Re: [PATCH v2 3/3] mm: huge_memory: use folio_skip_prot_numa() for pmd folio

[Kefeng Wang <wangkefeng.wang@huawei.com>]

On 2025/10/15 19:30, David Hildenbrand wrote:
> On 15.10.25 13:04, Kefeng Wang wrote:
>>
>>
>> On 2025/10/15 17:54, David Hildenbrand wrote:
>>> On 15.10.25 11:21, Kefeng Wang wrote:
>>>>
>>>>
>>>> On 2025/10/15 15:30, Dev Jain wrote:
>>>>>
>>>>> On 14/10/25 5:03 pm, Kefeng Wang wrote:
>>>>>> Rename prot_numa_skip() to folio_skip_prot_numa(), and remove
>>>>>> ret by directly return value instead of goto style.
>>>>>>
>>>>>> The folio skip checks for prot numa should be suitable for pmd
>>>>>> folio too, which helps to avoid unnecessary pmd change and folio
>>>>>> migration attempts.
>>>>>>
>>>>>> Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com>
>>>>>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>>>>>
>>>>> In the review of my mprotect pte batching series, reviewers had
>>>>> noted that the branch "if (folio_use_access_time(folio))" in
>>>>> folio_skip_prot_numa() did not belong there - it should be done
>>>>> outside of the function. But I see that that would duplicate a line
>>>>> now that this function has two users. So in light of that, would you
>>>>> mind changing the name of this function to
>>>>> folio_skip_or_process_prot_numa()?
>>>>>
>>>>
>>>> The name is a bit long, and it only update access_time not change the
>>>> pte, so maybe we leave it as is?
>>>
>>> Any such name might make the return value weird (which indicates whether
>>> to skip) I'm afraid.
>>>
>>> We could invert the meaning and call it something like
>>>
>>>       folio_apply_prot_numa()
>>>
>>> And return whether we have to protect it.
>>>
>>> Maybe that's better? Other naming suggestions welcome :)
>>>
>>
>> That's better, or folio_needs_prot_numa() as there are already some
>> similar names with folio_needs_ prefix ?
> 
> Yeah, why not. Then we can also add kerneldoc to describe that it will 
> adjust the access time if the function returns true.
> 

Done in v3.