxxxxxxx> wrote:
>>>
>>>> If two or more threads call get_huge_zero_page concurrently, THP_ZERO_PAGE_ALLOC
>>>> may increased two or more times. But actually, this should only count
>>>> as once since the extra zero pages has been freed.
>>> Well, for better of for worse,
>>> Documentation/admin-guide/mm/transhuge.rst says
>>>
>>> thp_zero_page_alloc
>>> 	is incremented every time a huge zero page is
>>> 	successfully allocated. It includes allocations which where
>>> 	dropped due race with other allocation. Note, it doesn't count
>>> 	every map of the huge zero page, only its allocation.
>>>
>>> If you think this interprtation should be changed then please explain
>>> why, and let's be sure to update the documentation accordingly.
>>>
>>> .
>> Thanks for your explanation. I misunderstood the meaning of thp_zero_page_alloc before.
>> Although the rules are clearly explained in the documentation, I think that this variable
>> should only incremented when a huge zero page used for thp is successfully allocated and
>> the pages dropped due race should skip increment. It seems strange to count in all allocations.
>>
>> If there's something I still misunderstand, please point it out, thanks.
> It seems strange to me also.  Perhaps there's a rationale buried in the
> git and mailing list history.
>
> .
I didn't find previous discussion about this point. I update document in v2.
Kirill, what do you think about this change?

https://lore.kernel.org/linux-mm/20220908035533.2186159-1-liushixin2@xxxxxxxxxx/