Re: [PATCH 1/1] mm/khugepaged: reduce process visible downtime by pre-zeroing hugepage

[David Hildenbrand <david@redhat.com>]

On 14.03.24 15:19, Lance Yang wrote:
> Another thought suggested by Bang Li is that we record which pte is none
> in hpage_collapse_scan_pmd. Then, before acquiring the mmap_lock (write mode),
> we will pre-zero pages as needed.

Here is my point of view: we cannot optimize the common case where we 
have mostly !pte_none() in a similar way.

So why do we care about the less common case? Is the process visible 
downtime reduction for that less common case really noticable?

Or is it rather something that looks good in a micro-benchmark, but 
won't really make any difference in actual applications (again, where 
the common case will still result the same downtime).

I'm not against this, I'm rather wonder "do we really care". I'd like to 
hear other opinions.

>>>>>
>>>>> So my question is: do we really care about it that much that we care to
>>>>> optimize?
>>>>
>>>> IMO, although it may not be our main concern, reducing the impact of
>>>> khugepaged on the process remains crucial. I think that users also prefer
>>>> minimal interference from khugepaged.
>>>
>>> The problem I am having with this is that for the *common* case where we
>>> have a small number of pte_none(), we cannot really optimize because we
>>> have to perform the copy.
>>>
>>> So this feels like we're rather optimizing a corner case, and I am not
>>> so sure if that is really worth it.
>>>
>>> Other thoughts?
>>
>> Another thought is to introduce khugepaged/alloc_zeroed_hpage for THP
>> sysfs settings. This would enable users to decide whether to avoid unnecessary
>> copies when nr_ptes_none > 0.

Hm, new toggles for that, not sure ... I much rather prefer something 
without any new toggles, especially for this case.

-- 
Cheers,

David / dhildenb